Biosensors & Bioelectronics 6 (1991) 201-214

Sensitivity enhancement of optical immunosensors by the use of a surface plasmon resonance fluoroimmunoassay J . W. Attridge, P. B . Daniels, J . K. Deacon, G . A . Robinson & G . P. Davidson Serono Diagnostics Ltd ., Unit 21, Woking Business Park, Albert Drive, Woking, Surrey GU21 5JY, UK (Received 4 May 1990 ; revised version received 7 August 1990; accepted 9 August 1990)

Abstract : Optical immunosensors employing evanescent wave techniques have the potential to address the requirements of the `alternative site' market ; however, this potential has yet to be realised . The development of `direct' sensors, such as those using surface plasmon resonance (SPR), has been hampered by problems of non-specific binding and poor sensitivity to small molecules. `Indirect' sensors (for example, those employing a fluorescently labelled reagent) overcome many of the problems of direct sensors but require more sophisticated instrumentation because of the low light levels detected . In an attempt to combine the best features of the two techniques, an indirect SPR fluoroimmunoassay (SPRF) technique has been investigated . The surface field intensity enhancement produced by SPR is used to boost the emission from a fluorescently labelled immunoassay complex at a metal surface . The potential of the method is demonstrated by assaying for human Chorionic Gonadotrophin (hCG) in serum. Enhanced sensitivity over conventional total internal reflection fluorescence (TIRF) and SPR techniques was achieved. Keywords : surface plasmon resonance, immunosensor, fluorescence, fluoroimmunoassay, total internal reflection, human chorionic gonadotrophin, biosensor, evanescent field .

INTRODUCTION

(c) instrumentation that is inexpensive, robust, reliable and simple to use; (d) assay controls or references which give the user confidence in the results obtained.

Much effort in recent years has concentrated on the development of optical immunosensors for the `alternative site' market, i .e. outside the clinical laboratory. In order to address the requirements of this market any sensing system will need to have :

Optical techniques using evanescent wave phenomena have been identified as attractive approaches to meeting these requirements . In general, immunosensors fall into one of two categories : `direct' and `indirect' sensors. With a direct immunosensor the antigen/antibody reaction is monitored without the use of labelled reagents, whereas for an indirect sensor a label such as an

(a) cheap and disposable sensing elements ; (b) one-step assay protocols which require no sample manipulation, reagent addition or wash/separation steps by the user ; 201

Biosensors & Bioelertronics 0956-5663/91/$03 .50 0 1991 Elsevier Science Publishers Ltd, England . Printed in Great Britain

202 enzyme or fluorophore is used to provide the detected signal . Direct optical sensing techniques include : ellipsometry (Vroman & Adams, 1969), attenuated total internal reflection (Sutherland et al., 1984a), surface plasmon resonance (Liedberg et al., 1983) and, more recently, approaches using monomode dielectric waveguides (Nellen et al., 1988). Following the work of Liedberg et al. in 1983, the field of SPR immunosensors has been much developed (Flanagan & Pantell, 1984 ; Cullen et al., 1987/88; Daniels et al., 1988 ; Mayo & Hallock, 1989 ; Parry et al., 1989). In an SPR immunosensor light is coupled resonantly into electronic oscillations, called surface plasmons, at a metal surface . These oscillations give rise to a non-propagating evanescent field which extends from the metal surface into the sample solution to a depth of less than a micron. Localised refractive index changes, produced by the formation of, for example, an antibody/antigen complex immobilised at the metal surface, perturb this field and alter the propagation characteristics of the surface plasmons, thus moving the sensor off-resonance . This is detected through the resultant change in the reflectivity of the metal surface . The attraction of SPR is its relative simplicity ; no reagent addition is required and, being a reflectance technique, there is plenty of light available to enable simple instrumentation to be used. However, because of the need to produce a significant surface refractive index change it is most suited to the detection of large analytes with molecular weights typically greater than 100 000 Daltons . A test for small molecules, e.g. haptens, requires careful choice of assay format and/or the use of a'refractive index label' (Drake et al., 1988). The measurement of refractive index makes these sensors susceptible to interferences from nonspecific binding, temperature fluctuations and sample absorbance. Only recently have some of the problems of using direct SPR immunosensors in realistic environments been appreciated (Cullen & Lowe, 1990 ; Sambles, 1988). Even with careful referencing, detection in biological samples is difficult unless wash/separation steps are employed . Indirect optical immunosensors using fluorescent evanescent wave coupling phenomena overcome many of the limitations of the direct sensors and offer attractive routes to improving sensitivity . The perceived disadvantage of reagent addition can be overcome through careful `packaging' of the sensor to make the assay

J. W Attridge et al.

procedure appear 'reagentless' to the user (Deacon et al., 1990) . These techniques are considerably more sensitive than the direct sensors but, because of the low light levels detected, require more sophisticated instrumentation . There are, however, few restrictions on the assay format, analyte size or sample type available to such sensors. In addition, fluorescence is less susceptible to the interferences and perturbations experienced by direct sensors. A number of indirect optical immunosensor device formats have been reported . Fluorescently labelled reagents, immobilised at a glass surface as a result of assay complex formation, can be excited either by direct illumination or by the evanescent field associated with total internal reflection of light at the glass/sample interface . Fluorescence emission can be detected both in the sample and through evanescent coupling into the substrate (Kronick & Little, 1975 ; Sutherland et al ., 1984a; Place et al., 1985 ; Andrade et al., 1985; Badley et al., 1987; Parry et al., 1990). A number of approaches have been suggested to improve the signal levels generated by TIRF immunosensors including the use of multiple reflections within a waveguide to increase the number of interactions of the evanescent field with the surface (Sutherland et al., 1984b) and, more recently, the use of planar monomode waveguides (Sloper et al., 1990). In the latter case the evanescent field is continuous along the length of the waveguide surface allowing the integration of the emission signal over a larger surface area . However, in both cases, the detection of fluorescence must be in the same plane as the excitation light to obtain the benefit of the multiple interaction with the surface . Whilst theoretically the emission signal can be discriminated from the excitation light both in angle and wavelength, in practice this geometry can cause problems within the instrumentation as the excitation light must be filtered completely to prevent interference with the signal being detected. Detection of fluorescence emission back in the direction of excitation overcomes this problem as only scattered excitation light need be filtered out, however, this can lead to complex optical configurations (Bluestein, 1989) . It would be advantageous to detect the fluorescence emission in a plane other than that of the incident radiation . Ultimately, the sensitivity of these immunosensors is limited by the amount of labelled reagent that remains unbound within the evanescent field . Reducing

203

Sensitivity enhancement of optical immunosensors

the concentrations of labelled reagent will improve assay sensitivity but can also reduce signal levels to an unacceptable degree . In an attempt to improve the performance of fluorescence immunosensors, an indirect SPR fluoroimmunoassay (SPRF) technique has been investigated. Surface plasmon resonance excitation is used to enhance the emission from fluorescently labelled reagent immobilised at a metal surface . Emission couples from the fluorophore back into the substrate via a surface plasmon at the fluorophore emission wavelength . The technique right-angled excitation/detection enables geometries and efficient detection optics to be used. By enhancing the detected signal SPRF offers the possibility of improving the signal-tobackground ratio of an assay through a reduction of the concentration of fluorophore labelled reagent used.

THEORY The resonant coupling of SPR focuses energy into the evanescent field giving rise to an enhancement in intensity of the surface field over the intensity of the input light by as much as two orders of magnitude (Weber & Ford, 1981) . This enhancement can be calculated using the established `characteristic matrix' approach to the analysis of multilayer systems (Born & Wolf, 1980). Using the optical configuration shown in Fig. 1 the surface field enhancement has been calculated for SPR and total internal reflection (TIR) using typical values for substrate and sample refractive indices (Fig . 2). For silver the SPR field is approximately 40 times more intense than the input radiation ; it is this that gives rise to the enhanced emission from fluorophore within the evanescent field . For TIR there is no resonant coupling and an enhancement factor of only 3 arises from the standing wave generated by the interference between incident and reflected light. SPR excitation occurs at angles well removed from the critical angle where the evanescent field penetration depth, defined here as the distance from the surface at which the field intensity drops to 1/e of its surface value, is a slowly varying function of angle . The field penetration depth is a function of the angle of incidence 0, the wavelength A, and the refractive indices of substrate (n ;) and sample (nt ) and is given by the following expression :

A Pd = 4n[n 2 sin2 0 - nt J l/2 The penetration depth is more usually defined in terms of field amplitude (Harrick, 1979), but using the field intensity gives a better indication of the coupling probabilities; the latter being proportional to field intensity rather than amplitude. It is desirable to maximise the surface field intensity to enhance fluorescence emission whilst at the same time minimising the field penetration depth in order to avoid unnecessary excitation of any fluorescently labelled reagent remaining in solution. A'figure of merit', defined as the ratio of the surface field enhancement to penetration depth, can be used to optimise the excitation conditions. A comparison of SPR and TIR excitation is given in Table 1 . For TIR a compromise has to be reached because both field penetration and intensity attain maximum values at the critical angle. For SPR no such compromise is necessary and so the excitation figure of merit, at the resonance angle, is almost twenty times greater than for TIR . An excited fluorophore within the evanescent field close to a metal surface can excite a surface plasmon at the Stoke's shifted wavelength which, in turn, gives rise to emitted light radiated into the substrate over a narrow range of angles (Benner et al., 1979). The angular distribution of this emission is determined by the fluorophore emission spectrum and the surface plasmon dispersion characteristics . In Fig. 3 a theoretical emission profile for a rhodamine derivative has been calculated by integrating the surface field intensity profile across the fluorophore emission spectrum . This simplified analysis makes the assumption that the proportion of fluorophore emission coupled back into the substrate at any wavelength is related directly to the surface field intensity for excitation at that wavelength . As might be expected from the surface plasmon dispersion relationship, fluorescence emission is at a maximum at smaller angles relative to the surface normal compared with the excitation resonance angle. By integrating the emission signal over a narrow range of angles it is possible to reduce the background contribution from the radiated component of the fluorophore emission scattered into all angles within the substrate . Attenuation by the metal also helps to reduce this contribution .



204

J. W. Attridge et al .

4

Sample (ne)

1

Labelled Antibody

Layer 2 (n2, t2)

f &r .

Immobilized Antibody

f

Layer 1 (n 1 , t 1 )

Substrate (n)

I

Fig. 1. Schematic representation of hCG sandwich assay and optical configuration used for theoretical calculations .

45

1

. . . .. . . . . . . .. .. .. .. ... . . . . .. . . . . .. . . . . .



. . . . . . . .. . .

- 0 .8 35 m

30 - 0 .6

C

25-

a

W V

ii m U

7

20-

CD -

l0

0.4

15 10 5-

- 0.2

I

60

62

64

66

68

I

70

I

I

72

I

I

74

I

1

1

76

78

0 80

Angle of Incidence

Fig. 2. Theoretical surface field enhancement for SPR and TIR• n; = 1 .5171, n, = 0.053 + 13.968, n2 = 1.46, n1 = 1 .35, t2 = 10 nm, A = 594 .1 nm. SPR (t1 = 55 nm);- - - -, TIR (t, = 0 nm) SPR reflectance response.



205

Sensitivity enhancement of optical immunosensors TABLE 1

Theoretical comparison of TIR and SPR excitation methods Method

Angle (degrees)

Pd (pm)

TIR SPR

67-01 71-21

0.13 0.10

Field Figure of enhancement merit (pm -r) 2-00 42-11

15-38 421-10

immunoassay for human Chorionic Gonadotrophin (hCG) with a comparison being made between SPR and TIR excitation at each stage . A SPR reflectance immunoassay for hCG in serum was also performed for comparison with the fluorescence techniques . Instrumentation

Critical angle = 62.86° . Because of the signal enhancement it is also possible to optimise assays for sensitivity by reducing the concentration of fluorescently labelled conjugate (see the Discussion below) .

EXPERIMENTAL TECHNIQUES The enhancement of emission from fluorophore solutions using SPR excitation has already been demonstrated (Benner et al., 1979) but, to our knowledge, its application to immunoassay has not yet been reported. A series of experiments was undertaken which developed the earlier work with fluorophores solutions to immobilised fluorophore layers and ultimately to a full

Surface plasmon resonance reflectance measurements were performed using the system illustrated in Fig. 4. This is a modification of the configuration described by Oda & Fukui (1986) and enables scans to be taken over a 10° angle range using a single rotating element A 632-8 nm helium-neon laser was used for all the reflectance work. For indirect SPR and TIR fluorescence measurements the arrangement illustrated in Fig . 5 was used. The excitation assembly comprised : (a) a helium-neon laser (either 543 .5 nm or 594. 1 nm), (b) a light chopper, (c) a cube polariser to provide transverse magnetic (TM) polarised light, and (d) a short pass interference filter to block long wavelength emission from the laser . Light was refracted through a side face of a truncated hemicylindrical lens (BK7 glass, 25 mm radius of

Angle of Excitation/Emission

Fig. 3. Theoretical comparison of excitation and emission profiles for SPR enhancedfuorescence (dimensions and indices same as in Fig. 2): excitation ; , emission.

206

J. W. Attridge et al. polarising beamsplitter

rotating hemicylindrical lens

lens

HeNe laser

chopper

reference photodiode

lock-in amplifier

P

fixed lens and photodiode assembly

divider

00 Fig. 4. Experimental apparatus used for SPR reflectance measurementss Illumination Area Hemicylindrical lens

Bandpass filter Aperture Lens Photomultiplier

Stepper motor driven rotation of detector assembly Manual rotation of laser assembly

Fig. 5. Experimental apparatus used for SPR and TIR fluorescence measurements.

curvature) to give total internal reflection on the top planar surface. Capillary fill device (CFD) test cells, similar to those described by Badley et al. (1987) and Parry et al. (1990), containing the reagents under study, were coupled optically to the lens using index matching liquid (n d = 1 . 53). Comparisons between SPR and TIR excitation were made by choosing an appropriately coated capillary cell. Fluorescence emission coupled into the lens substrate was scanned in a plane at right-angles to that of the excitation light . The detection assembly comprised : (a) a bandpass interference filter, chosen for the fluorophore under study, which blocked any scattered excitation

light, (b) an aperture which defined an angular resolution of 0. 5° and (c) a lens which focussed the emission onto the photocathode of the photomultiplier detector (Hamamatsu 1P28A). Materials Mouse monoclonal antibodies against hCG were obtained and characterised as described by Siddle et al. (1984) . Rhodamine fluorophores were obtained from Molecular Probes Inc ., Eugene, USA. Rhodamine labelled antibody conjugates were prepared using the method described by Rattle et al. (1984). Standard solutions of hCG



207

Sensitivity enhancement of optical immunosensors 1 0.9 0.8 0.7 0 .6 m U

C fd

U a) a)

lr

0 .5 0.4 0 .3 0 .2 0.1 -

1

1

68

I

70

I

I

72

1

74

Angle of Incidence

, anti-hCG capture layer in Fig. 6. SPR reflectance assay for hCG in serum : bufferhCG + anti-hCG antibody in serum ; ---, anti-hCG antibody in serum .

were prepared from a freeze-dried preparation of hCG (Biodata SpA, Milan, Italy), calibrated against the first International Reference Preparation (75/537) . Horse serum was supplied by Seralab, Crawley Down, Sussex, UK Index matching liquid was supplied by McCrone Research Associates Ltd ., London NW3 5BG, UK Aminopropyltrimethoxysilane was obtained from Aldrich Chemical Co ., Gillingham, UK

EXPERIMENTS AND RESULTS An SPR sandwich assay for hCG in serum The problems associated with using direct SPR immunosensors in untreated biological samples were illustrated with an assay for hCG in serum . With a molecular weight of 40 000 Daltons, hCG is too small to generate a significant shift in resonance minimum on its own . A second antibody was used in a sandwich assay format to try to improve signal discrimination . CFDs (30 mm X 15 mm) with a volume of 35 pl were fabricated after a 55 nm silver layer had been evaporated onto one of the glass surfaces. Reflectance scans performed in air

buffer,• ,

established the uniformity of coating from cell to cell. A layer of monoclonal antibody against hCG was adsorbed onto the silvet surface of one CFD using a HEPES solution with an antibody concentration of 35 pg/ml . Reflectance scans indicated that within 20 s the resonance minimum had shifted by 0.41 ° compared with a CFD filled with buffer (Fig. 6). The shift in resonance can be attributed to protein adsorption as the resonance response remained unchanged after washing out the CFD with buffer. A chemically immobilised antibody layer will typically achieve a loading in the order of 0. 5 pg/cm2. From theoretical considerations, this translates to a resonance shift of about 0 . 8° . The magnitude of the shift seen in Fig. 6, therefore, suggests that a reasonable protein coverage has been achieved. Following washing through the CFD with 60pl of HEPES buffer (0. 1 M, pH 7.0), a solution of horse serum containing 2500 mIU/ml hCG (equivalent to 313 ng/ml or 7.44 nM) and a second monoclonal antibody against hCG at 10 pg/ml, was introduced into the CFD. A second CFD was treated in the same manner to the first except hCG was not added to the serum . Both cells showed a large shift in resonance position of 1 . 0° after a 15 min



208

J. W Attridge et al .

incubation period, which can be attributed to a combination of non-specific binding and the 0 . 01 increase in refractive index of serum with respect to buffer. The discrimination between the responses is, however, poor with the solution containing hCG showing a resonance minimum displacement of only 0. 08° from the zero standard (Fig. 6). It should also be noted that the reflectance responses from the two assays do not have the same shape (see Discussion). The second antibody used in this experiment was labelled with rhodamine allowing a comparison to be made between this technique and SPRF. Fluorophore solution concentration study Initial experiments were conducted to compare the sensitivity of SPR and TIR fluorescence excitation using solutions of rhodamine-6G in water/TWEEN. A CFD was used in which one of the inner surfaces had been half coated with 55 nm of silver then fully coated with 5 nm of silicon dioxide (Fig. 7, inset) . This ensured that the contribution from adsorbed fluorophore would be the same for both SPR and TIR excitation.

Light from a 543 . 5 nm helium-neon laser was fixed at the SPR angle and a fluorescence emission scan was performed for SPR excitation . The cell was then moved so that a TIR scan could be taken at the same angle of incidence, thus ensuring that the evanescent field profiles were the same for both measurements . The cell was then flushed through with increasingly concentrated fluorophore solutions with scans being taken at each concentration. The typical emission response (Fig. 7) agrees well with theory (Fig . 3). At the angle of peak emission, 71 . 75°, the fluorescence signal has been enhanced by a factor of 32 for SPR compared with TIR excitation. In Fig . 8 the signal as a function of fluorophore concentration has been plotted. The SPR signal was determined by integrating the emission response over 5° centred at 71 .75° . The TIR signal has been integrated over the same 5° range and also over a range of 20° . The greater sensitivity of the SPR technique is seen clearly; for SPR excitation the signal can immediately be discriminated from the background level and is consistently over an order of magnitude greater than for TIR excitation . In the latter case the signal does not become significantly larger than the background until the fluorophore

45

40-

Si0 2 Ag iii Glass

r

35-

30-

20 -

15 -

10 -

5 -

0

I

50

60

I

I

70

80

90

Angle of Emission

Fig. 7. Fluorescence emission comparison for TIR and SPR excitation of rhodamine-6G solution (1 .3 SPR; TIR

x 10-6 M).

.





209

Sensitivity enhancement of optical immunosensors 3.5 a 3-

2.5 -

7

2., c o

05

1 .5 -

Ile

1-

C

0.5 -

0-

-0.5

I

I

-10

I

1

-8

1

-4

-6 Log (Molar Concentration)

Fig. 8 Integrated fluorescence emission response as a function of rhodamine-6G concentration : -0 integration); , TIR (5° integration); ---0 , TIR (20° integration).

. SPR (5°

50 45 -

35 -

n

30 -

15 10 -

I

-7 .4

I

-7

I

I

-6.6

I

I

-6.2

I

I

-5.8

I

1

-5.4

1

-5

Log (Molar Deposition Concentration) Fig. 9. Comparison of fluorescence emission response from immobilised rhodamine as a function of deposition concentration: -p , SPR excitation; , TIR excitation .



210

J. W Auridge et al.

concentration has increased by a factor of about 50. Broadening the integration angle range for TIR excitation increases signal levels but does not improve sensitivity as the background signal level is also raised .

for SPR excitation. A significant decrease in signal is observed for SPR excitation at high concentrations due to quenching effects occurring at the surface . This effect is also observed with fluorophore solutions when absorption at the excitation wavelength becomes sufficiently large to disrupt the optimum resonance conditions .

Immobilised fluorophore study With the ultimate aim of performing a true immunoassay, the intermediate step from working with free fluorophore solutions is to consider the case when the fluorophore is chemically bound to the sensor surface directly rather than immunologically. To achieve this a layer of aminopropyl trimethoxysilane was formed on silver coated and plain glass plates . Solutions of an isothiocyanate derivative of rhodamine ranging in concentration from I X 10-5 m to 6. 3 X 10-8 m, were then coupled to the silanised surface of the plates which were then washed and dried before being fabricated into CFDs. SPR and TIR excitation scans were performed on ultrapure water filled cells using a 594 . 1 nm helium-neon laser (to match the longer wavelength rhodamine fluorophore used on this occasion) and concentration profiles for the two techniques were produced (Fig . 9). A typical signal enhancement factor of about 15 is observed

0

I

60

I

62

64

I

1

66

SPR and TIR fluoroimmunoassay for hCG in buffer The previous two experiments clearly illustrate the potential for the SPR fluorescence technique and the next step was to perform a full immunoassay for hCG using the model system (Fig . 1). Silvered and unsilvered capillary cells were incubated for 15 min with a 35 pg/ml solution of anti-hCG antibody in HEPES buffer. Following a HEPES wash the cells were then filled with solutions containing 2500 mIU/ml hCG to act as a top signal . Following a further wash through with buffer a rhodamine labelled anti-hCG antibody, which reacts with different hCG epitope than the first antibody, was flushed through the cell. Angle scans were performed after a 15 min incubation. The experiment was then repeated without hCG to dermine a zero signal . The results (Fig . 10) again show a significant

1

68

1

1

1

72 70 Angle of Emission

I

1

74

76

I

78

80

Fig. 10. Comparison of emission responses from SPR and TIRfluoroimmunoassays for hCG in HEPES buffer. , SPR (2500 mtUlml), - - -, SPR (0 miU/ml),• TIR (2500 miUlml);• TIR (0 m1Ulml).



Sensitivity enhancement of optical

211

immunosensors

enhancement of signal obtained with the SPR technique. Signal integration yields a 15-fold enhancement of the signal-to-background ratio for SPR over TIR excitation . SPR and TIR fluo in serum

TABLE 2 Summary of results for TIR and SPR fluoroimmunoassays for hCG in serum Method

unoassay for hCG

The above experiment was repeated using horse serum as the sample matrix . However, to give a closer comparison with a true sensing system, the hCG and labelled antibody were premixed in the serum before being introduced into CFDs in which the capture antibody had been previously adsorbed . Three or four SPR and TIR cells were scanned for each of seven hCG standard concentrations after 15 min incubation without subsequent wash or separation steps . Emission profiles of similar form to those shown in Fig. 10 were obtained. Standard curves were generated by integrating the emission signals from 69 .68° to 72.68° . A comparison of SPR and TIR excitation techniques clearly shows that the signal enhancements observed in the previous experiments translates to a significant improvement in assay performance (Fig. 11). The assays' sensitivities (calculated by determining the concentration equivalent to the

Standard Sensitivity Sensitivity (net) signal deviation (mlUlml) Zero 71 .6 97.3

TIR SPR

2.2 1.6

500