affinity-mediated protein modification and recovery · doctor of philosophy in chemistry university...

Affinity-mediated protein modification and recovery

by

Richard Lee Kwant

A dissertation submitted in partial satisfaction of the

requirements for the degree of

Doctor of Philosophy

in

Chemistry

in the

Graduate Division

of the

University of California, Berkeley

Committee in charge:

Professor Matthew B. Francis, ChairProfessor Phillip GeisslerProfessor Wenjun Zhang

Fall 2015


Copyright 2015by

Richard Lee Kwant

1

Abstract


by

Richard Lee Kwant

Doctor of Philosophy in Chemistry

University of California, Berkeley

Professor Matthew B. Francis, Chair

As one of the core building blocks of life, proteins play a central role in our efforts tounderstand, control, and engineer the natural world. Chemical modification of proteinsfacilitates these efforts by allowing manipulation of protein properties and creation of newmaterials. A variety of methods for the modification of proteins exist, and collectivelythese methods have enabled the synthesis of drugs, the study of biological systems, and thecreation of new materials. Unfortunately the exquisite control provided by many of thesemethods does not extend to oligomeric proteins, which are found ubiquitously in nature.These proteins often have symmetries and structures that make them particularly appealingfor use in materials science and pharmaceutical applications, yet their multivalency makestheir controlled modification difficult. This thesis describes the development of a methodfor protein modification in which control over the number of modification sites is exertedthrough noncovalent interactions. This method is able to control the level of modificationof both monomeric and small oligomeric proteins. Further development revealed that thismethod also discriminates between different modification sites, such that it could be used tosite-specifically modify proteins or monitor bioconjugation reactions. Finally, the principlesused in the construction of these bioconjugation strategies were applied to the recovery ofenzymes from industrial reactions.

i

To my parents

ii

Contents

Contents ii

List of Figures v

List of Tables vii

I Introduction, scientific motivation, and theory 1

1 Introduction 21.1 Importance of protein bioconjugates . . . . . . . . . . . . . . . . . . . . . . . 31.2 Traditional methods for modification of proteins . . . . . . . . . . . . . . . . 4

1.2.1 Modification of amino acid side chains . . . . . . . . . . . . . . . . . 51.2.2 Modification of unique sites on proteins . . . . . . . . . . . . . . . . . 61.2.3 Introduced sites for chemical modification . . . . . . . . . . . . . . . 7

1.3 Affinity-directed methods for modification of proteins . . . . . . . . . . . . . 91.4 Second-order methods for protein modification and their potential for devel-

opment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111.5 Cyclodextrin and its applications . . . . . . . . . . . . . . . . . . . . . . . . 11

1.5.1 Chemical modification of cyclodextrins . . . . . . . . . . . . . . . . . 121.5.2 Uses of cyclodextrins . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.6 Project design and thesis overview . . . . . . . . . . . . . . . . . . . . . . . . 131.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2 Theory of chromatography 222.1 Motivation for theoretical calculations . . . . . . . . . . . . . . . . . . . . . 232.2 Existing theories of chromatography . . . . . . . . . . . . . . . . . . . . . . . 232.3 Basic feasibility of chromatography-mediated bioconjugation . . . . . . . . . 252.4 Plate-based model of affinity chromatography . . . . . . . . . . . . . . . . . 25

2.4.1 Kinetic binding model . . . . . . . . . . . . . . . . . . . . . . . . . . 262.4.2 Analytic binding model . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

iii

2.5.1 Choice of eluent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292.5.2 Potential to separate multiple modification levels . . . . . . . . . . . 312.5.3 Application to lateral flow assays . . . . . . . . . . . . . . . . . . . . 322.5.4 Web-based simulation . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.6 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

II Experimental work 37

3 Chromatography-mediated bioconjugation 383.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393.2 Design of system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.2.1 Design and synthesis of the affinity handle . . . . . . . . . . . . . . . 393.2.2 Design and synthesis of the resin . . . . . . . . . . . . . . . . . . . . 40

3.3 Purification of proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423.4 Elaboration of modified protein . . . . . . . . . . . . . . . . . . . . . . . . . 433.5 Construction of well-defined light harvesting mimics . . . . . . . . . . . . . . 443.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483.7 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3.7.1 General methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483.7.2 Instrumentation and sample analysis . . . . . . . . . . . . . . . . . . 483.7.3 Small molecule synthesis . . . . . . . . . . . . . . . . . . . . . . . . . 493.7.4 Resin synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 553.7.5 Determination of resin loading and binding constants . . . . . . . . . 563.7.6 Protein expression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573.7.7 Protein modification . . . . . . . . . . . . . . . . . . . . . . . . . . . 583.7.8 Chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

3.8 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593.9 Additional figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

4 Monolithic chromatography-mediated bioconjugation 664.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674.2 Choice of monolithic support . . . . . . . . . . . . . . . . . . . . . . . . . . . 674.3 Modification and characterization of monolithic silica columns . . . . . . . . 68

4.3.1 Initial modification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 684.3.2 Optimization of modification . . . . . . . . . . . . . . . . . . . . . . . 714.3.3 Characterization of protein folding . . . . . . . . . . . . . . . . . . . 74

4.4 Purification of modification levels . . . . . . . . . . . . . . . . . . . . . . . . 754.4.1 Site-specific purification of Mth1491 . . . . . . . . . . . . . . . . . . . 754.4.2 Site-specific purification of ubiquitin . . . . . . . . . . . . . . . . . . 77

4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

iv

4.6 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 804.6.1 General methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 804.6.2 Instrumentation and sample analysis . . . . . . . . . . . . . . . . . . 804.6.3 Modification of silica monoliths . . . . . . . . . . . . . . . . . . . . . 804.6.4 Small molecule synthesis . . . . . . . . . . . . . . . . . . . . . . . . . 824.6.5 Determination of binding constants to β-cyclodextrin immobilized on

silica gel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834.6.6 Chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834.6.7 Protein expression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834.6.8 Protein modification . . . . . . . . . . . . . . . . . . . . . . . . . . . 834.6.9 Protein digestion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83


5 Affinity-based recovery of enzymes 895.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 905.2 Design of system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 915.3 Results from low-affinity ligands . . . . . . . . . . . . . . . . . . . . . . . . . 925.4 High-affinity ligands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

5.4.1 High-affinity ligands of modified β-cyclodextrin . . . . . . . . . . . . 955.4.1.1 Results of pulldowns with cysteamine-β-cyclodextrin 5.2 . . 95

5.4.2 High-affinity cholesterol-based ligands of native β-cyclodextrin . . . . 975.4.2.1 Results of pulldowns using lithocholic acid . . . . . . . . . . 97

5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 985.6 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

5.6.1 Instrumentation and sample analysis . . . . . . . . . . . . . . . . . . 995.6.2 Small molecule synthesis . . . . . . . . . . . . . . . . . . . . . . . . . 1005.6.3 Resin synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025.6.4 Pulldowns of EGPh . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1035.6.5 Determination of in-solution binding constants . . . . . . . . . . . . . 1035.6.6 Protein modification . . . . . . . . . . . . . . . . . . . . . . . . . . . 1035.6.7 Protein expression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104


A Simulation code 108A.1 Kinetic binding model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108A.2 Analytic binding model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112A.3 Lateral flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114A.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

v

List of Figures

1.1 Common methods for the modification of lysine side chains . . . . . . . . . . . . 41.2 Common methods for the modification of cysteine side chains . . . . . . . . . . 51.3 Chemical methods for the modification of N- and C-termini of proteins . . . . . 71.4 Examples of “click” methods for protein modification . . . . . . . . . . . . . . . 81.5 Histogram of cyclodextrin binding constants, produced from Rekharsky and Inoue. 121.6 Overview of chromatography-mediated bioconjugation. . . . . . . . . . . . . . . 14

2.1 Heatmap of ∆r as a function of binding constant and resin loading . . . . . . . 262.2 Heatmaps showing elution of unmodified, singly-modified, and doubly-modified

proteins with different eluents. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302.3 Elution profile of 20 modification states . . . . . . . . . . . . . . . . . . . . . . . 312.4 Schematic diagram of LFA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332.5 Results of signal-to-background calculations for LFA . . . . . . . . . . . . . . . 34

3.1 Proposed method for chromatography-mediated bioconjugation and overview ofazo modification chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3.2 Functionalization of 90 µM Sepharose CL-4B with β-cyclodextrin . . . . . . . . 413.3 Reconstructed ESI-TOF mass spectra of crude and purified bioconjugates . . . . 433.4 Handle-assisted purification results . . . . . . . . . . . . . . . . . . . . . . . . . 443.5 Modification and fluorescence of Mth1491 . . . . . . . . . . . . . . . . . . . . . 463.6 Absorbance spectrum of azo 3.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 613.7 Absorbance spectrum of azo 3.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . 613.8 Binding of azo 3.3 to β-cyclodextrin . . . . . . . . . . . . . . . . . . . . . . . . 623.9 Binding of azo 3.6 to β-cyclodextrin . . . . . . . . . . . . . . . . . . . . . . . . 623.10 Pulldown experiments with azo 3.3 and resins 1-3 . . . . . . . . . . . . . . . . . 633.11 Elution of azo 3.3 and Mth1491 modified to 30% with azo maleimide 3.5 from

alkyne- and β-cyclodextrin-terminated resins . . . . . . . . . . . . . . . . . . . . 643.12 Attempted modification of endogenous cysteines in wt-Mth1491 . . . . . . . . . 653.13 Absorbance spectra of protein samples discussed in the main text. . . . . . . . . 65

4.1 Initial synthetic modification of silica rod column . . . . . . . . . . . . . . . . . 684.2 Elution of azo 3.3 and Mth1491 from monolith 4.2 . . . . . . . . . . . . . . . . 69

vi

4.3 Elution of azo 3.3 and Mth1491 from monolith 4.3 . . . . . . . . . . . . . . . . 704.4 Elution of azo 3.3 from monoliths 4.3, 4.6, and 4.7 . . . . . . . . . . . . . . . . 724.5 Elution of Mth1491 from monoliths 4.3, 4.6, and 4.7 . . . . . . . . . . . . . . . 734.6 Purification of proteins labeled with azo NHS-ester 3.4. . . . . . . . . . . . . . . 764.7 Analysis of Mth1491 modified in two different locations. . . . . . . . . . . . . . 774.8 Crystal structure and sequence of bovine ubiquitin . . . . . . . . . . . . . . . . 784.9 Identification of modification sites of ubiquitin . . . . . . . . . . . . . . . . . . . 864.10 Determination of binding constants to silica gel modified with β-cyclodextrin . . 874.11 Fluorescence of GFP before and after purification . . . . . . . . . . . . . . . . . 88

5.1 Scheme for recovery of EGPh from crude reaction mixtures . . . . . . . . . . . . 915.2 Heatmaps showing the efficacy of pulldowns for different binding constants . . . 935.3 Pulldown results for adamantyl EGPh . . . . . . . . . . . . . . . . . . . . . . . 945.4 Synthesis and binding of stilbene 5.5. . . . . . . . . . . . . . . . . . . . . . . . . 965.5 Modification of EGPh with lithocholic acid derivative 5.7 and pulldown of RNAse

A modified with lithocholic acid derivative 5.7 . . . . . . . . . . . . . . . . . . . 985.6 Low-salt pulldown of EGPh modified with an adamantyl group . . . . . . . . . 1075.7 Pulldown and titration of EGPh modified with an adamantyl group . . . . . . . 107

vii

List of Tables

3.1 Resins with varying Clinker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

viii

Acknowledgments

It would be impossible to go through graduate school without an extraordinary amount ofhelp. The majority of learning happens outside of the classroom, while puzzling throughproblems with friends and colleages, while thinking about something that someone said,or while trying to help another. A complete acknowledgements section would therefore beimpossible to generate, but here’s an attempt.

First and foremost I’d like to thank Matt Francis. Matt is one of the most supportivePIs that I know, and I don’t think that there was ever a moment that he did not encourageme to follow my interests, whether that be coding or chemistry. Matt has a real knack forpresenting information and figures, and I think this ability is one of the fundamental skillsthat I picked up in the lab. I’ll use it for years to come. Beyond that, Matt actively createda positive work environment that was supportive and educational. While being in lab waswork, it was often fun.

I am heavily indebted to the members of the Francis group for teaching me many ofthe lab skills that I know. Many of those who did the most teaching are those who leftthe earliest—Zac, Leah, and Michel helped me get my footing in the lab. They taught meabout lab activities, included me in outings, and never hesitated when I needed to learna new instrument. I think we were all a little sad that Michel graduated when he did,because everything got fixed quite a bit more slowly after he left. And of course there’sGary: always the life of the party, full of interesting and provocative questions. I think Garysingle-handedly kept us entertained on multiple car rides to and from Tahoe. And withGary comes Chris, a great chemist who is also really good at organizing stock rooms. I thinkWesley and Kanna were also in this class. We didn’t see them as much, but when we didtheir presence was always appreciated because of their sharp questions and good advice.

The next class included Kristen, Mike, Troy, Dan, Amy, and Allie, and this was reallythe class that defined what the Francis group was for me. Kristen introduced me to life in743 with a colorful selection of alternative music. Her synthetic skills were impressive, andI was always amazed at how quickly she turned out complicated molecules. Mike combinedan impressive knowledge of biology, physics, and lasers. He was one of the most helpfulmembers of the lab, and was always ready to give some solid advice when you needed it.Troy and Amy were quite elusive when I joined the lab because they sat on a different floor.However, I’m glad that I eventually got to know them better when they moved up to the7th floor. Dan is an exceedingly meticulous chemist when it comes to the preparation ofmodified proteins, and I learned a lot about protein modification from him. He also was oneof those people who would randomly decide that a given day needed to be a little more fun,and he would play a prank or make some amusing drawings of animals. And lastly there’sAllie, to whom I owe a huge thanks for teaching me so much about oxidative couplings,protein modification, and science in general. Allie was one of the most talented and mosthelpful members of the lab, and she did a lot to make the Francis group the friendly, helpfulplace that it is.

ix

Then there were KMac, Jeff, Stacy, and Farkas. Katherine taught me almost everythingI know about EGPh and the recovery of cellulases. She was both really good at carrying outcomplicated procedures over long periods of time and at baking. We didn’t see much of Jeffbecause he sat with the DTE lab, but when we did he was usually helping us with eitherMS2 or the HPLCs. Stacy showed us an impressive work ethic and set of skills—she wasalmost always in the lab before and after everyone else, and she worked on a range of projectsthat involved a number of different skills. Despite all of her hard work, Stacy remained acheerful labmate who was always ready to help out. And lastly Farkas, who wasn’t really inthis class but we’ll include her anyway. I best remember Farkas for her strong predilectionsfor certain activities, like softball, beer, or antibodies. She really helped to get the lab doinginteresting things, often other than science.

Abby came in a class all by herself, but she more than occupied the position. From beingthe person who knew how all of the servers and computers worked to being the lab experton peptoids, she always excelled at what she did.

The biggest thanks to my class, who helped out with countless experiments, lab events,qual practices, and meetings. Jake was there from the very beginning when we had toshare a desk. He helped get me started on my first project, and I always appreciated hisknowledgeable insight on protein modification and synthesis in general. I probably saw themost of Jim, from whom I stole Leah’s desk and with whom I shared 743 for years. I willnever forget Jim’s excessive beard, or the hundreds of great conversations that we had aboutscience and not science throughout the years. Kanwal gave me invaluable advice on manymatters more biological than I preferred, and also helped in a thousand different ways, fromreading manuscripts and this thesis to grabbing sushi and life advice. Ioana is a somewhatof an intimidating presence in 733 because of her serious demeanor, but that just hides thejokester within that comes out when you least expect it. Her ability to make and delivercomplex bioconjugates on a schedule is also quite impressive. Jenna was a late addition, butshe’s done a lot to make the lab the fun place that it is. You can always count on her. I guesswe’ll adopt Jess, who joined late as well. I will best remember Jess for her delicious-lookingfood that she would heat up in the group room.

Then there are Matt, Joel, and Am, who have all changed the lab for the better in theirown ways. Matt, for showing up at all hours of the day, randomly popping in to have aconversation about something and then walking away in the middle, and talking a lot aboutsharks. I’m also thankful for the opportunity to collaborate with Matt for a while becauseit was a very interesting project. Joel is probably the most important person in the labbecause he runs the coffee maker. It has also been great hanging out with him, literally,while climbing. And Am, mysteriously silent much of the time, but admirable for her abilityto make so much progress so quickly on her first project.

Rachel, Kristen, Adam, and Emily: keep up the good work. It’s been great getting toknow you over the past year and a half, and I’m looking forward to seeing the great sciencethat you’ll do. Honestly I could say more but I really just want to turn this in.

And I can’t forget the set of awesome postdocs that we’ve had throughout the years. Ayoand Christian, always awesome to hang out with, even if you don’t want to do Karaoke. I had

x

the opportunity to work fairly closely with Christian on two of my projects, and am thankfulfor his contributions and his interesting perspective on protein modification. The projectsthat I worked on with Christian got their start with Meera, who taught me quite a bit oforganic chemistry and was also really fun to work with. She could always be counted on for anenlightening conversation on the happenings of the lab. Henrik’s stay in the Francis lab wasbrief, but it was great getting to know him and collaborating on resin modification. Carsonalways provided great examples of independent thinking in project design and practicalproject management, and Rafi was always there to provide cheerful conversation. And Adel,who is both and impressive organic chemist and an impressive thinking. It’s often challengingkeeping up with the places Adel’s mind goes, and he provided tons of great feedback duringmy PhD that I would have been nowhere without.

During my time as a grad student I had the opportunity to work with a stellar undergrad,Pete. Pete helped me quite a bit get both of my projects off the ground, and his contributionswere invaluable. Pete also helped me become a better teacher and learn to work as a team.

And lastly my family, who always supported me during my graduate studies. My parentsinstilled in me the curiosity to pursue an academic career, and gave me the resources andskills to do so. And Kathryn, she’s just awesome :)

1

Part I

Introduction, scientific motivation,and theory

2

Chapter 1

Introduction

Abstract

As one of the core building blocks of life, proteins play a central role in our efforts to under-stand, control, and engineer the natural world. Chemical modification of proteins facilitatesmany of these efforts because it allows manipulation of protein properties and creation ofnew materials. A large variety of methods exist to chemically modify proteins. These meth-ods vary widely in their applicability and specificity, and in many cases it is possible to finda chemical method that can accomplish a desired chemical modification. However, thereremain a subset of proteins whose modification is difficult because they contain more thanone potential modification site. Such proteins include oligomeric complexes and large pro-teins like antibodies. In such cases, additional noncovalent interactions can be used to directchemical modification and increase specificity. This chapter discusses the potential to usecyclodextrins to noncovalently direct chemical modification of proteins.

Haiku

Proteins do a lotMuch more, when programmableBy appending stuff

CHAPTER 1. INTRODUCTION 3

1.1 Importance of protein bioconjugates

As a central element of life and a class of molecules with seemingly boundless structure andfunction, proteins have found countless uses in medicine, biotechnology, and science. Chem-ical modification of these molecules further expands their utility, for modification changesprotein function,1 allows construction of new materials,2,3 and enables the study of proteinstructure and function.4–6 For example, the chemical modification of viral capsids withinthe Francis group has lead to applications that range from the diagnosis and treatment ofdisease2,7,8 to the study of energy transfer.3 In these cases, the proteins serve primarily asscaffolds, and most of the rich functionality of these materials derives from the ability toorganize spatially a variety of chemical groups. The construction of such materials would notbe possible without a complete set of chemical tools to allow this modification. In anotherexample, Clarke et al. synthetically modified the pore protein α-haemolysin to reduce thesize of its β barrel.1 This reduction in pore size allowed for the sequencing of DNA passingthrough the pore under an applied potential.

One particularly interesting class of targets for bioconjugation is antibodies. Antibodiesconsist of four polypeptide chains—two light 25 kDa chains and two heavy 50 kDa chainsjoined by multiple disulfide bonds—whose purpose is to recognize specific antigens and in-teract with effector cells.9 Because of their exquisite selectivity in binding specific targetswithin a complex mixture of biomolecules, antibodies have found numerous therapeutic ap-plications, particularly for autoimmune diseases and cancer.10 In 2008, antibodies were thehighest grossing biologic and produced over $15 billion in revenue.11 However, antibodiesdo not always elicit the desired therapeutic effect, and there has been great interest in con-jugating them with drugs to broaden their therapeutic potential.12 There are over thirtyantibody-drug-conjugates (ADCs) currently in clinical trials, but these conjugates remaintechnically challenging to construct.13 The drug loading on these constructs is known toaffect their efficacy, yet the controlled modification of antibodies remains difficult.14,15 Atthe heart of the issue is the size and complexity of antibodies, for each antibody has overone-hundred lysines and multiple disulfide bonds whose modification would disrupt the pro-tein structure. These proteins defy traditional methods for protein modification, and thedevelopment of strategies to modify antibodies in a controlled manner is ongoing.

Oligomeric proteins represent another class of proteins that has defied existing strategiesfor bioconjugation. In fact, most proteins form complexes with at least two subunits, anda survey of proteins from E. coli. found that only 19% of proteins were monomeric.16

Oligomeric proteins represent particularly attractive scaffolds for the construction of newmaterials, for their symmetries can be used to precisely template molecules and to createcompartments of well-defined size. Oligomeric proteins are the basis on which the researchprogram of the Francis group has been built, and these proteins have been used extensivelyto study energy transfer,3,17,18 construct platforms for drug delivery,2,8, and to make agentsfor biological imaging.8,19,20 However, the very feature that makes these proteins interestingmakes their chemical modification difficult. Most methods for the modification of proteinsare selective at the level of functional groups, and therefore they cannot distinguish between


HN

NHO

HN

R1

R2 R4

NH

R3

HN

NHO

HN

R1

R2 O

R3

HN

NHO

HN

R1

R2

S R3

O O

HN

NO

HN

R1

R2

R3

HN

NHO

HN

R1

R2

R3

OH

HN

NHO

HN

R1

R2

R3

R3NCR6

R6=S or O

O

O

R3N

O

O

ClS R3

O O

O

H R3

R3

O

O

O

R3

O

O

O

R3

N

HNR5

R4

HN

NH2O

HN

R1

R2

Reduction

NHS-ester, anhydride, carbodiimide

Aldehyde

Epoxide

Isothiocyanate and isocyanate

Sulfonyl chloride

Figure 1.1: Common methods for the modification of lysine side chains. Lysines can be modifiedvia a variety of chemistries to give water-stable conjugates. See text for details.

multiple chemically identical sites that are repeated on an oligomeric complex. Thereforecontrolling the number of chemical modifications per complex or the ratio of different kindsof chemical modifications has proven difficult, and in most cases these bioconjugates consistof statistical mixtures with different levels of modification.

1.2 Traditional methods for modification of proteins

The development of effective methods for bioconjugation is often quite challenging becausethe chemical modification of proteins is subject to many constraints. Not only do proteinshave a limited set of functional groups that are usually repeated many times, but they aresoluble almost exclusively in water and usually near neutral pH. Most proteins also havelimited stability and are sensitive to factors such as high and low temperatures, stirring, andoxidation. Because of these difficulties, numerous chemical methods for the modificationof proteins exist, with each having its own particular advantages. These methods can begrouped into three main categories: methods that target amino acid side chains, methodsthat target unique sites on protein surfaces, and methods that introduce reactive sites forchemical modification.

The most common targets for the modification of native amino acid side chains are ly-sine and cysteine. Lysine, an amino acid with an abundance of 5.9%,21 is commonly foundon protein surfaces; because of this fact, most proteins can be modified at their lysine side


HNS

O

HN

R2

R1 SR3

HNS

O

HN

R2

R1 N

O

O

R3

HNS

O

HN

R2

R1 R3

O

R4 SS R3

N

O

R3

O

IR3

O

HNSH

O

HN

R2

R1 Disul�de

Maleimide

Iodoacetamide

Figure 1.2: Common methods for the modification of cysteine side chains. See text for details.

chains. In fact, a search of characterized proteins in the Uniprot database revealed that98.8% of proteins contain at least one lysine.22 Lysines are typically modified at slightlybasic pH (7.5-9), most commonly using activated esters (often NHS-esters), carbodiimides,anhydrides, aldehydes and glyoxals, epoxides, isothiocyanates, isocyanates, and sulfonyl chlo-rides (Figure 1.1).23 These reactions commonly reach completion in a matter of hours to givewater-stable acylation products and secondary amines. Given the abundance of lysines, thesemodification strategies commonly result in mixtures of proteins bearing multiple numbersand different locations of modification.24

In cases where the limited selectivity of lysine chemistry becomes an issue, often a goodalternative is the modification of cysteine residues. Cysteine residues are found in lowerabundance (1.9%) and are often occupied in disulfide bonds as cystines.21 As a result, proteinstypically have few cysteines on their surface, and when present these residues can be targetedquite selectively. Historically, these residues have been commonly modified through disulfide-exchange, in which an activated thiol forms a mixed disulfide with a cysteine side chain(Figure 1.2).24 This reaction has the benefit of being tolerant of a range of pH and bufferconditions, and it is also reversible.23 Cysteines are also commonly modified using maleimidesand haloacetyls (e.g. iodoacetamides) (Figure 1.2).23 These reactions can be carried out atneutral pH and result in moderately stable conjugates with half lives on the order of days,a time scale which is sufficient for many applications.25

1.2.1 Modification of amino acid side chains

In addition to strategies for the modification of the side chains of lysine and cysteine, avariety of chemical techniques exist for the modification of other side chains, including thoseof tyrosine, serine and threonine, glutamic and aspartic acid, and tryptophan.23


1.2.2 Modification of unique sites on proteins

Another strategy for the modification of proteins has been to target unique functional groupson proteins, rather than rare amino acids. Typically, this approach focuses on the N- andC- termini, for these moieties occur only once on each protein chain (Figure 1.3).

The development of chemistries to N-terminally modify proteins has been a particularinterest of the Francis group, and it has seen active development during the past ten years.The N-terminus differs from primary amines found on lysine side chains in that its pKa andthe pKa of the carbon alpha are comparatively lower. As a result, it was observed thatcertain aldehydes were capable of transaminating the N-terminus to form either a ketone oran aldehyde, which can be further modified by condensation with alkoxyamine or hydrazinederivatives.26 The earliest form of this method used pyridoxal-5’phosphate to accomplish thetransamination,26, and it was later shown that the transamination reaction is particularlyefficient on proteins whose first three amino acids are AKT.27 More recently, it was shownthat N -methylpyridinium-4-carboxaldehyde can also be used to effect this transamination,but with a complementary N-terminal preference in which proteins starting with negativelycharged amino acids lead to greater conversion.28,29 These two methods enable the modi-fication of a large number of proteins to high conversion under fairly mild conditions (pH5.5-6.5). The installation of an aldehyde on the N-terminus of a protein can also be ac-complished through oxidation of the N-terminus with sodium periodate, provided that theN-terminal amino acid is either threonine or serine. The resulting aldehyde can be modifiedas discussed above.

During the course of this work it was also observed that certain pyridinecarboxaldehydesformed condensation products with proteins. Of these, 2-pyridinecarboxaldehyde was foundto form a cyclic imidazolidinone as a result of nucleophilic attack of the first amide nitrogenon the imine condensation product. As a result, 2-pyridinecarboxaldehyde derivatives can beused to modify protein N-termini to high conversion under mild conditions (pH 7.5, 37 ◦C)in one step.30 While the resulting bioconjugates are less water stable than those formedvia transamination, the simplified one-step reaction often leads to higher overall yields ofprotein.

Another fortuitous observation laid the groundwork for an alternative N-terminal mod-ification strategy. During the development of an oxidative coupling reaction (discussed inSection 1.2.3) between aniline and o-aminophenol derivatives, a small amount of backgroundreactivity was observed. This background reactivity resulted from the reaction of modifiedprotein N-termini with o-aminophenol derivatives oxidized with K3Fe(CN)6. Further reac-tion development revealed that protein N-termini could be selectively modified as a result oftheir low pKa’s.31 Yields of 70-100% can be commonly achieved within 15 min with as fewas 5-10 equivalents of reagent.

While the C-terminus has received less attention because its properties are less uniquethan those of the N-terminus, native chemical ligation offers a robust method for its modifi-cation.32 This method works in a manner identical to intein splicing, in which a C-terminalthioester undergoes a trans-thioesterification reaction with the N-terminal cysteine of an-


NH

ON

R2

R1 R4R3

NN

NR3

NH

N

R2

R1

O

NH

ON

R2

R1

O

R3

OH

O

NH

R5

R7

R3

HS

NH

ON

H

R1 R4R3

NH

ONH2

R2

R1

NH

ONH2R1

OH

NH

ONH2

R2

R1

NH

ONH2

R2

R1

O

SR6R5

R7

N

OH

HOOPO3

2-

N+

OH

PhSO3-

NH

OO

R2

R1

NH

OO

H

R1

NH

O

NN

R3

NH2OH

R3

H2N

HS

R3

Transamination

OxidationSodium periodate

R3R4H2N

R4=O or N

R3R4H2N

R4=O or N

N-terminal modi�cation with 2-pyridine carboxaldehyde

Oxidative coupling

Native chemical ligation

Oxime/hydrazone formation

Oxime/hydrazone formation

N-terminal modi�cation chemistry

C-terminal modi�cation chemistry

Figure 1.3: Chemical methods for the modification of N- and C-termini of proteins. See text fordetails.

other peptide.23. The resulting thioester is then displaced by the N-terminal nitrogen to givea stable amide bond. This reaction proceeds at physiological pH under mild conditions andoften reaches high yields and can be used for demanding couplings.33 Its primary limitationis the incorporation of a thioester into the C-terminus of a protein or a peptide, and thisstep is often accomplished by synthesizing a protein or peptide by hand on a solid support.

1.2.3 Introduced sites for chemical modification

Perhaps the area of most concentrated effort in the field of protein modification has beenthe development of reactions that target functional groups that are not typically found inproteins. This effort has been enabled, in part, by the development of methods to incorpo-rate unnatural amino acids into proteins, primarily through amber codon suppression andreassignment.34–36 These methods enable the incorporation of a range of functional groups,provided that the introduced side chain is not significantly different in size or shape from


ProteinN

Protein

NN

NN

N

R1

R1

ProteinHN

O

R1POPh

Ph

ProteinN

NNR1

ProteinNHN

R1

R1

Protein

N

O

R1

OH

ProteinN3

Protein

ProteinN3

ProteinN3

Protein

NH2

Protein

R1

R1N3

R1

MeO

O

Ph2P

R1

NN N

N

R1

R1

R1

OHNH2

Staudinger ligation

Copper(I)-catalyzed alkyne-azide cycloaddition

Copper(I)-catalyzed alkyne-azide cycloaddition

Strain-promoted alkyne-azide cycloaddition

Tetrazine ligation

Oxidative coupling

Figure 1.4: Examples of “click” methods for protein modification. Schema do not represent thefull scope of each reaction, and many variants exist.

a canonical amino acid. The bioconjugation community responded to the ability to addadditional functional groups to proteins by developing an entirely new class of reactionsthat were termed “bioorthogonal.”Such reactions minimally affect functional groups foundin common biomolecules and can be used in biological contexts with minimal disruption.Many bioorthogonal reactions are also referred to as click reactions because they are “mod-ular, wide in scope, give very high yields, and generate only inoffensive byproducts.”37

Saxon and Bertozzi developed the term “bioorthogonal ”to describe a new reaction inwhich the traditional Staudinger reduction is co-opted for the purposes of bioconjugation(Figure 1.4).38 In this reaction, the bioconjugation reagent bears a phosphine that firstreduces an organic azide. The resulting aza-ylide is trapped by an ester also present onthe reagent to give a water-stable amide adduct. The reaction was first used to investigateglycosylation patterns on cells, and it has also been modified to yield a bioconjugationproduct without a pendant phosphine group.39,40 However, the reaction takes place at modestrates, the reagents are poorly water soluble, and they can be difficult to handle because ofsusceptibility to oxidation in air.41,42


Perhaps the most well-known click reaction is the 1,3-dipolar cycloaddition between pri-mary alkynes and organic azides to form 1,2,3-triazoles. The reaction was popularized byK. Barry Sharpless, and relies on the presence of a Copper(I) catalyst.43 It is tolerant of anarray of different functional groups and can be carried out in water at room temperature.The kinetics of the reaction are quite fast, and the reaction can be found ubiquitously inmany fields, including protein modification. However, Copper(I) is cytotoxic, so this reactionhas limited use in biological contexts, and the copper salts can also be difficult to removefrom solutions of proteins. As a result, the reaction has found limited use within our lab.

To address the limitations of copper(I)-catalyzed alkyne-azide cycloadditions, Bertozziand coworkers developed a strain-promoted [3+2] azide–alkyne cycloaddition.42 The need fora catalyst in this reaction is obviated by incorporating an alkyne into an eight-membered ringto create ring strain. Further optimization of the reaction yielded a set of cyclooctynes withimproved water solubility and rates approaching 1 M−1 s−1.44–47 These reactions are com-patible with living cells, although they do show minor background reactivity with cysteines.Another click reaction that makes use of cyclooctynes is the tetrazine ligation, in which atrans-cyclooctyne reacts with tetrazine via an inverse-demand Diels Alder reaction.48 Afterthe initial reaction, nitrogen gas is liberated via an inverse Diels Alder reaction to afford astable product. The reaction has extraordinarily fast kinetics, with rate constants reportedat 2000 M−1 s−1.

Within the Francis group, effort toward click-type reactions has focused on oxidative cou-plings, in which anilines can be conjugated to a phenylenediamines, aminophenols, catechols,and anisidines under oxidizing conditions.49–52 The most common oxidants used are sodiumperiodate and potassium ferricyanide, with potassium ferricyanide exhibiting greater func-tional group tolerance. A light-activated variant of this reaction also allows conjugation ofazidophenols to anilines without oxidant.53 These reactions occur rapidly, with rate constantsof 10-100 M−1 s−1, and they can be used to conjugate bulky substituents onto proteins, oftenwith only a ten-fold excess of reagent. These reactions do exhibit some cross-creativity withamines and thiols, and they are not compatible with substrates that are easily oxidized.

1.3 Affinity-directed methods for modification of

proteins

All of the methods for protein modification described so far have been chemically directed—selective and efficient chemistry is used to target as few potential modification sites as possibleto create a well-defined bioconjugate. In practice, cases remain where the application of thesemethods is difficult. A notable example from the Francis lab is the modification of multi-meric proteins, particularly viral capsids.3,54 These proteins are often used to template dyes,serve as artificial light harvesting mimics, or form scaffolds for diagnostics and therapeutics.Because these proteins consist of a number of chemically identical subunits, it is impos-sible to control which subunits are modified, or the exact ratios of multiple modifications.


Improvements to the chemical selectivity, kinetics, or efficiency of existing bioconjugation re-actions will not help in such cases, because these methods cannot be influenced by a protein’squaternary structure. Modification of these protein targets requires methods that considerglobal protein structure, not simply the reactivity of a single modification site.

Several methods have already been developed that consider more than just the reactivityof a potential modification site. Notable examples are aptamer-based affinity labeling,55,56

ligand-directed modification,57 and His-tag-directed protein labeling.58,59 These methods usenoncovalent interactions to position a reagent close to a particular area of the protein surface.The reagent then reacts with the most probable—usually the closest—reactive site to forma stable covalent bond. These methods often use lysine chemistry because of the prevalenceof lysine residues, but they also commonly use photochemistry because of its ability to labelproteins at many locations. While these methods do not generalize well because the reagentsused are designed to target a specific site, they highlight the power of noncovalent chemistryin targeting chemical modification of protein surfaces.

This approach was recently shown to be more general by Rosen et al., who took ad-vantage of the intrinsic metal-binding character of some IgG1 proteins to direct chemicalmodifications.60 In this method, a DNA strand with a metal-binding moiety binds noncova-lently to a protein surface, usually in an area with a large surface concentration of histidine.A complementary DNA strand with an NHS-ester is then added, and it hybridizes withthe protein-bound template strand. Due to increased proximity to the protein surface, thisconstruct then reacts with a small number of lysines near the hybridization site to yield aprotein-DNA conjugate modified in a defined location.

An alternative method used by pharmaceutical companies exerts minimal control overthe location and degree of modification, but instead uses chromatography to select pro-teins with the desired number of modifications. This approach has been taken most oftenwith antibodies, which are traditionally very difficult bioconjugation targets because of theirlarge size, multiple disulfide bonds, and multiple chains. While chromatographic separa-tion of proteins that are as large as antibodies is quite challenging, hydrophobic interactionchromatography has been used in a few cases where antibodies are modified with largehydrophobic molecules.61,62 In these cases it is possible to separate proteins with differentnumber of chemical modification, but it is not possible to discriminate between modificationsites. An affinity-based version of this technique has also been developed by several groups,including ours.63,64 In these techniques, the broad affinity of cyclodextrins are used to purifyproteins that are modified with their ligands. Nguyen, Joshi, and Francis used this methodin a batch-mode to isolate proteins modified with certain dyes, and Chung et al. used thismethod to isolate prenylated proteins.


1.4 Second-order methods for protein modification

and their potential for development

With the recent advent of click chemistry and methods to modify protein N-termini, apowerful toolbox of reactions is available to the modern bioconjugator. These methodsprovide a set of chemistries with broad scope and dense coverage, wherein one method canbe used if another fails, and it is usually possible to find an effective modification strategyfor a given protein and its specific application.

These methods could be described as first-order, for they discriminate between modi-fication sites based on one main variable: chemical reactivity. Although such first-orderreactions cannot be used in nontraditional contexts like the controlled modification of pro-tein multimers, they provide a foundation on which to build more complicated second-ordermethods. For example, the method developed by Rosen et al. adds an additional selectivityconstraint by using noncovalent interactions to guide lysine acylation chemistry. The resultis a second-order method that has the broad applicability of lysine chemistry but with amuch narrower selectivity.

In the development of second-order methods for protein modification, affinity-based in-teractions are a natural choice to further control selectivity. Not only do affinity interactionsadd a new set of capabilities, but they provide a way to interact with proteins without funda-mentally changing them. A broad range of affinity interactions exist, with their occurrenceparticularly prevalent in biology. There exist a set of well-defined host-ligand pairs that areparticularly amenable to use with different contexts because of their modularity; these in-clude cyclodextrins,65,66 cucurbiturils,67,68 biotin-streptavidin, and calixarenes.69,70 Of these,cyclodextrins are well-suited for use with proteins.

1.5 Cyclodextrin and its applications

Cyclic polymers of glucose, the most common cyclodextrins are α-cyclodextrin, β-cyclodextrin,and γ-cyclodextrin, which are composed of six, seven, and eight 1,4-anhydroglucopyranosideunits, respectively. These molecules are toroidal in shape, but one face is larger than theother such that the molecule has roughly the same outer shape as a conical section. Cy-clodextrins are produced enzymatically from starch, and as a result their prices are quitelow, with α-cyclodextrin, β-cyclodextrin, and γ-cyclodextrin costing $20-25/kg, $3-4/kg, and$80-100/kg, respectively, in 2002.71 These low prices have enabled the use of cyclodextrinsin many fields on an industrial scale, and in 2002 the global consumption was 6000 metrictons.71

Broad interest in cyclodextrins derives from their amphiphilic character, wherein their in-ternal cavities are hydrophobic yet their exteriors are hydrophilic.65 The external hydrophilic-ity of cyclodextrins makes them water soluble, with β-cyclodextrin, the least water soluble,having a water solubility of 16 mM, and α-cyclodextrin and γ-cyclodextrin having solubili-ties in excess of 100 mM.65 In aqueous environments, cyclodextrins interact favorably with a


0 1 2 3 4 5 6 70

70

60

50

40

30

20

10

Log10(KA / M-1)

Cyclodextrin Ligand Binding Constants

α-Cyclodextrinβ-Cyclodextrinγ-Cyclodextrin

Num

ber

Figure 1.5: Histogram of cyclodextrin binding constants, produced from Rekharsky and Inoue.

broad range of hydrophobic molecules by complexing them within their hydrophobic cavities.These complexes usually have 1:1 cyclodextrin:ligand stoichiometries, although ratios of 1:2,2:1, and 2:2 have also been reported.65 The thermodynamics of these complexes have beenthoroughly investigated, with the binding constants and thermodynamics of several hundredligands in different solution conditions reported.66 Histograms of the binding constants be-tween cyclodextrins and their ligands (Figure 1.5) reveal that they range over four to fiveorders of magnitude. This dynamic range means that systems incorporating cyclodextrincan have their thermodynamics tuned through the use of different ligands. Moreover, thecavity diameters of α-cyclodextrin and β-cyclodextrin (5.2 A and 6.6 A, respectively) aresuch that molecules with the width of a benzene ring usually interact favorably.65 As a result,a wide variety of organic molecules bind to cyclodextrins, and to some extent it is possibleto rationally design ligands.

1.5.1 Chemical modification of cyclodextrins

Another feature of cyclodextrins that has enabled their use in many fields is their synthetictractability. These modifications roughly fit into two categories: modifications that changetheir properties and modifications to attach them to other functionalities.

The first category has been used primarily to change the water solubility and complex-ation thermodynamics of cyclodextrins. For example, randomly methylating or hydrox-ypropylating the hydroxyl groups of β-cyclodextrin significantly improves their solubility.72

Other modifications such as sulfation improve solubility as well. These modifications arethought to improve the solubility of β-cyclodextrin by disrupting the stacking of cyclodex-trins into long rods in aqueous solutions.73 Modification of the 6-position hydroxyl groups ofβ-cyclodextrin has also been used to study the recognition of ionic guests in cyclodextrins.74

This work resulted in the discovery of guests of modified β-cyclodextrin with unusually tightbinding constants.

Perhaps the most useful chemical modification of cyclodextrins is their monosubstitution.


One of the earliest reports of monosubstituted cyclodextrins came from Melton and Slessor,in which α-cyclodextrin was reacted with p-toluenesulfonylchloride in pyridine to give 67% ofthe monotosylated α-cyclodextrin. In practice this reaction tends to be difficult to performselectively, and there is commonly doubly modified cyclodextrin found in the product mix-ture. Fortunately, two similar methods have been developed in which β-cyclodextrin is mod-ified in basic water with either p-toluenesulfonylchloride or 1-(p-toluenesulfonyl)imidazole togive monosubstituted β-cyclodextrin in 25-40% yield.76–78 These reactions are quite reliable,and enable the generation of large amounts of well-defined chemically modified cyclodextrinson relatively large scale. The ability to monofunctionalize β-cyclodextrin makes possible thecontrolled incorporation of β-cyclodextrin and therefore affinity-based interactions into awide variety of contexts.

1.5.2 Uses of cyclodextrins

Because of their water solubility, synthetic tractability, and complexation of a large numberof small molecules, cyclodextrins have found use in a wide range of fields. Cyclodextrinsare nontoxic when consumed orally, and a number of chemically modified cyclodextrins alsoappear to be nontoxic when administered interperenterally.79 As a result, they have beenused in consumer and healthcare products to deliver drugs, increase the solubility of poorlysoluble drugs, mask unfavorable odors and smells, control the release of volatile fragrances,protect flavors, and remove cholesterol.79

Within the academic research community cyclodextrins have found many uses as well.A SciFinder search for “cyclodextrin” returns over 79000 publications, and new papers arecontinually appearing with new applications for these molecules. Cyclodextrins have beenused to mimic enzymes,80 control protein assembly,81 aid in DNA sequencing,82 form thebasis of chiral chromatography,79 create dendrimers,83 and add noncovalent interactions tomany other systems. In these contexts, it is the modularity of cyclodextrins that makesthem so useful because they can easily be incorporated into other systems without changingtheir properties. As a result, cyclodextrins have become a go-to chemical functionality foradding relatively weak but tunable noncovalent interactions to a system.

1.6 Project design and thesis overview

The majority of this work concerns the development of chromatography-mediated biocon-jugation, a second-order technique for protein modification. Specifically, control over thenumber of modifications and the location of those modifications is exerted by chromatog-raphy, as shown in Figure 1.6. In some cases, a desired bioconjugate cannot be accessedthrough traditional bioconjugation chemistry (Figure 1.6.i) because multiple potential mod-ification sites exist, or because the reaction results in side products. In these cases, analternative is to first modify the protein with a purification handle (Figure 1.6.ii) This mod-ification results in the same product mixture as with the desired bioconjugate. However, the


+ + +

i. TraditionalModification

iii. Affinity chromatography

iv. Modification

ii. Modification= puri�cation handle

+

+

+

Desiredconjugate

Figure 1.6: Overview of chromatography-mediated bioconjugation. (i) Modification with first ordertechniques results in an inseparable product mixture. (ii) Instead, a first-order technique is used toattach a chemical handle to a protein. (iii) This handle is used to purify the desired bioconjugate(second-order step), and (iv) the isolated bioconjugate is then modified using click chemistry.

purification handle has two properties that allow selective modification. First, the handleis a ligand of β-cyclodextrin and allows separation of the components of the crude reactionmixture when used in conjugation with a β-cyclodextrin-modified chromatography column(Figure 1.6.iii). After the desired bioconjugate is isolated, the purification handle can befurther elaborated using selective and efficient chemistry to access the desired bioconjugate(Figure 1.6.iv). Assuming that the chemistry in step iv is sufficiently efficient, this methodresults in a pure sample of protein with the desired modification attached to the proteinwith a short linker.

Chapter 2 discusses the theoretical concerns for the design of such a system, and it dis-cusses the use of this theory in additional contexts. Chapter 3 describes the development ofchromatography-mediated bioconjugation using traditional media for affinity chromatogra-phy. As an example of the capabilities of this technique, a well-defined light harvesting mimicis synthesized using a protein homotrimer as a template. Chapter 4 describes the develop-ment of chromatography-mediated bioconjugation using monolithic columns and highlightsthe advantages of monolithic columns. Finally, Chapter 5 describes the use of materials fromChapter 3 in an industrial context to recover enzymes from reaction mixtures.


1.7 References

1. James Clarke, Haichen Wu, Lakmal Jayasinghe, Alpesh Patel, Stuart Reid, and HaganBayley. “Nanopore DNA sequencing.” Nat. Nanotechnol. 4, 2009, 265–270.

2. Wesley Wu, Sonny C. Hsiao, Zachary M. Carrico, and Matthew B. Francis. “Genome-free viral capsids as multivalent carriers for taxol delivery.” Angew. Chem., Int. Ed.Engl. 48, 2009, 9493–7.

3. Ying-Zhong Ma, Rebekah A. Miller, Graham R. Fleming, and Matthew B. Francis.“Energy transfer dynamics in light-harvesting assemblies templated by the tobaccomosaic virus coat protein.” J. Phys. Chem. B 112, 2008, 6887–92.

4. Steven H. L. Verhelst, Marko Fonovic, and Matthew Bogyo. “A mild chemically cleav-able linker system for functional proteomic applications.” Angew. Chem., Int. Ed. Engl.46, 2007, 1284–6.

5. William P. Heal, T. H. Tam Dang, and Edward W. Tate. “Activity-based probes:discovering new biology and new drug targets.” Chem. Soc. Rev. 40, 2011, 246–257.

6. Yao-Wen Wu and Roger S. Goody. “Probing protein function by chemical modifica-tion.” J. Pept. Sci. 16, 2010, 514–523.

7. Nicholas Stephanopoulos, Gary J. Tong, Sonny C. Hsiao, and Matthew B. Francis.“Dual-surface modified virus capsids for targeted delivery of photodynamic agents tocancer cells.” ACS Nano 4, 2010, 6014–6020.

8. Adel M. El Sohly, Chawita Netirojjanakul, Ioana L. Aanei, Astraea Jager, Sean Ben-dall, Michelle E. Farkas, Garry P. Nolan, and Matthew B. Francis. “Syntheticallymodified viral capsids as versatile carriers for use in antibody-based cell targeting.”Bioconjug. Chem. 26, 2015, 1590–1596.

9. Charles A. Janeway Jr., Paul Travers, Mark Walport, and Mark J. Shlomchik. Immuno-biology: The Immune System in Health and Disease. 5th Editio. New York: GarlandScience, 2001.

10. Olive Leavy. “Therapeutic antibodies: past, present and future.” Nat. Rev. Immunol.10, 2010, 297.

11. Saurabh Aggarwal. “What’s fueling the biotech engine.” Nat. Biotechnol. 27, 2009,987–993.

12. Laurent Ducry and Bernhard Stump. “Antibody-drug conjugates: linking cytotoxicpayloads to monoclonal antibodies.” Bioconjug. Chem. 21, 2010, 5–13.

13. Asher Mullard. “Maturing antibody-drug conjugate pipeline hits 30.” Nat. Rev. DrugDiscov. 12, 2013.


14. Kevin J. Hamblett, Peter D. Senter, Dana F. Chace, Michael M.C. Sun, Joel Lenox,Charles G. Cerveny, Kim M. Kissler, Starr X. Bernhardt, Anastasia K. Kopcha, RogerF. Zabinski, Damon L. Meyer, and Joseph A. Francisco. “Effects of drug loading onthe antitumor activity of a monoclonal antibody drug conjugate.” Clin. Cancer Res.10, 2004, 7063–7070.

15. Jagath R. Junutula, Helga Raab, Suzanna Clark, Sunil Bhakta, Douglas D. Leipold,Sylvia Weir, Yvonne Chen, Michelle Simpson, Siao Ping Tsai, Mark S. Dennis, YanmeiLu, Y. Gloria Meng, Carl Ng, Jihong Yang, Chien C. Lee, Eileen Duenas, Jeffrey Gor-rell, Viswanatham Katta, Amy Kim, Kevin McDorman, Kelly Flagella, Rayna Venook,Sarajane Ross, Susan D. Spencer, Wai Lee Wong, Henry B. Lowman, Richard Van-dlen, Mark X. Sliwkowski, Richard H. Scheller, Paul Polakis, and William Mallet.“Site-specific conjugation of a cytotoxic drug to an antibody improves the therapeuticindex.” Nat. Biotechnol. 26, 2008, 925–932.

16. David S. Goodsell and Arthur J. Olson. “Structural symmetry and protein function.”Annu. Rev. Biophys. Biomol. Struct. 29, 2000, 105–53.

17. Tara L. Schlick, Zhebo Ding, Ernest W. Kovacs, and Matthew B. Francis. “Dual-surfacemodification of the tobacco mosaic virus.” J. Am. Chem. Soc. 127, 2005, 3718–3723.

18. Rodrigo Noriega, Daniel T. Finley, John Haberstroh, Phillip L. Geissler, Matthew B.Francis, and Naomi S. Ginsberg. “Manipulating excited-state dynamics of individuallight-harvesting chromophores through restricted motions in a hydrated nanoscale pro-tein cavity.” J. Phys. Chem. B, 2015, 6963–6973.

19. Allie C. Obermeyer, Stacy L. Capehart, John B. Jarman, and Matthew B. Francis.“Multivalent viral capsids with internal cargo for fibrin imaging.” PLoS One 9, 2014,e100678.

20. Krishnan K. Palaniappan, R. Matthew Ramirez, Vikram S. Bajaj, David E. Wemmer,Alexander Pines, and Matthew B. Francis. “Molecular imaging of cancer cells usinga bacteriophage-based 129Xe NMR biosensor.” Angew. Chemie - Int. Ed. 52, 2013,4849–4853.

21. Russell F. Doolittle. “Redundancies in protein sequences.” Predict. Protein Struct.Princ. Protein Conform. 1989. Ed. by Gerald D. Fasman, 599–623.

22. The Uniprot Consortium. “UniProt: a hub for protein information.” Nucleic Acids Res.43, 2014, D204–D212.

23. Greg T. Hermanson. Bioconjugate Techniques. Third Edit. Elsevier Inc., 2013.

24. Nicholas Stephanopoulos and Matthew B. Francis. “Choosing an effective protein bio-conjugation strategy.” Nat. Chem. Biol. 7, 2011, 876–884.

25. Shaun D. Fontaine, Ralph Reid, Louise Robinson, Gary W. Ashley, and Daniel V.Santi. “Long-term stabilization of maleimide-thiol conjugates.” Bioconjug. Chem. 26,2015, 145–152.


26. Joshua M. Gilmore, Rebecca A. Scheck, Aaron P. Esser-Kahn, Neel S. Joshi, andMatthew B. Francis. “N-terminal protein modification through a biomimetic transam-ination reaction.” Angew. Chemie - Int. Ed. 45, 2006, 5307–5311.

27. Leah S. Witus, Troy Moore, Benjamin W. Thuronyi, Aaron P. Esser-Kahn, Rebecca A.Scheck, Anthony T. Iavarone, and Matthew B. Francis. “Identification of highly reactivesequences for PLP-mediated bioconjugation using a combinatorial peptide library.” J.Am. Chem. Soc. 132, 2010, 16812–16817.

28. Kanwal S. Palla, Leah S. Witus, Katherine J. Mackenzie, Chawita Netirojjanakul, andMatthew B. Francis. “Optimization and expansion of a ste-selective N -methylpyridinium-4-carboxaldehyde-mediated transamination for bacterially expressed proteins.” J. Am.Chem. Soc. 137, 2015, 1123–1129.

29. Leah S. Witus, Chawita Netirojjanakul, Kanwal S. Palla, Ellen M. Muehl, Chih-HisangWeng, Anthony T. Iavarone, and Matthew B. Francis. “Site-specific protein transam-ination using N -methylpyridinium-4-carboxaldehyde.” J. Am. Chem. Soc. 135, 2013,17223–9.

30. James I. MacDonald, Henrik K. Munch, Troy Moore, and Matthew B. Francis. “One-step site-specific modification of native proteins with 2-pyridinecarboxyaldehydes.”Nat. Chem. Biol. 11, 2015, 326–331.

31. Allie C. Obermeyer, John B. Jarman, and Matthew B. Francis. “N-terminal modifica-tion of proteins with o-aminophenols.” J. Am. Chem. Soc. 136, 2014, 9572–9579.

32. Philip E. Dawson. “Native chemical ligation combined with desulfurization and dese-lenization: a general strategy for chemical protein synthesis.” Isr. J. Chem. 51, 2011,862–867.

33. Philip E. Dawson and Stephen B.H. Kent. “Synthesis of native proteins by chemicalligation.” Annu. Rev. Biochem. 69, 2000, 923–60.

34. J. Normanly, L. G. Kleina, J. M. Masson, J. Abelson, and J. H. Miller. “Construc-tion of Escherichia coli amber suppressor tRNA genes. III. Determination of tRNAspecificity.” J. Mol. Biol. 213, 1990, 719–726.

35. Hee-Sung Park, Michael J. Hohn, Takuya Umehara, Li-Tao Guo, Edith M. Osborne,Jack Benner, Christopher J. Noren, Jesse Rinehart, and Dieter Soll. “Expanding thegenetic code of Escherichia coli with phosphoserine.” Science 333, 2011, 1151–1154.

36. Jianming Xie and Peter G. Schultz. “Adding amino acids to the genetic repertoire.”Curr. Opin. Chem. Biol. 9, 2005, 548–554.

37. Hartmuth C. Kolb, M. G. Finn, and K. Barry Sharpless. “Click chemistry: diversechemical function from a few good reactions.” Angew. Chem. Int. Ed. Engl. 40, 2001,2004–2021.

38. E. Saxon and C. R. Bertozzi. “Cell surface engineering by a modified Staudinger reac-tion.” Science 287, 2000, 2007–2010.


39. Eliana Saxon, Sarah J. Luchansky, Howard C. Hang, Chong Yu, Sandy C. Lee, andCarolyn R. Bertozzi. “Investigating cellular metabolism of synthetic azidosugars withthe Staudinger ligation.” J. Am. Chem. Soc. 124, 2002, 14893–14902.

40. E. Saxon, J. I. Armstrong, and C. R. Bertozzi. “A ‘traceless’ Staudinger ligation forthe chemoselective synthesis of amide bonds.” Org. Lett. 2, 2000, 2141–2143.

41. Fiona L. Lin, Helen M. Hoyt, Herman Van Halbeek, Robert G. Bergman, and CarolynR. Bertozzi. “Mechanistic investigation of the Staudinger ligation.” J. Am. Chem. Soc.127, 2005, 2686–2695.

42. Nicholas J. Agard, Jennifer A. Prescher, and Carolyn R. Bertozzi. “A strain-promoted[3 + 2] azide-alkyne cycloaddition for covalent modification of biomolecules in livingsystems.” J. Am. Chem. Soc. 126, 2004, 15046–7.

43. Hartmuth C. Kolb and K. Barry Sharpless. “The growing impact of click chemistry ondrug discovery.” Drug Discov. Today 8, 2003, 1128–1137.

44. Ellen M. Sletten and Carolyn R. Bertozzi. “A hydrophilic azacyclooctyne for cu-freeclick chemistry.” Org. Lett. 10, 2008, 3097–3099.

45. Ellen M. Sletten, Hitomi Nakamura, John C. Jewett, and Carolyn R. Bertozzi. “Di-fluorobenzocyclooctyne: synthesis, reactivity, and stabilization by β-cyclodextrin.” J.Am. Chem. Soc. 132, 2010, 11799–11805.

46. Julian A. Codelli, Jeremy M. Baskin, Nicholas J. Agard, and Carolyn R. Bertozzi.“Second-generation difluorinated cyclooctynes for copper-free click chemistry.” J. Am.Chem. Soc. 130, 2008, 11486–93.

47. Chelsea G. Gordon, Joel L. Mackey, John C. Jewett, Ellen M. Sletten, K. N. Houk,and Carolyn R. Bertozzi. “Reactivity of biarylazacyclooctynones in copper-free clickchemistry.” J. Am. Chem. Soc. 134, 2012, 9199–208.

48. Melissa L. Blackman, Maksim Royzen, and Joseph M. Fox. “Tetrazine ligation: fast bio-conjugation based on inverse-electron-demand Diels-Alder reactivity.” J. Am. Chem.Soc. 130, 2008, 13518–13519.

49. J. M. Hooker, Aaron P. Esser-Kahn, and M. B. Francis. “Modification of aniline con-taining proteins using an oxidative coupling strategy.” J. Am. Chem. Soc. 128, 2006,15558–15559.

50. Christopher R. Behrens, Jacob M. Hooker, Allie C. Obermeyer, Dante W. Romanini,Elan M. Katz, and Matthew B. Francis. “Rapid chemoselective bioconjugation throughoxidative coupling of anilines and aminophenols.” J. Am. Chem. Soc. 133, 2011, 16398–16401.

51. Allie C. Obermeyer, John B. Jarman, Chawita Netirojjanakul, Kareem El Muslemany,and Matthew B. Francis. “Mild bioconjugation through the oxidative coupling of ortho-aminophenols and anilines with ferricyanide.” Angew. Chem. Int. Ed. Engl. 53, 2013,1057–61.


52. Adel M. ElSohly and Matthew B. Francis. “Development of oxidative coupling strate-gies for site-selective protein modification.” Acc. Chem. Res. 2015, 1971–1978.

53. Kareem M. El Muslemany, Amy A. Twite, Adel M. ElSohly, Allie C. Obermeyer,Richard A. Mathies, and Matthew B. Francis. “Photoactivated bioconjugation betweenortho-azidophenols and anilines: a facile approach to biomolecular photopatterning.”J. Am. Chem. Soc. 136, 2014, 12600–12606.

54. Rebecca A. Scheck and Matthew B. Francis. “Regioselective labeling of antibodiesthrough N-terminal transamination.” ACS Chem. Biol. 2, 2007, 247–51.

55. Jan L. Vinkenborg, Gunter Mayer, and Michael Famulok. “Aptamer-based affinitylabeling of proteins.” Angew. Chemie - Int. Ed. 51, 2012, 9176–9180.

56. Yoichiro Koshi, Eiji Nakata, Masayoshi Miyagawa, Shinya Tsukiji, Tomohisa Ogawa,and Itaru Hamachi. “Target-specific chemical acylation of lectins by ligand-tetheredDMAP catalysts.” J. Am. Chem. Soc. 130, 2008, 245–251.

57. Shinya Tsukiji, Masayoshi Miyagawa, Yousuke Takaoka, Tomonori Tamura, and ItaruHamachi. “Ligand-directed tosyl chemistry for protein labeling in vivo.” Nat. Chem.Biol. 5, 2009, 341–343.

58. Sho-hei Uchinomiya, Hiroshi Nonaka, Sho-hei Fujishima, Shinya Tsukiji, Akio Ojida,and Itaru Hamachi. “Site-specific covalent labeling of His-tag fused proteins with areactive Ni(II)-NTA probe.” Chem. Commun. (Camb). 2009, 5880–5882.

59. Gavin D. Meredith, Hayley Y. Wu, and Nancy L. Allbritton. “Targeted protein func-tionalization using his-tags.” Bioconjug. Chem. 15, 2004, 969–982.

60. Christian B. Rosen, Anne L. B. Kodal, Jesper S. Nielsen, David H. Schaffert, CarstenScavenius, Anders H. Okholm, Niels V. Voigt, Jan J. Enghild, Jørgen Kjems, ThomasTørring, and Kurt V. Gothelf. “Template-directed covalent conjugation of DNA tonative antibodies, transferrin and other metal-binding proteins.” Nat. Chem. 6, 2014,804–809.

61. Aditya Wakankar, Yan Chen, Yatin Gokarn, and Fredric S. Jacobson. “Analytical meth-ods for physicochemical characterization of antibody drug conjugates.” MAbs 3, 2011,161–172.

62. Lan N. Le, Jamie M. R. Moore, Jun Ouyang, Xiaoying Chen, Mary D. H. Nguyen, andWilliam J. Galush. “Profiling antibody drug conjugate positional isomers: a system-of-equations approach.” Anal. Chem. 84, 2012, 7479–7486.

63. Trung Nguyen, Neel S. Joshi, and Matthew B. Francis. “An affinity-based methodfor the purification of fluorescently-labeled biomolecules.” Bioconjug. Chem. 17, 2006,869–872.


64. Jinhwa A. Chung, James W. Wollack, Marisa L. Hovlid, Ayse Okesli, Yan Chen,Joachim D. Mueller, Mark D. Distefano, and T. Andrew Taton. “Purification of preny-lated proteins by affinity chromatography on cyclodextrin-modified agarose.” Anal.Biochem. 386, 2009, 1–8.

65. Kenneth A. Connors. “The stability of cyclodextrin complexes in solution.” Chem.Rev. 97, 1997, 1325–1358.

66. Mikhail V. Rekharsky and Yoshihisa Inoue. “Complexation thermodynamics of cy-clodextrins.” Chem. Rev. 98, 1998, 1875–1918.

67. Kimoon Kim, Narayanan Selvapalam, Young Ho Ko, Kyeng Min Park, Dongwoo Kim,and Jeeyeon Kim. “Functionalized cucurbiturils and their applications.” Chem. Soc.Rev. 36, 2007, 267–79.

68. Jason Lagona, Pritam Mukhopadhyay, Sriparna Chakrabarti, and Lyle Isaacs. “Thecucurbit[n]uril family.” Angew. Chem. Int. Ed. Engl. 44, 2005, 4844–70.

69. Julius Rebek Jr. “Host-guest chemistry of calixarene capsules.” Chem. Commun. 2000,637–643.

70. Angela F. Danil De Namor, Robert M. Cleverley, and Mariel L. Zapata-Ormachea.“Thermodynamics of calixarene chemistry.” Chem. Rev. 98, 1998, 2495–2526.

71. A. Biwer, G. Antranikian, and E. Heinzle. “Enzymatic production of cyclodextrins.”Appl. Microbiol. Biotechnol. 59, 2002, 609–617.

72. L. Szente and J. Szejtli. “Highly soluble cyclodextrin derivatives: chemistry, properties,and trends in development.” Adv. Drug Deliv. Rev. 36, 1999, 17–28.

73. Anthony W. Coleman, Ioannis Nicolis, Nelly Keller, and Jean Pierre Dalbiez. “Aggre-gation of cyclodextrins: an explanation of the abnormal solubility of β-cyclodextrin.”J. Incl. Phenom. Mol. Recognit. Chem. 13, 1992, 139–143.

74. Gerhard Wenz, Christian Strassnig, Carolin Thiele, Annegret Engelke, Bernd Morgen-stern, and Kaspar Hegetschweiler. “Recognition of ionic guests by ionic β-cyclodextrinderivatives.” Chemistry 14, 2008, 7202–11.

75. L. D. Melton and K. N. Slessor. “Synthesis of monosubstituted cyclodexaamyloses.”Carbohydr. Res. 18, 1971, 29–37.

76. Hoe-Sup Byun, Ning Zhong, and Robert Bittman. “6-O-p-toluenesulfonyl-β-cyclodex-trin.” Org. Syn. 77, 2000, 225.

77. Bernadette Brady, Nuala Lynam, Thomas O’Sullivan, Cormac Ahern, and RaphaelDarcy. “6-O-p-toluenesulfonyl-β-cyclodextrin.” Org. Syn. 10, 2000, 220.

78. Weihua Tang and Siu-Choon Ng. “Synthesis of cationic single-isomer cyclodextrinsfor the chiral separation of amino acids and anionic pharmaceuticals.” Nat. Protoc. 2,2007, 3195–200.


79. E. M. Martin Del Valle. “Cyclodextrins and their uses: a review.” Process Biochem.39, 2004, 1033–1046.

80. Ronald Breslow and Steven D. Dong. “Biomimetic reactions catalyzed by cyclodextrinsand their derivatives.” Chem. Rev. 98, 1998, 1997–2012.

81. Dana A. Uhlenheuer, Dorothee Wasserberg, Hoang Nguyen, Li Zhang, Christian Blum,Vinod Subramaniam, and Luc Brunsveld. “Modulation of protein dimerization by asupramolecular host-guest system.” Chemistry 15, 2009, 8779–90.

82. Hai-Chen Wu, Yann Astier, Giovanni Maglia, Ellina Mikhailova, and Hagan Bayley.“Protein nanopores with covalently attached molecular adapters.” J. Am. Chem. Soc.129, 2007, 16142–8.

83. Hassan Namazi and Abolfazl Heydari. “Synthesis of β-cyclodextrin-based dendrimeras a novel encapsulation agent.” Polym. Int. 63, 2014, 1447–1455.

22

Chapter 2

Theory of chromatography

Abstract

Any chromatographic system relies on the balance of a number of parameters whose opti-mization can be laborious. To determine whether affinity chromatography could be used toseparate proteins based on their degree of chemical modification, several models of affinitychromatography were constructed by combining pre-existing methods from the literature.These methods were used to determine the optimal column loading and binding constant.Additional features, including the identity of the eluent, were then considered to aid withengineering the actual system.

Haiku

Crunch text, fan spinningFlowing the ones and zerosYour column should work

CHAPTER 2. THEORY OF CHROMATOGRAPHY 23

2.1 Motivation for theoretical calculations

The development of any chromatographic method involves the optimization of a numberof parameters under constraints. In the case of chromatography-mediated bioconjugation,the primary concern was to design a system that was capable of separating proteins withdifferent numbers of chemical modifications based on an affinity interaction with the chemicalmodification. This method needed to be compatible with many different types of proteinsin aqueous media. While this sort of system could have been designed experimentally, wedecided to model this system theoretically to determine whether this project was feasible andto understand the effects of various changes during development. Modeling lead to a moredirected development process, and also enabled the exploration of several different types ofsystems that will be described in Section 2.5.

2.2 Existing theories of chromatography

One of the first theoretical treatments of column chromatography was performed by Martinand Synge, wherein a chromatographic column was treated as a series of plates.1 In thistreatment, each plate in the column is assumed to be at equilibrium, and diffusion betweenplates is assumed to be negligible. Solutes within the column were also assumed not tointeract with one another and to be in the linear range of their binding curves, such that theproportion of solute bound to the stationary phase was always constant.

These simplifying assumptions lead to a tractable analytical model in which solutes iter-atively travel from plate to plate. During each iteration, a constant fraction of the materialin the ith plate moves to the (i+ 1)th plate. This fraction is defined by δv/V , where δv is aninfinitesimal volume that has flowed past and V , the total column void volume, is definedas

V = h(AL − αAS) (2.1)

where AL is the cross-sectional area of the mobile phase, AS is the cross-sectional area of thestationary phase, h is the height of a plate, and α is the partition coefficient, defined as theratio of solute bound to the stationary phase to solute in the mobile phase. If we assumethat the amount of material in the first plate is 1 at time 0, then after nδv volumes havepassed, the amount of material in the rth plate is given by

Qr+1 =n!(1 − δv/V )n−r(δv/V )r

r!(n− r)!(2.2)

This expression is equivalent to the binomial distribution, and finding its maximal valuegives the distance that the solute has traveled. The ratio of the distance that the peak hastraveled to the distance that the solvent front has traveled (R) is given by

R =AL + AS + AI

AL + αAS

(2.3)


where AI is the inaccessible dead-volume. Although quite simple, this theory captures thekey aspects of chromatography. It is able to predict when a compound will elute from thecolumn, that bands will spread out as they travel through the column, and that their shapeswill be roughly Gaussian. Because of its simplicity, this model helps build an intuitiveunderstanding of chromatography and provides a minimum working example upon which tobuild a more complicated model.

A more nuanced view of affinity chromatography was developed by Arnold, Schofield, andBlanch to account for nonlinearity of partition curves.2 In this theory, the elution volume ofa pulse of solute is considered to be

Ve = V0 + KavVs (2.4)

where Ve is the elution volume, V0 is the column void volume, Kav is the partition coefficient,as defined above, and Vs is the volume of column solids. This term is equivalent to thedenominator in the equation derived by Martin and Synge above (Eq. 2.3). This expressioncan be broken down into a more detailed view by considering that solute can be eitheradsorbed to the surface of particles, or trapped within their pores. Taking this partitioninginto account, and then substituting the Langmuir equilibrium relation, Arnold, Schofield,and Blanch obtain

Ve =

(ε+ (1 − ε)β +

(1 − ε)ρpQmaxKA

1 +KAcp

)v (2.5)

where ε is the void fraction of the column, β is the void proportion of particles, ρp is theparticle density in units of g L−1, Qmax is the number of sites in units of mmol/g, KA is thebinding constant between a ligand L and a solid-phase binding site P , defined as

KA =[LP ]

[L][P ](2.6)

and cp is the concentration of P in the liquid in units of M. This model adds the next layer ofcomplexity to the model developed by Martin and Synge because it does not assume linearpartitioning of the analyte between the solid and liquid phase. As a result this model candiscern how elution from a column will change when the concentration of analyte approachesthe concentration of binding partner on the solid phase. This knowledge can be used todetermine the practical limits of a column.

Many more models of chromatography have been developed that account for more compli-cated features of columns such as porosity, particle shape, diffusion, and kinetics. However,each further refinement results in successively less intuitive understanding for a considerablylarger amount of computational work. As we will see, the work of Martin and Synge andArnold, Schofield, and Blanch is sufficient to develop a rich model of affinity chromatographythat can guide the physical design of a system.


2.3 Basic feasibility of chromatography-mediated

bioconjugation

The key step in chromatography-mediated bioconjugation is to purify proteins based on thenumber of ligands to which they are conjugated using a resin onto which is immobilizedβ-cyclodextrin. A rough metric for the success of the chromatography step is that singlymodified proteins should elute at a volume that is at least twice the volume at which un-modified proteins elute (i.e., the void volume). The expression derived by Arnold, Schofield,and Blanch can be used to evaluate the conditions under which this separation is possible.

The first two terms, ε + (1 − ε)β, in the parentheses in Eq. 2.5, sum to give the voidvolume, and this expression is usually between 0.5 and 1 given typical values for ε and β.3

The last term within the parenthesis provides the additional retention volume that resultsfrom binding to the solid phase. Because the first two terms are close to 1, this last termgives the ratio of the retention of singly modified proteins to the retention of unmodifiedproteins, and it will be defined as the retention ratio (∆r).

∆r =(1 − ε)ρpQmaxKA

1 +KAcp

(2.7)

Substitution of reasonable values into this expression should outline the regions of parameterspace for which chromatographic separation is possible. Proteins used within laboratorysettings are usually at concentrations around 1 µM, and almost never above 100 µM. Becauseeq. 2.5 is sensitive to large values of cp, a value of 100 µM was used to serve as an upperlimit. Based on reported literature values of ε for Sepharose, a resin commonly used forprotein chromatography, a value of 0.5 was used.3 The resulting expression is a function oftwo variables: ρpQmax, the concentration of β-cyclodextrin on the resin, and KA, the bindingconstant between the ligand and β-cyclodextrin (see eq. 2.7).

A heat map showing the value of this expression was generated for binding constants of101 M−1 to 106 M−1, a range that reflects the possible binding constants based on Figure 1.5,and for resin loadings between 1 µM and 10 mM, a range that reflects typical modificationlevels of Sepharose found in the product literature. The result (Figure 2.1) indicates param-eter ranges for which chromatographic separation should be possible. The black line showsthe contour along which ∆r = 1, and points below this line should have better separation.Given that most ligands of β-cyclodextrin have KA values between 102 M−1 and 104 M−1,and a resin loading of around 1 mM is reasonable, the area highlighted by the dashed linesshould provide a chromatographic system with the desired properties. As we will see laterin Chapter 3, these predictions prove correct.

2.4 Plate-based model of affinity chromatography

Having predicted the feasibility of chromatography-mediated bioconjugation, we sought amodel to understand how factors such as column length, gradient, and multiple modifications


1 2 3 4 5 6log10(Ka / M)

−6.0

−5.5

−5.0

−4.5

−4.0

−3.5

−3.0

−2.5

−2.0

log 10

(CD

Con

cent

ratio

n / M

)

Values of ∆r

0.0001

0.001

0.01

0.1

1

10

100

1000

10000

Figure 2.1: Heatmap of ∆r as a function of binding constant and resin loading. The black linedenotes the contour along which ∆r = 1, and the region denoted by dashed lines is physicallyreasonable for common binding constants to β-cyclodextrin and resin loadings of commonly usedresins.

affect retention. There exist many analytical models that can answer these questions, butthey tend to be mathematically difficult and inflexible to changes.4–6 On the other hand,simulation-based approaches are less elegant, but they provide a flexible context in which totest the affects of many parameters. Moreover, the increasing speed of personal computers,the availability of on-demand computing resources from companies like Amazon, and theavailability of a number of linear algebra libraries makes this computation accessible.

2.4.1 Kinetic binding model

The conceptually simplest treatment of affinity chromatography is to use the plate model ofMartin and Synge that includes an explicit treatment of binding equilibria. Martin and Syngeassumed that the partitioning between the mobile and stationary phases could be describedby a constant partition coefficient, but this treatment neglects the loading of the columnand disallows the possibility of binding interactions with stochiometries different from 1:1.However, from a kinetic perspective these interactions are conceptually quite simple, andthey can be modeled as sets of first order differential equations.

For the sake of simplicity, we will assume that we are analyzing a protein with a maximum


of two ligands on its surface. As a result, the following equilibria are possible:

P + R PR (2.8)

P + S PS (2.9)

PS + S PS,S (2.10)

PR + R PR,R (2.11)

PR + S PR,S (2.12)

PS + R PR,S (2.13)

where P is free protein, R is a binding unit (β-cyclodextrin) on the resin, S is the competingbinding unit in solution (i.e., free β-cyclodextrin), and the subscripts denote what each of thetwo binding sites is bound to. For the purpose of this analysis, PR,S and PS,R are identical.

Assuming that proteins are dilute and there are no geometric constraints on the binding,these equilibria can be modeled as follows:

∂P

∂t= − kon,1PR + koff,1PR − kon,SPS + koff,SPS (2.14)

∂PR

∂t=kon,1PR− koff,1PR − kon,2PRR + koff,2PR,R − kon,SPRS + koff,SPR,S (2.15)

∂PR,R

∂t=kon,2PRR− koff,2PR,R (2.16)

∂PS

∂t=kon,SPS − koff,2PS − kon,SPSS + koff,SPS,S − kon,1PSR + koff,1PR,S (2.17)

∂PS,S

∂t=kon,2PSS − koff,2PS,S (2.18)

∂PR,S

∂t=kon,SPRS − koff,SPR,S + kon,1PSR− koff,1PR,S (2.19)

∂R

∂t= − kon,1PR + koff,1PR − kon,1PSR + koff,1PR,S − kon,2PRR + koff,2PR,R (2.20)

∂S

∂t= − kon,SPS + koff,SPS − kon,SPRS + koff,SPR,S − kon,SPRS + koff,SPR,S (2.21)

where the italicized variables represent the concentrations of the corresponding species, kon,1

and koff,1 are the on and off rates for equation 2.8, kon,2 and koff,2 are the on and off rates forequation 2.11, and kon,S and koff,S are the on and off rates for equation 2.9

After taking into account the kinetics of binding, the only remaining consideration is theflow of material from plate to plate. Following the lead of Martin and Synge, the dependenceof the concentration of material in plate i as a function of time is described by

∂Qi

∂t=

u

Vp(Qi−1 −Qi) (2.22)

where Q is the concentration of an arbitrary species, u is the volumetric flow rate of thecolumn, and Vp is the volume of each plate. Moreover, a gradient of the eluent S can be


specified by adding a term to the rate equation of S in the first plate. These flow-dependentterms can be added to the equilibrium terms described above to yield a full model for affinitychromatography.

This model was implemented in IPython7 using the numerical integration package odeintfrom SciPy,8,9 and the results were visualized with Matplotlib.10 The source code is availablein Section A.1.

2.4.2 Analytic binding model

While the model described above is conceptually simple, it is tedious to implement and doesnot generalize well for species with more than a few ligands. Moreover, it is computationallyslow because each species requires its own differential equation, and the number of speciesincreases as the number of ligands on the protein increases.

It is possible to reduce the complexity of the model by analytically calculating the concen-tration of bound and unbound protein. Nichol, Ward, and Winzor showed that a satisfactorysolution can be found under acceptable approximations.11 They began by assuming that aprotein could independently bind to a maximum of f sites, either on a solid-phase matrixor in solution. In the case of protein binding to sites on a resin, this assumption meant thatthere were a series of f equations wherein

(PR)i + R (PRi+1) (2.23)

where

KA =(PRi+1)

(PRi)R(2.24)

The binding constant for each additional binding event is assumed to be same, and thereforethe cooperativity12 is assumed to be zero. Using reacted-site probability theory, it can beshown that

Psolution =2−f P

fKAP

(q +

√−4fKAP (−1 −KSS) + q2

)f

(2.25)

where q = −1+fKAP −KAR−KSS, an overbar denotes the total concentration of a speciesin a plate, and KS is the binding constant between the protein and the eluting species insolution. In the above expression, S is the concentration of eluting species in solution, but incases where the protein concentration is significantly lower than the eluent concentration, Scan be replaced by S. As will be seen in Chapter 3, typical concentrations of protein are inthe micromolar range, whereas typical concentrations of eluent are in the millimolar range,so this approximation is reasonable.

With this result in-hand, the algorithm described in Section 2.4.1 becomes a two-stepalgorithm. In the first step, each plate is equilibrated and the concentrations of all speciesare determined. In the second step, the column is flowed according to Eq. 2.22. These twosteps are repeated until the desired elution volume has passed through the column.

This model was implemented using the same packages as the one described in Sec-tion 2.4.1, and the source code is available in Section A.2.


2.5 Applications

Much of the work that follows was performed in tandem with the experimental work thatis described in Chapter 3. As a result, this section will reference Chapter 3 extensively, andmany of the parameters used in the models will reflect values that were experimentally deter-mined at the time. Unless otherwise noted, the same default parameters were used for eachsimulation. Specifically, columns were modeled to have a volume of 4 mL with dimensionsof 25 cm by 0.5 cm inner diameter. Based on preliminary work with Sepharose, a value of2.2 plates per cm was used, giving each column 55 plates. The loading of β-cyclodextrinon the resin was assumed to be 1 mM. Binding constants between singly modified proteinsand β-cyclodextrin were set to 103.5 M−1. In some contexts this binding constant will bereferred to as KA,P because it refers to the binding constant between the 1:1 protein:ligandconjugate and β-cyclodextrin. Columns were assumed to be loaded with 100 µM of proteinin their first plate. In the case of the column described above, this loading corresponds to36 µL of protein.

2.5.1 Choice of eluent

A first consideration when designing chromatography-mediated bioconjugation was whateluent to use. Pre-existing methods13,14 to purify proteins modified with ligands of β-cyclodextrin used either ligands of β-cyclodextrin (1-adamantane carboxylic acid) or salts(NH4Cl) as eluents. However, it was not clear that these choices would produce the bestresults, and there was little precedent for using β-cyclodextrin itself as an eluent. Moreover,initial attempts at using salts to elute proteins proved unsuccessful, and the use of competingligands did not produce satisfactory results.

Rather than extensively test individual eluents, two plate models with kinetic bindingmodels were used to understand the effects of eluent choice. In the case of β-cyclodextrin aseluent, a linear gradient of β-cyclodextrin was applied to the column between 2 and 27 mL,and the final concentration of the β-cyclodextrin was varied between 100 µM and 100 mM.This range is reasonable considering that the solubility limit of β-cyclodextrin is 16 mM.

In the case of a competing ligand as eluent, the parameters of interest were both thebinding constant of the ligand (here defined as KA,CL) and its concentration. First, a lineargradient of 0 to 10 mM was applied to the column between 2 and 27 mL and KA,CL wasvaried between 101 M−1 and 105 M−1. The final concentration of the ligand was based onthe β-cyclodextrin loading of the resin. At concentrations around 1 mM and below, there isnot enough ligand to saturate the β-cyclodextrin on the column, and therefore the eluent isineffectual. A concentration of 10 mM is enough to effect elution of proteins, yet reasonablylow when considering that many ligands of β-cyclodextrin are poorly water soluble. Basedon these results, an optimal KA,CL of 102.75 M−1 was used to construct an additional heatmapin which the final concentration of ligand was varied between 100 µM and 100 mM.

The results of these calculations were stacked into two-dimensional arrays, with the x -axis representing the elution volume and the y-axis representing the variable being changed.


0 5 10 15 20 25 30Volume / ml

−4.0

−3.5

−3.0

−2.5

−2.0

−1.5

−1.00 5 10 15 20 25 30

Volume / ml

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0lo

g 10(K

a / M

-1)

log 10

(�na

l [CD

] / M

-1)

log 10

(�na

l [lig

and]

/ M

-1)

i. β-CD eluent ii. Competing ligand iii. Competing ligand

0 5 10 15 20 25 30Volume / ml

−4.0

−3.5

−3.0

−2.5

−2.0

−1.5

−1.0

Figure 2.2: Heatmaps showing elution of unmodified, singly-modified, and doubly-modified proteinswith different eluents. The left-most line corresponds to unmodified protein, the middle line corre-sponds to singly-modified protein, and the right-most line corresponds to doubly-modified protein.(i) Heatmap showing variation of β-cyclodextrin concentration between 100 µM and 100 mM. Thedashed line represents the solubility limit of β-cyclodextrin. (ii) Heatmap showing variation of theKA of the competing ligand. (iii) Heatmap showing the variation of the final concentration of acompeting ligand with a KA of 102.75 M−1.

In the case of β-cyclodextrin as eluent (Figure 2.2.i), increasing eluent concentrations leadto smaller elution volumes, with the elution volumes varying smoothly as a function ofconcentration. This behavior indicates that elution can be tuned by varying only the final β-cyclodextrin concentration. Moreover, because β-cyclodextrin serves as both the eluent andthe affinity element on the matrix, the ratio of the on-resin and in-solution binding constantsstays constant when different ligands are conjugated to the protein or when conjugation ofthe ligand to the protein changes the binding constant. As a result, the qualitative shape ofthe elution profiles in Figure 2.2.i stays the same regardless of the ligand used.

Using a competing ligand as eluent results in more complicated behavior. When lookingat elution volume as a function of KA,CL, the peaks corresponding to singly and doubly mod-ified protein converge to a point, and both elute together after the KA,CL exceeds the KA,P

(Figure 2.2.ii). This convergence results from the fact that the column becomes completelysaturated with competing ligand when KA,CL exceeds KA,P. As a result, the competingligand must be selected such that KA,CL is always less than KA,P. But once a ligand withan appropriate KA,CL is selected—in this case 102.75 M−1—the retention can be smoothlyadjusted by varying the final concentration of competing ligand (see Figure 2.2.iii).

Both paradigms can be used to achieve the desired elution characteristics, but elutionwith a competing ligand is more complicated because it depends on a parameter that inturn depends on KA,P. For this reason β-cyclodextrin was chosen to be the eluent. It shouldbe noted, however, that in cases where binding of the protein-ligand conjugate to the resinis stronger than binding to free β-cyclodextrin, elution with β-cyclodextrin becomes lesseffective because a concentration higher than 16 mM cannot be used. This type of behavior


[Pro

tein

]

02

4

6

810

[BCD

] / m

M

[Pro

tein

]

02

4

6

810

[BCD

] / m

M

0 20 40 60 80 100 120 140Volume / ml

[Pro

tein

]

02

4

6

810

i. Separation of multiply modi�ed species; 55 plates

ii. Separation of multiply modi�ed species; 55 plates, slower gradient

ii. Separation of multiply modi�ed species; 250 plates

[BCD

] / m

M

Figure 2.3: Simulation of the elution of a mixture of proteins with up to 20 modifications. (i)Elution with a 55 plate column separates three modification levels. (ii) Elution with a more gradualgradient gives better resolution, but at the cost of broader peaks. (iii) Use of a column with 250plates results in significantly improved resolution.

is seen to some extent in Chapter 3 and to a larger extent in Chapter 4. While use of acompeting ligand has not yet been investigated in Chapter 4, it could enable faster elutionof highly modified proteins.

2.5.2 Potential to separate multiple modification levels

Another concern when beginning this project was whether it would be possible to separateproteins based on the valency of their modification, and to what extent it would be possibleto distinguish between modification levels. Work with the kinetic model described in Sec-tion 2.5.1 indicated that it would indeed be possible to separate singly and doubly modifiedproteins. However, it is common to modify proteins more than two times, and many pro-teins also exist as homomultimers with many individual subunits. For example, the tobacco


mosaic virus coat protein that is commonly used in the Francis group forms disks of 17proteins, and 41% of the proteins in E. coli are found as trimers or higher oligomers.15 Asdiscussed in Section 3.1, controlled modification of these types of protein is of great interest,yet practically impossible.

The model described in Section 2.4.2 was used to explore the ability of this techniqueto separate proteins with between 0 and 20 modifications. Based on the parameters of thesystem designed in Chapter 3 it is reasonable to separate proteins with up to 3 chemicalmodifications (Figure 2.3.i). Separation can be improved by running a slower gradient (Fig-ure 2.3.ii), but this strategy gives only a marginal increase in separation at the expense ofsignificantly broadening more highly modified peaks. A more fruitful method to increase thenumber of modification levels that can be resolved is to increase the number of theoreticalplates. This feat can be accomplished by changing the chromatographic medium, as dis-cussed in Chapter 4. A similarly shaped column with a monolithic silica architecture wouldhave at least 250 plates, and it could separate up to about 9 chemical modifications usingthe gradient shown (Figure 2.3.iii). In practice, the separation of multiple modification onlarge oligomeric proteins may be limited by their large size and small diffusion coefficients.16

2.5.3 Application to lateral flow assays

Chromatography pervades the fields of organic chemistry, biochemistry, and biology. Thus,the ability to model chromatography is useful in many different contexts. One such contextis the design of lateral flow assays (LFAs).17,18 The most widely known LFA is the homepregnancy test, in which the presence of human chorionic gonadotropin (hCG) is detected ina urine sample. These tests are essentially specialized thin-layer chromatography systems,often made out of nitrocellulose, onto which various affinity molecules have been patterned.In a typical LFA, a sample is added to a sample pad that contains a recognition elementthat consists of a colored species like a nanoparticle.17 This mixture of species is then wickedalong the test strip, and they usually pass two lines that are labeled with affinity groups.One line represents a control line that shows that the assay is working, and the other linerepresents a test line that indicates the presence of the analyte. These lines concentratethe recognition element based on affinity interactions with the analyte and the recognitionelement, and the results of the test are visually read by the presence or absence of coloron each line. LFAs have been used in many contexts, including the detection of bacteria,their toxins, viruses, drug residues, and hormones.18 These tests are often low in price andproduce results in a matter of minutes.

Within the context of our lab, we are interested in using LFAs to detect the presence ofendocrine disruptors, particularly estrogenic compounds. Endocrine disruptors are moleculesthat have similar shapes to hormones found in the human body, and therefore they have thepotential to interact with the endocrine system.19,20 These interactions may cause disorderssuch as low semen quality, genital malformation, cancers, early puberty, obesity, and behav-ioral disorders.20 There are almost 800 known endocrine disruptors, yet their detection is not


Flow

Signal

Flow

FlowFlow

Flow Flow

Background

i. Positive test ii. Negative test

Figure 2.4: Schematic diagram of LFA. (i) Example of a positive result. (ii) Example of a negativeresult.

trivial.20 A low-cost, fast LFA for the detection of endocrine disruptors would be of greatuse.

As a proof of concept, a minimum working example of an LFA was constructed (Fig-ure 2.4). In this LFA, estrogen receptors are conjugated along a narrow strip. Estradiol-coumarin conjugates are then pre-bound to the estrogen receptors. When a sample of en-docrine disruptor is applied to the test strip (Figure 2.4.i), this analyte disrupts the dye-protein complexes and elutes the dye from the strip. The movement of this dye is thendetected as a signal. However, during the initial development of this method, it was notedthat the application of a negative control also resulted in the elution of some dye (Fig-ure 2.4.ii). The presence of this background decreased the overall sensitivity of the assay bylowering the turn on rate.

To better understand how to increase the signal to background ratio and decrease thelimit of detection, this system was modeled computationally (see Section 2.6 for details).The assay was simulated for two different protein concentrations (40 µM and 1 µM), and thesignal-to-background ratio was calculated as a function of both analyte concentration andthe fraction of protein sites initially occupied by dye. The results are plotted as heatmaps(Figure 2.5).

Two key points stand out from these calculations. First, decreasing the initial occupancyof the protein on the LFA increases the signal-to-background ratio. This result makes in-tuitive sense if one considers that the fraction of ligand bound decreases as more ligand isadded. The protein saturates asymptotically, meaning that the concentration of unbounddye is higher for higher dye loadings. Therefore, when one equivalent of dye is bound to the


-9.0

0

-8.4

7

-7.9

5

-7.4

2

-6.8

9

-6.3

7

-5.8

4

-5.3

2

-4.7

9

-4.2

6

0.05

0.16

0.26

0.37

0.47

0.58

0.68

0.79

0.89

1.00

Frac

tion

occu

pied

-9.0

0

-8.4

7

-7.9

5

-7.4

2

-6.8

9

-6.3

7

-5.8

4

-5.3

2

-4.7

9

-4.2

6

log10([Analyte] / M)log10([Analyte] / M)

0.05

0.16

0.26

0.37

0.47

0.58

0.68

0.79

0.89

1.00

Frac

tion

occu

pied

1.00

1.85

3.41

6.31

11.66

21.54

39.81

73.56

135.94

251.19i. Signal-to-background for 40 μM protein ii. Signal-to-background for 1 μM protein

Figure 2.5: Results of signal-to-background calculations for LFA as a function of analyte concentra-tion and the fraction of proteins that are bound with an estradiol-dye conjugate. (i) Calculationsfor 40 µMprotein. (ii) Calculations for 1 µM protein. Color scale is logarithmic.

protein on the strip, some will always wash off regardless of the presence of the analyte. Thisundesired leakage can be reduced by adding less estradiol-coumarin to the test strip, and inthe case of a 40 µM protein loading, even a small (10%) decrease in loading leads to a largeincrease in the signal-to-background ratio. The second feature of these heatmaps is that thelimit of detection (as measured by when the signal-to-background ratio is greater than 1)depends on the protein concentration. This result again makes intuitive sense because oneestradiol-dye molecule is displaced for every molecule of analyte. As a result, the proteinand analyte concentrations need to be around the same order of magnitude to result in anoticeable signal. However, the lower limit of detection is limited by the binding constantbetween estradiol-coumarin and the protein. At concentrations less than about 1 µM, thedye falls off the protein as a result of being too dilute, and the background increases signif-icantly. It is for this reason that the signal-to-background ratios are significantly lower forthe LFA with 1 µM protein.

This type of analysis has been used to explore other questions related to LFAs, includingwhether other formats would provide lower limits of detection and to determine the limit ofdetection for various binding constants. This work is ongoing and will not be reported here.


2.5.4 Web-based simulation

The model described in Section 2.4.2 is simple enough that a single run can be accomplishedon a personal computer in about 100 ms. This speed is fast enough that this type ofmodel could reasonably be implemented within a web browser in JavaScript such that itis accessible to anyone with a modern web browser. Too seldom are scientific results andtools made accessible to a wide range of people, and often scientific software is difficult touse because it is complicated and has many dependencies. While some of these difficultiesare unavoidable, the development of software for the web ensures that almost everyone witha computer can use the software. As a result, the model described in Section 2.4.2 wastranslated into JavaScript and is accessible via the internet at chartograph.com.

2.6 Materials and methods

Code used to simulate LFAs can be found in Section A.3. Parameters used for the simulationare listed below:

Kd between coumarin-estradiol and the protein: 20 nMKd between analyte and the protein: 1 nMStopping volume for the flow was twice the column volume.

2.7 References

1. J. P. Martin and R. L. M. Synge. “A new form of chromatogram employing two liquidphases.” Biochem. J. 35, 1941, 1358–1368.

2. F. H. Arnold, S. A. Schofield, and H. W. Blanch. “Analytical affinity chromatography I.local equilibrium theory and the measurement of association and inhibition constants.”J. Chromatogr. 355, 1986, 1–12.

3. P.M. Boyer and J.T. Hsu. “Experimental studies of restricted protein diffusion in anagarose matrix.” AIChE J. 38, 1992, 259–272.

4. F. H. Arnold, H. W. Blanch, and C. R. Wilke. “Liquid chromatography plate heightequations.” J. Chromatogr. 330, 1985, 159–166.

5. F.H. Arnold and H.W. Blanch. “Analytical affinity chromatography II. rate theory andthe measurement of biological binding kinetics.” J. Chromatogr. 355, 1986, 13–27.

6. D. A. Bergman and D. J. Winzor. “Quantitative affinity chromatography: increasedversatility of the technique for studies of ligand binding.” Anal. Biochem. 153, 1986,380–6.

7. Fernando Perez and Brian E. Granger. “IPython: a system for interactive scientificcomputing.” Comput. Sci. Eng. 9, 2007, 21–29.

http://www.chartograph.com/chroma


8. Eric Jones, Travis Oliphant, Pearu Peterson, et al. SciPy: Open Source Scientific Toolsfor Python.

9. Stefan van der Walt, S. Chris Colbert, and Gael Varoquaux. “The NumPy array: astructure for efficient numerical computation.” Comput. Sci. Eng. 13, 2011, 22–30.

10. John D. Hunter. “Matplotlib: a 2D graphics environment.” Comput. Sci. Eng. 9, 2007,90–95.

11. L. W. Nichol, L. D. Ward, and D. J. Winzor. “Multivalency of the partitioning speciesin quantitative affinity chromatography. Evaluation of the site-binding constant forthe aldolase-phosphate interaction from studies with cellulose phosphate as the affinitymatrix.” Biochemistry 20, 1981, 4856–60.

12. Vijay M. Krishnamurthy, Lara A. Estroff, and George M. Whitesides. “Multivalencyin ligand design.” Fragm. Approaches Drug Discov. Ed. by R Mannhold, H Kubinyi,and G Folkers. WILEY-VCH Verlag GmbH & Co., 2006. Chap. 2, 11–53.

13. Trung Nguyen, Neel S. Joshi, and Matthew B. Francis. “An affinity-based methodfor the purification of fluorescently-labeled biomolecules.” Bioconjug. Chem. 17, 2006,869–872.

14. Jinhwa A. Chung, James W. Wollack, Marisa L. Hovlid, Ayse Okesli, Yan Chen,Joachim D. Mueller, Mark D. Distefano, and T. Andrew Taton. “Purification of preny-lated proteins by affinity chromatography on cyclodextrin-modified agarose.” Anal.Biochem. 386, 2009, 1–8.

15. David S. Goodsell and Arthur J. Olson. “Structural symmetry and protein function.”Annu. Rev. Biophys. Biomol. Struct. 29, 2000, 105–53.

16. J.J. van Deemter, F.J. Zuiderweg, and A. Klinkenberg. “diffusion and resistance tomass transfer nonideality in chromatography.” Chem. Eng. Sci. 5, 1956, 271–289.

17. Geertruida A. Posthuma-Trumpie, Jakob Korf, and Aart van Amerongen. “Lateralflow (immuno)assay: its strengths, weaknesses, opportunities and threats. A literaturesurvey.” Anal. Bioanal. Chem. 393, 2009, 569–82.

18. Babacar Ngom, Yancheng Guo, Xiliang Wang, and Dingren Bi. “Development andapplication of lateral flow test strip technology for detection of infectious agents andchemical contaminants: a review.” Anal. Bioanal. Chem. 397, 2010, 1113–1135.

19. Ronald Melnick, George Lucier, Mary Wolfe, Roxanne Hall, George Stancel, Gail Prins,Michael Gallo, Kenneth Reuhl, Shuk-Mei Ho, Terry Brown, John Moore, Julian Leakey,Joseph Haseman, and Michael Kohn. “Summary of the national toxicology program’sreport of the endocrine disruptors low-dose peer review.” Environ. Health Perspect.110, 2002, 427–31.

20. Ake Bergman, Jerrold Heindel, Susan Jobling, Karen Kidd, and R. Thomas Zoeller.State of the science of endocrine disrupting chemicals, 2012. Tech. rep. World HealthOrganization, 2012.

37

Part II

Experimental work

38

Chapter 3

Chromatography-mediatedbioconjugation

Abstract

This chapter describes a method to control the level of protein modification in cases wherethere exist multiple potential modification sites. A protein is first tagged with a handleusing any of a variety of modification chemistries. This handle is used to isolate proteinswith a particular number of modifications via affinity chromatography, and then the handleis elaborated with a desired moiety using an oxidative coupling reaction. This methodresults in a sample of protein with a well-defined number of modifications, and we find itparticularly applicable to systems like protein homomultimers for which there is no way todiscern between chemically identical subunits. We demonstrate the use of this method inthe construction of a protein-templated light-harvesting mimic, a type of system which hashistorically been difficult to make in a well-defined manner.

Haiku

Little donuts stickTo proteins with precisionTo give the good stuff

The majority of this chapter has been reported in a separate publication.1 It appears herein a modified form and is licensed under a CC BY 3.0 license.

http://creativecommons.org/licenses/by/3.0/

CHAPTER 3. CHROMATOGRAPHY-MEDIATED BIOCONJUGATION 39

3.1 Introduction

As discussed in Chapter 1, a rich set of first-order methods for bioconjugation exist. However,these methods cannot be used in certain challenging situations, like the controlled modifica-tion of multimeric proteins. The development of second-order methods could facilitate thesechallenging bioconjugations, because such methods could discriminate between modificationsites based on more than just the reactivity of that site. For example, Chapter 2 demon-strated that it is possible to separate proteins and protein complexes that are modified withdifferent numbers of affinity handles. The incorporation of this chromatographic techniqueinto a strategy for bioconjugation would result in a method with the ability to discriminateamong proteins modified to different extents.

Chromatography-mediated bioconjugation (ChroMB) can be implemented by tagging aprotein with an affinity handle that also serves as a site for further modification (Figure 3.1.i).In this approach, proteins tagged with a desired number of chemical handles are first iso-lated from a crude reaction mixture using affinity chromatography. After purification, thechemical handles are selectively elaborated to access a sample of well-defined bioconjugatemodified with an arbitrary chemical moiety. Such a method allows the controlled modifi-cation of a protein that has more potential modification sites than the desired number ofmodifications. In this chapter, we introduce such a technique and demonstrate its abil-ity to control the modification levels of several monomeric proteins that are modified withNHS-ester chemistry. We then apply this methodology to a particularly challenging biocon-jugation target—a synthetic light-harvesting mimic with a precise ratio of dyes templatedby a protein homotrimer.

3.2 Design of system

3.2.1 Design and synthesis of the affinity handle

A successful realization of chromatography-mediated bioconjugation depends on the affinityhandle, which must participate in an affinity interaction with a chromatographic column andmust also have chemical functionality to allow elaboration with an arbitrary chemical moietyto high conversion. Elaboration of the affinity handle to high conversion can be accomplishedusing many of the bioconjugation strategies discussed in Section 1.2.3; we decided to focus onhandles that could participate in oxidative couplings.2,3 In its most common form, oxidativecoupling occurs between anilines and o-aminophenols in less than two minutes. This chem-istry is particularly well-suited to the construction of well-defined bioconjugates because itusually reaches quantitative conversion and has been used successfully for demanding cou-plings.4 Moreover, when ferricyanide is used as an oxidant, this reaction displays excellentfunctional group tolerance and is compatible with a variety of protein substrates.

As discussed in Section 1.5, β-cyclodextrin is particularly well-suited for incorporationinto a column intended for protein purification. We identified azo compounds as a promising


Modify+ +

Modify Purify Cleave

DesiredBioconjugate

= desired adduct

i.

ii.

= handle

3 3 33.3: 3.4: 3.5: 3.7:3.6: 3.8:

OHNN

SO3-

O R3

OHNH2

O R3

OHO

O R3

N

R4R4

NH2

Figure 3.1: Proposed method for chromatography-mediated bioconjugation and overview of azomodification chemistry. (i) Modification scheme in which the protein is first modified with a handle.Proteins with the desired level of modification are isolated based on an affinity interaction withthis handle. Innate reactivity in the handle is unmasked, and the handle is modified with a desiredbioconjugate. (ii) Azo handles. Azo 3.3 and azo 3.6 bind β-cyclodextrin with a binding constantsufficient for purification. Its cleavage under mild reducing conditions affords an aniline derivativethat can be coupled to many moieties through highly efficient chemistry.

class of handles. Not only is the binding of azo dyes to β-cyclodextrin well-characterized,5–7

but azo dyes with o-hydroxy groups can be cleaved under mild reducing conditions to affordanilines and o-aminophenol.8 Azo compounds 3.3 and 3.6 were identified as potential affinityhandles (Figure 3.1) and were synthesized in one step each from commercially availablematerials. Their measured binding constants of 103.35 M−1 and 103.47 M−1, respectively, andtheir water solubility at concentrations above 1 mM suggested that both would work readilyas affinity handles.

Given the relative ease of synthesis of azo 3.3, we synthesized it first and used it as a toolcompound to aid with the design of an appropriate resin. Later it became apparent thatplacing the o-aminophenol coupling partner on the protein could lead to a small amount ofcross reactivity between the oxidized o-aminophenol and neighboring residues. As a result,azo 3.6 was used for all subsequent experiments.

3.2.2 Design and synthesis of the resin

The resin component of this system was designed to display at least 1 mM of β-cyclodextrinand to interact minimally with unmodified proteins. Early work indicated that the ion-


OH OHN

ONH

O

5

6-azido-β-cyclodextrin, 2,6-lutidine, 2,2’-bipyridine,Cu(I)Br, sodium ascorbateDMF, 15 h

1. CDI, dioxane, 15 min2. Aminocaproic acid water, pH 10, 15 h3. 0.5 M propargylamine 0.1 M EDC, pH 4.5, 15 h

Alkyne-terminated resin β-cyclodextrin-terminated resin

OHN

ONH

O

N5 N N

OHN

ONH

O

53.18

Figure 3.2: Functionalization of 90 µm Sepharose CL-4B with β-cyclodextrin. Sepharose CL-4Bwas first functionalized with CDI and reacted with aminocaproic acid to afford a resin functionalizedwith 5 to 30 mM of carboxylic acid. This acid was then coupled to propargyl amine. Onto thisresin was added azido β-cyclodextrin 3.2 via CuAAC to give the final material. Because theconcentration of attached β-cyclodextrin never exceeded 2 mM, all resins contain an excess ofalkyne-terminated linker on their surfaces.

Table 3.1: Resins with varying Clinker

Clinker / mM CCD / mM log10(Ka/M−1)

A 9 2.3±1 3.4±0.1B 17 2.4±0.5 3.9±0.2C 32 1.8±0.9 3.9±0.3

exchange character of ECH Sepharose 4B and NHS-Sepharose 4B, which contain secondaryamines and isourea groups, respectively, impeded the elution of negatively charged proteins.Moreover, uncharged epoxy-activated Sepharose 4B did not couple sufficient β-cyclodextrinto effect separation. As a result, we turned to Sepharose CL-4B, which contains no ionizablegroups and can be efficiently activated with 1,1’-carbonyldiimidazole (Figure 3.2).9 ActivatedSepharose CL-4B was coupled with aminocaproic acid to yield a water-stable resin withbetween 5 and 30 mM of carboxylic acid functionality. To this resin was coupled propar-gylamine using EDC. Azido β-cyclodextrin was then installed using the copper(I)-catalyzedalkyne-azide cycloaddition to afford the final separation support.10

We explored the importance of linker concentration (Clinker) by synthesizing a series ofresins with varying Clinker (Table 3.1). For each resin, the concentration of β-cyclodextrin(CCD) and the KA to azo 3.3 were measured by performing pulldowns of azo 3.3. All resinshad β-cyclodextrin loadings of 2 mM within error, and it is presumed that sterics limitedfurther modification. It was noted that the KA increased with increasing Clinker. Theseresults suggested that azo 3.3 may favorably interact with some part of the linker betweenthe β-cyclodextrin and the resin to increase the overall binding constant.

To verify this hypothesis, the elution of azo 3.3 from these resins was characterizedbefore and after their modification with β-cyclodextrin (Figure 3.11a and b). Azo 3.3 elutedmore slowly from resins with higher Clinker, suggesting that this handle did indeed havesome affinity for the linker itself (Figure 3.11). Modification with β-cyclodextrin created anadditional favorable interaction and increased retention times further (Figure 3.11b).


To characterize the interaction of these resins with modified protein, the thermostablehomotrimer Mth1491 originally identified in Methanobacterium thermoautotrophicum wasobtained via expression in E. coli.11 To introduce sites for chemical modification, cysteineresidues were added to position 92 on each monomer. Optimization studies showed thatthis site could be modified selectively over the two endogenous cysteines in positions 70 and72 (Figure 3.12). For use in initial chromatographic separations, Mth1491 was modified toabout 30% with azo maleimide 3.5.

Elution of this protein mixture from the alkyne-terminated resins revealed affinity inter-actions between the linker and both the azo 3.3 handle and the protein (Figure 3.11c). Whileresin B slightly separated unmodified and singly modified protein, the higher linker concen-tration of resin C resulted in the retention of all protein species. However, modification withβ-cyclodextrin masked this indiscriminate binding to unmodified protein (Figure 3.11d). β-cyclodextrin-terminated resins B and C showed negligible affinity for unmodified protein, yetthey both bound the azo 3.3 handle and separated unmodified and singly modified protein.The strength of this binding interaction appeared to be a function of Clinker, such that thedegree of separation between unmodified and singly modified protein could be tuned.

It was determined that the Clinker of resin B was optimal to separate proteins withoutexcessive retention times. In the remainder of this chapter, a resin with a Clinker of 15 mMis discussed.

3.3 Purification of proteins

We first explored the potential of this technique to control the number of modifications onmonomeric proteins. NHS-ester chemistry was selected to perform the initial tagging withazo 3.6 because it is frequently used, yet it often results in overmodification. Lysozyme,myoglobin, and RNAse A were tagged using azo NHS-ester 3.7, and LCMS analysis of theproducts indicated that all three showed characteristic product mixtures with up to threecopies of the azo handle on some proteins (Figure 3.3a0, b0, c0). These samples were thenpurified by elution from a 25 cm column packed with resin 3.18 using a linear gradient of0-10 mM β-cyclodextrin. This procedure resulted in separation of the protein bioconjugatesbased on their degree of modification with azo NHS-ester 3.7 (Figure 3.4). Subsequent LCMSanalysis of selected fractions revealed that in all cases it was possible to isolate samples ofsingly modified protein (Figure 3.3a1, b1, c1), and for myoglobin and RNAse A it was alsopossible to isolate doubly modified protein (Figure 3.3b2, c2). The purity of these sampleswas calculated by comparing the areas of the peaks corresponding to each modification levelin Figure 3.3, and they ranged from 91% in the case of doubly modified RNAse A to 98% inthe case of singly modified lysozyme. As an example of a typical yield for this process, wequantified the amount of singly modified RNAse A that was recovered. The crude samplecontained 35% singly modified protein. Purification yielded 72% of the theoretical maximumamount of singly modified protein as determined by UV/vis analysis. Given that some singlymodified RNAse A eluted in fractions with unmodified or doubly modified protein, this yield


m/z

m/z

m/z

Lysozyme

Myoglobin

RNAse A

+0

+1

+2

13000 14000 15000 16000

+1

+0

+1 +2+3

+1

16000 1800017000 19000

+2

+1

+0

+1+2

+3

12000 14000 16000

+2

a0)

a1)

b0)

b1)

b2)

c0)

c1)

c2)

c3)

c4)

cleaved

modified

(i) 50 mM dithionite

(ii) 10 eq. aminophenol, 100 eq. ferricyanide

NH2

NN

HO

SO3-

N

OHO

ONH

Figure 3.3: Reconstructed ESI-TOF mass spectra of crude (a0, b0, c0) and purified bioconjugatestagged with azo NHS-ester 2. Using handle-assisted purification, it was possible to isolate singlymodified samples of each protein (a1, b1, c1), and doubly modified samples of myoglobin and RNAseA (b2, c2). As a demonstration of the potential for the modification of these tagged proteins, theazo handle on singly modified RNAse A was cleaved (i) and coupled to adamantane aminophenolusing an oxidative coupling (ii).

highlights that little protein is lost on the column during purification. The overall yield ofsingly modified protein was 25% of the total protein.

3.4 Elaboration of modified protein

Previous work in our lab has shown that these isolated samples can be elaborated with anarbitrary moiety to high conversion.2,3 As an example, a sample of singly tagged RNAseA was exposed to sodium dithionite to unmask the aniline functionality of the azo handle(Figure 3.3i). Cleavage of the azo took place in less than one minute and resulted in completeconversion without the reduction of any of the four disulfide bonds of RNAse A. This samplewas then exposed to adamantane o-aminophenol under oxidizing conditions to install oneadamantyl group on each protein (Figure 3.3ii). The reaction reached 94% conversion, andthe overall purity of the singly modified conjugate was 89%. Given that the initial purityof RNAse A modified with one azo moiety was about 92%, these results highlight that


13.5k

15.1k

13.5k

15.1k

16.8k

18.2k

16.8k

18.2k

14.0k

15.3k

12.5k

13.5k

12.5k

13.5k

0 5 10 15 20 25 30

0 5 10 15 20 25 30

0 5 10 15 20 25 30

0 5 10 15 20 25 30

0 5 10 15 20 25 30

0 5 10 15 20 25 30 35

0 5 10 15 20 25 30 35

0 5 10 15 20 25 30 35

Volume / mlVolume / ml

12.5k

13.5k

a

b

c

d

e

f

g

h

+0 +0

+0+0+1

+2

+0 +1 +2

+2

+2

+3

+0 +1

+1 +1

+1

+1

+2

A280 (protein) A330 (azo) 0-10 mM β-CD

Figure 3.4: Purification results for (a) RNAse A modified with 3.7, (b) transaminated RNAse Amodified with deprotected 3.10, (c) myoglobin modified with 3.7, (d) transaminated myoglobinmodified with deprotected 3.10, (e) lysozyme modified with 3.7, (f) the Mth1491 trimer modifiedwith 3.8, (g) re-analysis of the isolated singly-modified Mth1491 species, (h) re-analysis of theisolated doubly-modified Mth1491 species. Below each trace appear reconstructed ESI-TOF massspectra of selected fractions, rotated 90◦ clockwise.

elaboration of the cleaved azo handle can be accomplished at high enough yield to maintainthe purity of isolated samples after the purification step.

3.5 Construction of well-defined light harvesting

mimics

Figure 3.3 illustrates that this method can be used to isolate proteins with a particularnumber of modifications. However, this method discerns poorly between proteins that aremodified in different locations. For example, inspection of the elution profile of RNAse A


reveals that each modification level is resolved, but that singly modified protein elutes asa combination of at least three overlapping peaks between 5 and 15 mL (Figure 3.4a). Incomparison, purification of N-terminally modified RNAse A—which can be singly modifiedat only one location—results in only one Gaussian peak corresponding to singly modifiedprotein (Figure 3.4b). These results, and similar results from myoglobin (Figure 3.4c,d),indicate that proteins modified in different locations can have different effective bindingconstants, likely as a result of secondary interactions between the resin and the local proteinenvironment around the handle. Improvement in the resolution of the affinity step couldallow the selection of proteins modified in a particular location or subset of locations, andwork toward this goal is discussed in Chapter 4.

Of particular interest to us was the use of this technique to control the modification ofprotein homomultimers. Our research group has had a longstanding interest in the modifica-tion of these proteins because they can be used to mimic natural light-harvesting systems.12,13

Such mimics are typically composed of pairs of fluorescent dyes that participate in Forsterresonance energy transfer (FRET) and whose arrangement is templated by proteins such asthe homotrimer Mth1491, the MS2 viral capsid,14 or tobacco mosaic virus coat protein.12,13

Because the modification sites on these proteins have identical reactivity, it is impossible tocontrol the arrangement and ratio of dyes on each mutimeric assembly at the single moleculelevel. For example, in the case of the homotrimer Mth1491, an attempt to produce a sam-ple of protein with a 1:2 ratio of dyes would result in a statistical mixture of proteins with0:3, 1:2, 2:1, and 3:0 ratios of dyes, with the amount of each species following the binomialdistribution. As a result, our studies have been limited to ensemble averages of statisticalmixtures of products.

Handle-assisted protein modification could allow controlled modification of multimericcomplexes, especially if they are composed of a small number of monomers and thereforehave a limited number of potential modification sites. We set out to explore this possibilityby using handle-assisted protein modification to create homogeneous samples of Mth1491with 1:2 and 2:1 ratios of two FRET pairs. A column packed with resin 3.18 allowed theseparation of a product mixture of Mth1491 that was modified to about 50% completion withazo maleimide 3.8 (Figure 3.4f). This chromatogram highlights the difficulties traditionallyassociated with the modification of these complexes. Each distinct peak corresponds to oneof the four distinct modification states of the trimer—unmodified, singly modified, doublymodified, and triply modified—and each species is present in significant abundance. Becausethe trimer disassembles upon LCMS analysis, the mass spectrum of each peak shows a ratio ofunmodified to singly modified monomer that is consistent with each modification state. Forthis reason it is also difficult to characterize the homogeneity of these types of samples, andthis chromatographic analysis is one of the few methods that can enumerate the modificationstates of a multimeric protein. To exemplify typical yields during this process, we quantifiedthe amount of doubly modified Mth1491 trimer isolated from one purification. From a1000 µL sample of 100 µM protein modified to about 50% completion, we isolated 27% ofthe original protein from the doubly modified peak. To confirm the purity and stabilityof each isolated sample of Mth1491, these samples were repurified (Figure 3.4g,h). The


3:0

2:1

1:2

0:3

3:0

2:1

1:2

0:3

3:0

2:1

1:2

0:3

3:0

2:1

1:2

0:3

3:0

2:1

1:2

0:3

d)

b)

a)

c)

+ + +

Azo handlePurify

Initial mod.

1. Cleave2. +AF

+OG

AF OG e)AF:OG

m/z

12700

13172 (+azo)

1500012000 13000 14000

13178(+AF) 13632 (+OG)

13172

13165 13175 13185

13178

1.0

0.8

0.6

0.4

0.2

0N

orm

aliz

ed I 44

0 / A

365

wavelength / nm400 450 500 550

Figure 3.5: a) Schematic representation of dual modification of Mth1491 wih AF350 and OG514via handle-assisted purification and conjugation. b-c) Reconstructed ESI-TOF mass spectra for themodification of doubly tagged Mth1491 (b) with AF350 and OG514 (c). d) Histograms represent-ing the composition of the samples whose fluorescent properties were measured. The two samplesprepared using a handle-assisted strategy consist of a single species (solid borders), whereas thesamples prepared using conventional strategies are composed of a statistical mixture. e) Emissionspectra of protein-templated dyes upon excitation at 365 nm. In both cases the homogeneous sam-ples prepared using the handle-assisted strategy exhibit greater quenching, which is characteristicof greater proximity between the two dyes.

resulting chromatograms illustrate the reliability of this technique in isolating singly anddoubly tagged Mth1491 and indicate that the protein stayed folded and assembled duringpurification, handling, and storage.

With singly and doubly tagged Mth1491 in hand, we continued the handle-assisted modi-fication of these proteins with Alexafluor 350 (AF350, donor) and Oregon green 514 (OG514,acceptor) as outlined in Figure 3.5a. After cleavage of the azo handle, the remaining un-modified cysteines at position 92 were modified with AF350 maleimide. OG514 was thencoupled to the resulting anilines using an oxidative coupling. We have shown previously thatcysteines are modified by oxidized o-aminophenols.3 In the case of Mth1491, the presence oftwo endogenous cysteines results in double modification of 18 to 35% of monomers when 5 to10 equivalents of o-aminophenol are used. To prevent this undesired overlabeling, the pro-tein was protected with Ellman’s reagent prior to oxidative coupling, and then subsequentlydeprotected with TCEP after the reaction. To verify that this protection did not affect theassembly of Mth1491, dynamic light scattering was used to determine that the average par-


ticle size increased from 5.5±0.3 nm to 7.4±0.9 nm during the protection step. This increasein diameter is consistent with the modification of the the surface of the protein and indicatesthat the protein was still assembled as a trimer. Fortunately, many proteins will not requirethis protection step because their cysteines are buried or participate in disulfide bonds.

Illustrative mass spectra for the construction of the 1:2 AF350:OG514 sample are shownin Figure 3.5b and c. These spectra indicate high conversion for the addition of AF350 andOG514 (the peak at 13172 Da completely disappears, see inset). We presume that changesto the ionizability of the monomers as a result of conjugation to the dyes are responsiblefor the change in the ratios of the peak heights.∗ Measurement of the absorbance spectraindicated that AF350 and OG514 were present in the expected ratios and supported thisinterpretation.

The fluorescence properties of these systems were then characterized by considering thequenching of the AF350 donor. The concentration of AF350 dye was first calculated fromthe absorbance spectra of these samples after removing the contributions of OG514 and aminor amount of scatter. Emission spectra were measured upon excitation at 365 nm, andthese data were smoothed and corrected for baseline artifacts. These emission spectra werenormalized by the concentration of AF350 dye, and the data were plotted relative to theemission spectrum of a Mth1491 trimer bearing three AF350 dyes (Figure 3.5d-e, solid lines).Both the 2:1 an the 1:2 AF350:OG514 systems show quenching of the AF350 donors that isindicative of energy transfer, with efficiencies of 73 and 84%, respectively. Assuming randomorientations of the dyes, we computed their Forster radius to be 4.9±0.05 nm. These valuesfor efficiency correspond to distances of 4.1 and 3.7 nm between the dyes. Such lengthsare consistent with the fact that the dyes are templated by Mth1491, which has a distancebetween its cysteines at position 92 of 4.1 nm.

For the purpose of comparison, we constructed systems whose dye content was the sameas for the 2:1 and 1:2 AF350:OG514 samples, but without the purification step shown inFigure 3.5a. Obtaining protein initially modified with the correct amount of azo maleimide3.8 proved challenging, and we eventually resorted to preparing a number of samples withdifferent modification levels and selecting the correct one. This exercise alone illustrates thedifficulties associated with obtaining protein bioconjugates with precise levels of modifica-tion, for ∼80% of the protein modified during this process was not used. Construction ofthe two-dye systems in this way resulted in statistical mixtures of trimers with the four pos-sible combinations of AF350 and OG514 that result from traditional strategies for proteinmodification (Figure 3.5d, dashed lines). The emission spectra of these samples indicatereduction in efficiency of energy transfer by 14 and 8%, respectively (Figure 3.5d-e, dashedlines). Reduction of FRET efficiency is consistent with the increased separation between theAF350 donor and the OG514 acceptor that would result from larger population of the 3:0 and

∗In our experience these protein-dye conjugates ionize poorly compared to the unmodified and aniline-bearing proteins, with total ion counts that are about an order of magnitude lower. Of the minor peaks foundalong the baselines, two were identified as the unmodified Mth1491 monomer and the Mth1491 monomerbearing an aniline moeity. However, the abundance of these species was not quantified in light of the poorionizability of these conjugates.


2:1 AF350:OG514 systems. These results illustrate the difficulties associated with the con-trolled modification of homomultimeric proteins and underscore the utility of handle-assistedprotein modification in producing well-defined nano-scale materials.

3.6 Conclusions

This chapter reports a method for the construction of well-defined protein bioconjugatesthrough chromatography-assisted protein modification. This method relies on the tagging ofa protein with a specific affinity handle that allows purification and subsequent modificationof the protein. Through this procedure it is possible to control the degree of modificationof both monomeric and multimeric proteins, even when the proteins are modified with non-specific reagents such as NHS-esters. In the case of homomultimeric proteins, this methodis to our knowledge the only way to control the number of modifications. We anticipatethat this method will be of substantial synthetic utility in making protein-based materialsof increasing complexity.


3.7.1 General methods

Unless otherwise noted, the chemicals and solvents used were of analytical grade and wereused as received from commercial sources. Purifications by flash chromatography were per-formed using EM silica gel 60 (230-400 mesh). Chromatography solvents were used with-out distillation. All organic solvents were removed under reduced pressure using a rotaryevaporator. Water (dd-H2O) used as a reaction solvent was deionized using a BarnsteadNANOpure purification system. Centrifugations were performed with an Eppendorf MiniSpin Plus (Eppendorf, Hauppauge, NY).

3.7.2 Instrumentation and sample analysis

NMR. 1H and 13C spectra were recorded with a Bruker AVB-400 (400 MHz, 100 MHz) or aBruker AV-600 (600 MHz, 150 MHz). 1H NMR chemical shifts are reported as δ in units ofparts per million (ppm) relative to residual CH3OH (δ, 3.31, pentet), DMF (δ 8.03, singlet),or CHCl3 (δ, 7.24, singlet). Multiplicities are reported as follows: s (singlet), d (doublet),t (triplet), dd (doublet of doublets) or m (multiplet). Coupling constants are reported as aJ value in Hertz (Hz). The number of protons (n) for a given resonance is indicated as nHand is based on spectral integration values. 13C NMR chemical shifts are reported as δ inunits of parts per million (ppm) relative to DMF−d7 (δ 163.15, triplet), MeOH−d4 (δ 49.15,septet), or CDCl3 (δ, 77.23, triplet).Mass Spectrometry. Matrix-assisted laser desorption/ionization time-of-flight mass spec-trometry (MALDI-TOF MS) was performed on a Voyager-DE system (PerSeptive Biosys-


tems, USA). Small molecule samples were co-crystallized with α-cyano-4-hydroxycinnamicacid in 1:1 acetonitrile (MeCN) to H2O with 0.1% trifluoroacetic acid (TFA). Cyclodextrinswere co-crystallized with 2,5-dihydroxybenzoic acid in 1:1 MeCN to H2O with 0.1% TFA.Protein bioconjugates were analyzed using an Agilent 1200 series liquid chromatograph (Ag-ilent Technologies, USA) that was connected in-line with an Agilent 6224 Time-of-Flight(TOF) LC/MS system equipped with an electrospray ion source. Extracted mass spectrawere plotted using chartograph.com/ms.High Performance Liquid Chromatography. HPLC was performed on Agilent 1100 Se-ries HPLC Systems (Agilent Technologies, USA). Sample analysis for all HPLC experimentswas achieved with an in-line diode array detector (DAD) and in-line fluorescence detector(FLD). Analytical reverse-phase HPLC of small molecules was accomplished using a C18stationary phase and a H2O / MeCN with 0.1% TFA gradient mobile phase.Fast protein liquid chromatography. FPLC was performed on an Akta Pure M (GEHealthcare, USA) at 8 ◦C. Sample analysis was performed with an in-line UV monitor andan in-line conductance monitor. All injections were performed manually with a 1 mL syringefit to the top of the column.UV-Visible Spectrometry. UV-visible spectrometry was performed using quartz cuvetteswith a Varian Cary 50 spectrophotometer (Agilent, USA). Small-scale UV-visible spectrom-etry was performed using a Nanodrop 1000 (Thermo Scientific, USA). Absorbance mea-surements of samples in plates were obtained with a SpectraMaxM2 (Molecular Devices,Sunnyvale, CA).Fluorescence. Fluorescence measurements were obtained on a Fluoromax-4 spectrofluo-rometer equipped with automatic polarizers and a Peltier temperature controller (ISA In-struments, USA). Slit widths were set to 1.0 nm for excitation and 1.0 nm for emission.Fluorescence emission was monitored with a 0.5 s integration time. For three-dimensionalexcitation/emission measurements, the excitation wavelength was scanned from 200 – 600nm in 4 nm increments and the fluorescence emission was monitored from 290 – 600 nm in2 nm increments.

3.7.3 Small molecule synthesis

O

OHHO

OH

O

O

OH

HOOTsO

OOH

OH

OH

O

O

OHOH

OH

OO

OH

OH

HOO

OOH

OHHO

O

OOH

HO

HO

O

Synthesis of 6-O-p-toluenesulfonyl-β-cyclodextrin (3.1). Prepared on a 50 g scale asdescribed by Byun, Zhong, and Bittman15

chartograph.com/ms


O

OHHO

OH

O

O

OH

HON3

O

OOH

OH

OH

O

O

OHOH

OH

OO

OH

OH

HOO

OOH

OHHO

O

OOH

HO

HO

O

Synthesis of 6-azido-β-cyclodextrin (3.2). This procedure was adapted from Nielsenet al.16 Cyclodextrin 3.1 (7.2 g, 5.6 mmol, 1 eq.) was dissolved in 49 mL of DMF in a 250mL round bottom flask. To this solution was added sodium azide (6.3 g, 97 mmol, 17.3eq.). The reaction was stirred at 75 ◦C under N2 for 18 h. The reaction was cooled to roomtemperature, filtered through Celite, and poured into 800 mL of vigorously stirred acetone.The white precipitate was collected with a Buchner funnel. To remove excess sodium azide,the crude product was dissolved in 60-70 mL of water at 80 ◦C, cooled on ice to roomtemperature, and again poured into 800 mL of vigorously stirred acetone to precipitate theproduct. After filtration, this procedure was repeated an additional three times, at whichpoint it was determined that the amount of sodium azide in the product was less than 0.3%by mass17. A total of 6.4 g of material was recovered (>90%).

N

NaO3S

N

HO

OH

O Synthesis of azo 3.3. Sulfanilic acid (1.7 g, 9.83 mmol, 1 eq.)was added to 150 mL of concentrated aqueous HCl in a 250 mLErlenmeyer flask at 0 ◦C. To this solution was added 0.75 g (10.87mmol, 1.1 eq.) of sodium nitrite in 10 mL of water, and the

reaction was stirred for 30 min. This solution was added dropwise over 1 h to a solution ofvigorously stirred 3-(4-hydroxyphenyl)propionic acid (1.8 g, 10.87 mmol, 1.1 eq.) and sodiumcarbonate (105 g, 1 mol) in 250 mL of 4:1 water:methanol at 0 ◦C. During the addition,the reaction bubbled vigorously and turned bright reddish orange. After the addition wascomplete, the reaction was stirred for an additional hour at room temperature. The reactionwas acidified with concentrated aqueous HCl, at which point the product precipitated as afine reddish brown solid that was collected via filtration with a 0.2 µm Teflon filter. Theresulting material was recrystallized from 3:1 ethanol:water to afford 2.5 g (73%) of material.An absorbance spectrum is shown in Figure 3.6. 1H NMR (400 MHz, DMF−d7) δ 12.43 (s,1H), 11.67 (s, 1H), 7.97 (d, J = 1.5 Hz, 4H), 7.79 (d, J = 2.2 Hz, 1H), 7.40 (dd, J = 8.5, 2.2Hz, 1H), 7.02 (d, J = 8.4 Hz, 1H), 2.95 (t, J = 7.6 Hz, 2H), 2.69 (t, J = 7.5 Hz, 2H). 13CNMR (100 MHz, DMF−d7) δ 174.80, 153.70, 152.92, 152.22, 139.03, 135.11, 133.95, 128.07,126.92, 122.99, 119.16, 36.46, 30.66. HRMS (ESI) calculated for C15H13N2O6S– ([M-Na]−)349.0500, found 349.0496.


O

ON

O

OHO

NN

NaO3S Synthesis of azo 3.3 NHS-ester (3.4). Azo 3.3 (52.9mg, 0.220 mmol, 1 eq.) and N -hydroxysuccinimide (46.0 mg,0.4 mmol, 2 eq.) were dissolved in 1 mL of DMF. N,N ’-

Diisopropylcarbodiimide (50.4 mg, 0.4 mmol, 2 eq.) was added and the reaction was stirredfor 3 days. The solvent was removed under reduced pressure, and the product was takenup in methanol and adsorbed to 1 mL of silica gel. The product was purified via silica gelchromatography with a gradient of 0 to 50% methanol in dicholormethane with 0.1% aceticacid. 1H NMR (400 MHz, Methanol−d4) δ 8.02 (d, J = 8.7 Hz, 2H), 7.98 (d, J = 8.7 Hz,2H), 7.87 (d, J = 2.4 Hz, 1H), 7.35 (dd, J = 8.5, 2.3 Hz, 1H), 6.98 (d, J = 8.5 Hz, 1H),3.10 (t, J = 8.9, 6.7 Hz, 2H), 3.01 (t, J = 7.0, 1.6 Hz, 2H), 2.85 (s, 4H).

N

NaO3S

N

HO

NH

ON

O

O Synthesis of azo 3.3 maleimide (3.5). Azo 1 (70 mg,0.2 mmol, 1 eq.) and N -hydroxysuccinimide (46.0 mg, 0.4mmol, 2 eq.) were dissolved in 1 mL of DMF. N,N ’-Diisopropylcarbodiimide (50.4 mg, 0.4 mmol, 2 eq.) was added

and the reaction was stirred for 72 h. To this solution, 2-maleimido-ethylamine (28.0 mg,0.2 mmol, 1 eq.) was added, and the reaction was stirred for 5 h. The solvent was removedunder reduced pressure, and the solid taken up in 5 mL of methanol and adsorbed to 1 mLof silica gel. The product was purified using silica gel chromatography with a gradient of5 to 20% methanol in dichloromethane with 0.1% acetic acid to afford 17 mg of material(18%). 1H NMR (600 MHz, DMF−d7) δ 11.63 (s, 1H), 8.07 (t, J = 6.2 Hz, 1H), 8.02 7.91(m, 4H), 7.71 (d, J = 2.2 Hz, 1H), 7.33 (dd, J = 8.5, 2.3 Hz, 1H), 7.02 (m, 3H), 3.56 (d, J =5.9 Hz, 2H), 3.35 (q, J = 6.0 Hz, 2H), 2.90 (t, J = 7.8 Hz, 2H), 2.45 (t, J = 7.8 Hz, 2H). 13CNMR (150 MHz, DMF−d7) 172.96, 172.22, 153.69, 152.45, 152.36, 139.08, 135.58, 135.06,134.31, 128.09, 126.62, 123.00, 119.14, 38.55, 38.48, 38.31, 31.36. HRMS (ESI) calculatedfor C21H19N4O7S– ([M-Na]−) 471.0980, found 471.0970.

NN

OH

NaO3S

O

OH

Synthesis of azo 3.6. To 29 mL of 1 M aqueous HCl was added3-(4-aminophenyl)propionic acid (1 g, 6.06 mmol, 1 eq.), and thesolution was stirred until all solids had dissolved. This solution wascooled to 0 ◦C, and 0.42 g (6.06 mmol, 1 eq.) of sodium nitrite in 5

mL of water was added dropwise. After 30 min, this solution was added dropwise to sodium4-hydroxybezenesulfonate dihydrate (1.41 g, 6.06 mmol, 1 eq.) and sodium carbonate (6.03g, 57.4 mmol) in 14.4 mL of water at 0 ◦C. After the addition was complete, the reaction waswarmed to room temperature and was stirred for an additional hour. During this time thereaction mixture turned dark red/black. The solvent was removed under reduced pressure,and the crude reaction mixture was dissolved in methanol and filtered through Celite. Thisprocedure was repeated a second time, at which point the product mixture was adsorbedto 20 mL of silica gel and purified via silica gel chromatography with a gradient of 0-70%methanol in dichloromethane to give 1 g (44%) of material. A portion of this was purified byreverse-phase chromatography with a gradient of 5-95% acetonitrile in water with 0.1% TFAover 1 h. An absorbance spectrum is shown in Figure 3.7. This portion was used for binding


studies and characterization. 1H NMR (400 MHz, DMF−d7) δ 8.25 (d, J = 2.2 Hz, 1H),7.98 (d, J = 8.35 Hz, 2H), 7.84 (dd, J = 8.6, 2.2 Hz, 1H), 7.52 (d, J = 8.35 Hz, 2H), 7.10(d, J = 8.5 Hz, 1H), 3.02 (t, J = 7.5 Hz, 2H), 2.72 (t, J = 7.5 Hz, 2H). 13C NMR (100 MHz,DMF−d7) δ 174.56, 156.23, 151.10, 146.47, 140.57, 138.16, 131.91, 130.44, 123.79, 123.14,118.69, 35.93, 31.55. HRMS (ESI) calculated for C15H13N2O6S– ([M-Na]−) 349.0500, found349.0497.

ON

O

ON

N

OH

NaO3S

O Synthesis of azo 3.6 1 NHS-ester (3.7). Flash columnchromatography-purified azo 2 (180 mg, 0.484 mmol, 1 eq.) andN -hydroxysuccinimide (66.8 mg, 0.581 mmol, 1.2 eq.) were dis-solved in 0.5 mL of DMF. N,N ’-Diisopropylcarbodiimide (73.3 µL0.581 mmol, 1.2 eq.) was added and the reaction was stirred for

15 h. The solvent was removed under reduced pressure, and the product was taken up inmethanol and adsorbed to 1 mL of silica gel. The product was purified via silica gel chro-matography with a gradient of 0 to 50% methanol in dicholormethane with 0.1% acetic acidto afford 60 mg of material (26%). 1H NMR (400 MHz, DMF−d7) δ 11.74 (s, 1H), 8.25 (d, J= 1.7 Hz, 1H), 7.99 (d, J = 8.3 Hz, 2H), 7.83 (dd, J = 8.6, 2.1 Hz, 1H), 7.60 (d, J = 8.3 Hz,2H), 7.07 (d, J = 8.5 Hz, 1H), 3.52 (s, 2H), 3.15 (s, 4H), 2.94 (s, 4H). 13C NMR (100 MHz,DMF−d7) δ 170.46, 168.81, 155.14, 150.64, 143.98, 140.96, 137.40, 131.47, 129.90, 123.06,122.75, 117.69, 31.88, 30.30, 25.87. HRMS (ESI) calculated for C19H16N3O8S– ([M-Na]−)446.0664, found 446.0658.

NH

N

O

ON

N

OH

NaO3S

O Synthesis of azo 3.6 maleimide (3.8). Flash columnchromatography-purified azo 2 (27.3 mg, 0.073 mmol, 1 eq.)and N -hydroxysuccinimide (10.1 mg, 0.088 mmol, 1.2 eq.) weredissolved in 0.5 mL of DMF. N,N ’-Diisopropylcarbodiimide(13.6 µl, 0.088 mmol, 1.2 eq.) was added and the reaction

was stirred for 15 h. To this solution, 2-maleimido-ethylamine (17.6 mg, 0.073 mmol, 1 eq.)was added, and the reaction was stirred for 30 min. The solvent was removed under reducedpressure, and the solid was taken up in 5 mL of methanol and adsorbed to 1 mL of silicagel. The product was purified using silica gel chromatography with a gradient of 0 to 50%methanol in dichloromethane with 0.1% acetic acid to afford 2.5 mg of material (7%). 1HNMR (600 MHz, DMF−d7) δ 8.25 (d, J = 2.2 Hz, 1H), 8.08 (t, J = 5.8 Hz, 2H), 7.96 (d,J = 8.3 Hz, 2H), 7.81 (dd, J = 8.5, 2.2 Hz, 1H), 7.45 (d, J = 8.4 Hz, 2H), 7.05 (d, J =8.5 Hz, 1H), 7.02 (s, 2H), 3.58 (t, J = 5.8 Hz, 2H), 3.36 (q, J = 5.9 Hz, 2H), 2.97 (t, J= 7.8 Hz, 2H), 2.48 (t, J = 7.9 Hz, 2H). 13C NMR (150 MHz, DMF−d7) δ 172.79, 172.25,155.49, 151.06, 146.85, 142.49, 138.06, 135.61, 132.19, 130.42, 124.14, 123.75, 118.28, 38.50,38.35, 38.05, 32.18. HRMS (ESI) calculated for C21H19N4O7S– ([M-Na]−) 471.0980, found471.0977.

NH

HN

NN

OH

NaO3S

O

Boc

Synthesis of azo 3.6 Boc-amine (3.9). Compound 3.7(60 mg, 0.128 mmol, 1 eq.), N -Boc-ethylenediamine (20.5 mg,


0.128 mmol, 1 eq.), and triethylamine (15.5 mg, 0.154 mmol,1.2 eq.) were added to 1 mL of DMF and stirred at room

temperature for 30 min. The solvent was removed, and the product was purified via silicagel chromatography with a gradient of 0 to 50% methanol in dicholormethane with 0.1%acetic acid. Trace amounts of N -Boc-ethylenediamine remained, so the product mixturewas added to 3 mL of Amberlite IR-120 Na form resin in 5 mL of water and turned end-over-end on a laboratory rotisserie for 30 min. After removal of the resin, the solvent wasremoved under reduced pressure to afford 58 mg of material (86%). 1H NMR (400 MHz,MeOH−d4) δ 8.26 (d, J = 2.3 Hz, 1H), 7.82 (d, J = 2.3 Hz, 1H), 7.78 (d, J = 8.5 Hz, 2H),7.38 (d, J = 8.3 Hz, 2H), 7.09 (d, J = 8.7 Hz, 1H), 3.20 (t, J = 6.1 Hz, 2H), 3.08 (t, J =5.6 Hz, 2H), 2.97 (t, J = 7.4 Hz, 2H), 2.73 (s, 2H), 2.55 (t, J = 7.5 Hz, 2H), 1.36 (s, 9H).13C NMR (100 MHz, MeOH−d4) δ 178.55, 175.70, 154.23, 145.36, 136.24, 130.34, 129.46,127.15, 122.44, 118.07, 99.98, 79.72, 39.10, 37.02, 31.26, 27.52, 25.03, 22.13. MALDI-TOFMS calculated for C22H27N4O7S– ([M-Na]−) 491.16, found 491.40.Deprotection of of azo 3.6 Boc-amine (3.9). The Boc group was removed by addi-tion of 0.5 mL of trifluoroacetic acid to the compound, followed by removal of the volatilecomponents under reduced pressure.

NH

HN

NN

OH

-O3S

O

OO

HN

Boc

H+

N

Synthesis of azo 3.6 Boc-alkoxyamine (3.10).To 0.85 mL of dichloromethane was added N -hydroxysuccinimide (19.5 mg, 0.17 mmol, 1.44 eq.)and (Boc-aminooxy)acetic acid (27.1 mg, 0.142 mmol,1.2 eq.). After sonication to mostly dissolve the N -

hydroxysuccinimide, N,N ’-dicycohexylcarbodiimide (35 mg, 0.17 mmol, 1.44 eq.) was added,and the reaction was stirred at room temperature for 15 min. The reaction mixture wascooled to 0 ◦C, filtered through Celite, and added to 3.9 (58 mg, 0.118 mmol, 1 eq.) andtriethylamine (23.8 mg, 0.236 mmol, 2 eq.) in 0.85 mL of DMF. The reaction was stirredfor 30 min, at which point the solvent was removed, and the crude product was adsorbedto 1 mL of silica gel. The product was purified using silica gel chromatography with a gra-dient of 0 to 50% methanol in dichloromethane with 0.1% acetic acid to afford 41.3 mg ofmaterial (60%). 1H NMR (400 MHz, MeOH−d4) δ 8.30 (d, J = 2.3 Hz, 1H), 7.81 (m, J =5.3, 2.4 Hz, 3H), 7.40 (d, J = 8.0 Hz, 2H), 7.08 (d, J = 8.6 Hz, 1H), 4.26 (s, 2H), 3.32 (s,4H), 3.18 (q, J = 7.3 Hz, 8H), 3.00 (t, J = 7.7 Hz, 2H), 2.57 (t, J = 7.7 Hz, 2H), 1.43 (s,9H), 1.30 (t, J = 7.3 Hz, 12H). [Note: excess triethylamine was observed and could not beseparated from the compound.] 13C NMR (100 MHz, MeOH−d4) δ 174.40, 170.65, 158.36,154.36, 149.31, 145.41, 136.70, 136.32, 130.32, 129.36, 127.05, 122.41, 117.91, 82.27, 74.97,46.54, 38.42, 37.05, 31.23, 27.19, 24.80, 7.97. HRMS (ESI) calculated for C24H30N5O9S–

([M-Triethylammonium]−) 564.1770, found 564.1761.Deprotection of azo 3.6 Boc-alkoxyamine (3.10). The Boc group was removed byaddition of 0.5 mL of trifluoroacetic acid to the compound, followed by removal under reducedpressure to afford the alkoxyamine.


NH2

HONO2 Synthesis of o-nitrotyramine (3.12). This procedure was adapted from

Waser and Sommer18. Tyramine (0.57 g, 4.16 mmol, 1 eq.) in 4 mL of waterin a 20 mL scintillation vial was cooled to 0 ◦C with an ice bath. Nitric acid (2ml) was added dropwise, and the solution turned reddish brown immediately.

After the addition, the reaction was warmed to room temperature and stirred for 10 min. Thereaction mixture then stood uncovered at room temperature overnight, at which point theyellow precipitate was collected with a Buchner funnel. This crude product was recrystallizedfrom 3-5 mL of boiling water to afford 0.25 g of papery yellow material (34%).

O

HN

OH

NO2 Synthesis of adamantane o-nitrotyramine (3.13). 1-adamantanecarboxylic acid (1 g, 5.55 mmol, 1 eq.) was dissolved in 5.2 mL thionylchloride (8.6 g, 72 mmol, 13 eq.) and stirred for 2.5 hr at 80 ◦C under

nitrogen. The reaction mixture was cooled to room temperature, and excess thionyl chloridewas removed under reduced pressure by azeotropic distillation with toluene. The resultingsolid (20 mg, 0.1 mmol, 1 eq.) was immediately added to a mixture of pyridine (73 µL, 0.9mmol, 9 eq.) and 3.12 (18.38 mg, 0.1 mmol, 1 eq.) in 1 mL of DMF. The reaction wasturned on a laboratory rotisserie for 30 min, at which point excess solvent was removed underreduced pressure. The crude product was taken up in 10 mL of 1 M HCl, and extracted twicewith two 10 mL portions of DCM. The combined organic fractions were dried with MgSO4,and the solvent was removed under reduced pressure. This crude solid was then purified viasilica gel chromatography with a gradient of 33 to 100% ethyl acetate in hexanes to afford6 mg of yellow solid (17%). The compound was stored as a 17 mM solution in DMSO. 1HNMR (600 MHz, CDCl3, 63:37 ratio of rotamers) δ 10.46 (s, 1H), 7.89 and 7.84 (rotamers,d, J = 2.2 Hz), 7.41 and 7.43 (rotamers, dd, J = 8.2, 2.1 Hz, 1H), 7.09 (dd, J = 8.4, 2.4 Hz,1H), 5.63 and 5.66 (rotamers, t, J = 6.6, 5.8 Hz, 1H), 3.44 and 3.475 (rotamers, q, J = 6.7,6.2 Hz, 2H), 2.79, 2.87 (rotamers, t, J = 6.9 Hz, 2H) 2.08 1.62 (rotamers, m, 15H). 13C NMR(150 MHz, CDCl3) δ 178.30, 154.01, 138.58, 138.37, 135.08, 131.78, 125.92, 125.48, 124.75,120.38, 77.44, 77.23, 77.02, 40.88, 40.43, 39.51, 39.49, 38.91, 38.79, 36.69, 36.60, 35.11, 34.80,28.31, 28.06. HRMS (ESI) calculated for C19H23N2O –

4 ([M-H]−) 343.1663, found 343.1664.Synthesis of adamantane o-aminotyramine. To a 0.5 mL solution of 100 mMsodium dithionite in 200 mM pH 6.5 phosphate buffer was added 1-3 µL of 17 mM 3.13.The reaction was stirred for 10 min, at which point excess dithionite was removed using aC18 Sep-Pak according to the manufacturer’s instructions. The eluent was concentrated todryness, and the product was dissolved in 10 mM pH 6.5 phosphate buffer. The productwas used immediately without characterization.

O OHO

F F

O

OH

FF

F

SO

HN

HO

O2N

Synthesis of Oregon green 514 o-nitrophenol (3.15). Ore-gon green 514 NHS-ester (2.46 mg, 0.004 mmol, 1 eq.) was dis-solved in 1 mL of DMF. Compound 3.12 (1.31 mg, 0.0097 mmol,2.4 eq.) and triethylamine (8.15 mg, 0.081 mmol, 20 eq.) wereadded and the reaction was turned end-over-end for 30 min ona laboratory rotisserie. The solvent was removed under reduced


pressure, and the crude product mixture was dissolved in 250 µL of methanol. After filtrationthrough a 0.2 µm filter, the o-nitrophenol product was purified via HPLC using an AlltechEconosil C18 10 µM column with a gradient of 5-95% acetonitrile in water with 0.1% TFAover 1 h. The solvent was removed under reduced pressure, and the dye was stored as a 5mM solution in 50 µL of DMSO.Synthesis of Oregon green 514 o-aminophenol. To a 0.5 mL solution of 100 mMsodium dithionite in 200 mM pH 6.5 phosphate buffer was added 10-30 µL of 5 mM 3.15.The reaction was stirred for 10 min, at which point excess dithionite was removed using aC18 Sep-Pak according to the manufacturer’s instructions. The eluent was concentrated todryness, and the product was dissolved in 10 mM pH 6.5 phosphate buffer. The productwas used immediately without characterization.

3.7.4 Resin synthesis

OHN

OOH

O

5

Carboxy Sepharose CL-4B (3.16). Settled Sepharose CL-4B (8 mL) was rinsed succes-sively with 10 mL each of with water, 3:1 water:dioxane, 1:3 water dioxane, and dioxane.Depending on the desired level of modification, 100-700 mg of 1,1’-carbonyldiimidazole in4 mL of dioxane was added to this resin. This suspension was rotated end-over-end on alaboratory rotisserie for 15 min, at which point it was rinsed with 10 mL each of dioxane,1:3 water dioxane, 3:1 water:dioxane, and water. Once the resin had fully drained, 1.3 g ofaminocaproic acid was added in 10 mL of water adjusted to pH 10 with 1 M NaOH, andthe resin was turned end-over-end on a laboratory rotisserie for 15 h. The resin was washedrepeatedly with 0.1 M pH 4 sodium acetate buffer containing 0.5 M sodium chloride and 0.1M pH 8 TRIS buffer containing 0.5 M sodium chloride, in alternation. The resin was finallywashed with several portions of water.

OHN

ONH

O

5

Alkyne-terminated resin (3.17). After the carboxylic acid content ofthe resin 3.16 was determined, it was rinsed thoroughly with water andallowed to drain completely. To this resin was added propargyl amine

(0.46 g, 5 mmol) in 5 mL of water and N -(3-Dimethylaminopropyl)-N ’-ethylcarbodiimidehydrochloride (0.766 g, 4 mmol) in 5 mL of water, both pH adjusted to 4.5 using HCl. Aftermixing, the pH of the reaction mixture was checked and adjusted again to pH 4.5, if necessary.The reaction rotated end-over-end on a laboratory rotisserie for 1 h, at which point the pHwas again adjusted to 4.5 using NaOH. After 18 h, the resin was washed repeatedly with 0.1M sodium acetate buffer containing 0.5 M sodium chloride buffer, pH 4, and 0.1 M TRISbuffer containing 0.5 M sodium chloride, pH 8, in alternation. The resin was finally washedwith several portions of water.


OHN

ONH

O

N5 N N CD

OHN

ONH

O

5

β-cyclodextrin-terminated resin (3.18). This procedure hasbeen adapted from Punna, Kaltgrad, and Finn10 Resin 3.17 wasrinsed successively with 10 mL each of water, 3:1 water:DMF, 1:3water:DMF, and DMF. To the drained resin was added 2,6-lutidine(64.3 mg, 0.6 mmol) in 0.5 mL DMF, 2,2’-bipyridine (93.7 mg, 0.6

mmol) in 1 mL DMF, 6-azido-β-cyclodextrin(348 mg, 0.3 mmol) in 1.5 mL of DMF, copper(I)bromide (43 mg, 0.3 mmol) in 1 mL of DMF, and sodium ascorbate (119 mg, 0.6 mmol) in1 mL of water. This reaction mixture was turned end-over-end on a laboratory rotisseriefor 15 h. The resin was rinsed with 10 mL each of DMF, 1:3 water:DMF, 3:1 water:DMF,and water, followed by repeated washings with 0.1 M pH 4 sodium acetate buffer containing0.5 M sodium chloride and 0.1 M pH 8 TRIS buffer containing 0.5 M sodium chloride, inalternation. The resin was finally washed with several portions of water.Three different resins with varying linker concentrations were synthesized (Table 3.1). Aftercharacterization (Section 3.7.5), the ability of these resins to separate Mth1491 modifiedwith the azo 3.6 handle was assessed (Figure 3.11). Based on these results, we determinedthat a resin with a total linker concentration of 15 mM and a β-cyclodextrin concentrationof 2 mM was optimal to effect separation between different levels of protein modificationwithout excessive retention times.

3.7.5 Determination of resin loading and binding constants

Determination of in-solution binding constants. A solution of 100 µM azo 1 or 2was prepared in 10 mM pH 6.5 phosphate buffer (henceforth called azo buffer). A 10 mMsolution of β-cyclodextrin was then prepared by dissolving β-cyclodextrin in the previouslyprepared azo buffer (henceforth called CD buffer). The CD buffer was then titrated into0.75 mL of the azo buffer, and the absorbance of the solution was measured three timesafter each addition. After the titration, the absorbance values at 375 nm were plotted, anda single-site binding model was fit to the data using the nonlinear curve fitting package inSciPy:

A375 = εazo

Fazo − FCD −Kd +√

(Fazo + FCD +Kd)2 + 4FazoKd

2

+ εCD-azo

Fazo + FCD +Kd −√

(Fazo + FCD +Kd)2 − 4FazoFCD

2(3.1)

where Fcd is the total concentration of β-cyclodextrin in units of M, Fazo is the total con-centration of azo in units of M, Kd is the dissociation constant in units of M, εazo is the 1cm molar extinction coefficient for the uncomplexed azo dye in units of M−1, and εCD-azo isthe 1 cm molar extinction coefficient for the complexed azo dye, also in units of M−1. Thisprocedure was performed three times each for azo 3.3 and for azo 3.6 and the results areshown in Figures 3.8 and 3.9.Determination of carboxylic acid concentration on resins. Resin 3.17 or 3.18 (0.5mL, settled) was measured into a 3 mL fritted column and washed thoroughly with 1 mM


HCl until the pH of the eluate was 3 as determined by pH paper. This resin was thenwashed three times each with 1 mL of 1 mM HCl with 100 µm phenolphthalein. After it hadcompletely drained, the resin was transferred to a 4 mL scintillation vial with an additional0.5 mL of 1 mM HCl with 100 µm phenolphthalein, such that the vial contained 0.5 mL ofresin in a total volume of 1 mL. A solution of 100 mM NaOH was then added until the resinsuspension displayed the first hint of a persistent pink color. This process was repeated fora control sample of unmodified Sepharose CL-4B, and the amount of excess 100 mM NaOHrequired to neutralize the modified resin was used to calculate its concentration of carboxylicacid.Determination of resin loading and binding constant. Resin 3.18 (0.5 mL, settled)was measured into a 3 mL fritted column. This resin was washed thoroughly with 10 mMpH 6.5 phosphate buffer and transferred into a 20 mL scintillation vial with an additional 0.5mL of buffer, such that the vial contained 0.5 mL of resin in 1 mL total volume. A solutionof 550 azo 3.3 in 10 mM pH 6.5 phosphate buffer was then titrated into this resin, and theabsorbance of the supernatant was measured three times after every addition. A 1:1 bindingmodel was then fit to the data, and the binding constant and resin loading were determined.Because binding constant and resin loading are correlated variables during the fitting, thistitration was performed three times for every resin, and the reported value is the average ofthe three trials. Results are shown in Figure 3.10.

3.7.6 Protein expression

Expression and purification of Mth1491-V92CMth1491-92C was expressed and purified according to a modified literature procedure.BL21(GoldλDE3) cells were transformed with a pJexpress404 plasmid containing the ampi-cillin-resistant gene ampR and the Mth1491-92C gene with an N-terminal His6-tag followedby a thrombin cleavage site (DNA2.0 Inc., Menlo Park, CA). Colonies were selected for inoc-ulation of lysogeny broth cultures with 100 µg/L ampicillin. When cultures reached mid-logphase as determined by OD600, expression was induced by addition of 10 mM IPTG. Cul-tures were grown for 14 to 18 h at 37 ◦C and cells were isolated by centrifugation. Cellswere resuspended in lysis buffer (50 mM NaH2PO4, pH 8.0 with 300 mM NaCl and 10mM imidazole). Cells were lysed by sonication (Fisher Scientific, USA) and cleared by cen-trifugation. The His6-tagged Mth1491-92C was purified from the cleared lysate by affinitychromatography with a Ni-NTA agarose kit, following the manufacturers protocol (Qiagen,USA). The purified protein was buffer-exchanged into cleavage buffer (50 mM Tris-HCl, pH8.0 with 10 mM CaCl2) and the His6-tag was cleaved using a Thrombin CleanCleaveTM Kit(Sigma-Aldrich, USA), incubating for 24 h at 4 ◦C. Mth1491 was isolated from the cleavedHis6-tag and residual His6-Mth1491-92C by buffer exchange into lysis buffer and performinga pull-down with Ni-NTA agarose. Collection of the supernatant yielded 20 to 40 mg of pro-tein per liter of culture. Isolated protein was exchanged into 25 mM potassium phosphatebuffer, pH 6.5 with 0.01% NaN3 and stored at 4 ◦C.


3.7.7 Protein modification

General procedure for protein transamination. To a solution of protein in 10 mMpH 6.5 phosphate buffer was added an equal volume of a 200 µM solution of pyridoxal-5’-phosphate (PLP). This solution was incubated at 37 ◦C for 1 h. Excess PLP was removedby repeated centrifugal filtration with a 10 kDa molecular weight cutoff (MWCO) membraneand 50 mM pH 5.5 phosphate buffer.General procedure for oxime formation. Transaminated protein in 50 mM pH 5.5 phos-phate buffer was added to an equal volume of 44 mM deprotected 3.10 in 50 mM pH 5.5phosphate buffer. After mixing, the reaction was rotated end-over-end on a laboratory rotis-serie at room temperature for 3 d. Excess alkoxyamine was removed by repeated centrifugalfiltration against a 10 kDa MWCO membrane.General procedure for NHS-ester coupling. To a solution of protein in 25 mM pH 8phosphate buffer was added 0.5-8 equivalents of 3.7 in DMSO. After mixing, the reactionstood at room temperature for 1-3 h, at which point excess reagent was removed by repeatedcentrifugal filtration against a 10 kDa MWCO membrane.Modification of Mth1491 with Alexafluor 350 maleimide. To 100 µL of 50 µMMth1491 in 25 mM pH 7.2 phosphate buffer was added 5 equivalents of AF350 maleimide.The reaction stood at room temperature for 1 h, at which point excess reagent was removedby repeated centrifugal filtration against a 10 kDa MWCO membrane.Cleavage of purification handles. To 100 µL of 50 µm Mth1491 in 25 mM pH 7.2phosphate buffer was added and equal volume of 50 mM sodium dithionite in 25 mM pH 7.2phosphate buffer. Prior to use, sodium dithionite was stored in a dessicator under vacuum.After 1 min, excess sodium dithionite was removed with a NAP-5 Sephadex size exclusioncolumn. Any remaining sodium dithionite was removed by repeated centrifugal filtrationagainst a 10 kDa MWCO membrane.Modification of Mth1491 with azo 2 maleimide. To 100 µL of 50 µM Mth1491 in 25mM pH 7.2 phosphate buffer was added 0.3-5 equivalents of 3.8. After mixing, the reactionstood at room temperature for 1 h, at which point excess reagent was removed by repeatedcentrifugal filtration against a 10 kDa MWCO membrane.Protection of free cysteines with Ellman’s reagent. To a solution of 100 µL of 50µM protein in 10 mM pH 6.5 phosphate buffer was added an equal volume of 10 mMDTNB in 25 mM pH 7.2 phosphate buffer with 1 mM EDTA. The reaction was incubated atroom temperature for 3 h, at which point excess reagent was removed and the protein wasexchanged into 10 mM pH 6.5 phosphate buffer by repeated centrifugal filtration against a10 kDa MWCO membrane.Oxidative coupling of RNAse A. A 10 µL portion of 100 µM RNAse A was diluted to250 µL with 25 mM pH 6.0 bis-tris. To this solution was added 50 µL of 10 mM potassiumferricyanide (100 eq. relative to protein) and 200 µL of 250 µM adamantane o-aminophenol(reduced 3.13, 10 eq. relative to protein). After 25 min, excess reagent was removed byrepeated centrifugal filtration against a 10 kDa MWCO membrane.Oxidative coupling of Mth1491. A 100 µL portion of a 50 µM Mth1491 monomer in


solution was diluted to 250 µL with 10 mM pH 6.5 phosphate buffer. To this solution wasadded 50 µL of 10 mM potassium ferricyanide (100 eq. relative to protein) and 200 µL of200 µM o-aminophenol OG514 (8 eq. relative to protein). After 25 min, excess reagent wasremoved by repeated centrifugal filtration against a 10 kDa MWCO membrane.Deprotection of free cysteines. To 100 µL of 50 µL protein was added 5 µL of 0.5 M tris-(2-carboxyethyl)phosphine (TCEP). After 10 min excess reagent was removed by repeatedcentrifugal filtration against a 10 kDa MWCO membrane.

3.7.8 Chromatography

Packing resin. Resin was packed into a 25 cm by 0.5 cm inner diameter HPLC column.Resin was first suspended in 20% ethanol in water, and this slurry was poured into a columnwith a packing adaptor fitted to the top. This suspension was then packed by flowing 20%ethanol through the column at 0.4 mL/min until the resin had settled. This procedure wasrepeated until the column was full of packed resin, and then two additional column volumeswere flowed through the column to ensure that it was packed.Column chromatography. All chromatography was performed using 20 mM pH 6.5 phos-phate buffer and 30 mM pH 6.5 phosphate buffer with 10 mM β-cyclodextrin as the eluent.Columns were washed at 0.3 mL/min, and all chromatography was performed at 0.2 mL/minat 8 ◦C.

3.8 References

1. Richard L. Kwant, Jake Jaffe, Peter J. Palmere, and Matthew B. Francis. “Controlledlevels of protein modification through a chromatography-mediated bioconjugation.”Chem. Sci. 6, 2015, 2596–2601.


3. Allie C. Obermeyer, John B. Jarman, and Matthew B. Francis. “N-terminal modifica-tion of proteins with o-aminophenols.” J. Am. Chem. Soc. 136, 2014, 9572–9579.

4. Wesley Wu, Sonny C. Hsiao, Zachary M. Carrico, and Matthew B. Francis. “Genome-free viral capsids as multivalent carriers for taxol delivery.” Angew. Chem., Int. Ed.Engl. 48, 2009, 9493–7.

5. Noboru Yoshida, Akitoshi Seiyama, and Masatoshi Fujimoto. “Stability and structureof the inclusion complexes of alkyl-substituted hydroxyphenylazo derivatives of sul-fanilic acid with α- and β-cyclodextrins.” J. Phys. Chem. 94, 1990, 4254–4259.


6. Noboru Yoshida, Hiroyuki Yamaguchi, and Miwako Higashi. “Induced circular dichro-ism spectra of α-, β-, and γ-cyclodextrin complexes with sodium 4’-hydroxy-3’-isoprop-yl azobenzene-4-sulfonate and sodium 4’-hydroxy-3’, 5’-diisopropyl azobenzene-4-sulf-onate.” J. Chem. Soc., Perkin Trans. 2, 1994, 2507–2513.

7. Noboru Yoshida and Katura Hayashi. “Dynamic aspects in host-guest interactions.Part 2. Directional inclusion reactions of some azo guest molecules with β-cyclodextrin.”J. Chem. Soc., Perkin Trans. 2, 1994, 1285.

8. Steven H. L. Verhelst, Marko Fonovic, and Matthew Bogyo. “A mild chemically cleav-able linker system for functional proteomic applications.” Angew. Chem., Int. Ed. Engl.46, 2007, 1284–6.

9. Geoffry S. Bethell, John S. Ayers, and T. W. Hearn. “A novel method of activationof cross-linked agaroses with 1,1’-carbonyldiimidazole which gives a matrix for affinitychromatography devoid of additional charged groups.” J. Biol. Chem. 254, 1979, 2572–2574.

10. Sreenivas Punna, Eiton Kaltgrad, and M. G. Finn. “‘Clickable’ agarose for affinitychromatography.” Bioconjugate Chem. 16, 2005, 1536–41.

11. Dinesh Christendat, Vivian Saridakis, Youngchang Kim, Ponni A. Kumar, Xiaohui Xu,Anthony Semesi, Andzrej Joachimiak, Cheryl H. Arrowsmith, and Aled M. Edwards.“The crystal structure of hypothetical protein MTH1491 from Methanobacterium ther-moautotrophicum.” Protein Sci. 11, 2002, 1409–1414.

12. Rebecca A. Scheck and Matthew B. Francis. “Regioselective labeling of antibodiesthrough N-terminal transamination.” ACS Chem. Biol. 2, 2007, 247–51.

13. Ying-Zhong Ma, Rebekah A. Miller, Graham R. Fleming, and Matthew B. Francis.“Energy transfer dynamics in light-harvesting assemblies templated by the tobaccomosaic virus coat protein.” J. Phys. Chem. B 112, 2008, 6887–92.

14. Nicholas Stephanopoulos, Zachary M. Carrico, and Matthew B. Francis. “Nanoscale in-tegration of sensitizing chromophores and porphyrins with bacteriophage MS2.” Angew.Chem. Int. Ed. Engl. 48, 2009, 9498–502.

15. Hoe-Sup Byun, Ning Zhong, and Robert Bittman. “6-O-p-toluenesulfonyl-β-cyclodex-trin.” Org. Syn. 77, 2000, 225.

16. Thorbjørn Terndrup Nielsen, Veronique Wintgens, Catherine Amiel, Reinhard Wim-mer, and Kim Lambertsen Larsen. “Facile synthesis of β-cyclodextrin-dextran polymersby ‘Click’ chemistry.” Biomacromolecules 11, 2010, 1710–5.

17. V. Christova-Bagdassarian and M. Atanassova. “Spectrophotometric determination ofsodium azide in workplace air.” J. Univ. Chem. Technol. Metall. 42, 2007, 311–314.

18. E. Waser and H. Sommer. “Untersuchungen in der phenylalanin-reihe II. Synthese des3,4-dioxy-phenylathylamins.” Helv. Chim. Acta 6, 1923, 54–61.


3.9 Additional figures

250 300 350 400 450 500Wavelength / nm

0.0

0.5

1.0

1.5

2.0

Abs

orba

nce

Figure 3.6: Absorbance spectrum of azo 3.3.

250 300 350 400 450 500Wavelength / nm

0.0

0.2

0.4

0.6

0.8

Abs

orba

nce

Figure 3.7: absorbance spectrum of azo 3.6.


10-5 10-4 10-3 10-20.65

0.70

0.75

0.80

0.85A

bsor

banc

e at

375

nm

Residuals

log10(Ka / M-1) = 3.37±0.0103.34±0.0153.34±0.028

Concentration of β-cyclodextrin / M

Figure 3.8: Binding of azo 3.3 to β-cyclodextrin.

log10(Ka / M-1) = 3.46±0.107

log10(Ka / M-1) = 3.45±0.061

log10(Ka / M-1) = 3.49±0.068

Concentration of β-cyclodextrin / M10-5 10-4 10-3 10-20.545

0.5500.5550.5600.5650.5700.5750.580

1.741.761.781.801.821.841.86

0.72

0.73

0.74

0.75

Abs

orba

nce

at 3

64nm

Residuals

Figure 3.9: Binding of azo 3.6 to β-cyclodextrin.


−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0log10(Volume of Sample / ml)

−0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Abs

orba

nce

at 3

30nm

Pulldown with resin 1residuals

[β-CD] / mM = 1.962.622.25

log10(Ka / M-1) = 3.383.383.46


−0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Abs

orba

nce

at 3

30nm

a


[β-CD] / mM = 2.162.502.57

log10(Ka / M-1) = 3.923.813.81


−0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Abs

orba

nce

at 3

30nm


[β-CD] / mM =1.481.682.18

log10(Ka / M-1) = 3.973.863.73

Figure 3.10: Pulldown experiments with azo 3.3 and resins 1-3.


unmodi�ed

singlymodi�ed

0 5 10 15 20 25

Alk

yne

resi

n

Azo S3

0 5 10 15 20 25

Mth1491+0 and +1 of 3

0 5 10 15 20 25Volume / mL

0 5 10 15 20 25Volume / mL

β-CD

resi

n

Alk

yne

resi

nβ-

CD re

sin

a c

db

A (9 mM linker) B (17 mM linker) C (32 mM linker)

Figure 3.11: Elution of azo 3.3 and Mth1491 modified to 30% with azo maleimide 3.5 from alkyne-and β-cyclodextrin-terminated resins. (a) Azo 3.3 elutes more slowly from resins with higherdegrees of alkyne substitution. (b) The attachment of β-cyclodextrin to the resin increases thebinding further. (c) Mth1491 interacts nonspecifically with resins that have at high alkyne sub-stitution levels. (d) The attachment of β-cyclodextrin masks these nonspecific interactions andprovides separation between unmodified and singly modified proteins. The dashed lines representa gradient of 0 mM to 10 mM β-cyclodextrin that was applied during the separation.


(a)

(c)

(b)

14000 1600013000 13500 14500 15000 15500

unmod.14117

molecular weight (Da)

unmod.14117

unmod.14117

Figure 3.12: Attempted modification of the two endogenous cysteines in His6-wt-Mth1491. Decon-voluted ESI-TOF-MS spectra of His6-wt-Mth1491 (25 uM) reacted with (a) 0, (b) 0.1, or (c) 3.0equivalents of coumarin 343-ethyl-maleimide in 25 mM potassium phosphate buffer, pH 6.5 at roomtemperature for 1 h. The molecular weight of unmodified, singly modified, and doubly modifiedprotein is 14117 Da, 14524 Da, and 14931 Da, respectively. No modification was observed.

200 300 400 500 600 700Wavelength / nm

0.00

0.02

0.04

0.06

0.08

0.10

1 cm

abs

orba

nce

Pure 2:1

Pure 1:2

0:3

3:0

Stat. 2:1

Stat. 1:2

Absorbance spectra of 2-dye systems

Figure 3.13: Absorbance spectra of protein samples discussed in the main text.

66

Chapter 4

Monolithic chromatography-mediatedbioconjugation

Abstract

Monolithic columns alleviate many of the problems associated with traditional protein chro-matography because of their lack of interpartical spaces. This chapter describes the adap-tation of commercial monolithic columns for use with chromatography-mediated biocon-jugation. Use of a monolithic silica column greatly increases the resolution and speed ofseparation. Moreover, this increase in resolution allows the separation of proteins modifiedin different locations in some cases. Ubiquitin is used as a case study to demonstrate thepotential of this method for site-specific protein modification.

Haiku

One big moleculeBetter than many small onesFaster and more exact

Portions of the Materials and Methods portion of this chapter were taken from the supportinginformation of a separate publication by Kwant et al.1 These sections appear here in amodified form and are licensed under a CC BY 3.0 license.


CHAPTER 4. MONOLITHIC CHROMATOGRAPHY-MEDIATEDBIOCONJUGATION 67

4.1 Introduction

As discussed in Chapter 3, we developed an alternative method for the modification ofproteins in which chromatography is used to control the number of modifications. In thismethod, a protein is first modified with a chemical handle consisting of an azobenzenederivative. The crude reaction mixture is purified via affinity chromatography based onan affinity interaction with the handle. The handles are then elaborated with a variety ofchemical moieties via an oxidative coupling.2,3

The purity of bioconjugates resulting from this method depends on the resolution ofthe chromatography step. We were able to separate proteins with different numbers ofmodifications based on the multivalency of their interaction with the resin. In some casesthere were indications that proteins were also separated based on the local environmentaround the modification site (see Figure 3.4), for some singly modified peaks appeared toelute as multiple overlapping peaks. However, we were unable to resolve these species wellenough to identify them. Nevertheless, these results encouraged us to explore alternativechromatographic media to improve the resolution of the separation.

One limitation of chromatography with large biomolecules is their slow mass transfer ki-netics. Traditional chromatographic media, including Sepharose, consist of small particles, sothe partitioning of large molecules like proteins into and out of these particles is slow. Theseslow mass transfer kinetics result in peak broadening and a decrease in the overall resolutionof the method. This problem can be circumvented by using media that do not have parti-cles, specifically monolithic polymers. Monolithic columns consist of one extended moleculepermeated by a series of through-pores with well-defined size. Thus they lack extrapartic-ulate space and have excellent mass transfer characteristics, even for large biomolecules.4

Practically speaking, they allow higher resolution separations for larger molecules at higherflow rates and lower pressure drops. In this chapter, we show that adoption of monolithicmedia enables significantly higher resolution separations during the chromatographic stepof chromatography-mediated bioconjugation. This higher resolution enables the isolation ofproteins with increased purity, and it likely allows separation of proteins modified in differ-ent locations. As a case study, we investigate the potential of this method to monitor theprogress of bioconjugation reactions performed on ubiquitin.

4.2 Choice of monolithic support

Monolithic columns are a relatively recent arrival for the field of chromatography, withmuch of their development taking place in the 1990s.4 Three broad categories of monolithicmedia have been developed: silica monoliths,5 hydrophilic polymer monoliths,6,7 and acrylatepolymer monoliths.8–13 We first focused on acrylate polymer monoliths because of theirprior use with proteins in a broad range of applications, their ease of synthesis, and theirdevelopment at UC Berkeley and Lawrence Berkeley National Lab (LBNL).

In collaboration with LBNL, a series of methacrylate and glycidyl methacrylate poly-


NH2 NH

NH2

O

NH

O

NH

HN

O

NH

O

O4.1 4.2 4.3

Figure 4.1: Initial modification of silica monolith. The monolith was first silanized withaminopropyltrimethoxysilane to afford monolith 4.1 modified with primary amino groups. Tothese amines was then coupled carboxy-β-cyclodextrin to afford monolith 4.2 modified with β-cyclodextrin. Initial characterization of the column indicated that ionic interactions between thestationary phase and the mobile phase may contribute to biomodal peaks, so amines were cappedwith acetic anhydride to afford monolith 4.3.

mers were synthesized. Modification with β-cyclodextrin was attempted using a varietyof chemistries, including amide bond formation, epoxide opening, and disulfide formation.Despite precedent for the modification of acrylate polymers with β-cyclodextrin,14–17 littleconvincing evidence was generated to show substantial modification of the polymers withβ-cyclodextrin, and no selective retention of proteins modified with azobenzene derivativeswas demonstrated.

Based on these results, attention was shifted to monolithic silica rods. Very few typesof silica monoliths exist, with essentially one patented format available for preparatory usefrom either Millipore or Phenomenex.4 These columns are also difficult to prepare in alaboratory setting, so a large portion of the literature focuses on uses and modificationsof these commercially available silica monoliths. As a result, modification of silica rodmonoliths is well-understood, and robust methods exist for their modification.4 Silica rodcolumns are commonly used to separate enantiomers, and there is literature precedent forthe modification of their surfaces with β-cyclodextrin to make chiral stationary phases.18–20

Therefore despite the limited use of silica rod monoliths for the purification of proteins,we investigated the ability of these materials to separate proteins based on their chemicalmodification.

4.3 Modification and characterization of monolithic

silica columns

4.3.1 Initial modification

Commercially available 10 cm by 4.6 mm inner diameter columns were purchased fromboth Phenomenex and Millipore, and their initial modification was carried out accordingto procedures described by Lubda et al. (Figure 4.1).18 Lubda et al. demonstrated that itwas possible to uniformly modify a pre-cladded silica rod with aminopropyltrimethoxysilaneto a final concentration of about 2.8 µmol/m2. Based on the manufacturer’s specifications


0 2 4 6 8 10

0 5 10 15 20

Volume / ml

A280A330%B

Volume / ml

i. Elution of azo from monolith

ii. Elution of partially modi�ed Mth1491 from monolith

Figure 4.2: Elution of azo 3.3 and Mth1491 from monolith 4.2. (i) Elution of azo 3.3 with amobile phase of 10 mM β-cyclodextrin. The azo compound did not elute when no β-cyclodextrinwas added to the mobile phase. Elution of the azo after the void volume (1.4 mL) indicates that thedye binds more strongly to the stationary phase than to the mobile phase. (ii) Elution of Mth1491partially modified with azo 3.5. A linear gradient of 0 to 10 mM β-cyclodextrin was applied tothe column over 10 min. Increasing absorbance at 330 nm corresponds to more highly modifiedprotein.

that each column is roughly 300 m2 and 0.5 g with a void volume of 1.4 mL, this valuecorresponds to a concentration of about 300 mM, which is more than sufficient according tocalculations in Chapter 2.

Initial attempts at modification of these columns using Click chemistry (as in Chapter 3)did not produce columns with acceptable properties. These columns did show evidence forincorporation of β-cyclodextrin, but the unfortunately also retained unmodified proteins.However, modification using only amide bond-forming reactions proved more fruitful (Fig-ure 4.1). Amino monolith 4.1 was reacted with carboxy-β-cyclodextrin 4.9 to afford amonolithic support modified with β-cyclodextrin (compound 4.2). This resin showed a se-lective retention of azo 3.3 that could be disrupted by eluting with 10 mM β-cyclodextrin(Figure 4.2.i). The chromatogram revealed multimodal binding characteristics, suggestingthat the column was interacting with the dye in at least two different ways. Given thatazo 3.3 is negatively charged and could interact with the positively charged monolith, thisbehavior could be explained by additional ionic interactions. To further characterize themonolith, Mth1491 92C was partially modified with azo maleimide 3.5 and was eluted fromthe column with a linear gradient from 0 to 10 mM of β-cyclodextrin. Despite the appear-ance of bimodal peaks in this chromatogram as well, several distinct peaks elute, each withan increasing ratio of the absorbance at 330 nm to the absorbance at 280 nm. These changesin absorbance are characteristic of increasing modification with the azo dye and indicate theability of the resin to separate the modification levels of Mth1491 92C. Still more encourag-


0 2 4 6 8 10

0 5 10 15 20

Volume / ml

Volume / ml

i. Elution of azo from monolith

ii. Elution of partially modi�ed Mth1491 from monolith

A280A330%B

A280A330

Unmodi�ed

Singly modi�ed Doubly modi�ed

Figure 4.3: Elution of azo 3.3 and Mth1491 from monolith 4.3. (i) Elution of azo 3.3 frommonolith 4.3 with 10 mM β-cyclodextrin. Retention past the void volume (1.4 mL) indicates thatthe dye interacts more strongly with the β-cyclodextrin on the monolith than the β-cyclodextrinin solution. (ii) Elution of Mth1491 92C partially modified with azo maleimide 3.5 from monolith4.3 under a linear gradient of 0-10 mM β-cyclodextrin. The column separates modification levelscompletely and exhibits excellent peak shapes.

ing, the unmodified and singly modified peaks, which elute around 1.5 mL, were significantlynarrower than peaks that eluted from Sepharose-based columns.

Encouraged by these results, we decided to cap the remaining amino groups on themonolith to remove the potential for ionic interactions between the stationary and mobilephases. Monolith 4.2 was reacted with acetic anhydride to acylate any remaining free amines,and subsequently exposed to an excess of ethylenediamine to cleave esters formed during theacylation step to yield monolith 4.3.

This monolith was then again tested with azo 3.3 and Mth1491 92C partially modifiedwith azo 3.5 (Figure 4.3). Acylation of the monolith eliminates the multimodality seen inFigure 4.2, and it also increases the selective retention of azo compounds and their conjugates.In the case of azo 3.3 (Figure 4.3.i), retention increases from about 2.5 mL to 4 mL as thepeak becomes significantly narrower. The elution of the azo dye at a volume greater thanthe void volume (1.4 mL) indicates that the dye interacts more strongly with the stationaryphase than with the β-cyclodextrin in the mobile phase. Elimination of possible charge-charge interactions actually increases the affinity for the stationary phase, potentially dueto charge-dipole interactions with the acyl groups on the resin. This phenomenon wasalso seen in Chapter 3 with Sepharose-based stationary phases. Purification of partiallymodified Mth1491 92C indicates that monolith 4.3 is able to separate the modification statesof Mth1491 with significantly higher resolution than Sepharose-based resins (Figure 4.3.ii,compare to Figure 3.4.f). Each modification state is well-separated, and exhibits a peak


shape that is near-Gaussian. As with azo 3.3, singly modified Mth1491 elutes later than itwould if it interacted with the stationary and mobile phases with the same strength. In thiscase, the increased strength of the interaction with the stationary phase causes the triplymodified protein not to elute even after application of 10 mM β-cyclodextrin to the columnfor 10 min.

4.3.2 Optimization of modification

Before continuing, we wanted to optimize the characteristics of the monolithic stationaryphase. Acylation of monolith 4.2 does increase the retention of unmodified Mth1491 92C by0.14 min. While slight, this increase can correspond to a 1-2 order of magnitude increase inthe affinity constant of the non-specific interaction with the stationary phase. Moreover, thestrong preference of the azo dye for the stationary phase could prevent separation of morehighly modified species, so it could be preferable to attenuate this favorable interaction.

Characterization of resins in Chapter 3 indicated that steric effects most likely limit themaximum amount of β-cyclodextrin that can be incorporated onto the stationary phase,and the same likely holds true for monoliths as well. Based on the results of Lubda et al.,the surface concentration of amines is far in excess of the possible surface concentrationof β-cyclodextrin. Therefore rather than focus on methods to change the amount of β-cyclodextrin on the monolith, we focused our intention on the identity of the acylatingreagent and the overall concentration of amines on the monolith. Given the expense ofmonolithic silica columns ($600-$1100/column) and the amount of labor required to modifythem (2-3 days/column, difficult to parallelize without purchasing multiple syringe pumps),the number of experiments that could be reasonably performed was limited to three at most.

To gain insight into potential modifications, silica gel was used as a proxy for silicamonoliths. A synthetic scheme similar to the one depicted in Figure 4.1 was carried out,but acetic acid, glycolic acid, and 2,5-dihydroxybenzoic acid were used as acylating agents.These agents were chosen because after their incorporation, the silica surface should betterresemble Sepharose, which is primarily covered in hydroxyl groups and is known not tointeract significantly with most proteins. These resins were used to perform pulldowns ofazo 3.3, and binding curves were fit to the results. The fits (Figure 4.10) indicate that thebinding of the azo does depend somewhat on the identity of the blocking group, with thefit KA values ranging from 103.7 M−1 in the case of glycolic acid to 104.0 M−1 in the caseof 2,5-dihydroxybenzoic acid. However, incorporation of glycolic acid instead of acetic acidweakens the KA by only 0.1 on a log scale (103.8 M−1 to 103.7 M−1), and this slight decreasewould not have a significant effect on the strong retention of highly modified proteins.

This limited evidence suggested that the amide groups on the monolith surface maycontribute to its increased binding of azo 3.3. This hypothesis is consistent with the resultsseen in Chapter 3, in which the presence of excess propargyl amide groups on the Sepharosesurface was shown to have a concentration-dependent effect on the binding constant betweenazo 3.3 and the resin. But unlike with Sepharose, it was not possible to directly change theoverall modification of the monolith because it could not be derivatized in a batch setting.


0 2 4 6 8 10Volume / ml

A280A330

i. Monolith4.3

Elution of azo from monolithic columns, 10 mM BCD

ii. Monolith4.6

iii. Monolith4.7

Figure 4.4: Elution of azo 3.3 from monoliths 4.3 (i), 4.6 (ii), and 4.7 (iii) with an isocratic of 10mM β-cyclodextrin. The strength of the interaction between the azo and the monolith can be tunedby the polarity of the blocking group. The most polar blocking group, cyanopropyltrimethoxysilane(monolith 4.7), results in the weakest retention, whereas the most polar blocking group, propy-ltrimethoxysilane (monolith 4.6), results in the strongest retention. Note that cyano monolith 4.7showed bimodal elution profiles and was likely cracked or not uniformly modified.

Instead, monoliths were modified with a 1:1 v/v mixture of aminopropyltrimethoxysilaneand either propyltrimethoxysilane or cyanopropyltrimethoxysilane. This approach ensuredthat the monolith surface was uniformly modified yet still allowed variation in the concen-tration of amines. These two silanes were chosen because they were similar to aminopropy-ltrimethoxysilane in size but exhibit different polarities that could allow further tuning of thecolumn properties. These columns were then modified as was monolith 4.3 to give monoliths4.6 (propyltrimethoxysilane-modified) and 4.7 (cyanopropyltrimethoxysilane-modified).

These monoliths were first characterized by their retention of azo 3.3 with both 0 and10 mM β-cyclodextrin in solution as eluent. In all cases the columns retained azo 3.3longer than 10 mL with no eluent present. When 10 mM β-cyclodextrin was added to thebuffer, all columns eluted the dye within 10 min, indicating that retention of the dye couldbe disrupted by β-cyclodextrin and was likely mediated by a specific binding interaction(Figure 4.4). However, the retention of azo 3.3 varied among the three monoliths, with theretention roughly proportional to the polarity of the blocking silane. Cyano monolith 4.7,which contained the polar cyano silane as a blocking group, showed the weakest retention ofazo 3.3, with an elution volume of about 3.5 mL. In contrast, propyl monolith 4.6 whichcontained the non-polar propyl silane as a blocking group, retained azo 3.3 more stronglythan both monolith 4.3 and 4.7, with an elution volume of almost 6 mL. These results


0 5 10 15 20Volume / ml

Elution of partially modi�ed Mth1491 92C from columns

i. Monolith4.3

ii. Monolith4.6

iii. Monolith4.7

A280A330

Figure 4.5: Elution of Mth1491 92C partially modified with azo maleimide 3.5 from monoliths 4.3(i), 4.6 (ii), and 4.7 (iii) with a linear gradient of 0-10 mM β-cyclodextrin. Two characteristicswere used to evaluate the performance of the columns: elution volume of the unmodified peak(left-most) and elution volume of the singly modified peak (second from the left). The elutionvolume of the unmodified peak can be compared to the void volume (small negative dip in the330 channel). Elution volumes at or less than the void volume indicate columns with little to nonon-specific affinity for proteins. Retention of the singly modified protein is largely mediated bythe interaction between the azo and the monolith, and tracks with the results seen in Figure 4.4.

are consistent with intuition, which would suggest that the mostly non-polar azo 3.3 wouldinteract more strongly with a surface that is itself non-polar. More importantly, these resultsconfirm the hypothesis that it is possible to tune the interaction strength of azo dyes withthe monolithic columns by varying the chemical moiety that back-fills the space not occupiedby β-cyclodextrin.

These columns were then characterized by their retention of Mth1491 92C that was par-tially modified with azo 3.5 (Figure 4.5). Two features were considered—retention of singlymodified protein and retention of unmodified protein. The retention of singly modified pro-tein tracked with the retention of azo 3.3 in that less polar monoliths retained the singlymodified protein more. This result is consistent with the fact that retention of the protein ismediated by a specific interaction between azo 3.3 and the monolith. Retention of unmodi-fied protein indicates non-specific interactions between the protein and the monolith surface,and ideally these interactions should be minimized. The degree of non-specific retention caneasily be seen by comparing the elution volume of the first peak to the small negative peakin the 330 nm channel, which marks the void volume of the column. Retention of unmodifiedproteins was more complicated than retention of azo 3.3, with both the propyl monolith 4.6


and the cyano monolith 4.7 having smaller retention volumes. This result suggests that theexcess of amide groups on the monolith surface interacts favorably with Mth1491. Betweenpropyl monolith 4.6 and cyano monolith 4.7, cyano monolith 4.7 shows the weakest non-specific interactions with Mth1491, and the protein actually elutes before the void as a resultof size-exclusion.

It should be noted that all three of the columns exhibited excellent retention character-istics and could potentially be used to purify chemically modified proteins. Monolith 4.3and propyl monolith 4.6 were both used in further experiments with additional proteins.Propyl monolith 4.6 was judged as the most promising because it combined strong specificretention of azo 3.3 with with moderately weak non-specific retention of Mth1491. How-ever, it is worth noting that the weak non-specific interactions of cyano monolith 4.7 withMth1491 may indicate that this column can be used with the largest number of proteins. Asdiscussed in Section 4.3.3, non-specific interactions may prevent the purification of some pro-teins. Unfortunately, because cyano monolith 4.7 was either cracked or unevenly modifiedand showed poor resolution, this modification chemistry was not explored further.

4.3.3 Characterization of protein folding

One concern when developing a new chromatographic technique for proteins is whether thattechnique denatures proteins. In many cases, like reverse-phase HPLC, proteins will elute,but their fold will not be the same as when the protein was injected onto the column.This concern was particularly salient because unmodified silica columns typically denatureproteins, even if there are only a few patches of exposed silica.

Two experiments provided evidence that these columns do not denature at least someproteins. First, Mth1491 92C that is partially modified with azo maleimide 3.5 elutes fromthese column as four distinct peaks. These peaks are indicative of the four modificationlevels of the protein, and they would not appear if the protein were misfolded because thenoncovalent homotrimer would disassemble.

This finding was confirmed by comparing the fluorescence of green fluorescent protein(GFP) before and after elution from a modified monolithic column. When GFP denatures, itaggregates and its fluorescence is quenched.21 The absorbance and fluorescence of a sample ofGFP was characterized before and after elution from propyl monolith 4.6 (Figure 4.11). Thenormalized fluorescence of the protein does not change significantly, and therefore indicatesthat the protein remains folded.

However, while investigating the ability of these columns to separate the modification lev-els of proteins (discussed in Section 4.4), some proteins—rabbit aldolase, α-chymotrypsinogen,and to a lesser extent papain and creatine kinase—eluted as broad peaks from the column.These proteins were used as-is from Sigma-Aldrich without further characterization or pu-rification, so it is not clear whether the proteins were impure or misfolded, or whether theproteins denatured on the column or showed strong non-specific affinity for the columns.The answers to these questions will help determine the general applicability of this method.However, here we focus on demonstrating a proof of concept with a limited set of proteins.


4.4 Purification of modification levels

With the synthesis and preliminary characterization of three monoliths complete, we char-acterized their ability to separate complex mixtures of proteins with different numbers ofchemical modifications. RNAse A and ubiquitin were selected as model proteins becausethey have relatively few lysines (7 and 10, respectively) and are moderately sized globularproteins. These proteins were labeled with 1-2 equivalents of azo NHS-ester 3.4, and subse-quently purified using either propyl monolith 4.6 (ubiquitin, Figure 4.6.i) or monoliths 4.3and 4.6 connected in series (RNAse A, Figure 4.6.ii). Selected fractions from these runswere then analyzed via LCMS-TOF to determine the number of chemical modifications oneach protein.

For both proteins it is possible to isolate fractions that contain protein with a givenmodification level. The sample of ubiquitin consists of primarily singly and doubly modifiedprotein, and both elute as multiple well-defined peaks with peak widths less than 2 mL. Thesample of RNAse A is more heavily modified and consists of all modification levels betweenunmodified and triply modified. These peaks again elute roughly in order of number ofmodifications, with more modified protein eluting later than less modified protein.

The resolution of these chromatograms is significantly better than required to separatemodification levels. This improvement in resolution reveals an underlying set of peaks foreach modification level, with at least four distinct peaks of ubiquitin and five distinct peaksof RNAse A corresponding to singly modified protein. Because each modification level elutesas a set of multiple peaks, there must be additional factors beyond the absolute number ofchemical modifications that affect the elution volume of a modified protein. For example,the local chemical environment around a chemical modification site could restrict accessto the β-cyclodextrin-modified monolith surface, or itself interact with the surface. Thesephenomena would lead to different effective binding constants for the azo handles at eachpotential modification site.

Moreover, in both chromatograms in Figure 4.6 the second peak corresponding to modi-fied protein contains doubly modified protein. Based on the modeling described in Chapter 2,this relatively early elution can be explained only by invoking negative coopertivity, in whichthe second binding event is weakened by the first binding event. Negative coopertivity couldresult from steric limitations, in which two modifications on opposite sides of the protein areseldom able to both interact with the monolith surface at the same time, or two modificationsites are so close to one another that the binding of one blocks the binding of the other.

4.4.1 Site-specific purification of Mth1491

Regardless of the underlying mechanisms that produce the sets of peaks seen in Figure 4.6,it is clear that monolithic columns separate chemically modified proteins based on morethan just their absolute number of modifications. We set out to test whether monolithiccolumns were able to separate proteins that are modified in different locations by modifyingtwo different mutants of Mth1491: one with a cysteine at position 30 and another with a


0 10 20 30 40Volume / ml

0 5 10 15 20Volume / ml

+1+2

+1+0

+2+3

A280A330

i. Separation of modi�cation levels of ubiquitin

ii. Separation of modi�cation levels of RNAse AA280A330

Figure 4.6: Purification of proteins labeled with azo NHS-ester 3.4. Spectra under each chro-matogram correspond to LCMS-TOF traces of selected fractions rotated 90 degrees clock-wise.Dashed lines correspond to discrete modification levels of unmodified, singly modified, doublymodified, and triply modified. (i) Ubiquitin purified with propyl monolith 4.6 with a linear gra-dient of 0-10 mM β-cyclodextrin. (ii) RNAse A purified with monoliths 4.3 and 4.6 connected inseries with a linear gradient of 0-10 mM β-cyclodextrin.

cysteine at position 92. These two locations have distinct environments (Figure 4.7). Position92 is on a flat face of the protein and is therefore quite solvent accessible. On the other hand,position 30 is on one end of a groove that cuts down the side of the protein. As a result, thislocation is more sterically hindered. Naıvely, one might expect proteins modified at position92 to be retained more strongly because an azo dye at this location could better interact withβ-cyclodextrin immobilized on a surface. Purification of these proteins partially modifiedwith azo maleimide 3.5 validates this hypothesis (Figure 4.7). The singly modified peakof Mth1491 30C elutes nearly 2 mL before the singly modified peak of Mth1491 92C. Thisresult indicates that the chemical environment around a modification site can play a largerole in the retention volume of that protein species, and as a result this method is able toseparate proteins that are modified in different locations.


280330

0 5 10 15 20ml

0 5 10 15 20ml

i. MTH1491 30C

ii. MTH1491 92C

Position 30

Position 92

Groove

Figure 4.7: Analysis of Mth1491 modified in two different locations with propyl monolith 4.6. (i)Mth1491 30C modified with azo maleimide 3.5. (ii) Mth1491 92C modified with azo maleimide3.5. Elution volumes of singly modified peaks are marked with dashed lines. The elution of thesingly modified peaks at different volumes indicates that retention volume is dependent on the siteof chemical modification. The additional peaks likely result from impurities in the sample, such asfree azo dye.

4.4.2 Site-specific purification of ubiquitin

We hypothesized that each of the peaks seen in Figure 4.6 corresponded to protein thatwas modified in different locations, and used ubiquitin as a model protein to investigate thishypothesis. Ubiquitin was selected because of its small size (76 amino acids, 8.5 kDa) andrelatively small number of lysines (7) (Figure 4.8). The reactivity of each amine on the surfaceof ubiquitin has been well-characterized, with multiple groups reaching similar conclusions.Jabusch and Deutsch performed digests of ubiquitin modified with p-nitrophenyl acetateand isolated individual peptides with HPLC.22 Edman degradation of the isolated productsrevealed the identity of each peptide. This analysis indicated the following order of reactivity:Lys 6 > Lys 11 ≈ Lys 33 ≈ Lys 48 ≈ Lys 63 > Lys 27 ≈ Lys 29. Novak et al. were ableto validate these results and determine the order of reactivity more finely by using Fouriertransform mass spectrometry to characterize the order in which the amines on ubiquitin aremodified. They found the following order of reactivity: Met 1 ≈ Lys 6 ≈ Lys 48 ≈ Lys 63


MQIFVKTLTGKTITLEVEPSDTIENVKAKIQDKEGIPPDQQRLIFAGKQLEDGRTLSDYNIQKESTLHLVLRLRGG1 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75

Lys 6

Lys 11

Lys 63

Met 1 Met 1

Lys 29

Lys 33Lys 27

Lys 27

Lys 29

Lys 48

Figure 4.8: Crystal structure and sequence of bovine ubiquitin. The eight amines (seven lysinesand the N-terminus) are highlighted in red, and the sequence is included below the protein forreference.

> Lys 33 > Lys 11 > Lys 27, Lys 29.23 The authors also note that these reactivities trackwith the electrostatic potentials of each amine, although there are some discrepancies thatmay be accounted for by taking into account sterics. During this analysis, modification oflysine 11 was observed only after four other lysines had been modified, and modificationof lysines 27 and 29 was not observed. Thus for ubiquitin modified with one equivalent ofazo NHS-ester 3.4, it would be reasonable to see five or fewer peaks corresponding to singlymodified protein, and the four peaks seen in Figure 4.6 are encouraging.

To determine whether these monolithic columns were able to separate proteins modified indifferent locations, we purified samples of ubiquitin modified with either azo NHS-ester 3.4 orazo NHS-ester 3.7 and focused our attention on several of the fractions. These fractions wereanalyzed in three different ways—full peptides were analyzed via LCMS-TOF, fragmentedpeptides were analyzed via LCMS-triple quadrupole, and fragmented peptides were alsoanalyzed via LC-MS/MS through the UC Berkeley Proteomics/Mass Spectrometry Facility.In all cases the data were noisy, and it was difficult to reach a conclusive answer due towidely varying ionizabilities of many peptides. However, the sum of these results presented aconsensus picture that suggested that different peaks are indeed proteins modified in differentlocations, and two of these experiments are presented here.

In the first experiment, ubiquitin was modified with azo NHS-ester 3.7, purified withmonolith 4.3 and propyl monolith 4.6 connected in series, cleaved with sodium dithionite,and trypsinized. The cleavage step resulted in peptides modified with an aniline moiety toincrease their overall ionization. The resulting peptides were analyzed intact using an LCMS-TOF (Figure 4.9.i). While the ionization of modified peptides was low, the results suggestthat the major peak contained protein modified at or near its N-terminus. This result isconsistent with previous findings that the most reactive lysines are near the N-terminus of theprotein. The peaks near 15 and 17 mL did not contain any modified peptides that ionized.


However, a fraction centered on 21 mL suggested that the protein was likely modified atposition 48 or position 33.

These results were replicated by another experiment in which ubiquitin was modifiedwith azo NHS-ester 3.4. The same procedure was followed as in the previous experiment,except the azo was not cleaved after purification, and the protein was purified with twopropyl monolith 4.6 columns connected in series, such that the retention volumes are signif-icantly larger. Aminophenols are rather unstable because of their proclivity for oxidation, sotheir cleavage could make the analysis more difficult. These samples were then analyzed viaLC-MS/MS. These results were more difficult to analyze because of the presence of manymodified peptides in each fraction. However, three modification sites stood out, and thesesites were consistent with the results from the previous experiment. Two parameters wereused to judge the presence of a modified peptide: the number of ion counts and the prob-ability score. The number of ion counts corresponds to the number of ions observed thatbelong to a particular peptide, and the probability score is a measure of the confidence in thepresence of those peptides. For each fraction, one modification site stood out that had boththe largest ion count and the highest probability score. As with the TOF results, the largestpeak contains predominantly protein that is modified near the N-terminus of the protein.Similarly, the last peak corresponds to protein that was likely predominantly modified atposition 48. The middle peak likely corresponds to protein modified at position 63. The sizesof these peaks are consistent with the orders of reactivity reported previously. To verify thatthe abundance of each species was to some degree proportional to the observed ion counts,peaks 1 and 2 were mixed together and reanalyzed. This analysis revealed ion counts thatwould be achieved roughly by adding the two ion counts of the two samples together.

Taken together, these results do not prove that each peak corresponds to a pure species.This claim is difficult to prove, especially with mass spectrometry data, which notoriouslyhas a high background and is not quantitative unless standards are prepared. However,these results do indicate that the predominant location of modification changes as specieselute. These results, the results in Section 4.4.1, and the presence of distinct peaks eachcorresponding to the same modification level indicate that these monolithic columns separateproteins based on their degree of modification and the local chemical environment aroundthe modification site.

4.5 Conclusions

This chapter describes the development of monolithic columns for chromatography-mediatedbioconjugation and shows their potential to increase the speed and resolution of the chro-matography step. These columns increase resolution enough to separate proteins that aremodified to the same extent but in different locations. This method has the potential toenable site-selective modification of proteins whose controlled modification is challenging,and it could prove general if the chromatography step is compatible with a wide range ofproteins. Ongoing work in the Francis lab will evaluate these questions.



4.6.1 General methods

Unless otherwise noted, the chemicals and solvents used were of analytical grade and wereused as received from commercial sources. Water (dd-H2O) used as a reaction solvent wasdeionized using a Barnstead NANOpure purification system. Centrifugations were performedwith an Eppendorf Mini Spin Plus (Eppendorf, Hauppauge, NY).


Mass Spectrometry. Matrix-assisted laser desorption/ionization time-of-flight mass spec-trometry (MALDI-TOF MS) was performed on a Voyager-DE system (PerSeptive Biosys-tems, USA). Small molecule samples were co-crystallized with α-cyano-4-hydroxycinnamicacid in 1:1 acetonitrile (MeCN) to H2O with 0.1% trifluoroacetic acid (TFA). Cyclodextrinswere co-crystallized with 2,5-dihydroxybenzoic acid in 1:1 MeCN to H2O with 0.1% TFA.Protein bioconjugates were analyzed using an Agilent 1200 series liquid chromatograph (Ag-ilent Technologies, USA) that was connected in-line with an Agilent 6224 Time-of-Flight(TOF) LC/MS system equipped with an electrospray ion source. Extracted mass spectrawere plotted using chartograph.com/ms. Digested peptides were also analyzed on an AB-SCIEX 3200 Qtrap equipped with an ESI Turbo V ion source. Samples were also submittedto the Vincent J. Coates Proteomics/Mass Spectrometry facility for LC-MS/MS analysis.High Performance Liquid Chromatography. HPLC was performed on Agilent 1100 Se-ries HPLC Systems (Agilent Technologies, USA). Sample analysis for all HPLC experimentswas achieved with an in-line diode array detector (DAD) and in-line fluorescence detector(FLD). Analytical reverse-phase HPLC of small molecules was accomplished using a C18stationary phase and a H2O / MeCN with 0.1% TFA gradient mobile phase.UV-Visible Spectrometry. UV-visible spectrometry was performed using quartz cuvetteswith a Varian Cary 50 spectrophotometer (Agilent, USA). Small-scale UV-visible spectrom-etry was performed using a Nanodrop 1000 (Thermo Scientific, USA). Absorbance mea-surements of samples in plates were obtained with a SpectraMaxM2 (Molecular Devices,Sunnyvale, CA).Fluorescence. Fluorescence measurements were obtained on a Fluoromax-4 spectrofluo-rometer equipped with automatic polarizers and a Peltier temperature controller (ISA In-struments, USA). Slit widths were set to 1.0 nm for excitation and 1.0 nm for emission.Fluorescence emission was monitored with a 0.5 s integration time.

4.6.3 Modification of silica monoliths

Unless otherwise noted, all solvents used in this section were purified with a solvent purifi-cation system and filtered through a 0.2 µm syringe filter prior to use. Modification wascarried out using a KdScientific syringe pump loaded with 5-20 mL syringes.

chartograph.com/ms


NH2 Synthesis of amine-functionalized silica monolith (4.1). This pro-cedure was adapted from Lubda et al.18 The monolithic columns were un-capped and heated at 100 ◦C in an oven for at least 18 hrs to remove

adsorbed surface water, even though the manufacturer specifically states that the maximumoperating temperature of the column is 45 ◦C. The column was then attached to a syringepump and 15 mL of toluene was flowed through the column at 0.5 mL/min. The columnassembly, while still attached to the syringe pump, was placed into a water bath held at45 ◦C. A 40% v/v solution of aminopropyltrimethoxysilane in toluene was then pumpedthrough the column at 0.5 mL/min for 2 mL and then at 0.1 mL/min for 1 h. The columnwas then rinsed with 20 mL of toluene at 0.2 mL/min.

NH

NH2

O

NH

O Synthesis of β-cyclodextrin-functionalized silica mono-lith (4.2). Monolith 4.1 was attached to a syringe pumpand flushed with 9 mL of pyridine at 0.2 mL/min. A so-lution of 500 mg of β-cyclodextrin 4.9, 124 µL of N,N ’-

diisopropylcarbodiimide, and 46.1 mg of N -hydroxysuccinimide in 6 mL of pyridine and4 mL of DMF was prepared. This solution was pumped through the column at 0.2 mL/minfor 2 mL, then 0.5 mL/hr for 8 mL while the column was heated to 30 ◦C. The column wasthen washed with 9 mL of pyridine at a flow rate of 0.2 mL/min.

NH

HN

O

NH

O

O

Synthesis of acylated β-cyclodextrin-functionalized silicamonolith (4.3). Monolith 4.2 was flushed with 9 mL of pyridineand subsequently heated to 45 ◦C in a water bath. A solutionof 15% v/v acetic anhydride in pyridine was pumped through

the column at 0.1 mL/min for 1 h. The column was rinsed with 5 mL of pyridine at 0.5mL/min, followed by 15% v/v ethylenediamine in pyridine at 0.1 mL/min for 1 h. Thecolumn then stood at room temperature overnight. The following day, it was washed with 5mL of pyridine and 10 mL of water, each at 0.1 mL/min.

NH2

50%

50%Synthesis of amine- and propyl-functionalized silica monolith(4.4). This procedure was identical to the synthesis of monolith 4.1.However, during the silanization of the column, a solution of 20% v/v

aminopropyltrimethoxysilane, 20% v/v propylmethoxysilane in toluene was used.

NH2

N 50%

50%

Synthesis of amine- and cyano-functionalized silica monolith(4.5). This procedure was identical to the synthesis of monolith 4.1.However, during the silanization of the column, a solution of 20% v/v

aminopropyltrimethoxysilane, 20% v/v cyanopropylmethoxysilane in toluene was used.

NH

HN

O

NH

O

O

Synthesis of acylated β-cyclodextrin- and propyl-functionalized silica monolith (4.6). This monolith was pre-pared in the same way as monolith 4.3, except that monolith 4.4was used at the beginning of the synthetic scheme.


NH

HN

O

NH

O

O

N

Synthesis of acylated β-cyclodextrin- and cyano-functionalized silica monolith (4.7). This monolith was pre-pared in the same way as monolith 4.3, except that monolith 4.5was used at the beginning of the synthetic scheme.


O

OHHO

OH

O

OOH

HONH2

O

OOH

OH

OH

O

OOH

OH

OH

OO

OH

OH

HOO

OOH

OHHO

O

OOH

HO

HO

O

Synthesis of 6-amino-β-cyclodextrin (4.8). This procedure was adapted from Muder-awan et al.24 A 25 mL round bottom flask was charged with 2.42 g of 6-azido-β-cyclodextrin3.2 in 12 mL of DMF. To this solution was added 0.6 g (1.1 eq.) of triphenylphosphine, andthe reaction was stirred at room temperature for 2 h. Water (0.522 mL) was added, and thereaction was heated to 100 ◦C for 45 min. The reaction was cooled to room temperatureand stirred for an additional 2 h. The product was precipitated by slow addition of the crudereaction mixture to 800 mL of vigorously stirred acetone. This suspension was allowed tosettle for approximately 1 h, and the solid product was collected using a Buchener funnel.The product was dried under vacuum to give 1.9 g (80%).

O

OHHO

OH

O

OOH

HOHNO

OOH

OH

OH

O

OOH

OH

OH

OO

OH

OH

HOO

OOH

OHHO

O

OOH

HO

HO

OO O

OH

Synthesis of carboxy-β-cyclodextrin (4.9). 6-amino-β-cyclodextrin 4.8 (795 mg, 1eq.) was dissolved in 4 mL of DMF. To this solution was added 104 mg (1.3 eq.) of glutaricanhydride and 70.7 mg (97.71 µL, 1 eq.) of triethylamine. The reaction was stirred overnightat room temperature, and then poured slowly into 800 mL of vigorously stirred acetone. Theresulting suspension was allowed to settled for 2 h. The product was isolated by filtrationthrough a Buchener funnel and dried under vacuum to yield 700 mg (80%) of material.MALDI-TOF MS calculated for C47H77NNaO +

37 ([M+Na]+) 1270.41, found 1270.08.


4.6.5 Determination of binding constants to β-cyclodextrinimmobilized on silica gel

Modified silica gel (0.5 mL, settled) was measured into a 3 mL fritted column. This resinwas washed thoroughly with 10 mM pH 6.5 phosphate buffer and transferred into a 20 mLscintillation vial with an additional 0.5 mL of buffer, such that the vial contained 0.5 mL ofresin in 1 mL total volume. A solution of 550 azo 3.3 in 10 mM pH 6.5 phosphate bufferwas then titrated into this resin, and the absorbance of the supernatant was measured threetimes after every addition. A 1:1 binding model was then fit to the data. Because both thebinding constant and the resin loading are highly correlated variables, the resin loading wasfixed and only the binding constant was fit to the data.

4.6.6 Chromatography

Unless otherwise mentioned, all chromatography was performed using 30 mM pH 7.2 phos-phate buffer with 500 mM NaCl and 30 mM pH 7.2 phosphate buffer with 500 mM of NaCland 10 mM of β-cyclodextrin as eluent. A flow rate of 1 mL/min was used for all samples.


Expression of Mth1491 proteins was carried out as described in Section 3.7.6.


Modification of proteins was carried out as described in Section 3.7.7.

4.6.9 Protein digestion

Selected fractions were buffer exchanged into 100 mM ammonium bicarbonate buffer usinga 3,000 MWCO spin filter. After concentration to about 20 µL, the concentrations of thesesamples ranged from 1-20 µM. These samples were placed into a heating block kept at 100 ◦Cfor 10 min and then cooled to room temperature. To each sample was added 7 µL of 1 mg / 5mL Promega sequencing-grade trypsin, and the samples were incubated at 37 ◦C overnight.The resulting peptides were concentrated using C18 ZipTips according to the manufacturersdirections.

4.7 References




3. Allie C. Obermeyer, John B. Jarman, Chawita Netirojjanakul, Kareem El Muslemany,and Matthew B. Francis. “Mild bioconjugation through the oxidative coupling of ortho-aminophenols and anilines with ferricyanide.” Angew. Chem. Int. Ed. Engl. 53, 2013,1057–61.

4. Georges Guiochon. “Monolithic columns in high-performance liquid chromatography.”J. Chromatogr. A 1168, 2007, 101–168.

5. Oscar Nunez, Kazuki Nakanishi, and Nobuo Tanaka. “Preparation of monolithic silicacolumns for high-performance liquid chromatography.” J. Chromatogr. A 1191, 2008,231–52.

6. Christer Ericson, Jia Li Liao, Ken’ichi Nakazato, and Stellan Hjerten. “Preparation ofcontinuous beds for electrochromatography and reversed-phase liquid chromatographyof low-molecular-mass compounds.” J. Chromatogr. A 767, 1997, 33–41.

7. Stellan Hjerten. “Standard and capillary chromatography, including electrochromatog-raphy, on continuous polymer beds (monoliths), based on water-soluble monomers.”Ind. Eng. Chem. Res. 38, 1999, 1205–1214.

8. Eric C. Peters, Miroslav Petro, Frantisek Svec, and J. M. Frechet. “Molded rigid poly-mer monoliths as separation media for capillary electrochromatography. 2. Effect ofchromatographic conditions on the separation.” Anal. Chem. 70, 1998, 2296–2302.

9. Michael D. Slater, J. M. J. Frechet, and Frantisek Svec. “In-column preparation of abrush-type chiral stationary phase using click chemistry and a silica monolith.” J. Sep.Sci. 32, 2009, 21–28.

10. Michael Slater, Marian Snauko, Frantisek Svec, and Jean M. J. Frechet. “‘Click chem-istry’ in the preparation of porous polymer-based particulate stationary phases forµ-HPLC separation of peptides and proteins.” Anal. Chem. 78, 2006, 4969–4975.

11. Vladimir Smigol and Frantisek Svec. “Synthesis and properties of uniform beads basedon macroporous copolymer glycidyl methacrylate-ethylene dimethacrylate. A way toimprove separation media for HPLC.” J. Appl. Polym. Sci. 46, 1992, 1439–1448.

12. David Sykora, Frantisek Svec, and J. M. J. Frechet. “Separation of oligonucleotideson novel monolithic columns with ion-exchange functional surfaces.” J. Chromatogr. A852, 1999, 297–304.

13. Yongqin Lv, Zhixing Lin, and Frantisek Svec. “Thiol-ene click chemistry: a facile andversatile route for the functionalization of porous polymer monoliths.” Analyst 137,2012, 4114.


14. Yingjie Li, Chunhui Song, Lingyi Zhang, Weibing Zhang, and Honggang Fu. “Fabri-cation and evaluation of chiral monolithic column modified by β-cyclodextrin deriva-tives.” Talanta 80, 2010, 1378–1384.

15. Yongqin Lv, Timothy C. Hughes, Xiaojuan Hao, Danping Mei, and Tianwei Tan.“Preparation of monomeric and polymeric β-cyclodextrin functionalized monoliths forrapid isolation and purification of puerarin from Radix puerariae.” J. Sep. Sci. 2011,2131–2137.

16. Yongqin Lv, Danping Mei, Xinxin Pan, and Tianwei Tan. “Preparation of novel β-cyclodextrin functionalized monolith and its application in chiral separation.” J. Chro-matogr. B 878, 2010, 2461–2464.

17. Qiaoxuan Zhang, Jialiang Guo, Feng Wang, Jacques Crommen, and Zhengjin Jiang.“Preparation of a β-cyclodextrin functionalized monolith via a novel and simple one-pot approach and application to enantioseparations.” J. Chromatogr. A 1325, 2014,147–154.

18. D. Lubda, K. Cabrera, K. Nakanishi, and W. Lindner. “Monolithic silica columns withchemically bonded β-cyclodextrin as a stationary phase for enantiomer separations ofchiral pharmaceuticals.” Anal. Bioanal. Chem. 377, 2003, 892–901.

19. Dorothee Wistuba and Volker Schurig. “Enantiomer separation by capillary electrochro-matography.” Electrophoresis 21, 2000, 3152–3159.

20. Zhenbin Zhang, Minghuo Wu, Renan Wu, Jing Dong, Junjie Ou, and Hanfa Zou.“Preparation of perphenylcarbamoylated β-cyclodextrin-silica hybrid monolithic col-umn with ‘one-pot’ approach for enantioseparation by capillary liquid chromatogra-phy.” Anal. Chem. 83, 2011, 3616–3622.

21. Michael S. Lawrence, Kevin J. Phillips, and David R. Liu. “Supercharging proteins canimpart unusual resilience.” J. Am. Chem. Soc. 129, 2007, 10110–10112.

22. J. R. Jabusch and H. F. Deutsch. “Localization of lysines acetylated in ubiquitin reactedwith p-nitrophenyl acetate.” Arch. Biochem. Biophys. 238, 1985, 170–177.

23. Petr Novak, Gary H. Kruppa, Malin M. Young, and Joe Schoeniger. “A top-downmethod for the determination of residue-specific solvent accessibility in proteins.” J.Mass Spectrom. 39, 2004, 322–328.

24. I. Wayan Muderawan, Teng Teng Ong, Teck Chia Lee, David J. Young, Chi BunChing, and Siu Choon Ng. “A reliable synthesis of 2- and 6-amino-β-cyclodextrin andpermethylated-β-cyclodextrin.” Tetrahedron Lett. 46, 2005, 7905–7907.



0 10 20 30

1 2 3

0 5 10 15 20 25 30ml

1-11 16 0 5 48 8 0 52 63 12 1 13

1-11 7.5 0 5.9 48 5.4 0 7.2 63 5.7 7.6 5

Ion Counts:

Fraction:

Prob. Score:

Site

i. Ubiquitin modi�ed with azo NHS-ester 3.4 and puri�ed with monolith 4.3 + monolith 4.6

ii. Ubiquitin modi�ed with azo NHS-ester 3.7 and puri�ed with 2x propyl monolith 4.6

Figure 4.9: Identification of modification sites of ubiquitin. See text for details. (i) Ubiquitinmodified with azo NHS-ester 3.7. Black line segments correspond to detections of unmodifiedpeptides, and red line segments correspond to detections of modified peptides. The first fractionshowed modification at positions 1-11 (they could not be differentiated), and the last fragmentshowed modification at positions 33 or 48. (ii) Ubiquitin modified with azo NHS-ester 3.4. Ioncounts and probability scores for each fraction are listed below the chromatogram. The shapes ofthe major peaks differ from those in Figure 4.6 because the overall modification levels of ubiquitindiffer and because in this case two columns are used to purify the protein.


HN

O O

HN BCD

NH

O

HN

O O

HN BCD

NH

OHO

HN

O O

HN BCD

NH

OOH

HO

−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5Volume of Sample Cell (ml)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

Abs

orba

nce

at 3

40nm

log10(K a ) = 3.98 § 0.00R 2 = 0.997

−2.0−1.5−1.0−0.50.0 0.5 1.0Volume of Sample Cell (ml)

−0.005−0.004−0.003−0.002−0.001

0.0000.0010.0020.0030.004

Exp

- Cal

c

Residuals


0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

Abs

orba

nce

at 3

40nm

10( .7 0.000.


−0.003−0.002−0.001

0.0000.0010.0020.0030.004

Exp

- Cal

c

Residuals


0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

Abs

orba

nce

at 3

40nm

log10(K a ) = 3.70 § 0.00R 2 = 0.999


−0.005−0.004−0.003−0.002−0.001

0.0000.0010.0020.0030.004

Exp

- Cal

c

Residuals

log10(Ka / M) = 3.79

log10(Ka / M) = 3.70

log10(Ka / M) = 3.98

Figure 4.10: Determination of binding constants to silica gel modified with β-cyclodextrin as de-scribed in Section 4.6.5.


300 400 500 600 700 800Wavelength / nm

0.00

0.05

0.10

0.15

0.20

Abs

i. Absorbance

450 500 550 600 650 700Wavelength / nm

0

50000

100000

150000

200000

250000

Coun

ts /

abs

ii. Normalized �uorescence

before

after

Figure 4.11: Fluorescence of GFP before and after purification. Fluorescence was normalized byabsorbance at 495 nm.

89

Chapter 5

Affinity-based recovery of enzymes

Abstract

Affinity interactions with proteins are of use in many diverse areas. This chapter describesthe extension of the materials and methodology developed in the previous two chapters tothe recovery of industrial enzymes. The conversion of cellulose to fermentable sugars requireslarge amounts of enzyme that greatly increase the end cost of the conversion. Recovery ofthese enzymes could reduce costs, and one way to accomplish this goal is through affinityinteractions with β-cyclodextrin. This chapter explores potential ligands that could enablethis recovery.

Haiku

Sticky sugar ballsTo clean up your reactionDamn you lignin goop!

Portions of the Materials and Methods portion of this chapter were taken from the supportinginformation of a separate publication.1 These sections appear here in a modified form andare licensed under a CC BY 3.0 license.


CHAPTER 5. AFFINITY-BASED RECOVERY OF ENZYMES 90

5.1 Introduction

As the need to curb the emission of greenhouse gases becomes more pressing, attention hasfocused on alternative sources of energy. One alternative source is the generation of ethanolor other alcohols from cellulose. Cellulose is abundant in nature and often generated as awaste product; one report from the Department of Energy and the Department of Agricul-ture suggested that as many as 1.3 billion tons of cellulose could be generated every year.2

This supply could be enough to produce 130 billion gallons of ethanol—the energy equiva-lent of roughly 59% of annual US gas consumption—at 90% conversion efficiency.3 However,significant doubts exist concerning the energy efficiency of production and processing of cel-lulosic feedstock, with some suggesting that the process requires input of at least 50% of theoutput energy in the form of fertilizer, farming, and transportation.4 Moreover, the processof converting cellulose to ethanol is also inefficient and costly. Depolymerization of celluloseis most commonly performed using cellulases, enzymes which produce fermentable sugars.Such depolymerization reactions require unusually large amounts of cellulase, and the cost ofthe cellulases is the largest contributer to the cost of ethanol production.3 Estimates for thecost of cellulases vary and depend greatly on the design of the plant in which depolymeriza-tion is carried out, but cellulases likely account for 20-40% of the costs of ethanol productionby one estimate, or $0.50/gallon of ethanol ($0.84/equivalent of 1 gallon of gasoline).5,6

Given the high costs of cellulases, there has been substantial effort to reduce the overallamount of enzyme required by recycling enzymes after use. A number of strategies7–13 havebeen developed to allow the recycling of cellulases by attaching them to insoluble particles orsurfaces. These immobilized enzymes perform both better than and worse than free enzymedepending on the conditions of hydrolysis. Immobilized enzymes tend to exhibit better sta-bility and therefore stay active longer, but at the cost of limited diffusional mobility. As aresult the enzymes cannot interact as much with insoluble cellulose fragments. Efforts fromour own lab have sought to overcome these diffusional issues by attaching an endoglucanasefrom Pyrococcus horikoshii (EGPh) to soluble thermo-responsive polymers that can be pre-cipitated to isolate the enzyme after a reaction.14 While this strategy proved successful andallowed recycling of the enzyme, subsequent work showed that addition of N -isopropyl acry-lamide (NIPAm) polymer without attachment of the enzyme was capable of reducing theamount of enzyme required for hydrolysis by 60%.15 While the mechanisms by which thesepolymers act are not entirely clear, there is evidence to suggest that they stabilize cellulases,prevent nonproductive binding of cellulases to lignin, and disrupt cellulose structure.

While the use of NIPAm as a polymer additive substantially improves the performanceof EGPh, it does not achieve our original goal of developing a method for the recyclingof cellulases. NIPAm could be used as both an additive and also as a vehicle for enzymerecovery, but an ideal recovery method would allow isolation of the protein from the polymeras well. Cellulases and NIPAm polymers have different useful lifespans, and therefore theability to replace them at separate times could further reduce costs. Thus, within thecontext of the work on polymer additives by Mackenzie and Francis,15 we were interested indeveloping a method to enable recycling of the enzyme without its covalent attachment to


N

Recycle

Recovery fo EGPh from heterogeneous reaction mixtures

i. Crude reaction mixture ii. A�nity-capture of proteins iii. Isolation of beads

iv. Removal of subtrate and productsv. Release of protein

Figure 5.1: Scheme for recovery of EGPh from crude reaction mixtures. After a reaction is complete,magnetic beads are added to isolate protein (ii). The beads are separated from other solids in thereaction mixture with a magnetic field (iii). After isolation (iv), the protein is liberated (v) for usein another reaction.

either a polymer or a surface. Such a strategy could take advantage of the beneficial effectsof the NIPAm polymer additive while still allowing full diffusional freedom of the proteinand separate isolation of the protein and the polymer.

5.2 Design of system

Recovery of EGPh from a crude reaction mixture could be accomplished using an affinity-based interaction, wherein the protein is modified with a small molecule handle that min-imally affects the properties of the protein. Such a recovery scheme would proceed as il-lustrated in Figure 5.1. EGPh would be modified with a small molecule handle so as tominimally affect its diffusional freedom (Figure 5.1.i). After a reaction is complete, addition


of beads bearing a binding partner of the affinity group on EGPh would remove the enzymefrom solution (Figure 5.1.ii). Because cellulases work on heterogeneous mixtures, isolationof the beads from remaining cellulase could be difficult, so the beads could be magnetizedto enable their separation from other insoluble components of the reaction mixture (Fig-ure 5.1.iii). The enzyme-bead complexes could then be isolated and the enzyme liberatedfor future use (Figure 5.1.iv-v)

A natural choice for the basis of such an affinity interaction is β-cyclodextrin. Asdiscussed in Chapter 1 and Chapter 3, β-cyclodextrin has many properties that make itamenable to use with proteins, and it is also produced on scales that enable its use in indus-trial contexts. Moreover, the resins developed in Chapter 3 provide a logical starting pointfor the development of this method, and Sepharose can be magnetized simply by mixingit with ferrofluid.16 As a result, any Sepharose-based resin can be adapted for use in thiscontext.

The fundamental limitation for this pulldown-based recovery is the binding constantbetween the protein and the resin. The value of this binding constant dictates how muchprotein is recovered at each step and how much resin must be used relative to the volumeof crude reaction mixture. Ideally, the binding constant should be high enough to allowrecovery of a substantial amount of protein after several successive rounds of capture usinga relatively small amount of resin.

To evaluate these constraints, a binding model was constructed to simulate three con-secutive recoveries performed on protein in crude reaction mixtures. The amount of proteinrecovered was assumed to be equal to the amount of protein that theoretically bound eachtime to the resin. Both the protein concentration and the ratio of resin volume to reac-tion volume were iterated over to construct heatmaps indicating the maximum theoreticalrecovery. This simulation was repeated for several binding constants to determine the mini-mum required binding constant for this system (Figure 5.2). The weakest binding constant,104.5 M−1, is sufficient for modest recovery (80-90%) at relatively low ratios of crude reactionmixture to resin (Figure 5.2.i). At binding constants at or above 105.5 M−1, the requirementfor low crude reaction mixture ratios is relaxed, and the pulldown results in high recoveriesof protein in excess of 90% (Figure 5.2.iii-iv). Based on these results, the absolute minimumbinding constant was deemed to be 104.5 M−1, with higher binding constants preferable.

5.3 Results from low-affinity ligands

One of the most commonly used ligands of β-cyclodextrin is 1-adamantane carboxylic acid,which has a relatively high affinity constant of about 104.5 M−1 and can be incorporated intomany different contexts via amide chemistry. While the binding constant of this ligand is atthe lower end of the acceptable range, 1-adamantane carboxylic acid was used to evaluate thefeasibility of this method because of its general availability and synthetic tractability. Twoadamantane-based modification reagents were prepared (adamantane aminophenol 3.14 andadamantane alkoxyamine 5.4), and their ability to singly modify the N-terminus of EGPh


0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.00.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.00.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

Reaction volume : resin volumeReaction volume : resin volume

Reaction volume : resin volumeReaction volume : resin volume

−6.0

−5.5

−5.0

−4.5

−4.0

−3.5

−3.0

log 10

([Pro

tein

] M)

−6.0

−5.5

−5.0

−4.5

−4.0

−3.5

−3.0

log 10

([Pro

tein

] M)

−6.0

−5.5

−5.0

−4.5

−4.0

−3.5

−3.0

log 10

([Pro

tein

] M)

−6.0

−5.5

−5.0

−4.5

−4.0

−3.5

−3.0

log 10

([Pro

tein

] M)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0i. log10(Ka / M-1) = 4.5 ii. log10(Ka / M-1) = 5.0

iv. log10(Ka / M-1) = 6.0iii. log10(Ka / M-1) = 5.5

Figure 5.2: Heatmaps showing the efficacy of pulldowns for different binding constants. Eachheatmap shows the percent recovery after three reactions as a function of the number of equivalentsof supernatant and the protein concentration. The number of equivalents of supernatant is thevolumetric ratio of the volume of the solution before adding the beads to the volume of the swollenbeads added. The section of the heatmaps that is always red results from the stoichiometric limitwhen the beads are saturated. The loading of β-cyclodextrin used in these calculations is the sameas was obtained in Chapter 3.

was evaluated (Figure 5.3.i-ii). Both reagents were capable of near quantitative conversionof the unmodified protein to modified protein, with a moderate amount of over-modification(Figure 5.3). In terms of efficacy, both modification routes produced modified protein withthe same binding characteristics, but modification with 3.14 was significantly faster andhigher yielding because it could be accomplished in one step instead of two. However, EGPhmodified with 5.4 was used for most of these studies because the chemistry to modify with3.14 was not known at the time.

Recovery of modified EGPh was simulated to determine the recovery efficiency of thismethod. A solution of 10 µM EGPh was added to the resin used for most of Chapter 3. Thisresin was then washed with several portions of buffer, and EGPh was eluted by applying10 mM β-cyclodextrin to the resin. Early optimization of this process indicated that EGPh


49224ii. Modification of P-EGPh with 3.13

iii. Recovery of modified EGPh modified with 5.4 with Sepharose-based resin

i. Modification of AKT-EGPh with 5.4

47K 48K 49K 50K 51K 52K

49427

Reconstructed Mass47K 48K 49K 50K 51K 52K

Reconstructed Mass

5.4 3.13

FT W1 W2 W3 W4 W5 E1 E2 E3 E4 E5

0.0

0.2

0.4

0.6

0.8

1.0

Nor

med

Abs

orba

nce

at 2

80 n

m

Elution of each fraction

Calc. elution of each fractionCalc. cumulative elution

Cumulative elution

HN

OOH

NH2

Fraction

NH

OO NH2

Figure 5.3: Pulldown results for adamantyl EGPh. (i-ii) Modification of EGPh with adamantylgroups. Both an alkoxyamine derivative and an aminophenol derivative were able to completelymodify the protein. (iii) High-salt pulldown of modified EGPh. To a solution of 10 µM proteinin buffer was added Sepharose beads modified with β-cyclodextrin. After mixing, the absorbanceof the supernatant was measured (FT), and the resin was washed five times (W1-5). The proteinwas then eluted with five fractions of buffer containing 10 mM β-cyclodextrin (E1-5). Blue linesshow the amount of protein that eluted in each fraction, and grey lines indicate the total amountof protein that had eluted. Dashed lines indicate the theoretical results assuming no nonspecificbinding to the resin.

exhibited some nonspecific affinity for the resin (see Figure 5.6). As a result, NaCl wasadded to all buffers to disrupt nonspecific interactions (Figure 5.3). Comparison of the resultsindicate that addition of the salt results in a 30% increase in the amount of protein recovered,with much of the increase in recovery occurring during the elution phase. The amount ofprotein recovered during the elution steps was about 60% of the protein applied to the resin.When these results are compared to their theoretical counterparts (see Figure 5.3, dashedlines), two issues emerge. First, elution of the protein from the resin is slower than if the onlyinteraction were between the adamantyl handle and the β-cyclodextrin. This artifact likelyresults from nonspecific interactions of the protein with the resin, and is not particularlysurprising given that the natural substrate of EGPh is cellulose. Both Sepharose and β-


cyclodextrin are composed of glucose monomers, and therefore could interact favorably withEGPh. Second, the binding between adamantane and β-cyclodextrin is not strong enoughto prevent a substantial portion (around 20% in this case) of the protein being lost in thewashes. Ideally, this binding could be improved to increase recovery of the enzyme.

5.4 High-affinity ligands

5.4.1 High-affinity ligands of modified β-cyclodextrin

As noted in Section 1.5.1, one of the appealing aspects of cyclodextrins is the ease of theirchemical modification. This feature has enabled their use in a wide range of applications,and has resulted in the synthesis of a myriad of cyclodextrin derivatives. Within the contextof this work, heptakis-[2-aminoethylsulfanyl]-β-cyclodextrin (cysteamine-β-cyclodextrin 5.2)was synthesized as a template for glycoclusters,17 as a host for camptothecin,18, and to studythe role of ionic interactions in molecular recognition by β-cyclodextrins.19 During the courseof their work, Wenz et al. noted that cysteamine-β-cyclodextrin 5.2 exhibited unusuallylarge binding constants to negatively charged derivatives of t-butyl benzene. Gomez-Biagi,Jagt, and Nitz realized that this cyclodextrin could serve as a host for other high-affinityguests, so they further explored the complexation thermodynamics of this cyclodextrin.20

They found that three features contribute to the binding thermodynamics of cysteamine-β-cyclodextrin 5.2—charge-charge interactions, a lengthening of the hydrophobic pocket, andat low pH a conformational change brought about by charge-charge repulsion between theammonium groups. These features increased the binding of 1-adamantane carboxylic acidto 105.64 M−1 at about neutral pH, and several long hydrophobic molecules with negativelycharged moieties on one end were found to bind with affinity constants in excess of 106 M−1.The strongest binding observed was that of lithocholic acid, whose binding constant was107.74 M−1. These binding constants are well beyond the normal range of binding constantsfor β-cyclodextrin and would be sufficient to enable affinity-based recovery of EGPh.

5.4.1.1 Results of pulldowns with cysteamine-β-cyclodextrin 5.2

The ability of cysteamine-β-cyclodextrin 5.2 to improve the recovery of cellulases was in-vestigated using a number of resins and ligands. Because cysteamine-β-cyclodextrin 5.2bears seven primary amino groups, its conjugation to resins is relatively simple and can beaccomplished in one or two steps using either amide-bond forming reactions (resin 5.8) orreductive amination (resin 5.9). While these resins are easy to make, their composition maybe heterogeneous given the potential for multiple amino groups to react with the resin. Asa result, several resins with more well-defined composition were also synthesized, but theirperformance was similar, and therefore they will not be discussed here.

Early work using coumarin derivatives as ligands proved difficult because of syntheticintractability, and these ligands were abandoned. Due to the increased binding affinityof 1-adamantane carboxylic acid for cysteamine-β-cyclodextrin 5.2 over β-cyclodextrin by


240 260 280 300 320 340 360Wavelength / nm

−0.4

−0.3

−0.2

−0.1

0.0

0.1

0.2

1 cm

Abs

orba

nce

ii. In-solution binding of 5.5 to cyclodextrin 5.2 iii. Binding of 5.5 to resin 5.7

O

OHNaO3S

5.5

−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5log10(Volume of Sample Cell / mL)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

Abs

orba

nce

at 3

40nm

Residuals

BrNaO3S

O

OH

log10(Ka / M-1) = 6.6 +/- 0.25

log10(Ka / M-1) = 4.19 +/- 0.25[CD] = 15.96 +/- 4.93 mM

R2=0.984

+Pd(OAc)2, Et3N

tri(o-tolyl)phospine

i.

Figure 5.4: Synthesis and binding of stilbene 5.5. (i) Synthesis of stilbene 5.5 from commericallyavailable materials. (ii) Binding of stilbene 5.5 to cysteamine-β-cyclodextrin 5.2. A simultaneousfit of the entire spectrum between 240 and 365 nm gave a binding constant of 106.6 M−1. (iii)Binding of stilbene 5.5 to resin 5.9. Binding of the stilbene to the resin was significantly weakerthan its binding in solution.

about an order of magnitude, the potential of adamantyl groups to serve as ligands for thismodified cyclodextrin was also investigated. The binding of EGPh modified with adamantylo-aminophenol 3.14 binding to resin 5.8 was evaluated in two ways: a pulldown and atitration (Figure 5.7). Both the pulldown and the titration indicated that this bioconjugatedid not bind to the resin particularly strongly, and fitting a model to the data gave abinding constant of 103.4 M−1. This value was far too low to enable an effective pulldown,and adamantyl ligands were not investigated further.

Based on these results, stilbene 5.5 was identified as a promising ligand because of itsrelatively long hydrophobic core, its ability to be modified in the same ways as azo 3.3 andazo 3.6, and the presence of a permanent negative charge on one end. This molecule wasaccessible from commercial starting materials in one step, and it could be purified throughrecrystallization (Figure 5.4.i). After its synthesis, the binding constant of the complexbetween this ligand and cysteamine-β-cyclodextrin 5.2 was measured to be 106.6 M−1, avalue more than sufficient to enable recovery of EGPh (Figure 5.4.ii).

With this high-affinity ligand in-hand, the binding of the ligand alone to resin 5.9 wasevaluated by performing a titration of the compound into a solution containing the resin. Thebinding constant was then calculated by fitting a theoretical model to the results. Despitethe high in-solution binding constant, the binding constant to the cyclodextrin on the resinwas weaker by more than two orders of magnitude (Figure 5.4.iii). This result held for other


resins with more well-defined linkers between the resin and the cyclodextrin, and is consistentwith the results seen earlier with adamantyl ligands. Regardless of the chemistry used tolink cysteamine-β-cyclodextrin 5.2 to Sepharose, it seems that the binding of its ligandsis reduced by about two orders of magnitude from the calculated or reported in-solutionbinding constants. The exact cause of this phenomenon was not extensively investigated.However, as noted earlier, the conformation of cysteamine-β-cyclodextrin 5.2 seems to playa role in increasing the binding constants of its ligands. It is possible that attachment to asurface disrupts this conformation and thereby weakens binding.

5.4.2 High-affinity cholesterol-based ligands of nativeβ-cyclodextrin

At least one high-affinity ligand of β-cyclodextrin has been known since the 1990s. Litho-cholic acid was shown to bind β-cyclodextrin with an affinity constant greater than 106 M−1

in buffer with 1% DMSO, and this ligand has been used to dimerize proteins at concentra-tions as low as 1 µM.21–23 Despite this large binding constant, lithocholic acid is difficult towork with because it is insoluble in water at concentrations above 1 µM.24 Initial attempts atusing this ligand focused on the addition of permanently charged groups to it to improve itswater solubility, although these synthetic modifications proved time-consuming and difficultand were eventually abandoned.

5.4.2.1 Results of pulldowns using lithocholic acid

Fortunately, β-cyclodextrin can serve to solubilize lithocholic acid at high µM concentrations,such that modification of proteins in aqueous solutions is possible. In collaboration with Dr.Meera Rao and Dr. Christian Rosen in the Francis lab, lithocholic acid derivative 5.6 wasprepared as a reagent to modify proteins via oxidative coupling. Optimization of the reactionconditions indicated that it was possible to modify P-EGPh to about 73% conversion, asshown in (Figure 5.5.i). EGPh proved difficult to work with for reasons discussed previouslyin this chapter: the protein exhibited a large degree of nonspecific affinity for the resin. Thistendency likely results from the chemical composition of the modified resin, and ameliorationof this problem may require the investigation of different resins and hosts. Work on this areais ongoing in the lab and will not be reported here.

However, we did use RNAse A as a model protein to test the ability of lithocholic acidderivative 5.6 to recover proteins that do not exhibit nonspecific sticking. RNAse A wasmodified to 54%, and its recovery from a resin was simulated (Figure 5.5.ii). The elutionprofile exhibits two characteristic peaks corresponding to protein that flowed through theresin and protein that was retained. The ratio of retained to total protein is 51%, indicatingthat almost all of the modified protein bound to the column and was subsequently released.Notably, the cumulative sum of the elution profile exceeds 100% of the protein. This featureis commonly seen with these pulldown experiments and likely results from undeterminederrors in the UV/Visible measurements of protein concentration.


48000 48500 49000 49500 50000 50500 51000

27 7310 eq.

FT W1 W2 W3 E1 E2 E3 E4Fraction

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Perc

ent o

f pro

tein

elu

ting

i. Modi�cation of EGPh with lithocholic acid

ii. Pulldown of RNAse A modi�ed with lithocholic acid

Figure 5.5: Modification of EGPh with lithocholic acid derivative 5.7 and pulldown of RNAse Amodified with lithocholic acid derivative 5.7. (i) Modification of EGPh results in 73% conversionof the protein. Integrals of the mass spectrum are shown above the trace. (ii) Pulldown of RNAseA modified to 54% conversion results in recovery of 51% of the protein during the elution phase.The gray line represents a cumulative sum of the protein eluting from the resin, whereas the blueline represents the protein eluting in each fraction.

5.5 Conclusions

This chapter discusses the the development of an affinity-based system for the recovery ofenzymes from industrial reactions. Recovery of enzymes like EGPh could improve efficiencyof the production of biofuels. β-cyclodextrin and its ligands were investigated for potentialuse, and it was determined that the majority of ligands of β-cyclodextrin do not bind stronglyenough to enable recovery of cellulases. However, lithocholic acid did show promise as a high-affinity ligand of β-cyclodextrin that could enable the recovery of enzymes at concentrationsof 1 µM. While this ligand-host pair ultimately did not exhibit ideal characteristics withthe recovery of EGPh, it has shown promise with other enzymes. The application of thistechnique to other industrially relevant enzymes is ongoing.



Unless otherwise noted, the chemicals and solvents used were of analytical grade and wereused as received from commercial sources. Purifications by flash chromatography were per-formed using EM silica gel 60 (230-400 mesh). Chromatography solvents were used with-out distillation. All organic solvents were removed under reduced pressure using a rotaryevaporator. Water (dd-H2O) used as a reaction solvent was deionized using a BarnsteadNANOpure purification system. Centrifugations were performed with an Eppendorf MiniSpin Plus (Eppendorf, Hauppauge, NY).


NMR. 1H and 13C spectra were recorded with a Bruker AVB-400 (400 MHz, 100 MHz) or aBruker AV-600 (600 MHz, 150 MHz). 1H NMR chemical shifts are reported as δ in units ofparts per million (ppm) relative to residual CH3OH (δ, 3.31, pentet), DMF (δ 8.03, singlet),or CHCl3 (δ, 7.24, singlet). Multiplicities are reported as follows: s (singlet), d (doublet),t (triplet), dd (doublet of doublets) or m (multiplet). Coupling constants are reported as aJ value in Hertz (Hz). The number of protons (n) for a given resonance is indicated as nHand is based on spectral integration values. 13C NMR chemical shifts are reported as δ inunits of parts per million (ppm) relative to DMF−d7 (δ 163.15, triplet), MeOH−d4 (δ 49.15,septet), or CDCl3 (δ, 77.23, triplet).Mass Spectrometry. Matrix-assisted laser desorption/ionization time-of-flight mass spec-trometry (MALDI-TOF MS) was performed on a Voyager-DE system (PerSeptive Biosys-tems, USA). Small molecule samples were co-crystallized with α-cyano-4-hydroxycinnamicacid in 1:1 acetonitrile (MeCN) to H2O with 0.1% trifluoroacetic acid (TFA). Cyclodextrinswere co-crystallized with 2,5-dihydroxybenzoic acid in 1:1 MeCN to H2O with 0.1% TFA.Protein bioconjugates were analyzed using an Agilent 1200 series liquid chromatograph (Ag-ilent Technologies, USA) that was connected in-line with an Agilent 6224 Time-of-Flight(TOF) LC/MS system equipped with an electrospray ion source. Extracted mass spectrawere plotted using chartograph.com/ms.High Performance Liquid Chromatography. HPLC was performed on Agilent 1100 Se-ries HPLC Systems (Agilent Technologies, USA). Sample analysis for all HPLC experimentswas achieved with an in-line diode array detector (DAD) and in-line fluorescence detector(FLD). Analytical reverse-phase HPLC of small molecules was accomplished using a C18stationary phase and a H2O / MeCN with 0.1% TFA gradient mobile phase.Fast protein liquid chromatography. FPLC was performed on an Akta Pure M (GEHealthcare, USA) at 8 ◦C. Sample analysis was performed with an in-line UV monitor andan in-line conductance monitor. All injections were performed manually with a 1 mL syringefit to the top of the column.UV-Visible Spectrometry. UV-visible spectrometry was performed using quartz cuvetteswith a Varian Cary 50 spectrophotometer (Agilent, USA). Small-scale UV-visible spectrom-etry was performed using a Nanodrop 1000 (Thermo Scientific, USA). Absorbance mea-

chartograph.com/ms


surements of samples in plates were obtained with a SpectraMaxM2 (Molecular Devices,Sunnyvale, CA).


O

OHHO

Br

O

OOH

HOBrO

OOH

OH

Br

O

OOH

OH

Br

OO

OH

OH

BrO

OOH

OHBr

O

OOH

HO

Br

O

Synthesis of Per-6-[(2-aminoalkyl)amino]-6-deoxy-β-cyclodextrin (5.1). Preparedon a 3.4 g scale as described by Vizitiu et al.25 Resulted in a yield of about 52%.

O

OHHO

S

O

OOH

HOSO

OOH

OH

S

O

OOH

OH

S

OO

OH

OH

SO

OOH

OHS

O

OOH

HO

S

O

NH2

NH2

NH2

H2N

H2N

H2N

NH2

Synthesis of Per-6-[(2-aminoalkyl)amino]-6-deoxy-β-cyclodextrin (5.2). Preparedon a 1 g scale as described by Gomez-Garcıa et al.17 Resulted in a yield of about 40%.

NH

OO N

HO

O Synthesis of adamantane Boc-alkoxyamine (5.3). Boc hydroxyl-amine acetic acid (0.5 g, 2.62 mmol, 1 eq.) and N -hydroxysuccinimide

(0.36 g, 3.14 mmol, 1.2 eq.) were dissolved in 16 mL of DCM in a 50 mL round bottom flask.After sonication to dissolve most of the solids, 0.648 g of dicyclohexylcarbodiimide (3.14mmol, 1.2 eq.) was added. After approximately 5 min, a precipitate began to form, andthe reaction mixture was cooled to 0 ◦C with an ice bath to induce further precipitation.The crude reaction mixture was filtered through Celite. To the filtrate was added 0.395 g(2.62 mmol, 1 eq.) of 1-adamantylamine, and the mixture was stirred overnight at roomtemperature. After removal of the solvent under reduced pressure, the crude product waspurified via silica-gel chromatograph with a gradient of 4:1 hexanes:ethyl acetate to 1:9hexanes:ethyl acetate. The product eluted from the column as the first major peak, andremoval of the solvent under reduced pressure yielded 600 mg of material (71%). 1H NMR(400 MHz, CDCl3) δ 8.25 (d, J = 2.2 Hz, 1H), 7.51 (s, 1H), 7.46 (s, 1H), 4.16 (s, 2H),2.02-2.06 (m, 9H), 1.66 (m, 6H), 1.47 (s, 9H).


NH

OO NH2

Synthesis of adamantane alkoxyamine (5.4). To a mixture of 1 mLof trifluoroacetic acid in 1 mL of DCM was added 600 mg of adamantane

Boc-alkoxyamine 5.3. The mixture was stirred for 30 min at room temperature, and thesolvent was removed under reduced pressure. To the remaining oil was added 1 mL oftrifluoroacetic acid, which was subsequently removed under reduced pressure. This processwas repeated three times. The material was then dissolved in 2 mL of methanol, and thesolvent was removed under reduced pressure. Finally, the material was sonicated in 2 mLof water, and the solvent was removed under reduced pressure. The resulting white powderwas used without further characterization.

O

OHNaO3S

Synthesis of stilbene (5.5). This synthesis is based on a modifiedprocedure from Yamamoto et al.26 A 100 mL round bottom flask with

24 mL of DMF under N2 was degassed by twice freezing with dry ice, applying a vacuum, andthawing. To this flask was added 1 g (4.85 mmol, 1.1 eq.) of sodium 4-vinylbenzenesulfonate,1 g (4.36 mmol, 1 eq.) of 3-(4-bromophenyl)propionic acid, and 1.1 g (10.9 mmol, 2.5 eq.,1.5 mL) of triethylamine. This mixture was degassed again, and palladium(II) acetate (24.5mg, 0.11 mmol, 0.025 eq.) and tri-o-tolylphosphine (66.4 mg, 0.22 mmol, 0.05 eq.) wereadded. This mixture was stirred at 120 ◦C for 16 h. After the crude product mixture wascooled to room temperature, a blackish precipitate was filtered off using a Buchener funnel.This precipitate was recrystallized twice from 3:1 ethanol:water to yield an off-white powder.1H (600 MHz, Methanol-d4) δ 7.79 (d, J = 8.4 Hz, 2H), 7.61 (d, J = 8.4 Hz, 2H), 7.49 (d,J = 8.2 Hz, 2H), 7.27 7.21 (m, 3H), 7.15 (d, J = 16.4 Hz, 1H), 2.91 (t, J = 8.06, 1H), 2.53(t, J = 8.6, 7.2 Hz, 1H). MALDI-TOF calculated for C17H15O5S– ([M-Na]−) 331.06, found331.277.

HOO

HN

H

H HOH

NO2 Synthesis of lithocholic o-nitrophenol (5.6). This com-pound was prepared by Dr. Meera Rao. To 4 mL of DMF

was added 250 mg (1 eq., 0.664 mmol) of lithocholic acid, 133 mg (1.1 eq., 0.730 mmol)of nitrotyramine 3.12, 373 mg (3 eq., 1.95 mmol) of EDC, 78 mg (0.7 eq., 0.462 mmol) ofHOBt, and 0.3 mL (3.2 eq., 2.15 mmol) of triethylamine. The reaction mixture was stirred atroom temperature overnight, and isolation of the product via flash chromatography yielded93 mg (50%) of yellow powder. 1H (400 MHz, CDCl3) All peaks did not completely resolve,and rotamers were observed. δ 10.52 (s, 1H), 7.94 (dd, J = 8.4, 2.1 Hz, 2H), 7.50 (ddd, J= 16.5, 8.5, 2.2 Hz, 2H), 7.17 (dd, J = 18.9, 8.4 Hz, 2H), 5.67-5.6 rotamers (t, J = 6.0 Hz,1H), 3.65 (dq, J = 10.8, 5.4, 4.8 Hz, 3H), 3.53 (dt, J = 13.4, 6.9 Hz, 4H), 2.97 2.91 (m,1H), 2.85 (t, J = 7.1 Hz, 2H), 2.72 (ddd, J = 15.1, 10.1, 4.7 Hz, 1H), 2.58 (ddd, J = 15.8,9.4, 6.4 Hz, 1H), 2.23 (tt, J = 10.1, 4.9 Hz, 2H), 2.13 0.82 (m, 73H), 0.70 (s, 2H), 0.66 (s,5H).

HOO

HN

H

H HOH

NH2 Synthesis of lithocholic o-aminophenol (5.7). To a solu-tion of 7.9 mg of sodium dithionite in 1 mL of water and 0.5mL of methanol was added 3.76 µL of 16.6 mM lithocholic acid


5.6 in DMSO. This solution was stirred for 10 min, at which point excess sodium dithionitewas removed using a C18 Sep-Pak according to the manufacturer’s instructions. The prod-uct was first eluted with 0.5 mL of methanol followed by 0.5 mL of THF. The solvent wasremoved from the eluent under reduced pressure, and the product was dissolved in 10 mMpH 8 phosphate buffer containing 10 mM β-cyclodextrin. The product was used immediatelywithout further characterization.

5.6.3 Resin synthesis

NH

O

O

OHHO

S

O

OH

HOSO

OOH

OH

S

O

OOH

OH

S

OO

OH

OH

SO

OOH

OHS

O

OOH

HO

S

O

NH2

NH2

NH2

H2N

H2N

NH2

Synthesis of resin 5.8. ECH Sepharose 4B (2 mL) was thoroughly washed with water anddrained. To the drained resin was added 100 mg (0.064 mmol) of cysteamine-β-cyclodextrin5.2 and 60 mg (0.3129) of 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide in 3 mL of waterthat had been pH adjusted to 4.5 using HCl. After mixing, the pH was again checked andadjusted to 4.5 if necessary. The reaction was rotated end-over-end on a laboratory rotisseriefor 1 hr, and the pH was adjusted to 4.5 using NaOH. After 18 h of rotation, the resin waswashed thoroughly with water.

O

OHHO

S

O

OH

HOSO

OOH

OH

S

O

OOH

OH

S

OO

OH

OH

SO

OOH

OHS

O

OOH

HO

S

O

NH2

NH2

NH2

HN

H2N

H2N

NH2

Synthesis of resin 5.9. Sepharose CL-4B (2 mL) was thoroughly washed with water anddrained. A 10 mL solution of 50 mM sodium periodate was prepared, and 3 mL was added tothe swollen Sepharose. The mixture was turned end-over-end for 30 min on a lab rotisserie,followed by addition of 10 µL of ethylene glycol to quench unreacted sodium periodate.The Sepharose was washed thoroughly with several portions of water and drained. A 3 mLsolution of 20 mM cysteamine-β-cyclodextrin 5.2 and 50 mM NaCNBH3 in 25 mM pH 7.2


potassium phosphate buffer with 150 mM NaCl was prepared, and the pH was adjustedback to 7.2 with 1 N HCl. This solution was added to the resin, and the mixture was turnedfor three days on a lab rotisserie. The resin was then washed thoroughly with water anddrained.

5.6.4 Pulldowns of EGPh

To a fritted 2 mL tube was added 0.5 mL of resin. This resin was washed thoroughly withbuffer. A 0.5 mL solution of 10 µM protein was prepared and added to the column. Thecontents of the column were agitated to create a slurry of the resin. After the resin hadsettled, the excess supernatant was either removed with a pipette or by allowing the liquidto flow through the resin bed until the column had drained. The column was washed withseveral (2-6) 0.5 mL portions of buffer using the method described above. The protein wasthen eluted using several (2-6) 0.5 mL portions of buffer containing 10 mM of β-cyclodextrin.Each fraction was collected and the protein concentration later determined using UV/Visiblespectrometry.

5.6.5 Determination of in-solution binding constants

A 15 µM solution of stilbene 5.5 was prepared in 25 mM pH 8.0 potassium phosphate buffer.A 0.5 mL portion of this was placed in a 1.5 mL quartz cuvette, and a solution of 15 µMstilbene with 0.5 mM β-cyclodextrin in 25 mM pH 8.0 potassium phosphate buffer was addedin portions. After each addition, the solution was mixed and the absorbance spectrum wasmeasured. A binding constant was determined by a simultaneous least-squares fitting.


Transamination of EGPh. This procedure was adapted from Mackenzie and Francis.14

To a solution of 10-100 µM AKT-EGPh in 10 mM pH 4.5 sodium acetate buffer was addedan equal volume of a 200 µM solution of pyridoxal-5’-phosphate (PLP). This solution wasincubated at 37 ◦C for 1 h. Excess PLP was removed by repeated centrifugal filtration witha 30 kDa molecular weight cutoff (MWCO) membrane and 50 mM pH 5.5 phosphate buffer.General procedure for oxime formation. Transaminated AKT-EGPh in 50 mM pH 5.5phosphate buffer was added to an equal volume of 100 mM 5.4 in 50 mM pH 5.5 phosphatebuffer. After mixing, the reaction was rotated end-over-end on a laboratory rotisserie atroom temperature for 3 d. Excess alkoxyamine was removed by repeated centrifugal filtrationagainst a 10 kDa MWCO membrane.Oxidative coupling of adamantane o-aminophenol 3.14 to P-EGPh. A 0.5 mLsolution of 200 µM EGPh was prepared in 25 mM pH 8.0 potassium phosphate buffer. Tothis solution was added 0.5 mL of 10 mM potassium ferricyanide (50 eq. relative to theprotein) and 4 mL of 87.5 µM adamantane o-aminophenol (3.5 eq. relative to the protein)


in the same buffer. The reaction stood at room temperature for 5 min, at which point excessreagent was removed by repeated centrifugal filtration against a 30 kDa MWCO membrane.Oxidative coupling of 5.7 to P-EGPh. To a 50 µL portion of a 50 µM P-EGPh in25 mM pH 8 phosphate buffer with 10 mM β-cyclodextrin was added 19 µL of 10 mMpotassium ferricyanide (75 eq. relative to protein). To this solution was added 181 µL of 104µM lithocholic o-aminophenol 5.7 (7.5 eq. relative to the protein) in 25 mM pH 8 phosphatebuffer with 10 mM β-cyclodextrin. After 10 min, excess reagent was removed by repeatedcentrifugal filtration against a 30 kDa MWCO membrane.


EGPh was expressed and purified as described by Mackenzie and Francis.14

5.7 References


2. Robert D. Perlack, Lynn L. Wright, Anthony F. Turhollow, Robin L. Graham, Bryce J.Stokes, and Donald C. Erback. Biomass as feedstock for a bioenergy and bioproductsindustry: the technical feasibility of a billion-ton annual supply. Tech. rep. Oak Ridge,Tennessee: Oak Ridge Natl. Lab, US DOE, USDA, 2005.

3. Andrew Carroll and C. R. Somerville. “Cellulosic biofuels.” Annu. Rev. Plant Biol. 60,2009, 165–82.

4. Hartmut Michel. “Editorial: the nonsense of biofuels.” Angew. Chem. Int. Ed. Engl.51, 2012, 2516–8.

5. Manuel B. Sainz. “Commercial cellulosic ethanol: The role of plant-expressed enzymes.”Vitr. Cell. Dev. Biol. - Plant 45, 2009, 314–329.

6. David B. Wilson. “Cellulases and biofuels.” Curr. Opin. Biotechnol. 20, 2009, 295–299.

7. Jason Jordan, Challa S. S. R. Kumar, and Chandra Theegala. “Preparation and char-acterization of cellulase-bound magnetite nanoparticles.” J. Mol. Catal. B Enzym. 68,2011, 139–146.

8. Wenjuan Liang and Xuejun Cao. “Preparation of a pH-sensitive polyacrylate am-phiphilic copolymer and its application in cellulase immobilization.” Bioresour. Tech-nol. 116, 2012, 140–146.


9. Ahmet Ince, Gulay Bayramoglu, Bunyamin Karagoz, Begum Altintas, Niyazi Bicak,and Mehmet Yakup Arica. “A method for fabrication of polyaniline coated polymermicrospheres and its application for cellulase immobilization.” Chem. Eng. J. 189-190,2012, 404–412.

10. Renee Han-Yi Chang, Jen Jang, and Kevin C.-W. Wu. “Cellulase immobilized meso-porous silica nanocatalysts for efficient cellulose-to-glucose conversion.” Green Chem.13, 2011, 2844.

11. Dian Andriani, Changshin Sunwoo, Hwa Won Ryu, Bambang Prasetya, and Don HeePark. “Immobilization of cellulase from newly isolated strain Bacillus subtilis TD6using calcium alginate as a support material.” Bioprocess Biosyst. Eng. 35, 2012, 29–33.

12. Jason S. Lupoi and Emily a. Smith. “Evaluation of nanoparticle-immobilized cellu-lase for improved ethanol yield in simultaneous saccharification and fermentation re-actions.” Biotechnol. Bioeng. 108, 2011, 2835–2843.

13. Thais L. Ogeda, Igor B. Silva, Ludmila C. Fidale, Omar a. El Seoud, and DeniseF. S. Petri. “Effect of cellulose physical characteristics, especially the water sorptionvalue, on the efficiency of its hydrolysis catalyzed by free or immobilized cellulase.” J.Biotechnol. 157, 2012, 246–252.

14. Katherine J. Mackenzie and Matthew B. Francis. “Recyclable thermoresponsive polymer-cellulase bioconjugates for biomass depolymerization.” J. Am. Chem. Soc. 135, 2013,293–300.

15. Katherine J. Mackenzie and Matthew B. Francis. “Effects of NIPAm polymer additiveson the enzymatic hydrolysis of Avicel and pretreated Miscanthus.” Biotechnol. Bioeng.111, 2014, 1792–1800.

16. Klaus H. Mosbach. Magnetic Polymer Particles. 1982.

17. Marta Gomez-Garcıa, Juan M. Benito, David Rodrıguez-Lucena, Jian-Xin Yu, Kaz-imierz Chmurski, Carmen Ortiz Mellet, Ricardo Gutierrez Gallego, Alfredo Maestre,Jacques Defaye, and Jose M. Garcıa Fernandez. “Probing secondary carbohydrate-protein interactions with highly dense cyclodextrin-centered heteroglycoclusters: theheterocluster effect.” J. Am. Chem. Soc. 127, 2005, 7970–1.

18. Andreas Steffen, Carolin Thiele, Simon Tietze, Christian Strassnig, Andreas Kamper,Thomas Lengauer, Gerhard Wenz, and Joannis Apostolakis. “Improved cyclodextrin-based receptors for camptothecin by inverse virtual screening.” Chem. - A Eur. J. 13,2007, 6801–6809.

19. Gerhard Wenz, Christian Strassnig, Carolin Thiele, Annegret Engelke, Bernd Morgen-stern, and Kaspar Hegetschweiler. “Recognition of ionic guests by ionic β-cyclodextrinderivatives.” Chemistry 14, 2008, 7202–11.


20. Rodolfo F. Gomez-Biagi, Richard B. C. Jagt, and Mark Nitz. “Remarkably stableinclusion complexes with heptakis-[6-deoxy-6-(2-aminoethylsulfanyl)]-β-cyclodextrin.”Org. Biomol. Chem. 6, 2008, 4622–6.

21. Zhiwei Yang and Ronald Breslow. “Very strong binding of lithocholic acid to β-cyclodex-trin.” Tetrahedron Lett. 38, 1997, 6171–6172.

22. Li Zhang, Yaowen Wu, and Luc Brunsveld. “A synthetic supramolecular constructmodulating protein assembly in cells.” Angew. Chem. Int. Ed. Engl. 46, 2007, 1798–802.

23. Dana A. Uhlenheuer, Dorothee Wasserberg, Hoang Nguyen, Li Zhang, Christian Blum,Vinod Subramaniam, and Luc Brunsveld. “Modulation of protein dimerization by asupramolecular host-guest system.” Chemistry 15, 2009, 8779–90.

24. S. H. Yalkowsky and R. M. Dannenfelser. The AQUASOL dATAbASE of AqueousSolubility. Fifth ed. Tucson, AZ: University of Arizona, College of Pharmacy, 1992.

25. Dragos Vizitiu, Caroline S. Walkinshaw, Borio I. Gorin, and Gregory R. J. Thatcher.“Synthesis of monofacially functionalized cyclodextrins bearing amino pendent groups.”J. Org. Chem. 62, 1997, 8760–8766.

26. Atsushi Yamamoto, Tomofumi Hirukawa, Ichiro Hisaki, Mikiji Miyata, and NorimitsuTohnai. “Multifunctionalized porosity in zeolitic diamondoid porous organic salt: se-lective adsorption and guest-responsive fluorescent properties.” Tetrahedron Lett. 54,2013, 1268–1273.



FT W1 W2 W3 E1 E2 E3

0.0

0.2

0.4

0.6

Nor

med

Abs

orba

nce

at 2

80 n

m Normed data

Cumulative sum

Figure 5.6: Low-salt pulldown of EGPh modified with an adamantyl group.

−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0Volume of Sample Cell (ml)Fraction

−0.04

−0.02

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

Abs

orba

nce

at 2

80nm

K a = 3.35[EGPh]=9.32 uMR 2 = 0.998

−2 −1 0 1 2 3 4 5Volume of Sample Cell (ml)

−0.004−0.003−0.002−0.001

0.0000.0010.002

Exp

- Cal

c

Residuals

ft w1 w2 e1 e2 e3

0.0

0.2

0.4

0.6

0.8

1.0

Nor

med

Abs

orba

nce

at 2

80 n

m

Figure 5.7: Pulldown and titration of EGPh modified with an adamantyl group. Left: pulldown.A significant amount of protein elutes during the washes. Right: titration. Protein was titratedinto a solution containing resin, and a binding constant was determined by measuring the amountof protein remaining in solution.

108

Appendix A

Simulation code

This code was developed using Python 2.7 with NumPy 1.9.2, SciPy 0.15.1, and Matplotlib1.4.3.

A.1 Kinetic binding model

This code is intended to be run within an IPython notebook.

%pylab inline

from scipy.integrate import odeint

from IPython.parallel import Client

from scipy.signal import argrelmax

import time

species = 11

_P_i, _PRi, RP_i, _PSi, SP_i, RPSi, SPRi, SPSi, RPRi, Si, Ri = \

range(0,species)

def f(y, t, plate_flow, Ka1, k1on, Ka2, k2on, Ka3, k3on, gradient):

ar_y = y.reshape(-1, species)

out = zeros(ar_y.shape)

_P_ = ar_y[:,_P_i]

_PR = ar_y[:,_PRi]

RP_ = ar_y[:,RP_i]

_PS = ar_y[:,_PSi]

SP_ = ar_y[:,SP_i]

RPS = ar_y[:,RPSi]

SPR = ar_y[:,SPRi]

APPENDIX A. SIMULATION CODE 109

SPS = ar_y[:,SPSi]

RPR = ar_y[:,RPRi]

S = ar_y[:,Si]

R = ar_y[:,Ri]

k1off = k1on / Ka1

k2off = k2on / Ka2

k3off = k3on / Ka3

if k2on == 0: ind = 0

else: ind = 1.0

out[:,_P_i]= -(k1on*_P_*R - k1off*RP_) - ind*(k1on*_P_*R - k1off*_PR)\

- (k3on*_P_*S - k3off*SP_) - ind*(k3on*_P_*S - k3off*_PS)

out[:,_PRi]= ind*(k1on*_P_*R - k1off*_PR) - (k2on*_PR*R - k2off*RPR)\

- (k3on*_PR*S - k3off*SPR)

out[:,RP_i]= (k1on*_P_*R - k1off*RP_) - (k2on*RP_*R - k2off*RPR) -\

ind*(k3on*RP_*S - k3off*RPS)

out[:,_PSi]= ind*(k3on*_P_*S - k3off*_PS) - (k3on*_PS*S - k3off*SPS)\

- (k2on*_PS*R - k2off*RPS)

out[:,SP_i]= (k3on*_P_*S - k3off*SP_) - ind*(k3on*SP_*S - k3off*SPS)\

- (k2on*SP_*R - k2off*SPR)

out[:,RPSi]= ind*(k3on*RP_*S - k3off*RPS) + (k1on*_PS*R - k1off*RPS)

out[:,SPRi]= (k3on*_PR*S - k3off*SPR) + ind*(k1on*SP_*R - k1off*SPR)

out[:,SPSi]= (k3on*_PS*S - k3off*SPS) + ind*(k3on*SP_*S - k3off*SPS)

out[:,RPRi]= (k2on*_PR*R - k2off*RPR) + (k2on*RP_*R - k2off*RPR)

out[:,Si]= -(k3on*_P_*S - k3off*SP_) - ind*(k3on*_P_*S - k3off*_PS)\

- (k3on*_PS*S - k3off*SPS) - ind*(k3on*SP_*S - k3off*SPS)\

- (k3on*_PR*S - k3off*SPR) - ind*(k3on*RP_*S - k3off*RPS)

out[:,Ri]= -(k1on*_P_*R - k1off*RP_) - ind*(k1on*_P_*R - k1off*_PR)\

- (k2on*_PR*R - k2off*RPR) - (k2on*RP_*R - k2off*RPR)\

- (k1on*_PS*R - k1off*RPS) - ind*(k1on*SP_*R - k1off*SPR)

#operations between plates (rows)

p_i = _P_i

additional = hstack((0, ar_y[:,p_i]))

out[:,p_i] = out[:,p_i] - plate_flow*ar_y[:,p_i] + \

plate_flow*additional[0:-1]

p_i = _PSi





p_i = SP_i




p_i = SPSi




p_i = Si




s = out.shape

for grad in gradient:

(a, b, conc) = grad

out[0, Si] = out[0, Si] + plate_flow*conc*((0.5*(t-a)+\

0.5*abs(t-a))\

-(0.5*(t-(a+b))+0.5*abs(t-(a+b))))/(b)

return out.reshape(-1,1)[:,0]

Usage:

col_len = 25 #in cm

plates_per_cm = 2.2

col_diameter = 0.5

filling = 1

col_vol = 4#ml

flow = 0.20 #ml/min

flow = flow/60.0 #ml/s

plates = floor(plates_per_cm*col_len)

print ’Using’,plates,’plates’

plate_vol = col_vol / plates


plate_flow = flow/plate_vol

times_min = arange(0, 150, 0.01) #minutes

times = times_min*60.0 #seconds

volumes = flow*times

gradient = [(2*5*60, 25*5*60, 0.01)]

kinetic = 10**3.0

cd_conc = 1.0*(10**-3.0)

protein_conc = 100*(10**-6.0)

binding_constant = 10**3.0

if True:

start = time.time()

col = zeros((plates, species))

col[:, Ri] = cd_conc # cd concentration

col[0:1, _P_i] = protein_conc #starting protein

linear_col = col.reshape(-1,1)

results1 = odeint(f,

linear_col[:,0],

times,

args=(plate_flow, binding_constant, kinetic, \

binding_constant, kinetic, \

binding_constant, kinetic, \

gradient))

t1=results1[:,-species+_P_i]+results1[:,-species+_PSi]+\

results1[:,-species+SP_i]+results1[:,-species+SPSi]

print ’Protein in %0.7f’%(protein_conc*plate_vol)

print ’Protein out %0.7f’%(t1.sum()*(volumes[1]-volumes[0]))

stop = time.time()

print ’Completed in’,(stop-start),’s’

t1=1000000*t1

figure(1)

clf()

xlabel(’Volume / ml’)

title(’Column volume %0.2f ml’%(col_vol))

t1dist = t1/(sum(t1)*(volumes[1]-volumes[0]))

v1 = sum(volumes*volumes*t1dist)*(volumes[2]-volumes[1])


m1 = sum(volumes*t1dist)*(volumes[2]-volumes[1])

plot(volumes, t1, label=’Unmodified’)

a=gca()

a.set_ylabel(’[Protein] / uM’)

a2=a.twinx()

a2.plot(volumes, results1[:, -species+Si]+\

results1[:, -species+_PSi]\

+results1[:, -species+SP_i]+2.0*results1[:, -species+SPSi],

color=’Green’)

a2.set_ylim((0, 0.016))

a2.set_ylabel(’[BCD]’)

draw()

A.2 Analytic binding model

This code is intended to be run within an IPython notebook.

%pylab inline

from __future__ import division

def get_sol_conc(f, Fx, kax, Fa, kas, ms):

return 2.0**(-f)*Fa*( (1/(f*kax*Fa))*(-1.0+f*kax*Fa-kax*Fx-\

kas*ms+sqrt( -4.0*f*kax*Fa*(-1-kas*ms)+ \

((1-f*kax*Fa+kax*Fx+kas*ms)**2.0))) )**f

def run_column(col_vol, multiplicity, plates,

resin, kax, kas, protein_start, gradient):

vol_per_plate = col_vol / plates

vi, vs = 0, 150

flow=0.2 # ml/min

plate_flow = flow/vol_per_plate

timestep = 0.05 #minutes

iterations = int(floor(vs / (timestep*flow)))

volumes = array(range(0, iterations))*flow*timestep

solution = zeros(plates)


bound = zeros(plates)

eluent = zeros(plates)

solution[0] = protein_start

out=[]

outeluent=[]

for i in range(0, iterations):

current_volume = i*timestep*flow

#equilibrate the column

total = solution+bound

solution = get_sol_conc(multiplicity, resin,

kax, total, kas, eluent)

bound = total - solution

bound[isnan(bound)] = 0

solution[isnan(solution)] = 0

#flow the column

solution_change = plate_flow*solution.copy()*timestep

eluent_change = plate_flow*eluent.copy()*timestep

out.append(solution_change[-1]/(timestep*plate_flow))

outeluent.append(eluent_change[-1]/(timestep*plate_flow))

solution = solution - solution_change

solution[1:] = solution[1:] + solution_change[:-1]

eluent = eluent - eluent_change

eluent[1:] = eluent[1:] + eluent_change[:-1]

for grad in gradient:

(a, b, conc) = grad

eluent[0] = eluent[0] + plate_flow*timestep*conc*\

((0.5*(current_volume-a)+\

0.5*abs(current_volume-a))-\

(0.5*(current_volume-(a+b))+\

0.5*abs(current_volume-(a+b))))/(b)

return (array(volumes), array(out), array(outeluent))

def get_multiples(upto, col_vol, plates, resin, kax, kas, scale, \

protein_start=0.00001, gradient=[(2.5, 0.01, 0.002), \

(9,0.01,0.0025), (21,0.01,0.005)]):


for multiplicity in range(0, upto):

if multiplicity ==0:

vol,out,outeluent = run_column(col_vol,

multiplicity, plates,

resin, kax, kas,

protein_start, gradient)

else:

out=out+run_column(col_vol, multiplicity, plates, resin, kax,\

kas, protein_start, gradient)[1]

return vol,out*scale,outeluent

Usage:

resin = 0.001

kax = 10**3.5

kas = 10**3.5

#time iteration

gradient = [(2, 200, 0.01)]

vol,trace,gradient = get_multiples(20, 4, 100*25/10,\

0.001, kax, kas, 1, 0.00001, gradient)

figure(1, figsize=(6,1.5))

clf()

plot(vol, trace*1000000)

xlabel(’Volume / ml’)

ylabel(’[Protein] / uM’)

ylim(ymin=0)

na=gca().twinx()

na.plot(vol,gradient*1000, ’--’, color=rcParams[’axes.color_cycle’][1])

ylabel(’[BCD] / mM’)

xlim((0, 150))

ylim((0,10))

draw()

savefig(’thesis-num-mod-states-4-grad2.pdf’)

A.3 Lateral flow

This code is intended to be run within an IPython notebook. The binding model used isbased on work by Wang.1


%pylab inline

from scipy.integrate import odeint

from __future__ import division

import time

import matplotlib.animation as animation

def equilibrate(fa, fb, fp, ka, kb):

ka, kb = 1.0/ka, 1.0/kb

a = ka + kb + fa + fb - fp

b = kb*(fa-fp) + ka*(fb-fp) + ka*kb

c = -ka*kb*fp

theta = arccos((-2.0*a*a*a+9.0*a*b-27.0*c)/\

(2*sqrt((a*a-3.0*b)**3.0)))

p = -(a/3.0) + (2.0/3.0)*sqrt(a*a-3.0*b)*cos(theta/3.0)

pa = (fa*(2*sqrt(a*a-3*b)*cos(theta/3.0)-a)) / \

(3.0*ka+(2*sqrt(a*a-3*b)*cos(theta/3.0)-a))

pb = (fb*(2*sqrt(a*a-3*b)*cos(theta/3.0)-a)) / \

(3.0*kb+(2*sqrt(a*a-3*b)*cos(theta/3.0)-a))

a = fa - pa

b = fb - pb

fa / (a+pa)

return a, b, p, pa, pb

species = 5

# protein, dye, analyte, control

prot_i, dye_i, anal_i, prot_dye_i, prot_anal_i = range(0,species)

def run_column(protein_start=10**-10.0,

anal_start=10**-9.0,

dye_start=10**-5.5,

kanal=10**10.0,

kdye=10**8.0,

timestep=0.1,

verbose=False,

dtype=longdouble,

stop=2.0):

#width, length, thickness = 8, 50.0, 0.15#units of mm

#porosity = 0.7

col_vol = 0.2 #porosity*width*length*thickness / 1000.0 #ml

if verbose: print ’Column volume is %0.2f ul’%(col_vol*1000)


flow = 0.01 #ml/min

flow = flow/60.0 #ml/s

plates = 100

plate_vol = col_vol / plates

plate_flow = flow/plate_vol

if verbose: print ’Plate volume is %0.2f ul’%(plate_vol*1000)

vstart, vstop = 0, col_vol*stop

iterations = int(floor(vstop/(timestep*flow)))

if verbose: print iterations

volumes = array(range(0, iterations))*flow*timestep

if verbose: print ’pf timestep product:’,plate_flow*timestep

col = zeros((plates, species), dtype=dtype)

col[:, prot_i] = protein_start

col[:, dye_i] = dye_start

col[0, anal_i] = anal_start

start = time.time()

out = zeros((plates, species, iterations))

for i in range(0, iterations):

out[:,:,i] = col

current_volume = i*timestep*flow

prot = col[:,prot_i]

dye = col[:,dye_i]

anal = col[:,anal_i]

prot_dye = col[:,prot_dye_i]

prot_anal = col[:,prot_anal_i]

col[:,anal_i], col[:,dye_i], col[:,prot_i], \

col[:,prot_anal_i], col[:,prot_dye_i] =\

equilibrate(anal+prot_anal, dye+prot_dye, \

prot+prot_dye+prot_anal, kanal, kdye)

col[isnan(col)] = 0

#flow the column--things that flow are the analyte

#and the capture (dye) reagent

for key in [anal_i, dye_i]:


change = col[:,key].copy()*plate_flow*timestep

col[:,key] = col[:,key] - change[:]

col[1:, key] = col[1:, key] + change[:-1]

#continuous input of the analyte at the beginning

if key == anal_i: col[0, key] = col[0, key] + \

anal_start*plate_flow*timestep

out[:,:,i] = col

if verbose: print "Calculation took %0.2fs"%(time.time()-start)

return out, iterations, volumes

Usage:

#2D heatmap

#analyte concentration

xstart, xstop, xdim = -9, -4.0, 20

#protein concentration

ystart, ystop, ydim = 0, 1, 20

kdye, kanalyte = 1.0/(20*10**-9.0), 1.0/(100*10**-9.0)

out = zeros((ydim, xdim))

dyepow = log10(1*10**-6)

xs, ys = linspace(xstart, xstop, xdim), linspace(ystart, ystop, ydim)

f=FloatProgress(min = 0, max = xdim*ydim)

display(f)

for i,analpow in enumerate(xs):

for j,frac in enumerate(ys):

f.value=ydim*i+j+1

rs = []

for multiplier in [0.0, 1.0]:

result, iterations, volumes = run_column(

anal_start = multiplier*10**analpow,

dye_start = frac*10**dyepow,

protein_start = 10**dyepow,

kdye=kdye,

kanal=kanalyte,

timestep=1.0,

stop=2.0)


rs.append(result)

out[j,i] = rs[1][-1, dye_i-species, :].sum() / \

rs[0][-1, dye_i-species, :].sum()

Plotting:

figure(4, figsize=(5,5))

clf()

vmin, vmax = 0, 2.4

ypad, xpad = 0.5*((ystop-ystart)/(ydim-1)), 0.5*((xstop-xstart)/(xdim-1))

ymin, ymax, xmin, xmax = ystart-ypad, ystop + ypad, \

xstart-xpad, xstop+xpad

imshow(log10(out), interpolation=’none’, cmap=’rainbow’, \

extent=(xmin, xmax, ymax, ymin), vmin=vmin, vmax=vmax,\

aspect=(xstop-xstart)/(ystop-ystart))

xticks(xs, ["%0.2f"%i for i in xs], rotation=’vertical’)

yticks(ys, ["%0.2f"%i for i in ys])

ylabel(’Fraction occupied’)

xlabel(’[Analyte]’)

xlim((xmin, xmax))

ylim((ymin, ymax))

ticks = linspace(vmin, vmax, 10)

cbar = colorbar(ticks=ticks)

cbar.ax.set_yticklabels([’%0.2f’%(10**i) for i in ticks])

tight_layout()

draw()

A.4 References

1. Zhi X. Wang. “An exact mathematical expression for describing competitive bindingof two different ligands to a protein molecule.” FEBS Lett. 360, 1995, 111–114.

affinity-mediated protein modification and recovery · doctor of philosophy in chemistry university...

Documents