Anti-TFAM antibodies link mitochondrial damage with anti-phospholipid syndrome and thrombosis in SLE

Abstract

Objectives.

Mitochondria are a source of autoantigens and damage-associated molecular patterns (DAMPs) in systemic lupus erythematosus (SLE). Nucleoids carrying TFAM (transcription factor A, mitochondrial) and mitochondrial DNA (mtDNA) are important DAMPs in SLE. While mtDNA has been associated with anti-dsDNA antibodies and type I interferon (IFN-I), the immunogenic role of TFAM in SLE pathogenesis is unknown. Here, we characterized the clinical and transcriptional phenotypes linked to anti-TFAM antibodies in SLE.

Methods.

Anti-TFAM antibodies were discovered in an exploratory sample of 22 SLE patients and 9 healthy controls. To define the prevalence, clinical significance, and associations with transcriptional profiles and IFN levels, anti-TFAM antibodies were determined by ELISA in 98 healthy controls and 158 SLE patients. Sera from dermatomyositis, rheumatoid arthritis and primary anti-phospholipid syndrome (PAPS) were additionally tested.

Results.

Anti-TFAM antibodies were discovered in SLE while analyzing neutrophil autoantigens and confirmed by ELISA and immunoblotting. One-third of SLE patients (48/158) were positive for anti-TFAM antibodies. Unlike anti-dsDNA antibodies, anti-TFAM antibodies were not associated with disease activity or the IFN signature. Instead, anti-TFAM antibodies were associated with thrombosis, anti-phospholipid syndrome (APS) (OR 2.9 and 5.4, respectively), thrombosis-associated transcriptional profiles, and elevated IFN-III. Anti-TFAM antibodies were also found in PAPS, supporting their role in APS but not SLE pathogenesis. Lupus anticoagulant increased the risk of thrombosis associated with anti-TFAM antibodies (OR 8.71), indicating that they are markers of independent pro-thrombotic pathways.

Conclusions.

Anti-TFAM antibodies identify a distinct clinical and transcriptional disease subset associated with mitochondrial damage, thrombosis, and APS in SLE.

INTRODUCTION

Mitochondria are evolutionary endosymbionts derived from bacteria,¹ which contain potent immunological activators via damage-associated molecular patterns (DAMPs).² TFAM (transcription factor A, mitochondrial) is a member of the high-mobility group (HMG) family of DNA-binding proteins, which is essential for mitochondrial DNA (mtDNA) transcription and for packaging mtDNA into a nucleoprotein complex denoted the nucleoid.^3–5 In addition, TFAM is also a DAMP. During mitochondrial stress, TFAM re-localizes from mitochondria into cytosol, where it enhances cyclic GMP-AMP synthase (cGAS) activation by pre-structuring cytosolic DNA to nucleate or stabilize cGAS dimers.⁶ TFAM also promotes NLRP3 inflammasome activation in macrophages by facilitating oxidized (Ox) mtDNA formation or release.⁷

Extracellularly, TFAM released from necrotic cells binds RAGE (receptor for advanced glycation end-products) on plasmacytoid dendritic cells (pDCs),^8–10 functioning as a nucleoprotein carrier that facilitates the delivery of mtDNA into TLR9 endosomes, promoting the production of type I interferon (IFN-I) and tumor necrosis factor alpha (TNFα),^9,10 Recombinant TFAM has also been shown to drive cytokine secretion by cultured human monocytes and primary microglial cells,¹¹ and a TFAM polymorphism has been linked to the risk of venous thromboembolism (VTE) recurrence.¹² In systemic lupus erythematosus (SLE), an autoimmune disease characterized by high titer antibodies to nucleoprotein self-antigens,^13,14 nucleoids containing Ox mtDNA and TFAM are extruded from IFN-I activated neutrophils upon stimulation with ribonucleoprotein (RNP) immune complexes (ICs), serving as potent interferogenic stimuli for pDCs.¹⁵ Moreover, mtDNA, particularly Ox mtDNA in nucleoids, is thought to be an immunogenic trigger for anti-dsDNA antibodies in SLE.^15,16 Thus, as a carrier of Ox mtDNA, TFAM may play a dual role in SLE pathogenesis, inducing both IFN-I and anti-dsDNA antibodies.

Here, we identified that TFAM is also a common autoantigen in SLE. However, contrary to autoantibodies targeting dsDNA and nucleoprotein complexes,^17–20 antibodies to TFAM were neither associated with markers of disease activity or the IFN signature. Instead, these novel autoantibodies showed a striking association with thrombosis, anti-phospholipid syndrome (APS), thrombosis-associated transcriptional profiles and increased activity levels of IFN-III. Interestingly, anti-TFAM antibodies are also found in patients with primary APS – implying that these antibodies are related to APS but not to SLE pathogenesis – and the risk of thrombosis associated with anti-TFAM antibodies is similar to and independent of lupus anticoagulant (LAC), indicating that these antibodies are markers of thrombotic pathways distinct from those associated with LAC.

METHODS

Sex as a biological variable.

Male and female subjects were included in this study, although females predominate (9:1 female:male ratio) due to the disease demographics.

Study design and patients.

The study included an exploratory sample of 9 healthy controls and 22 SLE patients, and a validation sample of 98 healthy controls [age, median (IQR) 37 (30,45); female sex 71/98, 72%] and 158 SLE patients from the “Study of biological Pathways, Disease Activity and Response markers in patients with Systemic Lupus Erythematosus” (SPARE) cohort, which has been previously described.^21,22 Briefly, adult patients (age 18 to 75 years-old) who met the definition of SLE per the revised American College of Rheumatology classification criteria were enrolled into the study.²³ Patients were seen as outpatients at the Johns Hopkins Lupus Center and treated according to standard clinical practice. Disease activity was assessed using the Safety of Estrogens in Lupus Erythematosus: National Assessment (SELENA) version of the Systemic Lupus Erythematosus Disease Activity Index (SLEDAI)²⁴ and physician global assessment (PGA).²⁵ C3, C4, anti-dsDNA (Crithidia), complete blood cell count and urinalysis were performed at every visit. Study participants also underwent whole blood gene expression analysis using the Affymetrix GeneChip HT HG-U133+,^21,22 as well as quantification of circulating IFN-I, IFN-II and IFN-III activity levels.²⁶ Antibodies to TFAM were also determined in 36 patients with rheumatoid arthritis (RA), and 40 patients with dermatomyositis (DM), and 50 patients with primary antiphospholipid syndrome (PAPS).

Study approval.

All samples were obtained under approval by the Institutional Review Board at the Johns Hopkins University School of Medicine and Sapienza University of Rome.

TFAM cloning, protein expression and structure.

cDNA encoding human mature TFAM (amino acids 43–246) was synthesized using RNA from human PBMCs and cloned into pET-28a(+). Recombinant N-terminal his-tagged TFAM was expressed in E. coli BL21 (DE3) and purified by Ni-NTA affinity chromatography. TFAM structure was predicted using AlphaFold Protein Structure Database https://alphafold.com/.

Detection of antibodies to TFAM.

Anti-TFAM antibodies were measured in serum/plasma by homemade ELISA. Briefly, Nunc Maxisorp plates were coated with 200 ng/well of recombinant TFAM. The plates were blocked for one hour with phosphate buffered saline plus 0.1% Tween-20 (PBST) with 3% non-fat milk. Serum/plasma was diluted 1:1000 in PBST plus 1% non-fat milk and assayed in duplicate using antigen-conjugated plates and plates without antigen for background subtraction. For competition assays, diluted SLE serum/plasma was pre-incubated with 10 μg/mL of human recombinant HMGB1 (Sino Biological) or β2-glycoprotein I (β2GPI) purified from human plasma (Arotec Diagnostics) for 1 hr at room temperature. HRP conjugated goat anti-human IgG was used as a secondary antibody (Diluted at 1:10,000 in PBST 1% nonfat milk). TFAM antibody arbitrary units (AU) were calculated using a standard curve made of a serial diluted serum from a high-titer SLE patient. Anti-TFAM positivity cutoff was set as the mean + 2SD of healthy control anti-TFAM antibodies (AU).

Indirect immunofluorescence.

Peripheral blood human neutrophils were purified as previously described,²⁷ attached to microscope slides using HistoGrip^™ (cat. 008050 ThermoFisher Scientific), fixed with 4% paraformaldehyde and permeabilized with acetone. After blocking with 2% bovine serum albumin (BSA) and incubation with SLE serum diluted at 1:100 in PBS or commercial anti-TFAM antibody (clone C9, cat. sc-376672, Santa Cruz Biotechnologies) at 1:10, human IgG and commercial anti-TFAM antibodies were visualized using goat anti-human IgG Alexa Fluor 488 or rabbit anti-mouse Alexa Fluor 594, respectively. DNA was detected using DAPI. In addition, TFAM detection using the commercial anti-TFAM antibody (clone C9) and anti-TFAM positive SLE sera was performed using commercially available HEp-2 cell slides (NOVA Lite ^® HEp-2 ANA Kit with DAPI, cat. 708102, Werfen) according to manufacturer instructions. For HEp-2 testing, serum was diluted at 1:80 and the commercial antibody at 1:10. Human IgG and commercial anti-TFAM antibodies were visualized using goat anti-human IgG Alexa Fluor 488 or rabbit anti-mouse Alexa Fluor 594, respectively. For competition assays, diluted SLE serum was pre-incubated with 40 μg/mL of human recombinant mature TFAM for 1 hr at room temperature.

Gene expression analyses.

Gene expression analysis from the SPARE cohort was previously described.²² CEL files were subjected to RMA background correction, and quantile normalization, using the Oligo package.²⁸ To select only expressed genes in whole blood, we filtered out transcripts that had a raw signal < 100 in less than 10% of samples with the genefilter R package. All calculations and analyses were performed using R (ver 4.0.2) and Bioconductor (ver 3.13).²⁹

Feature selection using the fisher score algorithm and PCA analysis.

To select the most relevant transcripts to distinguish patients positive vs. negative for anti-TFAM antibodies, thrombotic events, or anti-dsDNA antibodies, we employed an algorithm known as the Fisher score.³⁰ Briefly, a Fisher score is calculated for each individual transcript by considering the number of samples, mean and standard deviation on each category (e.g., anti-TFAM positive or anti-TFAM negative) and each transcript is then ranked according to its score. For the principal component analysis (PCA), we selected the top 1000 genes since this number of transcripts is useful to identify biologically relevant pathways using the fisher score algorithm.³⁰ Fisher scores were calculated using the Rdimtools R package.³¹ PCA was performed using base R functions.

Gene set enrichment analyses.

To perform gene set enrichment analyses, we determined the transcripts with the highest loading on the anti-TFAM antibodies, thrombosis and anti-dsDNA antibodies PC1 by calculating the Pearson’s R correlation coefficient. P values were adjusted with the Benjamini-Hochberg method. Transcripts with an adjusted p value ≤0.01 were considered as significantly correlated. In order to compare the anti-TFAM, thrombosis and anti-dsDNA related transcriptomes, gene set enrichment analysis was performed using the metanalysis function of Metascape.³²

Statistical analyses.

Comparisons of continuous variables between groups were done using Student’s T test and ANOVA test with Tukey’s posthoc test as indicated. The Mann-Whitney’s U test and Kruskal-Wallis’s test was used for group-wise comparisons of non-normally distributed variables Fisher’s exact test and χ² tests were used for univariate analysis on SPARE cohort variables, as appropriate. Exact2×2 package in R version 3.5.1 was used for binary variables to obtain p-value, OR, and 95% CI. Correlations between variables were calculated using Pearson’s R. Comparisons of paired samples were done using Wilcoxon signed-rank test. Multivariate analyses were carried out using multivariate logistic regression. Statistical significance was set at p < 0.05. The statistical analyses were carried out with the R software version 4.3.3.

RESULTS

TFAM is an autoantigen in SLE

Autoantibodies in SLE bind to different antigens in human neutrophils, generating distinct cellular patterns when analyzed by indirect immunofluorescence, including nucleolar, nuclear and cytoplasmic patterns (Supplemental Figure 1). Coincidentally, while working on parallel studies with neutrophils and TFAM, we noticed that commercial antibodies to TFAM generate a cellular pattern similar to SLE sera targeting cytoplasmic antigens (Figure 1A), leading to the hypothesis that TFAM is an autoantigen in SLE. TFAM is a nuclear-encoded protein of 204 amino acids (24 kDa).⁴ It is translated as a pre-protein containing a mitochondrial targeting sequence (MTS, amino acids 1–42) that is cleaved upon mitochondrial import (Figure 1B). Mature TFAM (amino acids 43–246) consists of two HMG boxes separated by a linker and followed by a short tail at the C-terminus (Figure 1B).⁴

Figure 1. TFAM is a novel autoantigen in SLE.

(A) Healthy control neutrophils were stained with a SLE serum detecting cytoplasmic antigens (green), anti-TFAM monoclonal antibody C9 (red) and DAPI (blue), and viewed by immunofluorescence microscopy. Individual and merged images are shown. (B) TFAM structure predicted by AlphaFold. Amino acids (aa) 1–42 form the mitochondrial targeting sequence (MTS), while aa 43–246 comprise the mature form of TFAM, including two high mobility group domains (HMG1 and HMG2), a linker and a C-terminal tail (C-tail). (C) Serum levels of anti-TFAM antibodies in SLE (n=22) and healthy controls (n=10). The p value was obtained using the Wilcoxon Rank Sum test. (D) Representative immunoblots showing the reactivity of healthy control serum (n=4) and anti-TFAM positive SLE serum against human recombinant TFAM (n=4). (E) Serum levels of anti-TFAM antibodies before and after blocking with human recombinant high mobility group box 1 (HMGB1) protein (N = 9). Comparisons were done using a paired samples Wilcoxon signed-rank test. (F) HEp2 cells were stained using anti-TFAM positive SLE serum (green) and anti-TFAM monoclonal antibody C9 (red), and viewed by immunofluorescence microscopy. Individual and merged images are shown. (G) HEp2 cells were stained with anti-TFAM positive SLE serum (green) in the absence (left) and presence (right) of blocking with recombinant TFAM (rTFAM). Representative images of 12 (F) and 8 (G) anti-TFAM positive SLE serum are shown.

To address whether TFAM is a target of autoantibodies in SLE, we developed an ELISA using human recombinant mature TFAM as substrate and screened sera from an exploratory cohort sample of patients with SLE (n=22) and healthy controls (n=9). Patients with SLE had significantly elevated serum levels of anti-TFAM antibodies compared with healthy controls (P < 0.0004) (Figure 1C). Notably, SLE sera associated with a cytoplasmic pattern in neutrophils (either alone or concomitant with a nuclear pattern) (Figure 1A and Supplemental Figure 1) were those with the highest levels of anti-TFAM antibodies detected by ELISA. We further confirmed the presence of antibodies to TFAM in SLE by immunoblotting (Figure 1D), and that these antibodies were not cross-reactive with HMGB1 (Figure 1E), a protein analogous to TFAM that is also an autoantigen in SLE.^4,33 In HEp-2 cells, anti-TFAM positive SLE serum colocalize with TFAM staining, which shows a cytoplasmic reticular pattern consistent with the AC-21 ANA pattern (Figure 1F). In contrast to neutrophils (Figure 1A), however, TFAM also shows some nuclear localization in HEp-2 cells, which may be explained by the fixation/permeabilization method or the expression of a TFAM variant lacking the MTS.³⁴ The antibody pattern that colocalize with TFAM was completely blocked by preincubating SLE serum with recombinant mature TFAM (Figure 1G).

Antibodies to TFAM are associated with APS and thrombotic events in SLE

To define clinical features and transcriptional pathways associated with anti-TFAM antibodies in SLE, we studied samples from the SPARE cohort for which extensive clinical and serologic variables are available, as well as whole blood gene expression data.^21,22 Demographic, clinical and laboratory features of the SLE cohort are summarized in Supplementary Table 1. The presence of antibodies to TFAM were also addressed in serum from patients with PAPS, RA and DM. Among the different disease groups, only patients with SLE had significantly higher levels of anti-TFAM antibodies compared to healthy controls [mean (SD), 126.1 (236.4) vs. 32.2 (50.5), p<.0001] (Figure 2A). After determining the cut-off level of two standard deviations above the mean anti-TFAM antibody level in healthy sera, 30.3% (48/158) of SLE patients tested positive for antibodies to TFAM compared to 4% (4/98) in healthy controls (P < 0.0001) (Figure 2B–C). In addition to SLE, PAPS was linked with an increased frequency of anti-TFAM antibodies, 16% (8/50) vs. healthy controls (p<0.05). Anti-TFAM antibodies were not significantly increased in RA or DM compared to healthy controls (Figure 2A–C). In SLE, anti-TFAM antibodies were stable over time. Although anti-TFAM levels varied between visits (intra class-correlation of 0.622, p=0.002), none of the samples became negative for the autoantibody, and the mean anti-TFAM levels between was not statistically different between visits (mean ± SD, 500 ± 282 AU/mL vs. 432 ± 392 AU/mL, p = 0.555) (Supplemental Figure 2A).

Figure 2. Anti-TFAM antibodies are prevalent in SLE.

(A) Serum levels of anti-TFAM antibodies in SLE patients (n=158) from the SPARE lupus cohort compared to healthy controls (HC, n=98), primary anti-phospholipid syndrome (PAPS, n=50), rheumatoid arthritis (RA, n=36) and dermatomyositis (DM, n=40). Comparisons between groups were done using one-way ANOVA and Tukey’s posthoc test. (B) Fequency of Anti-TFAM antibodies in each group from A. (C) P values indicating significant differences between the frequency in B of anti-TFAM in each group from A. P values were obtained with Fisher’s exact test. (D) Association between anti-dsDNA positivity and anti-TFAM antibodies. The association was estimated using Fisher’s exact test. (E) SLEDAI score at the time of visit in SLE patients according to anti-TFAM positivity. The p value was obtained with Student’s T test.

Since mtDNA has been implicated in the production of anti-dsDNA antibodies in SLE,^15,16 and TFAM is in complex with mtDNA,^3–5 we anticipated that antibodies to TFAM would be mechanistically linked to anti-dsDNA antibodies and disease activity in SLE. However, only 36.2% (17/47) of anti-TFAM positive patients were also positive for anti-dsDNA antibodies (Figure 2D), and there was a mild correlation between anti-TFAM and anti-dsDNA levels (r = 0.311, p < 0.005). Moreover, anti-TFAM antibodies only showed a mild association with SLEDAI score (Figure 2E, and Supplemental Table 1). Instead, anti-TFAM antibodies were significantly associated with renal insufficiency (but not active nephritis) and features of APS (Figure 3A, and Supplemental Table 1 and 2), in contrast to anti-dsDNA antibodies, which showed prominent association with features linked to IC deposition, such as nephritis and low complement (Figure 3B). Notably, SLE patients positive for anti-TFAM antibodies were more likely to have a history of leg ulcers, cardiac murmur, any thrombotic event, and be classified with APS [OR (95% CI), 22.1 (1.2, 491.9), 2.7 (1.3, 6.0), 2.9 (1.4, 6.2), and 5.4 (2.5, 11.9), respectively] (Figure 3A, and Supplemental Table 2). Anti-TFAM antibodies were also associated with increased frequency of antibodies to β2GPI, cardiolipin and dsDNA [OR (95% CI), 2.6 (1.3, 5.6), 2.2 (1.0, 5.0), and 4.4 (1.8, 10.6), respectively]. Given the strong association between APS and anti-TFAM antibodies, we further discarded that anti-TFAM antibodies in SLE were cross-reactive with β2GPI (Supplemental Figure 2B). Since patients with PAPS are defined by the presence of APS-associated antibodies, it not surprising that anti-TFAM antibodies were not associated with any particular antibody in patients with PAPS (Supplemental Table 3).

Figure 3. Anti-TFAM antibodies are associated with APS and thrombotic events in SLE.

Clinical and immunological characteristics of SLE patients according to anti-TFAM (A) and anti-dsDNA antibody positivity (B). The graphs represent data for 157 SLE patients, including 48 positive and 109 negatives for anti-TFAM antibodies. Each dot represents a clinical manifestation. Only clinical manifestations significantly associated (p < 0.05) with anti-TFAM and anti-dsDNA antibodies were labeled. Quantities under each label correspond to the odds ratio (OR) and 95% confidence interval (CI) of anti-TFAM (+) vs. anti-TFAM (−) (A) and anti-dsDNA (+) vs. anti-dsDNA (−) (B) calculated using a 2 × 2 table. P values were obtained by Fisher’s exact test. Renal insuff., renal insufficiency; LAC, lupus anticoagulant; Sm, anti-Sm antibodies; Ro52, anti-Ro52 antibodies; ACL, anti-cardiolipin antibodies; dsDNA, anti-dsDNA antibodies; B2GPI, anti-β2-glycoprotein-I antibodies; FPRPR, false positive rapid plasma reagin test; CH50, 50% Hemolytic complement test; ESR, erythrocyte sedimentation rate; DVT, deep vein thrombosis; VT, venous thrombosis; AT, any thrombotic event; APS, anti-phospholipid syndrome; MS, musculoskeletal; CV, cardiovascular; P, pulmonary; GI, gastrointestinal.

Anti-TFAM antibodies are predictive of thrombotic events in SLE

To analyze the predictive value of anti-TFAM antibodies and possible covariates, we classified SLE patients in two groups: 1) subjects with any thrombotic event (n = 54, 34%), and 2) subjects without thrombosis (n = 102, 65.4%) (Supplemental Table 4). Univariate analysis revealed that patients who had thrombotic events had longer disease duration (23.7 vs. 18.5 years, OR = 1.04 per year of disease duration, p = 0.01) (Figure 4A). Notably, only anti-TFAM antibodies and LAC (determined by increased dilute Rusell viper venom test, dRVVT) were significantly associated with any thrombotic event [46% (25/54) vs. 23% (23/102), OR = 3.74, p = 0.003; and 50% (27/54) vs. 27% (28/102), OR =3.46, p = 0.007] (Figure 4A and Supplemental Table 4). Similar to other studies in SLE³⁵, however, anti-β2GPI or anti-cardiolipin (ACL) antibodies were not associated with thrombotic events in the SPARE cohort (Supplemental Table 4).

Figure 4. Anti-TFAM antibodies are predictive of thrombotic events in SLE.

(A) Univariate logistic regression showing the predictive value of anti-TFAM antibodies and significantly associated factors from Supplemental Table 3 for any thrombotic events in patients with SLE. (B-C) Multivariate logistic regression models analyzing the predictive value of anti-TFAM antibodies after adjustment for disease duration, lupus anticoagulant (LAC), smoking, alcohol use, disability, and photosensitivity. (D-E) Logistic regression models showing the additive value of a history of smoking or LAC with anti-TFAM antibodies in SLE. LAC, lupus anticoagulant. The graphs represent data for 157 SLE patients, including 48 positive and 109 negatives for anti-TFAM antibodies.

Among traditional cardiovascular risk factors, prior smoking was significantly associated with thrombotic events [59% (32/54) vs. 26.5% (27/102), OR = 4.1, p < 0.001] (Figure 4A and Supplemental Table 4), and photosensitivity was more frequent in patients with a history of thrombosis amongst clinical manifestations not known to be related with thrombosis [72.2% (39/54) vs. 48% (49/102), OR=2.71, p = 0.004] (Figure 4A and Supplemental Table 4). Patients with thrombosis were also more likely to have some degree of disability (OR = 3.66) (Figure 4A). After adjustment for LAC, photosensitivity, and smoking, SLE patients positive for anti-TFAM antibodies were 2.82 to 3.34 times more likely to have a thrombotic event during their clinical course than anti-TFAM negative patients (Figure 4B and C). Interestingly, whereas the likelihood of having a thrombotic event was similar among patients positive for anti-TFAM antibodies or a smoking history (OR = 4.92 and 5.14, respectively), we found an additive effect when both conditions were present (OR = 13.78) (Figure 4D). Similarly, the presence of LAC had an additive effect on the risk of thrombosis in SLE patients positive for anti-TFAM antibodies (OR = 8.71) (Figure 4E).

Antibodies to TFAM are associated to transcriptional fingerprints linked to thrombosis in SLE

Since the extracellular release of nucleoids from activated neutrophils has been associated with the induction of IFN-I in SLE,¹⁵ we addressed whether anti-TFAM antibodies are linked to unique transcriptional profiles – particularly the IFN signature – using gene expression data from blood collected in parallel with the samples used to measure anti-TFAM antibodies. To uncover gene expression signatures associated to anti-TFAM antibodies, we used principal components analysis (PCA) on the top 500 transcripts ranked by their Fisher score between SLE patients positive and negative for anti-TFAM antibodies (i.e., the anti-TFAM transcriptome) (Supplemental File 1). The first principal component (PC1) stratified SLE patients on a gradient rather than specific clusters (Figure 5A). PC1 was significantly higher in SLE patients who were positive for anti-TFAM antibodies and had a history of thrombotic events (Figure 5B).

Figure 5. Anti-TFAM antibodies are linked to a thrombotic transcriptome in SLE.

(A) Principal component (PC) analysis of the top 500 transcripts ranked by their Fisher score between in anti-TFAM positive vs. anti-TFAM negative SLE. (B) PC1 levels in patients with SLE according to the presence of both anti-TFAM antibodies and a history of any thrombotic event (AT). Comparisons were done using ANOVA and Tukey’s test as post-hoc. (C-D) Correlation between Anti-TFAM PC1 and anti-dsDNA PC1 (C) and anti-TFAM PC1 and thrombosis PC1 (D). Correlations were estimated with Pearson’s r correlation coefficient. (E) Enrichment analysis of significantly correlated transcripts with anti-TFAM, thrombosis, and anti-dsDNA PC1. Significantly, correlated genes were determined by Pearson’s r correlation coefficients. Only, genes with an adjusted p value < 0.01 (Benjamini-Hochberg’s method) were considered in the analyses. The enrichment analysis was conducted using the Metanalysis function from Metascape.org. For this data set, we analyzed a total of 156 microarrays, of which 47 and 109 correspond to anti-TFAM positive and anti-TFAM negative SLE. (F-H) Activity levels of IFN-I, IFN-II and IFN-III in patients with SLE according to anti-TFAM positivity. Anti-TFAM positive, n = 47; anti-TFAM negative, n = 109. Comparison of IFN levels was done with Student’s T test.

To compare the transcriptome of anti-TFAM antibodies to those associated with thrombosis and activation by anti-dsDNA antibodies, we performed PCA analysis using the top 500 transcripts ranked by their Fisher score between patients with and without history of thrombosis (i.e., thrombosis transcriptome) and between patients positive and negative for anti-dsDNA antibodies (i.e., anti-dsDNA transcriptome) (Supplemental File 1), respectively. Interestingly, while anti-TFAM PC1 showed only mild correlation with anti-dsDNA antibodies PC1 (r = 0.230, p = 0.004), it had a strong correlation with thrombosis PC1 (r = 0.869, p = 0.001) (Figure 5C and D, respectively). There was no correlation between antibodies to dsDNA and thrombosis PC1 (r = 0.01, p = 0.914). Together, these data imply that PC1 represents transcripts and pathways associated with anti-TFAM-related thrombosis rather than anti-dsDNA mediated immune activation.

To further determine the biological significance of anti-TFAM, anti-dsDNA and thrombosis PC1, we performed enrichment analysis on the transcripts that were significantly correlated with each component. We found 1799 and 246 transcripts that were positively and negatively correlated with anti-TFAM PC1 (Supplemental File 1). Positively enriched transcripts were linked to pathways associated with immune-mediated activation and hemostasis, such as response to virus and cytokine stimuli, neutrophil degranulation, activation of the innate immune response, vascular endothelial growth factor A (VEGFA)-VEGFR2 signaling, responses to bacteria, cellular response to stress, hemostasis, and autophagy, among others. In contrast, transcripts that were negatively correlated were mainly associated with pathways linked to RNA and DNA metabolism, including protein translation, RNA localization and splicing, chromatin organization, RNP complex biogenesis, cell cycle, and mitochondrial gene expression, among others (Figure 5E).

As indicated by the strong correlation between anti-TFAM antibodies and thrombosis PC1 (Figure 5C), the enrichment of both components was essentially identical, with the exception of two pathways related to response to virus and cytokine stimuli (Figure 5E). Notably, pathways involving hemostasis and VEGFA-VEGFR2 signaling, as well as RNA and DNA metabolism (except for RNA translation and protein modification) were exclusively linked to anti-TFAM antibodies and thrombosis, whereas type I IFN was only associated with anti-dsDNA antibodies (Figure 5E). Other immune-mediated activation pathways showed a significant similarity between anti-TFAM antibodies and thrombosis with anti-dsDNA antibodies (Figure 5E), which is most likely explained by the 30% overlap between anti-dsDNA and anti-TFAM antibodies. Collectively, the data support that anti-TFAM antibodies are mechanistically related to thrombosis in SLE, but are independent to hallmark features, such as disease activity, anti-dsDNA antibodies and the IFN signature. Consistent with these findings, circulating activity levels of IFN-I or IFN-II were not associated with anti-TFAM antibodies (Figure 5F and G, respectively). Unexpectedly, however, IFN-III was significantly elevated in patients positive for antibodies to TFAM (p = 0.0001) (Figure 5H).

DISCUSSION

Mitochondria appear to play at least two immunogenic and potentially independent roles in SLE pathogenesis. First, nucleoids – particularly containing Ox mtDNA – have been linked to the induction of IFN-I by pDCs and mtDNA is considered an immunogenic candidate for the production of anti-dsDNA antibodies in SLE.^15,16 In this context, mitochondria may have a mechanistic role in clinical phenotypes associated with the IFN signature and disease activity features related to anti-dsDNA antibodies, such as complement activation and lupus nephritis. Second, antibodies to mitochondrial proteins (e.g., 60-kd heat-shock protein, HSP60) and cardiolipin-protein complexes have been associated with thrombosis,^36,37 supporting a role for mitochondrial non-DNA-associated components in thrombotic events. Because TFAM is the main nucleoprotein involved in the packing of mtDNA into nucleoids,^3–5 and Ox nucleoids have been linked to the induction of IFN-I by pDCs in SLE,¹⁵ we were particularly interested in anti-TFAM antibodies and their relationship to antibodies to DNA and the IFN signature. Collectively, our clinical and transcriptional data show that anti-TFAM antibodies are not mechanistically linked to antibodies to DNA or IFN-I induced activation, but rather constitute a novel biomarker associated with APS and thrombosis in SLE. Indeed, the finding that antibodies to TFAM are also present in patients with primary APS supports the notion that these antibodies are generated in response to mechanisms of mitochondrial damage other than those potentially associated with anti-dsDNA and IFN-I induction.

While antibodies to TFAM and dsDNA overlap in about a third of patients with SLE, they identify two distinct clinical and transcriptional subsets within SLE. Anti-dsDNA antibodies are associated with mechanisms related to IC formation, including complement activation, nephritis, toll-like receptor (TLR) signaling and IFN-I production. In contrast, anti-TFAM antibodies are linked to thrombotic events, APS and thrombosis-associated transcriptional fingerprints, but not to pathways activated by IFN-I. Indeed, unlike SLE-associated autoantibodies such as anti-dsDNA, anti-Sm, anti-RNP, anti-DNase1L3 and anti-Ro52Ex4, which are strongly associated with high levels of circulating IFN-I,^26,38 antibodies to TFAM showed no relationship with IFN-I levels. These data are consistent with the lack of correlation between APS-associated autoantibodies and IFN-I levels or the IFN-I signature.^18,26,38 Thus, although TFAM is the major carrier of mtDNA in nucleoids, our findings indicate that the immune response to TFAM in SLE is independent of any potential role of mtDNA in anti-dsDNA antibody production, and anti-TFAM antibodies appear to play no role in the induction of IFN-I by cell-free nucleoids. Nevertheless, it is worth noting that antibodies to TFAM were associated with increased activity levels of IFN-III. Although it is unclear whether IFN-III promotes the production of anti-TFAM antibodies or these antibodies induce IFN-III expression, and whether IFN-III is relevant for thrombosis, an association between IFN-λ1 and thrombotic events has been suggested in SLE, but it was not statistically significant.³⁸ The mechanistic relationship between anti-TFAM antibodies, IFN-III, and APS has yet to be determined.

Notably, the risk of thrombosis associated with anti-TFAM antibodies is similar to and independent of LAC, the stronger risk factor for thrombosis in SLE.^35,39–41 Moreover, a history of smoking or LAC has an additive effect on the risk of thrombosis in SLE patients positive for anti-TFAM antibodies, implying that these antibodies are markers of thrombotic pathways distinct from those associated with smoking and LAC. Indeed, while the prothrombotic effect of LAC is mediated by targeting the coagulation pathway,^37,42 the presence of antibodies to TFAM indicates thrombotic mechanisms involving cellular and mitochondrial damage.

Although the study of neutrophils provided initial clues for the search of antibodies to TFAM, the cellular source of TFAM that drives the production of autoantibodies may have different origins in SLE, including neutrophils, erythrocytes and platelets. For instance, TFAM released from IFN-I activated neutrophils in response to RNP IC stimulation,¹⁵ as well as by forms of cell death involving mitochondrial membrane permeabilization, such as pyroptosis, apoptosis and necrosis,^43,44 are potential sources of immunogenic TFAM in SLE. Interestingly, although mtDNA is found in neutrophil extracellular traps (NETs),⁴⁵ TFAM has not been detected in NETs by mass spectrometry analysis.^46,47 The abnormal accumulation of erythrocytes carrying mitochondria in SLE and their clearance by macrophages may also trigger the production of antibodies to TFAM⁴⁸. Lastly, extracellular mitochondria released from IC-activated platelets have been proposed as a key source of mitochondrial antigens in SLE.⁴⁹

Interestingly, mitochondria are essential for platelet function and therefore, mitochondrial damage or dysfunction can affect platelet survival and increase the risk of thrombosis.⁵⁰ In this case, the production of anti-TFAM antibodies may represent markers of mitochondrial dysfunction associated with thrombosis rather than direct drivers of thrombotic events in SLE. Alternatively, these antibodies may promote thrombosis by binding TFAM-DNA complexes released from activated neutrophils, degranulating platelets, apoptotic cells undergoing secondary necrosis, or from other forms of cell death. In either scenario, as markers of prothrombotic mitochondrial dysfunction or direct triggers of thrombosis, the discovery of antibodies to TFAM provides a novel tool independent of traditional APS-associated antibodies for identifying SLE patients at risk of thrombosis, where mitochondrial damage is likely to play a pathogenic role.

Limitations of the study

Because the study only examined cross-sectional samples, it is unclear whether anti-TFAM seroconversion occurs synchronously or metachronously with vascular events in SLE, as well as their stability over time. Similarly, it is important to further define the significance of anti-TFAM antibodies in other groups of patients with thrombotic events who do not have SLE or PAPS. Furthermore, it would be interesting to determine if anti-TFAM antibodies are prevalent among patients with anti-phosphatidylserine/prothrombin complex antibodies, which similar to anti-TFAM antibodies, are associated with increased risk of thrombosis in subjects negative for ACL or anti-β2GPI.⁵¹

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Mitochondrial damage has been implicated in the pathogenesis of systemic lupus erythematosus (SLE), particularly in relation to the production of anti-dsDNA antibodies and the induction of type I interferon (IFN-I).

• Thrombotic events and anti-phospholipid syndrome (APS) in SLE are immune-mediated conditions associated with the production of autoantibodies, particularly lupus anticoagulant.

WHAT THIS STUDY ADDS

• This is the first study to identify the mitochondrial nucleoprotein TFAM (transcription factor A, mitochondrial) as an autoantigen in SLE.

• Anti-TFAM antibodies are significantly associated with thrombotic events and APS in SLE patients, independent of traditional APS-associated antibodies. Notably, antibodies to TFAM increase the risk of thrombosis associated with lupus anticoagulant, indicating that they are markers of independent prothrombotic pathways.

• Gene expression analysis reveals unique transcriptional signatures associated with anti-TFAM antibodies, implying a mechanistic link with immune activation pathways associated with thrombosis in SLE.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Anti-TFAM antibodies are potential biomarkers for identifying SLE patients at risk of thrombosis, independent of traditional APS-associated antibodies. These novel autoantibodies may aid in disease evaluation and treatment strategies for SLE, as well as provide novel insights into disease mechanisms, particularly those related to thrombotic events and APS.

DATA AVAILABILITY

Microarray data are available from the Gene Expression Omnibus with accession numbers GSE45291 and GSE121239. Any additional data presented in this work can be obtained from the authors upon request.