Mutual reinforcement of lymphotoxin-driven myositis and impaired autophagy in murine muscle

Abstract

Inclusion body myositis (IBM) is a progressive muscle disorder characterized by inflammation and degeneration with altered proteostasis. To better understand the interrelationship between these two features, we aimed to establish a novel preclinical mouse model.

First, we used quantitative PCR, in situ hybridization and immunohistochemistry to determine the expression of pro-inflammatory chemokines and cytokines including lymphotoxin (LT)-signalling pathway components in human skeletal muscle tissue diagnosed with myositis. Based on these results, we generated a mouse model that we analysed at the histological, ultrastructural, transcriptional, biochemical and behavioural level. Lastly, we subjected this model to anti-inflammatory treatments.

After confirming and extending previous data on activation of LT-signalling in human myositis, we generated distinct transgenic mouse lines co-expressing LTα and -β in skeletal muscle fibres. Transgenic mice displayed chronic myositis accompanied by dysregulated proteostasis, including an altered autophagolysosomal pathway that initially showed signs of activation and later exhaustion and decreased flux. To enhance the latter, we genetically impaired autophagy in skeletal muscle cells. Autophagy impairment alone induced a pro-inflammatory transcriptional state, but no obvious cellular inflammation. However, the combination of LT-driven myositis with autophagy impairment induced the full spectrum of characteristic molecular and pathological features of IBM in skeletal muscle, including protein inclusions with typical ultrastructural morphology and mild mitochondrial pathology. Our attempts to treat the pathology by subjecting these mice to corticosteroids or anti-Thy1.2 antibodies mirrored recent treatment failures in humans, i.e. none of these treatments resulted in significant clinical improvement of motor performance or the transcriptional profile of muscle pathology.

In summary, these data provide evidence that inflammation and autophagy disruption play a synergistic role in the development of IBM-like muscular pathology. Furthermore, once established, IBM-like pathology in these mice, as in human IBM patients, cannot be reverted or prevented from progression by conventional means of immunosuppression. We expect that this novel mouse model will help to identify future treatment modalities for IBM.

Inclusion body myositis (IBM) is a progressive muscle disorder characterized by inflammation and degeneration with altered proteostasis. Bremer et al. establish a novel mouse model that reveals a synergistic role for chronic lymphocytic inflammation and dysregulated autophagy in the development of IBM-like muscular pathology.

Introduction

Inclusion body myositis (IBM) belongs to a heterogenous group of idiopathic inflammatory myopathies (IIM)¹ that affects between 25 and 46 per million people, mostly above the age of 45 years.^2,3 Clinically, patients with IBM usually present with progressive asymmetric weakness and atrophy often affecting the quadriceps femoris and/or finger flexors.⁴ Histopathologically, IBM is characterized by two main features, inflammation and degeneration.⁵ Initial diagnostic criteria allowed the diagnosis of IBM solely based on these pathological features,⁶ while more recent diagnostic criteria also take clinical features into account³ and define the pathological feature ‘inflammation’ as endomysial inflammatory infiltrates and upregulation of major histocompatibility complex class I (MHCI).³ CD8⁺ cytotoxic T cells usually predominate in these endomysial inflammatory infiltrates and some of these T cells invade non-necrotic myofibres.^6,7 The second diagnostic criterion, ‘degeneration’, is defined as the light microscopic presence of rimmed vacuoles and/or protein accumulation or 15–18 nm filaments on electron microscopy.³ These rimmed vacuoles consist of autophagic vacuoles and ubiquitin-positive multiprotein inclusions, containing numerous proteins associated with neurodegenerative disorders, such as amyloid-β, phosphorylated tau and phosphorylated TAR DNA-binding protein 43 (TDP-43), prion protein, while p62.^8-11 At the cellular level, different biochemical pathways that can ameliorate protein accumulations, including endoplasmic reticulum (ER) stress/unfolded protein response (UPR), autophagolysosomal pathway and proteasomal degradation are severely dysregulated in IBM. Accumulation of misfolded proteins in the ER leads to ER stress, which in turn activates the UPR as a protective mechanism. Increased expression of factors characteristic of UPR (ATF4, CHOP, GRP78, spliced XBP1) were observed in IBM but not in vacuolar myopathy due to GNE mutation.^12-14 Activities of lysosomal enzymes cathepsin D and B were decreased and autophagosome maturation, i.e. LC3-II, was increased in human IBM muscle fibres, suggesting disturbed autophagosomal activity.⁸ Even though expression of proteasome subunits was increased, one proteasome subunit was biochemically interacting with protein inclusions in IBM and proteasomal proteolytic activities were reduced, suggesting that the proteasomal degradation system is also dysfunctional in IBM.¹⁵ Furthermore, mitochondrial alterations, including cytochrome c oxidase (COX)-deficient fibres, accumulation of abnormal mitochondria, sometimes reflected in the presence of ragged red fibres, and mitochondrial DNA deletions are frequently found in IBM.^7,16,17

The simultaneous occurrence of inflammation and degenerative pathological features in IBM has triggered considerable debate regarding which is the primary cause and driver of the disease. To solve this ‘chicken and egg’ problem, i.e. whether inflammation drives myodegeneration or whether primary degeneration leads to inflammation,^7,18,19 it is essential to understand the interaction between these two pathologies. Over the past few decades, considerable efforts have been made to resolve this conundrum. Evidence supports that IBM is an autoimmune/inflammatory disease: genetic studies show strong genetic linkage to autoimmunity-associated human leukocyte antigen (HLA) variants²⁰ and IBM is associated with other autoimmune diseases such as Sjogren's syndrome.²¹ Previous studies have shown that muscle fibre-invading CD8⁺ T cells had restricted T-cell receptor sequences, suggesting that they are clonally expanded,²² but a muscle antigen has remained unknown. Autoantibodies have been detected in around 25% [anti-valosin-containing protein (VCP)] to up to 50% [anti-cytosolic 5′-nucleotidase 1A (NT5c1A)] of IBM patients, with 15% being positive for both antibodies. These antibodies are, however, not specific for IBM^23-25 and the clinical value of anti-VCP antibodies in particular remains to be determined.²⁶ Few studies have shown links between inflammatory mediators and signs of disturbed proteostasis. Notably, the exposure of human myotubes to interleukin (IL)-1β caused upregulation of the amyloid beta precursor protein (APP) with subsequent intracellular aggregation of amyloid-β, suggesting that pro-inflammatory mediators can induce amyloid-β-associated muscle cell degeneration, thus supporting inflammation as the primary event.²⁷

However, there are also compelling arguments in favour of a primarily degenerative pathogenesis. For example, protein inclusions, i.e. paired helical filaments in IBM strikingly resemble those in neurodegenerative diseases such as Alzheimer's disease.^10,11 APP can contribute to myopathy since its overexpression caused mitochondrial abnormalities and COX negativity in cultured human muscle fibres.²⁸ In addition, muscle cell-specific APP overexpression induced late-onset mild myopathic changes characterized by an increased number of internally located myonuclei and mostly granulocytic, but not lymphocytic inflammation in mice.²⁹ When Mck-driven App overexpression was genetically combined with an Alzheimer's disease-associated mutation in presenilin 1, mice displayed T cellular inflammation and increased amyloid-β deposits in muscle fibres, but vacuoles, muscle fibre degeneration or typical inclusions like in IBM were still not described.³⁰ Nuclear loss and the cytoplasmic accumulation of TDP-43 into inclusions observed in IBM muscle^9,10 resemble pathological features typical for amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD).³¹ Furthermore, ER stress was shown to activate pro-inflammatory NF-κB signalling^32,33 and exposure of human myoblasts and C2C12 mouse skeletal muscle cells to amyloid-β peptides lead to increased IL-6 expression that may in turn stimulate CD8⁺ T cells.³⁴ These findings show that disturbed protein homeostasis can induce inflammation in skeletal muscle. MHCI upregulation in IBM is usually regarded as evidence in favour of a primarily inflammatory disease, but skeletal muscle-specific overexpression of MHCI in mice induced myopathic changes with protein accumulations, vacuolation and ER stress/UPR induction in the absence of lymphocytes³⁵—inflammatory infiltrates in this model were of macrophage/myeloid lineage only without detectable T or B cells.³⁶ The first compelling argument for a primary degenerative pathogenesis, however, is that immunotherapies, which are successful in treating other inflammatory myopathies, lack efficacy in IBM. Moreover, T-cell depletion by alemtuzumab (anti-CD52 antibody) resulted in only transient clinical stabilization in selected IBM patients.³⁷ However, this effect was inconsistent and there was no significant difference in molecular parameters in the muscle of alemtuzumab-treated IBM patients.³⁸ The second main argument in favour of a primary degenerative pathology is that prominent inflammatory infiltrates in addition to vacuolar myopathy can be found in the context of certain gene mutations causing vacuolar myopathy, including some cases of multisystem proteinopathy. Examples are facioscapulohumeral muscular dystrophy (FSHD),³⁹ and cases with mutations in the valosin-containing protein (VCP) gene, a multifunctional ATPase regulating autophagy and other cellular processes,^40-43 the TARDBP (TDP43) gene⁴⁴ and the glucosamine (UDP-N-acetyl)-2-epimerase/N-acetylmannosamine kinase (GNE) gene that is required for sialic acid biosynthesis.⁴⁵ Thus, gene defects, including those involved in proteostasis, can rarely trigger vacuolar myopathy with inflammation, presenting as ‘hereditary myositis’. However, this is not a consistent finding in genetic myopathies. Furthermore, this secondary inflammation differs markedly from myositis, T cells do not typically invade intact muscle fibres and clonal expansion remains to be determined.

An animal model with the two cardinal IBM-like features, lymphocytic inflammation and characteristic inclusions is unavailable to date. Given the importance of the inflammatory pathology in disease pathogenesis, we aimed to generate a mouse model with chronic lymphocytic inflammation to determine the consequence of inflammation on the proteostasis machinery and muscle fibre integrity in vivo. Besides the involvement of interferon signalling in human myositis,⁴⁶ there is evidence of induced LT signalling and activation of its downstream mediator NF-κB^47,48 in IIM,⁴⁹ including IBM, in which normal-appearing myofibres express LTβ, possibly reflecting early myofibre damage.⁵⁰ ER stress, observed in IBM can induce NF-κB in cultured human muscle fibres.³² Interestingly, the autophagy regulator and cargo adapter p62, which accumulates in IBM inclusions is known to crosstalk with NF-κB signalling.^51,52 Here, we provide further evidence that LTs and their target genes are upregulated in human IIM. By expressing LTα and -β specifically in skeletal muscle fibres, we established a transgenic mouse line (HSA-LTα/β) with chronic myositis. Chronic myositis was accompanied by dysregulated proteostasis, including the autophagolysosomal pathway. The genetic combination of LT-driven inflammation and autophagy impairment in mice lead to characteristic molecular and myopathological features resembling IBM, including lymphocytic inflammation and protein inclusions with typical ultrastructural morphology.

Materials and methods

Mouse housing and permits

Animal procedures were performed in accordance with EU and Swiss animal laws and approved by the government of Zurich, Bavaria, Baden-Württemberg and Niedersachsen, respectively (reference numbers 200/2007, 35-9185.81/G-113/17, 33.9-42502-04-18/2916). Human specimens were provided by the Armed Forces Institute of Pathology (AFIP), Washington D.C., after approval by the local ethics committee and then processed anonymously. ATG5<tm1(flox)> mice (here referred to as Atg5^flox/flox) were kindly provided by Noboru Mizushima (University of Tokyo, Japan). B6.Cg-Tg(ACTA1-cre)79Jme/J (here termed ACTA-Cre) and B6.FVB(129S4)-Tg(Ckmm-cre)5Khn/J (here termed Ckmm-Cre) were obtained from The Jackson Laboratory. Analysed mice were of mixed gender.

Generation of tg(HSA:LTα_HSA:LTβ) mice

LTs were amplified from plasmid DNA and cloned between human α-skeletal actin (HSA) promoter (−1905 to +239), kindly provided by Edna Hardeman⁵³ and SV40 poly A tail. Linearized HSA-LTα and HSA-LTβ were co-injected into pro-nuclei and embryos were transferred into pseudo-pregnant C57BL/6J at Biomodels Austria GmbH (Biat) by Thomas Rülicke. Animals were maintained on C57BL/6J background under specific pathogen-free (SPF) conditions, see also the Supplementary material.

Genotyping

DNA was extracted from tail or ear biopsies. For primer sequences see the Supplementary material.

Quantitative real-time PCR

Messenger RNA (mRNA) was extracted and reversely transcribed. Quantitative real-time PCR (qPCR) was performed using SYBR Green. For experimental details see the Supplementary material. Results are displayed by heat maps, created using the ComplexHeatmap package in Bioconductor/ RStudio Version 4.3.0.^54,55

Histology, immunohistochemistry and muscle fibre size determination

For detailed protocols see the Supplementary material. Muscle fibre size distribution was determined on haematoxylin and eosin (H&E)-stained cross sections of the quadriceps femoris muscle, using ImageJ.

Electron microscopy

Samples were fixed in 3% glutaraldehyde in Sörensen buffer, incubated in 1% osmium, then further processed and embedded in Epon® resin as described in the Supplementary material.

Behavioural tests before and during drug treatments

Tests were performed as previously described,⁵⁶ see also the Supplementary material.

In situ RNA hybridization

RNA in situ hybridization was performed using the RNAscope 2.0® Assay on formalin-fixed paraffin-embedded (FFPE) tissue for murine and the RNAscope Assay-BROWN for detection on the BOND system (Advanced Cell Diagnostic/Leica) for human tissue according to the manufacturer's protocol. See also the Supplementary material.

Mitochondrial DNA sequencing

Mitochondrial DNA was amplified using established primers.⁵⁷ Long-range PCR products were purified. Libraries were prepared, pooled and sequenced on a MinION sequencer. The resulting BAM files were aligned to the GRCm38 reference genome using minimap2, and structural variations were identified with Sniffles2.⁵⁸ See the Supplementary material for details.

RNA sequencing

RNA processing, library preparation and sequencing (RNA-Seq) are described in the Supplementary material. Data were aligned to the GRCh38p14 genome and counted with STAR Aligner,⁵⁹ further analysed and visualized with BioJupies.⁶⁰ Integrated pathway analysis and visualization was done with Pathview.⁶¹

Analysis of published human RNA sequencing data

We downloaded datasets from the Gene Expression Omnibus (GEO) database: GSE151757,⁶² GSE220915.⁶³ Differentially expressed genes (DEGs) were determined in R. Adjusted P-values were calculated using DESeq2.⁶⁴ For details see the Supplementary material.

Drug treatments

Mice were subjected to continuous behavioural tests from 90 days of age. Treatment was initiated 30 days later (at 120 days), see also the Supplementary material.

Statistical analysis

Statistical analysis was mostly performed using GraphPad Prism 9.3.1 or GraphPad Prism 10.5.0 (GraphPad Software, Boston, USA). See figure legends for applied tests. For categorial immunohistochemical data, we used Fisher's exact test (GraphPad online tool). The chi-squared test was used to compare the fibre size distribution. All displayed graphs show mean and standard deviation.

Additional methods

For additional details on forelimb grip strength, hanging wire test, muscle MRI, enzyme-linked immunosorbent assays (ELISA) and western blot see the Supplementary material.

Results

Lymphotoxin and target gene expression in human idiopathic inflammatory myopathies

IIMs in humans are characterized by signs of inflammation and skeletal muscle fibre necrosis and regeneration as well as atrophy (Fig. 1A). Endomysial infiltrates mainly consist of T cells in polymyositis (PM) and IBM and CD4⁺ T cells and B cells in dermatomyositis (DM).⁶⁵ In PM and IBM, some T cells invade non-necrotic muscle fibres. Typically, MHCI is upregulated on the sarcolemma. In IBM, the light microscopic correlate of ‘degeneration’ is the presence of abnormal autophagic vacuoles (‘rimmed vacuoles’). Immunohistochemically, there is accumulation of the autophagy regulator and cargo adaptor p62⁵² combined with the aggregation of proteins such as phospho-TDP43 (Fig. 1A). To identify candidate pathomechanisms that could facilitate the establishment of suitable animal models, we examined the expression of several cytokines and chemokines in human inflammatory myopathies using qPCR. We found significant upregulation of several cytokines and chemokines, i.e. LTα was upregulated in PM, while LT/NF-κB target genes^47,48,66-69 were upregulated in DM (CCL19, CXCL10 and BAFF) and IBM (CCL5 and CCL19). The activation of NF-κB pathways was in line with previous investigations on human myositis^49,50,70 (Fig. 1B and C). Using in situ hybridization and immunohistochemistry, we observed LTα and LTβ as well as LTβR expression in inflammatory infiltrates as well as in muscle cells in IBM (Fig. 1D and E). Previously published RNA-Seq data of human myositis did not specifically uncover enhanced LT signalling in IBM.^62,63 We integrated both published datasets, analysed the expression of selected LT signalling-associated genes and observed a strong and significant upregulation of LTα, LTβ and LT target genes in IBM (Fig. 1F). Other IIMs,⁶³ such as immune-mediated necrotizing myopathy (IMNM), anti-synthetase syndrome (ASyS) and DM also showed upregulation of LTα. In ASyS and DM, we also observed a significant, albeit weaker upregulation of LTβ (Supplementary Fig. 1). A previous study demonstrated a significant increase of CXCL10 and CCL5, but not of CXCL13 and CCL19, in IBM compared with other inflammatory myopathies, suggesting that LT target genes CXCL10 and CCL5 are particularly upregulated in IBM.⁷¹ These results confirm and extend previous data^47-50 demonstrating that LT signalling and its downstream targets are activated in human IIM.

Figure 1.

Histological features and chemokine as well as cytokine expression in human inflammatory myopathies. (A) Trichrome staining showing endomysial inflammatory infiltrates (white arrowheads) as well as muscle fibre atrophy and endomysial fibrosis (black arrowheads), which is most severe in this case of inclusion body myositis (IBM). Yellow arrowheads: rimmed vacuoles. Immunohistochemistry (brown signal): CD3⁺ T cells occasionally invading intact muscle fibres in polymyositis (PM) and IBM (blue arrowheads); CD20⁺ B cells are most frequent in dermatomyositis (DM), but are occasionally also seen in PM and IBM; CD68⁺ macrophages are frequent in all of these idiopathic inflammatory myopathies (IIMs); phospho-TDP43-, p62- and ubiquitin-positive inclusions in IBM (black arrowheads). (B) Heat map of chemo- and cytokine expression in PM, DM and IBM compared with controls. P-values were determined using the two-tailed Mann–Whitney test and are shown if the expression was significantly different from control after Bonferroni correction (P < 0.17). Upregulation of lymphotoxin (LT)α and LT-signalling target genes including CCL19 and CXCL10. (C) Schematic drawing of LTα/LTβ heterotrimer and LTα homotrimer activation of both tumor necrosis factor (TNF) and LTβ receptor, leading to the translocation of nuclear factor-kappa B (NF-κB) into the nucleus. (D) In situ hybridization of LTα and LTβ receptor (LTβR). Controls (i and iii) show no LTα signal and little LTβR signal in muscle fibres (arrows in enlarged images shown below in iii). In IBM, little LTα is detected in inflammatory infiltrates (black arrowheads) as well as in myonuclei (black arrows in enlarged images shown below in ii). In IBM, LTβR is also detected in both, inflammatory infiltrates (black arrowheads) as well as in myonuclei (black arrows, in enlarged images shown below in iv as well as in another area in v). (E) Immunohistochemistry (brown signal) of LTβ in muscle cells of IBM cases (ii–iv) compared with controls (i). Bottom row shows enlarged areas of i–iv. LTβ is detected in inflammatory infiltrates (black arrowheads) in IBM. In muscle fibres, the overall LTβ signal is slightly increased in IBM with a particular enrichment in rimmed vacuoles (black arrows) and where lymphocytes infiltrate intact muscle fibres (white arrowheads). (F) Normalized read counts of LTα, LTβ as well as LT target genes CXCL10, CXCL13, CCL5 and CCL19 compared with the housekeeping gene HPRT1 in controls and IBM samples derived from two previously published RNA-Sequencing datasets, showing a strong and significant increase of LT-related gene expression.

Lymphotoxin expression in skeletal muscle causes chronic myositis in mice

Considering that LT signalling is prominent in human IIM and its role in inducing chronic inflammation in other tissues, including the liver,⁷² kidney,⁷³ pancreas^73,74 and brain,⁷⁵ we generated a mouse model of chronic myositis by co-expressing LTα and LTβ in skeletal muscle. We cloned both LTα and LTβ separately behind the HSA promoter and co-injected both transgenic constructs into pronuclei derived from C57BL/6J mice (Fig. 2A). We obtained four founder mice that contained both transgenes, HSA-LTα and HSA-LTβ and transmitted these together through the germline, suggesting concatemeric genomic insertion (Fig. 2B). Line #8 did not breed sufficiently and could not be maintained. Lines #8, #22 and #44 displayed strong ectopic expression in other organs, including the heart, brain, kidney and lung. Therefore, Line #19 was selected for further study and is termed HSA-LTα/β hereafter. In situ mRNA hybridization and LTα ELISA showed strong and sustained LTα and LTβ expression in skeletal muscle (quadriceps femoris) of 3-, 6- and 10-month-old HSA-LTα/β transgenic mice (Fig. 2C and D). Histology revealed numerous foci of endomysial, often perivascular inflammatory infiltrates composed of CD4⁺ and CD8⁺ T cells as well as frequent B220⁺ B cells and CD68⁺ macrophages. MHCI showed focal sarcolemmal upregulation. Especially in older mice (9 months), we observed endomysial fibrosis and the proliferation of fat tissue (Fig. 2E). Quantitative PCR (qPCR) and RNA-Seq analysis of quadriceps femoris muscle confirmed the strong and sustained upregulation of LTα/β. We also observed strong and significant upregulation of several pro-inflammatory cytokines and chemokines, including LT target genes, such as Ccl19 and Cxcl10, in line with our previous observations in human IIMs (Fig. 2F and Supplementary material). LTα/β upregulation was restricted to the skeletal muscle compared with other organs in HSA-LTα/β Line #19 (Fig. 2G). No obvious difference in pathology was observed between different muscles (Fig. 2H).

Figure 2.

Generation, histology and chemokine and cytokine expression of HSA-LTα/β transgenic mice. (A) Schematic drawing of the transgenic constructs of the coding sequences (CDS) of lymphotoxin (LT)α and LTβ both being cloned downstream of the human skeletal muscle actin (HSA) promoter (−2000 to +239) followed by an SV40 poly A site. (B) Four founder mice carrying both transgenes, HSA-LTα and HSA-LTβ, and transmitting them through the germline were obtained (Lines #8, #19, #22, #44). (C and D) LTα and LTβ transgene expression was detected in quadriceps muscle fibres of HSA-LTα/β #19 mice by RNA in situ hybridization of paraffin sections (brown signal) in the muscle fibres at 3, 6 and 10 months of age and also in inflammatory infiltrates in skeletal muscle, shown at 10 months of age (C). LTα ELISA with skeletal muscle tissue homogenate from indicated time points and genotypes. Data are shown as nanogram LTα3/LTα1β2/LTα2β1 protein per miligram total protein. Statistical significance was tested using the two-tailed Student's t-test. (E) Histologically, HSA-LTα/β transgenic mice show endomysial inflammatory infiltrates and occasional muscle fibre necrosis. Especially at the age of 9 months, there is fibrosis and partial replacement by fatty tissue within the endomysium. Fibre size variation is also increased in transgenic mice [haematoxylin and eosin (H&E) and trichrome]. In addition to the normal MHCI localization on endomysial capillaries seen in wild-type mice, transgenic mice display focal sarcolemmal major histocompatibility complex (MHC)I upregulation (red signal). The inflammatory infiltrates are mostly composed of B220⁺ B cells and CD68⁺ macrophages and—to a lesser extent—of CD4⁺ and CD8⁺ T cells (brown signals). (F) Heat map of chemo- and cytokine expression in HSA-LTα/β transgenic mice determined by quantitative PCR of quadriceps femoris muscle tissue at 3, 6 and 10 months of age. P-values were determined using two-tailed Mann–Whitney tests. In the case of significant differences between wild-type and transgenic mice, the P-values are displayed between the group of wild-type and transgenic line of the respective age group. (G) Quantitative PCR determining LTα and LTβ expression in different organs of HSA-LTα/β transgenic mice. P-values were determined using the ANOVA test with Šídák's multiple comparison test, significant differences compared with wild-type are displayed. (H) H&E-stained sections of three different muscles from 9-month-old HSA-LTα/β mice.

Quantification of muscle fibre size distribution revealed considerable widening of the muscle fibre calibre spectrum with increased atrophic and hypertrophic muscle fibres in HSA-LTα/β mice compared with controls at 3 and 6 months as well as a significantly larger fraction of atrophic fibres at 10 months of age (Fig. 3A), resembling observations in human IIM (Fig. 1A). This was accompanied by a reduction in the overall body weight of both male and female mice (Fig. 3B). The quadriceps femoris, gastrocnemius and triceps brachii muscles showed a significantly lower weight in HSA-LTα/β transgenic mice compared with controls (Fig. 3C). Relative muscle weight loss was higher in the lower extremity (51 ± 12% for the quadriceps, 46 ± 10% for the gastrocnemius) compared with the upper extremity (36 ± 13% for the triceps brachii, at 10 months of age). Accordingly, motor functions, i.e. grip strength (Fig. 3D) and performance in the hanging wire test (Fig. 3E) were significantly impaired in HSA-LTα/β transgenic mice compared with wild-type. These data demonstrate that skeletal muscle-specific expression of LTα and LTβ leads to chronic myositis in mice that shares histological, molecular and clinical features of human IIM.

Figure 3.

Muscle fibre size distribution, muscle hypotrophy and motor impairment in HSA-LTα/β transgenic mice. (A) Muscle fibre diameters were examined morphometrically using cross sections of the quadriceps femoris muscle [at 3 months, 596 fibres of n = 4 wild-type (wt) and 840 fibres of n = 4 human skeletal muscle actin (HSA)-lymphotoxin (LT)α/β transgenic mice (LT); at 6 months: 981 fibres of n = 4 wt mice and 1075 fibres of n = 5 LT mice; at 10 months: 1266 fibres of n = 5 wt mice and 1732 fibres of n = 5 LT mice]. For statistical analysis, we used the chi-squared test and grouped fibres into atrophic (<50 µm), normal (50–109 µm) and hypertrophic (>109 µm). Fibre size distribution was broader in HSA-LTα/β transgenic mice at 3 and 6 months, with both, more atrophic and hypertrophic fibres; there were more atrophic fibres at 10 months of age. (B) Body weight was determined every 4 weeks and is displayed separately for female and male mice. (C) Weight of quadriceps femoris, gastrocnemius and triceps brachii muscles. Two-sided unpaired Student's t-test was used for statistical analysis. Differences were significant for all analysed time points for all muscles; ****P < 0.0001. Motor performance of wild-type and HSA-LTα/β transgenic mice was determined by the grip strength test (D) and by the hanging wire test (E). Groups were compared using the two-sided unpaired Student's t-test. P-values are shown in case of a significant difference.

Chronic myositis induces organelle stress, alters autophagy and mitochondrial gene expression

Skeletal muscles of HSA-LTα/β transgenic mice significantly upregulate genes associated with ER stress/UPR (Fig. 4A). Upon accumulation of unfolded proteins, binding immunoglobulin protein (BiP) dissociates from the ER intraluminal domains of three ER transmembrane receptors: (i) type I transmembrane protein inositol requiring 1 (IRE1α); (ii) protein kinase R (PKR)-like endoplasmic reticulum kinase (PERK), which phosphorylates eukaryotic initiation factor 2α (eIF2α); and (iii) activating transcription factor 6 (ATF6) to activate pathways, which in turn reduce accumulated unfolded proteins by increasing their folding, increasing their degradation and/or decreasing the transcription and/or translation.⁷⁶

Figure 4.

Organelle stress and signs of altered autophagolysosomal pathways in chronic myositis. (A–C) Gene expression in human skeletal muscle actin (HSA)-lymphotoxin (LT)α/β transgenic mice determined by quantitative PCR of quadriceps femoris muscle tissue at 3, 6 and 10 months of age. Expression of genes related to endoplasmic reticulum (ER) stress, autophagolysosomal pathway, heat shock response (A) as well as oxidative stress-related genes (B) are displayed in heat maps. P-values were determined using two-tailed Mann–Whitney tests. In the case of significant differences between wild-type and transgenic mice, the P-values are displayed between the group of wild-type and transgenic line of the respective age group. To visualize the gene expression changes of the marked upregulated autophagy-related gene Atg5 at 6 months of age, we also display it together with another autophagy-related gene Becn1 and lysosomal protease Ctsb in transgenic compared with wild-type mice in C. Both autophagy-related genes are upregulated at 3 months (Atg5) and 6 months (Atg5, Becn1, see A), but downregulated (Atg5) along with lysosomal Ctsb at 10 months. (D) Downregulation of ATG5, BECN1 and HSPB1 as well as upregulation of CTSB in human inclusion body myositis (IBM) muscle tissue compared with controls determined by analysing published RNA-Seq data. (E and F) We determined amounts of LC3-I and LC3-II proteins in the muscle of 6- and 10-month-old HSA-LTα/β transgenic compared with wild-type mice by western blot (E) and observed a significant reduction of total LC3 and LC3-II/ α-tubulin ratio in transgenic mice at 6 months and a significant increase in total LC3 and a decrease in the LC3II/ LC3-I ratio at 10 months. Quantification of western blot bands (F); values for total LC3 and LC3II are relative to wild-type values; n = 5 biological replicates at 10 months; one to two technical replicates from three to four biological replicates at 6 months. P-values were determined using two-tailed Student's t-tests. Dotted line shows where membranes were cut; solid line shows cropping for representation in this figure; full size membranes are shown in the Supplementary Figs 7 and 8. (G) In line with disturbed proteostasis, we occasionally observed focal accumulation of ubiquitin in muscle fibres of HSA-LTα/β transgenic mice at 6 and 10 months of age (immunohistochemistry). Distribution of LAMP2⁺ lysosomes was even in wild-type mice, but uneven with less, but larger lysosomes in HSA-LTα/β transgenic mice (immunohistochemistry). Scale bar = 50 µm. (H) Ultrastructural examination at 9 months of age did not show any IBM-like inclusions in HSA-LTα/β transgenic mice, but mitochondria showed slight swelling. The chi-squared test was used for statistical testing.

While we did not detect an increase of spliced Xbp1, a surrogate marker of IRE1 activation, we found transcriptional evidence of activation of the other UPR pathways. In the second pathway, PERK phosphorylates eIF2α, which in turn reduces protein synthesis of most mRNAs except Atf4 and Chop whose translation increases. ATF4 then induces expression of the transcription factors Chop and Atf3.⁷⁶ CHOP, in turn, increases transcription of Ero1l.⁷⁷ We observed significant upregulation of Ero1l at 3 months, of Atf3 at 3 and 6 months and of Chop at 10 months (Fig. 4A). ATF3 and ATF4 both bind to the promotor region of Gadd34, which was significantly upregulated at 3 months of age. In the third pathway, ATF6 translocates to the Golgi, where it is cleaved and activated; cleaved ATF6 is then transported to the nucleus, where it activates transcription of target genes such as Grp78/BiP.⁷⁶ We observed upregulation of Grp78 at 6 months, but downregulation at 10 months (Fig. 4A), suggesting a transient induction of the ATF6 pathway (Supplementary Fig. 2).

The significant upregulation of several glutathione peroxidases: cytosolic Gpx1, plasma Gpx3, phospholipid hydroperoxide Gpx4 and Gpx8 (Fig. 4B) suggested induction of oxidative stress in HSA-LTα/β mice. Gpx family members are antioxidants that detoxify hydroperoxides by reducing them to the corresponding alcohols by means of glutathione (GSH). However, specific functions of individual Gpx have been shown. Gpx4 regulates inflammatory signalling: NF-κB activation by IL-1, reduced leukotriene and prostanoid biosynthesis, prevents cyclooxygenase-2 (COX-2) expression. Superoxide dismutase 3 (SOD3) is a known regulator of signalling pathways during inflammation and was found to be upregulated in HSA-LTα/β mice (Fig. 4B). Nrf2, a transcription factor that regulates expression of antioxidant proteins that protect against oxidative damage,⁷⁸ was initially down- (6 months) and later upregulated (10 months; Fig. 4B).

In addition, we observed alterations in the expression of genes related to autophagolysosomal pathways (Fig. 4A, C and D). Autophagy related 5 (ATG5) is essential for the extension of the phagophoric membrane in the autophagic vesicle. It forms a complex with ATG12 and ATG16LL1, which is necessary for conjugation of LC3-I to phosphatidylethanolamine (PE) to form LC3-II. Atg5 was significantly and strongly upregulated at 3 and 6 months in HSA-LTα/β mice. Likewise beclin 1 (Becn1), another important regulator of autophagy was significantly increased at 6 months. Interestingly, at 10 months, we observed downregulation of Atg5 and lysosomal cathepsin B (Ctsb), indicative of a disturbed autophagolysosomal pathway in HSA-LTα/β mice. Similarly, analysis of published human IBM RNA-Seq data^62,63 revealed downregulation of autophagy-related genes ATG5 and BECN1, CTSB was upregulated and HSPB1 was downregulated in IBM (Fig. 4D). While altered expression of BECN1 and CTSB was also observed in other IIMs, the regulation of ATG5 appears to be specific for IBM, and HSPB1 was upregulated in other IIMs (Supplementary Fig. 1).⁶³

The quantity of LC3-II protein has been suggested as a suitable indicator of autophagic activity.⁷⁹ We therefore measured LC3-II in skeletal muscle of HSA-LTα/β and wild-type mice by western blot analysis (Fig. 4E and F) at 6 and 10 months of age.

At 6 months, HSA-LTα/β muscle showed significantly reduced levels of total LC3 and LC3-I, while the LC3-II/LC3-I ratio remained unaltered, i.e. LC3-I was reduced along with LC3-II (Fig. 4E and F), as reported after prolonged starvation.⁷⁹ Combined with the transcriptional upregulation of autophagy-related genes (Atg5 and Becn1) described above, this suggests prolonged/sustained autophagic activity in middle-aged HSA-LTα/β. The LC3/LC3-I accumulation and decreased LC3-II/LC3-I ratio in HSA-LTα/β muscle at 10 months combined with the transcriptional downregulation of autophagy- and lysosmal-related gene expression (Atg5 and Ctsb) suggest an impaired autophagic flux (Fig. 4E and F and Supplementary Figs 2 and 3), most likely reflecting exhaustion due to overload. These changes were similar, but less pronounced compared with ‘controls’ with skeletal muscle-specific autophagy disruption by depleting Atg5—by combining the creatine kinase, M-type (Ckmm) promoter driven Cre expressing transgene (Ckmm-Cre) with the floxed Atg5 allele (Atg5^fl/fl), later termed HSA-LT;CreAtg5 (see later).

In line with disturbed proteostasis and autophagolysosomal pathway, we detected occasional accumulation of ubiquitinated proteins in some muscle fibres of HSA-LTα/β mice compared with controls and altered distribution of LAMP2⁺ lysosomes (Fig. 4G), similar to observations in humans.⁸⁰

To understand the cellular and molecular pathophysiological mechanisms that lead to the observed shifts with ageing, we performed RNA-Seq of wild-type and HSA-LTα/β muscles, both at 6 and 9 months. Beside the strong upregulation of pro-inflammatory genes in HSA-LTα/β, including Tbx21, an established marker of highly differentiated T cells specific for IBM,⁷¹ along with Il2, another marker of Tc1 cytotoxic T cells, but not of DM signature genes.⁷¹ Gene set enrichment analysis (GSEA) revealed profound upregulation of mitochondria-associated genes with ageing in wild-type muscle, which was not observed in HSA-LTα/β mice. Instead, HSA-LTα/β mice showed significantly reduced mitochondria-associated gene expression compared with wild-type (Supplementary material). Furthermore, in line with previous observations,⁸¹ UPR-related gene expression was dysregulated in ageing wild-type muscle. Hence, age-related changes in proteostasis- and mitochondria-associated gene expression might contribute to the apparent shifts in transcriptional alterations in HSA-LTα/β muscle. Immunoproteasome subunits Psmb8–10, but not subunits 1–7, were upregulated in HSA-LTα/β muscle (Supplementary material), similar to the specific upregulation of PSMB9 and 10, previously detected in human IBM compared with other inflammatory myopathies.⁷¹ GSEA further revealed downregulation of ubiquitination- and mammalian target of rapamycin (mTOR) signalling-associated genes in HSA- LTα/β muscle, pointing to possible mechanisms that could alter the autophagolysosomal pathway activity.

Based on the failure of HSA-LTα/β muscle to upregulate mitochondria-associated genes with ageing, the transcriptional signs of oxidative stress in HSA-LTα/β mice and in light of the prominent mitochondrial pathology in human IIM, especially in IBM, we examined the ultrastructure of mitochondria in these mice. We found considerable mitochondrial swelling, but no definitive signs of major chronic mitochondrial pathology such as paracrystalline inclusions, major structural abnormalities of cristae including concentric forms or abnormal mitophagy (Fig. 4H) at this point.

Together, these results suggest that chronic myositis in HSA-LTα/β mice is associated with unspecific alterations of proteostasis and autophagy. However, there are no major, potentially self-perpetuating and self-reinforcing autophagy disturbances or other major degenerative pathologies such as abnormal protein inclusions.

Genetically impairing autophagy in lymphotoxin-induced myositis adds IBM-like inclusions to the phenotype

We hypothesized that in LT-driven chronic myositis, temporary induction of the autophagic flux prevents the formation of IBM-like protein inclusions. In older HSA-LTα/β mice and in human IBM muscle biopsy tissue, Atg5/ATG5 was downregulated in line with impaired autophagy reported in human IBM.^82,83 Therefore, we aimed to speed up and enhance its downregulation and at the same time determine the role of autophagy in chronic inflammation by specifically depleting Atg5 in the skeletal muscle of HSA-LTα/β transgenic mice. To do so, we combined the Ckmm-Cre transgene with the floxed Atg5 allele (Atg5^fl/fl; Fig. 5A). Double genetically modified mice (HSA-LTα/β⁺ Ckmm-Cre⁺ Atg5^fl/fl), hereafter termed HSA-LT;CreAtg5 mice displayed muscular atrophy compared with mice lacking Atg5 only (Ckmm-Cre⁺ Atg5^fl/fl), hereafter termed CreAtg5. Muscle volumes were determined by muscle MRI (Fig. 5B and C). HSA-LT;CreAtg5 mice showed histological signs of inflammation similar to HSA-LTα/β mice, including endomysial inflammatory infiltrates and MHCI upregulation (Fig. 5D–F), while CreAtg5 mice did not show inflammation. HSA-LT;CreAtg5 mice were weaker, i.e. had reduced grip strength and shorter hanging time, compared with CreAtg5 controls (Fig. 5G and H). In addition, HSA-LT;CreAtg5 mice showed disrupted distribution of LAMP2⁺ lysosomes, and displayed numerous ubiquitin- and p62-positive and occasional phospho-TDP43-positive inclusions (Fig. 6A). Electron microscopy of skeletal muscle of HSA-LT;CreAtg5 mice revealed partial disintegration of sarcomeres and abundant accumulation of granular and fibrillar material often resembling tubulofilamentous inclusions characteristic of human IBM. Mitochondria were focally increased and often showed transitions in abnormal osmiophilic, myelin-like autophagic material, indicative of abnormal mitophagy (Fig. 6B). Although we did not detect mitochondrial DNA deletions in ‘bulk’ long-read sequencing (Fig. 6C), combined COX and succinate dehydrogenase (SDH) enzyme histochemistry of the quadriceps muscle revealed mitochondrial abnormalities in HSA-LTα/β and especially in HSA-LT;CreAtg5 mice that were not observed in wild-type animals. COX-positive mitochondria showed a slightly more irregular distribution in HSA-LTα/β compared with wild-type (Fig. 6D). The distribution of COX-positive mitochondria was even more irregular in HSA-LT;CreAtg5 and single muscle fibres showed mosaicisms, i.e. focal subpopulations of COX-negative mitochondria next to COX-positive mitochondria in the same fibre. Irregularities were found in all and COX mosaicism was found in six of eight HSA-LT;CreAtg5 mice. This mosaicism was absent in the HSA-LTα/β and wild-type mice tested (Fig. 6D). Hence mitochondrial pathology is present, but minimal in our model.

Figure 5.

Behavioural and histological consequences of autophagy depletion in lymphotoxin (LT)-induced chronic myositis. (A) Breeding scheme to obtain human skeletal muscle actin (HSA)-LTα/β⁺ Ckmm-Cre⁺ Atg5^fl/fl (HSA-LT;CreAtg5) mice. (B and C) Muscle volume was determined by MRI at 3 and 6 months of age of male and female HSA-LT;CreAtg5 compared with Ckmm-Cre⁺ Atg5^fl/fl (CreAtg5) mice. Representative fast low angle shot MRI (FLASH) images and 3D reconstructions of calf muscles are shown along with volumetric quantification. Muscle atrophy was observed in HSA-LT;CreAtg5 compared with CreAtg5 mice. Two-sided unpaired Student's t-test was used to determine P-values (****P < 0.0001). (D–F) Histological and immunohistochemical analysis for inflammatory markers revealed chronic myositis in HSA-LT;CreAtg5, but not in CreAtg5 mice, characterized by endomysial infiltrates composed of CD4⁺ and CD8⁺ T cells, B220⁺ B cells and CD68⁺ macrophages (brown signals) as previously observed in HSA-LTα/β transgenic mice. Sarcolemmal upregulation of major histocompatibility complex I (MHCI) was detected in HSA-LT;CreAtg5, but MHCI was either not (one of four) or expressed weakly/only on single fibres (three of four) in CreAtg5 mice (red signal). Quantification of most affected fields of view is shown in F. P-values were determined using the ANOVA test with Šídák's multiple comparison test. (G and H) Motor performance of CreAtg5 and HSA-LT;CreAtg5 mice was determined using the grip strength test (G) and the hanging wire test (H). Groups were compared using the two-sided unpaired Student's t-test. P-values are shown.

Figure 6.

Consequences of autophagy depletion in lymphotoxin (LT)-induced chronic myositis on proteostasis, ultrastructure and mitochondria. (A) Immunohistochemistry for ubiquitin, LAMP2, p62 and phospho-TDP43 (brown signals). Human skeletal muscle actin (HSA)-LT;CreAtg5 mice show numerous ubiquitin- and p62-positive inclusions in all muscle fibres. Ubiquitin-positive inclusions were not observed in wild-type mice. One mouse showed little focal p62-positivity (n = 6 of 6 HSA-LT;CreAtg5, but only focally in n = 1 of 6 wild-type; P = 0.0152, Fisher's exact test). The distribution of LAMP2⁺ lysosomes in HSA-LT;CreAtg5 compared with wild-type with less, but often larger lysosomes in most fibres and increased lysosomal density in other fibres. Occasionally, phospho-TDP43-positive inclusions are observed in individual fibres of HSA-LT;CreAtg5 mice (n = 6 of 6 HSA-LT;CreAtg5, Fisher's exact test when compared with wild-type: P = 0.0152). (B) Electron microscopy at 6 months of age revealed disintegration of sarcomeric structure/myofibrils (i), accumulation of granular and fibrillar material (between black arrows in ii and iii, and black arrowheads in vii), probably corresponding to myofibrillar fragments, often containing inclusions resembling tubulofilamentous inclusions characteristic for human inclusion body myositis (IBM; between black arrows in iv and vii), diameter measured in iv: 106 tubulofilaments showed a mean diameter of 16.9 ± 2.3 nm in n = 3 HSA-LT;CreAtg5 mice. Subsarcolemmal accumulation of mitochondria (v) next to accumulation of granular material (black arrowheads in v) with foci of mitochondria in different stages of abnormal mitophagy (vi) with some mitochondria showing almost normal structure of cristae (black arrows in vi) and those with cristae remnants in double membranes characteristic for autophagosomes (abnormal mitophagy, white arrowheads in vi). Focal deposits of abnormal myelin-like phospholipid are frequently seen (white arrowheads in vii). Pyknotic, abnormally invaginated myonucleus (viii). Scale bars = 500 nm. (C) Long-read nanopore sequencing of PCR-amplified mitochondrial DNA, showing no obvious deletions in any of the genotypes tested. Coverage is displayed in grey. Single nucleotide substitutions are shown in green (adenine), orange (guanine), blue (cytosine) and red (thymine). (D) Combined cytochrome c oxidase (COX, brown signal) and succinate dehydrogenase (SDH, blue signal) enzyme histochemistry. Even distribution of COX-positive mitochondria in wild-type (i) with higher density in type 1 fibres (brown signal). Focal irregularities in the distribution of COX-positive mitochondria in HSA-LTα/β (ii), more pronounced in HSA-LT;CreAtg5 mice that also display single fibres with mosaic COX-positive (brown) and COX-negative mitochondria (blue, due to preserved SDH activity, black arrowheads, iii and iv, with enlarged images of these fibres to the right of iii and iv). Several fibres also focally lack both, COX and SDH activity (black arrows).

The upregulation of chemokines and cytokines at 6 months of age in HSA-LT;CreAtg5 mice was similar to that observed in HSA-LTα/β mice (Fig. 7A). In line with our expectation, we observed transcriptional signs of disturbed proteostasis in skeletal muscle with Atg5 depletion alone (CreAtg5; Fig. 6B), including increased expression of UPR-associated genes like Grp78 and Chop (Fig. 7B). Notably, we observed cytokine and chemokine upregulation, including LT/ NF-κB target genes Ccl5⁸⁴ and Cxcl10 (IP-10)^67,68 in CreAtg5 (Fig. 7A). This demonstrates that chronic inflammation not only leads to autophagy dysregulation, but also vice versa, autophagy impairment leads to pro-inflammatory signalling, i.e. activation of the LT/NF-κB signal transduction pathway and only caused sarcolemmal MHCI upregulation in some cases. This pro-inflammatory transcriptional milieu in CreAtg5 mice was, however, not sufficient to induce histological signs of inflammation such as endomysial lymphocytic infiltrates (Fig. 5D and E). Besides, we confirmed the upregulation of autophagy-related genes Atg5, Atg12, Becn1, observed Ctsb upregulation and oxidative stress-related upregulation of Gpx1 and Sod3 as well as downregulation of Ncf2 in HSA-LTα/β mice at 6 months in this independent experiment (Fig. 7, in line with previous results in Fig. 4A and B). Tumorous imaginal disc 1 (Tid1), also called DnaJ homolog subfamily A member 3 (DnaJA3) was the only gene specifically downregulated in HSA-LT;CreAtg5 mice (Fig. 7B). Tid1 is a multifunctional protein that acts as a co-chaperone of mtHsp70. Of note, Tid1 is a regulator of autophagy⁸⁵ and several signalling pathways including Wnt and tropomyosin receptor kinase (Trk). Interestingly, it also represses activity of NF-κB through interaction with inhibitor of kappa B (IkB).⁸⁶ In fact, Tid1 was shown to control various biological processes, including muscle energy homeostasis, development, myogenesis⁸⁷ and synapse formation by binding to muscle-specific kinase (MuSK),⁸⁸ mitochondrial fragmentation,⁸⁹ amyloid-β production by controlling beta-site amyloid precursor protein cleaving enzyme 1 (BACE1) levels and mediating amyloid-β induced reactive oxygen species (ROS) generation and neuronal cell death in Alzheimer's diseases⁹⁰—a disease associated with cellular protein inclusions similar to IBM. We also observed upregulation of App and crystallin alpha B (Cryab), previously observed in human myositis, especially in IBM.^91,92

Figure 7.

Transcriptional consequences of depleting autophagy in lymphotoxin (LT)-induced chronic myositis mice. (A–D) Heat maps of gene expression in wild-type, CreAtg5, human skeletal muscle actin(HSA)-LTα/β and HSA-LT;CreAtg5 mice determined by quantitative PCR of quadriceps femoris muscle tissue at 6 months of age. P-values were determined using Mann–Whitney U-test and are displayed above the group of mice in case of a significant difference after Bonferroni correction (P-values <0.016). Cytokine and chemokine expression is shown in A, endoplasmic reticulum (ER) stress and autophagolysosomal gene expression is displayed in B. (C) Genes related to muscle de- and regeneration and Alzheimer disease/neurodegeneration including App and Cryab. (D) Genes related to oxidative stress (gene functions overlap).

Muscle-specific RING finger protein-1 (MuRF1) is an E3 ubiquitin ligase in cardiac and skeletal muscles that is upregulated during skeletal muscle atrophy and mediates ubiquitin proteasome system-mediated degradation of sarcomeric muscle proteins^93-95 and may contribute to muscle fibre atrophy observed in HSA-LTα/β mice. Furthermore, TNF-related weak inducer of apoptosis (Tweak) was upregulated in CreAtg5, HSA-LTα/β and HSA-LT;CreAtg5 mice as previously described in human IBM muscle where TWEAK is supposed to reduce activation and differentiation of muscle precursor cells and may induce muscle atrophy.⁹⁶

IBM-like myositis in mice is resistant to anti-inflammatory treatments

In humans, anti-inflammatory treatments, including steroids, intravenous immunoglobulins (IVIG), anti-CD52 (alemtuzumab), have shown efficacy in PM and DM. Unfortunately, none of these approaches have shown any long-lasting therapeutic effects in IBM.^37,38,97-99 We tested the efficacy of these therapies, including prednisolone, and lymphocyte-depleting anti-Thy1.2 in our mouse model. To determine motor function in these mice, behavioural performance was continuously monitored in untreated and treated mice, starting at 90 days of age—before and during treatment from 120 to 160 days of age (Fig. 8A). None of these treatments improved motor performance in mice with chronic myositis (Fig. 8 and Supplementary Figs 4 and 5). To verify the effect of these treatments on inflammatory cells, we determined T cell densities in the liver of these mice and found a significant reduction after treatment (Fig. 8B and C). To detect milder effects on the skeletal muscle upon treatment, we analysed transcriptional profiles. We observed minor alterations following prednisolone (Fig. 8F), but no significant effects on gene expression in anti-Thy1.2-treated mice (Supplementary Fig. 6). This shows that once initiated, chronic myositis is resistant to anti-inflammatory treatment in HSA-LTα/β and HSA-LT;CreAtg5 mice.

Figure 8.

Inclusion body myositis (IBM)-like inflammatory myopathy in mice is resistant to anti-inflammatory treatments. (A) Timeline of treatment. Density of CD3⁺ T cells in the liver of untreated mice and after treatment; for anti-Thy1.2 (B) and prednisolone (C) treatment. P-values were determined using two-tailed Students t-tests. Grip strength of mice at Day 150. The results show genotype-dependent differences, but no treatment effect after prednisolone (D) and anti-Thy1.2 (E) treatment. (F) The heat map shows expression examined by quantitative PCR (qPCR) relative to the respective untreated controls (treated compared with untreated wild-type, human skeletal muscle actin (HSA)-lymphotoxin (LT)α/β and HSA-LT;CreAtg5, respectively)—here in the case of prednisolone treatment. Except for a significant downregulation of Cxcl10, Ncf2 and Gpx2 following prednisolone treatment of HSA-LTα/β mice compared with untreated HSA-LTα/β and non-significant trends for some other genes, there was a counterintuitive upregulation of Ccl5 in treated compared with untreated HSA-LT;CreAtg5 mice, but otherwise stable gene expression. n.d. = not determined. Mann–Whitney U-test was performed as a statistical test since some values were not normally distributed. P-values are only displayed in the case of statistical significance between untreated and treated mice. Behavioural phenotypes and gene expression patterns following these treatments were stable (Supplementary material).

Discussion

Lymphotoxin-driven inflammation with inclusions in mice resembles several but not all features of human IBM

Our mouse model of chronic inflammatory and vacuolar myopathy recapitulates some of the key features seen in human IBM, including lymphocytic endomysial inflammation and chronic necrotizing and vacuolar myopathy with prominent alterations of proteostasis and autophagy associated with tubulofilamentous inclusions and abnormal autophagic vacuoles. However, inflammatory myopathy in HSA-LT;CreAtg5 mice also showed features that differ from typical human IBM cases. First, B cells outnumbered T cells in the lymphocytic infiltrates in mouse muscle, in contrast to mostly T cells, especially CD8⁺ cytotoxic T cells in human IBM.^6,7 Still, B-cell-rich inflammation has been described in rare cases of human IBM,¹⁰⁰ and terminally differentiated B cells—plasma cells—as well as immunoglobulin gene transcripts are well established adjacent to scattered B cells in IBM as well as in PM,^101,102 suggesting that they contribute to pathogenesis. Furthermore, the B-cell marker B220 that was applied in our study also labels plasma cell precursors, the plasmablasts.¹⁰³ In our model, B cells likely infiltrate the muscle because LT induces Cxcl13 expression (Fig. 7A) that in turn is known to attract B cells.¹⁰⁴ Of note, CXCL13 is also upregulated in human IBM (Fig. 1E) and a recent paper using single nucleus RNA-Seq combined with spatial transcriptomics revealed a B-cell-rich inflammatory niche in human IBM.⁸²

Mitochondrial abnormalities are frequently observed in IBM in humans.^7,16,17 These findings include COX-negative fibres, mitochondrial DNA deletions, paracrystalline inclusions and structural abnormalities of cristae such as their concentric arrangement, as well as signs of dysfunctional mitophagy.^105,106 Whereas abnormal mitophagy associated with transcriptional signs of mitochondrial oxidative stress are prominent features in HSA-LTα/β and HSA-LT;CreAtg5 mice, anomalies of cristae architecture are minimal (Figs 4B, H and 6B). Paracrystalline inclusions have not been observed. However, we are not aware of paracrystalline inclusions and other major structural mitochondrial changes consistently being described in mouse models of mitochondrial myopathies. For example, Polg mutant mice ultrastructurally display dissolution of mitochondrial cristae and swelling in cardiac muscle at the ultrastructural level.¹⁰⁷ These findings are similar to the widened spaces between mitochondrial cristae observed by us in HSA-LTα/β mouse skeletal muscle fibres (Fig. 4H). Of note, paracrystalline inclusions are a feature of human patients with POLG mutations,¹⁰⁸ but were not reported in Polg mutant mice. Hence, mice might be protected from mitochondrial paracrystalline inclusion formation, possibly due to a higher volume of mitochondria in murine compared with human skeletal muscle.¹⁰⁹ Despite the lack of paracrystalline inclusions, and mitochondrial DNA deletions in bulk sequencing, mitochondria are involved in chronic inflammatory and vacuolar myopathy in our mice. This is demonstrated by the failure of the inflamed muscle in our mouse model to upregulate mitochondria-associated gene expression with ageing based on the RNA-Seq data (Supplementary material) and by the irregular distribution of COX-positive mitochondria in transgenic mice and especially by the single mosaic fibres with COX-negative mitochondria in otherwise COX-positive fibres in HSA-LT;CreAtg5 mice (Fig. 6D).

In this context, it is interesting to note that PM with mitochondrial pathology (PM-Mito¹¹⁰), more recently referred to as inflammatory myopathy with mitochondrial pathology in muscle (IM-Mito), may be considered a precursor or early form of IBM with little or no pTDP-43 aggregation,¹¹¹ implying that mitochondrial alterations could be an early event in the pathophysiological cascade leading to the full-blown picture of IBM.^111,112 Alterations in autophagic degradation pathways have been found in both, IBM and PM with mitochondrial alterations, suggesting a link between mitochondrial and autophagy dysregulation in myositis in humans.¹¹¹ On a molecular level, there are possible links between ER stress, oxidative stress and autophagy impairment in HSA-LTα/β mice. UPR-induced Ero1l observed in HSA-LTα/β mice is a H₂O₂-producing enzyme that has been shown to increase oxidative stress upon ER stress.¹¹³ In mesenchymal stem cells (MSCs), depletion of autophagic proteins such as LC3B or BECN1 increased susceptibility of cells to oxidative stress.¹¹⁴ However, how ER stress pathways and mitochondria intersect in myositis pathogenesis remains to be determined.¹¹⁵

Interdependence of inflammation and impaired proteostasis

Our expression data describe an upregulation of LT and LTβ receptor target genes in human inflammatory myopathies, confirming and extending previous observations.^49,50,70 Targeted expression of LTs in muscle fibres of our transgenic HSA-LTα/β mice induced chronic inflammatory myopathy that shares characteristic clinical and pathological features with human myositis, including necrotizing myopathy with muscular atrophy and weakness, lymphomonocytic endomysial infiltrates, upregulation of chemokine and cytokine transcription and of sarcolemmal MHCI and endomysial fibrosis. LT-driven chronic lymphomonocytic myositis in these mice alters autophagolysosomal pathways, induces ER stress and leads to focal accumulation of ubiquitin-positive inclusions as well as oxidative stress and mild mitochondrial alterations, in line with observations in human IBM.¹¹⁶ This suggests that activation of the NF-κB pathway and inflammation perturb proteostasis and induce a mild, unspecific degenerative pathology. Chronic myositis did not suffice to induce the full spectrum of IBM pathology. We hypothesized that activation of the autophagolysosomal flux partially compensates and impedes further dysregulation of proteostasis and protein aggregation in chronic myositis in middle-aged HSA-LTα/β mice (6 months). To speed-up and enhance autophagy impairment, which we only observed later in older HSA-LTα/β mice (10 months), we genetically impaired autophagy in mice with LT-driven myositis and then observed the full spectrum of IBM-associated pathology—a myopathic phenotype with inflammation and IBM-like degeneration (Figs 5 and 6). If the additional autophagy impairment is actually a requirement, i.e. the straw that breaks the camel's back or if this occurs when chronic myositis persists longer still remains unknown. Autophagy impairment alone did not induce myositis. Although there was significant upregulation of some chemokine and cytokines, suggesting a pro-inflammatory molecular milieu, and signs of ER stress, we did not observe endomysial lymphomonocytic cell infiltrates or consistent MHCI upregulation. This suggests that autophagy impairment and ER stress can at most contribute to inflammation but not independently induce the complete picture of myositis. Nevertheless, our data underscore that inflammation and autophagy impairment act synergistically and jointly contribute to potentially self-perpetuating and self-reinforcing disturbances, eventually resulting in IBM-like abnormal protein inclusions and disturbed mitophagy.

Molecular links between inflammation and degenerative features

Several previous observations indicate that inflammation and degenerative features in IBM are intertwined. MHCI, which is upregulated on the sarcolemma of muscle fibres, was shown to induce ER stress, which in turn leads to a release of chemokine and cytokines from C2C12 myoblasts.¹¹⁷ ER stress is well known to activate the ER overload response (EOR) that involves the upregulation of the NF-κB pathway and modulation of the inflammatory response. In fact, ER stress/EOR are not exclusive to IBM, but also observed in PM and DM.⁷⁰ Different pathways regulating proteostasis are linked, including ER stress and autophagy. In cultured human muscle fibres, ER stress induction leads to increased levels of LC3-II and increased maturation of autophagosomes, but decreased activities of lysosomal cathepsins D and B.⁸

Our results support these observations and point to previously unrecognized crosstalk pathways. In mice with impaired autophagy in skeletal muscle, we observed upregulated ER stress-related genes and chemokine and cytokines including NF-κB target genes, probably through the EOR. Activation of the NF-κB pathway in the muscles of HSA-LTα/β mice in turn induced ER stress. We further observed strong upregulation of Tribbles homolog 3 (Trib3), a regulator of ER stress-induced apoptosis and NF-κB signalling¹¹⁸ in CreAtg5 mice, possibly linking ER stress with NF-κB signalling and inflammation.

In addition, we found dysregulated expression of further genes that may be involved in these interactions. We observed significant upregulation of superoxide dismutase Sod3 in HSA-LTα/β and HSA-LT;CreAtg5 mice (Figs 4 and 7). All SODs possess antioxidative properties. Within this family, SOD3 is secreted into the extracellular matrix, where it ameliorates oxidative stress- and inflammation-induced tissue damage. The anti-inflammatory function of SOD3 is in part exerted by downregulating NF-κB signalling.¹¹⁹ Inflammatory cytokines and influenza A virus infection have been shown to enhance SOD3 expression.^120,121 Overexpression of SOD3, in turn, was found to enhance autophagy.¹²² SOD3 may therefore link and regulate inflammation, oxidative stress and autophagy in our mouse model and potentially exert a protective function.

Subunits Psmb9 and 10 of the immunoproteasome that breaks down intracellular proteins for antigen presentation by MHCII,¹²³ were upregulated in HSA-LTα/β mice similar to human IBM.⁷¹ This potentially links inflammation, proteostasis regulation and mitochondrial dysfunction as PSMB9 is induced to regulate cellular proteostasis upon mitochondrial dysfunction.¹²⁴

Although we noted at least a temporary upregulation of several genes involved in UPR and oxidative stress, including Grp78, Chop, Atf3, Ero1l and Gpx genes (Figs 4 and 7), we observed early and strong downregulation of Hsp27/Hspb1 in HSA-LTα/β mice (Fig. 4). HSP27 has been shown to act as a chaperone and antioxidative protein and ameliorates fibrotic processes in kidney disease.¹²⁵ Its downregulation may perturb proteostasis and induce endomysial fibrosis in HSA-LTα/β mice. Similarly, downregulation of XBP1 promotes fibrosis in injured kidneys¹²⁶ where overwhelming ER stress results in selective Xbp1u and Xbp1s downregulation,¹²⁷ possibly contributing to endomysial fibrosis in HSA-LTα/β mice.

Failure of anti-inflammatory treatments and self-perpetuation of chronic inflammatory and vacuolar myopathy

The resistance to therapeutic T-cell depletion by prednisolone and anti-Thy1.2 observed in our murine model is in line with the lack of a consistent, at the most, subtle clinical improvement following anti-inflammatory treatments⁹⁷ as well as the absence of significant differences in molecular parameters in alemtuzumab/anti-CD52 antibody-treated IBM patients.³⁸ Recently, the further characterization of T cells in human IBM muscle revealed an increased number of highly differentiated CD8⁺ cytotoxic T cells expressing killer cell lectin-like receptor G1 (KLRG1).⁷¹ These CD8⁺ KLRG1⁺ T cells may be refractory to immunotherapy and may explain the resistance of IBM patients to anti-inflammatory therapies.^71,128 However, the persistent loss of TDP-43 function and rimmed vacuoles after T-cell depletion in mice with xenografts of human IBM tissue suggested that T-cell depletion cannot alter muscle degenerative pathology in IBM,¹²⁹ even when CD8⁺ KLRG1⁺ T cells are removed. This suggests that inflammation is either secondary or required to initiate pathology, however, not essential for disease progression and self-perpetuation. The selective, antibody-mediated depletion of KLRG⁺ T cells in an ongoing human clinical trial in IBM will provide further insight (ClinicalTrials.gov ID NCT05721573).

An alternative explanation for treatment failure in chronic inflammatory and vacuolar myopathy in our mouse model and human IBM is impaired muscle regeneration after degeneration. In neuromuscular diseases, satellite cells, the muscle stem cells proliferate and differentiate for repair and regeneration.¹³⁰ Satellite cells may be impaired or exhausted in neuromuscular diseases. Paired box 7 (PAX7) is expressed in satellite cells and is required for their regenerative function in adult skeletal muscle.¹³¹ The significant reduction of Pax7 in CreAtg5, and even more in HSA-LTα/β and HSA-LT;CreAtg5 mice suggests fewer satellite cell numbers and therefore decreased regenerative potential. Our finding may appear to contrast the increase of Pax7⁺ satellite cells previously described in human PM/DM and IBM.^132,133 However, they are in line with evidence of reduced satellite cell plasticity shown in long-term IBM patients.¹³⁴ In addition, the resistance of IBM patients to the myostatin blocker bimagrumab, which increases the volume of normal muscle,¹³⁵ argues for a defective regenerative response in IBM. The lack of response to treatment could also be due to hitherto unknown factors that we also do not know in human patients with IBM.

In the future, our mouse model of chronic inflammatory and vacuolar myopathy can be used to test novel therapeutic strategies including combinatorial strategies of anti-inflammatory treatment and molecules targeting protein homeostasis.

Data availability

The authors confirm that the data supporting the findings of this study are available within the article and its Supplementary material.

Funding

M.H. was supported by a grant of the Helmholtz-Gemeinschaft (Future topic "Immunology and inflammation" Z-0027), a grant provided by the Rainer Hoenig Stiftung, a grant provided by The Research Foundation Flanders ("Fonds Wetenschappelijk Onderzoek – Vlaanderen" (FWO); EOS 30826052), and a seed-funding from HI-TRON (Helmholtz-Institute for Translational Oncology Mainz). The study was supported by intramural research funding from the University Medical Center Göttingen (UMG), Germany.

Competing interests

M.J.W. was employed by Roche Diagnostics GmbH, this author declares no conflict of interest that pertain to this work. The other authors declare no competing interests.

Supplementary material

Supplementary material is available at Brain  online.