WHAT IS ALREADY KNOWN ON THIS TOPIC

Autoantibodies targeting intracellular proteins are common in various autoimmune diseases.

In myositis, the pathological significance of these autoantibodies has been questioned due to the assumption that autoantibodies cannot enter living muscle cells.

WHAT THIS STUDY ADDS

In patients with myositis autoantibodies, antibodies accumulate inside myofibres in the same subcellular compartment as the autoantigen.

Muscle biopsies from patients with autoantibodies targeting transcriptional regulators exhibit transcriptomic patterns consistent with dysfunction of the autoantigen.

Introducing patient antibodies into cultured muscle cells recapitulates the transcriptomic effects observed in human disease.

Further supporting evidence suggests that myositis autoantibodies recognising other autoantigens may also disrupt the function of their targets.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

In myositis, autoantibodies are internalised into living cells, causing biological effects consistent with the disrupted function of their autoantigen.

Treatments to reset, decrease the half-life or reduce the production of endogenous antibodies can be effective in various forms of myositis.

Introduction

Myositis is a family of autoimmune diseases variably affecting multiple organs, including muscle, skin, lungs and/or joints. These diseases have been traditionally grouped into clinical syndromes such as dermatomyositis, the antisynthetase syndrome, immune-mediated necrotising myopathy, overlap myositis and inclusion body myositis.

Most myositis patients have a known autoantibody targeting an intracellular autoantigen. For example, dermatomyositis patients may have autoantibodies recognising either nuclear (ie, Mi2, TIF1g and NXP2) or cytoplasmic (ie, MDA5 or SAE) proteins; antisynthetase syndrome patients have autoantibodies recognising one of the cytoplasmic aminoacyl-tRNA synthetases; immune-mediated necrotising myopathy patients harbour autoantibodies against either the signal recognition particle (SRP) or HMG-CoA reductase, both of which are found in the cytoplasmic compartment; and patients with overlap myositis may have autoantibodies (ie, anti-PM/Scl) recognising nuclear RNA exosome components 9 and 10, which are localised to the nucleolus.

While each myositis autoantibody is associated with a unique clinical phenotype,1 it remains unknown whether these autoantibodies play a functional role in myositis pathogenesis. Indeed, because the myositis autoantigens are located inside the cell, it has been assumed that myositis autoantibodies cannot reach their target to have a functional effect.2–4

We recently showed that in muscle tissue from patients with anti-Mi2 autoantibodies, antibodies are deposited in the nuclei of muscle fibres, which is also where the Mi2 autoantigen is located. Furthermore, we found that muscle biopsies from these patients have a unique transcriptomic profile, characterised by the overexpression of genes normally repressed by the Mi2-containing nucleosome remodelling and deacetylation (Mi2/NuRD) complex.5 These findings suggested that anti-Mi2 autoantibodies reach the Mi2/NuRD complex within muscle cell nuclei and disrupt its function, leading to aberrant expression of genes normally repressed by that complex. However, direct evidence that antibodies were causing the dysfunction of the Mi2/NuRD complex in anti-Mi2-positive patients was lacking.

In the current study, we have analysed muscle tissue from myositis patients with a wide variety of myositis autoantibodies to determine whether antibodies accumulate in the same subcellular compartment as the relevant autoantigen. Moreover, we sought to identify transcriptomic profiles in muscle tissues obtained from patients with each type of autoantibody-defined myositis. Finally, we studied the transcriptomic profiles of cultured muscle cells following the internalisation of antibodies obtained from myositis patients with different myositis autoantibodies. Taken together, these studies show that myositis autoantibodies reach their intracellular targets. Furthermore, we provide strong evidence that autoantibodies against Mi2 and exosome components 9 and 10 (anti-PM/Scl autoantibodies) disrupt the normal function of these proteins.

MethodsPatients

Muscle tissue from patients who underwent diagnostic muscle biopsies and healthy volunteers at several centres specialised in neuromuscular diseases underwent bulk RNAseq analysis (online supplemental methods and online supplemental table 1).

A subset of these muscle biopsies was used for immunofluorescence studies. For antibody internalisation experiments, immunoglobulin was purified from the serum of patients with myositis and healthy controls.

RNA sequencing

Bulk RNAseq was performed on frozen muscle biopsy specimens as previously described.5–10 Briefly, muscle biopsies underwent immediate flash freezing and were stored at −80°C across all contributing centres. Samples were then transported in dry ice to the NIH and processed uniformly to prepare the library and conduct the analysis. RNA was extracted with TRIzol. Libraries were either prepared with the NeoPrep system according to the TruSeq Stranded mRNA Library Prep protocol (Illumina, San Diego, California, USA) or with the NEBNext Poly(A) mRNA Magnetic Isolation Module and Ultra II Directional RNA Library Prep Kit for Illumina (New England BioLabs, ref. #E7490 and #E7760).

Histopathology and immunofluorescence

Muscle biopsy sections processed for clinical purposes were stained for H&E, oil-red O, Gömöri trichrome, CD56 (NCAM), membrane attack complex, NADH and COX, and then microscopic images were digitised using a Leica Slide Scanner SCN400F.

For immunofluorescence, 10 μm unfixed sections (one normal biopsy and two cases per autoantibody group) were immunoreacted overnight at 4°C in a humidified chamber using the primary antibodies included in online supplemental table 2. The sections were then washed in phosphate-buffered saline and immunoreacted using the appropriate fluorochrome-conjugated secondary antibodies included in online supplemental table 3. Of note, a fluorochrome-conjugated secondary antibody was used to stain human IgG in the muscle biopsies. The sections were then washed in washing buffer, incubated for 15 min at room temperature in Hoechst 33 258 (Abcam, ab228550, 1:4000) to stain the cell nuclei, rinsed in phosphate-buffered saline and cover-slipped using Prolong Diamond Antifade mountant. Additional sections were stained exclusively with Goat-anti-Rabbit IgG and Goat-anti-Rat IgG secondary antibodies, without the addition of any primary antibody, resulting in no detectable signal. All sections were imaged using a high-resolution immunofluorescence confocal microscope Leica SP8.

Culture of differentiating human skeletal muscle myoblasts and treatment with different types of interferon

Normal human skeletal muscle myoblasts were cultured according to the protocol recommended by the supplier (Lonza). When 80% confluent, myoblasts were induced to differentiate into myotubes by replacing the growth medium with a differentiation medium (DMEM-F12 (Lonza, ref. 12-719F) supplemented with 2% horse serum (Gibco, ref. 16050-122), insulin-transferrin-selenium (Gibco, ref. 41400-045) and penicillin-streptomycin-L-glutamine (Gibco, ref. 10378-016). Cells were harvested before differentiation and then daily after differentiation for 6 days.

To examine the effect of different types of interferon on gene expression, we treated the cells daily with 100 U/μL and 1000 U/μL of IFNA2a (R&D, ref. 11100-1), and IFNB1 (PeproTech, ref. 300-02BC), respectively, for 7 days. Then, the cells were harvested for RNA extraction and RNA sequencing.

Electroporation of antibodies into human muscle cells

Human immunoglobulin G was purified and concentrated from serum using protein G Agarose (Millipore, ref. 16-266) and the Amicon Pro Purification System (Millipore, ref. ACS500024) with a 30 kDa molecular weight cut-off Amicon Ultra Centrifugal Filter (Millipore, ref. UFC503024).

Normal human skeletal muscle myoblasts were cultured in growth medium and nucleofected with purified immunoglobulins according to the protocol recommended by the supplier (Lonza) and using the P5 Primary Cell 4D-Nucleofector X Kit L (Lonza, ref. V4XP-5024). Nucleofected cells were plated in differentiation medium and harvested for RNA extraction and subsequent RNA sequencing 24 hours after unless otherwise indicated.

Statistical and bioinformatic analysis

For RNAseq analysis, sequencing reads were demultiplexed using bcl2fastq/2.20.0 and preprocessed using fastp/0.21.0. The abundance of each gene was determined using Salmon/1.5.2. Counts were normalised using the Trimmed Means of M values from edgeR/3.34.1 for graphical analysis. Differential expression was performed using limma/3.48.3. The Benjamini-Hochberg correction was used to adjust for multiple comparisons if appropriate. Pathway enrichment analysis employed a one-sided Fisher’s exact test.

To define the specific set of genes associated with each group of interest, we calculated the intersection of the differentially overexpressed genes (q value <0.01) between the group of interest and each of the other comparator groups. Venn diagrams were used to represent graphically these analyses.

For immunofluorescence image analysis, individual muscle fibres and nuclei were segmented using Cellpose/2.1.1,11 employing the neural network model ‘Cytoplasm 2.0’. The intensity of human immunoglobulin was quantified using ImageJ2/2.9.0.

ResultsAntibodies accumulate within myofibres in myositis muscle

Immunoglobulin G staining revealed that antibodies were deposited inside the muscle fibres of patients with each of the different myositis-specific autoantibodies. Myofibres of patients with anti-HMGCR, anti-SRP, anti-MDA5 and anti-Jo1 autoantibodies, all of which recognise cytoplasmic autoantigens, displayed a cytoplasmic pattern of immunoglobulin deposition. In contrast, biopsies from those with anti-Mi2 autoantibodies, which target nuclear proteins, and those with anti-PM/Scl autoantibodies, targeting proteins in the nucleolus, exhibited predominantly nuclear and nucleolar immunoglobulin deposition, respectively. Interestingly, although anti-NXP2 and anti-TIF1g autoantibodies target nuclear antigens, muscle fibres from these patients had a predominant cytoplasmic pattern of immunoglobulin deposition (figure 1, online supplemental figures 1–27). Of note, in patients with anti-NXP2 autoantibodies, the cognate autoantigen was aberrantly localised in the cytoplasm of muscle cells, as will be discussed below.

Figure 1

Immunoglobulin localisation in myositis muscle. Confocal immunofluorescence of human IgG in different autoantibody-defined types of myositis shows antibody deposition in the nuclei of muscle fibres from anti-Mi2-positive patients (A–C); the nucleoli of anti-PM/Scl-positive patients (D–F) the cytoplasm of anti-MDA5-positive patients (G–I); and the cytoplasm of anti-HMGCR-positive patients (J–L). The square boxes contain higher magnification images of one nucleus.

Internalisation of antibodies from anti-Mi2 patients causes the derepression of Mi2/NURD-regulated genes

Anti-Mi2 autoantibodies target the Mi2/NuRD complex,12 which is a transcriptional repressor located in the nucleus.13 Here, using our expanded muscle biopsy RNAseq dataset, we refined our previously identified anti-Mi2-specific gene set by calculating the intersection of the differentially expressed genes (q value cut-off <0.01) between anti-Mi2-positive patients and each of the other comparator groups. This analysis revealed more than 100 genes (eg, SCRT1) that are exclusively overexpressed in anti-Mi2 muscle biopsies (figure 2, online supplemental figures 28 and 29, online supplemental table 3). This specific gene set was strongly associated with transcriptomic markers of disease activity (figure 3A), was present both in male and female patients (online supplemental figure 30) and, as previously reported,5 was more intense in biceps and deltoid compared with quadriceps (online supplemental figure 31). Significantly, the expression of CHD3 and CHD4 was correlated with this specific gene set and demonstrated higher levels in the biceps and deltoid muscles compared with the quadriceps (online supplemental figure 32). Of note, we previously established that the anti-Mi2-specific gene set is highly enriched for genes known to be transcriptionally repressed by the Mi2/NuRD complex.5 Interferon-induced transmembrane protein 5 (IFITM5) was identified as one of the genes specifically overexpressed in anti-Mi2 DM muscle biopsies. While its name suggests otherwise, IFITM5 has not been shown to be interferon inducible.14

Figure 2

Expression of the anti-Mi2-specific gene set in different types of myositis. Top 12 specifically overexpressed genes in muscle biopsies from anti-Mi2-positive patients compared with all the other groups included in the study. Each dot represents the gene expression value of a single patient. AS, antisynthetase syndrome; DM, dermatomyositis; IBM, inclusion body myositis; IM, inflammatory myopathies; NT, histologically normal muscle biopsies; SRP, signal recognition particle.

Figure 3

Correlation of autoantibody-specific gene sets with markers of disease activity. The heat maps depict correlations of the anti-Mi2-specific gene set (A), the anti-PM/Scl-specific gene set (B), and the antisynthetase-specific gene set (C) with transcriptomic markers of type 1 interferon-inducible genes (ISG15, MX1), type 2 interferon-inducible genes (GBP2, IFI30), T-cell markers (CD3E, CD4, CD8), macrophage markers (CD14, CD68), markers of muscle differentiation (NCAM1, MYOG, MYH3, MYH8) and structural mature muscle proteins (ACTA1, MYH1, MYH2).

Internalisation of purified antibodies from anti-Mi2-positive patients into cultured muscle cells induced the overexpression of the same set of anti-Mi2-specific genes that are observed in the muscle biopsies of dermatomyositis patients with anti-Mi2 autoantibodies (figure 4, online supplemental figures 33–35). Among the anti-Mi2-specific genes, 76.1% exhibited differential expression in muscle cells electroporated with antibodies from anti-Mi2-positive patients, compared with cells electroporated with antibodies from other patients at a q value <0.01. In contrast, only 5.2% of the non-anti-Mi2-specific genes displayed differential expression in muscle cells electroporated with antibodies from anti-Mi2-positive patients (76.1% vs 5.2%, p<2.2e−16). Of note, incubation of purified antibodies without electroporation had no transcriptomic effect in any of the autoantibody groups that we studied (online supplemental figure 36).

Figure 4

The expression of anti-Mi2-specific genes in cultured muscle cells electroporated with antibodies. Standardised expression levels (Z-scores) of the more than 100 anti-Mi2-specific genes in human skeletal muscle cells electroporated with purified antibodies from different myositis patients (each column corresponds to a different serum from a different patient). Only cells electroporated with antibodies from anti-Mi2-positive patients have strong expression of anti-Mi2-specific genes.

Antibodies from anti-PM/Scl-positive patients cause the accumulation of transcripts normally degraded by the nuclear RNA exosome complex

Anti-PM/Scl autoantibodies recognise EXOSC9 and EXOSC10, which are key components of the nuclear RNA exosome complex. A key biological function of the exosome complex is to degrade various types of RNA, including long noncoding RNAs and divergent transcripts.15 16

We performed bulk RNAseq on 19 muscle biopsies from patients with anti-PM/Scl autoantibodies and compared them to the rest of the samples included in the study. We defined the set of anti-PM/Scl-specific genes by calculating the intersection of the differentially expressed genes (q value cut-off <0.01) between anti-PM/Scl-positive patients and each of the other comparator groups (figure 5, online supplemental figures 37 and 38, online supplemental table 4). This analysis identified 236 overexpressed RNAs and 1 underexpressed gene. Most overexpressed genes were long noncoding RNAs and divergent transcripts, suggesting a dysfunction of the nuclear RNA exosome complex exclusively in anti-PM/Scl-positive patients. The expression of these genes had a modest but consistent correlation with transcriptomic markers of disease activity (figure 3B), was present both in male and female patients (online supplemental figure 39) and was similar across different muscle groups (online supplemental figure 40).

Figure 5

Expression of the anti-PM/Scl-specific gene set in different types of myositis. Top 12 specifically overexpressed genes in muscle biopsies from in muscle biopsies from anti-PM/Scl-positive patients compared with all the other groups included in the study. Each dot represents the gene expression value of a single patient. AS, antisynthetase syndrome; DM, dermatomyositis; IBM, inclusion body myositis; IM, inflammatory myopathies; NT, histologically normal muscle biopsies; SRP, signal recognition particle.

To verify that antibodies from anti-PM/Scl patients induce the expression of the anti-PM/Scl-specific gene set, we electroporated purified antibodies from anti-PM/Scl patients into human myoblasts. Indeed, this treatment induced the overexpression of the same set of anti-PM/Scl-specific genes that are observed in the muscle biopsies of anti-PM/Scl-positive patients (figure 6 and online supplemental figures 35 and 41). Of the genes identified in muscle biopsies as anti-PM/Scl-specific, 76.3% displayed differential expression in muscle cells electroporated with antibodies from anti-PM/Scl-positive patients, in comparison to cells electroporated with antibodies from other patients at a q value <0.01. In contrast, 1.7% of the genes not specific to anti-PM/Scl patient muscle biopsies exhibited differential expression in muscle cells electroporated with antibodies from anti-PM/Scl-positive patients (76.3% vs 1.7%, p<2.2e−16).

Figure 6

The expression of anti-PM/Scl-specific genes in cultured muscle cells electroporated with antibodies. Standardised expression levels (Z-scores) of the more than 200 anti-PM/Scl-specific genes in human skeletal muscle cells electroporated with purified antibodies from different myositis patients (each column corresponds to a different serum from a different patient). Only cells electroporated with antibodies from anti-PM/Scl-positive patients have strong expression of anti-PM/Scl-specific genes.

Supporting evidence in other autoantibody groups

While the effects of internalised antibodies on transcriptomic profiles were readily apparent for those autoantibodies targeting proteins with broad effects on transcription or RNA metabolism (such as the Mi2/NuRD complex for anti-Mi2 autoantibodies and the nuclear RNA exosome complex for anti-PM/Scl autoantibodies), only a single overexpressed gene was identified among all the other non-anti-Mi2 and non-anti-PM/Scl autoantibody groups (online supplemental figures 42 and 43). However, we observed less conclusive or indirectly supportive evidence suggesting the biological effects of autoantibodies on their target autoantigens in muscle biopsies from patients with other myositis autoantibodies.

IFN1 overexpression in dermatomyositis muscle is driven by IFNB1

It is well established that type I interferon-stimulated genes are highly overexpressed in dermatomyositis muscle tissue.7 17 18 However, the type I IFN family includes multiple different IFN alpha genes and one IFN beta gene. Here, we used our muscle biopsy transcriptomic dataset to determine which of these is most likely to be driving the expression of type I interferon-stimulated genes in dermatomyositis muscle. Among the various type I interferons, IFNB1 had the highest expression levels in dermatomyositis and could be detected in more than 60% of the muscle biopsy samples from patients with DM. In contrast, IFNA5 was observed in less than 40% of samples and the other IFN alpha genes were detected in less than 20% of cases. Furthermore, our analysis revealed a correlation between the abundance of IFNB1 and the expression of positive (eg, MX1) and negative (eg, TTN) IFN1-inducible genes in dermatomyositis muscle (online supplemental figures 44–46). These results suggest that IFNB1 drives the overexpression of type I IFN-inducible genes in dermatomyositis muscle.

Interestingly, the function of several dermatomyositis autoantigens is to regulate the expression of individual type-I interferon proteins. For instance, NXP219 and TIF1g19 20 inhibit IFNB1 expression; dysfunction of these dermatomyositis autoantigens should result in increased IFNB1 expression rather than IFNA, aligning with our observations.

To define the set of genes that are inducible by IFNB1, we treated cultured myoblasts with different types of type 1 interferon. Interestingly, IFNB1 treatment also stimulated its own expression in cultured differentiated human myoblasts (online supplemental figure 47). This suggests the possibility that the overexpression of IFNB1 can persist through a self-sustaining loop and confirms the results from publicly available datasets suggesting the same phenomenon in other cell types (monocytes (GSE34627)21).

Antibodies from anti-MDA5 patients induce overexpression of IFNB1

Anti-MDA5 autoantibodies bind to the helicase domains of the MDA5 protein.22 MDA5 is primarily a cytoplasmic sensor of viral double-stranded RNA. Binding to double-stranded RNA activates the MDA5 protein, which ultimately induces the transcription of type I interferon.23 Interestingly, internalisation of purified immunoglobulin from three of five anti-MDA5 patients into human myoblasts induced a robust overexpression of IFNB1 and IFNB1-inducible genes (online supplemental figures 35 and 48). This suggests the possibility that anti-MDA5 autoantibodies bind and activate MDA5.

Among the differentially expressed genes identified by comparing biopsies from anti-MDA5-positive patients with all other biopsies at a q value <0.01, 13.7% were also differentially expressed in muscle cells electroporated with antibodies from anti-MDA5-positive patients, compared with cells electroporated with antibodies from other patients at a q value <0.01. In contrast, only 0.2% of the non-differentially expressed genes in anti-MDA5 demonstrated differential expression in muscle cells electroporated with antibodies from anti-MDA5-positive patients (13.7% vs 0.2%, p<2.2e−16).

Internalisation of antibodies from antisynthetase patients induces a transcriptional phenotype consistent with dysfunction of aminoacyl-tRNA synthetase

Anti-synthetase autoantibodies recognise members of the aminoacyl-tRNA synthetase family of proteins. These enzymes load the appropriate amino acid into its corresponding tRNA for eventual integration onto an elongating polypeptide. The most common of these autoantibodies recognise the histidyl-tRNA synthetase (anti-Jo1). Of note, it has been reported that anti-Jo1 autoantibodies inhibit the function of its target protein in vitro.24

Anti-synthetase autoantibodies recognise members of the aminoacyl-tRNA synthetase family of proteins. These enzymes load the appropriate amino acid into its corresponding tRNA for eventual integration onto an elongating polypeptide. The most common of these autoantibodies recognise the histidyl-tRNA synthetase (anti-Jo1). Of note, it has been reported that anti-Jo1 autoantibodies inhibit the function of its target protein in vitro.24

Previously, we reported that a set of genes including CAMK1G, EGR4 and CXCL8 are overexpressed in the muscle biopsies of patients with anti-Jo1 autoantibodies.6 In this study, we have expanded our cohort by including muscle biopsies from patients with non-Jo1 antisynthetase autoantibodies (eg, anti-PL7 and anti-PL12). These tissue samples also overexpress genes that were elevated in muscle biopsies from anti-Jo1-positive patients6 and a set of antisynthetase syndrome-specific genes could be defined using a similar approach to the one used for anti-Mi2 or anti-PM/Scl-positive samples (online supplemental figures 49 and 50, online supplemental table 5). However, the overexpression of the antisynthetase syndrome-specific genes was less dramatic than the one observed for the anti-Mi2 or anti-PM/Scl-specific gene sets. This set of specific genes had a moderate correlation with transcriptomic markers of disease activity (figure 3C).

Internalisation of purified immunoglobulin from anti-Jo1 patients into cultured human myoblasts induced overexpression of some of the same genes that we identified in muscle biopsies from patients with antisynthetase autoantibodies (online supplemental figure 35). However, there were no genes significantly differentially expressed in muscle cells electroporated with antibodies from anti-Jo1-positive patients compared with cells exposed to antibodies from other patients at a q value <0.01. That being said, by relaxing the criteria to a q value of 0.05, 13.3% of differentially expressed genes in muscle biopsies from anti-Jo1 patients (compared with all other biopsies) were also significantly overexpressed in muscle cells electroporated with antibodies from anti-Jo1-positive patients. By comparison, only 0.07% of genes that were not differentially expressed in anti-Jo1 patients were overexpressed in myoblasts electroporated with antibodies from anti-Jo1-positive patients (13.3% vs 0.07%, p=6.2e−5).

To explore whether this transcriptional programme may be the consequence of aminoacyl tRNA-transferase dysfunction, we compared the set of antisynthetase syndrome-specific genes with a publicly available gene expression dataset from HepG2 human hepatoma cells treated with histidinol, an inhibitor of histidyl-tRNA synthetase.25 This dataset showed overexpression of genes specifically overexpressed in patients with the antisynthetase syndrome (eg, EGR4, CXCL8). In addition, pathway enrichment analysis demonstrated a significant association between the genes in this dataset and those differentially expressed in patients with antisynthetase autoantibodies (p<0.001).

Taken together, our observations that (a) immunoglobulin accumulates in the cytoplasm of muscle cells in anti-Jo1-positive patients, (b) internalisation of antibodies from anti-Jo1 patients into cultured muscle cells recapitulates the transcriptomic profile seen in muscle biopsies from anti-Jo1 patients and (c) pharmacological inhibition of histidyl-tRNA synthetase recapitulates the transcriptomic profile seen in muscle biopsies from anti-Jo1 patients suggest that antibodies from anti-Jo1-positive patients can enter live muscle cells and inhibit the function of its target autoantigen.

Lipids accumulate in the muscle fibres of patients with anti-HMGCR autoantibodies

Anti-HMGCR autoantibodies recognise the rate-limiting enzyme of the cholesterol biosynthetic pathway. Recent studies demonstrated that mutations disrupting the enzymatic activity of HMGCR cause a genetic myopathy characterised by myofibre necrosis. This suggests the possibility that disruption of HMGCR by autoantibodies could lead to the same pathogenic abnormalities.26 27

Muscle biopsies from immune-mediated necrotising myopathy patients with anti-HMGCR autoantibodies do not have a unique transcriptomic profile, as seen in patients with anti-Mi2 or anti-PM/Scl autoantibodies. Indeed, the only specific differentially expressed gene that we could identify was APOA4.6 However, internalisation of antibodies from anti-HMGCR-positive patients into cultured myoblasts did not lead to an increase in APOA4 expression.

Interestingly, on review of all available muscle biopsies, we noted accumulations of lipids in myofibres of patients with anti-HMGCR autoantibodies (online supplemental figures 51–53). Specifically, 90% (18/20) of muscle biopsies from anti-HMGCR-positive patients had lipid accumulation compared with 8.3% (10/110) of patients with other types of myositis (p<0.001). As myofibres from patients under pharmacological inhibition of HMGCR by statins may also have prominent lipid accumulations,28 we hypothesise that autoantibody-mediated HMGCR dysfunction may induce the same effect by causing the accumulation of acetyl-CoA and the subsequent production of excess lipids. Of note, only one of the patients whose muscle biopsies were reviewed had been exposed to statins at the time of the muscle biopsy.

In patients with anti-NXP2 autoantibodies, NXP2 protein is aberrantly localised to the cytoplasm

Both NXP2 and TIF1g proteins inhibit IFNB1 expression by binding to regulatory regions of this gene within the nuclei of cells.19 20 Thus, if autoantibodies recognising these proteins were to disrupt their function, one would expect to see the overexpression of IFNB1 and IFNB1-inducible genes just as we have observed in the muscle tissue of dermatomyositis patients with these autoantibodies (online supplemental figures 44 and 45).7

Interestingly, our immunofluorescence microscopy studies showed that in muscle biopsies from those with autoantibodies recognising NXP2 or TIF1g, immunoglobulin was localised to the cytoplasm rather than the nucleus, where these two proteins normally reside (online supplemental figures 14–17). Of note, muscle biopsies from anti-NXP2-positive patients costained for NXP2 protein and immunoglobulin showed that both proteins were predominantly colocalised in the cytoplasm (online supplemental figures 19 and 20). In contrast, NXP2 staining was predominantly localised to the nucleus in muscle biopsies from patients without anti-NXP2 autoantibodies (online supplemental figure 21). The fact that the NXP2 protein is mislocalised to the cytoplasm suggests that the anti-NXP2 autoantibodies may sequester their autoantigen in the cytoplasm of muscle cells. This could cause increased expression of IFNB1 in the absence of inhibition by nuclear NXP2.

Electroporation of immunoglobulin from patients with anti-NXP2 and anti-TIF1g autoantibodies into myoblasts did not activate the IFNB1 pathway. However, it should be noted that electroporation generates pores in the nuclear membrane allowing antibodies to enter the cell nuclei. This could preclude anti-NXP2 autoantibodies from retaining NXP2 protein in the cytoplasm of the electroporated cells.

We could not find an effective commercial anti-TIF1g antibody histochemical reagent to replicate the same study in anti-TIF1g muscle biopsies.

Discussion

This study reveals two novel features in myositis. First, our immunofluorescence analyses demonstrate that antibodies are internalised within living human muscle cells from patients with a variety of myositis autoantibodies and that they localise to the site of the corresponding target autoantigen. Second, using antibody internalisation experiments along with transcriptomic analyses, we have shown that these internalised autoantibodies disrupt the normal function of their protein targets (figure 7). Taken together, these findings disprove the conventional assumption that myositis autoantibodies are incapable of reaching their intracellular target and exerting a functional effect.

Figure 7

A model of pathogenic myositis autoantibody internalisation. In myositis, autoantibodies are internalised into the muscle fibres, disrupting the normal biological function of their autoantigen, which mediates the pathogenesis of the disease (A). For instance, anti-Mi2 autoantibodies (B) interfere with the Mi2/NuRD complex inducing the derepression of more than 100 genes. Similarly, anti-PM/Scl autoantibodies (C) cause a dysfunction of the nuclear RNA exosome complex, impairing the normal degradation of various types of RNA.

In patients with autoantibodies targeting autoantigens involved in transcriptional regulation or RNA metabolism, notably those linked to the Mi2/NuRD complex in anti-Mi2 individuals or the nuclear RNA exosome complex in patients with anti-PM/Scl autoantibodies, we observed robust effects on transcriptomic profiles consistent with dysfunction of their targets. Conversely, muscle biopsies from patients with autoantibodies targeting proteins functionally related to IFNB1 (MDA5, TIF1g, NXP2) showed a transcriptional pattern restricted to IFNB1 and IFNB1-inducible genes.

Muscle biopsies from patients with autoantibodies targeting proteins involved in protein synthesis (aminoacyl-tRNA synthetases), protein trafficking (SRP) or lipid metabolism (HMGCR), exhibited less distinct transcriptional patterns. Nonetheless, we found limited and indirect evidence of amino acid deprivation in antisynthetase syndrome muscle biopsies, and lipid dysregulation in muscle biopsies from anti-HMGCR-positive patients. These findings are also consistent with a disease mechanism in which myositis autoantibodies reach their autoantigen inside of the cell and disrupt their normal function (figure 7).

Our investigation has several limitations. First, we only studied antibody internalisation in muscle cells. Additional studies will be required to determine whether autoantibodies enter other cells within the muscle tissue, such as endothelial cells and fibroblasts, where they could also cause functional effects. Second, we did not study other organ systems frequently affected in myositis patients, such as the lungs or skin, to determine whether autoantibody internalisation may play a pathophysiological role in these locations. Finally, although we show that myositis autoantibodies can enter living muscle cells and disrupt the normal function of their targets in either the cytoplasm or the nucleus, the mechanisms mediating autoantibody internalisation remain to be elucidated. Potential mechanisms for autoantibody internalisation into human cells in other autoimmune diseases include via Fc receptors29 and the brush border myosin 1 protein.30 Additionally, in muscle fibres, mechanical forces may also play a role in disrupting the cell membrane and facilitating autoantibody internalisation.

Our study was limited to studying autoantibodies found in patients with myositis. While we did not study other autoimmune diseases, our current findings in myositis are in line with older studies demonstrating intracellular immunoglobulin deposition in tissues from patients with scleroderma, lupus and mixed connective tissue disease.29 31 32 Indeed, we propose that the role of pathogenic intracellular autoantibody deposition in other systemic autoimmune diseases, including systemic sclerosis, vasculitis and lupus, should be re-examined in light of our findings in myositis.

In conclusion, our study shows that autoantibodies are deposited inside live muscle cells where they disrupt the normal function of their target proteins. Beyond myositis muscle, this pathophysiological mechanism may be relevant in other tissues and other autoimmune diseases.

Data availability statement

Data are available on reasonable request. Any anonymised data not published within the article will be shared by request from any qualified investigator.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

All biopsies were from subjects enrolled in institutional review board (IRB)-approved longitudinal cohorts in the National Institutes of Health, the Johns Hopkins, the Clinic Hospital, the Vall d’Hebron Hospital, the Mayo Clinic and the Charité-Universitätsmedizin Berlin. Participants gave informed consent to participate in the study before taking part.

Acknowledgments

This work is dedicated to the memory of Dr. Josep Maria Grau. Special thanks to Julie Thompson for her invaluable help maintaining the NIH Natural History Protocol, the NIAMS Sequencing Core and its members.