Abstract
In contrast to traditional singleplex assays that provide values for only a single analyte in a single biological sample, multiplex assays are a time and resource efficient high-throughput approach that provides the opportunity to determine numerous analytes within a single and small sample volume. In this editorial on a paper by Dorn et al. in this issue of Psychosomatic Medicine, we provide a brief description of the advantages and challenges related to multiplex assays. While the use of multiplexing as a tool has been relatively limited in biobehavioral research, more recent studies are taking advantage of this technology to obtain deeper insight into regulatory patterns in health and disease states. Multiplex approaches range from several targets to global target profiling that importantly enable unbiased biomarker and pathway discovery.
In this issue of Psychosomatic Medicine, Dorn et al. (1) report novel and interesting cytokine patterns in a group of healthy adolescent girls. The group utilized a multiplex platform to simultaneously assess and model 13 different biomarkers representing adaptive Th1 and Th2 as well as innate immune pathway cytokines. In addition to the importance of the data revealing the innate variation of cytokines in healthy adolescent girls and their associations with mood and anxiety, the paper also includes a valuable examination of two different data reduction approaches to simultaneously analyzing multiple cytokine data [variable-centered (Principal Factor Analysis) and person-centered data (Latent Profile Analysis)], while discussing the merits and shortcomings of each. These statistical approaches may reduce biases related to multiple testing and statistical type I error (2). While much attention in biomarker discovery has focused on discrete disease states, it can be argued that equivalent biomarkers of health are equally important in order to develop improved disease prevention strategies. The study by Dorn et al. (1) is significant in this regard.
We have come to understand that disease, particularly chronic disease, involves dysregulation of multiple pathways that often vary among afflicted individuals. As “no man is an island”, so too no active biological compound exists unto its own in its respective milieu, and while the reporting of a single or very few biomarkers in a scientific study does possess value, that value would be markedly higher when information is added about the other numerous important integrated regulatory factors. For cytokines for example (such as Interleukein-6), as noted by Kingsmore (3), the other important regulatory factors include ‘the dynamic aggregate of multiple tissue-dependent agonist and antagonist cytokines, associated modifier proteins, receptors and receptor antagonists’. Multiplexing then is an efficient bioassay tool for measuring multiple analytes in a single biological sample. This is in contrast to traditional singleplex assays that provide measurements for only a single analyte in a single biological sample.
For many decades enzyme-linked immunosorbent assays (ELISAs) have been a reliable workhorse of numerous laboratories and generally considered the gold standard for protein assessment. As ELISA platforms moved from singleplex to multiplex assays, singleplex ELISAs were used to test the accuracy and reproducibility of the multiplex platform ELISA (4). With advances in the reliability and validity of multiplexing platforms, the biomedical literature has in parallel shown a significant increase in multiplexing biomarker manuscripts across diverse types of biomedical research domains (Figure 1). The literature reporting on multiple global ‘omics’ strategies, including genomics, epigenomics, transcriptomics, proteomics and metabolomics, has too grown significantly. This advancement to simultaneously assess numerous biomarkers within systems has obviously deepened our capacity to understand and model regulatory pathways and other phenomena across numerous complex biological systems in health and disease states.
Figure 1.
Number of publications appearing in PubMed by 5-year increments from 1975 to the present using the search terms “multiplexing, assay, biomarker”.
It would be reasonable to say that while ELISAs made a relatively easy transition to multiplexing, other multiplexing approaches such as bead array assays were initially more problematic in terms of their validity and reliability. More recent performance evaluations of this platform, however, have been more favorable (5, 6). Dorn et al. (1) used a Luminex Lincoplex multiplex immunoassay panel comprised of a high sensitivity 13-plex antibody bead array that permitted simultaneous measurement of multiple cytokines in a single, small volume blood sample. Other biobehavioral studies have used other multiplexing platforms, including the Meso Scale Discovery (Rockville, MD) ELISA platform (7–9).
As the number of analytes to assess has increased in multiplexing, vendors have grouped them into “panels” for specific areas of interest. Analytes are typically grouped within themed panels and strategically paired based on compatibility of the chemistry as well as sample dilution needs. For example, cytokine, cancer genome, and kinome panels are often employed in biobehavioral research (1, 10). Flow cytometry is regularly used in biobehavioral research to simultaneously quantify numerous cell surface and intra-cellular cytokines, not so much circulating levels. Depending on the vendor and platform, typical multiplex panels can provide simultaneous assessment of up to many dozens of biomarkers in a single sample of plasma, serum, urine or saliva. Of course, the potential to expand the quantity of multiplexed analytes is as easy as combining values across a number of different panels. For example, cardiovascular disease researchers might combine proinflammatory, vascular injury, and chemokine panels in order to more thoroughly assess risk profile of a certain cardiovascular disease population or to assess the cardiovascular health related benefits of a behavioral intervention. Arnold et al. (11), for example, combined several different cardiovascular and metabolic disease multiplex panels to yield a total of 190 assessed protein biomarkers that provided 80% accuracy to classify individuals into their correct depressive diagnoses.
What are pros and cons to employing multiplex assays in biobehavioral research? As far as concerns with multiplexing, many of these are the same that hold for singleplex assays, including the primary need for consistency and reliability of antibody quality across lots. This merits significant attention as antibody quality can be poor (12, 13) and has been cited as a contributor to the current reproducibility crisis in biomedical science (14). Also, while there are generally good correlations among analyte levels across different multiplexing platforms, there are invariably significant differences in absolute levels (5). Thus, and consistent with singleplex assays, it is important not to change multiplexing platforms midstream within a given study. Finally, there are noted limitations too to the proliferation of both diagnostic and prognostic protein biomarkers across many fields (15–18). Although these potential limitations are not a focus of this editorial, whether we are considering biomarkers for risk prediction or screening and diagnosis, limitations include issues such as validity, reliability, sensitivity, and specificity. In addition, the undermining effects of inherent high biological variability on the utility of commonly used biomarkers have been noted (17).
As exemplified in the Dorn et al. study (1), there are both opportunities and challenges for data reduction and pattern recognition analysis that correspond to the increasing the number of biomarker outcomes. Their approach to the cytokine data included both variable-centered and person-centered methodologies; the former to ascertain which cytokines to factor together while the latter to ascertain which study subjects to group together. The major cons related to global approaches include the lack of off-the-shelf systems biology data analysis packages and the challenge of integrating multiple data types. Such approaches are also not hypothesis-driven but are often useful for hypothesis generation and have shown great potential in disease modeling, diagnostics and providing a greater understanding of systems, pathways, and their functionally dynamic interactions.
While the use of multiplexing for proteins and neurohormones in biobehavioral research is increasing, their use has lagged far behind as compared to other fields. In addition, what constitutes ‘multiplexing’ varies across fields of research and by orders of magnitude across platforms and substrates. In proteomics and metabolomics platforms, for example, which are less commonly used in biobehavioral research, the multiplexing scale increases approximately 10 to 1,000 fold (19).
Future directions for biobehavioral research will include the increased use of multiplex platforms. While it was historically challenging to find platforms that could adequately detect biomarkers at inherently low biological levels (i.e., below the lower limits of assay detection) and typically easier to assess levels that were elevated in disease states, newer higher sensitivity platforms and kits do a better job at detecting levels in healthy individuals. Future studies will benefit the literature by reporting sensitivity ranges for the different analytes being assessed. Studies like that of Dorn et al. (1) will continue to be undertaken on a larger and larger scale and the continued parallel development of statistical approaches will allow one to effectively address concomitant multivariate problems. Given the multi-component, multi-pathway nature of human disease, global datasets and powerful multivariate analysis procedures are in better alignment to understand the complexities of both health and disease. The integration of -omics approaches such as global transcriptional profiling and transcription factor motif analysis in biobehavioral research has generated profound insights such as increased support for an immune system involvement in depression and fatigue (20). Genome-wide association studies, proteomics, metabolomics, RNA-sequencing, and both exome and whole genome sequencing have been employed successfully in other fields and efforts to integrate large data sets are rapidly evolving, e.g. the Cancer Genome Atlas (21). In cancer research, for example, studies often integrate both genetic and environmentally influenced pathway data. Similarly, the increased use of multiple -omics approaches and integration of those datasets may further enrich biobehavioral research. For example, approaches may include metabolomics of peripheral blood and data integration with observed gene, gene mutation, and protein expression patterns. Changes in gene expression patterns and metabolite concentration can be linked to behavioral changes and interpreted in a biological context. Future studies that include measurement of biomarkers of health may further advance the development of disease prevention strategies. The integration of these data types with clinical observations will improve our understanding of the molecular basis of disease as well as disease management and prevention.
Acknowledgments
The work is supported in part by a grant from the National Institutes of Health (Paul Mills). The authors indicate that they have no conflicts of interest to disclose.
References
- 1.Dorn LD, Gayles JG, Engeland CG, Houts R, Cizza G, Denson LA. Healthy Adolescent Girls and Heterogeneous Cytokine Patterns: Using Variable- and Person-Centered Statistical Strategies to Capture Distinct Characteristics of Immune Function Psychosom Med. 2016. in press. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Lindquist MA, Mejia A. Zen and the art of multiple comparisons. Psychosom Med. 2015;77:114–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kingsmore SF. Multiplexed protein measurement: technologies and applications of protein and antibody arrays. Nat Rev Drug Discov. 2006;5:310–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Bastarache JA, Koyama T, Wickersham NE, Ware LB. Validation of a multiplex electrochemiluminescent immunoassay platform in human and mouse samples. J Immunol Methods. 2014;408:13–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Breen EC, Reynolds SM, Cox C, Jacobson LP, Magpantay L, Mulder CB, Dibben O, Margolick JB, Bream JH, Sambrano E, Martinez-Maza O, Sinclair E, Borrow P, Landay AL, Rinaldo CR, Norris PJ. Multisite comparison of high-sensitivity multiplex cytokine assays. Clin Vaccine Immunol. 2011;18:1229–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Elshal MF, McCoy JP. Multiplex bead array assays: performance evaluation and comparison of sensitivity to ELISA. Methods. 2006;38:317–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Breitbart W, Rosenfeld B, Tobias K, Pessin H, Ku GY, Yuan J, Wolchok J. Depression, cytokines, and pancreatic cancer. Psychooncology. 2014;23:339–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Mills PJ, Wilson K, Iqbal N, Iqbal F, Alvarez M, Pung MA, Wachmann K, Rutledge T, Maglione J, Zisook S, Dimsdale JE, Lunde O, Greenberg BH, Maisel A, Raisinghani A, Natarajan L, Jain S, Hufford DJ, Redwine L. Depressive symptoms and spiritual wellbeing in asymptomatic heart failure patients. J Behav Med. 2015;38:407–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.von Kanel R, Mausbach BT, Mills PJ, Dimsdale JE, Patterson TL, Ancoli-Israel S, Ziegler MG, Allison M, Chattillion EA, Grant I. Longitudinal relationship of low leisure satisfaction but not depressive symptoms with systemic low-grade inflammation in dementia caregivers. J Gerontol B Psychol Sci Soc Sci. 2014;69:397–407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lyon D, Walter J, Munro CL, Schubert CM, McCain NL. Challenges in interpreting cytokine biomarkers in biobehavioral research: a breast cancer exemplar. Biological research for nursing. 2011;13:25–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Arnold SE, Xie SX, Leung YY, Wang LS, Kling MA, Han X, Kim EJ, Wolk DA, Bennett DA, Chen-Plotkin A, Grossman M, Hu W, Lee VM, Mackin RS, Trojanowski JQ, Wilson RS, Shaw LM. Plasma biomarkers of depressive symptoms in older adults. Transl Psychiatry. 2012;2:e65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Baker M Reproducibility crisis: Blame it on the antibodies. Nature. 2015;521:274–6. [DOI] [PubMed] [Google Scholar]
- 13.Egelhofer TA, Minoda A, Klugman S, Lee K, Kolasinska-Zwierz P, Alekseyenko AA, Cheung MS, Day DS, Gadel S, Gorchakov AA, Gu T, Kharchenko PV, Kuan S, Latorre I, Linder-Basso D, Luu Y, Ngo Q, Perry M, Rechtsteiner A, Riddle NC, Schwartz YB, Shanower GA, Vielle A, Ahringer J, Elgin SC, Kuroda MI, Pirrotta V, Ren B, Strome S, Park PJ, Karpen GH, Hawkins RD, Lieb JD. An assessment of histone-modification antibody quality. Nat Struct Mol Biol. 2011;18:91–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Collins FS, Tabak LA. Policy: NIH plans to enhance reproducibility. Nature. 2014;505:612–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Mayeux R Biomarkers: potential uses and limitations. NeuroRx. 2004;1:182–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Hedrick VE, Dietrich AM, Estabrooks PA, Savla J, Serrano E, Davy BM. Dietary biomarkers: advances, limitations and future directions. Nutr J. 2012;11:109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Alexander KS, Kazmierczak SC, Snyder CK, Oberdorf JA, Farrell DH. Prognostic utility of biochemical markers of cardiovascular risk: impact of biological variability. Clin Chem Lab Med. 2013;51:1875–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Lainscak M, Anker SD. Prognostic factors in chronic heart failure. A review of serum biomarkers, metabolic changes, symptoms, and scoring systems. Herz. 2009;34:141–7. [DOI] [PubMed] [Google Scholar]
- 19.Fuhrer T, Zamboni N. High-throughput discovery metabolomics. Current opinion in biotechnology. 2015;31:73–8. [DOI] [PubMed] [Google Scholar]
- 20.Felger JC, Cole SW, Pace TW, Hu F, Woolwine BJ, Doho GH, Raison CL, Miller AH. Molecular signatures of peripheral blood mononuclear cells during chronic interferon-alpha treatment: relationship with depression and fatigue. Psychol Med. 2012;42:1591–603. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Cancer Genome Atlas Research N. Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature. 2008;455:1061–8. [DOI] [PMC free article] [PubMed] [Google Scholar]