Eur. J. Biochem. 230, 416-423 (1995) 0 FEBS 1995

Real-time monitoring of antigen-antibody recognition on a metal oxide surface by an optical grating coupler sensor AndrC BERNARD and Hans Rudolf BOSSHARD

Eiochemisches Institut der Universitat, Zurich, Switzerland

(Received 22 February 1995) - EJB 95 0290/3

Real-time monitoring of intermolecular interactions can provide a direct and rapid estimate of the affinity and kinetics of interactions between biomolecules. Optical methods based on the measurement of changes of refractive index in the immediate vicinity of a liquid-solid interface are particularly convenient because they require no radioactive, fluorescent or other labelling of the molecules under study. In the present work we have followed the specific interaction of protein molecules on a SiO,/TiO, surface with the help of the optical grating coupler sensor instrument BIOS-1. This instrument allows the determination of the absolute mass of protein adsorbed to the sensor surface and, therefore, the calculation of the molar ratio of the components partaking in an intermolecular interaction. For example, about 3 ng avidirdmm’ surface area could be adsorbed. This amount closely corresponds to a monolayer composed of densely packed globular avidin molecules. A dimeric, biotinylated leucine zipper peptide was bound to this avidin layer at a molar ratio of 1 : 1 (1 peptide molecule/4 biotin binding sites of tetrameric avidin). An average of 1J2.6 peptides was recognized by a peptide-specific monoclonal antibody. Even though avidin was not covalently bound to the sensor surface, the avidin-coated chip could be used repeatedly to measure the time course of antibody binding as a function of the concentration of the antibody. From such measure- ments it was possible to calculate the association and dissociation rate constants assuming that the interac- tion of the antibody with the surface-bound antigen can be described by a simple Langmuir binding model. The limits of the Langmuir model are discussed. The same antigen-antibody reaction was also analyzed by a surface plasmon resonance biosensor (BIAcoreTM, Pharmacia). The results obtained with the two instruments, which register different optical phenomena and employ different surface chemistry, were in good agreement.

Keywords. Grating coupler sensor ; biosensor ; surface plasmon resonance ; leucine zipper; antigen-anti- body recognition.

Label-free, real-time optical detection methods are of increasing importance in the characterization of macromolecular interactions (Hodgson, 1994). Currently available instruments, such as BIAcoreTM from Pharmacia Biosensor, IAsysTM from Fi- sons Applied Sensor Technology, and BIOS-1 from Artificial Sensing Instruments AS1 AG, analyze the behaviour of light at boundaries between media of different refractive indices, for ex- ample at a solid-liquid interface. In general, a ligand such as an antigen is attached to a solid surface, called the sensor sur- face. An analyte (e.g. a buffered antibody solution) is passed over the sensor surface at a defined flow rate. Binding of the analyte to the ligand gives rise to an increase of mass, leading to a change of refractive index next to the sensor surface. Asso- ciation and dissociation rates are derived from the time course of the refractive index change.

Different optical principles are employed to measure the change of the refractive index, the most widely used technique

Correspondence to H. R. Bosshard, Biochemisches Institut der Uni-

Fax; +41 1 363 7947. Abbreviations. AFM, atomic force microscope; BSA, bovine serum

albumin; GCS, grating coupler sensor; NaCVP,, phosphate-buffered sa- line; SPR, surface plasmon resonance; TE, transverse electric; TM, transverse magnetic.

versitat, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland

being surface plasmon resonance (SPR) (Kretschmann and Raether, 1968; Raether, 1988; Chaiken et al., 1992). Pharmacia Biosensor has commercialized the SPR technology in their in- strument called BIAcore. An alternative technique is based on integrated optics and measures the phase shifts of a light wave coupled into a waveguide (Tien, 1977 ; Tiefenthaler and Lukosz, 1989; Ramsden, 1993a). The shift is caused by the adsorption of particles (molecules) to the surface of the waveguide, which consists of a metal oxide film of high refractive index. From the phase shifts of the transverse electric (TE) and transverse mag- netic (TM) guided modes one can calculate the absolute mass of material adsorbed to the sensor surface. This principle is em- ployed in the grating coupler sensor (GCS) instrument BIOS-1 that was used in the present study to measure the deposition of proteins to a SiO,/TiO, surface. After a short review of the the- ory of grating coupler-based instruments and a description of the instrument BIOS-1, the chemisorption of avidin and protein A to the sensor chip and the reaction of a monoclonal antibody with an avidin-bound biotinylated antigen will be presented. Each of a series of consecutive reaction steps on the surface of the sensor chip could be quantified in terms of the absolute mass of bound protein (ng/mm’), wherefrom the molar ratio of the interacting components was calculated. The time course of the signal change provided values of k,,, k,,, and Kd, assuming a

Bernard and Bosshard (Eur: J. Biochern. 230) 417

cover C flow-through cell lyte at early time points following the change to analyte-free buffer.) The surface coverage r in units of masstarea (e.g. ngt mm') can be calculated from the analyte concentration above the sensor chip surface (see Ramsden, 1993a):

a n a l y t e d

bulk solution 1 adlayer

waveguide

support

/ I

/;aserbeam I

Fig. 1. Schematic diagram of the optical grating coupler sensor. rz is the refractive index and t the thickness of the indicated layer. A laser beam impinges onto the grating at the angle of incidence a. Further details are explained in the text.

single bimolecular equilibrium reaction between the antigen and the antibody.

THEORY OF MEASUREMENT

Fig. 1 shows a schematic diagram of the experimental setup. A glass support S is covered with an optical waveguiding film F into which an optical grating is imprinted at the S/F interface. The thickness tF of F is about 200 nm and its refractive index nF is larger than the refractive index ns of S. A flow-through cell is mounted onto F to allow the analyte solution C to be pumped over the surface of F. The analyte C adsorbs to the surface of F to form the adlayer A of thickness tA and refractive index n,. The adlayer can be formed either through nonspecific adsorption (chemisorption) or by specific binding to a ligand that has been previously attached to the surface of F (not shown in the figure). The combination of waveguiding film, support and diffraction grating is called the sensor chip.

The net rate of adsorption of the analyte to the sensorchip is given by the difference between the rates of association and dissociation and can be written as

drldt = k,, Tm,, c - (ken c + koS) r (1) where r is the surface concentration (surface coverage) of bound analyte, r,, the maximum coverage, c the concentration of free analyte in the bulk solution, k,, the association rate con- stant and k,, the dissociation rate constant. Considering that c remains constant if the analyte is pumped continuously across the sensor chip, Eqn (1) can be integrated (initial conditions: r = 0 at t = to., = beginning of association phase) to

which, for t - and K = k,,/koff, reduces to the Langmuir equa- tion r = r,,, Kc/(Kc + l).

The association phase of the binding of analyte to the sensor chip depends both on the rate of association and the rate of dissociation (Eqn 2). The dissociation of the analyte can be studied by replacing the analyte solution by buffer (c = 0). The

r = (dn,/dc)-' (n, - n&A (4) where n, is the refractive index of the analyte at infinite distance from the sensor surface. For many aqueous protein solutions dnJdc has a value of 0.188 ml/g (de Feijter et al., 1978). Eqn (4) provides a relationship between the surface coverage 7 and optical parameters. Since, as will be explained now, tA and n, can be determined with the help of the GCS, r can be calculated from Eqn (4).

The arrangement shown schematically in Fig. 1 is called an integrated optical grating coupler and can be used to determine nA and t,. If light of wavelength 2 is directed at the angle a onto the optical grating and if nF>n,, n,, the light will be coupled into the waveguide, that means totally reflected at the interfaces FtS and FIA, provided the following relationship holds (Tien, 1977 ; Tiefenthaler and Lukosz, 1989; Ramsden, 1993a):

N = n,,,sina + L2tA ( 5 )

where N is the effective refractive index of the excited guided mode for total internal reflection and is defined as the ratio of the velocity of light in vacuum to the velocity of light in the composite waveguide structure comprising the support S, the waveguiding film F, the adlayer A and the bulk solution. nalr is the refractive index of air, L is the diffraction order (0, 1, 2, . . .), and A is the line spacing of the optical grating. To determine N, the light intensity of the guided wave is measured by a photo- detector situated as shown in Fig. 1, while varying the angle of incidence a. N is highly sensitive to the immediate environment of F because, under conditions of total internal reflection, the electromagnetic field of the guided wave travels some distance into the surrounding medium, the adlayer A and the bulk solu- tion. Thus, any change of nA and tA will result in a change of N. Guided waves of both transverse electric (TE) and transverse magnetic (TM) polarizations can be excited and it is possible to formulate so-called mode equations of the following general form :

tA = f [ @ = 0, n,, N = N(TE)] (6)

t A = f [ @ = I , f l A , N = N(TM)] ( 6 4 where f is a function symbol and Q is a parameter with a value of 0 for the TE mode and 1 for the TM mode. (For a full descrip- tion of the mode equations, see Tiefenthaler and Lukosz, 1989.) Subtracting Eqns (6) and (6a) yields

f [ @ = 0, nA9 N = N(TE)] - f [ @ = 1, n,, N = N(TM)] = 0. (7)

Since the GCS instrument BIOS-1 measures both N(TE) and N(TM), the refractive index n, of the adlayer can be determined by finding the root of Eqn (7), and tA is then calculated from one of Eqn (6) or (6a). n, and tA obtained in this way are then used to calculate r, the absolute mass of analyte adsorbed to the sensor surface, according to Eqn (4).

AN is the change of N during the association and dissociation of an analyte. Since for many situations the surface coverage r is to a good approximation proportional to AN(TE) or AN(TM), Eqns (2) and (3) can be rewritten as

integrated rate equation for dissociation is - ANt) = ANm,, c ( C + k d k o n ) - ' (1 - ~ x P [ - ( k n c + kod ( t - t0,Jl) ( 2 4

ANt) = ~ x P [ - koff ( t - t0,Jl . ( 3 4 r = To exp[- koff (t - t,,.,>l ( 3 )

where r,, indicates the surface coverage at the time to,d when the analyte solution is replaced by buffer. (Eqn 3 is an oversimplifi- cation as it does not take into account the rebinding of the ana-

AN(t) equals N(TE, t ) - Nbacellne (TE, t ) for the TE mode, and N(TM, t ) - Nbaseiine (TM, t ) for the TM mode. Hence, the asso-

41 8 Bernard and Bosshard (Eul: J . Biochern. 230)

LA

t v-

Fig. 2. Schematic arrangement of the components in the grating coupler sensor BIOS-1. The central element is the sensor chip (chip) mounted on the goniometer table (G), which is driven by the stepper motor (M) via the micrometer screw (MS) and the lever arm (LA). The angle encoder (AE) gives feedback information to the computer. The fluid-handling system comprises the switching valves (SV), the tubing (T), the flow-through cell (FC) and the peristaltic pump (PP). The laser beam (L) is deflected by the mirror (MI) onto the grating of the sensor chip. The beam travels in the x-direction along the waveguide and its intensity is measured by the two photodiodes (P).

ciation rate constant k,, and the dissociation rate constant k,, can be calculated from the sensorgrams N(TE) versus time or N(TM) versus time.

MATERIALS AND METHODS

The grating coupler sensor instrument BIOS-1. Fig. 2 de- picts schematically the principle components of the GCS instru- ment BIOS-1 built by Artificial Sensing Instruments (AS1 AG, Zurich). The light source (L) is a He-Ne laser of wavelength 632.8 nm. The laser beam is deflected by a mirror (MI) to the sensor chip. In order to determine the effective refractive index N (Eqn 5), the angle a at which the beam impinges onto the grating (Fig. 1) is varied. To this end, the table on which the sensor chip is mounted can be rotated. A stepper motor (M) drives a micrometer screw (MS) and thereby moves a lever arm (LA). The lever arm rotates a goniometer (G), resulting in a change of the angle of incidence of the light beam. The position of the stepper motor is precisely monitored by an optical protrac- tor. The angle encoder (AE) provides feedback information for the computer to control the change of a with a resolution of 1.25 pad. Photodiodes (P) are positioned at both ends of the goniometer table. For a complete scan of the TM and TE modes by the two photodiodes, the goniometer must move from the left-most to the right-most position and back again, which takes about 25 s. A monomode scan by one photodiode lasts 2.9 s. The monomode scan is used to analyze binding and dissociation kinetics. Switching valves (SV) allow for the pre-programmed automated flow (controlled by a peristaltic pump, PP) of analyte and buffer solutions through the tubing (T) to the flow-through cell (FC), which is made of silicone rubber and is mounted di- rectly onto the sensor chip surface. The cell has a semi-circular cross-section of 1.7 mm2 in the direction of flow and a volume

The sensor chip. The chip is a rectangular slab of dimen- sions 48X16X0.5 mm. The support S consists of glass or plas- tic. The waveguiding film F is about 200-nm thick and the grat-

of 12 pl.

Fig.3. Atomic force microscope image of the grating region of an AS1 1400 sensor chip.

ing covers an area of 1 6 x 2 mm at the center of the chip. For the experiments described the sensor chip AS1 1400, manufac- tured by Balzers AG (Fiirstentum Liechtenstein), was used. It consists of a waveguide ( t , = 190-200 nm) composed of Si,-,Ti,O, (x =0.4) on top of a glass slide (Corning C7659). The refractive indices are n, = 1.53151 and n, = 1.778. The grating period A is 833.33 nm (1200 linedmm). Fig. 3 shows an atomic force microscope image of the grating region of an AS1 1400 sensor chip.

Prior to an experiment, chips were wetted for 20 min in boil- ing water followed by immersion in hot detergent solution (2 % Deconex Universal 11, Borer Chemie, Switzerland) for 30 min, rinsing with distilled water, and drying in a stream of nitrogen gas. Treated chips were then soaked for 2 h in the appropriate buffer solution, mounted in the instrument and finally flushed with buffer until the baseline drift was dN/dt<5X10-* s-'. All the subsequent reaction steps were conducted in NaClP, and at ambient temperature (20 ? 2°C). NaCUP,, pH 7.2, was pre- pared by dissolving 8 g NaC1, 0.2 g KCI, 0.2 g KH,P04, 1.15 g Na,HPO,(H,O),, and 0.2 g NaN, in a final volume of 1000 ml.

Data analysis. Surface coverage r was calculated from the raw data provided by the instrument using a program written by Roger Kurrat (Chemical Engineering Laboratory, ETH, Zurich). Kinetic rate constants were obtained by fitting routines written in this laboratory and using the program Origin of MicroCal Software, Inc. (Northampton NJ, USA).

Measurement with SPR instrument. The BIAcore system, streptavidin-derivatized sensor chip SA5, and surfactant P20 were obtained from Pharmacia Biosensor AB (Uppsala, Swe- den). Measurements were performed following the instructions given by the manufacturer. Coverage of the sensor chip with biotinylated peptide antigen corresponded to 500 resonance units. Other conditions were as specified for measurements with BIOS-1. Kinetic parameters were calculated by our own fitting routines using the program Origin.

Atomic force microscope (AFM) imaging. AFM images were taken with a commercial instrument (Topometrix TMX2000) equipped with a 75-pm scanner. The scans were per- formed in the constant force mode using commercially available AFM tips made of silicon nitride. The sample was investigated at several randomly chosen locations to check its homogeneity. The procedure guarantees that the image shown is representative for the surface morphology of the optical grating of the AS1 1400 sensor chip.

Chemicals and reagents. Cytochrome c was type I11 from Sigma and was further purified as described (Saad et al., 1988).

Bernard and Bosshard (Eur: J. Biochent. 230) 41 9

- N

E E . cn c a, m

a, > 0

(u 0 m 't m

Y

2

-0.5 I I I, I I I I I 0 1000 2000 3000 4000

time [s]

Fig. 4. Physical adsorption of avidin to seven different sensor chips. Avidin (1.2 pM in NaCUP,) was pumped over the sensor surface at flow rates of 8.5 and 13 pl/min. Saturation was reached after approximately 40 min.

Biotinylated cytochrome c was prepared by reaction with the N- hydroxysuccinimide ester of biotin (Fluka) as before (Ngai et al., 1993). The derivative contained about 0.9 mol biotin/mol protein. The peptide antigen was a disulfide-linked dimer with the sequence biotin-GGGCGGGEYEALEKKLAALEAKLQA- LEKKLEALEHG-amide; the biotin group was linked to the a- amino group of the N-terminal residue. The peptide forms a very stable dimeric leucine zipper and was synthesized as described (Leder et al., 1994). mAb 15AF (subclass IgG,) against this peptide was obtained by the standard hybridoma technology (Leder, L. and Bosshard, H. R., unpublished). An mAb against the unrelated protein axonin 1 (Ruegg et al., 1989) was used to control for nonspecific binding. This mAb was kindly provided to us by Daniel Suter and Dr Peter Sonderegger of this labora- tory. The IgG fraction against horse cytochrome c corresponded to the antibody fraction N3 described by Schwab et al. (1993). Nonspecific binding was tested with rabbit IgG from Sigma.

Avidin (Fluka 11368), protein A (Pharmacia Bio Products) and BSA (Sigma A-3803) were used as supplied.

RESULTS AND DISCUSSION

Chemisorption of avidin to the sensor chip. Reproducibility of the physical adsorption of avidin to seven sensor chips is shown in Fig. 4. Surface coverage levelled off around 3-3.5 ng/ mm2 and was independent of the flow rate (4-80 pl/min) and of the concentration of the avidin solution (20-100 pg/ml). Sub- sequent washing with buffer removed only a small amount of loosely bound protein and thereafter the coverage remained con- stant following repeated washing with buffer (not shown). If one assumes the avidin molecule to have an ideally spherical shape, the surface area covered by a single molecule is

n (3m/4neP NL)z/3 (8) where m is the mol mass, eP the protein density and NL Avo- gadro's number. The number of molecules/unit surface area is

TNL/m (9) where T is the measured surface coverage. A compact, ordered monolayer of identical spherical molecules covers n/4 of an ide- ally flat surface. If one defines as 100% the coverage of n/4 of the total surface, the percentage surface coverage, r( %), is ob- tained from Eqns (8) and (9) according to:

0.6 , I I I

0.5- A buffer buffer

cytochrome c

N-

0.4 . m

2 0.2 0

0 1000 2000 3000 4000

time [s]

1 buffer ' B

N- 0.3 - E E .

a, 0.2 - 2 .

s

. m c I

m

a,

8 0.1 - a,

c L -

3 m 0.0 - 1 anti-cytochrome c IgG

I . I , I . I . I . I . I

0 500 1000 1500 2000 2500 3000

. m c I

m 2 a,

a, s c L

m 0.0 1 anti-cytochrome c IgG

500 1000 1500 2000 2500 3000

time [s]

Fig. 5. Reaction of biotinylated cytochrome c (500 nM in NaCVPJ with the avidin-coated sensor surface (A) followed by reaction with a rabbit IgG fraction (150 nM in NaCVP,) against cytochrome c (B). Flow rate 12 pl/min.

r(%) = 100x(4TNL/m) ( 3 m I 4 z ~ ~ NL)2". (10) If r is expressed in ng/mm2, m in glmol and eP in glcm?, Eqn (10) must be multiplied by lo-' to obtain the correct value of

The surface coverage with avidin at the plateau level of Fig. 4 is 80-90% (M, = 68000, ep = 1.3 g/cm'). In practice, the sensor chip surface is not ideally flat (Fig. 3) and the area available for adsorption of protein may be somewhat larger than that of an ideally flat surface. On the other hand, avidin is not completely spherical; dimensions of the crystal are 5.6X5X4 nm (hgliese et al., 1993). Taken together, we believe that a value of 80-90% surface coverage is compatible with a fairly compact, ordered monolayer of avidin, corresponding to approximately 45 fmol avidin/mmz.

r( %).

Reaction of avidin monolayer with biotinylated protein. The reaction of the avidin-coated chip with biotinylated cytochrome c (M, = 12000) followed a saturation curve (Fig. 5A). The pla- teau was reached after binding of approximately 40 fmol cyto- chrome c/mm2, which is about one molecule cytochrome clmole- cule avidin. Avidin is a tehamer composed of identical subunits, each with a single biotin-binding site. Thus, one in four sites

420 Bernard and Bosshard (Eur: J. Biochem. 230)

8 7

7-

a- 6- E . E h 5-

% 4 - E - F 3 -

8 2 -

z 1 -

u -

0 0 .

. t " - I NaCVP, e

4

I , , , , ,

mAb (3.8 ng/mrn2)

(0.3 ng/mm2)

o-2tavidin (3.2 ng/mm2) l ' l - l - l . , - , - ,

I

IgG (2.4 ng/rnm2)

r "'3 ZJ m 0 r; BSA (2 ng/mm2)

1

J f protein A (1.2 ng/mm*) -1 , c , . ) . , . , . , . ,

0 2500 5000 7500 10000 12500 15000

time [s]

Fig. 6. Reaction of rabbit IgG with a protein-A-coated sensor chip. The following steps were performed: coating with protein A (1 pM in NaCVP,), washing with NaCW,, blocking with BSA (10 mg/ml in NaCV PI), washing with NaCVP,, reaction with rabbit IgG (84 nM in NaCYP,), washing with NaCVP,, second reaction with IgG (1 pM) to demonstrate completeness of binding. Numbers in parentheses indicate the amount of bound material (0 after washing with NaCVP, to remove loosely hound material.

bound to a molecule of biotinylated cytochrome c. This stoichi- ometry seems plausible for a molecule of the size of cytochrome c (molecular diameter about 3.5 nm) binding to a monolayer of randomly oriented avidin molecules. Only a small fraction of the cytochrome c was bound nonspecifically and could be re- moved by washing with buffer. Weak nonspecific binding was also demonstrated by adding non-biotinylated cytochrome c (Fig. 5A). The same result was obtained when the order of addi- tion was reversed, biotin-free protein being added first, followed by biotinylated protein (not shown). Bound cytochrome c was recognized by the IgG fraction of a polyclonal rabbit serum against cytochrome c (Fig. 5B). The binding curve of Fig. 5 B extrapolates to 3.3 fmol IgG/mmZ corresponding to a stoichiom- etry of approximately 12 molecules biotinylated cytochrome c/ molecule IgG. We had found previously that direct binding of cytochrome c to plastic microtiter plates leads to considerable denaturation of the protein (Schwab and Bosshard, 1992). In the present experiment, coupling of the biotinylated protein to a monolayer of avidin adsorbed to the sensor chip may circumvent this problem. Nonspecific binding by control IgG was low (not shown).

Protein-A-immunoglobulin interaction. Fig. 6 demonstrates capturing of IgG by a protein-A-coated sensor chip. Protein A from Staphylococcus aureus binds to the F, part of IgG, leaving the antibody binding sites of IgG free to interact with antigen. The binding of protein A saturated at 1.2 ng/mm2, amounting to 38 % surface coverage (Mr of protein A is 42 000). The low cov- erage may indicate random adsorption of the protein (Feder, 1980; Ramsden, 1993b). That the coverage was incomplete was confirmed when the protein-A-covered chip was treated with BSA to yield an additional binding of 2 ng BSA/mmZ, equiva- lent to r(%) = 64%. The reaction with rabbit IgG showed that the surface-bound protein A was able to recognize the F, part of the antibodies. Binding saturated at 2.4 ng/mmz, equivalent to one molecule IgG/two protein A molecules. The captured IgG could be recognized specifically by goat anti-(rabbit IgG) (not shown).

An avidinhiotin-peptidelmonoclonal antibody system. A complete binding experiment with mAb 15AF directed against a dimeric leucine zipper peptide is presented in Fig. 7. The sen- sor chip was coated with avidin and any remaining free binding sites on the surface of the chip were blocked with BSA. r(%) was 85 % for avidin and 8 % for BSA, in agreement with com- plete coverage of the chip surface after the avidin and BSA treat- ment. The chip was thereafter reacted with the biotinylated pep- tide antigen, which bound at a ratio of 1 antigen molecule/3.6 biotin binding sites of avidin, a ratio similar to that seen for the binding of biotin-labelled cytochrome c in Fig. 5. Finally, under the prevailing conditions, one in 2.6 antigen molecules reacted, on average, with the peptide-specific mAb 15AF (3.8 ng/mm2 of mAb and 0.5 ng/mm2 of antigen, M, of mAb = 150000, M, of antigen = 7864). This ratio is again very plausible in view of the small size and high surface density of the antigen.

Bound mAb could be completely removed by brief treatment with 0.1 M glycineMC1 pH 2.7. After washing with BSA, the baseline was slightly lower than before the addition of the mAb. This indicated that the low-pH treatment removed primarily the mAb and very little of the avidin-antigen complex. Since the biotin-avidin complex is stable to treatment with glycine/HCI, the decrease in the baseline must have been due to the removal of avidin from the sensor chip surface. This was demonstrated by a series of six consecutive binding and regeneration experi- ments (Fig. 8A). After each treatment with glycine/HCl, the level of re-bound mAb decreased. However, despite of the pro- gressively lower surface coverage with the avidin-antigen com- plex, the apparent rate of binding of the mAb remained constant for the entire series of bindinghegeneration experiments (Fig. 8B). Even though avidin was not covalently bound to the chip surface, repeated binding and desorption of a mAb could be achieved and it was possible to follow the reaction of dif- ferent amounts of mAb with the same antigen-covered sensor chip.

Determination of association and dissociation rate constants. A kinetic analysis performed on results from a single sensor chip

Bernard and Bosshard ( E m J. Biochem. 230) 421

N- 2 0 -

1 rnAb - 0 . 5 ' ' ' ' ' ' ' ' '

0 500 1000 1500 2000 2500 3000

time [s]

1 .o U al N m

0 c

.- - c & 0.5

a, E

8 a, 0 m

0)

..- 5 0.0

l . l - l ' l ' l ~ l -

NaCIIP, buffer Y,

I mAb , , l . l . l . I . I .

0 500 1000 1500 2000 2500 3000

time [s]

Fig. 8. Overlay plot of six consecutive reactions performed on the same sensor chip. (A) The chip had been coated with avidin followed by biotinylated peptide antigen (see Fig. 7). Six bindinghegeneration cy- cles were performed. A cycle consisted of the reaction with 250 nM mAb, washing with NaCW,, regeneration with 0.1 M glycineiHC1, and re-equilibration with NaClP,. The last cycle (dotted line) was prolonged to demonstrate the lowered equilibrium level after six bindinghegenera- tion steps. (B) The six association phases were normalized and are pre- sented as an overlay plot to demonstrate that the rates of association did not change significantly even though some of the avidin-antigen com- plex was removed at each regeneration step. Normalization was achieved by assigning a value of 1 to the change of surface coverage between to,,, the beginning of the association phase, and the beginning of the dissociation phase.

is shown in Fig. 9A. Altogether eight different concentrations of mAb, from 20 nM to 500 nM, were reacted with the avidin- bound antigen (only three traces are shown for clarity in Fig. 9A). The binding traces were analyzed assuming a single bimolecular binding equilibrium between antigen and mAb, as represented by Eqn (2). To this end, the data points were fitted to Eqn (11)

N = Nbasellne + ANm {I - exp[- kbs ( t - t,,,>l) (11) where AN, is the maximum change of the effective refractive index at t - 00 and kobs = k,,c + kOfp If necessary, a small drift of the baseline was corrected for by

1.57490-

i5 1.57488-

al -0

.G 1.57486- > .- 1

2 1.57484- E a, 2 1.57482-

1.57480 - [mab] '10' M

1 ~ 1 ' I ' I ' l . I ' I I 0 100 200 300 400 500 600

time Is]

1 . 1 ' 1 ' I buffer ' I '

1.57490 - - 2 1.574881 ' I 1 z x

7 .- 1.57486

1.57484

1.57480

J l - l ~ l ' l ' l '

0 200 400 600 800 1000 1200

time [s]

Fig. 9. Reaction of mAb with biotinylated peptide antigen bound to an avidin-coated sensor chip. (A) Reaction of 50, 100 and 200nM mAb with biotinylated peptide antigen bound to an avidin-coated sensor chip as revealed by the time course of the change of N(TM). The chip was regenerated with 0.1 M glycine/HCl after each binding experiment (see Fig. 8A). The solid lines are best fits according to Eqn (11). Inset: plot of k,,, against the concentration of the mAb. k,,, and kOff were calcu- lated from the slope and intercept, respectively, and are shown in Ta- ble 1. (8) Association and dissociation phase of the reaction of 100 nM mAb with biotinylated peptide antigen bound to an avidin-coated sensor chip. The solid line is a best fit of Eqn (13) to the dissociation phase to obtain k,, = (4.5 t 0.2)X10-3 s-'. This value was used for a constrained best fit of the association phase by Eqn (11) to obtain k,. = (2.13t0.11)X104 M-'s-'.

where z is an arbitrary time preceding to,,, the beginning of the association phase. Eqn (11) is equivalent to Eqn (2a). A plot of kob\ versus [mAb] is shown in the inset of Fig. 9A and ken and kofl were calculated from this plot (Table 1). The data of Fig. 9A were also analyzed by a direct fitting procedure (O'Shannessy et al., 1993). Firstly, k,, was determined from the dissociation reaction according to

N = Nbaseitne + ANo exP[- k m (t - t o , J I (13)

where AN, is the maximum change of refractive index at time t = to,*, the beginning of the dissociation phase. Eqn (13) corres- ponds to Eqn (3a). kOfl obtained in this way was then used to constrain the analysis of the association phase, using Eqn (11). Since the direct fitting method does not depend on a lineariza- tion of the primary data, it is believed to yield more reliable results than the indirect fitting method (O'Shannessy et al.,

Nbarcline (t) = Nbaselirle (t = t) + [dNb,,,,,,,/dr] (t - z) (12) 1993). An example of a fit is shown in Fig. 9B. Values for k,,

422 Bernard and Bosshard (Eur. J. Biochem. 230)

Table 1. Kinetic constants for the binding of mAb 15AF to the di- meric leucine zipper peptide. Values in parentheses were obtained by BIAcore. In the indirect fitting procedure, values of k,,, obtained from the association phase according to Eqn (11) were plotted as shown in the inset of Fig. 9A to obtain k,, and k,,* In the direct fitting procedure (O’Shannessy et al., 1993), k,, was obtained first from a fit of Eqn (13) to the dissociation phase, this value was then used for a constrained fit of Eqn (11) to the association phase (Fig. 9B). Values arc the mean C SEM from 8 measurements in the concentration range of 20-500 nM mAb.

M-’ s- ’ s- ’ M

Indirect fitting 3.6+O.2X1O4 2.3?0.4X10-~’ 6.420.3X10-x (Fig. 9A) (6.O+0.4X1O4) (1.8C0.8X10-3) (3.01 1.3XlO-‘)

Direct fitting 4.9+1.5X104 2.9C1.2XlO-’ 5.9?2.5XlO-* (Fig. 9B) (4.5C1.4X104) (4.8*1.9X10-3) (11 -C IXIO-’)

and k,, obtained by the direct fitting method are shown in Table 1.

Comparison with data obtained by SPR-based instrument. The same antigen-antibody reaction was also studied in a BIA- core instrument. Here, the ligand was covalently immobilized on a dextran network deposited onto the gold film of the sensor chip. A pre-fabricated streptavidin-dextran chip was used to which the biotinylated antigen was bound. The kinetic parame- ters were obtained as described above for the BIOS-1 experi- ment and the results are also given in Table 1. The association rate constants were of similar magnitude as those measured with BIOS-1. Depending on the method of data analysis, the off-rates found with BIAcore were faster or slower than the off-rates from the GCS instrument.

The SPR technique does not allow the calculation of the absolute amount of adsorbed material. By an indirect estimate based on the assumption that 1000 SPR resonance units corre- spond to 1 ng/mm2 (Stenberg et al., 1991), 0.5 ng/mm2 of bio- tinylated antigen were bound to the dextran-coupled streptavi- din. This figure happens to be the same as that reported in Fig. 7. However, the surface structures of the sensor chips used in the two instruments are very different. In the case of the BIOS-1, the reaction took place at a monolayer of protein molecules ad- sorbed to a smooth metal oxide surface. In the BIAcore, the analyte molecules had to diffuse into and out of a mesh of dextran about 100 nm thick attached to the gold film of the sen- sor chip. The close correspondence of the kinetic data obtained by the two experimental systems is quite satisfactory and indi- cates that probably in neither system was the rate of mass diffu- sion from the bulk solution to the sensor chip surface affecting the rate of mAb binding (Glaser, 1993; O’Shannessy, 1994). In- deed, the k,,, values measured with the BIOS-1 instrument were unaffected by a tenfold change of the rate of the buffer flow, from 8 to 80 ,ul/min (not shown).

Simple Langmuir binding equilibrium versus more complex binding models. Close inspection of Fig. 9 shows that the sim- ple exponential binding model described by Eqns (11) and (13) did not fit to the experimental data very accurately. Also the Langmuir fit of the data obtained by the BIAcore was unsatisfac- tory (not shown). Indeed, this is not surprising since the simple Langmuir binding model is likely to fail the requirements of the processes taking place on the sensor surface, which was densely packed with peptide antigens. The stochastic process of complex formation on the sensor surface may be described more realisti-

cally by a set of several binding equilibria. These have to ac- count for differences in the accessibility of the surface-bound antigen to the antibody. Some avidin-bound antigens may be freely accessible and others partly or completely buried. Also, crowding of the antibodies, which could lead to direct interac- tion between neighbouring antibodies, can occur at high anti- body concentrations. Description of the increase of the analyte concentration close to the sensor surface by a simple step model (as assumed in deriving Eqn 4 ; Ramsden, 1993a) may not be adequate, and rebinding of the antibody during the early dissoci- ation phase has to be considered. There have been attempts to use biphasic and even multiphasic reaction kinetics to more ade- quately fit the experimental data from BIAcore measurements (Fagerstam et al., 1992; Malmborg et al., 1992; Zeder-Lutz et al., 1993; Morton et al., 1994; O’Shannessy, 1994). Indeed, an improved fit of the experimental data shown in Fig. 9A was obtained using a bi-exponential extension of Eqn (11) (data not shown)

Numerical fitting, based on a set of coupled differential equations, has been employed to describe a model with several interdependent consecutive equilibria between the analyte and the ligand at the sensor surface (Fisher et al., 1994). Although better fits to the experimental data can be achieved by such mod- els, all these attempts account primarily for peculiarities of the biosensor system and one is left with the problem of how the parameters pertaining to the reactions in the biosensor are to be correlated with the reactions as they occur in vivo. Forcing the Langmuir model onto the experimental data is certainly an over- simplification and may even produce parameters of doubtful utility. Whenever possible the parameters obtained from bio- sensor measurements should be compared with parameters from another experimental approach. In the present case, we have measured the equilibrium dissociation constant of the reaction of mAb 15AF with the leucine zipper antigen by isothermal tit- ration calorimetry (Wiseman et al., 1989). A value of Kd = (4.7 % 2.5) nM was obtained under the same buffer conditions as were used in the BIOS-1 and BIAcore experiments. The value from calorimetry is almost tenfold smaller (tighter binding) than that deduced from the kinetic constants obtained with either of the two biosensors. We have no explanation for this discrepancy except to note that one reaction took place in solution and the other at a solid- liquid interface.

CONCLUDING REMARKS

A main advantage of the GCS technology is that it allows direct measurement of the amount of material deposited on the sensor chip surface. The direct calculation of the adsorbed mass of material is not possible with the SPR method nor with the resonant mirror technique as employed in the IAsys instrument from Fisons Biosensor Technology. With these instruments the adsorbed mass may be estimated based on a calibration curve with radioactive material (Stenberg et al., 1991).

To perform a series of repetitive measurements with the same sensor chip, it was not necessary to covalently link the initial ligand to the waveguide film. The reason is that avidin adsorbed very tightly to the SiO,ITiO, surface. Other non-cova- lently adsorbed proteins may not survive the low-pH conditions necessary to regenerate the sensor chip surface for repetitive measurements. Covalent attachment of ligands to the GCS sur- face has been achieved through silanization (Jockers et al., 1993) or by photochemical linkage (Collioud et al., 1993; Gao et al., 1994). Alternatively, biospecific interaction can also be analyzed by anchoring proteins to a phospholipid bilayer deposited onto the GCS surface (Ramsden and Schneider, 1993).

Bernard and Bosshard

We are grateful to Drs Jeremy J. Ramsden and Kurt Tiefenthaler for very helpful discussions and critical reading of the manuscript. We thank Dr Fachri Atamny and Mr Ham H M a for taking AFM pictures of the sensor chip and for providing Fig. 3, Mr Roger Kurrat for providing the software to calculate the surface coverage, and Mr Lukas Leder for help with the isothermal titration calorimetry experiment. This work was sup- ported in part by the Kommissionfir Wissenscha#liche Forsckung (pro- ject 2690.1, Eureka project MEMOCS), by the Swiss National Science Foundation (grant 31-36149.92), and by the Kanton of Zurich.

REFERENCES Chaiken, I., RosC, S . & Karlsson, R. (1992) Analysis of macromolecular

interactions using immobilized ligands, Anal. Biochem. 201, 197 - 210.

Collioud, A,, Clemence, J. F., Sanger, M. & Sigrist, H. (1993) Oriented and covalent immobilization of target molecules to solid supports: synthesis and application of a light-activatable and thiol-reactive cross-linking reagent, Bioconjug. Ckem. 4, 528 -536.

de Feijter, J. A., Benjamins, J. & Veer, F. A. (1978) Ellipsometry as a tool to study the adsorption behaviour of synthetic and biopolymers at the air-water interface, Biopolymers 17, 1759- 1772.

Fagerstam, L. G., Frostell-Karlsson, A,, Karlsson, R., Persson, B. & Ronnberg, I. (1 992) Biospecific interaction analysis using surface plasmon resonance detection applied to kinetic, binding site and con- centration analysis, J. CkromtogE 597, 397-410.

Feder, J. (1980) Random sequential adsorption, J. Tkeos Biol. 87,237- 254.

Fisher, R. J., Fivash, M., Casas Finet, J., Erickson, J. W., Kondoh, A., Bladen, S . V., Fisher, C., Watson, D. K. & Papas, T. (1994) Real-time DNA binding measurements of the ETSl recombinant oncoproteins reveal significant kinetic differences between the p42 and p51 iso- forms, Protein Sci. 3, 257-266.

Gao, H., Kislig, E., Oranth, N. & Sigrist, H. (1994) Photolinker polymer mediated immobilization of monoclonal antibodies and F(ab’) frag- ments, Biotechnol. Appl. Biockem. 20,251 -263.

Glaser, R. W. (1993) Antigen-antibody binding and mass transport by convection and diffusion to a surface: A two-dimensional computer model of binding and dissociation kinetics, Anal. Biockem. 213, 152-161.

Hodgson, J. (1994) Light, angles, action, BioTechnology 12, 31 -35. Jockers, R., Bier, F. F. & Schmid, R. D. (1993) Specific binding of

photosynthetic reaction centres to herbicide-modified grating cou- plers, Anal. Ckim. Acta 280, 53-59.

Kretschmann, E. & Raether, H. (1968) Radiative decay of non-radiative surface plasmons excited by light, Z. Naturjorsch. A 23,2135-2136.

Leder, L., Wendt, H., Schwab, C., Jelesarov, I., Bomhauser, S . , Acker- mann, F. & Bosshard, H. R. (1994) Genuine and apparent cross- reaction of polyclonal antibodies to proteins and peptides, E m J. Biochem. 219, 73-81.

Malmborg, A. C., Michaelsson, A., Ohlin, M., Jansson, B. & Borre- baeck, C. A. K. (1992) Real-time analysis of antibody-antigen in- teraction kinetics, Scand. J. Immunol. 35, 643-650.

Morton, T. A., Bennett, D. B., Appelbaum, E. R., Cusimano, D. M., Johanson, K. O., Matico, R. E., Young, P. R., Doyle, M. & Chaiken, I. M. (1994) Analysis of the interaction between human interleukin- 5 and the soluble domain of its receptor using a surface plasmon resonance biosensor, J. Mol. Rec. 7, 47-55.

(Eus J. Biockem. 230) 423

Ngai, P. K., Ackermann, F., Wendt, H., Savoca, R. & Bosshard, H. R. (1993) Protein A antibody-capture ELISA (PACE): An ELISA for- mat to avoid denaturation of surface-adsorbed antigens, J. Immunol. Methods 158, 267-216.

O’Shannessy, D. J., Brigham-Burke, M., Soneson, K. K., Hensley, P. & Brooks, I. (1993) Determination of rate and equilibrium binding con- stants for macromolecular interactions using surface plasmon reso- nance: Use of nonlinear least squares analysis methods, Anal. Bio- ckem. 212,457-468.

O’Shannessy, D. J. (1994) Determination of kinetic rate and equilibrium binding constants for macromolecular interactions : A critique of the surface plasmon resonance literature, Cum Opin. Biotechnol. 5,65 - 71.

Pugliese, L., Coda, A., Malcovati, M. & Bolognesi, M. (1993) Three- dimensional structure of the tetragonal crystal f o p of egg-white avi- din in its functional complex with biotin at 2.7 A resolution, J. Mol. Biol. 231, 698-710.

Raether, H. (1988) Surjace plasmons on smooth and rough suqaces and on gratings, Springer-Verlag, Berlin.

Ramsden, J. J. (1993a) Review of new experimental techniques for in- vestigating random sequential adsorption, J. Stat. Pkys. 73, 853- 877.

Ramsden, J. J. (1993b) Concentration scaling of protein deposition ki- netics, Pkys. Rev. Lett. 71, 295-298.

Ramsden, J. J. & Schneider, P. (1993) Membrane insertion and antibody recognition of a glycosylphosphatidylinositol-anchored protein : an optical study, Biochemistry 32, 523-529.

Ruegg, M. A,, Stoeckli, E. T., Kuhn, T. B., Heller, M., Zuellig, R. & Sonderegger, P. (1989) Purification of axonin-1, a protein that is secreted from axons during neurogenesis, EMBO J. 8, 55-63.

Saad, B., Corradin, G. & Bosshard, H. R. (1988) Monoclonal antibody recognizes a conformational epitope in a random coil protein, Eur: J. Biockem. 178, 219-224.

Schwab, C . & Bosshard, H. R. (1992) Caveats for the use of surface- adsorbed protein antigen to test the specificity of antibodies, J. Im- munol. Methods 147, 125-134.

Schwab, C., Twardek, A,, Lo, T. P., Brayer, G. D. & Bosshard, H. R. (1993) Mapping antibody binding sites with synthetic peptides : Are results representative of the antigenic structure of proteins? Protein Sci. 2, 175-182.

Stenberg, E., Persson, B., Roos, H. & Urbaniczky, C. (1991) Quantitative determination of surface concentration of protein with surface plas- mon resonance using radio-labeled proteins, J. Colloid. Interface Sci. 143,513-526.

Tiefenthaler, K. & Lukosz, W. (1989) Sensitivity of grating couplers as integrated optical chemical sensors, J. Opt. Soc. Am. B6, 209-220.

Tien, P. K. (1977) Integrated optics and new wave phenomena in optical wave guides, Rev. Mod. Pkys. 49, 361-420.

Wiseman, T., Williston, S . , Brandts, J. F. & Lin, L.-N. (1989) Rapid measurment of binding constants and heats of binding using a new titration calorimeter, Anal. Biockem. 179, 131 - 137.

Zeder-Lutz, G., Altschuh, D., Geysen, H. M., Trifilieff, E., Sommer- meyer, G. & Van Regenmortel, M. H. V. (1993) Monoclonal antipep- tide antibodies: Affinity and kinetic rate constants measured for the peptide and the cognate protein using a biosensor technology, Mol. Immunol. 30, 145-155.