SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES
  8. Supporting Information

Objective

Subsets of patients with rheumatoid arthritis (RA) develop extraarticular complications that include interstitial lung disease (ILD). Because standardized algorithms for identification of RA patients at risk of developing clinically significant ILD are lacking, the purpose of this study was to elucidate unique serologic markers of RA-associated ILD (RA-ILD).

Methods

Sera from RA patients with ILD and from RA patients without ILD were used to immunoprecipitate citrullinated and uncitrullinated proteins derived from K562 cell extracts. Mass spectrometry was performed to facilitate identification of citrullinated proteins differentially immunoprecipitated by RA-ILD patient sera. These candidate proteins were then used as substrate antigens in custom enzyme-linked immunosorbent assays (ELISAs) for high-throughput screening of sera obtained from cohorts of patients with RA, RA-ILD mixed connective tissue disease (MCTD), or idiopathic pulmonary fibrosis (IPF).

Results

Differential immunoprecipitation and subsequent mass spectrometric sequencing identified citrullinated Hsp90α and citrullinated Hsp90β as candidate autoantigens in patients with RA-ILD. ELISAs incorporating uncitrullinated and citrullinated isoforms of Hsp90 as substrate antigens demonstrated that sera from patients with RA-ILD preferentially recognized citrullinated versions of Hsp90 with moderate sensitivity (range 20–30%) and great specificity (>95%) relative to sera derived from patients with RA alone (without ILD), patients with MCTD, or patients with IPF.

Conclusion

These studies demonstrate the utility of a novel autoantigen discovery method based on differential immunoprecipitation of citrullinated protein extracts. Application of these techniques identified citrullinated versions of Hsp90α and Hsp90β as autoantibody targets distinguishing RA-ILD from RA without ILD, MCTD, and IPF, suggesting that anti–citrullinated Hsp90α/β autoantibodies may serve as effective biomarkers for the potentially devastating pulmonary manifestations of RA-ILD.

Rheumatoid arthritis (RA) is the most common systemic autoimmune disease in the United States, affecting ∼1% of the adult population (1, 2). While articular complications represent a dominant feature of this disease, extraarticular manifestations often involve additional tissues that include the skin, eye, vasculature, peripheral nervous system, heart, and lungs. Within the spectrum of pulmonary complications, parenchymal involvement in the form of interstitial lung disease (ILD) represents a serious problem that can significantly impact morbidity and mortality (3–8). Prevalence estimates vary widely depending on the clinical, functional, and radiographic criteria used to define RA-associated ILD (RA-ILD), but generally prevalence ranges from 10% to 40% among patients diagnosed as having RA (1, 9–11). Included in this spectrum is clinically advanced, fibrotic lung disease (reflecting underlying usual interstitial pneumonia), but the evolution/natural history of this RA-ILD subtype remains undefined. In fact, the wide-ranging prevalence estimates and uncertain clinical outcomes of RA-ILD have generated a clear need for improved biomarkers that will permit earlier identification of lung disease and provide insight into the pathogenesis of this severe extraarticular complication.

Epidemiologic studies have demonstrated associations between smoking, lung disease, protein citrullination, possession of the shared epitope, and development of RA (12). Thus, several investigative groups have previously assessed correlations between anti–cyclic citrullinated peptide (anti-CCP) antibody levels and RA-associated extraarticular manifestations such as cardiovascular disease (13–15) and ILD (16, 17). The proposed link between protein citrullination and RA-ILD reflects a more general paradigm in which injurious stimuli, such as smoking, induce damage/inflammation that triggers posttranslational modification (citrullination) of proteins, thereby generating cryptic/neo-epitopes that promote the breakdown of tolerance as well as subsequent autoimmune responses against citrullinated proteins in relevant target tissues (18). Consistent with this hypothesis, immunohistochemical studies have demonstrated the presence of citrullinated proteins in lung explant tissue derived from patients with RA-ILD (19). Although these same studies have shown that protein citrullination can also occur in alternative processes, such as idiopathic pulmonary fibrosis (IPF) (19), the immune response to such posttranslationally modified proteins is clearly influenced by genetic factors that include the shared epitope—explaining the relatively specific association between anti–citrullinated protein antibody (ACPA) responses and RA.

Despite the epidemiologic and experimental support for this conceptual framework linking citrullination of lung tissue with the development of RA-ILD, previous studies assessing correlations between ACPA titers and RA-ILD have yielded conflicting results. While retrospective analysis of a large Greek cohort demonstrated that high titers of anti–CCP-2 (measured using the second-generation anti–CCP-2 test) were significantly associated with serositis and pulmonary fibrosis (16), another study found no significant differences in the frequency of anti-CCP antibody formation between patients with and those without ILD or follicular bronchiolitis (17). Ultimately, the protected identity of proteins/peptides contained within these proprietary enzyme-linked immunosorbent assays (ELISAs) precludes full interpretation of these findings, though neither study focused on specific citrullinated proteins that might distinguish patients with and those without lung disease.

Given these considerations and the relative paucity of lung tissue available for more direct analysis of protein citrullination patterns, we devised a technique in which serologic “fingerprints” can be used to define underlying tissue-specific profiles of immunogenic, citrullinated proteins. Based on differential immunoprecipitation of proteins from cell extracts subjected to in vitro deimination, this approach enabled us to isolate citrullinated proteins preferentially recognized by sera from patients with RA-ILD. Mass spectrometric analysis of electrophoretically separated protein bands ultimately revealed citrullinated Hsp90α and citrullinated Hsp90β as candidate target proteins associated with autoantibody responses in RA-ILD. Development of custom ELISAs employing citrullinated and uncitrullinated versions of these proteins as substrate antigens subsequently facilitated high-throughput screening of sera obtained from well-defined cohorts of RA patients with different stages of lung disease. These ELISA-based studies demonstrated that, despite displaying relatively modest sensitivity, antibodies targeting citrullinated versions of the highly homologous Hsp90α and Hsp90β subunits were remarkably specific for RA-ILD—supporting the value of these autoantibodies as biomarkers and further validating our “reverse immunophenotyping” approach.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES
  8. Supporting Information

Inclusion criteria and patient samples.

In accordance with approved Institutional Review Board protocols, serum samples were obtained from previously established registries incorporating cohorts of patients meeting the American College of Rheumatology 1987 classification criteria for RA (20). These serum/data repositories included the Veterans Administration Rheumatoid Arthritis database (21), Brigham and Women's Hospital (secondary source of samples originally reported as part of the NIH study described in ref.22), and the University of Pittsburgh Connective Tissue Database. Serum samples used for the comparator cohorts of patients with mixed connective tissue disease (MCTD) (defined by the Alarcon-Segovia and Cardiel diagnostic criteria [23]) and patients with IPF were obtained from registries at the University of Miami and Brigham and Women's Hospital, respectively.

Additional diagnostic criteria for RA-ILD included the presence of pulmonary symptoms (dyspnea, cough), restrictive physiologic findings on pulmonary function testing (forced expiratory volume in 1 second [FEV1], forced vital capacity [FVC], total lung capacity, and diffusing capacity for carbon monoxide <80% predicted, and FEV1/FVC >80% predicted), and radiographic abnormalities on chest radiographs or computed tomography scans (consisting of reticulation, septal thickening, traction bronchiectasis, honeycombing, and/or ground-glass opacification) (22, 24). Classification of subclinical RA-ILD also required the presence of these radiographic abnormalities (with or without restrictive physiologic findings on pulmonary function tests) in the absence of dyspnea, cough, or other clinical features of pulmonary disease.

In patients with RA-ILD, a number of demographic and clinical variables were assessed, including sex, racial background, HLA–DRB1 shared epitope status, presence of nodules, smoking history, medication usage, and parameters of disease activity. Measurement of disease activity included use of the Disease Activity Score in 28 joints (DAS28) and the Health Assessment Questionnaire (HAQ) (25, 26), as well as determination of rheumatoid factor (RF) titer, anti–CCP-2 titer, and erythrocyte sedimentation rate (ESR).

Protein citrullination.

In vitro citrullination of cell lysates and other recombinant proteins (as described below) involved overnight incubation with rabbit skeletal muscle peptidylarginine deiminase (PAD; Sigma-Aldrich), using predefined ratios of protein to enzyme, in a buffer containing 20 mM Tris (pH 8.8), 0.3M NaCl, 1 mM EDTA, 10 mM dithiothreitol, and 5 mM CaCl2.

Immunoprecipitation.

Serum samples (60 μl, obtained from representative patients with RA and patients with RA-ILD) bound to protein A–Sepharose beads were incubated with citrullinated protein extracts (2 mg in a total volume of 1 ml) generated through in vitro, PAD-catalyzed deimination of lysates prepared from K562 cells (an immortalized erythroleukemia cell line). Parallel serum samples bound to protein A–Sepharose beads were coincubated with uncitrullinated K562 cell extracts (at an equivalent protein concentration) to control for recognition of uncitrullinated epitopes. Following this coincubation step, Sepharose bead complexes were washed with immunoprecipitation buffer, suspended in 2× Laemmli sample buffer, and electrophoresed at 200V (in 1 dimension, without isoelectric focusing) on a standard-size 10% polyacrylamide gel for comparative analysis of banding patterns.

Mass spectrometric sequence identification.

Candidate citrullinated proteins preferentially immunoprecipitated by sera from patients with RA-ILD were localized by immunoblotting (using an anti–modified citrulline antibody; EMD Millipore), excised from corresponding 1-dimensional sodium dodecyl sulfate–polyacrylamide gels, digested with trypsin in situ, eluted, and subjected to tandem matrix-assisted laser desorption ionization–time-of-flight mass spectrometry for sequence analysis of derivative peptides (conducted at the University of Pittsburgh Genomics and Proteomics Core Laboratory). Use of the Sequest search engine and Proteome Discoverer software generated probability scores for the identity of electrophoretically separated parent proteins. Application of modified search criteria using the same software permitted identification of candidate residues likely to be converted from arginine to citrulline (based on mass shift).

ELISA.

Serum levels of IgG anti-Hsp90α or anti-Hsp90β antibodies were measured using standard, solid-phase ELISAs according to the following protocol. Ninety-six–well microtiter plates (Nunc) were coated with either citrullinated or uncitrullinated recombinant human Hsp90α (1.0 μg/ml; Cell Sciences), citrullinated or uncitrullinated recombinant human Hsp90β (1.0 μg/ml; Hsp90β-eukaryotic [Hsp90β-E] from Biovision and Hspβ-prokaryotic [Hspβ-P] from Sigma), or no antigen in carbonate buffer (100 mM NaHCO3/Na2CO3 [pH 9.6]), and incubated overnight at 4°C. After the wells were blocked with phosphate buffered saline (PBS)/Tween containing 4% goat whey and 15% goat serum, diluted serum samples (1:100–1:2,000) were added for 2 hours. Sequential incubations with horseradish peroxidase–conjugated, anti-human IgG secondary antibodies (0.04 mg/ml; Santa Cruz Biotechnology) and tetramethylbenzidine (Sigma-Aldrich) as substrate facilitated spectrophotometric measurement at an optical density of 450 nm. As a specificity control, the same sera were separately assessed with a commercial anti–CCP-2 ELISA kit (Axis-Shield Diagnostics) according to an established protocol. For limited assays involving anti–citrullinated Hsp90 antibody detection in bronchoalveolar lavage (BAL) fluid, samples of BAL fluid were diluted 1:1 with PBS prior to use in the customized ELISA system outlined above.

Statistical analysis.

Using specified thresholds for seropositivity, serum samples were classified as either reactive or nonreactive to various ELISA substrate antigens; rates of seropositivity were then compared by Fisher's exact test, with statistical significance based on 2-tailed P values of less than 0.05. While the Fisher's exact test also facilitated analysis of other categorical data, unpaired 2-sample t-tests permitted comparison of clinical/laboratory data involving continuous variables.

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES
  8. Supporting Information

Identification of citrullinated Hsp90 as a target autoantigen in RA-ILD.

To detect candidate autoantigens in RA-ILD, we established a comparative immunoprecipitation protocol involving sera from RA patients with ILD and RA patients without ILD. As illustrated in Figure 1, this approach allowed us to focus on citrullinated target proteins preferentially recognized by sera from patients with RA-ILD.

thumbnail image

Figure 1. Experimental strategy for identification of novel citrullinated autoantigens. This schematic summarizes the steps of in vitro protein citrullination (cell versus tissue extract), immunoprecipitation with patient sera, sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE; 1-dimensional [1D] versus 2-dimensional [2D]), Western blotting with anti–citrullinated protein antibodies (Ab), band excision from matching gel, and protein identification through tandem matrix-assisted laser desorption ionization–time-of-flight (MALDI-TOF) mass spectrometry (MS). RA = rheumatoid arthritis; ILD = interstitial lung disease.

Download figure to PowerPoint

When probed with anti–modified citrulline antibody, an immunoblot of electrophoretically separated immunoprecipitates revealed a dominant band from our index case of RA-ILD (Figure 2, boxed area). The protein band corresponding to this region of the immunoblot was prominent enough to be distinguished on a Coomassie blue–stained, 1-dimensional gel (results not shown), thus facilitating extraction, trypsin digestion, and subsequent analysis by mass spectrometry.

thumbnail image

Figure 2. Differential recognition of in vitro citrullinated proteins by rheumatoid arthritis (RA) sera. Following immunoprecipitation of citrullinated (C) and uncitrullinated (U) K562 cell extracts by sera obtained from RA patients with interstitial lung disease (ILD) (lanes 1–10) and RA patients without ILD (lanes 11–20), proteins were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis, transferred to a nitrocellulose membrane, and probed with an anti–modified citrulline antibody that specifically binds to citrulline residues. Paired immunoprecipitates of uncitrullinated and citrullinated K562 extracts are shown in lanes 1–20, with patient number and diagnosis listed above each pair. Lane 21, standard (ST). Asterisks indicate the coexistence of rheumatoid factor and additional autoantibodies targeting Jo-1 (subject P2) or SSA (subjects P4 and P8). Controls include total citrullinated (lane 22) and uncitrullinated (lane 23) protein extract. The boxed area in lane 6 represents an example of differential antigen recognition by serum obtained from a patient with RA-ILD (subject P3). Of note, the Coomassie blue–stained protein corresponding to this band (∼80 kd) was not seen with immunoprecipitation of uncitrullinated protein extract.

Download figure to PowerPoint

Of the candidate proteins possessing mass spectrometry profiles compatible with the observed spectral analysis, Hsp90β and Hsp90α attained the highest likelihood scores. Despite the high degree of sequence homology between these ∼85-kd subunits of Hsp90 (85% amino acid sequence identity), several of the derivative peptides identified by mass spectrometry profiling were nonoverlapping. Corresponding to these differences, the distribution of putative citrullinated residues (determined by mass spectrometric analysis of subtle mass shifts relative to parent peptides) was more extensive in immunoprecipitated Hsp90β relative to Hsp90α. Furthermore, even in shared sites of presumed citrullination, surrounding amino acid substitutions distinguished sequences of these Hsp90 isoforms (see Supplementary Figure 1, available on the Arthritis & Rheumatism web site at http://onlinelibrary.wiley.com/doi/10.1002/art.37881/abstract).

Detection of anti–citrullinated Hsp90β and anti–citrullinated Hsp90α antibodies in RA patients with ILD.

Based on these mass spectrometry profiling results, we established an ELISA-based screening system using uncitrullinated and citrullinated Hsp90 subunits as substrate antigens. As shown in Table 1 and Figure 3, a number of RA patients with ILD demonstrated elevated titers of antibodies specifically recognizing citrullinated versions of human Hsp90β derived from either a prokaryotic expression system (Hspβ-P; 14 of 58 patients) or a eukaryotic expression system (Hspβ-E; 11 of 58 patients), whereas only 1 of 27 patients with RA alone demonstrated low-titer antibody responses against either of these citrullinated Hsp90β subunits (1 of 27 patients positive for anti–Hsp90β-P antibodies and 0 of 27 positive for anti–Hsp90β-E antibodies).

Table 1. Anti-Hsp90β antibody reactivity in sera from patients with rheumatoid arthritis (RA)–associated interstitial lung disease (RA-ILD) (n = 58), RA alone (RA without ILD) (n = 27), mixed connective tissue disease (MCTD) (n = 41), or idiopathic pulmonary fibrosis (IPF) (n = 33), compared to normal control subjects (n = 21)*
 uncitHsp90β−/ citHsp90β+uncitHsp90β+/ citHsp90β+uncitHsp90β+/ citHsp90β−uncitHsp90β−/ citHsp90β−
  • *

    Reactivity with uncitrullinated (uncit) and citrullinated (cit) isoforms of Hsp90β was assessed, using human Hsp90β derived from either a prokaryotic expression system (Hsp90β-P) or a eukaryotic expression system (Hsp90β-E). Responses were designated as positive (+) or negative (−) based on cutoff values established for a reference RA-ILD serum sample.

  • P = 0.03 versus RA-ILD, by Fisher's exact test.

  • P = 0.01 versus RA-ILD, by Fisher's exact test.

  • §

    P = 0.002 versus RA-ILD, by Fisher's exact test.

  • P = 0.02 versus RA-ILD, by Fisher's exact test.

  • #

    P = 0.07 versus RA-ILD, by Fisher's exact test.

  • **

    P = 0.006 versus RA-ILD, by Fisher's exact test.

Hsp90β-P    
 RA-ILD143338
 RA without ILD10323
 MCTD23531
 IPF0§7323
 Normal control00318
Hsp90β-E    
 RA-ILD110146
 RA without ILD00126
 MCTD2#2037
 IPF0**1032
 Normal control00021
thumbnail image

Figure 3. Antibodies targeting citrullinated isoforms of Hsp90 distinguish rheumatoid arthritis (RA) patients with interstitial lung disease (RA-ILD) from RA patients without ILD. A–C, Scatterplots depict relative enzyme-linked immunosorbent assay–determined levels of anti– citrullinated Hsp90 antibodies (anti–Hsp90β-P [prokaryotic], anti–Hsp90β-E [eukaryotic], and anti-Hsp90α) in sera obtained from RA-ILD patients and RA patients without ILD. All sera were diluted 1:500 (with the exception of the anti–Hsp90β-E assessment, in which a serum dilution of 1:250 was used) prior to coincubation with various substrate antigens, including citrullinated versions of Hsp90β-P, Hsp90β-E, or Hsp90α (each at 1.0 μg/ml). Adjusted values for the OD450, calculated as the OD450 for substrate antigen minus the OD450 for no antigen, have been converted to relative units using a standard curve that was generated by serial dilutions of a reference serum (index case). Data points represent relative anti–citrullinated Hsp90α/β antibody levels obtained from individual patients, with open circles denoting samples recognizing both native/uncitrullinated and citrullinated versions of specified Hsp90 isoforms. Results are representative of multiple experiments in which all positive samples were confirmed by at least 2 independent assays. D, The dot plot illustrates relative anti–cyclic citrullinated peptide 2 (anti–CCP-2) reactivity in serum samples derived from RA-ILD patients with anti–citrullinated (cit) Hsp90α/β antibodies (n = 15) or those without anti–citrullinated Hsp90α/β antibodies (n = 34). The threshold for anti–CCP-2 positivity is 1 standard unit (equivalent to ∼30 units/ml of anti–CCP-2 reactivity).

Download figure to PowerPoint

Overall, these results, which did not differ substantially between the Veterans Administration (VA) and non-VA RA-ILD cohorts, indicated that the sensitivity of anti–citrullinated Hsp90β-P antibodies for identification of RA-ILD was 0.24 and the specificity was 0.96. For anti–citrullinated Hsp90β-E responses, the corresponding sensitivity and specificity values for RA-ILD were 0.19 and 1.0, respectively. Because anti–citrullinated Hsp90β-P and anti–citrullinated Hsp90β-E responses were not completely congruous in patients with RA-ILD (only 8 patients possessed antibodies recognizing citrullinated versions of both recombinant Hsp90β isoforms), we established a composite index based on autoantibody recognition of either citrullinated Hsp90β-P or citrullinated Hsp90β-E that increased the sensitivity of these assays (0.29 [17 of 58 patients]) without compromising specificity (0.96 [1 of 27 patients]).

Complementing these results, 6 of 58 patients with RA-ILD possessed anti-Hsp90α antibodies (Figure 3C)—demonstrating a serologic specificity that was not seen in the RA subgroup without ILD (0 of 27 patients). Because the sera from anti–citrullinated Hsp90α antibody–positive patients with RA-ILD also recognized at least one of the recombinant Hsp90β proteins, these results suggested that there was a significant (but not complete) cross-reactivity between antibodies targeting the highly homologous Hsp90α and Hsp90β isoforms. Although the findings were not statistically significant, a subset of RA patients with asymptomatic, radiographically defined ILD (RA with subclinical ILD) manifested a similar overlap of antibodies specifically targeting citrullinated versions of Hsp90β-P (2 of 10 patients), Hsp90β-E (1 of 10 patients), and/or Hsp90α (1 of 10 patients). Of note, the levels of anti–citrullinated Hsp90 antibodies did not consistently correlate with anti–CCP-2 titers in any of the RA subpopulations (Figure 3D and results not shown), highlighting key contributions of the underlying Hsp90 structure to the generation of unique citrullinated epitopes.

Use of anti–citrullinated Hsp90 antibody profiles to distinguish RA-ILD from MCTD and IPF.

Beyond RA, previous studies have suggested that a significant percentage of patients with MCTD possess antibodies recognizing native/uncitrullinated Hsp90 (27). We therefore assessed serum samples derived from a well-established cohort of MCTD patients for relative antibody titers against citrullinated and uncitrullinated versions of Hsp90. Tables 1 and 2 summarize the results of this analysis, demonstrating that several MCTD sera possessed antibodies targeting native/uncitrullinated Hsp90β-P (8 of 41 patients) and Hsp90β-E (2 of 41 patients)—with even fewer of the MCTD serum samples exclusively recognizing the citrullinated versions of Hsp90β-P (2 of 41 patients), Hsp90β-E (2 of 41 patients), or Hsp90α (0 of 41 patients). Of note, neither of the MCTD patients manifesting antibody responses against citrullinated Hsp90β-P/Hsp90β-E had documented evidence of ILD, although both had polyarthritis and one had elevated titers of anti–CCP-2 antibodies. Conversely, none of the 8 MCTD patients with definite ILD possessed antibodies recognizing citrullinated versions of Hsp90β or Hsp90α.

Table 2. Anti-Hsp90α antibody reactivity in sera from patients with rheumatoid arthritis (RA)–associated interstitial lung disease (RA-ILD) (n = 58), RA alone (RA without ILD) (n = 27), mixed connective tissue disease (MCTD) (n = 41), or idiopathic pulmonary fibrosis (IPF) (n = 33), compared to normal control subjects (n = 21)*
 uncitHsp90α−/ citHsp90α+uncitHsp90α+/ citHsp90α+uncitHsp90α+/ citHsp90α−uncitHsp90α−/ citHsp90α−
  • *

    Reactivity with uncitrullinated (uncit) and/or citrullinated (cit) isoforms of Hsp90α was assessed.

  • P = 0.17 versus RA-ILD, by Fisher's exact test.

  • P = 0.04 versus RA-ILD, by Fisher's exact test.

  • §

    P = 0.08 versus RA-ILD, by Fisher's exact test.

  • P = 0.19 versus RA-ILD, by Fisher's exact test.

RA-ILD61051
RA without ILD01026
MCTD00041
IPF0§0033
Normal control00021

Given these results supporting the relative specificity of anti–citrullinated Hsp90β and anti–citrullinated Hsp90α antibody responses for RA-ILD, when compared to RA without ILD as well as MCTD, we then examined a cohort of patients with IPF to further address the specificity of these antibodies for specific subtypes of ILD. As shown in Tables 1 and 2, a number of IPF patients demonstrated antibodies against native/uncitrullinated Hsp90β-P (10 of 33 patients) or Hsp90β-E (1 of 33 patients); however, none of the 33 IPF serum samples preferentially recognized citrullinated versions of either Hsp90β or Hsp90α.

Statistical comparison of anti–citrullinated Hsp90 antibody responses between RA-ILD and IPF cohorts revealed highly significant differences (P = 0.002, P = 0.006, and P = 0.08 for exclusive anti–citrullinated Hsp90β-P, anti–citrullinated Hsp90β-E, and anti–citrullinated Hsp90α seropositivity, respectively) and well-preserved specificity for RA-ILD (specificity of 1.0 for anti–citrullinated Hsp90β and anti–citrullinated Hsp90α)—suggesting that anti–citrullinated Hsp90 recognition was not merely an epiphenomenon of lung injury occurring outside the context of RA. At the same time, experiments demonstrating the presence of anti–citrullinated Hsp90 antibodies in a limited number of BAL fluid samples derived from patients with RA (2 of 5 patients) (results not shown) supported a lung-based origin for this RA-centric autoimmune response.

Clinical and demographic features of patients with RA-ILD separated by anti–citrullinated Hsp90 antibody status.

To determine whether the presence of anti–citrullinated Hsp90α/β antibodies was associated with distinctive clinical or demographic features, we compared a number of variables encompassing sex, racial background, HLA status, smoking history, medication usage, and disease activity in RA-ILD patients with this combined antibody specificity or without this antibody specificity (Table 3). Interestingly, the HLA–DRB1 shared epitope occurred more frequently in the RA-ILD patients possessing anti–citrullinated Hsp90 antibodies (80% of antibody-positive patients versus 53% of antibody-negative patients), although this association did not reach statistical significance (P = 0.11).

Table 3. Clinical and demographic characteristics of the patients with rheumatoid arthritis–associated interstitial lung disease*
 citHsp90+ (n = 15)citHsp90− (n = 34)P§
  • *

    TNF = tumor necrosis factor; RF = rheumatoid factor; anti–CCP-2 = anti–cyclic citrullinated peptide 2; ESR = erythrocyte sedimentation rate; HAQ = Health Assessment Questionnaire; DAS28 = Disease Activity Score in 28 joints.

  • Positive for anti–citrullinated (cit) Hsp90α and/or anti–citrullinated Hsp90β antibodies in the absence of antibodies targeting uncitrullinated Hsp90 isoforms.

  • Negative for both anti–citrullinated Hsp90α and anti–citrullinated Hsp90β antibodies.

  • §

    P > 0.05 was considered not significant (NS).

  • Based on HLA typing of the Veterans Administration Rheumatoid Arthritis cohort described in ref.41.

Sex, no. male:no. female15:033:1NS
Race, no. (%)   
 Caucasian15 (100)30 (88)NS
 African American0 (0)3 (9)NS
 Other0 (0)1 (3)NS
HLA–DRB1 shared epitope positive, no. (%)12 (80)18 (53)NS
Nodules, no. (%)9 (60)17 (50)NS
Smoking, no. (%)   
 Ever12 (80)31 (91)NS
 Never3 (20)3 (9)NS
Medication, no. (%)   
 Prednisone9 (60)23 (67)NS
 Methotrexate5 (33)14 (41)NS
 TNF inhibitor9 (60)13 (38)NS
RF titer, mean ± SD1,168.5 ± 1,519.8303.0 ± 319.80.003
Anti–CCP-2 titer, mean ± SD613.6 ± 560.7225.4 ± 363.50.007
ESR, mean ± SD mm/hour35 ± 25.236.4 ± 26.9NS
HAQ score, mean ± SD1.07 ± 0.430.95 ± 0.53NS
DAS28, mean ± SD3.78 ± 1.263.59 ± 1.13NS

In concert with this apparent shift in genetic profile, RA-ILD patients with anti–citrullinated Hsp90 antibodies tended to have increased titers of RF and anti–CCP-2 antibodies compared to RA-ILD patients without anti–citrullinated Hsp90 antibodies. Conversely, other parameters of disease activity (ESR, DAS28, and HAQ scores), medication profiles, and smoking histories did not differ significantly between these disease subsets (Table 3). Moreover, both anti–citrullinated Hsp90 antibody–positive and –negative subgroups of RA-ILD demonstrated a similar range of computed tomography abnormalities (consisting of ground-glass opacification, septal thickening, traction bronchiectasis, and/or honeycombing) as well as comparable levels of pulmonary function restriction (results not shown), suggesting that anti–citrullinated Hsp90 antibodies did not correlate with the histopathologic subtype or severity of ILD in this cohort.

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES
  8. Supporting Information

Through this study, we have demonstrated the power and feasibility of using differential immunoprecipitation to identify novel citrullinated autoantigens in RA-ILD. Among the candidate molecules identified by mass spectrometric sequencing of protein immunoprecipitates, citrullinated Hsp90α and citrullinated Hsp90β emerged as target antigens specifically recognized by sera from patients with RA-ILD. Comparative ELISA assessment of anti–citrullinated Hsp90 antibody responses in patients with MCTD and patients with IPF provided further evidence of both organ and disease specificity, indicating that antibodies exclusively recognizing citrullinated forms of Hsp90α and Hsp90β were largely restricted to RA-ILD.

Within the RA-ILD cohort that encompassed patients from several independent databases, cross-sectional analysis did not clearly identify unique demographic or environmental features associated with the development of anti–citrullinated Hsp90α/Hsp90β antibody responses. Although patients possessing these autoantibodies manifested greater levels of RF and higher (mean) anti–CCP-2 concentrations in conjunction with a trend toward shared epitope positivity, other parameters of articular and global disease activity did not differ significantly between these subgroups. Collectively, however, these data not only provided initial validation of our “reverse immunophenotyping” approach as a viable tool in biomarker discovery, but also supported an emerging paradigm featuring the lung as an important site of protein citrullination and immune-mediated injury in RA (19, 28, 29).

When compared to other disease- and/or organ-specific autoantibodies in disorders ranging from systemic lupus erythematosus and idiopathic inflammatory myopathy to systemic sclerosis (14, 30–35), composite anti–citrullinated Hsp90α/β antibodies demonstrated comparable levels of sensitivity (ranging between 20% and 30%) and specificity (>95%) for RA-ILD. While the lack of definitive prevalence data for subtypes of RA-ILD precludes accurate calculation of positive predictive value (PPV) and negative predictive value (NPV), coupling of the projected prevalence rates with the observed sensitivity and specificity of anti–citrullinated Hsp90 antibodies suggests that the PPV and NPV for RA-ILD may be relatively modest in an unselected population of RA patients.

A central question therefore concerns the determination of particular clinical features and/or additional serologic markers that might distinguish RA-ILD patients with this autoantibody specificity from those without this autoantibody specificity. Establishing these characteristics and enhancing the diagnostic utility of anti–citrullinated Hsp90 autoantibodies will clearly require prospective, longitudinal assessment of additional, well-defined cohorts of patients with RA-ILD (particularly non-veteran populations that would have a more balanced sex distribution), utilizing appropriate functional, radiographic, and histopathologic data. This type of analysis will not only define the relationship between anti–citrullinated Hsp90 seropositivity/titer and specific subsets of RA-ILD patients (segregated by severity of respiratory defect, ILD subtype, presence of additional extraarticular features, etc.), but will also facilitate assessment of correlations with genomic and proteomic profiles that clarify the underlying pathogenesis of this disorder.

From a methodologic perspective, refining these associations will also involve delineation of citrullinated Hsp90 epitope–specific recognition patterns that are suggested by comparison of anti–CCP-2 and anti–citrullinated Hsp90 antibody profiles in patients with RA-ILD. In short, although higher titers of anti–CCP-2 are statistically associated with a greater likelihood of anti–citrullinated Hsp90 seroconversion (note the prevalence of anti–citrullinated Hsp90 antibody positivity in different subsets of RA-ILD patients stratified by relative anti–CCP-2 titers [Figure 3D]), this correlation is not precise, given the frequent discordance in anti–CCP-2 and anti-Hsp90 titers that likely reflects specificity for citrullinated residues occurring in the context of unique Hsp90 structural motifs. Review of the mass spectrometry–derived sequence data (see Supplementary Figure 1, available on the Arthritis & Rheumatism web site at http://onlinelibrary.wiley.com/doi/10.1002/art.37881/abstract) actually reveals multiple sites of potential citrullination in Hsp90α/β immunoprecipitated from RA-ILD patient sera, highlighting a number of candidate epitopes. Although this sequence analysis reflects the byproduct(s) of in vitro citrullination and may be biased by the methods used for distinguishing deimination of arginine from deamidation of asparagine and/or glutamine in selected peptides, recognition of in vitro citrullinated Hsp90 by RA-ILD patient sera suggests that at least some of these epitopes are generated in vivo.

In terms of epitope recognition, results of the Hsp90α/β ELISAs reinforce the critical role of factors that influence protein folding and/or binding of substrate antigen. For example, although Hsp90β-P and Hsp90β-E differ only by glycosylation status (reflecting prokaryotic versus eukaryotic expression) and location of the histidine tag, these forms of recombinant human Hsp90β yield somewhat different recognition profiles among patients with RA-ILD. In fact, only 8 of 17 anti-Hsp90β–positive sera recognized both Hsp90β-P and Hsp90β-E, while 6 of 17 sera specifically targeted Hsp90β-P, and 3 of 17 sera exclusively bound Hsp90β-E. The fact that only a subset of Hsp90β antibody–positive sera recognized Hsp90α (6 of 17 samples)—despite the high degree of linear sequence homology between these Hsp90 isoforms—further underscores the likely contribution of subtle structural modifications to conformational epitopes. Beyond these direct effects on epitope formation in native Hsp90α/β isoforms, protein folding likely influences the accessibility of specific arginine residues to PAD and the subsequent generation of novel citrullinated epitopes responsible for the unique serologic profile of RA-ILD (note the differences in distribution of citrulline residues resulting from in vitro deimination of Hsp90α and Hsp90β, as shown in Supplementary Figure 1, available on the Arthritis & Rheumatism web site at http://onlinelibrary.wiley.com/doi/10.1002/art.37881/abstract).

Overall, the relative lack of redundancy in Hsp90α/β recognition profiles supports the existence of multiple, nonoverlapping epitopes and suggests that immune recognition of this putative autoantigen extends beyond simple cross-reactivity with citrullinated epitopes of an unrelated protein. Future use of complementary citrullinated and uncitrullinated peptides corresponding to defined regions of Hsp90α/β will begin to address this issue by facilitating more detailed epitope-mapping studies that should, in turn, enhance the sensitivity and specificity of our current ELISA system—particularly in individuals possessing antibodies targeting both citrullinated and native Hsp90 isoforms. Ultimately, demonstration of complex peptide recognition profiles (through standard, as well as competition, ELISAs) encompassing multiple epitopes in a given individual would provide presumptive evidence of intermolecular versus intramolecular epitope spreading and further substantiate the role of citrullinated Hsp90 as an autoantigen in RA-ILD.

Among the intriguing issues raised by these findings is whether citrullination of Hsp90 isoforms relates to the underlying pathogenesis of RA-ILD. In fact, previous analyses of RA and MCTD cohorts have shown statistical associations with antibodies against native/uncitrullinated Hsp90 (27), but the current study is the first to demonstrate a relationship between RA-ILD (or any autoimmune disorder) and autoantibodies targeting citrullinated versions of both Hsp90α and Hsp90β. Because this molecular chaperone is responsible for maintaining proper protein folding of multiple signaling molecules (36), the impact of posttranslational modifications such as citrullination on the structure as well as function of Hsp90 is likely to be highly significant, with ramifications for numerous signaling pathways in cells that remain viable following calcium-dependent activation of PAD isoforms (a scenario that is compatible with the findings from immunohistochemical studies demonstrating intracellular protein citrullination in various cell types ranging from lung-infiltrating mononuclear cells [19] to RA synoviocytes [37]).

At the same time, emerging evidence supporting the pleiotropic functions of extracellular Hsp90 (including innate immune signaling [38] as well as activation of matrix metalloproteinases potentially implicated in tissue remodeling [39, 40]) suggests several plausible mechanisms through which citrullination of Hsp90 released from dying/apoptotic cells could influence the development of ILD. Even if citrullination of Hsp90α/β subunits and the associated autoimmune responses are not directly responsible for initiating the disease processes in RA-ILD, the remarkably specific link between antibodies targeting citrullinated Hsp90 subunits and the development of RA-ILD is entirely consistent with a mechanistic paradigm in which injurious stimuli (e.g., smoking) elicit “danger responses” such as protein citrullination that, in turn, influence downstream signaling pathways as well as the host immune response. Future studies demonstrating relationships between RA-ILD and ACPA responses targeting additional molecules in specific Hsp90-mediated pathways will be crucial to fully validate this concept, particularly given the formal possibility that anti–citrullinated Hsp90 antibody responses are triggered by unrelated proteins possessing cross-reactive epitopes. Notwithstanding this caveat, the findings of this study provide an initial roadmap with which to focus future studies of aberrant protein expression and/or dysregulated signaling cascades in RA-ILD.

Establishing the link between RA-ILD and in vivo citrullination of Hsp90 subunits will ultimately require direct demonstration of citrullinated forms of this protein in lung tissue through immunohistochemistry. As a surrogate of in situ protein citrullination and tissue-specific immune responses, however, BAL fluid represents a more accessible source of biologic material with which to substantiate the role of citrullinated Hsp90α/β in disease pathways leading to RA-ILD. Preliminary studies have, in fact, demonstrated the presence of anti–citrullinated Hsp90 antibodies in BAL fluid (results not shown), although the number of specimens assessed thus far has not permitted statistical calculation of specificity for RA-ILD.

What is most interesting is that several BAL fluid specimens demonstrated anti–citrullinated Hsp90 antibody responses that were discordant with the serum responses, indicating that the presence of these autoantibodies in BAL fluid does not simply result from breach of the blood–alveolar barrier. Moreover, because at least one of the BAL fluid samples with evidence of anti–citrullinated Hsp90 antibody responses was derived from a patient with RA alone, BAL fluid may ultimately represent a more sensitive early indicator of RA patients prone to future development of ILD. While surveillance screening of BAL fluid from asymptomatic individuals is clearly not feasible, these preliminary findings are compatible with a sequence of pathogenic events in which in situ immune responses to citrullinated Hsp90 may predate the development of anti–citrullinated Hsp90 antibodies in the serum and/or clinically evident ILD.

Beyond issues specifically related to citrullinated Hsp90 and the role of this putative autoantigen in the pathogenesis of RA-ILD, this analysis highlights the power and versatility of our methodologic approach in defining unique autoantibody specificities that reflect tissue-specific alterations in protein expression and/or molecular signaling events. Particularly in relatively inaccessible organs such as the lung, the use of immunoprecipitation profiles as a surrogate measure of posttranslational protein modifications could prove to be quite valuable in unraveling disease pathogenesis. In fact, the potential applications extend beyond RA-ILD to encompass other extraarticular complications of RA that may be associated with unique ACPA specificities. Moreover, through the use of differential in-gel electrophoresis, this comparative immunoprecipitation approach can be applied even more broadly to various extract preparations (derived from different tissue-specific cell lines) and is therefore not limited to analysis of citrullinated autoantigens. Although longitudinal studies will be required to formally validate any observed associations between clinical phenotypes and novel autoantibody specificities delineated through these techniques, the current study provides important proof of principle that such an approach is feasible for development of composite biomarker profiles that provide prognostic, as well as mechanistic, insight relevant to diseases such as RA-ILD.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES
  8. Supporting Information

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. Ascherman 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. Oddis, Ascherman.

Acquisition of data. Harlow, Rosas, Gochuico, Mikuls, Dellaripa, Oddis, Ascherman.

Analysis and interpretation of data. Rosas, Ascherman.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
ART_37881_sm_SupplFigure1.docx1294KSupplementary Figure 1

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.