The major bottleneck in translating newly discovered biomarkers to the clinic is the validation of these discoveries in controlled clinical samples . We and others are developing a novel high-throughput platform to meet the challenge of biomarker validation in human blood samples, using reverse-phase protein microarrays (RPPMs). Although RPPMs for cell and tissue lysates are used extensively and protocols are available [18–20] there are few reports on the use of RPPMs for serum and plasma samples and no specific protocol has been defined [8, 9, 21, 22]. In the present study we developed an optimal spotting protocol, including the choice of a printing buffer and surface chemistry, for the printing of plasma and serum in the reverse array format, and we assessed the applicability of RPPMs for the measurement of a mid-abundant plasma biomarker, clusterin.
The type of substrate and printing buffer used for microarray fabrication, in addition to sample properties such as viscosity and surface tension, have a major influence on array quality parameters including spot morphology, background and protein affinity binding . In the preparation of reverse phase protein microarrays for cell and tissue lysates, nitrocellulose remains the most commonly used substrate [24–27]. This surface allows proteins to bind non-covalently (hydrophobic interaction) to the surface and also provides high binding capacity allowing the detection of very low abundant analytes when using catalysed signal amplification detection methods [5, 18]. However, one of the drawbacks of nitrocellulose is its high intrinsic fluorescence that results in high background levels, thereby limiting the sensitivity of the assay when fluorescently labelled antibodies are used for detection . In our hands, plasma binding to nitrocellulose was limited, and, together with the higher background levels, it resulted in a lower dynamic range for the measurement of an endogenous protein, clusterin, when compared to epoxy coated slides. Although nitrocellulose performs very well for printing denatured cell and tissue lysates, we did not find it yield satisfactory results for serum and plasma RPPMs when using a fluorimetric detection method compared to epoxy, therefore no further development to deal with the autofluorescence issue of nitrocellulose was pursued. We also found that MaxiSorp™ black polymer slides, which perform well in forward-phase antibody microarrays [29–31], showed the poorest binding results and spot morphology with both of the buffers tested on our reverse phase arrays, suggesting that plasma samples may be incompatible with the hydrophilic chemistry of the substrate. Our best results were obtained when epoxy coated slides and non-denaturing PPB buffer were used in combination. Epoxy surfaces offer several advantages including covalent immobilization, a more diffuse distribution of the proteins on the slide, and low non-specific background [32, 33]. The more diffuse binding pattern of proteins on the slide may increase the chance of the antibody binding to its epitope since blood proteins are spotted in their native conformation with the PPB buffer.
In the present study we provide proof of principle for the detection of human endogenous clusterin, a mid-abundant blood protein, in plasma and serum using RPPMs and fluorescently labelled secondary antibodies. The sensitivity and specificity of any antibody based assay is largely dependent on antibody performance. We found that the polyclonal antibody targeted against clusterin showed a dynamic range 2-fold greater than that of the monoclonal antibody. Interestingly, spot variability was also reduced when slides were probed with the polyclonal antibody. It is worthwhile noting that both antibodies had demonstrated the pre-requisite specificity, recognizing a single band at the expected molecular weight in western blot analysis of the same complex sample used for the RPPM experiment. As with any use of reverse phase protein microarrays, careful screening of the specificity of antibodies is necessary; in addition, we suggest screening for dynamic range before selecting the optimal antibody to be used on the RPPM platform.
The sensitivity of any biomarker validation technology needs to be sufficiently high to enable the detection of low abundant protein biomarkers. Here we report a limit of detection (LOD) in the ng/mL range (780 ng/ml), which is approximately 100 fold less than the plasma levels of the endogenous mid abundant serum/plasma protein clusterin, using RPPMs in the absence of signal amplification detection methods. Although the LOD reached was essentially identical to that of the ELISA assay used in our validation studies (750 ng/mL), the sensitivity of our RPPM platform was significantly higher if we consider that we could detect this low amount of analyte from only 0.7 nL of biological sample. Our results suggest that the RPPM platform can be used for the high throughput and reproducible detection of mid abundant blood protein biomarkers. It is possible that the limit of detection can be lowered using different amplification techniques such as TSA amplification or the use of nanoparticles such as quantum dots. However, amplification methods were clearly not required for the accurate detection of clusterin on our platform.
We focused the development of our platform for the detection of clusterin. Blood clusterin levels have been shown to change in different disease states including numerous cancers  and systemic lupus erythematosus  as well as in diabetic type II patients and in patients with developing coronary heart disease, or myocardial infarction . Preliminary evidence also suggests that measurement of clusterin in serum and plasma may be useful for the early detection of colorectal cancer  and for monitoring pVHL-defective renal carcinomas . Moreover, changes in clusterin levels are being studied as surrogate biomarkers for treatment efficacy in clinical trials of anti-clusterin therapy in breast, lung and prostate cancer [39, 40]. In our platform, the minimal difference of 8 μg of clusterin that can be detected between samples is clearly in the range of clinically relevant changes in plasma clusterin. Moreover, the good correlation with ELISA values and the low spot to spot variability and array to array variability observed when large numbers of clinical samples are screened further demonstrates that RPPMs for the detection of clusterin hold potential for clinical applicability. We feel that these results are all the more remarkable as they were obtained using only samples from healthy volunteers, in which clusterin levels showed very limited variation. In fact, our RPPM platform was useful to assess whether clusterin levels changed when using different blood processing protocols. We could observe a relationship between the levels of clusterin and the type of blood collection tube used. We used CTAD tubes which contain citrate-theophylline-adenosine-dipyridamole and inhibit platelet activation . As clusterin is highly expressed in platelets , inhibiting platelet activation and thus the release of clusterin from platelets, is likely to explain the lower clusterin levels observed in CTAD plasma compared to serum. Our results are similar to those from a previous study which had recommended the use of citrate plasma for the measurement of clusterin , but this is, to our knowledge, the first report measuring plasma clusterin levels in CTAD tubes. Furthermore, the comparison of both CTAD specimens from the same individual showed that the levels of clusterin do not change from the first and third blood draw, thus suggesting that, when measuring clusterin, it may not be necessary to discard the first tube as sometimes recommended in blood collection guidelines. The results from our RPPM analysis and further validation with ELISA assays further confirm that blood collection protocols require careful standardization and that the differences in serum and plasma levels need to be taken into consideration in the course of clinical studies measuring clusterin levels.
Ideally, a clinically useful biomarker should be detected in serum or plasma. Prior to the clinical adoption of biomarker-based tests, putative biomarkers need to be rigorously validated in thousands of clinical samples. With RPPMs it is possible to spot hundreds if not thousands of serum and plasma samples on a single slide and analyze a biomarker of interest in one single experiment. Although the platform is presently limited by antibody availability and specificity, this is likely to change in the near future thanks to global initiatives such as the Human Antibody Initiative from HUPO and the Swedish Human Proteome Resource (HPR) program, which focus on the development of high quality antibodies against all human proteins .
One alternative method for biomarker validation is multiple reaction monitoring mass spectrometry (MRM-MS) . This method is very attractive since it does not require antibodies for the detection and measurement of biomarkers; however, it has not yet reached the required sensitivity for low abundant proteins in complex samples such as serum and plasma. Moreover, although many biomarkers can be simultaneously monitored in a sample, it has a limited throughput since only one sample can be analyzed at a time. From a practical perspective, if an antibody is available for a biomarker of interest, antibody based methods such as RPPMs, are more easily adopted as validation tools than mass spectrometry based methods .