Biosensors and Bioelectronics 63 (2015) 80–85

Contents lists available at ScienceDirect

Biosensors and Bioelectronics

http://d0956-56

n CorrE-m

journal homepage: www.elsevier.com/locate/bios

Leukemic marker detection using a spectro-polarimetric surfaceplasmon resonance platform

Mathieu Maisonneuve a, Chiara Valsecchi b, Chen Wang c, Alexandre G. Brolo b,Michel Meunier a,n

a Department of Engineering Physics, École Polytechnique de Montreal, Montreal, QC, Canadab Department of Chemistry, University of Victoria, Victoria, BC, Canadac Department of Pathology and Lab Medicine, Mount Sinai Hospital, University of Toronto, Toronto, ON, Canada

a r t i c l e i n f o

Article history:Received 26 March 2014Received in revised form10 June 2014Accepted 10 June 2014Available online 14 June 2014

Keywords:SPR sensorSpectro-polarimetric imagingAntibodies detectionLeukemia screening

x.doi.org/10.1016/j.bios.2014.06.01863/& 2014 Elsevier B.V. All rights reserved.

esponding author. Tel.: þ1 514 340 4711x497ail address: michel.meunier@polymtl.ca (M. M

a b s t r a c t

In this paper, we present a proof of concept screening for monoclonal immunoglobulin as a leukemiatumor marker using a surface plasmon resonance (SPR) bio-sensing platform. This screening method isbased on measurements of immunoglobulin levels in human serum and the determination of the relativeconcentrations of kappa and lambda light chains. The kappa/lambda ratio is used to determine thepresence of monoclonal immunoglobulin. Tests have been performed using standard solutions ofimmunoglobulins and serum samples from patients with known leukemic diagnoses. This platformhas a resolution of 5�10�7 refractive index unit (RIU) per channel, which is up to 10 times better thanother SPR imaging systems for multi-sensing applications. The results obtained with this technique are inagreement with those acquired using conventional methods for immunoglobulin detection, indicatingthat our polarimetric SPR platform should be suitable for a cheap and efficient tool for early leukemiabiomarker screening and monitoring applications.

& 2014 Elsevier B.V. All rights reserved.

1. Introduction

Leukemic proliferation of B-cell tumors, including plasma cellneoplasms, Waldenström's disease, chronic lymphocytic leukemia(CLL) and lymphomas, is often associated with monoclonal immu-noglobulin. In Canada, about 5800 people are diagnosed and 2600die yearly of leukemic malignancies (Canadian Cancer Statistics,2013). According to the American cancer society (American CancerSociety, 2013), there is no standard screening technique for earlydetection of leukemia which is generally diagnosed by morpholo-gical and phenotypic examinations when clinical symptomsappear. In addition, techniques currently used for leukemia diag-nosis, such as blood cell morphology, flow cytometry or bonemarrow biopsy, are not suitable for screening as they are expen-sive, need a long analysis time and require highly qualifiedprofessionals.

In the last two decades, important studies have been per-formed on serum components, such as immunoglobulin (Bradwellet al., 2013) and their light chain (Campbell et al., 2013; Katzmannet al., 2002; Rajkumar et al., 2005), in a range of plasma cell andB-cell neoplastic disorders in order to discover new biomarkers for

1.eunier).

leukemia diagnosis. In the specific case of CLL, a large proportionof articles in the literature suggests that a comparative analysis ofthe amounts of kappa and lambda light chains should provide astrong indication of the presence of the disease (Maurer et al.,2011; Pratt et al., 2009; Rajkumar et al., 2005).

Immunoglobulin classes and the different light chains, such asIgGκ, IgGλ, IgMk and IgMλ can be identified by specific antibodies.A skewed κ–λ ratio may be found in patients with asymptomaticmonoclonal immunoglobulin. By using the serum protein electro-phoresis method, a κ–λ ratio interval between 0.6 and 4.2 maypredict significant monoclonal components in patients with clin-ical disease (Bergón et al., 2005). More recently, free light chains inserum are used as a sensitive biomarker in screening, monitoringand risk stratifying of patients with myeloma, CLL and B-celllymphomas. Abnormal free light chain ratio has been shown as aprognostic factor for reduced survival of patients with CLL. Thepatients with an abnormal light chain ratio are more likely to havegenetic changes associated with aggressive disease progressionand require early treatment (Charafeddine et al., 2012).

The aim of this study is to propose and validate a cost-effectivemethod for real-time determination of relative concentrations ofdifferent types of immunoglobulins in human serum for leukemiascreening. Concentrations of these markers were monitored usingsurface-plasmon resonance (SPR) bio-sensing. SPR has becomeone of the dominant technologies for bio-detection, as it enables

M. Maisonneuve et al. / Biosensors and Bioelectronics 63 (2015) 80–85 81

real-time and label-free detection of biological species withoutrequiring a large amount of samples (low volume of analyte(Valsecchi and Brolo, 2013)). An SPR-based method also providesgood sensitivity and low limit of detection (Homola, 2006).

The reported SPR biosensor was able to selectively detect lightchain immunoglobulin from both test solutions and samples ofhuman sera obtained from leukemia patients. To assure specificadsorption of the analytes, an existing technique of antibodyimmobilization has been transferred to gold surface (Johnsonand Mutharasan, 2012; Saha et al., 2003).

Finally, this SPR-based biosensing platform should provide acheap and efficient tool for early leukemia screening. This tech-nology should also be useful for monitoring the progression ofleukemia patients during therapy.

2. Principle and instrumentation

2.1. Principle

The principle of leukemic marker detection is summarized inFig. 1. Mature B-cell leukemia, lymphoma, myeloma and otherplasma cell disorders are clonal proliferation of immunoglobulin-producing cells. Monoclonal overexpression results in a dominanttype of immunoglobulin that are composed of a specific heavychain for IgG or IgM (rare occasions IgA, IgD or IgE) and restrictedto either κ or λ light chain, as illustrated in Fig. 1. Therefore, thepresence of monoclonal immunoglobulin can be detected by askewed concentration ratio (relative to their usual distribution inthe blood of healthy individuals) between the kappa and lambdalight chains (Jain and Rai, 2011). This measurement of the κ–λ lightchain ratio in human serum may then be used for leukemiascreening, as well as a monitoring tool for the disease responseduring treatment. The κ–λ light chain ratio for non-leukemicpatients has been determined to fall in a range from 1.4 to 2,whereas ratio for the leukemic patients is greater than 2 for κoverexpressed light chain and much lower than 1 for λ (Katzmannet al., 2002).

2.2. Instrumentation

A biosensor suitable for leukemia screening needs to fulfillseveral requirements. First, a wide dynamic range is necessarysince the light chain concentrations in serum vary extensivelyfrom a patient to another. It ranges from 0.1 to 1 mg/mL, accordingto data acquired with conventional techniques (turbidimetryassays and serum protein electrophoresis) on samples used in thispaper. Taking into account molecular size, surface chemistry,

Fig. 1. Chart illustration of a normal (healthy) person and leukemic patientspresenting over expression of either κ or λ light chains.

volume and surface concentration, the system is required tooperate with a resolution of about 10�6 refractive index unit(RIU), and a dynamic range greater than 10�3 RIU. In order to getreliable and comparable data, biomarkers analysis must be per-formed with the exact same experimental condition. To answerthis need, the system has to include a multi-sensing scheme tosimultaneously measure the κ–λ light chain ratio in a sample ofhuman blood. The polarimetric method using spectral polarizationreading has been chosen as it provides good resolution and widedynamic range. Fig. 2 shows a schematic of the apparatus. It isbased on the measurement of p-polarized reflected light, affectedby the SPR coupling (Raether, 1988), referenced to s-polarizedlight, which is not affected by SPR, as seen in the experimentalcurve in Fig. 2 inset.

The ratio of these polarizations gives the tangent value of thepolarization vector (Hecht, 2002). The spectral position of theresonance, defined by the minimum of the measured reflectivitycurve (see Fig. 2), is tracked with a polynomial interpolation androots calculation. The details of the optical setup for the SPRsystem used in these experiments are presented in Fig. 2a.A white-light beam incoming from a LED (Thorlabs warm white450–750 nm – MWWHL3) was focused in a multimode fiber(Thorlabs MM core 105 μm – M15L01). Light, at the output of thefiber, was then brought to an achromatic lens (L1) and the beamwas collimated before passing through a plate polarizer. Thisparallel beam was directed to the sensing block, which consistsof a microscope plate covered with a 50 nm gold layer in contactwith the tested solution on one side (two channels – the flow cellback view is also presented in Fig. 2a) and with oil immerse BK7prism on the other side for coupling condition. Reflected lightpassed through an analyzer before going to the spectro-imagingsystem. A CCD camera (Andor Newton 971) acquired the lightdispersed from the monochromator and the generated image wasthen processed to get useful spectral information from the twochannels.

Finally, characterization tests have been realized on the systemto determine its performance. The resolution was determined byusing different solutions of ethanol in water with concentrationsof 0.1, 0.2 and 0.5 vol% as shown in Fig. 2b. Using the refractiveindex changes of the ethanol solutions (Lide, 2005) and themeasured signal-to-noise ratio, the setup resolution was calcu-lated to be 5�10�7 RIU, which is better than the neededresolution for leukemia screening. In terms of dynamic range,the spectral window offered by the instrument is about 44 nmwhich is equivalent to a change of 1.5�10�3 RIU and thusrespecting the needed specifications.

3. Materials and methods

3.1. Materials and samples

All the tests have been performed in 10 mM Phosphate BufferSaline (pH 7.4 PBS 10� , Sigma-Aldrich P5493 diluted 10 times in18 MΩ cm deionized water). Lyophilized recombinant Protein G(Sigma-Aldrich P4689) from E. coli (MW�20 kDa) was used in allthe experiments as well as Bovine Serum Albumin (BSA) fromSigma-Aldrich (A7906). Goat polyclonal anti-human kappa(Sigma-Aldrich K3502) and goat polyclonal anti-human lambda(Sigma-Aldrich L1645) were used respectively to bind to kappaand lambda light chains.

Patient serum samples were collected at the Mount-SinaïHospital, Toronto. The 4 serum samples in this study included anormal control and 3 samples from patients presenting an over-expression of a monoclonal type of immunoglobulin (G-kappa,M-lambda and G-lambda respectively). Total concentrations of

a b

Fig. 2. (a) Optical setup used in the experiments; inset: spectral curve acquired on p-polarization (black continuous), s-polarization (red continuous) and the ratio betweenthe p and the s polarizations (black dashed). (b) Characterization test using sequentially a 0.1, 0.2 and 0.5 vol% ethanol solution. (For interpretation of the references to colorin this figure legend, the reader is referred to the web version of this article.)

Table 1Monoclonal and polyclonal immunoglobulins in serum samples of normal controland patients.

Samples Monoclonalconcentration(μg/ml)

Polyclonalconcentration(μg/ml)

Estimatedratio κ:λ

Error of theestimatedratio

Control 0 120 1.70 0.03G-kappa 120 70 6.24 0.42M-lambda 120 35 0.164 0.013G-lambda 120 200 0.641 0.004

M. Maisonneuve et al. / Biosensors and Bioelectronics 63 (2015) 80–8582

immunoglobulin, shown in Table 1, have been obtained fromindependent turbidimetric assays. Monoclonal concentrationswere determined by serum protein electrophoresis (Dispenzieriet al., 2009).

The 4 tested serum samples were then diluted to get similareffective concentrations where resulting concentrations are shownin Table 1. Total concentration of immunoglobulin in blood serumwas defined by the addition of polyclonal (normal part of Ig) andmonoclonal (overexpressed Ig related to B-cell disorder) concen-trations. An estimation of the range of kappa and lambda con-centrations in human sera was extracted and their ratio ispresented in Table 1. This estimation was related to the fact thatthe κ–λ ratio in the polyclonal part ranged from 1.4 to 2 and led toerrors shown in Table 1.

3.2. Surface chemistry

Our method for leukemia screening is based on SPR. Theselectivity of an SPR experiment is provided by the proper surfacechemistry. The immobilization of species on the gold surface thatspecifically captures the biological molecules of interest from thesolution side is then an important aspect of the sensor develop-ment. Antibodies specific to human kappa or lambda light chainswere immobilized to the gold surface by direct attachment to aprotein G layer. Protein G specifically binds to the crystallizablefragment (Fc) of an antibody, which allows the fragment antigen-binding site (Fab) to be available for binding (Goward et al., 1990).This surface modification method presents some advantages overother techniques, such as the NHS/EDC chemistry. In that case,antibodies are randomly attached on the surface, which couldpossibly block their active binding sites. Moreover, the antibodiesused in our experiments were from goat, and protein G is moresuitable than other similar proteins (like protein A) because it has

a better affinity for this species (Millipore Technical Library, 2013;Piercenet Tech Tips, 2013). Also, according to Johnson andMutharasan (2012), protein G has other advantages, one of whichis its pH-dependent orientation which will maximize the bindingcapacity of antibodies at pH 7.2 because it makes the active region(Fab) available to the tested solution. Another advantage of theprotein G is its good affinity to gold surfaces (Johnson andMutharasan, 2012; Saha et al., 2003), due to a chemisorptionbinding mechanism which occurs at room temperature involvingAu–N interactions (Johnson and Mutharasan, 2012).

3.3. Experimental protocol

First, microscope glass slides were cleaned by using succes-sively acetone and isopropanol (IPA) bath and were dried with N2.Glass slides were then immersed in a PIRANHA (H2SO4:H2O2)solution for 15 min to remove all remaining organic compounds,followed by rinsing with water and drying with N2. Afterwards,a chromium adhesion layer (2 nm) and a gold layer (50 nm) wereevaporated onto the glass substrate using an electron beam(e-beam) evaporator. The layer thicknesses were verified by usinga quartz micro-balance and an ellipsometer. Deposition of aphotoresist was done before dicing the microscope glass slide intosquare samples of 1.25�1.25 cm2 to protect against the depositionof dust and organic particles on the gold surface. The photoresistwas then removed by successive baths of acetone, IPA and waterfor 10 min in a sonicator. Samples were then washed by usingoxygen plasma for 5 min (50 sccm O2) to remove remainingphotoresist, dust and organic particles.

SPR chips were then placed onto a fluidic cell (made inPlexiglas) with two channels and liquid were brought using aperistaltic pump with a flow rate per channel of 5 μL/min. Thesurface functionalization was then performed to make the goldsurface specific to human kappa or lambda light chain, as sche-matically illustrated in Fig. 3. Initially, the gold surface wasexposed to PBS with a flow rate of 5 μL/min for 20 min. Then asolution of protein G in PBS (50 μg/ml) was introduced in thefluidic cell for 20 min at the same flow rate (step 1 in Fig. 3).

As it was mentioned previously, there is no need to use anadhesion layer since protein G spontaneously chemisorb to thegold surface through their amine groups. Following a 10 minrinsing step, a solution of BSA (1 mg/mL) was then flown intothe fluidic cell to fill empty sites on the surface and avoid non-specific adsorption (step 2 in Fig. 3). Kinetics of these two stepscould be observed in Fig. 4a. Then, as shown on Fig. 4b, a new

M. Maisonneuve et al. / Biosensors and Bioelectronics 63 (2015) 80–85 83

baseline was performed for 20 min before passing a solution ofanti-human kappa and lambda respectively on two separatedchannels for 5 min. The pump was then stopped, leaving thesurface in contact with the antibody solution for 35 min. PBS wasflown again to rinse the antibody solution from the surface (step3 in Fig. 3).

Finally, a solution of human serum or pure immunoglobulin inPBS was injected in the two flowing channels to measure the

Fig. 3. Schematic illustration of the experimental protocol.

a b

c d

Fig. 4. Surface functionalization steps. (a) Adsorption of protein G and BSA. Steps 1 anddata of control sample (normal case – step 4 in Fig. 3) of human serum in contact to botstep in Fig. 3) with the G-λ human serum sample (leukemia patient, see Table 1). (For inthe web version of this article.)

relative concentration between kappa and lambda light chain(step 4 in Fig. 3). In order to prevent incorrect data processing,the sensorgrams were scaled respectively to the amount of surfacereceptors for both channels (SPR shift for anti-κ and anti-λadsorption on protein G in Fig. 4b).

The sensorgrams shown in Fig. 4c and d are examples of typicalresults obtained with human serum samples. In these figures, theblack lines represent adsorption on surfaces that are specific tolambda light chain and the red lines represent the adsorption onsurfaces specific to kappa light chain.

The κ–λ ratio obtained in Fig. 4c by SPR dynamic experimentswith the control sample was equal to 1.45 (compared to a ratio of1.7, quoted in Table 1, obtained with conventional analyticalmethods for the determination of immunoglobulin). Fig. 4d showsresults obtained with the G-λ sample (information about thesesamples is given in Table 1), where the κ–λ ratio evaluated by SPRwas close to 0.65 (compared to 0.641 in Table 1).

4. Results

4.1. Validation

Surface validation tests were primarily done to verify the anti-kappa and anti-lambda surface binding capacity and specificity.Binding capacity was quantified by looking at the SPR responsewith pure κ and λ light chain immunoglobulin solutions in PBS,with concentrations ranging from 100 ng/ml to 100 μg/ml. Resultsof this test revealed that a linear SPR response was achieved forconcentrations varying from 1 μg/ml to 10 μg/ml. This narrow

2 in Fig. 3. (b) Antibodies adsorption on protein G (step 3 in Fig. 3). (c) SPR kineticsh anti-kappa- and anti-lambda-modified surfaces. (d) Same experiments (the sameterpretation of the references to color in this figure legend, the reader is referred to

a b

Fig. 5. (a) Surface validation results with tested standard κ and λ light chain immunoglobulin on respectively anti-κ and anti-λ surfaces. (b) κ–λ ratios – in blue: ratiosobtained by the standard analytical method for the determination of immunoglobulin (used in hospital); – in red: κ–λ ratios obtained with the proposed SPR method. (Forinterpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

M. Maisonneuve et al. / Biosensors and Bioelectronics 63 (2015) 80–8584

range of concentration can be explained by the surface chemistry,which limits the amount of bound light chains immunoglobulin(surface saturation). The surface specificity was then tested.2.4 μg/ml solutions (in the linear range of the SPR response) ofpure kappa or lambda light chain immunoglobulins (in PBS) werealternatively injected on both anti-kappa and anti-lambda chan-nels of the SPR system to simultaneously measure specific andnon-specific responses. For each solution, the tests were per-formed 3 times to get reliable data. The results are presented inFig. 5a. In order to make the data set from the two channelscomparable, the specific adsorptions were set to 1 and, as shownin Fig. 5a, the functionalized surfaces provide rather good speci-ficity, with a specific signal up to 3 times greater than the non-specific signal. The non-specific adsorption could be explained bytwo different mechanisms: the physisorption of immunoglobulinon empty sites of the gold surface and/or the direct adsorption ofimmunoglobulin on free protein G sites (without anti-lambda oranti-kappa antibodies). In any case, non-specific adsorption is acommon limitation in biosensors. Although our surface chemistrydid not eliminate this problem, we were able to keep it to arelative low level.

4.2. Screening test with patient's samples

Finally, tests using real human sera have been performed tovalidate the reliability of the proposed screening method. Samplesused for the experiments are the ones described in Section 3.1(Table 1), but diluted 100 times to reach the linear range of the SPRinstrument. Examples of dynamic data are shown in Fig. 4c and d,where the tested samples are the control (ratio estimated at 1.7) inFig. 4c and the G-λ sample (ratio�0.65) in Fig. 4d. Tests on eachhuman sera sample have been done three times to validate theobtained data and to make statistical significant measurements ofthe ratios.

The results obtained are illustrated in Fig. 5b and are alsopresented in Supplementary data (Table). It could be noted thatthe ratios measured by SPR follow a similar trend than the onesobtained with a standard analytical technique for immunog-lobulin determination. Moreover, error bars for the measuredratios show that there was a low dispersion of the data, whichindicates that the proposed technique is quite reliable for the ratiomeasurements.

An under-estimated ratio for the M-lambda and G-kappapatients could be observed in Fig. 5b. This is explained by eitherthe fact that these samples had the highest concentration of

immunoglobulin leading to surface saturation and, as it wasnoticed in the surface validation part, by non-specific adsorptionlimiting the linear range of the sensor. Nevertheless, even if aperfect quantitative match was not yet achieved, these resultsshowed that SPR was able to identify leukemic patients.

5. Discussion

As it could be observed in both Fig. 5 and results presented inSupplementary data, the SPR system offered good accuracy whilestill suffered from non-specific binding (up to 40% of the signal inFig. 5a). This issue led to either an overestimation of the measuredratio (for those less than 1) or to an underestimation (for thosegreater than 2). However, since the threshold for diagnostic isdefined by a ratio, these errors have not altered the resultssignificantly. In that sense, the SPR platform in its current formcould already be used as a front line screening method.

The sources of this non-specificity arise mainly from theadsorption of different kinds of protein in the serum on the goldsurface, and as well from non-specific binding of Ig onto thesurface by either binding directly on gold or on protein G whichdid not catch anti-kappa/lambda antibodies.

Investigations are underway to improve those measurementsby working on both surface chemistry and measurements strategy,which would consist in simultaneous measurement of the rawratio and non-specific adsorption in order to correct in real timethe measured ratio.

Validation of a diagnostic technique is usually done usinglikelihood ratio (LR) testing based on the statistics of the false/true positive/negative. In our case, we did not notice any falsepositive nor negative, which makes an evaluation difficult for thesensitivity and specificity as well for the likelihood ratio. However,since the presented results are quite in good agreement withclinical ones, we can estimate that the likelihood ratios should beclose to those recommended for a conventional diagnostic tech-nique, i.e. LRþ410 and LR�o0.1.

6. Conclusion

In this paper, we have presented a proof-of-concept mono-clonal immunoglobulin screening method using the relative con-centration ratios between kappa and lambda light chains. Theproposed optical setup coupled with proper surface chemistry has

M. Maisonneuve et al. / Biosensors and Bioelectronics 63 (2015) 80–85 85

shown that measured ratios correspond to the ones measuredwith a conventional technique used in diagnostic laboratories.Although further improvement in the surface chemistry couldprovide a better measurement of ratios (mainly the ones with veryhigh over expression), the results presented here clearly demon-strated the potential of polarimetric SPR as a relative low cost toolfor fast disease diagnostic.

Acknowledgments

The authors acknowledge the financial contribution from theNSERC Strategic Network for Bioplasmonic Systems (Biopsys) andNanoQuébec. They also acknowledge the fruitful discussion withDr. Sergiy Patskovsky and the team members of Pr. Meunier group(André-Pierre Blanchard-Dionne, Anne-Marie Dallaire and LaurentDoré-Mathieu).

Appendix A. Supplementary information

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.bios.2014.06.018.

References

American Cancer Society, 2013. Leukemia-Chronic Lymphocytic. American cancersociety.

Bergón, E., Miravalles, E., Bergón, E., Miranda, I., Bergón, M., 2005. Clinical ChemicalLaboratory Medicine p. 32.

Bradwell, A., Harding, S., Fourrier, N., Mathiot, C., Attal, M., Moreau, P., Harousseau,J.L., Avet-Loiseau, H., 2013. Leukemia 27 (1), 202–207.

Campbell, J.P., Cobbold, M., Wang, Y., Goodall, M., Bonney, S.L., Chamba, A.,Birtwistle, J., Plant, T., Afzal, Z., Jefferis, R., Drayson, M.T., 2013. J. Immunol.Methods 391 (1-2), 1–13.

Canadian Cancer Statistics, 2013. Canadian Cancer Society’s Advisory Committee onCancer Statistics. Canadian Cancer Society, Toronto, ON.

Charafeddine, K.M., Jabbour, M.N., Kadi, R.H., Daher, R.T., 2012. Am. J. Clin. Pathol.137 (6), 890–897.

Dispenzieri, A., Kyle, R., Merlini, G., Miguel, J.S., Ludwig, H., Hajek, R., Palumbo, A.,Jagannath, S., Blade, J., Lonial, S., Dimopoulos, M., Comenzo, R., Einsele, H.,Barlogie, B., Anderson, K., Gertz, M., Harousseau, J.L., Attal, M., Tosi, P.,Sonneveld, P., Boccadoro, M., Morgan, G., Richardson, P., Sezer, O., Mateos, M.V., Cavo, M., Joshua, D., Turesson, I., Chen, W., Shimizu, K., Powles, R., Rajkumar,S.V., Durie, B.G., 2009. Leukemia 23 (2), 215–224.

Goward, C.R., Murphy, J.P., Atkinson, T., Barstow, D.A., 1990. Biochem. J. 267 (1),171–177.

Hecht, E., 2002. Optics, 4th ed. Addison-Wesley, San Francisco, CA, USA, pp. 325–384.

Homola, J., 2006. Surface Plasmon Resonance Based Sensors. Springer, New York,NY, USA.

Jain, P., Rai, K.R., 2011. Leukemia. Demos Medical Publishing, New York, NY, USA.Johnson, B.N., Mutharasan, R., 2012. Langmuir 28 (17), 6928–6934.Katzmann, J.A., Clark, R.J., Abraham, R.S., Bryant, S., Lymp, J.F., Bradwell, A.R., Kyle, R.

A., 2002. Clin. Chem. 48 (9), 1437–1444.Lide, D.R., (Ed), 2005. Handbook of Chemistry and Physics. CRC Press, Boca Raton,

FL, pp. 8.58–58.84.Maurer, M.J., Cerhan, J.R., Katzmann, J.A., Link, B.K., Allmer, C., Zent, C.S., Call, T.G.,

Rabe, K.G., Hanson, C.A., Kay, N.E., Slager, S.L., Witzig, T.E., Shanafelt, T.D., 2011.Blood 118 (10), 2821–2826.

Millipore Technical Library, 2013. Binding properties of protein A&G for immuno-globulins. Millipore Technical Library, pp. 1–2.

Piercenet Tech Tips, 2013. Pierce ™ Ig Binding Proteins. Piercenet tech tips.Pratt, G., Harding, S., Holder, R., Fegan, C., Pepper, C., Oscier, D., Gardiner, A.,

Bradwell, A.R., Mead, G., 2009. Br. J. Haematol. 144 (2), 217–222.Raether, H., 1988. Surface Plasmons on Smooth and Rough Surfaces and on

Gratings. Springer.Rajkumar, S.V., Kyle, R.A., Therneau, T.M., Melton, L.J., Bradwell, A.R., Clark, R.J.,

Larson, D.R., Plevak, M.F., Dispenzieri, A., Katzmann, J.A., 2005. Blood 106 (3),812–817.

Saha, K., Bender, F., Gizeli, E., 2003. Anal. Chem. 75 (4), 835–842.Valsecchi, C., Brolo, A.G., 2013. Langmuir 29 (19), 5638–5649.