Multiplex Immunoassay Analysis of Cytokines in Idiopathic Inflammatory Myopathy

Inclusion body myositis (IBM), polymyositis (PM), and dermatomyositis (DM) are diseases of unknown etiology that are collectively known as idiopathic inflammatory myopathies (IIMs).1 The 3 diseases are thought to differ mechanistically and are known to differ in terms of treatment and prognosis.1–14 Muscle injury in PM is thought to be related to direct CD8-positive T-cell mediated attack on muscle fibers that aberrantly express major histocompatibility complex type I. Immunohistochemical identification of CD8-positive lymphocytes and major histocompatibility complex type I–positive muscle fibers aids in making a tissue diagnosis. In contrast, DM is thought to proceed through antibody- and complement-mediated endothelial injury, with muscle damage because of microvascular injury. CD4-positive T cells and B cells are found on immunohistochemical analysis of DM specimens, but CD8-positive T cells and major histocompatibility complex type I expression on muscle fibers are not expected. Inclusion body myositis has many histologic features in common with PM, but in addition, so-called rimmed vacuoles are identified on frozen sections and characteristic inclusions are identified by electron microscopy.15–20 

Identification of these classic histologic patterns of inflammation and mechanistically important molecules by light microscopy and immunohistochemistry are the cornerstones of tissue diagnosis of IIM but are neither completely sensitive nor specific for the diseases.21 Inadequate sampling during surgical biopsy or histologic sectioning, incomplete concordance between clinical diagnosis and histopathology, and interpretative differences between pathologists complicate the process of diagnosis. A quantitative assay for IIM performed on bulk tissue could, in principle, overcome 2 of these problems. First, because a quantitative assay could be done on an entire tissue specimen, more than just a thin slice of tissue could be sampled, and pathologic changes in all parts of the biopsy specimen would be subject to analysis. Second, because a quantitative assay yields a numerical result, if appropriate studies are done to validate reference ranges, aberrant values could be defined relatively or absolutely, which should substantially reduce interobserver variability.

Even as several inflammatory mediators (ie, cytokines) have been mechanistically implicated in IIM7,22 and identified in diseased muscle tissue,3,22–28 there is no clinical assay currently in use for IIM based on measuring relative or absolute cytokine concentrations. With the goal of establishing such an assay, we analyzed the cytokine composition of several muscle biopsy specimens from patients with and without IIM using multiplex bead-based immunoassays to identify which, if any, cytokines may be good targets for clinical assays of IIM.

All human tissue used in this study was obtained from patients who underwent muscle biopsy between January 2001 and March 2007. Only excess muscle tissue not used in making the original diagnosis was used, and all experiments were approved by the local institutional review board. Tissue was generally snap-frozen in pentane cooled with liquid nitrogen at the time of receipt by the pathology department and then stored at −80°C per a standard muscle biopsy protocol. Samples were chosen from patients diagnosed clinically with PM, DM, and IBM but not other collagen vascular diseases; some samples from patients with other disorders that can be in the differential diagnosis of patients undergoing muscle biopsy were also included (Table 1). Too few examples of viral or focal myositis were available to include these for group analysis. Staff neuropathologists rendered diagnoses in all cases, using light microscopy of frozen section and fixed tissue, immunohistochemistry, and electron microscopy (if indicated). Three of the biopsies, from patients with eosinophilic fasciitis extending into muscle, Becker muscular dystrophy, and very mild but nonspecific chronic inflammation, were not included in the statistical analysis as either IIM or control specimens because the diagnoses did not fit into either category. Patient demographics are summarized in Table 2.

Portions from all 64 frozen muscle biopsy specimens were obtained through manual dissection of the frozen original biopsy sample in a −20°C cryostat. When possible, cubes of tissue measuring 3 mm per side were harvested. The size of the muscle portion taken was 30 ± 10 mg (range, 10.7–59.3 mg), with the variation in weight because of differences in the sizes of the frozen archived specimens.

Samples were analyzed with the investigator blinded to diagnoses. Three hundred microliters of ice-cold phosphate-buffered saline (pH 7.4), supplemented with 1:100 diluted Protease Inhibitor Cocktail (P8340, Sigma, St Louis, Mo), were added to the frozen muscle samples in a 1.5-mL plastic centrifuge tube. The tissue and solution were thawed on ice briefly and then homogenized using a plastic pestle fitted with a battery-operated motor (Pellet Pestle, Kontes, Vineland, NJ) until no visible pink tissue remained in suspension. A small amount of tough, tan connective tissue remained after homogenization of most samples. Samples were thereafter centrifuged at 15 000g for 30 minutes at 4°C in a microcentrifuge, and the pink supernatants were aliquoted into 96-well plates. Supernatant samples were then further clarified by filtering through 96-well plates fitted with 0.2-μm nylon filters (MSBVN-1250, Millipore, Billerica, Mass) and stored in multiple aliquots at −80°C to avoid multiple freeze-thaw cycles. Four percent to 20% gradient SDS-PAGE (Biorad, Hercules, Calif) and Lowry-method protein quantification (Dc Protein Assay, Biorad) were performed on all samples. Precision-Plus All Blue molecular weight standards (Biorad) were used for molecular weight estimation.

Multiplex microbead-based human cytokine immunoassay kits for a Luminex instrument (Liquichip, Qiagen, Valencia, Calif) were purchased from the following vendors: interleukin (IL) 1α, IL-1β, IL-1ra, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12 (p70), IL-15, IL-17, interferon-γ, granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), tumor necrosis factor α (TNF-α), monocyte chemotactant protein 1 (MCP-1), soluble CD40 ligand (sCD40L), macrophage inflammatory protein 1α (MIP-1α), and vascular endothelial growth factor (VEGF; Linco Research, St Charles, Mo); intercellular adhesion molecule 1 (ICAM-1; Biorad); and transforming growth factor β1 (TGF-β1; Biosource, Camarillo, Calif). Immunoassays were run on clarified, undiluted muscle lysate supernatants according to manufacturer's specifications. Cytokine assays from Linco Research were combined (“multiplexed”), and the remaining analytes were analyzed singly. Assay results in picograms per milliliter were normalized by dividing by the lysate total protein concentration in milligrams per milliliter, yielding final concentrations in picograms of analyte per milligram of muscle protein.

A small subset of the tissue samples was analyzed prior to the whole set of samples to discover which cytokines could be detected. Analyte concentrations in these initial analyses were not assumed to be normally distributed and so were compared using Mann-Whitney U tests or the Kruskal-Wallis 1-way analysis of variance. Analyte concentrations obtained from analyzing all samples appeared to be log-normally distributed and so were log-transformed prior to comparison using Student t tests or 1-way analysis of variance. Posttests comparing intergroup differences were performed with Bonferroni correction for multiple comparisons. Receiver operating characteristic curves were constructed manually using a Microsoft Excel spreadsheet (Microsoft Corp, Redmond, Wash).

The tissue extraction protocol yielded 39 ± 12 mg protein per gram of muscle. We found that different homogenization methods yielded different amounts of protein but verified that our fairly mild homogenization conditions (no detergent or sonication) were indeed efficient in lysing cells in the sample by measuring myocyte lysis (data not shown).

All analytes for which kits were purchased were quantified in a subset of the muscle biopsies (6 each of PM, DM, and IBM and 10 nonmyositis controls). Of the analytes tested, only IL-1ra, IL-7, IL-15, MCP-1, sCD40L, VEGF, and ICAM-1 were detected in most specimens (Table 3). Analyte yield was not significantly increased by addition of 0.5% Triton X-100 to the homogenization buffer (data not shown). The concentrations of ICAM-1, IL-1ra, IL-7, and MCP-1 differed significantly between IIM and nonmyositis samples, and concentrations of ICAM-1, IL-1ra, and MCP-1 differed significantly among the PM, DM, IBM, and non-IIM groups (Table 3).

The entire collection of 64 samples was then reanalyzed for IL-1ra, IL-7, IL-15, MCP-1, MIP-1α, and ICAM-1, using a second portion of muscle for those samples that were part of the initial subset. The concentrations of ICAM-1, IL-1ra, and MCP-1 were significantly different between IIM and nonmyositis samples (Figure 1). IL-7, IL-15, and MIP-1α were not consistently detected in all samples and, when detected, were found at concentrations near or at the manufacturer's suggested assay limit of detection.

Not all cases with inflammation on biopsy had similar results on testing, but several patterns were observed. ICAM-1 levels varied significantly among the different diagnoses of IIM (Figure 2). ICAM-1 was highest in PM and DM samples, with all of the very highest results obtained in PM samples. Similar results were obtained for IL-1ra (Figure 3) and MCP-1 (Figure 4), for which DM and PM samples were significantly elevated compared with non-IIM samples. Inclusion body myositis samples had elevated mean levels of all 3 analytes compared with the non-IIM samples, but this difference was not statistically significant. ICAM-1 in IBM samples was in fact significantly different (lower) than in PM samples.

Receiver operating characteristic analysis demonstrated that tests for ICAM-1, IL-1ra, and MCP-1 had relatively high sensitivity and specificity for the diagnosis of IIM (Figure 5). Areas under the receiver operating characteristic curve for the 3 analytes were 0.90, 0.87, and 0.85, respectively. A cutoff of 1240 pg ICAM-1 per milligram muscle protein was 83% sensitive and 91% specific for the diagnosis of IIM.

ICAM-1, IL-1ra, and MCP-1 levels in 3 samples excluded from the statistical analysis of IIM versus non-IIM are shown in Table 4. In 1 case with eosinophilic inflammation, most likely because of eosinophilic fasciitis extending into muscle, we found high analyte values, and in another case with mild, nonclassifiable chronic inflammation we found low analyte levels. A final case with Becker muscular dystrophy had mildly elevated analyte levels.

Because the hallmarks of IIM can be difficult to identify or even absent in histologic sections of a tissue biopsy sample, tissue diagnoses cannot always be rendered definitively. In this study, the concentrations of several cytokines and adhesion molecules were assayed in muscle biopsy homogenates from patients with and without IIM to investigate the possibility that quantitative differences in these markers could serve as the basis of a diagnostic assay. Many of the analytes measured in this study have been implicated in the pathogenesis of IIM or identified immunohistochemically in biopsy sections, but to our knowledge, none have been quantified directly in bulk biopsy tissue for this purpose. The analytes in this study were included if they had been implicated in IIM in the literature and if a bead-based immunoassay kit was commercially available.

One problem with validating any new assay as a diagnostic marker for IIM is that there is currently no definitive methodology for making the diagnosis. Microscopic examination of tissue obtained at biopsy is currently the standard for confirming IIM, but it is not thought to be totally sensitive or specific for the diagnosis. Until a gold standard diagnostic test is developed, therefore, one is left to compare new tests to microscopic methods, as is done in this study.

Our high correlation between analyte levels and final diagnoses of IIM suggest that a high analyte level in a biopsy from a patient clinically suspected of IIM would support the clinical diagnosis in the absence of diagnostic histologic findings, but it is not conclusive. Indeed, because we did not knowingly include any “normal” muscle biopsies from patients with IIM, we cannot directly address whether our analysis would identify cases with false-negative histologic findings. In addition, although it is interesting to consider comparing analyte levels with the inflammatory infiltrate observed histologically, these are complementary techniques, that is, quantitation in tissue homogenate versus localization in tissue slices, and as such are difficult to compare directly.

The increase in ICAM-1 observed in IIM in this study is consistent with the findings of published immunohistochemical studies.28,29 One difference, however, is that the kit used in this study measured levels of soluble ICAM-1, the soluble form of the adhesion molecule, whereas previous studies measured the membrane-bound form of the molecule. The regulation of soluble ICAM-1 production from membrane-bound ICAM-1 is not entirely clear, but there is evidence that soluble ICAM-1 levels elaborated by endothelium simply reflect surface expression.30 Therefore, our bulk measurements probably support the earlier immunohistochemical findings, as well as validate the approach of making bulk tissue measurements.

Soluble ICAM-1 is found in several inflammatory states and in several types of body fluids.31 In this study, 2 non-IIM patients used as controls had polymyalgia rheumatica and rheumatoid arthritis (1 patient) and a 20-year history of an ill-defined systemic autoimmune disorder (a second patient) and also had modestly elevated levels of ICAM-1 (1373 and 1639 pg ICAM-1 per milligram muscle, 2 of the highest control values and within the range seen in IIM patients). Although similar systemic inflammatory states were not reported in the clinical history provided for other control patients with modestly elevated ICAM-1, this may hint at a cause of at least some of the false-positive results of the assay. Although extremely high ICAM-1 levels may indicate localized inflammation associated with PM or DM, more modest elevations may simply indicate systemic inflammation with elevated serum levels.

IL-1ra32 and MCP-125 have also been identified in IIM muscle biopsy specimens by immunohistochemistry or RNA expression analysis. In contrast to ICAM-1, which is found in capillaries by immunohistochemistry, IL-1ra and MCP-1 are associated with leukocytes. This demonstrates that the methodology used in this study can quantify multiple signals originating from different cell types simultaneously, which is not routinely possible in histopathologic diagnosis. This study did not identify any added value of IL-1ra and MCP-1 levels in predicting IIM compared with only using ICAM-1 levels, but perhaps a larger study could reveal if there was significance in the relative elevations of all 3 analytes. For example, we investigated whether there were differences in analyte profiles among the different IIMs. Although we observed trends, that is, extremely high ICAM-1 is found in PM and sometimes DM but not in IBM, the differences among these analytes was not sufficient to permit reliable distinction among IIM subclasses.

The results in Table 4 indicate that the cytokine and adhesion molecule increases found in this study are not entirely specific for IIM. Very high levels of ICAM-1, IL-1ra, and MCP-1 were identified in a patient with eosinophilic fasciitis in whom eosinophilic inflammation spread into muscle, suggesting that the markers may only be specific for increased muscle inflammation, without much selectivity for the type of inflammation present. In a second case, a mild increase in ICAM-1 (with IL-1ra and MCP-1 levels in the range seen in control samples) was observed in a young male patient with muscular dystrophy, but no inflammation was identified on microscopic examination of the biopsy. This may be because of chance alone, as only 1 muscular dystrophy patient was included in the sample, but it may also indicate that ICAM-1 is increased in muscular dystrophy without concomitant observable inflammatory infiltrates or that patchy inflammatory changes were present in the biopsy specimen but not adequately sampled histologically. Finally, a biopsy with very mild, nonspecific inflammation had very low levels of all 3 analytes, close to the mean of the control specimens, indicating that the microscopic finding may have been a “false-positive” and that IIM was probably not the true diagnosis. Long-term clinical follow-up information on this patient was not available.

There are several possible reasons that cytokines previously associated with IIM were not detected in this study. First, it is possible that some of histologic diagnoses of our cases were erroneous, as some non-IIM conditions can mimic the histologic appearance of IIM. Because all patients diagnosed with IIM had both the clinical and the pathologic picture of IIM, however, it seems unlikely that these patients had other disorders showing similar changes on biopsy. Second, our study used archived muscle tissue that had, in some cases, been frozen for up to 6 years and was not originally collected for the purpose of this experiment. Immediate analysis of fresh tissue collected for the express purpose of cytokine analysis might therefore have yielded results more consistent with previous work. Third, some prior studies of cytokine levels in IIM have measured messenger RNA levels for cytokines not protein concentrations, and the 2 types of analytes may not have perfect concordance. Finally, the local high concentrations but overall low concentration of molecules in tissue may allow immunohistochemical identification but preclude identification in bulk homogenates. Concentration gradients of molecules apparent in immunohistochemical studies may also not correspond to a bulk concentration difference when larger portions of muscle are analyzed. Despite these shortcomings, robust signals for at least 2 cytokines and an adhesion molecule were obtained.

Based on our findings, we believe that a quantitative assay for cytokines and soluble ICAM-1 performed on a bulk muscle specimen could be useful in cases in which diagnostic histopathologic findings are lacking, yet the clinical impression strongly favors IIM. For example, after failing to find any histopathologic alterations in several sections made from a muscle biopsy specimen from a patient in whom PM is strongly suspected, a further finding of low tissue ICAM-1 would be strong evidence that additional histopathologic analysis of that biopsy sample would be unlikely to yield a diagnosis of PM. The cost of this strategy is an important consideration. Clinical assays using the Luminex platform are used currently in clinical immunology laboratories for autoantibody testing, and these tests are similar in cost and complexity to other common immunoassays. The time associated with the assay is comparable to immunohistochemistry (1 day), and the unit cost of reagents is similar as well.

The strategy of diagnostic quantitation of proteins of interest in solid tissue lysates, already common in the setting of metabolic disorders, may therefore be expanded to the diagnosis of inflammatory disorders because of the ready availability of highly sensitive, commercially available, multiplexed immunoassay kits for the determination of important pathologic biomarkers.

Financial support was obtained from the Nancy and Buster Alvord Endowment and from the Departments of Pathology and Laboratory Medicine, University of Washington, Seattle.