Dr. Weinblatt has received research grant support from Biogen Idec, Crescendo Bioscience, Millennium Pharmaceuticals, and the Raymond P. Lavietes Foundation, and he has served as a consultant to Biogen Idec and Crescendo Bioscience.
Oklahoma Medical Research Foundation, Oklahoma City
Serum rheumatoid factor (RF) and other heterophilic antibodies potentially interfere with antibody-based immunoassays by nonspecifically binding detection reagents. The purpose of this study was to assess whether these factors confound multiplex-based immunoassays, which are used with increasing frequency to measure cytokine and chemokine analytes in patients with rheumatoid arthritis (RA).
We performed multiplex immunoassays using different platforms to measure analyte concentrations in RA patient samples. Samples were depleted of RF by column-based affinity absorption or were exposed to agents that block heterophilic binding activity.
In RA patients with high-titer RF, 69% of analytes demonstrated at least a 2-fold stronger multiplex signal in non–RF-depleted samples as compared to RF-depleted samples. This degree of erroneous signal amplification was less frequent in low-titer RF samples (17% of analytes; P < 0.0000001). Signal amplification by heterophilic antibodies was blocked effectively by HeteroBlock (≥150 μg/ml). In 35 RA patients, multiplex signals for 14 of 22 analytes were amplified erroneously in unblocked samples as compared to blocked samples (some >100-fold), but only in patients with high-titer RF (P < 0.002). Two other blocking agents, heterophilic blocking reagent and immunoglobulin-inhibiting reagent, also blocked heterophilic activity.
All multiplex protein detection platforms we tested exhibited significant confounding by RF or other heterophilic antibodies. These findings have broad-reaching implications in the acquisition and interpretation of data derived from multiplex immunoassay testing of RA patient serum and possibly also in other conditions in which RF or other heterophilic antibodies may be present. Several available blocking agents effectively suppressed this erroneous signal amplification in the multiplex platforms tested.
Rheumatoid arthritis (RA), the most common form of autoimmune inflammatory arthritis, is a heterogeneous disease, both in clinical presentation and in response to therapies. Early effective therapy substantially improves long-term functional status in RA patients (1). Accordingly, considerable recent research has focused on molecular profiling of patient cohorts in efforts to define the molecular pathology of RA and to identify biomarkers useful for disease diagnosis and for prediction of therapeutic responses (2–5).
Multiplex immunoassay methods have emerged as a powerful high-throughput means of quantifying pathogenic pathways and identifying candidate biomarkers in disease cohorts (6). These techniques typically use multiple combinations of analyte-specific antibodies derived from non-human species. Although these methods have provided substantial insight into the biology of RA, there are technical reasons to be concerned about the use of this antibody-based technology in disease states such as RA, in which there is a high prevalence of serum rheumatoid factor (RF). RF is an antibody with specificity for antigenic epitopes on Fc portions of IgG. RF has the potential to interfere with immunoassays by binding to the antibodies used to detect them, resulting in analyte-independent signals. Nonspecific effects of RF may be increased in multiplex immunoassays relative to monoplex assays because increased antibody concentrations and diversity present increased opportunity for nonspecific RF binding.
Because of these concerns, we evaluated the potential of RF-driven interference in several multiplex immunoassays (both solid-phase and bead-based platforms) in patients with RA. We report herein our findings of a substantial and idiosyncratic impact of RF on this methodology as well as a means to suppress such interference.
PATIENTS AND METHODS
Serum samples and RF testing.
Serum samples were obtained from patients with RA and from healthy volunteer donor controls. All RA patients were either part of the Brigham Rheumatoid Arthritis Sequential Study (BRASS) cohort (7) or met the American College of Rheumatology 1987 criteria for RA (8). Samples were tested for the presence of RF by either enzyme-linked immunosorbent assay (ELISA; TheraTest Laboratories) or nephelometry, as specified in the Results section. Samples were then allocated to one of several experiments in which different multiplex immunoassay platforms were used to assay for the presence of prespecified cytokine and chemokine analytes.
The study was conducted in accordance with protocols approved by the Institutional Review Boards of Partners Healthcare System and Oklahoma Medical Research Foundation.
Column depletion of RF.
Column affinity absorption was used to deplete completely RF from 9 RA patient serum samples. Baseline RF values were measured by nephelometry (range 13–680 IU) (9). Each sample was then apportioned into 2 fractions. One fraction was depleted of RF by column affinity absorption against human IgG-conjugated Sepharose. Serum samples were diluted 1:1 with Tris buffered saline and incubated for >4 hours at 4°C with an equal volume of IgG Sepharose 6 Fast Flow (GE Healthcare). The second fraction was processed in parallel, except that it was mock-depleted of RF by column absorption against unconjugated Sepharose. With these techniques, all RF-depleted fractions were depleted of detectable RF, as confirmed by nephelometry. In total, 18 fractions were processed (9 RF-depleted and 9 RF–mock depleted).
Multiplex immunoassay platforms.
The following commercial multiplex immunoassay platforms were used according to the manufacturers' protocols: RayBiotech (RayBiotech), SearchLight (Thermo Scientific Pierce), and Luminex (Invitrogen). Luminex assays were performed in house. RayBiotech and SearchLight analyses were performed by the commercial vendors (RayBiotech and Thermo Scientific Pierce, respectively). The specific analytes measured are described below.
RF blocking reagents.
The following heterophile-blocking commercial reagents were used in multiplex assays at the concentrations described below: HeteroBlock (Omega Biologicals), heterophilic blocking reagent (HBR; Scantibodies), and immunoglobulin-inhibiting reagent (IIR), which consisted, in part, of antiidiotype antibodies (Bioreclamation).
For the column affinity absorption experiment, we calculated the ratio of analyte concentrations measured in RF–mock depleted and RF-depleted fractions. The resultant M:D ratios were not normally distributed and thus were ranked before applying an analysis of variance (ANOVA) in which data for the 3 RF tertiles were subjected to a Kruskal-Wallis test (10, 11). For the Luminex platform experiment performed on sera from 35 RA patients, the mean fluorescence intensity (MFI) results were likewise not normally distributed and thus were ranked prior to a two-way ANOVA in which the samples were paired on the blocking variable. For all statistical calculations, a Bonferroni-corrected alpha level of 0.05 was considered statistically significant.
Association of strongly positive analyte signals with RF positivity in a multiplex antibody–based platform using sera from patients with RA.
An initial assessment of RF interference on multiplex immunoassay platforms was done on a highly multiplexed solid-phase immunoassay that measures 507 RA disease-relevant proteins (RayBiotech). Serum samples from 6 RF-positive RA patients, 6 RF-negative RA patients, and 6 healthy volunteer donor controls were analyzed. Remarkably, in 5 of 6 RF-positive samples, >70% of all measured analytes registered a “strongly positive” signal, defined here as an MFI that was at least 10 times that of the corresponding negative control. Only RA patients with high-titer serum RF values (>100 IU by ELISA) exhibited this unlikely multiplex profile (Figure 1). These results raised concern that, in RF-positive RA patients, signals were inappropriately driven by RF.
Erroneous amplification of multiplex immunoassay signals by serum RF.
We next tested the effect of RF depletion on a bead-based multiplex platform (SearchLight) in serum samples from 9 RA patients. Baseline RF values were measured by nephelometry, and serum samples were stratified into RF-low, RF-mid, and RF-high tertiles. Each serum sample was then apportioned into RF-depleted or RF–mock depleted fractions. Each RF-depleted and RF–mock depleted fraction was then tested in duplicate for the presence of 16 analytes: acute-phase serum amyloid A (A-SAA), E-selectin, intercellular adhesion molecule 1, interferon-γ (IFNγ), interleukin-1 receptor antagonist (IL-1Ra), IL-2R, IL-6, IL-7, matrix metalloproteinase 1 (MMP-1), MMP-3, MMP-9, RANTES, tissue inhibitor of metalloproteinases 1 (TIMP-1), TIMP-2, tumor necrosis factor receptor type I (TNFRI), and TNFRII. Three of the 16 analytes (IFNγ, IL-6, and IL-7) were present at levels below the detection threshold in all fractions. Results for 1 analyte (A-SAA) did not replicate.
For the remaining 12 analytes, the concentrations measured by multiplex immunoassay were frequently much greater in RF–mock depleted fractions than in their corresponding RF-depleted fraction (Figure 2) (Data on the depletion of serum RF and reduction of the analyte signal in multiplex-based immunoassays for the 6 remaining analytes, TIMP-1, RANTES, TNFRI, TNFRII, IL-1Ra, and IL-2R, are available online at http://www.brighamandwomens.org/research/labs/todd/.) Interestingly, RF depletion affected the measured concentration of most, but not all, analytes (e.g., MMP-9, E-selectin). Analytes with highest residual serum concentrations after antibody depletion appeared to be least affected, suggesting that interference effects are dependent on the relative levels of heterophilic activity and analyte.
These data were then analyzed for the effects of RF on multiplex signal amplification in RF–mock depleted fractions. For each of the 12 measurable analytes per patient, we calculated the ratio of analyte concentrations measured in RF–mock depleted and RF-depleted fractions (M:D ratio). Hence, an M:D ratio of 2 represents a 2-fold increased signal in the RF–mock depleted fraction. For samples in the RF-high tertile, M:D ratios were >2 in 69% of measurable analytes, and M:D ratios were >3 in 25% of measurable analytes. In contrast, M:D ratios were >2 in only 17% of RF-mid and RF-low tertiles, and were >3 in only 6% and 0% of RF-mid and RF-low tertiles, respectively (data available online at http://www.brighamandwomens.org/research/labs/todd/). The mean ± SEM M:D ratio was 3.41 ± 0.67 for the 36 measurable analytes in the 3 RF-high samples. This was significantly greater than the mean M:D ratios in the RF-low (1.50 ± 0.08) and the RF-mid (1.46 ± 0.13) samples, which were not statistically different from one another, although they were significantly different from the RF-high samples (P < 0.0000001 for each comparison).
Collectively, these multiplex immunoassay data demonstrate that in patients with RA, many measured analyte concentrations are erroneously amplified. This effect is most pronounced in patients with the highest RF titers, and it is abrogated by RF depletion.
Heterophilic binding and multiplex signal amplification in patients with RA.
We next assessed the confounding activity of RF in the most commonly used multiplex immunoassay platform (Luminex), assaying for a set of chemokine analytes commonly studied in RA (2). In this series, in addition to using the Luminex platform in a traditional manner, assays were run such that only signals from heterophilic interference could be detected. This was done by using primary and secondary detection antibodies that were not matched for analyte binding. In this context, a positive signal could only be derived from heterophile (RF) activity in each subject's serum. Results were compared to the signals obtained from standard multiplex immunoassays (paired detection antibodies). For these studies, we used a single heterophilic blocking agent that has been shown to block RF interference in RA serum samples (HeteroBlock) (2).
In our initial studies, we sought to optimize use of HeteroBlock to suppress heterophile activity in RA patient serum. Samples from 3 RA patients were incubated with 0, 1.5, 15, 150, or 1,500 μg/ml of the blocking agent, and the effects of the blocking reagent were assessed in a Luminex multiplex assay that included 5 chemokines: macrophage inflammatory protein 1α (MIP-1α), MIP-1β, monocyte chemotactic protein 1 (MCP-1), eotaxin, and RANTES. RANTES was not quantifiable in these assays. In the absence of blocking agent, 2 of the 3 patient samples demonstrated strongly positive signals for MIP-1α, MIP-1β, MCP-1, and eotaxin. In these patients, a substantial fraction of analyte reactivity was suppressed by the addition of HeteroBlock, suggesting a confounding signal from heterophilic binding (Figure 3).
To assess this further, we performed parallel multiplex assays with mismatched primary and secondary detection antibodies, such that only nonspecific signal from RF and heterophilic antibody activity could be detected. Signals in mismatched and standard assays were similar for the 2 patient serum samples with strongly positive analyte activity, confirming that significant nonspecific signal is observed in serum from these patients. Furthermore, nonspecific signals in sera from these 2 patients were suppressed to undetectable only with an adequate amount of blocking agent (≥150 μg/ml), a concentration at which true analyte signal remained readily detectable in MIP-1α, MCP-1, and eotaxin (Figure 3).
Interestingly, despite having a high titer of RF, the third patient sample did not demonstrate nonspecific signal, and the analyte-specific signal was not altered by blocking agent. These results confirmed that heterophilic binding activity accounts for erroneous signal amplification in some, but not all, RA patients.
Association of multiplex signal amplification with serum RF in patients with RA.
Having determined the optimal concentration of HeteroBlock, we proceeded to assess the reproducibility of RF-associated aberrant signal amplification, testing for heterophilic signal amplification in serum samples from a replication cohort of 35 patients with RA. Serum RF values were measured by ELISA, and samples were assigned to 1 of 3 groups: RF-negative (IgM, IgA, and IgG RFs all <25 IU), RF-low (any IgM, IgA, or IgG RF >25 IU and all <100 IU), and RF-high (any IgM, IgA, or IgG RF >100 IU).
Each sample was either incubated with 150 μg/ml of blocking agent (blocked) or vehicle control (unblocked). Multiplex bead-based (Luminex) assays were used to measure the concentrations of 23 cytokines that are also commonly analyzed in RA studies, 4 of which (IL-17, IL-8, IFNα, and TNFα) are shown in Figure 4. (Data on the association of multiplex signal amplification with serum RF in RA patients for 18 analytes, MIP-1α, MIP-1β, MCP-1, eotaxin, granulocyte–macrophage colony-stimulating factor (GM-CSF), IFNγ, IL-13, IFNγ-inducible 10-kd protein (IP-10), IL-1Ra, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-10, IL-12, and IL-15 are available online at http://www.brighamandwomens.org/research/labs/todd/.) Again, RANTES values could not be quantified. As a control, a serum sample without measurable heterophilic activity was “spiked” with recombinant analyte and then assayed.
In the 7 RF-negative and 11 RF-low patients, there were very few analytes for which the measured concentrations differed between blocked and unblocked samples. In contrast, many of the unblocked samples from the 17 RF-high patients were markedly influenced by erroneous signal amplification when measured for most analytes, some >100-fold. Using ranked 2-way ANOVA and a cutoff P value of < 0.002 as a Bonferroni correction for multiple comparisons, we identified statistically significant differences between unblocked and blocked RF-high samples when assayed for GM-CSF, IL-1Ra, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-10, IL-13, IL-17, MIP-1α, MIP-1β, and TNFα. Data for IL-17, IL-8, IFNα, and TNFα are shown in Figure 4; see online at http://www.brighamandwomens.org/research/labs/todd/ for the remaining data. There were trends toward significance for IFNα (P = 0.0055) and IL-15 (P = 0.0234). There were no significant differences for eotaxin, IFNγ, IL-8, IL-12, IP-10, and MCP-1. The “spiked” sample consistently showed no differences in measured analyte concentrations between blocked and unblocked samples across all analytes. Therefore, at 150 μg/ml, the blocking agent itself did not interfere with the ability of the multiplex immunoassay to detect genuine analyte concentrations. These data indicate that heterophilic binding activity causes significant signal amplification across many (but not all) analytes, and this effect was almost exclusively limited to patients with high-titer RF values.
Effective blockade of heterophilic multiplex signal amplification by other agents.
We next assessed the blocking capabilities of two other commercially available blocking agents: heterophilic blocking reagent and immunoglobulin-inhibiting reagent. To test HBR, RA patient serum was first mixed with a separate analyte-spiked control (with no heterophilic activity), and aliquots of this were then incubated with 0, 40, 400, or 1,600 μg/ml of HBR. Each of the admixtures was then tested for 23 analytes by bead-based (Luminex) multiplex assays (Figure 5A). Unblocked, unspiked RA patient serum was used as a control. In addition, serum from a healthy volunteer donor control was processed in parallel with the RA patient serum for all assays.
For the RA patient only, many of the analytes demonstrated signal amplification in unblocked serum samples, whether spiked or not. Supporting this, HBR effectively blocked the erroneously amplified signal down to the correct spiked cytokine concentrations observed in the volunteer donor control serum, which showed no inappropriate signal amplification. Figure 5A shows examples of MIP-1α and MIP-1β, which are representative of the amplification patterns also observed for IL-1β, IL-1Ra, IL-2, IL-4, IL-6, IL-7, IL-12, IL-15, and TNFα (data not shown). Again, some analytes, eotaxin, IFNγ, IL-8 (Figure 5A), IL-13, and IP-10, were not erroneously amplified by heterophilic binding. Importantly, in the volunteer donor control serum, there was little if any unblocked signal activity, and there was no change in the spiked analyte signal despite increasing concentrations of HBR.
IIR was tested in a separate experiment in which serum samples from 3 patients with RA and 1 spiked volunteer donor control sample were each admixed with 0, 200, 400, or 600 μg/ml of IIR. These samples were tested for 23 analytes by multiplex bead-based (Luminex) assays (Figure 5B). All 3 unblocked RA samples demonstrated varying degrees of heterophilic signal amplification, which was effectively blocked by IIR (such as MIP-1α and MIP-1β in Figure 5B). Comparable patterns of signal amplification were also observed for GM-CSF, IFNα, IL-1β, IL-1Ra, IL-2, IL-4, IL-6, IL-7, IL-10, IL-12, IL-15, IL-17, and TNFα. Again, some analytes demonstrated no signal amplification: IFNγ, IL-8 (Figure 5B), and IP-10. Multiplex signals were unaffected in spiked serum samples from the healthy volunteer donor control (Figure 5B). Collectively, these 2 experiments confirm that erroneous heterophilic signal amplification of multiplex-based immunoassays can be blocked effectively by a number of commercially available agents without apparent interference from the blocking agent itself.
The data from the present study show that RF heterophilic antibodies adversely affected the ability of multiplex-based immunoassay platforms to measure serum cytokines and chemokines in RA patients. The signal augmentation by RF was unpredictable both in its magnitude and in its correlation with RF titer. The effect was also unpredictable for any given analyte, since some analytes were affected disproportionately more than others. Multiple agents designed to block heterophilic binding effectively dampened the erroneous signal amplification across all platforms tested. These findings have broad-reaching implications for the interpretation of multiplex-based immunoassay results obtained from patients with RA.
It has previously been recognized that RF and other heterophilic antibodies can cause erroneous amplification of signal (sometimes called heterophilic interference) in traditional immunoassays such as monoplex ELISA (12). In these and other sandwich-based immunoassays, heterophilic antibodies in the test sample bridge the capture and detection antibody reagents (13). This interaction bypasses the antigen specificity of the assay and gives rise to erroneously amplified results. A similar mechanism likely accounts for the RF-associated signal augmentation that we observed in multiplex immunoassays, although some analytes were much more adversely affected than others. One likely explanation for this is that multiplex reagent antibodies are differentially bound by RF as a function of their species of origin and by their Ig subclass (14), an effect that could be minimized by manufacturers' elimination of analyte-specific antibodies that show RF binding susceptibility. Until then, investigators will need to account for RF interference through the use of depletion or blocking protocols.
We used column-based human Ig techniques to deplete RF and were able to eliminate RF heterophilic antibody activity (Figure 2). A similar depletion technique using protein L–Sepharose beads has been used effectively by other investigators (2, 15, 16). Unfortunately, these RF depletion techniques are labor-intensive and costly, which limits their broad applicability.
Several commercially available agents claim to block heterophilic interference, which provides a more cost-effective and labor-favorable approach than RF depletion. We showed here that HeteroBlock, HBR, and IIR each effectively quenched heterophilic antibody activity in multiplex-based immunoassay testing of RA patient serum, without measurably altering the analyte signal in spiked control samples. Since blocking agents neither dampened nor amplified true analyte measurement, our findings suggest a method for effectively dealing with confounding RF when using multiplex immunoassay methods in RA patients (Figure 6).
Our results are congruent with those of previous studies by other investigators documenting that heterophilic activity confounds multiplex immunoassay measurements in RA patients (2, 15, 17). It is noteworthy that in our cross-bead study, HeteroBlock quenched heterophilic cross-bead binding at concentrations ≥150 μg/ml (Figure 3). This amount of HeteroBlock required to suppress heterophilic activity in our assays was substantially higher than that needed in a previous study (2). Although this difference may be attributable to our selection of patients with high RF titers, our results nonetheless underscore the need for use of appropriate amounts of heterophile-suppressing reagents in this disease population.
Multiplex immunoassay methods have been used in many studies of RA patients in an attempt to improve our understanding of disease pathophysiology and treatment effects. Indeed, a widespread up-regulation of serum cytokines and chemokines has been reported prior to RA disease onset, an effect that also correlated with RF-positive status (2–5, 18). Most of these studies did not account for heterophilic binding activity by RF (3–5, 18). Multiplex-based immunoassays have also been used to measure changes in cytokine and chemokine levels in proof-of-concept clinical trials in RA patients (5, 19–21). Again, most of these studies did not account for serum RF (5, 19, 21), which has been shown in separate studies to decrease in response to pharmacotherapeutic agents (22–25). Overall, our findings raise concern regarding the veracity of interpreting results derived from methodologies confounded by RF.
Our findings potentially have generalized applicability in disease states other than RA. Heterophiles interfere with immunoassay-based measurement of anti–double-stranded DNA antibodies in patients being treated with anti-TNFα pharmacologic agents for a variety of disorders, including RA, inflammatory bowel disease, and seronegative arthritis (17). Highly multiplexed platforms have been used to study cytokine and chemokine profiles in systemic lupus erythematosus (26, 27), Sjögren's syndrome (28, 29), and hepatitis C virus infection (30, 31). Notably, RF is often readily detectable in these conditions as well as in other chronic inflammatory disease states (e.g., subacute bacterial endocarditis) (32) and in a subset of otherwise healthy subjects. It is unknown, and largely unaddressed, whether RF has any contributing effect on multiplex immunoassay signals in these conditions. Furthermore, our data re-open the question of whether investigators must account for other non-RF heterophilic antibodies as potential confounders of multiplex-based immunoassays, especially since heterophilic antibodies are present in upward of 40% of otherwise healthy subjects (33).
In summary, the data presented herein argue strongly that investigators must be aware of, and account for, serum RF heterophilic activity when using multiplex immunoassays to measure cytokine and chemokine concentrations in patients with RA. In our experiments, serum RF was associated with erroneous signal amplification of many (but not all) of the analytes assayed. In some instances, the effect was profound, leading to a >100-fold amplified fluorescence reading. Further, even the RF-positive subset of patients was affected in an unpredictable manner. Our results therefore preclude post hoc correction for RF status and mandate protocols designed to deplete or block RF heterophilic activity prior to multiplex immunoassay–based quantitative methods in patients with RA. These considerations are crucial, given the ever-expanding search for biomarkers of RA disease activity and the proliferation of multiplex-based immunoassay techniques used to achieve this end.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Todd had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Todd, Knowlton, Frank, Schur, Izmailova, Roubenoff, Shadick, Weinblatt, Centola, Lee.
Analysis and interpretation of data. Todd, Knowlton, Schur, Izmailova, Roubenoff, Shadick, Weinblatt, Centola, Lee.
ROLE OF THE STUDY SPONSORS
Crescendo Bioscience and Millennium Pharmaceuticals provided financial sponsorship for commercial assay costs. Employees of these companies participated as coauthors of the manuscript by contributing to the study design, the acquisition, analysis, and interpretation of the data, and the manuscript preparation. The authors independently collected the data, interpreted the results, and had the final decision to submit the manuscript for publication. Crescendo Bioscience and Millennium Pharmaceuticals reviewed the manuscript prior to submission, but publication of this article was not contingent upon approval by these companies.
We acknowledge Alexander Parker and Michael Pickard for their input on the experimental design.