The biochip surface that we have developed (Nogues et al.
, submitted) consists of a 50 nm thick gold surface functionalised with hydroxyl-terminated tetra (ethylene glycol) (EG4
-OH) to which the proteins had been linked via cysteine residue 101 on GAD65 (). Typical images of the interaction of the antibody (4.8 nM) as it flows across the spots containing the immobilised GAD65 protein are shown in . The images are differential images and are taken every two seconds during the experiment. The background is low and this is also evident from curves in showing the % reflectivity changes for the wild type protein GAD65 containing spot and background spot as antibody passes over the surface. At relatively low concentrations of the antibody (4.8 nM) a simple binding curve was observed () from which apparent rate constants (shown in the legend to ) could be calculated. At these concentrations of antibody the protein GAD65-antibody interaction gave an overall apparent equilibrium dissociation constant (
) of 1.37 nM. We note that these values are somewhat different from values reported previously in the literature for this type of interaction 
; we consistently found lower equilibrium dissociation constants essentially due to more rapid association rates that we measure here. There may be many explanations for this related to the difference in the immobilisation procedure, the architecture of the SPR machine, and so forth. However we were surprised to note that at higher concentrations of antibody (48nM) the binding phase became multiphasic or at least biphasic (). This observation was clearly not due to non-specific binding to the surface ( shows a difference curve after subtraction of background), and indeed the observation of multiphasic binding is due precisely to the fact that we have very low non-specific binding to non target areas on the surface. We note that the concentrations of anti-GAD antibody used in the Biacore study of 
were in excess of 500nM, which is 10 to a 100 fold in excess of the values that we use in the current manuscript and that this strongly suggests that the values obtained in 
were somehow influenced by a degree of non specific binding, if not necessarily to the surface then perhaps to an abnormal form of the immobilised GAD.
Structural characteristics of GAD65.
Difference images obtained from SPRi of antibodies binding to GAD65 immobilised at a gold/GLISS surface.
Changes in % reflectivity as a function of time as GAD1 antibody reacts with immobilised GAD65 protein.
In order to illustrate the degree of antifouling conferred by the General Liquid Interface Specific Surfaces (GLISS) we in fact passed human serum across two types of surface, one consisting of a classical self assembling monolayer (SAM) constructed from undecanoic acid as described in 
, and the other using the GLISS surfaces used in the current GAD65 study. As can be seen in there is a qualitative and quantitative difference between non-specific binding to the two surfaces with a much reduced binding to the GLISS surface. The almost negligible amount of material retained at the GLISS surface compared to the classical SAM surface is even more evident from the SPR curves derived from the images.
Differential binding to SPRi surfaces.
The available crystallographic structure of GAD65 
allowed the selection of a suitable method of coupling of the protein to the chip. Specifically, coupling using surface-exposed cysteine residues in the N-terminal domains allowed the molecule to be immobilised such that the putative antibody epitopes in the C-terminal and PLP domains are exposed and accessible by an antibody molecule (). The dimeric architecture of GAD65 presents a challenge for immobilization, since the 2-fold symmetry may place equivalent residues in each monomer on opposite sides of the dimer. There are only two cysteine residues that are on or near the protein surface (Cys101 and Cys304). Cys304 is relatively buried, having only 22 Å2
solvent accessible surface area. In addition, its sidechain points inwards towards the body of the protein. Coupling would thus require a significant structural reorganisation, which is highly unlikely. Furthermore, Cys304 residues (from each monomer) are on opposite faces of the dimer, precluding coupling at both sites simultaneously without massive structural reorganisation. Although we cannot discount some coupling at only one Cys304 residue at a time (e.g., “half occupancy”), its low exposure, stereochemistry and location make it a highly unlikely candidate for coupling to the linker. Conversely, Cys101 is highly exposed on the surface of the protein (140 Å2
solvent accessible surface area). It is located in the N-terminal domain very close (within 0.15 nm) to its “symmetry-mate” in the dimer. The high exposure and location of Cys101, therefore, make it a highly likely candidate for linker coupling. The high probability of this mode of coupling now allows us to confidently pursue the immobilization of several GAD65 mutants on a single chip. For example we envisage that engineering epitopes on GAD65 may pave the way for improved diagnostic tests for type 1 diabetes using a panel of antibodies as well as human sera. The ability to couple multiple GAD65 proteins on the same chip and measure antibody binding in a reproducible and rapid fashion may ultimately lead to a novel SPRi-based diagnostic immunoassay for type 1 diabetes.
We clearly advocate the use of engineering surface exposed cysteines for immobilisation but recognize that this may not always be possible. A number of alternatives are available, and generally in SPR, coupling using amines, for the most part through accessible lysines, is advocated. It must be noted however, that this generally results in a reduced degree of activity of the immobilised target that may render quantitative analysis difficult. The GLISS surfaces used here may easily be functionalised with carboxyl, thiol or amine groups thus permitting a wide range of immobilisation techniques. However we would like to stress that although the expedient of engineering solvent accessible thiols is somewhat limiting it does optimise accessibility and that whilst this may restrict general applicability we strongly suggest that immobilisation strategies that aim at increasing accessibility and orientation be elaborated rather than blind immobilisation through solvent accessible amines for example. Alternatively one can immobilise on the GLISS surfaces, specific antibodies or haptens either through accessible cysteines or via other coupling techniques, that then allow mild capture of the target molecules.
Although the purpose of the present study was not to explore the detection levels of the SPRi technique we could detect and quantify GAD65-anti-GAD65 interactions between approximately 1010 molecules of target GAD65 on the surface and antibody at 4nM concentration in solution. Our limit of detection as discerned from the signal to noise ratio suggests that we can detect a change of approximately 0.01% reflectivity that corresponds to levels of detection of around 30 to 40 ng/ml of protein in solution.
We have characterised the binding of wild type GAD65 to the GAD1 monoclonal antibody using SPRi. Our data illustrate that we have effectively eliminated non-specific interactions with the surface containing the immobilised GAD65 molecules. The implications of this are far reaching; in short not only does this approach obviate the dubious process of background subtraction but gives access to more accurate kinetic and equilibrium values that tend towards more affine measurements since any multiphase behaviour can be separated from non-specific binding. On a broader level, an enhanced signal to noise ratio increases not only the sensitivity but also confidence in the use of SPR to generate kinetic constants that may then be inserted into van't Hoff type analyses to provide comparative ΔG, ΔS and ΔH values, making this an efficient, rapid and competitive alternative to ITC measurements used in drug and macromolecular-interaction mechanistic studies. Finally, and this is particularly evident here, the accuracy of the measurements allows the application of more intricate interaction models than simple Langmuir monophasic binding. The observation that monoclonal antibodies can use multiple binding modes is intriguing. We are currently applying the technology developed here to analyse monoclonal antibody binding to microarrays containing selected mutants of GAD65 (Buckle et al. in preparation). The implications are that intramolecular rearrangements associated with antibody binding may be involved.
The detection and measurement of antibody binding by the type 1 diabetes autoantigen GAD65 represents an example of an antibody-antigen interaction where good structural, mechanistic and immunological data are available. Using SPRi we were able to characterise the kinetics of the interaction in greater detail than ELISA/RIA methods. Furthermore, our data indicate that SPRi is well suited to a multiplexed immunoassay using GAD65 proteins, and may be applicable to other biomarkers.