| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Oak Ridge Conference |
1 Protein Science Department, Genomics Institute of the Novartis Research Foundation, San Diego, CA.
2 Department of Chemistry and the Skaggs Institute for Chemical Biology, The Scripps Research Institute, La Jolla, CA.
aAddress correspondence to this author at: Protein Science Department, Genomics Institute of the Novartis Research Foundation, 10675 John Jay Hopkins Dr., San Diego, CA 92121-1125. Fax 858-812-1746; e-mail geierstanger{at}gnf.org.
Abstract
Background: Conceptionally, antibody microarrays are simply multiplexed sandwich immunoassays in a miniaturized format. However, from the amounts of capture antibodies used, it is not apparent whether such assays are ambient analyte (Ekins. Clin Chem 1998;44:2015–30) or mass-sensing devices (Silzel et al. Clin Chem 1998;44:2036–43). We evaluated multiplexed microarray sandwich assays for 24 mouse serum proteins in these terms within the boundaries of our experimental setup and based on theoretical considerations of the law of mass action.
Methods: Capture antibodies for 24 mouse serum proteins were printed on planar microarray substrates. After incubation with mixtures of purified antigens for 1 or 18 h, mixtures of biotinylated detection antibodies were used. High assay sensitivity was achieved by use of resonance-light-scattering particles for signal generation. Titration curves were generated for assay volumes of 20, 40, and 80 µL, and detection limits were calculated and compared. The assays were modeled theoretically based on the amounts of capture antibodies and the assay volumes used.
Results: As predicted, experimental variations of the assay volume by up to fourfold did not appreciably affect detection. Even for the most sensitive assay, <2% of the analyte molecules present in the sample were captured and generated signal at the detection limit. However, increasing the sample incubation time from 1 to 18 h on average lowered the detection limit threefold.
Conclusions: In our experimental setup, all 24 sandwich microarray assays fulfill the criteria of the "ambient analyte" regime because depletion of analyte molecules from the assay volume is insignificant.
Over the last several years, protein and antibody array platforms have been developed as sensitive, multiplexed, and miniaturized assaying tools for proteomics (1)(2)(3)(4)(5)(6)(7)(8)(9)(10). In the process, several configurations of antibody microarrays have been implemented: Chemically labeled lysates can be applied to capture antibody arrays in a two-color ratiometric approach to obtain relative changes in protein abundance with respect to a reference sample (11)(12)(13)(14). Alternatively, arrays of spotted samples can be probed with individual antibodies, an approach that is particularly powerful for the detection of protein modifications such as phosphorylation (15). Microarrays of antigens have been used successfully for the detection of autoantibodies in the classification of different autoimmune diseases (16)(17).
Perhaps the most relevant approach for clinical diagnostics has been the development of antibody microarrays in sandwich format [for a recent review, see Ref. (18)]. The advantage over standard sandwich plate assays is that tens of analytes can be measured in parallel with sample amounts that in a traditional setup are sufficient only for a single measurement. We and others (19) have found that for practical reasons, largely based on the intrinsic low-level cross-reactivity of even the most specific antibodies, multiplex sandwich assays are limited to the parallel measurement of up to 50 analytes. Therefore, platforms that are designed for the simultaneous processing of several medium-sized microarrays arranged in standardized microtiter plate format will have practical advantages in terms of throughput and the flexibility to subdivide assays when necessary (18).
It has long been realized that miniaturized array-based assays should be more sensitive than conventional microtiter plate immunoassays (20)(21)(22) because the highest fractional occupancy of the capture antibody and highest signal per area can be achieved in microspots [for a recent summary of these concepts, see Ref. (10)]. Noncompetitive immunoassays such as sandwich microarray immunoassays can in general be classified as "ambient analyte" or "mass-sensing" assays (20)(23). The distinction between these two classes is based on the question of whether binding of the analyte to the immobilized capture antibody leads to a significant change in the concentration of the analyte in the sample solution. For the ambient analyte regime, which according to Ekins (20) is fulfilled for capture antibody concentrations [cAB] < 0.01/Ka, the fraction of bound analyte is <1.5%. This cutoff is somewhat arbitrary because the fraction of bound analyte steadily increases with increasing capture antibody concentration. In a truly mass-sensing device, the fraction of bound analyte reaches 1, i.e., the assay becomes an "affinity capture device". Although Silzel et al. (23) have shown evidence for mass sensing by protein microarrays, Ekins (20) has argued that immunoassays in microarray format, and microarray assays in general, should be considered ambient analyte assays. The distinction is relevant in terms of assay behavior. Although changes in assay volumes will have little effect on the detection limit or the analyte concentration reading of an ambient analyte assay, a mass-sensing device will yield a signal proportional to the total amount of analyte present in the sample solution. Variations in assay volume could be a source of experimental uncertainty in mass-sensing devices but should not affect the precision or the detection limit of ambient analyte assays. Detection limits of ambient analyte assays can be improved with higher-affinity antibodies, but for a given assay volume and capture antibody concentration, increasing affinity will eventually move an ambient analyte assay into the mass-sensing regime.
To detect the low number of bound molecules present in a microarray sandwich assay, labels of high intrinsic sensitivity are required. Resonance light scattering (RLS) gold particles (24)(25) coated with an anti-biotin antibody are extremely bright labels because they scatter white light back in a narrow wavelength range (26). Using RLS particles, we have previously achieved practical detection limits of as low as 0.3 ng/L, or 17 fmol/L with 10-µL samples in multiplexed measurements with as many as 45 different analytes (Saviranta et al. and Brinker et al., manuscripts in preparation). In this communication, we correlate calculations based on the law of mass action to microarray measurements of protein calibrators with different assay volumes and incubation times to determine whether sandwich microarray measurements in our format are operating in the ambient regime or are mass sensing. Such detailed understanding will have consequences for future improvements in microarray immunoassays.
Materials and Methods
The details and applications of our microarray platform as well as the optimization of an antibody microarray for the detection of 24 mouse serum proteins will be described in detail elsewhere (Saviranta et al., manuscript in preparation). In short, solutions of individual capture antibodies are deposited by contact printing as subnanoliter size droplets on amino-silane-coated, planar glass slides (GAPS II slides; Corning Inc.), producing spots 90–120 µm in diameter. The slides are blocked with a high-concentration protein solution and stored dry. Before the assay, which is outlined in Scheme 1, the slides are mounted in a slide holder that separates 48 subarrays on each slide into 3 x 3 mm areas, effectively creating a 384-well-type plate with microarrays of up to 48 different capture antibodies at the bottom of each well. The wells are washed and rehydrated by use of a modified plate washer before 20–80 µL of sample solution is added. After a 1- or 18-h incubation with rigorous, high-speed shaking, samples are washed away, and mixtures of biotinylated detection antibodies are added and incubated for 1 h (room temperature with shaking). After another wash step and a brief reblocking step, RLS particles, in essence colloid gold particles coated with an anti-biotin antibody (Invitrogen), are added for 1 h followed by a final wash step and the application of a polymer coating for improved scattering properties. The slides are then imaged at a resolution of 10 µm with a 16-bit charge-coupled device camera-based scanner optimized for light-scattering measurements (Invitrogen; in these particular experiments, images were obtained with an exposure time of 0.35 s). Images are then analyzed by ArrayVision, Ver. 8.0 (Imaging Research). RLS signals are reported as the median-trimmed mean signal density value for each spot without subtraction of adjacent local background readings.
Calibration curves (dose–response curves) were produced by use of combined data from eight independent 8-point titrations, performed on four slides with two replicate wells per slide for each analyte concentration. Thus a total of 64 data points were generated for each calibration curve. A data point represents three replicate spots in a single well, being the median of the measured signals. However, individual signals differing from the median by a factor of 1.5 or more were excluded as outliers; in these cases the mean for the remaining spots was used instead. The data were fitted to the four-parameter logistic equation: y = bottom + (top – bottom)/{1 + 10^[(logEC50 – x)^HillSlope]} by nonlinear fitting with 1/y2 weighting, using the Prism 4 for Windows (GraphPad Software, Inc.). The detection limit was defined as the analyte concentration corresponding to a signal 3 SD above the background signal, as calculated from the calibration curve.
To convert the measured RLS signal to the number of particles bound, we scanned a calibration slide with serial dilutions of RLS particles (Invitrogen) at the same exposure time as used for imaging the experimental slides. At low particle densities, there is a linear relationship of y = 1.15x + 25, where y is the RLS signal and x is number of particles per spot. A RLS signal of 150 corresponds to 
0.01 RLS particles per µm2.
For the theoretical calculations, the amount of the immobilized capture antibody is estimated as follows: Each printed droplet is 0.6 nL in volume (SMP3 pins; www.arrayit.com), and the antibody concentration is typically 1.6 µmol/L; thus, 1 fmol of antibody is deposited per spot, i.e., 3 fmol per array with three replicate spots. Because the amount of actively immobilized antibody is less than that, calculations are performed with the assumption that either 100% or 1% of the deposited antibodies remain able to bind an antigen. We further assume that equilibrium is reached and that the immobilized antibodies sample the whole assay volume. Because detection antibodies (most of them at 0.5 mg/L) and the RLS particles are applied in excess, the binding equilibria of detection and anti-biotin antibodies on the RLS particles can be ignored, and only the initial binding equilibrium of analyte and immobilized capture antibody is modeled. Finally, it is assumed that the antibodies used in our antibody microarray for mouse serum proteins are of high affinity with slow off-rates; thus, the dissociation of antibody–antigen complexes is not considered in the analysis.
Results
theoretical considerations of antibody microarray measurements
Sandwich microarray assays are multistep assays with several wash and incubation steps (Scheme 1
). Only analyte molecules that remain bound to the capture antibody throughout the process will generate an observable signal. Therefore, the detection limit is determined by the number of bound molecules, by the amount of signal generated per binding event, and by the ratio of signal vs background caused by nonspecific binding and other experimental contributions, such as instrument noise. In practice, reducing the background is most crucial and often requires intensive optimization of the blocking and wash conditions as well as optimization of reagent concentrations. This poses a special challenge for multiplexed sandwich assays as is discussed in detail elsewhere (Saviranta et al. and Brinker et al., manuscripts in preparation).
|
Previous antibody microarray experiments indicated that <100 000 molecules in a sample can be detected with use of RLS particles as labels (Saviranta et al. and Brinker et al., manuscripts in preparation). Detection limits of these multiplexed sandwich assays ranged from 15 x 10–15 mol/L to 1 x 10–10 mol/L. In our microarray format, <3 x 10–15 moles of capture antibody are in contact with 20–80 µL of sample at room temperature and under rigorous shaking. To model the assay behavior in the analyte-binding step (Scheme 1
) under different conditions, we used the law of mass action to calculate the number of analyte molecules bound to all three spots (Fig. 1
, A and C), the fractions of analyte molecules bound to the capture antibody, and the fraction of capture antibody occupied by an analyte molecule (Fig. 1
, B and D). In these calculations, it is assumed that the binding reaction approaches equilibrium and that the amount of active capture antibody is either 3 x 10–15 moles, which is the maximum possible, or 1/100 of this amount, which we arbitrarily used as an estimate of the active capture antibody molecules that are capable of binding an analyte molecule while attached to the surface. This estimate seems reasonable; because of orientation effects, denaturation, and steric hindrance, typically only a small percentage of the surface-attached capture antibodies will be able to form a sandwich with the analyte and the detection antibody (27)(28), and multiple wash steps in the preassay processing of the microarrays may remove weakly bound antibodies as well. Furthermore, to illustrate the general behavior of a noncompetitive immunoassay, the calculations are performed for different binding affinities of the analyte for its capture antibody as well as for assay volumes of 20, 80, and 300 µL (Fig. 1
, A and C).
|
In general, the assay becomes more sensitive, i.e., the same number of molecules are bound at lower analyte concentrations, as the affinity of the capture antibody increases and as the number of capture antibody molecules increases. For 3 x 10–15 moles of capture antibodies (Fig. 1
, A and B), the criterion for ambient analyte regime of [cAB] < 0.01/Ka (where Ka is the affinity constant) is fulfilled for capture antibodies with a Ka of 108 L/mol or less (at 20-µL assay volume). The fraction bound for an antibody of Ka of 1010 L/mol reaches 0.6 for analyte concentrations <10–11 mol/L (Fig. 1B
), i.e., 60% of all analyte molecules in a sample are bound to the capture antibody. Under these conditions, a microarray assay would deplete analyte molecules in the sample, as is also clearly indicated by the substantial increase of bound molecules with increasing assay volume (Fig. 1A
) for Ka values of 1010 and 1012 L/mol. For the latter, the number of bound molecules increases 3.9-fold for an increase in assay volume from 20 to 80 µL and 14-fold for an increase in assay volume from 20 to 300 µL.
For the "1% active" capture antibody regime (3 x 10–17 moles of actively immobilized antibody), the numbers are shifted by two orders of magnitude (Fig. 1
, C and D), and the cutoff for the ambient analyte regime is therefore at a Ka of 1010 mol/L. The assay volume has little effect on the number of bound molecules for Ka values as high as 1011 L/mol (Fig. 1C
). Only for a Ka of 1012 L/mol is an increase in the bound molecules apparent: 1.8-fold for an increase in assay volume from 20 to 80 µL and 2.3-fold for an increase in assay volume from 20 to 300 µL (identical to Ka = 1010 L/mol at 3 x 10–15 moles of capture antibody). For a Ka of 1011 L/mol, only 13% of all analyte molecules in the sample are bound to the capture antibodies, whereas this number increases to 60% for Ka = 1012 L/mol (Fig. 1D
).
The amount of active capture antibody largely distinguishes the assays in terms of ambient analyte or mass-sensing assays. This amount is hard to estimate, however. Once an assay is near mass sensing, the detection limit cannot be improved with a higher-affinity capture antibody because all or most analyte molecules are bound already (Fig. 1
, B and D). At that point, the detection limit can be improved only by increasing the number of capture antibody molecules (or viable binding sites; Fig. 1
, A vs C) or by increasing the sample volume (Fig. 1A
). For assay volumes between 20 and 300 µL, little difference is expected for Ka values of 1 x 1011 L/mol or less for the scenario in which 1% of the capture antibody molecules are active (Fig. 1C
). Only for high-affinity antibodies with a Ka of 1 x 1012 L/mol is there an improvement, but again because the fraction of bound analyte molecules approaches 1 under these conditions, the total number of molecules in the sample is limiting. Assuming that we in fact operate near equilibrium and near the 1% active regime, we expected little improvement with assay volume, perhaps with the exception of assays with antibodies of highest affinity.
antibody microarray assays with different assay volumes and incubation times
To test these predictions we performed and show data from microarray experiments with eight slides. On each slide, 48 identical microarrays are printed (Fig. 2A
), and each of these arrays is incubated with a mixture of recombinant protein calibrators diluted to seven different concentrations in diluent or with just the diluent. On each slide, two such 8-point titrations are performed with assay volumes of 20, 40, and 80 µL. Calibrator mixtures were incubated at room temperature with rigorous mixing for 1 or 18 h on four slides each, giving a total of eight titrations for each condition and each of the 24 analytes. Raw images (Fig. 2B
) suggest that an incubation time of 18 h produces a significantly increased signal. Titration curves of six representative analytes are shown in Figs. 3
and 4
. Fig. 3
shows the effect of the assay volume on the titration curves at either a 1-h (Fig. 3A
) or 18-h (Fig. 3B
) incubation time. The mean signal with standard deviation and a fitted curve derived from eight titrations for an assay volume of 20 µL (Fig. 3
, 
) is compared with that of the 80-µL assay (Fig. 3
, 
). The signal in the absence of any added analyte is displayed at the molar equivalent of 0.001 ng/L because of the log/log scale. With a 1-h incubation time (Fig. 3A
), the analytes interleukin-1B (IL1B), IL4, IL5, and IL6 exhibit significantly higher signals in the 20-µL assays compared with the 80-µL assays, whereas the analytes IL3 and troponin C appear not to be affected by the assay volume because their calibration curves completely overlap. No difference in background signal is observed between the two assay volumes. Increasing the incubation time to 18 h led to essentially overlapping curves for both assay volumes with all analytes (Fig. 3B
).
|
|
|
The data presented in Fig. 3
clearly indicate that there is little difference in the calibration curves for different assay volumes if the incubation time is sufficiently long, and 1 h is apparently not long enough for all analytes to reach equilibrium. Because the 80-µL assay appears to achieve equilibrium later, the results imply that the mixing of the larger volume in a 384-well microtiter plate is not as efficient as for an assay volume of 20 µL.
The effect of the incubation time on the titration curves is shown in Fig. 4
with an assay volume of either 20 µL (Fig. 4A
) or 80 µL (Fig. 4B
). The assays with an 18-h incubation time (Fig. 4
, ) generally have higher background signals than the 1-h assays (Fig. 4
, ), but the signals at each titration point are increased even further. In 11 of the 24 assays, this increased signal at a longer incubation time translated into an at least twofold improved detection limit (Fig. 5
and Table 1
). The detection limit throughout this discussion is defined as the analyte concentration corresponding to the signal in the absence of analyte plus 3 SD. The data in Fig. 5
show that the detection limit varies little with assay volume for each of the analytes for an incubation time of 18 h, whereas for the 1-h incubation time, titration curves for 80- and 40-µL assays generally seem to be lagging those of the 20-µL assay. The assays for cytochrome C failed at incubation times of 18 h, presumably because of degradation of the analyte. For 11 analytes, the longer incubation time improved the detection limit between 2- and 15-fold (Table 1
). The improvement was, on average, threefold for all 24 analytes and did not depend on the sensitivity range of the assay in general. The best assay, IL4, did not significantly benefit from an increased incubation time, possibly because it consists of very high-affinity antibodies. Its detection limit of 3 fmol/L (0.04 ng/L) translates into the detection of 36 000 IL4 molecules in an assay volume of 20 µL at a 1-h incubation time (Table 1
).
|
|
By calibrating the RLS signal to the number of particles (see Materials and Methods), we were able to estimate the number of bound antigen molecules at the detection limit. For IL4 (1-h incubation for a 20-µL assay), 3 SD of the background signal was 180, which corresponded to roughly 400 RLS particles on the three spots in an assay well. For IL1B, with 3 SD of 70, the detection limit corresponded to as few as 120 RLS particles. Assuming that each RLS particle corresponds to a single analyte-binding event, as few as 400 IL4 molecules of the 36 000 in the sample, or 1.1%, remained bound to the capture antibody spots throughout the assay procedure.
Discussion
Microarray measurements have been argued to be intrinsically more sensitive than microtiter well assays because they achieve the highest possible surface density of bound analytes (20)(21)(22). When the capture antibody concentration in the assay stays below 0.01/Ka, the free analyte concentration at equilibrium approaches the total analyte concentration in the sample, ensuring a maximum fractional occupancy of the antibody. In conventional microtiter plate assays, the goal is to collect as much analyte as possible to maximize the total signal, but this requires a high capture antibody concentration, which leads to a decreased fractional occupancy. Because the nonspecific background binding of the label is directly proportional to the amount of capture antibody present, a lower fractional occupancy also produces an inferior signal-to-background ratio. In the process of printing antibody microarrays, we use very high concentrations of capture antibody, and it was therefore not clear whether multiplexed sandwich assays would behave according to the ambient analyte principle (20) in the context of our microarray platform, especially because at least one report claimed mass-sensing conditions for a microarray assay (23). Here we characterized the behavior of 24 multiplexed sandwich assays performed in our microarray platform and show that they indeed operate in or near the ambient analyte regime.
To understand the differences between microarray assays within the experimental constraints of our platform and standard microtiter plate assays, it is helpful to compare the concentrations of antibodies in both: In a 384-well plate with a high coating capacity (e.g., Nunc Maxisorp plates at 650 ng/cm2), the capture antibody concentration can be on the order of 10–7 mol/L in a 20-µL assay volume. With antibody Ka values typically larger than 109 L/mol, these assays capture virtually all analyte molecules present in low-concentration samples, even if only a small percentage of the immobilized antibodies generally are active (27)(28). Within our current capabilities to manufacture microarrays and our desire to have three spots per assay and array to perform triplicate measurements, at most 3 x 10–15 moles of a given capture antibody can be present in the same volume of 20 µL in the microarray-based assays (see Materials and Methods). Depending on the fraction of actively immobilized antibodies, the binding site concentration in the assay can vary between 1.5 x 10–10 mol/L (100% active) and 1.5 x 10–12 mol/L (1% active), with the latter value, in most cases, being closer to reality. Because Ka values of antibodies typically range from 109 to 1010 L/mol, microarray-based assays are therefore unlikely to collect a significant proportion of the analyte molecules (see also Fig. 1
). The experimental results showing that an increased sample volume neither substantially increased the signal strength nor significantly improved the detection limit (Figs. 3
and 5
) support this conclusion. Furthermore, in a separate experiment it was shown that the signal intensities of individual spots were not affected by the number of spots of the same specificity in the well (data not shown). Finally, a direct estimation of the number of molecules bound in the IL4 assay at detection limit revealed that 1.1% of the total analyte molecules present in the sample were captured and retained for detection. For all other assays, the fraction of bound analyte molecules was even lower. Our assays thus appear to be consistent with the "1% active" rather than the "100% active" antibody scenario in the theoretical modeling (Fig. 1
) and to fall into the ambient analyte immunoassay regime of Ekins (20), i.e., they capture only a minor to vanishing fraction of the total analyte molecules present.
In a microarray assay, the number of bound and signal-generating analyte molecules will generally be lower than in a microtiter well assay, but they will all be concentrated into a few spots of high signal density. Indeed, there is a 3000- to 6000-fold difference in the surface area between a coated well (20 µL) in a 384-well plate and a single 90- to 120-µm diameter spot in the microarray. The background signals of the microarray assays in this study were, on average, 10-fold higher than the instrument background (i.e., internal noise of the scanner); thus, there is room for improving the sensitivity by reducing nonspecific binding. With the extremely high signal intensity of the RLS particles, only 15 particles per spot are required to produce a signal twice as strong as the signal from scanner noise. If nonspecific binding could be completely eliminated, one would potentially be able to detect 45 bound molecules on three spots, assuming that 1 particle corresponds to 1 bound analyte molecule at low concentrations. In a mass-sensing device, 45 molecules in a 20-µL assay volume would correspond to a detection limit of 3.7 x 10–18 mol/L. This would match the detection limit recently claimed for a protein detection scheme that combines immunocapture of prostate-specific antigen with PCR amplification of DNA bio-bar codes on nanoparticles followed by DNA detection with colloid gold particles (29). For an ambient analyte assay with 1.5% analyte depletion, the respective concentration would be 2.3 x 10–16 mol/L, which is 10-fold lower than our currently best detection limit. This discussion implies that the background caused by cross-reactivity between detection antibodies with the capture antibodies of a different analyte and background caused by other experimental variables is crucial in assay optimization.
In the absence of antibodies with higher binding affinities, the sensitivity of a microarray assay could be improved further only by significantly increasing the loading of active capture antibody on a smaller surface area. Spots smaller than 80 µm can currently be obtained only with surface-engineered substrates and a very specialized array printing process (30). Our antibody microarrays are prepared by passive adsorption and cannot provide the high binding capacity necessary for mass sensing. In fact, to retain all 3 x 10–15 moles of capture antibody deposited in three microarray spots, a loading capacity of between 1400 and 2500 ng/cm2 would be required, which is two- to fourfold above that of high-capacity plastic plates. Interestingly, Silzel et al. (23) reported to have achieved a coating density of 3500 ng/cm2 of active avidin on polystyrene films with the aid of covalent photo-cross-linking. With this high binding capacity, they could perform a true mass-sensing microarray measurement, i.e., to harvest and detect virtually all of the analyte molecules instead of just measuring their concentration. However, the results were based on titration with a small molecule, a near-infrared dye-labeled biotin derivative, and they cannot be directly translated to sandwich assays with protein antigens.
Attempts to improve the number of active capture antibodies through directional attachment strategies yielded little benefit for full-length antibodies but increased the number of active binding sites up to sixfold for Fab fragments (31). Three-dimensional "hydrogel" or gel matrix surfaces (32)(33)(34)(35) may have substantial advantages, but 100-fold improvements over planar surfaces may not be realistic. Multiplexed assays in microflow cells (36) or bead-based assays (37) may have advantages in terms of efficiency of mixing, but to date, to the best of our knowledge, suspension arrays have not reached the detection limits described here. Ultimately, the nonspecific background signal caused by antibody–antibody cross-reactivity will limit the sensitivity of any multiplexed immunoassay. Focusing on reducing nonspecific background caused by cross-reactivity and maintaining consistent spot morphology will likely yield the most dramatic improvements in the sensitivity and reliability of microarray immunoassays.
In conclusion, within the layout of our protein array platform, we showed that RLS particles are very sensitive labels for multiplexed sandwich immunoassays. The best assay detects 36 000 molecules in a sample of protein calibrators, with on the order of 400 analyte molecules bound and generating signal. In the future, we will further evaluate the assays in terms of analyte recovery with serum samples to which analyte is added. The experiments presented here were intended to probe the fundamental behavior of antibody array assays. None of the assays is depleting the analyte population in solution to a large extent and can therefore be considered to operate in the ambient analyte regime (20). The detection limits of these assays could therefore be improved with higher-affinity antibodies or by increasing the number of active capture antibodies in equilibrium with the sample. This will require surfaces with higher binding capacities and a significantly higher fraction of antibodies that retain their binding activity on surface attachment. To achieve the best sensitivity, a long sample incubation is required, implying that, contrary to the predictions for ambient immunoassays, the equilibrium is not reached instantaneously in our particular setup.
Acknowledgments
We thank Drs. Kate Rhodes and Steve Roman of Invitrogen (Carlsbad, CA) for helpful discussions, and Prof. Dr. Peter G. Schultz for continuous support. We thank the Academy of Finland for financial support (Grant 101064 to P.S.).
References
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  A. Claudon, P. Vergnaud, C. Valverde, A. Mayr, U. Klause, and P. Garnero New Automated Multiplex Assay for Bone Turnover Markers in Osteoporosis Clin. Chem., September 1, 2008; 54(9): 1554 - 1563. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Kverka, J. Burianova, R. Lodinova-Zadnikova, I. Kocourkova, J. Cinova, L. Tuckova, and H. Tlaskalova-Hogenova Cytokine Profiling in Human Colostrum and Milk by Protein Array Clin. Chem., May 1, 2007; 53(5): 955 - 962. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W. Kusnezow, Y. V. Syagailo, S. Ruffer, N. Baudenstiel, C. Gauer, J. D. Hoheisel, D. Wild, and I. Goychuk Optimal Design of Microarray Immunoassays to Compensate for Kinetic Limitations: Theory and Experiment Mol. Cell. Proteomics, September 1, 2006; 5(9): 1681 - 1696. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Sanchez-Carbayo Antibody Arrays: Technical Considerations and Clinical Applications in Cancer Clin. Chem., September 1, 2006; 52(9): 1651 - 1659. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. B. Haab Antibody Arrays in Cancer Research Mol. Cell. Proteomics, April 1, 2005; 4(4): 377 - 383. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. Y. H. Lin, M. Sabri, J. Alirezaie, D. Li, and P. M. Sherman Development of a Nanoparticle-Labeled Microfluidic Immunoassay for Detection of Pathogenic Microorganisms Clin. Vaccine Immunol., March 1, 2005; 12(3): 418 - 425. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
; G-CSF, granulocyte colony-stimulating factor; GM-CSF, granulocyte macrophage-colony-stimulating factor; TropC and TropI, troponin C and I, respectively; CytC, cytochrome c; TNFa, tumor necrosis factor 
; APO A2 and APO C3, apolipoprotein A2 and C3, respectively; MCP1, monocyte chemoattractant protein 1; TPA, tissue plasminogen activator.