Evaluation of Surface Plasmon Resonance (SPR) for Heparin Assay

JOURNAL OF COLLOID AND INTERFACE SCIENCE ARTICLE NO.

194, 373–378 (1997)

CS975098

Evaluation of Surface Plasmon Resonance (SPR) for Heparin Assay II. Polyethylene Imine (PEI) as an Affinity Surface Katharina Gaus and Elizabeth A. H. Hall 1 Institute of Biotechnology, University of Cambridge, Tennis Court Road, Cambridge CB2 1QT, United Kingdom Received April 21, 1997; accepted July 24, 1997

Heparin adsorption onto polyethylene imine (PEI) films was investigated. The adsorption was followed with time-dependent surface plasmon resonance (SPR) for PEI cast on gold films, and the thickness of the adsorbed heparin was determined by fitting the angular reflection curves to Fresnel’s equations. The average thickness of the PEI determined the heparin adsorption range and it was found that both heparin binding capacity and sensitivity increased with PEI thickness. At an appropriate thickness of 10 { 2 nm, adsorption of heparin in the clinically applicable range of 0–2 U/ml caused a linear concentration dependent change in the SPR reflectivity at a fixed angle, with excellent sensitivity (0.05 U/ml). The results are compared to heparin adsorption on immobilized protamine and the surface coverage and probable heparin distribution discussed. q 1997 Academic Press Key Words: heparin; SPR; biosensor; polyethylene imine.

INTRODUCTION

Heparin is clinically used as an anticoagulant, as it acts as a cofactor of antithrombin III (1–3). The treatment with heparin is monitored by an anti-clotting test which measures a parameter related to heparin activity (4–6). However, the heparin activity varies with the reagent used in these tests and also depends on the antithrombin III concentration in the blood samples (7–10). Additionally, it has been found that according to these tests heparin can have a higher activity after injection and subsequent separation from the blood than the original heparin! This is evidence of a significant difference between in vivo heparin levels and the anticoagulant activity in vitro as estimated by clotting type assays (8, 11, 12). It was therefore suggested that the absolute concentration of heparin should be monitored with a direct biosensor to overcome the dependency on antithrombin III and the assay reagents and to limit the differences of in vivo and in vitro heparin estimates (13). Heparin is a highly negatively charged glycosaminoglycan 1

To whom correspondence should be addressed.

with an ellipsoidal rod configuration exposing fully ionized carboxyl and sulfate groups along its length. It does not have a single structure, but is regarded as a heterogeneous family of fragments with molecular weights between 5000 and 30,000. In choosing an affinity surface for heparin, therefore, which can be used in a biosensor, the focus is toward a cationic ligand/surface able to accommodate the structural heterogeneity of heparin. We have previously reported the investigation of a heparin biosensor based on surface plasmon resonance (SPR) with protamine as an affinity surface (14). The heparin binding to protamine occurred specifically and the sensitivity could be modulated by the density and method of immobilization of the protamine underlayer. The greatest sensitivity for low concentrations of heparin was shown when the protamine layer was incomplete (the optical thickness was small), thus presenting an affinity surface with a large 3-D structure to the incoming heparin. However, protamine layers of this form may be difficult to control in a manufacturing environment and therefore limit the reproducibility of such a sensor, so other materials with similar properties and charge distribution as displayed by protamine but which would allow stable and reproducible immobilization on gold SPR surfaces would be preferred. Polyethylene imine (PEI) is used as an anion exchange polymer in high-performance chromatography (15–17). After polymerization, PEI is a basic polymer due to secondary amine groups. When placed on a surface, some portion of each PEI chain is ‘‘dissolved out’’ into the water phase (when in contact with water) and moves freely on the polymer surface. The ionic interaction sites of the PEI chains can therefore change according to the shape of the incoming protein or the analyte or the position of their ionic sites. Such dynamic surface arrangements without a clear boundary at the solid (polymer)/liquid (water) interface have typically a very low free energy, have a high hydrophilicity, and provide a high ‘‘active binding capacity.’’ The binding capacity for PEI is reported to be between 35 and 50 mg protein/ml gel (17).

373

AID

JCIS 5098

/

6g32$$$421

10-22-97 12:49:20

0021-9797/97 $25.00 Copyright q 1997 by Academic Press All rights of reproduction in any form reserved.

coida

374

GAUS AND HALL

The flexibility of the long PEI chain and its charge appears sufficient to enable heparin molecules to bind in spite of their structural heterogeneity. The electrostatic nature of the interaction between PEI and heparin should ensure that the heparin adsorption is independent of the heparin activity. PEI also has a different refractive index (nPEI Å 1.5290) in aqueous solution from heparin (nheparin Å 1.5) so that changes in both thickness and refractive index would be reflected in the SPR signal. This would therefore be expected to improve the signal resolution so that the absolute heparin concentration could be determined with good accuracy. This paper assesses PEI as an affinity ligand for heparin and compares the sensitivity of the SPR signal with that obtained previously with a protamine affinity ligand. MATERIALS AND METHODS

Heparin, sodium salt, 157 U /mg, was purchased from Sigma (Dorset, UK ) as well as polyethylene imine, MW 50,000. Polyethylene imine, MW 25,000, and ( 3-mercaptoproyl )trimethoxysilane were obtained from Aldrich ( Dorset, UK) . Pure gold ( 99.999%) was obtained from Alfa, Johnson Matthey ( Royston, UK ) . All experiments in buffer solution were performed in 0.05 M phosphate buffer at pH 7.4 with 0.15 M NaCl added to achieve a physiological ionic strength. Small samples of the phosphate buffer containing 100 U /ml heparin were diluted to the appropriate heparin concentration. Since heparin quantities are typically given in units (U) we follow this and give concentrations in U/ml. This refers to the absolute concentration and not to the heparin activity. The SPR configuration and system have been reported previously (18).

ml of a 5–20% (v/v) solution in 80:20% (v/v) H2O:acetone mixture. The slides were then placed in an oven at 357C for 15 min. For the spin-coating of the PEI film (MW 50 and 25 kDa), the PEI in aqueous solution was diluted with methanol to achieve PEI concentration between 5 and 25%. The volume of 25 ml and the same rotation velocity and time as above was used. The slides were again placed in the oven for 30 min at 357C. The thicknesses of the films were characterized by taking the angular reflection curve in phosphate buffer before and after the PEI immobilisation. The stability of the PEI films were determined by time dependent SPR for 30 min. The heparin adsorption in phosphate buffer or plasma was also followed by SPR and characterized by taking and fitting angular reflection curves before and after the adsorption. All reflection curves were fitted with a nonlinear least square fit to Fresnel’s equation (18) . It was assumed that the refractive index of the PEI film is n Å 1.5290 and of heparin is n Å 1.5. RESULTS AND DISCUSSION

The PEI Immobilization In order to determine the thickness of the PEI films, angular reflection curves were taken and fitted before and after the PEI immobilization. However, evaporated gold films of Ç55 nm have an inherent roughness, causing the SPR response to vary over the surface and thus introducing relatively large errors in the subsequent estimation of the thickness of the PEI films. For spin-coating 5 and 10% PEI solutions onto gold surfaces modified with glutaraldehyde, thicknesses of 4 { 1 nm and 10 { 2 nm were measured, respectively (Fig. 1). It has to be emphasized that these

Measurement Protocol PEI was immobilized on gold films which were coupled via (3-mercaptopropyl)trimethoxysilane (MPS) on the glass slides (14). To couple the gold to the glass the slides were stored in 0.5% (v/v) MPS solution in 95:5 (v/v) methanol:water for more than 24 h, subsequently sonicated in ethanol for 2 min, and stored at 1507C for at least 3 h. The metal was evaporated through a mask in the form of dots with 5 mm diameter, and the functional group of the silane (the thiol groups) was still accessible around the metal islands for ‘‘anchoring’’ of the PEI layer. For the immobilization of PEI, a glutaraldehyde layer was spin-coated onto the gold patterned glass slides. Glutaraldehyde may have two functions; it is likely to cross link the polymer via its secondary amine group (19); second, it may interact with the thiol groups of the silane (MPS) which are accessible at the edges of the gold island and therefore provide a link between the glass slide and the PEI film. The glutaraldehyde was spin coated at 2000 rpm for 20 s using 20

AID

JCIS 5098

/

6g32$$$422

10-22-97 12:49:20

FIG. 1. Angular reflection curves (scatter) for (a) plain gold surface, (b) after spin coating 25 ml of a 5% PEI solution, and (c) after spin coating a 10% PEI solution. The solid lines represent theoretical curves calculated for (a) a 49-nm gold layer, (b) an additional overlayer of 4.9 nm ( nPEI Å 1.290), and (c) an 11.2-nm-thick overlayer.

coida

AN SPR BASED HEPARIN BIOSENSOR II

FIG. 2. The behavior of two polymer films with an approximate thicknesses of 10 nm during five washing steps with phosphate buffer indicated by arrows. The molecular weights of the polymer were (a) 25 kDa and (b) 50 kDa.

thicknesses are average thicknesses as determined by optical thickness and thus refer to coverage. The stability of the films was tested by fixing an angle on the right-hand side of the angular reflection curve so that a change in the signal would indicate that modification of the film had occurred. Figure 2 shows the stability of two different films of PEI with different molecular weights during a washing procedure with phosphate buffer and suggests that the lower molecular weight (MW Å 25,000) produces a less robust layer, as shown by the loss in signal (this change in intensity could correlate with a shift in the position of the reflectance minimum to the left-hand side, which has to be interpreted as a decrease of the thickness of the polymer film or partial dissolution of the film). This behavior was found for all films of this molecular weight, irrespective of film thickness, and could not be overcome by increasing the glutaraldehyde cross-linking. On the other hand, for higher molecular weight (MW 50,000) the films were stable to repeated washing and for all the following experiments PEI with MW 50,000 was used.

375

FIG. 3. Heparin adsorption to a 10 { 2-nm-thick PEI film. Heparin concentrations between 0.5 and 10 U/ml were injected for 30 s and followed by a buffer wash. Since the heparin containing solution has a different bulk refractive index to the buffer, the signal measured includes both this ‘‘offset’’ and the binding information so the data obtained in buffer should be compared.

Quantitative evaluation. The response to heparin was determined for either increasing heparin concentrations, with alternating washing steps ( using the same polymer) , as shown in Fig. 3, or for the same heparin concentration for several different films. The agreement of the two data sets are within a 5% error, indicating that the change in reflectivity measured during heparin adsorption is concentration dependent. Figures 4 and 5 show the calibration curve of the heparin adsorption for two different average thicknesses of the polymer film: 10 and 4 nm, respectively. By comparing the two graphs it can be seen that with increasing thickness of the PEI film both the binding capacity and the sensitivity increased. For the thinner PEI layer no response could be seen below 0.1 U/ml, whereas for the thicker PEI layer a response could be seen at 0.05 U/ml in fixed angle time-dependent SPR. This suggests that polymer layers with higher average

Heparin Adsorption onto PEI Films in a Steady State Adsorption Qualitative evaluation. The adsorption of heparin to PEI was examined in phosphate buffer at physiological pH and ionic strength. After the characterization of the polymer film, 1 ml of a phosphate buffer containing heparin was injected and the adsorption process was followed by SPR. Figure 3 shows the typical step-like profile of the adsorption which occurs at all heparin concentrations, suggesting that the heparin adsorption to PEI is almost instantaneous. This interesting behavior was also seen when protamine was used as the affinity surface instead of PEI (14), suggesting a similar charge density interaction with protamine as with PEI. The adsorption of heparin binding also appears to be irreversible and stable for at least 20 min so that the results compare well with those previously reported for protamine.

AID

JCIS 5098

/

6g32$$$422

10-22-97 12:49:20

FIG. 4. Thickness of the adsorbed heparin versus heparin concentration in solution. The heparin adsorbs onto a 10 { 2-nm-thick polymer film immobilized onto a SPR device. A relative error of 10% is calculated from at least five measurements.

coida

376

GAUS AND HALL

FIG. 5. Thickness of the adsorbed heparin versus heparin concentration in solution. The heparin adsorption occurs onto a 4 { 1-nm-thick polymer film immobilized onto a SPR device. A relative error of 12% is calculated from at least four measurements.

thicknesses have a greater available cationic surface. Of particular use for the determination of heparin in a clinically useful range, the results for a 10 { 2-nm-thick polymer film show that the heparin adsorption is linear within the clinical concentration range of 0–2 U/ml. To calculate the coverage of the polymer surface by heparin two orientations of the heparin molecules have to be considered. Heparin is a rod-like molecule with an average length of 15 nm and a width of 1.6 1 1.7 nm (20). Vigil and Ziff calculated the maximum surface coverage of randomly adsorbed rods depending on the length/diameter ratio of the rods. Neglecting the repulsive forces due to the charge of heparin, an unordered adsorption of heparin on a planar surface should achieve a maximum £55% coverage (21). In the extreme case of all the heparin molecules being oriented perpendicular to the surface, a 55% coverage corresponds to an average heparin layer thickness of 8.4 nm, which is approximately the maximum obtained for a 10-nm PEI layer, as shown in Fig. 4. On the other hand, a thinner film fails to accommodate the same amount (e.g., in Fig. 5 a maximum thickness of 5.9 nm is determined which corresponds to a 39% coverage for a purely perpendicular orientation; a parallel orientation on a planar interface can be excluded since it corresponds to a multilayer of three to five layers). It is clear, however, that these results cannot necessarily be treated as interaction between planar distinct layers of PEI and heparin, but that penetration of the PEI by heparin and mixing of the layers occur. Comparable heparin concentrations using protamine as an affinity surface caused a much lower maximum surface coverage ( õ10% for a perpendicular orientation and a coverage õ80% for a parallel orientation). However, in contrast with the protamine affinity layer, the heparin sensitivity here improved with PEI thickness, suggesting that unlike protamine where significant mixing of the heparin/protamine layers was proposed only when the protamine layer was

AID

JCIS 5098

/

6g32$$$422

10-22-97 12:49:20

discontinuous and the electrostatic interactions between heparin and the cationic amino acid groups of the protamine could be maximized, even ‘‘thick’’ PEI layers have a sufficiently ‘porous’ distribution of PEI chains so that the layer volume (not just the surface) is available for interaction with heparin, resulting in higher binding capacity and a higher sensitivity with thickness. This ‘‘porosity’’ seems reasonable since electrostatic repulsion between the adjacent PEI chains will limit the chain-packing density. If these conclusions are correct then PEI layer is a permeable diffusion barrier to the incoming heparin, and the interaction between the PEI and heparin may involve a fast initial reaction with the surface of the PEI, followed by a slower penetration and interaction with the ‘‘interior.’’ In order to try to deconvolve this effect, the heparin adsorption protocol was changed so that the heparin sample remained in contact with the PEI modified SPR surface for less than 2 s and was displaced immediately by a buffer injection. (Figure 6 shows the typical binding process for different concentrations of heparin using this procedure.) Comparing Fig. 6 with Fig. 2, the former should allow the heparin adsorption process to be controlled mainly by convection and the addition of buffer no longer has an instantaneous effect on the measured signal (attributable to just a bulk dielectric change) but there is a comparatively slow ( Ç10–15 s) equilibration time, consistent with a surface desorption process. Although in such an apparently ‘‘fast’’ binding process it is difficult to resolve these effects fully, heparin at the surface may be bound first reversibly and fast, whereas in a second stage, that which has moved internally, it is irreversibly bound; it would thus be anticipated that the amount of irreversibly bound heparin would increase with exposure time to heparin (compare Fig. 2 and Fig. 6). The calibration curve given in Fig. 7 is also different from that for the equilibrium adsorption process (Fig. 4), suggesting that the heparin binding kinetics are not complete in the 2-s time frame, at least at higher concentrations of heparin. At low concentrations ( õ1.5 U/ml) there is a rea-

FIG. 6. Heparin adsorption onto Ç10 nm PEI layer with an incubation time below 2 s. The heparin sample was immediately displaced by buffer solution.

coida

377

AN SPR BASED HEPARIN BIOSENSOR II

necessary millisecond time resolution could not be achieved in the SPR system. Heparin does not show an affinity for either a plain gold surface or a gold surface modified with glutaraldehyde. In practical terms, however, for an SPR device for heparin for the clinically relevant range, the static system not only requires a simpler control mechanism, but also gives a greater sensitivity. CONCLUSION

FIG. 7. Thickness of adsorbed heparin versus heparin concentration. The incubation time of heparin was below 2 s and the adsorption was therefore mainly controlled by convection. Each data point is an average of four measurements, giving an relative error of 16%.

sonable correlation between the flowing and static system, but a reduction in heparin binding form 8.4 to 6.4 nm optical thickness is seen at 10 U/ml. Between 0 and 1 U/ml the observed coverage can be achieved by a mainly parallel orientation of heparin. For low concentrations this seems to be likely since parallel orientation links the heparin molecules more strongly to the surface due to a greater footprint area. The roughness of the PEI surface can also cause some of the heparin molecules to appear to have a more perpendicular orientation; for example, an average thickness of 1.8 nm can be achieved by a 50% coverage of parallel oriented heparin molecules and additional 8% coverage of perpendicular oriented molecules. These values approach the maximum coverage predicted for a surface according to the calculations of Vigil and Ziff. Above 1 U/ml, Fig. 7 shows a change in the curve, compared with the steady-state adsorption data (Fig. 4). In order to achieve further heparin binding in this region, the adsorbed molecules would be forced to rearrange or enter the PEI layer. The footprint of one parallel oriented molecule is equivalent to five perpendicular oriented molecules, so that if all molecules are reoriented perpendicular to the surface a maximum 33% coverage is predicted, which corresponds to an average thickness of 5 nm (5 U/ml). Further adsorption of heparin appears to become increasingly more difficult in this short time scale and the deviation between the data in Figs. 7 and 4 increases as the heparin concentration increases. This would be expected if reorientation and migration of the heparin molecules occurs and in the steady state data, the surface bound population reflects the equilibrium state, rather then the initial adsorption orientation seen at shorter times. In order to probe this further, it would be interesting to correlate the heparin binding with interaction time between PEI and heparin more quantitatively, but the

AID

JCIS 5098

/

6g32$$$422

10-22-97 12:49:20

The possibility of using PEI as an affinity surface for heparin in an SPR based sensor was investigated. PEI provides a cationic polymer with a dynamic surface arrangement with a part of the PEI chains moving freely on the polymer surface. The thickness of the PEI film immobilized on the SPR device determined the sensitivity and binding capacity of the heparin adsorption. With an average PEI thickness of 10 { 2 nm the heparin adsorption was within the clinically applicable range of 0–2 U/ml. In this format, the maximum heparin binding was obtained for 8–12 U/ml heparin and led to an adsorbed layer with a thickness of 8.4 nm. This is greater than the maximum predicted by Vigil and Ziff for random adsorption of rods at a surface and probably corresponds to a mixing of heparin into the PEI layer. Reducing the incubation time between heparin and PEI seemed to provide some evidence consistent with an initial surface adsorption, followed by rearrangement and migration/diffusion and revealed a change in the slope of the concentration curve at Ç1 U/ml heparin, compared with the steady state data. At this point, a surface coverage was achieved which approached the predicted maximum for an all parallel orientation of heparin (maximum footprint interaction) with respect to PEI. In a previous study using protamine as an affinity surface for heparin best sensitivity was seen when the protamine was physically adsorbed as a discontinuous layer. Thicker protamine layers had a better binding capacity for heparin, but the sensitivity was reduced. Effective interaction with heparin appears therefore to require a more three-dimensional interaction with the anionic groups which are distributed throughout the heparin molecule and are not focused in one area or to one side. This interaction is more readily achieved by the high active binding capacity of the porous PEI than in the less penetrable protamine layers. In comparing the results from protamine with these for PEI, the high binding capacity for heparin and the difference in the refractive index of PEI and heparin result in an improved accuracy and lower detection limit of 0.05 U/ml compared to the protamine system with the lowest detection limit and accuracy of 0.1 U/ml. It would appear, therefore, that PEI is a promising alternative as an affinity surface for a SPR based heparin biosensor and will offer an interesting comparison to protamine for study in complex clinical media

coida

378

GAUS AND HALL

such as blood, where interference from the adsorption of proteins and other sample compounds may be critical. ACKNOWLEDGMENT Katharina Gaus acknowledges financial support from the German National Scholarship Foundation (Studienstiftung des deutschen Volkes).

7.

8. 9. 10. 11.

REFERENCES 12. 13. 1. Damus, P. S., Hicks, M. S., and Rosenberg, R. D., Nature 246, 355 (1973). 2. Casu, B., in ‘‘Heparin, Chemical and Biological Properties, Clinical Applications’’ (D. A. Lane and U. Lindahl, Eds.), p. 25. Edward Arnold, London, 1989. 3. Howell, W. H., Am. J. Phys. 63, 434 (1923). 4. Abildgaard, U., in ‘‘Heparin, Chemical and Biological Properties, Clinical Applications’’ (D. A. Lane and U. Lindahl, Eds.), p. 495. Edward Arnold, London, 1989. 5. Barrowcliff, T. W., and Curtis, A. D., in ‘‘Haemostasis and Thrombosis’’ (A. L. Bloom and D. P. Thomas, Eds.). Churchill Livingstone, Edinburgh, 1987. 6. Barrowcliffe, T. W., in ‘‘Heparin, Chemical and Physical Properties,

AID

JCIS 5098

/

6g32$$$423

10-22-97 12:49:20

14. 15. 16. 17. 18. 19. 20.

21.

Clinical Applications’’ (D. A. Lane and U. Lindahl, Eds.), p. 393. Edward Arnold, London, 1989. Bjoerk, I., Olsen, S. T., and Shore, J. D., in ‘‘Heparin, Chemical and Biological Properties, Clinical Applications’’ (D. A. Lane and U. Lindahl, Eds.), p. 229. Edward Arnold, London, 1989. Bjornsson, T. D., and Wolfram, K. M., Eur. J. Clin. Pharm. 21, 491 (1982). Bounameaux, H., Marbet, G. A., La¨mmle, B., Eichlsberger, R., and Duckert, F., Am. J. Clin. Pathol. 74, 68 (1980). Holm, M. A., Abildgaard, U., Larsen, M. l., and Kalvenes, S., Throm. Res. 46, 669 (1987). Klein, M. D., Drongowski, R. A., Linhardt, R. J., and Langer, R. S., Anal. Biochem. 124, 59 (1982). Levy, S. W., and Jaques, L. B., Throm. Res. 13, 429 (1978). Jaques, L. B., Wice, S. M., and Hiebert, L. M., J. Lab. Clin. Med. 115, 422 (1990). Gaus, K., and Hall, E. A. H., J. Coll. Int. Sci. 193, 364 (1997). Jilge, G., Unger, K. K., and Esser, U., J. Chrom. 476, 37 (1989). Janzen, R., Unger, K. K., Mu¨ller, W., and Hearn, M. T. W., J. Chrom. 522, 77 (1990). Kitagawa, N., J. Chrom. 443, 133 (1988). Kessler, M., and Hall, E. A. H., J. Colloid Interface Sci. 169, 422 (1995). Richards, F. M., and Knowles, J. R., J. Mol. Biol. 37, 231 (1968). Nieduszynski, I., in ‘‘Heparin, Chemical and Biological Properties, Clinical Applications’’ (D. A. Lane and U. Lindahl, Eds.), p. 51. Edward Arnold, London, 1989. Vigil, R., and Ziff, R., J. Chem. Phys. 91, 2599 (1989).

coida