Coexisting autoantibodies against transcription factor Sp4 are associated with decreased cancer risk in dermatomyositis patients with anti-TIF1γ autoantibodies

Abstract

Objectives:

In dermatomyositis (DM), autoantibodies are associated with unique clinical phenotypes. For example, anti-TIF1γ autoantibodies are associated with an increased risk of cancer. The purpose of this study was to discover novel DM autoantibodies.

Methods:

Phage ImmunoPrecipitation Sequencing using sera from 43 DM patients suggested that transcription factor Sp4 is a novel autoantigen; this was confirmed by showing that patient sera immunoprecipitated full-length Sp4 protein. Sera from 371 Johns Hopkins myositis patients (255 with DM, 28 with antisynthetase syndrome, 40 with immune-mediated necrotizing myopathy, 29 with inclusion body myositis, and 19 with polymyositis, 80 rheumatologic disease controls (25 with Sjogren’s syndrome, 25 with systemic lupus erythematosus, and 30 with rheumatoid arthritis), and 200 healthy comparators were screened for anti-SP4 autoantibodies by an enzyme-linked immunosorbent assay. A validation cohort of 46 anti-TIF1γ-positive patient sera from the University of Pittsburgh were also screened for anti-Sp4 autoantibodies.

Results:

Anti-Sp4 autoantibodies were present in 27 (10.5%) DM patients and 1 (3.3%) rheumatoid arthritis patient but not in other clinical groups. In DM patients, 96.3% of anti-Sp4 autoantibodies were detected in those with anti-TIF1γ autoantibodies. Among 26 TIF1γ-positive patients with anti-Sp4 autoantibodies, none (0%) had cancer. In contrast, among 35 TIF1γ-positive patients without anti-Sp4 autoantibodies, 5 (14%; p=0.04) had cancer. In the validation cohort, among 15 TIF1γ-positive patients with anti-Sp4 autoantibodies, 2 (13.3%) had cancer. By comparison, among 31 TIF1γ-positive patients without anti-Sp4 autoantibodies, 21 (67.7%; p<0.001) had cancer.

Conclusions:

Anti-Sp4 autoantibodies appear to identify a subgroup of anti-TIF1γ-positive DM patients with lower cancer risk.

INTRODUCTION

The idiopathic inflammatory myopathies (IIM) are a heterogeneous family of diseases that includes dermatomyositis (DM), immune-mediated necrotizing myopathy (IMNM), the antisynthetase syndrome (ASyS), polymyositis (PM), and inclusion body myositis (IBM)¹. Most patients with IIM have a myositis-specific autoantibody (MSA). Among those with DM, approximately 70% have an MSA recognizing either TIF1γ, NXP2, Mi2, MDA5, or SAE. Importantly, each MSA is associated with a unique clinical phenotype. For instance, DM patients with anti-TIF1γ autoantibodies have a substantially increased risk of cancer² whereas those with anti-Mi2 autoantibodies do not³. Although MSAs are usually mutually exclusive, there are exceptions. For example, some anti-MDA5-positive DM patients develop a second MSA recognizing splicing factor proline/glutamine-rich (SFPQ); these patients have a decreased risk of arthritis compared to anti-MDA5-positive patients without anti-SFPQ autoantibodies⁴.

Phage ImmunoPrecipitation Sequencing (PhIP-Seq) is a programmable bacteriophage display based method for high throughput antibody binding analysis. Here we performed PhIP-Seq with a library of 274,207 overlapping 90 amino acid long peptides that tile across the human proteome^{5, 6} to identify novel autoantibodies in DM patients. This approach revealed novel autoantibodies recognizing transcription factor Sp4 in DM patients with co-existing anti-TIF1γ autoantibodies. Furthermore, we show that anti-Sp4 autoantibodies were more prevalent in two cohorts of TIF1γ-positive DM patients who do not have cancer.

PATIENTS AND METHODS

Patients and serum samples

The discovery cohort consisted of 43 patients enrolled in the Johns Hopkins Myositis Center Longitudinal study between 2002 and 2016 with a diagnosis of DM based on the criteria of Bohan and Peter^{7, 8} whose serum tested negative for all MSAs by the EUROLINE Autoimmune Inflammatory Myopathies 16 Ag (IgG) test kit, which includes the following antigens: Mi-2α, Mi-2β, TIF1γ, MDA5, NXP2, SAE1, Ku, PM-Scl100, PM-Scl75, Jo-1, SRP, PL-7, PL-12, EJ, OJ and Ro-52.

The screening cohort included myositis patients enrolled in the Johns Hopkins Myositis Center Longitudinal Cohort study between 2002 and 2018. This included patients with DM based on the criteria of Bohan and Peter^{7, 8}, ASyS defined by the presence of an antisynthetase autoantibody in patients with DM or PM according to the criteria of Bohan and Peter, IMNM defined by the presence of anti-SRP or anti-HMGCR autoantibodies in patients with proximal weakness and CK elevation as per 2018 ENMC criteria⁹, IBM defined by the Lloyd and Greenberg criteria¹⁰, as well as PM patients defined as those who fulfilled the criteria of Bohan and Peter for PM but who did not have ASyS or IMNM. Patients were considered positive for autoantibodies recognizing Mi2, NXP2, MDA5, Jo1, SRP, HMGCR, SAE, or PmScl if they tested positive by at least two immunologic techniques from among the following: ELISA, in vitro transcription and translation immunoprecipitation, line blotting (EUROLINE Autoimmune Inflammatory Myopathies 16 Ag (IgG) test kit), PhIP-Seq,¹¹ immunoprecipitation from S35-labeled HeLa cell lysates, or immunoprecipitation blotting.^12–14. All sera identified as anti-TIF1γ-positive were positive by anti-TIF1γ ELISA (MBL, RG-7854R). Except for one sample that did not undergo additional testing due to lack of availability, each ELISA-positive sera also tested positive by at least one additional anti-TIF1γ detection method (see Supplementary Figure 1).

A validation cohort included serum from 46 DM patients from the University of Pittsburgh who tested positive for anti-TIF1γ by ELISA (MBL anti-TIF1γ ELISA, RG-7854R); 23 of these had cancer and 23 of these did not have cancer.

Serum from 30 patients from the Instituto Nacional de Ciencias Medicas y de la Nutricion Salvador Zubiran with RA based on the 2010 EULAR/ACR classification criteria, as well as 25 patients with systemic lupus erythematosus (SLE) based on the 2019 EULAR/ACR criteria, 25 patients with Sjögren’s syndrome based on the 2016 EULAR/ACR classification criteria, and 200 healthy controls, all enrolled in studies at the National Institutes of Health, were also screened by ELISA for anti-Sp4 autoantibodies. Additional healthy control samples used for the analysis of peptidome data were from subjects self-reported to be free of autoimmune disease. Supplementary Table 1 summarizes demographic data for the following cohorts: myositis, healthy control, RA, SLE, and Sjogren’s syndrome.

Of note, sera were obtained from patients when they enrolled in a research study at one of the participating institutions. These patients may have been previously undiagnosed and untreated or already on immunosuppressive therapy for various lengths of time.

In the screening cohort, muscle strength was evaluated by the examining physician using the Medical Research Council scale. This scale was transformed to Kendalĺs 0–10 scale for analysis purposes as previously described.¹⁵ For the purposes of analyses, right- and left-side measurements for arm abduction and hip flexion strength were combined and the average was used for calculations (possible range 0–10). Skin manifestations (i.e., heliotrope rash or Gottron’s sign), weakness, symptoms of esophageal involvement, antisynthetase syndrome-associated clinical features (e.g. mechanics hands, Raynaud’s phenomenon, arthritis, fever), and other clinical features were documented both retrospectively at the onset of the disease (by asking patients about features present at the onset of disease) and prospectively at each visit. Interstitial lung disease was defined through a multidisciplinary approach as recommended by the American Thoracic Society.¹⁶ Cancer-associated myositis was defined as a malignancy occurring within 3 years either before or after the onset of myositis symptoms.

Standard protocol approvals and patient consents

This study was approved by the Institutional Review Boards at Johns Hopkins (IRB# 00235256), the National Institutes of Health (IRB# AR-0196), the University of Pittsburgh (IRB# MOD19090054–001), and the Instituto Nacional de Ciencias Medicas y de la Nutricion Salvador Zubiran (reference# 1243); written informed consent was obtained from each participant.

Screening for autoantibodies by PhIP-Seq

The standard PhIP-Seq procedure¹¹ was used to profile sera from 67 DM patients that were seronegative for Mi2, NXP2, TIF1γ or MDA5 by EUROLINE line blot (discovery cohort). Briefly, the IgG concentration of each serum sample was measured via ELISA assay, which allowed normalization of the IgG input (2 μg per reaction) into the PhIP-Seq assay. Serum antibody was mixed with the human peptidome library comprised of 274,207 90-aa peptides¹⁷ and incubated overnight. Antibody and antibody-bound phage were captured using protein A and protein G coated Dynal magnetic beads (Invitrogen #10002D & #10004D). The immunoprecipitated phage DNA library was then amplified using PCR, with sample-specific DNA barcodes added during a second PCR reaction. Pooled amplicons were sequenced using an Illumina NextSeq instrument. An informatics pipeline was used for sample demultiplexing and alignment. On each 96-well plate, eight wells were reserved for mock immunoprecipitations, in which the PhIP-Seq assay was performed without patient sera to account for background binding of the phage. Then, as described previously¹¹, patient data normalization was performed by comparison to mock immunoprecipitations. Briefly, peptide enrichment z-scores are calculated for each peptide and each sample based on the distribution of the corresponding peptide’s reads in the mock immunoprecipitation reactions. Z-scores greater than 7 are considered positive based on prior studies. Finally, the normalized PhIP-Seq profiles were compared against a previously established PhIP-Seq database of 683 healthy volunteers using a custom case-control analysis script to identify reactivities specifically associated with DM (https://brandonsie.github.io/phipcc/). For each peptide, Fisher’s exact test with Bonferroni correction is used to detect a peptide-disease association by considering the proportion of samples with z-scores above the 95^th percentile z-score among healthy controls. To identify potentially co-associated peptides, peptides with a significant disease association are subjected to hierarchical clustering with Ward agglomeration using Euclidean distance.

Immunoprecipitation using ³⁵S-labeled in vitro transcription/translated (IVTT) Sp4

DNA encoding full-length human Sp4 was purchased (Origene) and used in IVTT reactions (Promega), generating ³⁵S-labeled Sp4 protein. Immunoprecipitation of radiolabeled Sp4 was performed using patient serum or a mouse monoclonal anti-Sp4 positive control antibody (sc-515738, Santa Cruz Biotechnology), as previously described¹⁸. Immunoprecipitates were reduced, boiled, subjected to electrophoresis on 10% sodium dodecyl sulfated-polyacrylamide gels, and visualized using a Typhoon FLA 9500 scanner (GE Healthcare Life Sciences, PA, USA).

Anti-Sp4 and negative control ELISAs

ELISA plates (96-well) were coated overnight at 4°C with 100 ng of human recombinant Sp4 protein that included a GST tag (H00006671-P01, Abnova Corporation, Taipei, Taiwan) diluted in 100 μL of phosphate-buffered saline (PBS). After washing the plates with PBS including 0.05% Tween-20 (PBS-T) and blocking with 300 μl of 5% bovine serum albumin (BSA) in PBS-T for 1 hour at 37°C, the plates were washed with PBS-T. 100 μl of diluted human serum samples (1:400 with 1% BSA/PBS-T) was added to each well and incubated for 1 hour at 37°C. After washing with PBS-T, 100 μl of HRP-labeled goat anti-human IgG antibody that reacts with the whole human IgG molecule as well as with the light chains of other human immunoglobulins (1:10,000, catalog# 109–036-088, Jackson ImmunoResearch Lab, PA, USA) was added and incubated for 30 minutes at 37°C in the dark. After washing the plate with PBS-T and PBS, 100 μl of SureBlue TMB microwell peroxidase enzyme substrate kit (95059–286, KPL, MA, USA) was added. Reactions were stopped after 8 minutes. The absorbance at 450 nm was determined. Test sample absorbances were normalized to the sera of an arbitrary positive control sample, a reference sample included in every ELISA. The cutoff for a normal anti-Sp4 autoantibody titer (0.29 arbitrary units) was defined as the mean plus two standard deviations of the normalized absorbances of the 200 healthy comparators. This cutoff was determined to be optimal based on a graphical analysis of the normalized absorbances.

To control for potential reactivity against the GST tag included in the Sp4 protein used for ELISA, sera that tested positive for anti-Sp4 autoantibodies by ELISA were also screened for anti-GST reactivity by ELISA. Several sera samples were positive by ELISA for both Sp4-GST and GST alone; these were considered negative for anti-Sp4 autoantibodies.

Statistical analysis

Dichotomous variables were expressed as percentages and absolute frequencies, and continuous features were reported as means and standard deviations (SD). Pairwise comparisons for categorical variables between groups were made using the chi-square test or Fisher’s exact test, as appropriate. Studentś t-test was used to compare continuous variables among groups and paired t-test was used to compare the level of weakness of different muscle groups. CK, a highly positively skewed variable, was expressed as median, first, and third quartile for descriptive purposes, and was transformed through a base-10 logarithm for regression analysis.

Statistical analyses related to clinical variables were performed using Stata/MP 14.1. A 2-sided p-value of 0.05 or less was considered statistically significant with no adjustment for multiple comparisons.

Patient and Public Involvement statement

Patients and the public were not involved in the design, conduct, reporting, or dissemination plans of the present research.

Data availability statement.

All data relevant to the study are either included in the article or will be shared upon request.

RESULTS

Identification of Sp4 as a novel autoantigen

In an effort to discover novel DM-associated autoantibodies, we assembled a cohort of sera from DM patients who were negative for MSAs by EUROIMMUN line blot, which tests for DM-specific autoantibodies including anti-TIF1γ, anti-NXP2, anti-Mi2, anti-MDA5, and anti-SAE. These 43 serum samples were used in the PhIP-Seq assay with a T7 phage display library of 274,207 peptides spanning all open reading frames in the human genome as overlapping 90 amino acid peptides¹⁷. While PhIP-Seq does not detect antibodies directed against conformational or post-translational epitopes, the approach is unbiased and provides a higher resolution map of autoantibody binding specificities¹⁹. An average of 122 human peptides per sample (95% CI 84–159 peptides) were considered significantly reactive (see Methods). A case-control analysis compared PhIP-Seq profiles of the DM discovery cohort against profiles of 663 healthy controls to detect DM-associated antibodies (see Methods). Among the 43 serum samples analyzed, 13 recognized from 1 to 4 peptides corresponding to Transcription factor Sp4 (Table 1). Recognition of non-overlapping epitopes is indicative of a polyclonal antibody response that is more likely to be antigen driven.

Table 1.

Sp4 peptides identified by PhIP-Seq.

Peptide name	Amino acid position	Sequence
6	226–315	NQTVPVQIRPGVSIPLQLQTLPGTQAQVVTTLPINIGGVTLALPVINNVAAGGGTGQVGQPAATADSGTSNGNQLVSTPTNTTTSASTMP
10	406–495	QIQIQQPQQQIIQAIPPQSFQLQSGQTIQTIQQQPLQNVQLQAVNPTQVLIRAPTLTPSGQISWQTVQVQNIQSLSNLQVQNAGLSQQLT
11	451–540	PTQVLIRAPTLTPSGQISWQTVQVQNIQSLSNLQVQNAGLSQQLTITPVSSSGGTTLAQIAPVAVAGAPITLNTAQLASVPNLQTVSVAN
16	676–765	PFICNWMFCGKRFTRSDELQRHRRTHTGEKRFECPECSKRFMRSDHLSKHVKTHQNKKGGGTALAIVTSGELDSSVTEVLGSPRIVTVAA

To determine which sera had immunoreactivity against native Sp4, we generated full-length radiolabeled Sp4 protein by IVTT and used human serum or a rabbit anti-Sp4 positive control antibody to immunoprecipitate the protein (Figure 1). The positive control anti-Sp4 antibody and each of the 6 serum samples that recognized 3 or 4 distinct Sp4 peptides by PhIP-Seq (Figure 1, lanes 3–8) efficiently immunoprecipitated full-length Sp4 protein. In contrast, serum samples that recognized just 1 or 2 Sp4 peptides only weakly immunoprecipitated full-length Sp4 protein (lanes 9–15).

Figure 1. Human serum samples from DM patients immunoprecipitate full-length Sp4 protein.

Sera from DM patients that recognized at least one Sp4 peptide by PhIP-Seq were used to immunoprecipitate radiolabeled full-length Sp4 protein. The Sp4 peptides recognized by each serum using PhIP-Seq are indicated below each lane. Those serum samples recognizing 3–4 distinct Sp4 peptides (lanes 3–8) by PhIP-Seq immunoprecipitated full-length Sp4 protein more-efficiently than those serum samples that only recognized 1–2 Sp4 peptides by PhIP-Seq (lanes 9–15). The input Sp4 protein used for immunoprecipitation is shown in lane 1. A commercial rabbit anti-Sp4 autoantibody was used to immunoprecipitate radiolabeled Sp4 in lane 2. Healthy control sera did not immunoprecipitate full-length Sp4 protein (lanes 16–19).

Unexpectedly, since they had each tested negative for anti-TIF1γ autoantibodies by EUROIMMUN line blot, we noted that all 6 serum samples that efficiently immunoprecipitated full-length Sp4 protein also recognized a peptide corresponding to TIF1γ by the PhIP-Seq assay (data not shown). We subsequently confirmed that these 6 serum samples were anti-TIF1γ-positive by ELISA. This demonstrates the high sensitivity of PhIP-Seq and suggests that the line blot test may not be a sufficiently sensitive assay for detecting anti-TIF1γ autoantibodies. Indeed, a recent report showed that the EUROIMMUN line blot has good specificity but poor sensitivity for detecting anti-TIFγ autoantibodies compared to immunoprecipitation detection methods²⁰.

Screening for anti-Sp4 autoantibodies in patients with IIM, other rheumatologic conditions, and healthy controls

To screen patients rapidly for anti-Sp4 autoantibodies, we developed an ELISA using Sp4 protein with a GST tag. We defined a serum sample as being positive for anti-Sp4 autoantibodies if the relative absorbance was two standard deviations or higher than the mean value of 200 healthy control subjects and the serum sample was not reactive against the GST tag in an anti-GST ELISA (see Methods). Using this method, we found that 6 of the 13 samples recognizing Sp4 peptides by PhIP-Seq were also ELISA positive; these were the same 6 samples that recognized 3 or 4 Sp4 peptides and which most efficiently immunoprecipitated full-length Sp4 protein (Figure 1, lanes 3–8). In contrast, among the 30 samples that did not recognize Sp4 peptides by PhIP-Seq, 26 were available for further testing and each of these was negative for anti-Sp4 autoantibodies by ELISA.

The screening cohort consisted of 371 serum samples from myositis patients seen at the Johns Hopkins Myositis Center (255 with DM, 28 with ASyS [all anti-Jo1-positive], 40 with IMNM, 19 with PM, and 29 with IBM). Among these, 27 (7.3%) sera samples were anti-Sp4-positive by ELISA (Figure 2) and all of them had DM. Thus, 10.5% of DM patients were anti-Sp4-positive. 96.3% of the anti-SP4-positive DM samples were also TIF1γ-positive (n=26). Out of 194 anti-TIF1γ-negative DM patients, just one (0.5%), a single anti-NXP2-positive patient, had anti-Sp4 autoantibodies. In contrast, no patient with anti-MDA5, anti-Mi2, anti-SAE, or anti-PMScl autoantibodies also had anti-Sp4 autoantibodies.

Figure 2. Anti-Sp4 autoantibody titers determined by ELISA in sera from patients with myositis, other rheumatologic conditions, and healthy comparators.

Anti-Sp4 titers were determined by ELISA for sera from 371 adult myositis patients, 25 lupus patients, 25 Sjogren’s syndrome patients, 30 rheumatoid arthritis patients, and 200 healthy controls. The dotted line indicates the cut-off used to define anti-Sp4 positive sera. Blue dots represent samples that were anti-TIF1γ-negative and orange dots represent samples that were anti-TIF1γ-positive. Several serum samples reacted on ELISA with both the Sp4 protein (which has a GST tag) and with the GST protein alone. For clarity, these samples were excluded from this figure, but are included as anti-TIF1γ-negative samples in the subsequent analyses.

We also tested for anti-Sp4 autoantibodies in 25 patients with Sjogren’s syndrome, 25 patients with systemic lupus erythematosus, and 30 patients with rheumatoid arthritis. One (3.3%) of the rheumatoid arthritis patients was anti-Sp4-positive whereas the rest of the patients with other rheumatologic conditions were negative for these autoantibodies (Figure 2). Overall, compared to healthy controls (3%), there were significant increases in the prevalence of anti-Sp4 antibodies in the following groups: myositis (7.3%; p=0.02), DM (10.5%; p=0.0009), and anti-TIF1γ-positive DM (42.6%; p=1.5e-16). The prevalence of anti-Sp4 autoantibodies was the same in RA and the healthy control groups (~3%). Taken together, these results indicate that anti-Sp4 autoantibodies are strongly associated with anti-TIF1γ-positive DM.

The clinical features of TIF1γ-positive DM patients with and without co-existing anti-Sp4 autoantibodies.

Most demographic (Table 2) and clinical (Table 3 and Supplementary Table 2) features were similar between TIF1γ-positive DM patients with and without anti-Sp4 autoantibodies. However, patients with anti-Sp4 autoantibodies had measures of muscular strength that were significantly higher than those who were anti-Sp4-negative (Supplementary Table 2). We also noted that among 26 TIF1γ-positive patients from Johns Hopkins with anti-Sp4 autoantibodies, none had cancer. In contrast, among 35 TIF1γ-positive patients without anti-Sp4 autoantibodies, 5 (14%) had cancer (p=0.04). To confirm that anti-Sp4 autoantibodies are associated with absence of cancer in anti-TIF1γ-positive DM patients, we screened an additional cohort of 46 TIF1γ-positive DM patients from the University of Pittsburgh (Figure 3). Among 15 TIF1γ-positive patients with anti-Sp4 autoantibodies, 2 (13.3%) had cancer. By comparison, among 31 TIF1γ-positive patients without anti-Sp4 autoantibodies, 21 (67.8%; p<0.001) had cancer. Thus, anti-Sp4 autoantibodies are associated with a reduced risk of cancer in anti-TIF1γ-positive DM patients. Conversely, patients with anti-TIF1γ autoantibodies that do not co-express anti-Sp4 autoantibodies are at very high risk for cancer.

Table 2:

General features of anti-TIF1γ patients with and without anti-Sp4 autoantibodies.

	Anti-Sp4+ (n=26)	Anti-Sp4− (n=35)	p-value	Total (n=61)
Female sex	81% (21)	89% (31)	0.5	85% (52)
Race
White	85% (22)	86% (30)	1.0	85% (52)
Black	0% (0)	11% (4)	0.1	7% (4)
Other races	15% (4)	3% (1)	0.2	8% (5)
Age of onset (years)	43.0 (13.8)	50.6 (15.3)	0.05	47.4 (15.1)
Time of follow-up (years)	7.1 (4.3)	4.8 (2.7)	0.01	5.8 (3.7)
Number of visits per participant	10.2 (7.7)	10.9 (6.7)	0.7	10.6 (7.1)
Cancer associated myositis	0% (0)	14% (5)	0.04	8% (5)
Death during follow-up	0% (0)	9% (3)	0.3	5% (3)
Anti-Ro52	15% (4)	17% (6)	1.0	16% (10)
Treatments
Corticosteroids	81% (21)	71% (25)	0.4	75% (46)
Azathioprine	27% (7)	29% (10)	0.9	28% (17)
Methotrexate	62% (16)	54% (19)	0.6	57% (35)
Mycophenolate	46% (12)	46% (16)	1.0	46% (28)
IVIG	38% (10)	57% (20)	0.1	49% (30)
Rituximab	12% (3)	17% (6)	0.7	15% (9)

Dichotomous variables were expressed as percentage (count) and continuous variables as mean (SD). Bivariate comparisons of continuous variables were made using Studen’s t-test while bivariate comparisons of dichotomous variables were made either using chi-squared test or Fisher’s exact test, as appropriate.

Table 3:

Cumulative clinical features of anti-TIF1γ patients with and without anti-Sp4 autoantibodies.

	Anti-Sp4+ (n=26)	Anti-Sp4− (n=35)	p-value	Total (n=61)
Muscle involvement
Muscle weakness	85% (22)	80% (28)	0.7	82% (50)
Skin involvement
DM-specific skin involvement	100% (26)	100% (35)	1	100% (61)
Raynaud’s phenomenon	38% (10)	17% (6)	0.06	26% (16)
Mechanics hands	23% (6)	26% (9)	0.8	25% (15)
Calcinosis	8% (2)	17% (6)	0.4	13% (8)
Subcutaneous edema	8% (2)	17% (6)	0.4	13% (8)
Lung involvement
Interstitial lung disease	4% (1)	0% (0)	0.4	2% (1)
Esophageal involvement
Gastroesophageal reflux disease	27% (7)	34% (12)	0.5	31% (19)
Dysphagia	42% (11)	54% (19)	0.4	49% (30)
Joint involvement
Arthritis	19% (5)	17% (6)	1.0	18% (11)
Arthralgia	58% (15)	57% (20)	1.0	57% (35)
Systemic involvement
Fever	12% (3)	9% (3)	1.0	10% (6)

Clinical features include those that developed at any time during the disease course. Chi-squared or Fisher’s exact tests were used to compare each one of the clinical groups with the anti-Mi2 patients.

Figure 3. Anti-Sp4 autoantibody titers in the Pittsburgh cohort of anti-TIF1γ-positive subjects.

Serum samples from 23 anti-TIF1γ-positive subjects with cancer and 23 anti-TIF1γ-positive subjects without cancer were screened for anti-Sp4 autoantibodies by ELISA. The dotted line indicates the cut-off used to define anti-Sp4 positive sera.

We next analyzed the evolution of anti-Sp4 autoantibody titers during the disease course by analyzing longitudinally collected serum samples. We found that titers decreased in many patients but did not normalize after treatment of the DM (Supplementary Figure 2). Finally, to determine whether anti-TIF1γ-positive patients without anti-Sp4 autoantibodies at their initial visit to the Johns Hopkins Myositis Center might develop them during the course of the disease, we screened the most recently collected serum samples from 24 TIF1γ-positive and anti-Sp4-negative patients. During an average of 4.7 years (SD 2.3) between the collection of the first and most recent serum samples, only 2 (8.3%) of these became positive for anti-Sp4 autoantibodies.

DISCUSSION

In this study, we used PhIP-Seq, a bacteriophage display-based method, to identify Sp4 as a candidate autoantigen in DM patients; this finding was confirmed by immunoprecipitation of full-length Sp4 protein with patient sera. Using an anti-Sp4 ELISA, we screened several cohorts of myositis patients and controls and found that anti-Sp4 autoantibodies were found in 11% of DM, 3% of healthy controls, 3% of RA patients, but not in patients with SLE or Sjogren’s Syndrome. Unexpectedly, among DM patients, anti-Sp4 autoantibodies were found almost exclusively among those with co-existing anti-TIF1γ autoantibodies. Moreover, although anti-TIF1γ-positive DM patients have an increased risk of cancer, we discovered that the prevalence of cancer was relatively decreased in those with co-existing anti-Sp4 autoantibodies in two independent cohorts.

It is unclear why an immune response against TIF1γ would be associated with an increased prevalence of cancer in DM patients, while a co-existing immune response against Sp4 would be associated with a relatively reduced prevalence of malignancy compared to other anti-TIF1γ-positive patients. It has been proposed that in some patients, tumors elicit an immune response against autoantigens such as TIF1γ that is redirected to skin and muscle, causing DM, but is insufficient to eradicate the underlying tumor²¹; these patients would present with anti-TIF1γ-positive DM along with a clinically apparent malignancy. In other patients, we hypothesize that an anti-tumor immune response against TIF1γ becomes redirected to skin and muscle, causing DM, but that an additional immune response is raised against Sp4, which causes or is associated with effective eradication of the tumor; these patients would present with both anti-TIF1γ and anti-Sp4 autoantibodies, but no tumor. This speculative model remains to be validated.

Recently, autoantibodies against other targets, including the cell division cycle and apoptosis regulator 1 (CCAR1) protein, were also found to be associated with decreased cancer prevalence among anti-TIF1γ-positive DM patients²²; anti-CCAR autoantibodies were present in 36% of those without cancer and in 22% of those with cancer. When the two cohorts of anti-TIF1γ-positive patients used in the current study are combined, anti-Sp4 autoantibodies were present in 49.4% (39 of 79) of those without cancer and 7.1% (2 of 28) of those with cancer. Future studies looking at both anti-CCAR1 and anti-Sp4 autoantibodies in larger cohorts of DM patients will be required to determine which antibody or combination of antibodies has a stronger association with low cancer prevalence in anti-TIF1γ-positive DM patients.

This study has several limitations. First, PhIP-Seq only detects antibodies recognizing 90 amino acid-long peptides. Thus, autoantibodies recognizing conformational epitopes or post-translational modifications will not be detected using this method. Second, the Sp4 protein used for the anti-Sp4 ELISA also included a GST tag and we found that several sera recognized the GST tag. Although we overcame this limitation by identifying such sera using a secondary anti-GST ELISA, we plan to develop a more specific anti-Sp4 ELISA utilizing Sp4 protein without a tag. Third, our cohorts of disease control sera were relatively small and screening larger cohorts of patients with various autoimmune diseases would be of interest.

These limitations notwithstanding, we have shown that anti-Sp4 autoantibodies are predominantly found in anti-TIF1γ-positive DM patients without cancer. Future prospective studies including larger numbers of patients will be required to establish whether testing for anti-Sp4 autoantibodies may have clinical utility in identifying a subpopulation of anti-TIF1γ-positive patients who may be safely spared aggressive malignancy screening.

Supplementary Material

Supp2

Supplementary Figure 1. Summary of tests used to establish anti-TIF1γ seropositivity in the Hopkins screening cohort. All serum samples tested positive by the commercial MBL ELISA (“ELISA”). Except for one sample that was unavailable for further testing, all serum samples tested positive for anti-TIF1γ autoantibodies by at least one additional method including an in-house IP/blot validated at Hopkins (“IP_BLOT_1”)22, an IP/blot performed at the Oklahoma Medical Research Foundation (OMRF; “IP_BLOT_2”), IP performed at OMRF (“IP”), or by PhIP-Seq (as described in Methods).

Supp3

Supplementary Figure 2. Longitudinal analysis of anti-Sp4 titers. Longitudinal anti-Sp4 titers are shown for 13 anti-Sp4-positive subjects who had multiple serum samples collected over time. The dotted line indicates the cut-off used to define anti-Sp4 positive sera.

Supp4

Supplementary Figure 3: Raw image of gel used to generate Figure 1. The original image of the gel for Figure 1 captured by the Typhoon FLA 9500 scanner (GE Healthcare Life Sciences, PA, USA) is shown. Of note, molecular weight markers were included in the left-most lane but were not visualized using the scanner (which only detects radiolabeled proteins); these molecular weight markers were used to provide the molecular weights as shown in Figure 1.

KEY MESSAGES.

What is already known about this subject?

• Dermatomyositis patients with anti-TIF1γ autoantibodies have an increased risk of cancer.

What does this study add?

• Anti-Sp4 autoantibodies are present in ~11% of patients with dermatomyositis, ~3% of those with rheumatoid arthritis, and ~3% of healthy controls.

• Among those with dermatomyositis, 96% of anti-Sp4 autoantibody-positive patients had co-existing anti-TIF1γ autoantibodies.

• Anti-Sp4 autoantibodies were present in ~43% of dermatomyositis patients with anti-TIF1γ autoantibodies.

• In two cohorts, anti-TIF1γ-positive dermatomyositis patients with anti-Sp4 autoantibodies were significantly less likely to have cancer.

• Anti-Sp4 autoantibodies were not detected in patients with antisynthetase syndrome, immune-mediated necrotizing myopathy, or inclusion body myositis.

• Autoantibodies against Sp4 were not detected in those with systemic lupus erythematosus or Sjogren’s syndrome.

How might this impact on clinical practice?

Testing for anti-Sp4 autoantibodies may define a population of anti-TIF1γ-positive dermatomyositis patients without a substantially increased risk of cancer.