Semisynthetic glycoconjugate vaccine candidateagainst Streptococcus pneumoniae serotype 5Marilda P. Lisboaa,1,2, Naeem Khana,1, Christopher Martina,b,3, Fei-Fei Xua,b, Katrin Reppec, Andreas Geissnera,b,Subramanian Govindana,4, Martin Witzenrathc, Claney L. Pereiraa,2,5, and Peter H. Seebergera,b,5

aDepartment of Biomolecular Systems, Max Planck Institute of Colloids and Interfaces, D-14424 Potsdam, Germany; bDepartment of Chemistryand Biochemistry, Freie Universität Berlin, D-14195 Berlin, Germany; and cDepartment of Infectious Diseases and Pulmonary Medicine,Charité - Universitätsmedizin Berlin, 10117 Berlin, Germany

Edited by Michael L. Klein, Temple University, Philadelphia, PA, and approved September 8, 2017 (received for review April 25, 2017)

Glycoconjugate vaccines based on isolated capsular polysaccha-ride (CPS) save millions of lives annually by preventing invasivepneumococcal disease caused by Streptococcus pneumoniae.Some components of the S. pneumoniae glycoconjugate vaccinePrevnar13 that contains CPS antigens from 13 serotypes undergomodifications or degradation during isolation and conjugation,resulting in production problems and lower efficacy. We illustratehow stable, synthetic oligosaccharide analogs of labile CPS inducea specific protective immune response against native CPS usingS. pneumoniae serotype 5 (ST-5), a problematic CPS componentof Prevnar13. The rare aminosugar L-PneuNAc and a branchedL-FucNAc present in the natural repeating unit (RU) are essentialfor antibody recognition and avidity. The epitope responsible forspecificity differs from the part of the antigen that is stabilizedby chemical modification. Glycoconjugates containing stable,monovalent synthetic oligosaccharide analogs of ST-5 CPS RU in-duced long-termmemory and protective immune responses in rab-bits superior to those elicited by the ST-5 CPS component inmultivalent Prevnar13.

glycoconjugate | vaccine | S. pneumoniae | serotype 5 |carbohydrate chemistry

Pneumococcal infections continue to cause millions of fatali-ties among children and the elderly despite the widespread

use of glycoconjugate vaccines (Synflorix, Prevnar13) (1). Thesevaccines aim at inducing an immune response against bacterialcapsular polysaccharide (CPS) not present on human cells (2).Streptococcus pneumoniae is a Gram-positive human pathogencovered by CPS that is diverse and contains rare sugars (3, 4). Allcurrently marketed pneumococcal vaccines (5) are manufacturedusing CPS isolated from the surface of S. pneumoniae.S. pneumoniae serotype 5 (ST-5) is the fifth most prevalent

among more than 90 S. pneumoniae serotypes with differentCPS, causing invasive pneumococcal disease among young chil-dren globally (6, 7). Marketed glycoconjugate vaccines are notfully efficacious in preventing ST-5 infections (8). A change inthe CPS glycan structure during antigen isolation and purifica-tion such that the ST-5 antigens no longer sufficiently resemblethe native CPS may compromise vaccine efficacy (9, 10).Manufacturing glycoconjugate vaccines such as for ST-5 can beproblematic when CPS contains labile groups (11–13).The ST-5 repeating unit (RU) structure was assigned in 1985

following the initial classification in 1929 (3, 11–14). Thebranched pentasaccharide ST-5 CPS RU 1 contains a centralN-acetyl-L-fucosamine (L-FucNAc) amino sugar that is linked toD-glucose at C4 and to D-glucuronic acid at C3 (Fig. 1A). Tworare deoxyamino sugars, the ketoamino sugar 2-acetamido-2,6-dideoxy-D-xylose-hexos-4-ulose (Sugp) and N-acetyl-L-pneumosamine(L-PneuNAc), complete the RU.Marketed glycoconjugate vaccines are manufactured from ei-

ther native or depolymerized CPS (15) that is typically coupled toa carrier protein via reductive amination following isolation. Theketo group present in the rare sugar Sugp is partially or fully

reduced to form a mixture of ST-5 CPS components and degradesduring ST-5 glycoconjugate production (10). The complex CPSthus generated is characterized by variable RUs, leading to manu-facturing issues and decreased immunogenicity compared with thenative ST-5 CPS (10).Defined synthetic antigens are essential tools to identify pro-

tective glycan epitopes for the development of semisynthetic gly-coconjugate vaccines (2). Employing synthetic ST-5 glycans basedon the CPS RU provides valuable insights into how changes to thenatural ST-5 CPS may influence antigen stability and immuno-genicity. A flexible total synthesis approach had to be conceived toprovide access not only to the natural keto containing RU 1 butalso to the reduced forms of 1, particularly oligosaccharides 2, 3,and 4 (Fig. 1A). General aspects of vaccine design relating to theeffect of branching, length, and the role of unique sugars likeL-PneuNAc and Sugp on overall immunogenicity and protection inthe context of the RU 1 had to be considered. A series of oligo-saccharides related to the RU of ST-5 CPS equipped with a re-ducing end linker were synthesized to be fixed on glycan arrays

Significance

Each year, Streptococcus pneumoniae infections cause millionsof deaths worldwide. The capsular polysaccharide (CPS) basedglycoconjugate vaccine Prevnar13 prevents serious illness causedby 13 serotypes. S. pneumoniae serotype 5 (ST-5) is included inthe vaccine; however, it suffers from production problems dueto modifications or degradation during isolation and conjuga-tion. A medicinal chemistry approach helped to understand thestructural features of ST-5 CPS and design a stable semisyntheticoligosaccharide-based vaccine candidate. Oligosaccharide leadsfor immunological evaluations in vivowere identified employingglycan microarrays. The stable monovalent ST-5 oligosaccharideglycoconjugate vaccine candidate showed a superior immuneresponse in rabbits when compared with the ST-5 CPS present inthe multivalent vaccine Prevnar13.

Author contributions: P.H.S. designed research; M.P.L., N.K., C.M., F.-F.X., K.R., A.G., S.G.,and C.L.P. performed research; K.R. and M.W. contributed new reagents/analytic tools;M.P.L., N.K., A.G., C.L.P., and P.H.S. analyzed data; M.P.L., N.K., C.L.P., and P.H.S. wrote thepaper; and A.G. helped create figures.

Conflict of interest statement: P.H.S. declares a significant financial interest in Vaxxilon,the company that commercializes the Streptococus pneumoniae vaccine candidate de-scribed here. There is, however, no conflict of interest, as this is a purely scientific man-uscript that was not sponsored by the company.

This article is a PNAS Direct Submission.1M.P.L. and N.K. contributed equally to this work.2Present address: Vaxxilon Deutschland GmbH, 12489 Berlin, Germany.3Present address: Bachem AG, 4416 Bubendorf, Switzerland.4Present address: Department of Chemistry, Indian Institute of Technology Tirupati,Tirupati, Andhra Pradesh 517506, India.

5To whom correspondence may be addressed. Email: claney.pereira@vaxxilon.com orpeter.seeberger@mpikg.mpg.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1706875114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1706875114 PNAS | October 17, 2017 | vol. 114 | no. 42 | 11063–11068

CHEM

ISTR

YMED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

0

and conjugated to the carrier protein diphtheria toxin mutantCRM197, currently used in the vaccine Prevnar13 (16).A retrosynthetic analysis of the ST-5 RU revealed the need for

five differentially protected monosaccharide building blocks (5–9). The assembly of reduced ST-5 RU glycans 2 and 3 (Fig. 1B)

requires the procurement of the rare amino sugars L-pneumos-amine (9) via a novel synthetic route, as little attention hasbeen paid to this sugar (17–19). L- and D-fucosamine (5 and 8),the other two aminosugars, as well as D-glucose (6) andD-glucuronic acid (7) building blocks, will be synthesized via

D L D

β4 α3

β3

α2D

L

O4β4[ ]

O

ONHAc

OOHO O

HO

OH O

O

NHAc

OHO

HO OO

OHO

O

NHAcHOOH

O

L D

α3

β3

α2D

L

NH2β

OHOHO

O

O

OHO

ONHAc

O5

NH2

ONHAc

HO

O

O

NHAcHOOH

HO

4

Sugp

O

NAPO

HON3

O5

NBnCbz

ON3

HOOPMB

OBnOBnO SEt

LevO

OBnO

OTDS

I

8

5OBnO

BnO SEtLevO

OBn

6

7

A B

D L D

β4 α3

β3

α2D

L

NH2β

3

O

R1

ONHAc

OHOHO O

HO

OH O

O

NHAc

OHO

HO OO

OHO

O

NHAcHOOH

R2

5

NH2

2

D L D

β4 α3

β3

α2D

L

NH2β

1

D-GlcLegend:

FucNAc D-QuiNAcD-GlcA L-PneuNAc

L-PneuNAc

L-FucNAc

2 R1 = OH, R2 = H3 R1 = H, R2 = OH

ST-5 RU

II

IV

III

O

N3BnOOBn

O

NPh

CF3

9

Fig. 1. ST-5 CPS RU (1) and synthetic antigen targets (2–4). (A) Structure of the ST-5 CPS RU and synthetic antigen targets. (B) Retrosynthetic analysis ofST-5 reduced RU 2.

OBnO

BnO ORO

OBn O OTDSN3

OPMBO

BnOBnO O

BnO

OBn O OTDSN3

OH

25 R = Lev

26 R = Bn

OBnO

BnO OBnO

OBn O

OR

N3

OBnO

BnO OLevO

OBnO

OBnO

BnO OBnO

OBn ON3

OBnO

BnO ORO

OBnO

O

NAPO

ON3

O

5

NBnCbz

OBnO

BnO OBnO

OBn ON3

OBnO

BnO OO

OBnO

O

NAPO

ON3

O

5

NBnCbz

O

N3BnOOBn

5 + 6

2

NIS, TfOH

DCM, 78%

1) N2H4.H2O, py, AcOH, DCM, 82%2) BnBr, NaH, DMF, 90%

28 R = TDS 29 R = H

30 R = Lev31 R = H

N2H4.H2O, py,AcOH, DCM, 90%

9, TMSOTf

tol, 53%

27

32

DDQ

DCM, H2O 83%

7, NIS, TfOH,

DCM:tol (1:2), 76%

HF.py, py, DCM, 85%

1) CF3C(NPh)Cl, Cs2CO3, DCM

2) 8, TMSOTf DCM, 79%

1) CH3COSH, py, 98%

2) H2, Pd(OH)2, DCM,tBuOH, H2O, 50%

O

SePh

N3

AcOOAc

O

N3AcOOAc

SePh

O

AcO

OAc

5

9

A

10

Ph2Se2, PhI(OAc)2

TMSN3, DCM

11 (52%)

12 (< 25%)

1) NIS, THF, H2O2) TDSCl, Imid., DMF3) NaOMe, MeOH

4) Bu2SnO, tol, then PMBCl TBAB, 66% over 4 steps

1) NaOMe, MeOH2) BnBr, NaH, DMF

3) NIS, THF, H2O4) CF3C(NPh)Cl, Cs2CO3 DCM, 20% over 4 steps

B

Scheme 1. Synthesis of ST-5 reduced RU. (A) Synthesis of building blocks 5 and 9. (B) Synthesis of pentasaccharide 2. BnBr, benzyl bromide; DCM,dichloromethane; DDQ, 2,3-dichloro-5,6-dicyano-1,4-benzoquinone; Lev, levulinoyl group; NIS, N-iodosuccinimide, py, pyridine; TDS, thexyl-dimethylsilyl;TfOH, trifluoromethanesulfonic acid; TMSOTf, trimethylsilyl trifluoromethanesulfonate; Tol, toluene.

11064 | www.pnas.org/cgi/doi/10.1073/pnas.1706875114 Lisboa et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

0

modified literature protocols (20, 21) (Fig. 1B). The syntheticallychallenging branched 3,4-substituted L-fucosamine has to beinstalled early in the oligosaccharide syntheses to avoid issuesrelated to sterics and reactivity. D-Fucosamine and L-pneu-mosamine will be added sequentially at the end of the assemblyprocess (Fig. 1B). Finally, reduction and acetylation of theazido groups followed by global deprotection will provide thedesired glycans.The synthesis commenced with the procurement of the dif-

ferentially protected monosaccharide building blocks. While known

synthetic approaches to L-fucosamine building blocks are sufficient(22–24), the few routes to L-pneumosamine suffer from low yields(17–19). The azido-phenylselenation reaction was selected askey to access mainly L-fucosamine (8) and less L-pneumos-amine (9) from L-fucal (10) (22) (Scheme 1A). Protectinggroup manipulations of 11 and 12 provided nucleophile 5 andglycosylating agent 9, respectively (Scheme 1A). D-fucosamine8 was derived from D-fucal using the same approach (SI Ap-pendix, Scheme S1). The syntheses of differentially protectedD-glucose 6 (20) and D-glucuronic acid 7 building blocks followedestablished procedures (SI Appendix, Scheme S2).Using the building blocks that were designed for potential

access to the complete RU including the Sugp residue, severalsaccharides were synthesized for microarray analyses to identifyimmunodominant fragments of the RU (Fig. 2A). While thesynthetically challenging Sugp was included in this microarrayanalysis, the binding pattern seen after incubation with two dif-ferent sera strongly suggested that the GlcA-PneuNAc branch isthe most important substructure for tight antibody binding. Arabbit serum used to identify ST-5 strains in serotyping proce-dures showed strong reactivity to disaccharide 21 representingthe branch, followed by the PneuNAc monosaccharide 23 in

B C

D E

A

13 14 15 16

17 18 19 20

21 22 23 24

NH2

D

NH2

O4

NH2

D

NH2

D

NH2 NH2

NH2βNH2

αNH2NH2

D

O4

D

+ NH2NH2

D

NH2

No Inhibitor

13 14

21 22 23 24

10 μg/mL ST5 CPS

13 14

21 22 23 2413 14 15 16 17 18 19 20 21 22 23 24

0

2000

4000

6000

8000

Compound number

Mea

nFl

uore

scen

ceIn

tens

ity

No InhibitorST2 CPSST5 CPS

No Inhibitor

13 14

21 22 23 24

10 μg/mL ST5 CPS

13 14

21 22 23 2413 14 15 16 17 18 19 20 21 22 23 24

0

2000

4000

6000

8000

Compound number

Mea

nFl

uore

scen

ceIn

tens

ity

No InhibitorST2 CPSST5 CPS

0.2 mM 0.1 mMPrinting concentration

Fig. 2. Microarray screening to identify ST-5 oligosaccharide fragments ashits for vaccine development. (A) Microarray printing pattern with ST-5 RUfragments. Printed arrays were incubated with (B) rabbit typing serum (di-lution 1:1,800) or (D) human reference serum (007sp, dilution 1:40). Thebound antibodies were detected using fluorescently labeled secondary an-tibodies. For the inhibition study, sera were preincubated with ST-5 CPS(10 μg/mL) and ST-2 CPS (10 μg/mL) as a control. ST-2 CPS was used as controlbecause it resembles ST-5 CPS containing a branched linkage, a commonGlcA moiety, and three deoxy sugars. Printed arrays were incubated with theabove sera, and inhibition was analyzed by using fluorescently labeled sec-ondary antibodies. Mean fluorescence intensities (MFI) of inhibition assaywith (C) rabbit typing serum and (E) human reference serum (007sp) areshown. Data are represented as mean ± SD of duplicate determinations.

e

0 14 21 28 35

Serum collection (day)

Anti-CRM197-4

13 14 33 2

3 4

21

Anti-CRM197-2

13 14 33 2

3 4

21

A B

C

D

E

13 14

ST-5

CPS

(1)

CW

PS

33 2

3 4 15 16

17 18 19 20

21 22 23 24

2

NH2 NH2

4ImmunogenicityStability

F

CPS 5 2 3 4 13 14 33 21

0

500

1000

1500

2000

2500

3000

3500

4000

Compound number

Mea

nFl

uore

scen

ceIn

tens

ity

CRM197-4CRM197-2

Fig. 3. Microarray with rabbit anti–CRM197-2 and anti–CRM197-4 conju-gates sera. (A) Immunization and sera collection schedule. (B) Microarrayslide printing pattern. Rabbits were immunized with CRM197-2 andCRM197-4 conjugates in a prime boost manner on days 0, 14, and 28. Pre-immune and hyperimmune sera were collected at different time points fromindividual rabbits. (C and D) Microarray slides were incubated with rabbitsera (day 35) raised against CRM197-2 and CRM197-4 conjugates. (E) MFI ofcross-reactive spots are plotted as mean ± SD in duplicates. (F) Representa-tion of the monosaccharide moieties responsible for immunogenicity andstability within glycans 2 and 4.

Lisboa et al. PNAS | October 17, 2017 | vol. 114 | no. 42 | 11065

CHEM

ISTR

YMED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

0

alpha configuration with already comparatively lower signals(Fig. 2 B and C).Importantly, an inhibition assay using native ST-5 CPS to

capture ST-5-specific antibodies led to high degrees of signalsuppression for all recognized oligosaccharides. The humanreference serum pool 007sp (25) was used to confirm the findingsof the rabbit serum but revealed a less clear picture (Fig. 2 D andE). Analogous to the rabbit serum, the PneuNAc containingglycans 21, 23, and 24 showed higher degrees of signal sup-pression compared with the other oligosaccharides, suggestingthat the branch is also important for reactivity in humans, theultimate recipients of vaccines. After identifying the branch asimmunodominant, we focused the RU synthesis on the moreaccessible reduced glycans 2, 3, and 4 that were expected to bemore stable than a Sugp containing compound.The assembly of pentasaccharide 2 began with the synthesis of

branched trisaccharide 28 (Scheme 1B). Glycosylation of theaxial C4-hydroxy group in 5 with glycosylating agent 6 gaveβ-configured disaccharide 25. The participating levulinic esterwas exchanged for a benzyl ether before p-methoxy benzyl(PMB) ether cleavage to furnish nucleophile 27. Glycosylation of27 with thioglucoside 7 gave the desired trisaccharide 28 as asingle anomer. The anomeric linkages of trisaccharide 28 wereascertained unambiguously by 1D z-filtered total correlationspectroscopy (zTOCSY) NMR measurements upon removal ofthe levulinic ester (SI Appendix, Fig. S1).Trisaccharide imidate 29 was obtained from 28, by desilylation

and formation of the glycosyl imidate. Glycosylation usingD-FucNAc building block 8 furnished tetrasaccharide 30 (Scheme1B). Cleavage of the temporary levulinic ester protecting

group in 30 produced nucleophile 31. Glycosylation of 31 withL-pneumosazide 9 gave fully protected α-linked pentasaccharide 32.Branched pentasaccharide 2 was obtained by converting the azidesin 32 to N-acetyl moieties using thioacetic acid followed by globaldeprotection via hydrogenolysis (Scheme 1B).The synthesis of branched pentasaccharide 3 (Fig. 1A) con-

taining a reducing end N-acetyl-D-quinovosamine (D-QuiNAc)began with D-FucNAc building block 8 (SI Appendix, SchemeS1). The D-QuiNAc acceptor 37 was engaged in the synthesis ofpentasaccharide 3 in a manner analogous to nucleophile 8, fol-lowing the reaction sequences detailed for the synthesis of 2 (SIAppendix, Scheme S3). To better understand the role played bythe unique amino sugars and branching in the ST-5 RU, a seriesof related sequences, including the linear tetrasaccharide 4 (Fig.1A), were synthesized (SI Appendix, Scheme S4). Attempts tosynthesize the natural ketone-containing RU 1 did not meet withsuccess, as the intermediate ketone proved highly unstable andcould not be isolated in pure form.Reanalysis of the rabbit typing serum employing an extended

glycan array (SI Appendix, Fig. S2) illustrated that the epitoperecognized by anti–ST-5 CPS antibodies expanded beyond thedisaccharide branch, with much stronger recognition of penta-saccharides 2 and 3 compared with the previous lead, di-saccharide 21 (SI Appendix, Fig. S3). Tetrasaccharide 4 lackingthe D-Glc connected to the FucNAc showed an intermediatesignal. These larger oligosaccharides exhibited an encouragingintensity and inhibition pattern with human reference serum007sp, thus suggesting a role of the larger glycan epitopes inantipneumococcal immunity against ST-5 strains in a human

A B C

D F GCRM197-4 CRM197-2

1000

2000

4000

8000

1600

032

000

6400

012

8000

0.00.20.40.60.81.01.21.41.61.82.0

Ig titer

Abs

orba

nce

(OD 45

0)

CRM197CRM197-4CRM197-2

E CRM197-4

CRM197-2

0

5

10

15

20

25

30

Fold

chan

ge

*** **

CRM197

8 32 128

512

2048

8192

-100

102030405060708090

100

Ig titer

Bac

teria

lkill

ing

(%)

8 32 128

512

2048

8192

-100

102030405060708090

100

Ig titer

Bac

teria

lkill

ing

(%)

CRM197Prevnar 13Glycoconjugate (day 0)Glycoconjugate (day 35)

0 1 1.5 2 3 4 5 60.00.20.40.60.81.01.21.4

NH4SCN (M)

Abs

orba

nce

(OD 45

0)

Prevnar 13CRM197-4CRM197-2

CRM197-2

CRM197-4

Prevna

r 130

1

2

3

4

Avid

ityIn

dice

s(A

I)

100

101

102

103

104

FITC-A

0

20

40

60

80

100

% o

f Max

CRM197-2CRM197-4Preimmune seraSecondary antibody control

Fig. 4. ELISA and OPKA. (A) Pooled sera of rabbits (n = 3) before and after immunization with CRM197-2 and CRM197-4 conjugates were collected, the end pointtiter was analyzed by ELISA, and data were plotted as mean ± SD. (B) The antibody response of day 35 postimmune sera in relation to preimmune sera. Each dotrepresents an individual rabbit immune response, and data were analyzed by an unpaired t test. P values of <0.05 were considered statistically significant. (*P < 0.05;**P < 0.01; ***P < 0.001.) (C) UV inactivated pneumococci were incubated with anti–CRM197-2 and anti–CRM197-4 antibodies, and surface staining was analyzed byflow cytometry. (D) OPKAs were performed with differentiated HL-60 cells (N,N-dimethylformamide, 5 d) and incubated with pneumococci preopsonized withCRM197-4 conjugate-specific antibodies in the ratio of 1:400 (bacteria to effector cell) or with control sera. Bacterial survival was assessed after 45-min incubation.Percent killing of pneumococci was calculated based on viable pneumococcal colonies obtained relative to control sera. The experiment was repeated three times, andrepresentative values of one out of three independent experiments were plotted. Prevnar13 and CRM197 sera were used as positive and negative controls, re-spectively. Data of all experiments were represented as mean ± SD values of triplicates. (E) OPKA as described for D with CRM197-2 conjugate-specific antibodies. (F)The avidity of antibodies was obtained using increasing concentrations (0 M to 4 M) of NH4SCN to perturb antigen−antibody interactions. (G) The AI values werecalculated and plotted against the corresponding antigen labels. Data are represented as mean ± SD, triplicate determinations.

11066 | www.pnas.org/cgi/doi/10.1073/pnas.1706875114 Lisboa et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

0

vaccine setting. Glycan array analyses of anti-Prevnar13 rabbitserum gave similar results (SI Appendix, Fig. S4).Based on insights concerning epitope size and the immuno-

logical response, pentasaccharide 2 and linear tetrasaccharide 4were chosen for immunological evaluation, as they are signifi-cantly easier to synthesize than 3. Glycans 2 and 4 were conju-gated to the carrier protein diphtheria toxin mutant CRM197 (SIAppendix, Fig. S5A) using the p-nitrophenyl adipate ester con-jugation method (26) for single-point attachment and facile re-moval of excess coupling reagent, thus avoiding the formation ofunwanted linker antibodies (27, 28). The glycoconjugatesCRM197-2 and CRM197-4 contained 12 and 11 glycans perprotein, respectively (SI Appendix, Fig. S5 B and C). Any newglycococonjugate vaccine lead compound against pneumococcihas to induce a robust antibody response, as evidenced by aserological assay, antibody-mediated opsonic activity, and along-term memory response laid out in the World HealthOrganization guidelines for pneumococcal glycoconjugate vaccines(29–31). Mindful of these requirements, the vaccine potential ofglycoconjugates CRM197-2 and CRM197-4 was evaluated employ-ing a multiarm study in rabbits, a reliable species for pneumococcalvaccination (32). CRM197-2 and CRM197-4 (10 μg of glycan perimmunization) formulated with the adjuvant aluminum hydroxidewere injected s.c. (Fig. 3A), and the immune response was assessedusing glycan arrays (Fig. 3 B–E). CRM197-2 conjugate stimulatedmore cross-reactive antibodies than CRM197-4 against oligosac-charide 3 and the native ST-5 CPS. These results emphasize theimportance of the D-glucose residue, present in 2 but not in 4, as vitalfor antibody recognition and cross-reactivity. The ST-5 epitope re-sponsible for specificity (L-PneuNAc, D-Glc) differs from the re-ducing end sugar (D-FucNAc) modified to enhance stability (Fig.3F). The branched pentasaccharide 2 is more immunogenic and abetter mimic of the native ST-5 CPS, as conjugate CRM197-2 pro-duced CPS-specific cross-reactive antibodies titers, in contrastto linear tetrasaccharide conjugate CRM197-4 (33, 34) (Fig. 4A and B).Antibodies produced in response to glycoconjugate vaccines

are only protective when they bind to the native CPS on thepneumococcal surface. The phagocytic activity of anti-CPSantibodies is a major protective mechanism against pneumo-cocci that is governed by fixing the complement on the sur-face of bacteria (35). To this end, UV-inactivated pneumococcalbacteria were incubated with antibodies raised against CRM197-2(gray histogram, Fig. 4C) and CRM197-4 (black histogram) conju-gates and were analyzed by flow cytometry (Fig. 4C).Given the limited accessibility of an ST-5 animal challenge

model, the functional relevance of antibodies induced in re-sponse to immunization with CRM197-2 or CRM197-4 wastested using the standard pneumococcal opsonophagocytic kill-ing assay (OPKA) employing an HL-60 cell line as a phago-cyte source (36). Both CRM197-2 and CRM197-4 conjugatesgenerated opsonizing antibodies that promoted the killing ofpneumococci (Fig. 4 D and E) in an antibody titer-dependentfashion (Fig. 4A). Antibodies induced by monovalent CRM197-2and CRM197-4 were compared with anti-Prevnar13 antibodiesto address the question whether semisynthetic vaccines based onsynthetic glycan antigens can provide a more robust antibodyresponse and protection than traditional CPS conjugates. Anti-bodies induced by both of these monovalent semisynthetic gly-conjugates showed better opsonization than those raised againstthe CPS containing multivalent vaccine Prevnar13. CRM197-2antibodies showed a significantly greater response than CRM197-4antibodies (Fig. 4 D and E), thereby proving the potential of syn-thetic oligosaccharides to generate CPS-specific antibodies. Thus,the above OPKA data should correlate well with an ST-5 challengemodel, based on similar observations in the literature (28, 37).Opsonic activities needed for protection are not governed by titersalone but also by antibody avidity, which describes an entropy-

driven improvement of antibody binding due to multivalency (38).This effect can be probed by binding experiments in the pres-ence of a chaotropic, avidity-disrupting salt such as ammoniumthiocyanate (NH4SCN) (39) (Fig. 4F). To further validate theattributes of antibodies generated using synthetically modifiedglycans, the 50% avidity index (AI) was calculated for sera ofanimals immunized with glycoconjugates CRM197-2, CRM197-4,and Prevnar13 compared with the NH4SCN-free samples againstST-5 CPS (Fig. 4G). Antibodies raised against the CRM197-2conjugate exhibited a relatively high avidity (3.9 M), while thoseraised against CRM197-4 conjugate and Prevnar13 displayedrelatively low avidities (0.85 and 1.4 M, respectively) (36).These observations demonstrate that the monovalent semi-synthetic CRM197-2 is a better glycoconjugate vaccine can-didate than CRM197-4 and the ST-5 CPS glycoconjugatepresent in multivalent Prevnar13. Thus, synthetic ST-5 RU 2was identified as a lead compound for glycoconjugate vaccinedevelopment in preclinical studies.To compare ST-5 semisynthetic glycoconjugate CRM197-2

and ST-5 CPS glycoconjugate present in multivalent Prevnar13,rabbits were immunized with 2.2 μg of the glycan, a doseequivalent to that used in the Prevnar13. The ELISA resultssuggest that the CRM197-2 conjugate produces much higherantibody titers than Prevnar13 (Fig. 5A and SI Appendix, Fig.S6), with significantly higher opsonophagocytic killing proper-ties (Fig. 5B). Prevnar13 containing glycoconjugates of multipleserotypes was utilized in this study as a reference, with the knowl-edge that the presence of additional serotypes can decrease theefficacy of individual serotypes in the vaccine (40, 41).Effective vaccines induce a long-term protective memory

immune response that requires the presence of antibody-producing memory B cells. The memory B-cell response toCRM197-2 was determined over a period of 4 mo using theestablished prime boost regimen (Fig. 5C). Once antibodylevels dropped to baseline, a CRM197-2 booster dose was ad-ministered on day 119. The additional dose recalled antibody

A B

C D

500

1000

2000

4000

8000

1600

032

000

6400

00.00.20.40.60.81.01.21.4

Ig titer

CRM197

CRM197-2Prevnar 13

8 32 128

512

2048

8192

0

20

40

60

80

100

Ig titer

Bac

teria

lkill

ing

(%)

CRM197Prevnar 13CRM197-2 (day 0)CRM197-2 (day 35)

0 7 14 21 28 35 56 70 119126133

0.20.40.60.81.01.21.41.6

Abs

orba

nce

(OD 45

0)

8

32 128

512

2048

8192

0

20

40

60

80

100

Bac

teria

lkill

ing

(%)

Day 0Day 119Day 126

Abs

orba

nce

(OD 45

0)

Fig. 5. CRM197-2 conjugate produces very high opsonic antibody titer at 2.2 μgdose. (A) End point titer (total IgG) analysis performed on sera collected 1 wkafter second booster immunization (day 35). (B) Antibodies promote pneumo-coccal killing by phagocytosis using differentiated HL-60 cells. (C) Antibody re-sponse analyzed by ELISA at day 119, confirming the long-term memory B-cellresponse. (D) Opsonophagocytosis performed with memory response sera andcompared with preimmune sera (day 0) and after booster (day 126). Data arerepresented as mean ± SD, triplicate determinations.

Lisboa et al. PNAS | October 17, 2017 | vol. 114 | no. 42 | 11067

CHEM

ISTR

YMED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

0

levels to day 35 levels as assessed by analysis of CPS-specificantibodies by ELISA and OPKA. Clearly, the semisyntheticST-5 glycoconjugate CRM197-2 induced a robust immuno-logical memory (Fig. 5 C and D).In summary, a medicinal chemistry approach to vaccine dis-

covery yielded a promising candidate to protect from S. pneu-moniae ST-5. While ST-5 CPS is contained in the multivalentvaccine Prevnar13, it suffers from serious production problemsand varying efficacy that is caused by the presence of labilefunctional groups in the CPS. A series of oligosaccharides re-sembling the ST-5 RU were synthesized. Glycan array analysesof human reference sera and immunization experiments inrabbits identified the rare aminosugar L-PneuNAc, as well asbranching, as key to antibody recognition and avidity. Oligosac-charide 2 containing a secondary alcohol in place of the labileketone in native ST-5 CPS 1 resulted in improved antibody titersand opsonic activity compared with Prevnar13 that containsnatural ST-5 CPS. The medicinal chemistry approach allowed forthe identification of key epitopes and the replacement of non-essential, labile entities, that create production problems, withclosely related, stable functional groups. Synthetic chemistry

provides a solution to vaccine manufacturing problems associatedwith S. pneumoniae ST-5 vaccines by increasing stability, im-munogenicity, and protection.

Materials and MethodsOligosaccharide antigens were synthesized using standard protocols andconjugated to CRM197. Synthetic antigens were printed on NHS-activatedmicroarray slides. Immunization (approved by the Landesamt für Land-wirtschaft, Lebensmittelsicherheit und Fischerei Mecklenburg-Vorpommern)was carried out using Zikka rabbits, and the immune response was analyzedby microarrays and ELISA. The functional attribute of the immune responsewas monitored by OPKA using HL-60 cells. Detailed materials and methodscan be found in SI Appendix.

ACKNOWLEDGMENTS. We thank Dr. S. Parameswarappa for helping withthe conjugation chemistry; Dr. A. Calow for MALDI-TOF measurements; Prof.S. Hammerschmidt (University of Greifswald) for providing the S. pneumo-niae ATCC6305 strain; Dr. B. Schumann, T. Monroe, and Dr. A. Berger forcarefully revising the manuscript; and E. Settels and O. Niemeyer for techni-cal support. We acknowledge financial support from the Max Planck Societyand from grants from the German Research Foundation (SFB-TR 84, C3, andC6 to M.W. and C8 to P.H.S.) and from the German Federal Ministry ofEducation and Research (e:Med, CAPSys, FKZ 01ZX1304B/E to M.W.).

1. Anonymous (2009) Closing the GAPP on pneumonia. Nat Rev Microbiol 7:838.2. Schumann B, Anish C, Pereira CL, Seeberger PH (2013) Carbohydrate vaccines.

Biotherapeutics: Recent Developments using Chemical and Molecular Biology (R SocChem, London), pp 68–104.

3. How MJ, Brimacombe JS, Stacey M (1964) The pneumococcal polysaccharides. AdvCarbohydr Chem 19:303–358.

4. Kamerling JP (2000) Pneumococcal polysaccharides: A chemical view. Streptococcuspneumoniae: Molecular Biology andMechanisms of Disease, ed Liebert MA (Larchmont,New York), pp 81–114.

5. Feldman C, Anderson R (2014) Review: Current and new generation pneumococcalvaccines. J Infect 69:309–325.

6. Ginsburg AS, Alderson MR (2011) New conjugate vaccines for the prevention ofpneumococcal disease in developing countries. Drugs Today (Barc) 47:207–214.

7. Johnson HL, et al. (2010) Systematic evaluation of serotypes causing invasive pneu-mococcal disease among children under five: The pneumococcal global serotypeproject. PLoS Med 7:e1000348.

8. Cipollo JF (2009) Review of Manufacturing Process –Prevnar13 [Pneumococcal 13-Valent Conjugate Vaccine (Diphteria CRM197 protein)] (Food Agric Org, Rome).Available at https://www.fda.gov/downloads/BiologicsBloodVaccines/Vaccines/Ap-provedProducts/UCM570727.zip. Accessed September 22, 2017.

9. Sucher AJ, Chahine EB, Nelson M, Sucher BJ (2011) Prevnar 13, the new 13-valentpneumococcal conjugate vaccine. Ann Pharmacother 45:1516–1524.

10. Mistretta N, Danve E, Moreau M (2010) US patent 7,812,006 B2.11. Heidelberger M, Macleod CM, Markowitz H, Roe AS (1950) Improved methods for the

preparation of the specific polysaccharides of pneumococcus. J Exp Med 91:341–349.12. Jansson P-E, Lindberg B, Lindquist U (1985) Structural studies of the capsular poly-

saccharide from Streptococcus pneumoniae type 5. Carbohydr Res 140:101–110.13. Barker S, Bick SM, Brimacombe J, How M, Stacey M (1966) Structural studies on the

capsular polysaccharide of pneumococcus type V. Carbohydr Res 2:224–233.14. Cooper G, Edwards M, Rosenstein C (1929) The separation of types among the

pneumococci hitherto called group IV and the development of therapeutic antise-rums for these types. J Exp Med 49:461–474.

15. Ravenscroft N, Costantino P, Talaga P, Rodriguez R, Egan W (2015) Glycoconjugatevaccines. Vaccine Analysis: Strategies, Principles, and Control (Springer, New York), pp301–381.

16. Pichichero ME (2013) Protein carriers of conjugate vaccines: Characteristics, devel-opment, and clinical trials. Hum Vaccin Immunother 9:2505–2523.

17. Brimacombe JS, Cook MC (1964) 510. Synthesis of 2-amino-2,6-dideoxy-L-mannose(L-rhamnosamine). J Chem Soc 1964:2663–2666.

18. Banaszek A, Karpiesiuk W (1994) Studies of the stereoselective reduction of2-hydroxyimino-hexopyranosides: LiBH4-Me3SiCl, a mild reducing agent of oximesto amines. Carbohydr Res 251:233–242.

19. Panek JS, Zhang J (1993) Diastereoselectivity in the osmium tetraoxide promoteddihydroxylation of chiral (E)-crotylsilane: Asymmetric synthesis of silyl-functionalizedγ-lactones. J Org Chem 58:294–296.

20. Watt GM, Boons G-J (2004) A convergent strategy for the preparation of N-glycancore di-, tri-, and pentasaccharide thioaldoses for the site-specific glycosylation ofpeptides and proteins bearing free cysteines. Carbohydr Res 339:181–193.

21. Veselý J, et al. (2006) Preparation of ethyl 2-azido-2-deoxy-1-thio-β-d-mannopyranosides,and their rearrangement to 2-S-ethyl-2-thio-β-d-mannopyranosylamines. Synthesis2006:699–705.

22. Pereira CL, Geissner A, Anish C, Seeberger PH (2015) Chemical synthesis elucidates theimmunological importance of a pyruvate modification in the capsular polysaccharideof Streptococcus pneumoniae serotype 4. Angew Chem Int Ed Engl 54:10016–10019.

23. Danieli E, et al. (2012) Synthesis of Staphylococcus aureus type 5 capsular poly-saccharide repeating unit using novel L-FucNAc and D-FucNAc synthons and immu-nochemical evaluation. Bioorg Med Chem 20:6403–6415.

24. Gagarinov IA, Fang T, Liu L, Srivastava AD, Boons G-J (2015) Synthesis of Staphylo-coccus aureus type 5 trisaccharide repeating unit: Solving the problem of lactam-ization. Org Lett 17:928–931.

25. Goldblatt D, et al. (2011) Establishment of a new human pneumococcal standardreference serum, 007sp. Clin Vaccine Immunol 18:1728–1736.

26. Wu X, Ling C-C, Bundle DR (2004) A new homobifunctional p-nitro phenyl estercoupling reagent for the preparation of neoglycoproteins. Org Lett 6:4407–4410.

27. Möginger U, et al. (2016) Cross reactive material 197 glycoconjugate vaccines containprivileged conjugation sites. Sci Rep 6:20488.

28. Schumann B, et al. (2017) A semisynthetic Streptococcus pneumoniae serotype 8 gly-coconjugate vaccine. Sci Transl Med 9:eaaf5347.

29. World Health Organization (2007) Meeting Report. WHO Workshop onStandardization of Pneumococcal Opsonophagocytic Assay (World HealthOrg, Geneva).

30. World Health Organization (2013) Recommendations to assure the quality, safetyand efficacy of pneumococcal conjugate vaccines. WHO Expert Committee on Bi-ological Standardization. Sixtieth Report (World Health Org, Geneva), WHO Tech RepSer, No 977, Annex 3, pp 93−151.

31. World Health Organization (2005) Guidelines on Nonclinical Evaluation of Vaccines(World Health Org, Geneva), WHO Tech Rep Ser, No 927, Annex 1.

32. Caulfield MJ, Ahl PL, Blue JT, Cannon JL (2012) US Patent 8,192,746 B2.33. Phalipon A, et al. (2006) Characterization of functional oligosaccharide mimics of the

Shigella flexneri serotype 2a O-antigen: Implications for the development of achemically defined glycoconjugate vaccine. J Immunol 176:1686–1694.

34. Safari D, et al. (2008) Identification of the smallest structure capable of evoking op-sonophagocytic antibodies against Streptococcus pneumoniae type 14. Infect Immun76:4615–4623.

35. Khan N, Qadri RA, Sehgal D (2015) Correlation between in vitro complement de-position and passive mouse protection of anti-pneumococcal surface protein Amonoclonal antibodies. Clin Vaccine Immunol 22:99–107.

36. Birnie GD (1988) The HL60 cell line: A model system for studying human myeloid celldifferentiation. Br J Cancer Suppl 9:41–45.

37. Parameswarappa SG, et al. (2016) A semi-synthetic oligosaccharide conjugate vaccinecandidate confers protection against Streptococcus pneumoniae serotype 3 infection.Cell Chem Biol 23:1407–1416.

38. Usinger WR, Lucas AH (1999) Avidity as a determinant of the protective efficacy ofhuman antibodies to pneumococcal capsular polysaccharides. Infect Immun 67:2366–2370.

39. Romero-Steiner S, et al. (2005) Avidity determinations for Haemophilus influenzae type banti-polyribosylribitol phosphate antibodies. Clin Diagn Lab Immunol 12:1029–1035.

40. Khatun F, Stephenson RJ, Toth I (2017) An overview of structural features of antibacterialglycoconjugate vaccines that influence their immunogenicity. Chemistry 23:4233–4254.

41. Dagan R, Poolman J, Siegrist C-A (2010) Glycoconjugate vaccines and immune in-terference: A review. Vaccine 28:5513–5523.

11068 | www.pnas.org/cgi/doi/10.1073/pnas.1706875114 Lisboa et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

0