- Antibody arrays: removal of high abundance proteins is not required, but antibody availability limits protein detection.
- Mass spectrometry: “hypothesis-free” protein detection is affected by high abundance proteins and peptide ionization.
- Both high throughput technologies are capable of semi-quantitative and quantitative data.
- Antibody arrays and MS complement each other.
The protein profile reflects current cellular activity, with relevance in disease diagnosis, disease prognosis, and tracking treatment response. Studies comparing proteomic profiles of healthy and diseased patients have led to a paradigm shift in patient care for numerous types of diseases and conditions. CA125 and CA 27.29, for example, are used to monitor treatment in ovarian and breast cancer patients, respectively. [1, 2] The survival rates of HER2-positive breast and gastric cancer patients have increased with the help of HER2-directed immunotherapy. [3, 4] Liver dysfunction can be assessed by measuring the level of alanine aminotransferase (ALT).  Successful resection for hyperparathyroidism is evaluated by measuring the level of parathyroid hormone. 
Extensive protein profiling and biomarker research efforts have resulted in thousands of potential protein biomarkers being identified. [7, 8] However, only a subset of these biomarkers has been validated using orthogonal platforms, and even fewer still (< 25) are approved by the Food and Drug Administration (FDA) for clinical use. [8-11] The disconnect between biomarker discovery and biomarker validation has resulted in a lack of sufficient evidence regarding analytical validity and clinical utility. [11, 12]
In this blog, high throughput technology platforms for protein profiling and biomarker research, antibody arrays and mass spectrometry (MS), are discussed and compared. Data from published studies employing antibody arrays and MS are also reviewed, highlighting how using both platforms could streamline the biomarker discovery-validation pipeline to increase the number of clinically-approved biomarkers.
Antibody arrays immobilize “capture” antibodies to a solid substrate, such as a nitrocellulose membrane, glass slide, or beads (Figure 1). [13, 14] After a blocking step, samples are incubated with the arrays. Nonspecific proteins are then washed off, and the arrays are incubated with detection antibodies. Chemiluminescent or fluorescent signals on membrane- and glass-based arrays are visualized using a chemiluminescent detector or laser scanner, respectively. Bead-based arrays (i.e., Luminex) employ plate-based flow cytometers or fluorescent imagers. Signal is proportional to the amount of protein bound by the capture antibodies. By spotting or labeling different antibodies in an addressable format, protein identification is simple.
The largest antibody arrays on the market, available from RayBiotech, analyze as many as 2,000 human proteins and 800 murine proteins at one time. Various samples types have been successfully used with antibody arrays, including cell supernatant, serum/plasma, synovial fluid, urine, breast milk, and tissue lysate. [15-20] Sample processing is minimal since no denaturation, reduction, alkylation, or digestion are required. Moreover, the arrays are robust to numerous reagents and buffers, including sodium dodecyl sulfate (SDS; < 0.1% w/v), thus buffer exchange is rarely required.
Both relative (semi-quantitative) and absolute (quantitative) protein expression can be ascertained with antibody arrays. Semi-quantitative data are obtained with a labeled sample or sandwich antibody approach. For the former, sample proteins are first biotinylated before the sample is added to the array. Proteins bound by the capture antibody are then detected using a horse radish peroxidase (HRP)- or fluorophore-conjugated streptavidin molecule. For the latter, a detection antibody and an HRP- or fluorophore-conjugated secondary antibody are employed to detect sample proteins bound by the capture antibody. Quantitation is achieved by comparing the target signal with a standard curve generated from the purified target protein.
Arrays employing a nitrocellulose membrane substrate only require a chemiluminescent detector for analysis, a common laboratory instrument used to analyze western blots. Although only one sample can be applied to an array, arrays can be processed in parallel and analyzed simultaneously by placing the membranes side-by-side within the detection window of the chemiluminescent instrument. Compatible laser scanners to analyze glass-based arrays are not as common as chemiluminescent detectors, but they are still routinely used in numerous labs and core facilities (e.g., DNA microarray facilities). Multiple samples can be analyzed per slide by printing identical sub-arrays on the glass and then adhering a multi-well gasket (i.e., one well per sub-array) onto the surface. One sample is applied per well, and then the multi-well gasket is removed prior to signal detection. Both the chemiluminescent detector and laser scanner require minimal training. Bead-based arrays employ flow cytometers or fluorescent imagers that are specifically designed for the arrays. The instruments are more expensive and require more training than chemiluminescent detectors or glass-based laser scanners (Table 1).
Perhaps the strongest argument of using antibody arrays lies in its use of antibodies, which allows the detection of low abundance proteins without the removal of high abundance proteins. In other words, the detection of a target protein is independent of non-target proteins. Moreover, antibody arrays can analyze total protein content as well as post translational modifications (PTMs), like glycosylation and phosphorylation. 
Paradoxically, antibodies are also the main disadvantage of antibody arrays. First, a protein can only be detected if an antibody to that protein is available and compatible on the array platform. Second, sandwich-based antibody arrays use “paired” capture and detection antibodies that bind to different epitopes on the same target protein. This ultimately limits the throughput of sandwich-based antibody arrays because paired antibodies that are also compatible on the platform are often difficult to identify. Finally, antibody cross-reactivity (i.e., non-specific binding) must be methodically tested between every antibody pair before array panels can be assembled. As such, these antibodies should not be placed on the same array panel with non-target proteins. Not surprisingly, sandwich-based arrays that use paired antibodies to the same target have higher specificity than label-based arrays due to the possibility of antibody crossreactivity.
High throughput antibody arrays are particularly well-suited for protein profiling and biomarker discovery studies. However, smaller antibody arrays are often desired. For example, a researcher may only be interested in a specific protein class type, signaling pathway, or biological function. Hundreds of small pre-made arrays targeting 8 – 60 proteins are offered by RayBiotech that focus on a particular protein type (e.g., matrix metalloprotease), health condition (e.g., bone metabolism), or signaling pathway (e.g., AKT pathway).  In addition, small custom antibody arrays can be easily fabricated for biomarker validation as long as the antibodies are available.
MS comes in numerous different flavors, from the ionization technique to the mass spectrometer (e.g., time-of-flight, Orbitrap). In this blog, we will focus on “bottom-up” proteomic analysis via liquid chromatography (LC)-MS because of its widespread use in protein profiling and biomarker discovery studies. [12, 23, 24] Briefly, proteins are denatured, reduced, alkylated, and digested. Peptides are then separated online with LC and then interfaced to the mass spectrometer via electrospray ionization (ESI), during which peptides become ionized. Notably, only charged peptides are measured by the mass spectrometer. In MS/MS, peptides are fragmented into shorter sequences and weighed again for more accurate peptide identification. Using a compiled peptide-to-protein database, each peptide is assigned to a protein based on its mass and fragmentation.
Depending on the upstream sample preparation and matrix, more than 5,000 proteins can be identified with MS in a single analysis. [25, 26] Sample fractionation of proteins or peptides prior to LC-MS analysis simplifies sample complexity and increases protein coverage. Extended online LC or the use of multiple types of proteases during digestion can also improve proteome coverage. 
Highly abundant proteins can mask the detection of low abundance proteins by ion suppression, which is a phenomenon where peptide ionization – and therefore the detection of those peptides – is impeded by various factors, including the peptide’s amino acid composition, buffer, and presence of other peptide and non-peptide species.  In this case, highly abundant proteins may result in the ion suppression of less abundant proteins. Highly abundant proteins can also be problematic during MS/MS analysis when the top-most abundant peptides are selected for additional fragmentation and peptide identification.
Depletion, or removal, of highly abundant proteins using antibody-based affinity columns is primarily applied with serum or plasma samples because the proteome spans over 10 orders of magnitude with only 22 proteins constituting 99% of the total protein content. [9, 29] In one study, 5 protein standards were spiked into human plasma at 4 – 40 µg/ml prior to immunodepletion of the twelve most abundant proteins.  Only 1 of the spiked proteins was detected in the bound fraction whereas all 5 proteins were identified in the depleted fraction. Nonspecific binding of the depletion antibodies to less abundant and possibly biologically-relevant proteins may occur.  This has led some MS scientists to perform extensive fractionation rather than sample depletion to analyze serum and plasma samples.  The downside to this is that the time to process the samples and analyze the data increases proportionally to the number of collected fractions.
Protein or peptide enrichment is favored when there are few proteins-of-interest or a protein subset-of-interest. Antibody-based enrichment is primarily used during biomarker validation or clinical assessment when it’s important that target detection is not inhibited by the presence of non-target proteins (e.g., via ion suppression). Other types of enrichment, like titanium dioxide, enrich phosphorylated species.
Sample contamination of certain reagents common in the laboratory, such as SDS and polyethylene glycol (PEG), will result in high background as these molecules also become ionized during ESI and suppress the signal of analytes-of-interest. Other reagents that may plug the LC columns or affect sample pH will also affect the data negatively, and must be removed prior to analysis. Buffer exchange is thus often employed. Minimizing keratin contamination from the skin while obtaining and processing the samples is also important.
MS can perform both semi-quantitative and quantitative analyses. Semi-quantitative analyses include mass spectral counting, comparing relative peak intensities, and spiking a mixture of quantified proteins (e.g., Sigma Aldrich’s Universal Proteomics Standard 1) into the sample prior to digestion. Absolute quantitation employs an isotopically-labeled peptide representing the peptide-of-interest.  During sample analysis, the labeled peptides are then spiked into the sample at a known concentration. Quantification of hundreds of unique peptides within a sample is possible.
As many as 11 samples can be analyzed at the same time using Thermo Fisher’s tandem mass tag (TMT) system. In this semi-quantitative approach, peptides from different samples are labeled with a unique isobaric tag. This allows identical peptides from different samples to be eluted from the LC column and analyzed by the mass spectrometer simultaneously.
Both antibody arrays and MS can analyze thousands of proteins in one analysis. Antibody arrays are limited by the availability of good antibodies, yet the panels can be designed to target disease-specific proteins with documented roles in pathologic states. MS is “hypothesis-free” since it does not rely on antibodies and does not make any assumptions regarding which proteins will be of interest.  As such, the discovery of truly novel biomarkers is possible with MS; for example, unknown protein isoforms or PTMs for which no purified antibodies exist.
The detection of specific proteins-of-interest is more straightforward with antibody arrays than MS. Protein detection with antibody arrays is not impeded by ion suppression or the presence of non-target proteins. In other words, arrays can reliably measure disease-specific proteins that are typically in low abundance. However, protein detection with MS may require protein enrichment, depletion, or fractionation. Importantly, the protein(s) may still not be detectable after depletion or fractionation. Enrichment, especially for multiple targets, may not be cost- or time-feasible arrays as this requires testing and optimization (e.g., antibodies). In contrast, a targeted approach using commercially-available antibody arrays can be performed within a day. Antibody array pricing depends on the format (i.e., membrane, glass, bead) and number of proteins targeted, but are generally several hundred dollars per array. In one study, cytokines measured by antibody arrays were not detectable by MS in an experiment studying cytokine profiles of exosomes released by T cells infected with a virus. 
The time to process proteins for antibody arrays and MS is similar. Only protein is required for antibody arrays, yet processing the samples on the arrays, from array blocking to adding the detection reagent, is ~6 hours. On the other hand, bottom-up LC-MS analysis requires ≥ 5 hours of sample preparation, including protein denaturation, reduction, and digestion. Alkylation, which requires an additional ≥ 30 minutes, may also be performed. The entire processing time can be significantly decreased to minutes with sonication or pressure; however, these methods are not commonly employed. [35, 36] Other procedures, such as buffer exchange, fractionation, depletion, enrichment, and labeling, may also be required.
The instrument and the analysis time differ between antibody arrays and MS. Array data collection per glass slide with a laser scanner is completed within 5 – 20 minutes, depending on the size of the array and the scanning resolution. Data collection with a chemiluminescent detector, an inexpensive and common lab instrument, is even shorter (i.e., < 2 minutes). An advantage of chemiluminescent detectors and laser scanners is that they require minimal training to run. The length of MS data collection for protein profiling and biomarker discovery is, on average, 2 hours per specimen. However, data collection can range from minutes to days depending on the MS set-up and project objectives. Moreover, as many as 11 samples can be analyzed simultaneously. Mass spectrometers are expensive and are operated by researchers with specialized training. For laboratories that lack the training or accessibility to the required instruments, full testing services for antibody arrays and MS analyses are available from core facilities and companies.
Data analysis is quite different between antibody arrays and MS. For antibody arrays, regions-of-interest (i.e., spots) are first identified with array software, and then assigned to the protein target using a GenePix Assigned List (GAL) file. Due to the slight variations in spot deposition during printing, this step is often manual to ensure that the array grid is properly aligned spot-by-spot. Finally, the chemiluminescent or fluorescent signals for each spot are extracted. For MS, a chromatogram of peaks representing peptides and fragmented peptide ions (i.e., parent and daughter ions). The assignment of peptides to proteins using MS software and databases is routinely automated, with minimal manual intervention. However, protein identification is dependent on whether the protein is properly annotated in the protein database (e.g., Mascot) used for MS analyses.
Quantitative data is possible with both antibody arrays and MS. Quantitative antibody arrays are commercially available, with the appropriate reagents to create a standard curve. Spot signal is then extrapolated to the standard curve to determine protein concentration. Quantitative MS analysis is often performed during biomarker validation or assessment using isotopically-labeled peptides. Initial MS optimization to detect the peptides-of-interest can be labor- and time-intensive since the parameters will differ for each peptide. Once these parameters have been determined, they can be applied for routine quantitative analysis.
Platform sensitivity is dependent on various factors for both antibody arrays and MS. Antibody sensitivity largely dictates the sensitivity of the antibody array, ranging from sub-picogram to nanograms per milliliter. To a lesser degree, array sensitivity depends on the detection method (i.e., chemiluminescent, fluorescent, type of fluorophore). Depending on sample complexity and the mass spectrometer, the sensitivity of MS is higher than antibody arrays, measuring as low as femtograms per milliliter.
Data reproducibility is similar between antibody arrays and MS. The inter- and intra-coefficient of variation (CV) for arrays printed at the same time are typically < 20%, although ~10% is often observed. [37, 38] Each protein-of-interest can be analyzed multiple times within an array for more accurate results. Only hundreds of picograms of antibody is deposited per spot, yet it is inevitable that a new antibody lot will be required for large-scale antibody array production. Therefore, a new antibody lot must be tested and compared to previous lots regarding its performance (i.e., sensitivity, specificity). Importantly, the use of calibration curves can help normalize any lot-to-lot variations. Depending on the protein abundance, the CV for repeated MS analyses is also < 20%. 
Table 1. Advantages and disadvantages of antibody arrays and mass spectrometry
The advantages and disadvantages of antibody arrays and MS are compared in Table 1, with neither platform outshining the other. Antibody arrays are ideal for laboratories with:
- A limited budget since the detectors or detection are common or cheap
- Staff with basic laboratory skills since sample processing on both the semi-quantitative and quantitative arrays is simple
- An interest in a specific biological process or molecular function
- Proteins-of-interest with available antibodies, especially for proteins-of-interest in low abundance.
Antibody arrays and MS used in conjunction with each other can enrich protein profiling and biomarker discovery research. For example, proteome coverage will increase when both platforms are used in parallel during initial screening experiments. Potential biomarkers first identified by MS could be validated using an antibody array. In one study, serological protein profiles of aneurysmal subarachnoid hemorrhage patients with and without vasospasm were first analyzed with MS and then antibody arrays.  Using this two-pronged approach, 9 protein biomarkers of subarachnoid hemorrhage, vasospasm, and delayed ischemic neurological deficits were first identified with MS and validated with antibody arrays. In other study, MS and antibody arrays helped discover novel cancer-associated fibroblasts (CAFs) in the lysate and supernatant of CAFs or normal fibroblasts.  One CAF biomarker (CCL2) was observed by both platforms, with 5 or 4 biomarkers identified either by MS or antibody arrays, respectively. Notably, the 4 of the 5 unique biomarkers detected by MS were not present on the antibody array. In an experiment studying the protein profile of exosomes released from a mesenchymal stem cell line, a total of 857 proteins were detected using an antibody array and MS.  101 of the 507 proteins measured by the antibody array were detected in the samples. Only 10 of the 766 proteins detected by MS was also detected by the antibody array.
Numerous other MS studies have utilized enzyme-linked immunosorbent assays (ELISAs) for biomarker validation. Thus, it is worth noting that sandwich-based quantitative antibody arrays are multiplex ELISAs, which require less sample than multiple single-plex ELISA assays.
In conclusion, antibody arrays and MS both have unique advantages and disadvantages. Rather than one technology being better than the other, the data obtained from both platforms can complement each other in protein profiling and biomarker discovery studies. Moreover, potential biomarkers identified with MS can be validated using antibody arrays.
- Gupta, D. and C.G. Lis, Role of CA125 in predicting ovarian cancer survival – a review of the epidemiological literature. J Ovarian Res, 2009. 2: p. 13.
- Cen, P., et al., Increased cancer antigen 27.29 (CA27.29) level in patients with mycosis fungoides. J Am Acad Dermatol, 2008. 58(3): p. 382-6.
- Shitara, K., et al., Prognosis of patients with advanced gastric cancer by HER2 status and trastuzumab treatment. Gastric Cancer, 2013. 16(2): p. 261-7.
- Perez, E.A., et al., Trastuzumab plus adjuvant chemotherapy for human epidermal growth factor receptor 2-positive breast cancer: planned joint analysis of overall survival from NSABP B-31 and NCCTG N9831. J Clin Oncol, 2014. 32(33): p. 3744-52.
- Kim, W.R., et al., Serum activity of alanine aminotransferase (ALT) as an indicator of health and disease. Hepatology, 2008. 47(4): p. 1363-70.
- Sokoll, L.J., Measurement of parathyroid hormone and application of parathyroid hormone in intraoperative monitoring. Clin Lab Med, 2004. 24(1): p. 199-216.
- Paulovich, A.G., et al., The interface between biomarker discovery and clinical validation: The tar pit of the protein biomarker pipeline. Proteomics Clin Appl, 2008. 2(10-11): p. 1386-1402.
- Parker, C.E. and C.H. Borchers, Mass spectrometry based biomarker discovery, verification, and validation–quality assurance and control of protein biomarker assays. Mol Oncol, 2014. 8(4): p. 840-58.
- Geyer, P.E., et al., Revisiting biomarker discovery by plasma proteomics. Mol Syst Biol, 2017. 13(9): p. 942.
- Kirwan, A., et al., Glycosylation-Based Serum Biomarkers for Cancer Diagnostics and Prognostics. Biomed Res Int, 2015. 2015: p. 490531.
- Hayes, D.F., et al., Breaking a vicious cycle. Sci Transl Med, 2013. 5(196): p. 196cm6.
- Crutchfield, C.A., et al., Advances in mass spectrometry-based clinical biomarker discovery. Clin Proteomics, 2016. 13: p. 1.
- Wilson, J.J., et al., Antibody arrays in biomarker discovery. Adv Clin Chem, 2015. 69: p. 255-324.
- Lynch, H.E., et al., Development and implementation of a proficiency testing program for Luminex bead-based cytokine assays. J Immunol Methods, 2014. 409: p. 62-71.
- Luck, C., et al., Differential Regulation of Cardiac Function and Intracardiac Cytokines by Rapamycin in Healthy and Diabetic Rats.Oxid Med Cell Longev, 2017. 2017: p. 5724046.
- Dietze-Schroeder, D., et al., Autocrine action of adiponectin on human fat cells prevents the release of insulin resistance-inducing factors. Diabetes, 2005. 54(7): p. 2003-11.
- Vincent, F.B., et al., Effect of storage duration on cytokine stability in human serum and plasma. Cytokine, 2019. 113: p. 453-457.
- Goetz, J.E., et al., A clinically realistic large animal model of intra-articular fracture that progresses to post-traumatic osteoarthritis. Osteoarthritis Cartilage, 2015. 23(10): p. 1797-805.
- Charlton, J.R., et al., Evolution of the urinary proteome during human renal development and maturation: variations with gestational and postnatal age. Pediatr Res, 2012. 72(2): p. 179-85.
- Kverka, M., et al., Cytokine profiling in human colostrum and milk by protein array. Clin Chem, 2007. 53(5): p. 955-62.
- Cui, Y., et al., Identification of the receptor tyrosine kinases (RTKs)-oriented functional targets of miR-206 by an antibody-based protein array. FEBS Lett, 2015. 589(16): p. 2131-5.
- Chen, Z., et al., Current applications of antibody microarrays. Clin Proteomics, 2018. 15: p. 7.
- Gillet, L.C., A. Leitner, and R. Aebersold, Mass Spectrometry Applied to Bottom-Up Proteomics: Entering the High-Throughput Era for Hypothesis Testing. Annu Rev Anal Chem (Palo Alto Calif), 2016. 9(1): p. 449-72.
- Zhang, Y., et al., Protein analysis by shotgun/bottom-up proteomics. Chem Rev, 2013. 113(4): p. 2343-94.
- Zubarev, R.A., The challenge of the proteome dynamic range and its implications for in-depth proteomics. Proteomics, 2013. 13(5): p. 723-6.
- Geyer, P.E., et al., Plasma Proteome Profiling to Assess Human Health and Disease. Cell Syst, 2016. 2(3): p. 185-95.
- Guo, X., et al., Confetti: a multiprotease map of the HeLa proteome for comprehensive proteomics. Mol Cell Proteomics, 2014. 13(6): p. 1573-84.
- Furey, A., et al., Ion suppression; a critical review on causes, evaluation, prevention and applications. Talanta, 2013. 115: p. 104-22.
- Anderson, N.L. and N.G. Anderson, The human plasma proteome: history, character, and diagnostic prospects. Mol Cell Proteomics, 2002. 1(11): p. 845-67.
- Liu, T., et al., Evaluation of multiprotein immunoaffinity subtraction for plasma proteomics and candidate biomarker discovery using mass spectrometry. Mol Cell Proteomics, 2006. 5(11): p. 2167-74.
- Garbis, S.D., et al., A novel multidimensional protein identification technology approach combining protein size exclusion prefractionation, peptide zwitterion-ion hydrophilic interaction chromatography, and nano-ultraperformance RP chromatography/nESI-MS2 for the in-depth analysis of the serum proteome and phosphoproteome: application to clinical sera derived from humans with benign prostate hyperplasia. Anal Chem, 2011. 83(3): p. 708-18.
- Urban, P.L., Quantitative mass spectrometry: an overview. Philos Trans A Math Phys Eng Sci, 2016. 374(2079).
- Aebersold, R. and M. Mann, Mass-spectrometric exploration of proteome structure and function. Nature, 2016. 537(7620): p. 347-55.
- Jaworski, E., et al., Human T-lymphotropic virus type 1-infected cells secrete exosomes that contain Tax protein. J Biol Chem, 2014. 289(32): p. 22284-305.
- Lopez-Ferrer, D., et al., Rapid sample processing for LC-MS-based quantitative proteomics using high intensity focused ultrasound. J Proteome Res, 2008. 7(9): p. 3860-7.
- Olszowy, P.P., A. Burns, and P.S. Ciborowski, Pressure-assisted sample preparation for proteomic analysis. Anal Biochem, 2013. 438(1): p. 67-72.
- Huang, R., et al., A biotin label-based antibody array for high-content profiling of protein expression. Cancer Genomics Proteomics, 2010. 7(3): p. 129-41.
- Jiang, W., et al., Identification of five serum protein markers for detection of ovarian cancer by antibody arrays. PLoS One, 2013. 8(10): p. e76795.
- Tabb, D.L., et al., Repeatability and reproducibility in proteomic identifications by liquid chromatography-tandem mass spectrometry. J Proteome Res, 2010. 9(2): p. 761-76.
- Nyquist, P.A., H. Wang, and A.F. Suffredini, Protein biomarkers in patients with subarachnoid hemorrhage, vasospasm, and delayed ischemic neurological deficits. Acta Neurochir Suppl, 2013. 115: p. 23-5.
- Torres, S., et al., Proteome profiling of cancer-associated fibroblasts identifies novel proinflammatory signatures and prognostic markers for colorectal cancer. Clin Cancer Res, 2013. 19(21): p. 6006-19.
- Lai, R.C., et al., Proteolytic Potential of the MSC Exosome Proteome: Implications for an Exosome-Mediated Delivery of Therapeutic Proteasome. Int J Proteomics, 2012. 2012: p. 971907.
About the Sponsor:
Developing Innovative, Reliable Research Tools for the Post-Genomic Age. RayBiotech’s mission is to provide cutting-edge array technologies for the discovery of disease-related protein biomarkers and key factors, the identification of new drug targets, and the advancement of personalized medicine.