Introduction

Idiopathic inflammatory myopathies (IIM), also known as myositis, are a heterogeneous group of autoimmune disorders characterized by chronic inflammation of skeletal muscles, resulting in muscle weakness as a common feature1. IIM can be classified into several subgroups, including dermatomyositis, anti-synthetase syndrome, immune-mediated necrotizing myositis, inclusion body myositis, polymyositis and overlap myositis2. The incidence of IIM is estimated to range from 0.2 to 2 per 100,000 person-years, with prevalence from 2 to 25 per 100,000 people3. Despite the relatively low incidence of IIM, their heterogeneity poses persistent challenges in terms of accurate diagnosis, effective treatment, and evaluation of therapeutic efficacy.

The gut microbiome, which is mainly composed of the bacteriome, mycobiome, and virome, plays roles in various physiological processes such as metabolism, absorption, and immunity4,5,6. Changes in the composition and function of the gut microbiota, known as dysbiosis, exhibit a close association with diverse autoimmune disorders, including IIM7,8. Zhufeng et al9. demonstrated that individuals with IIM exhibit a significant reduction in the diversity of gut bacteriome compared to healthy controls. Additionally, they observed an increase in inflammatory-related bacteria and a significant decrease in certain butyrate-producing bacteria. Liang et al10. indicated a distinct alteration in the composition of gut bacteriome among patients with IMNM, characterized by an abnormal increase in Lactobacillus levels, which was accompanied by changes in clinical indicators. Similarly, a subsequent study also indicated that dysbiosis of gut bacteriome was associated with changes in serum inflammatory factors and oxidative stress markers in the IIM cohort11. Interestingly, the effective treatment for IIM can restore the imbalance of gut bacteriome9. All the evidence strongly suggests that dysbiosis of the gut microbiota plays a pivotal role in driving inflammation and immune dysregulation in IIM patients. Therefore, modulation of the microbiota offers potential for therapeutic intervention in IIM.

In addition to bacteria, the viral and fungal components within the gut also interact with the immune system and have an impact on human physiology. Recent studies suggest that the inter-kingdom associations among the gut bacteriome, mycobiome, and virome are closely related to the pathogenesis of autoimmune diseases such as rheumatoid arthritis, systemic lupus erythematosus, and osteoarthritis12,13,14. However, there is currently a lack of comprehensive understanding regarding the gut bacteriome, mycobiome, and virome in IIM patients.

Therefore, we conducted a comprehensive analysis of gut bacterial, fungal, and viral communities in 34 patients with IIM and 37 healthy controls using whole-metagenome shotgun sequencing of fecal samples. Our study meticulously investigated the associations among the multi-kingdom characteristics related to IIM, providing valuable insights into the role of gut microorganisms in the pathogenesis of IIM.

Materials and methods

Study design

A clinical cohort consisting of 34 IIM patients and 37 healthy controls was recruited at the Second Affiliated Hospital of Dalian Medical University. The IIM patients had not received any steroids, immunosuppressive drugs, or antibiotics within the preceding three months, and their diagnosis was based on the classification criteria suggested by EULAR/ACR in 201715. Healthy individuals did not experience arthralgia, heart failure, renal failure, autoimmune diseases, or inflammatory disorders. Moreover, individuals with the following conditions were excluded from the study: Firstly, patients with severe systemic diseases, malignancy, pyemia, cardiovascular, or metabolic disorders were not included. Secondly, those presenting symptoms of diarrhea were also excluded from participation. Additionally, individuals who had received antifungal, antibiotic, or probiotic treatment within the past month were deemed ineligible for the study. Furthermore, subjects with a history of excessive alcohol consumption were excluded to eliminate any potential confounding factors. Lastly, individuals who had consumed fermented milk products within the past week were also not included in the study. All participants were of Chinese ethnicity.

The study obtained approval from the Ethics Committee of the Second Affiliated Hospital of Dalian Medical University and was conducted in accordance with the principles outlined in the Declaration of Helsinki and the International Council for Harmonization Guidelines for Good Clinical Practice. Informed consent was obtained from all participants prior to their inclusion in the study.

Sample collection, DNA extraction and whole-metagenome shotgun sequencing

Fecal samples were collected by patients immediately after defecation in the hospital and promptly stored on dry ice. Subsequently, the samples were transported to the laboratory, where they were divided into two aliquots and transferred into cryotubes. All fecal samples were then stored at −80℃ for further analysis. The total DNA was extracted from fecal samples (170 mg per sample) using the Tiangen fecal DNA extraction kit (Tiangen, China) in accordance with the provided instructions. After the extraction, the concentration and purity of total DNA were evaluated using NanoDrop2000 and Qubit 4.0, respectively. The total DNA was fragmented to approximately 350bp with Covaris M220 (Gene Company Limited, China). Subsequently, each fragment was inserted into a 150 bp paired-end adaptor to prepare the library. All libraries were barcoded and pooled for whole-metagenome shotgun sequencing on the Illumina NovaSeq platform.

The metagenomic dataset was initially subjected to base calling using the default parameters on the sequencing platform. To ensure data quality, fastp16 was employed for independent quality control of raw sequencing reads. This involved removing low-quality bases (Q < 30) from the end of reads and filtering out N-containing, adapter-contaminated, or short (< 90-bp) reads to obtain high-quality sequences. The identification and removal of human reads were accomplished by aligning the high-quality reads against the human reference genome (GRCh38) using Bowtie217 alignment tool.

Gut bacteriome profiling

The gut bacteriome, which refers to the prokaryotic composition of the gut microbiome, was profiled using the MetaPhlAn4 algorithm in fecal metagenomes from all samples18. The relative abundances of prokaryotic species were determined by normalizing each sample. The relative abundances at the phylum and genus levels were subsequently obtained by aggregating the species abundances within each respective taxon. The functional composition of the fecal metagenomes was assessed using the KEGG Mapper19, which effectively and accurately characterizes the abundance of microbial function modules directly from metagenomic sequencing data.

Gut mycobiome profiling

The analysis of intestinal fungi was performed following the methodology described by Ren et al20., with minor modifications. First, a customized catalog of gut fungal genomes was constructed. A total of 16,634 raw fungal genomes were obtained from NCBI database on June 2024. Of these, 1384 were removed due to either extremely low assembly quality (N50 length <2,000 bp or number of scaffolds >10,000) or a mixture of multiple genomes21. The remaining 15,250 genomes were retained as a reference for further analyses. By incorporating methods from published articles, we updated the original database and identified a total of 1490 human-associated fungi species. These strains were then clustered into 317 non-redundant human-associated fungi species using an ANI threshold of 95%.

In order to mitigate the impact of non-specific read mapping to fungal genomes in subsequent analysis, as well as to exclude reads originating from human or prokaryotic sources, three distinct databases were utilized for mapping filtered reads: GRCh38 genome, UHGG collection22, and SILVA rRNA database23. The matched sequences were excluded, and the remaining reads in the samples were aligned against our customized fungal database using bowtie217. Subsequently, the read counts for each genome were calculated. The read counts were then converted into relative abundance by normalizing. The read count of each genome was divided by its genomic size for normalization, and the resulting normalized read count was then divided by the total sum of all normalized read counts in a sample. For different fungal taxa, the relative abundance of a taxon was determined by summing the relative abundances of all populations within the same taxa.

Gut virome profiling

The analysis of gut virome follow the protocol from Li et al24. A Chinese gut viral catalog (cnGVC), comprising 67,096 nonredundant viral operational taxonomic units (vOTUs) was utilized in our study25. The reads from samples were initially aligned to the UHGG database in order to eliminate potential bacterial contamination. Then the remaining high-quality reads were mapped to the cnGVC database using Bowtie2, based on a 95% nucleotide similarity threshold (a phylogenetic threshold for defining viral "species-level")26. The vOTUs abundance in each sample was determined by aggregating the reads mapped to each vOTU and normalizing it based on the total number of mapped reads in that particular sample. The relative abundance profile at family level was determined by aggregating the relative abundances of vOTUs belonging to the identical viral family.

Microbial correlation network analysis

Because metagenomic relative abundance data are inherently compositional, traditional correlation measures such as Pearson or Spearman can lead to spurious associations. To infer robust microbial interactions across bacteria, fungi, and viruses, we applied SparCC (Sparse Correlations for Compositional data), a method specifically developed to estimate linear relationships from compositional microbiome datasets. Prior to network construction, taxa with extremely low abundance (present in <10% of samples) were removed to reduce sparsity. SparCC was run for 20 inference iterations to generate the correlation matrix. To evaluate statistical significance, we performed 100 bootstrap permutations by random shuffling of the abundance table to generate a null distribution of correlation coefficients. For each taxon pair, a pseudo-p value was calculated as the proportion of bootstrap-derived correlations that exceeded the magnitude of the observed coefficient. Associations with |r| > 0.6 and p < 0.05 were considered significant and retained for network construction. The resulting microbial association networks were visualized using Cytoscape (v3.9.0).

Statistical analyses

The R v4.0.1 platform was utilized for performing the statistical analyses. For each sample, the richness of gut microbiota was assessed by determining the number of observed species-level taxa (i.e., bacterial species, fungal species, and vOTUs). Additionally, Shannon’s and Simpson’s indices were employed to estimate the diversity of the microbiome. These indices were computed utilizing the vegan package in R27, with a uniform number of reads (10 million) per sample. The vegan package was employed to conduct Principal coordinates analysis (PCoA) on Bray-Curtis distances. The permutational multivariate analysis of variance (PERMANOVA) analysis was performed using the adonis function from the vegan package, with p-values computed through 1,000 permutations. The differences in alpha diversity between IIM patients and healthy controls were assessed using the Wilcoxon rank-sum test, whereas taxonomic differences were evaluated using MaAsLin2. A Benjamini–Hochberg false discovery rate (FDR) correction was applied, and q-values < 0.05 were considered statistically significant. The randomForest package (with 1,000 trees) was utilized to train random forest models for discriminating between two cohorts based on the abundance profiles of differential bacteria, fungi, and vOTUs.

Results

Gut bacteriome: diversity, taxonomic composition, and function

Based on whole metagenome shotgun sequencing, we obtained a total of 449.74 Gbp of high-quality non-human data from the fecal samples of 34 patients with IIM and 37 healthy individuals, averaging at 6.33±2.26 per sample (Table S1). First, we mapped the sequencing reads from all samples to the UHGG database21 and obtained the gut prokaryotic profile (hereafter referred to as the “gut microbiome”), which includes a total of 3,007 bacterial and archaeal taxa, encompassing 13 phyla, 77 classes, 99 orders, 139 families, 476 genera, and 978 species. Rarefaction analysis showed that the richness of intestinal bacteria in the IIM group (measured by the number of observed species) was slightly higher than that in the HC group when both groups had an equal number of samples (Figure 1A). However, the analysis based on the number of observed species indicated that there was no significant difference between the two groups (Wilcoxon rank-sum test, p = 0.22; Figure 1B). Compared to the healthy control group, there were no significant differences in the Shannon diversity index, Simpson index, and the number of observed bacteria in the gut microbiota between patients with IIM and the control group (Figures 1C-D). These findings suggest that IIM exerts a minimal impact on bacterial diversity within gut samples. Next, we performed PCoA and PERMANOVA to further investigate the differences in gut bacterial communities between the two groups. The results showed a significant difference between the IIM and HC groups (PERMANOVA, p = 0.001), indicating dysbiosis in the gut microbiota of the IIM group (Figure 1E).

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

Difference in the gut bacteriome between IIM patients and healthy controls. (A), Rarefaction curves showing the number of observed species in the two groups. (B-D), Boxplots of alpha diversity at the species level, including the number of observed bacterial species (B), Shannon index (C), and Simpson index (D). (E), Principal coordinates analysis (PCoA) based on Bray–Curtis distances at the species level, illustrating the compositional separation between the two groups. (F), Relative abundance of gut bacterial genera in healthy controls and IIM patients. (G), The top 20 differentially abundant bacterial species between the two groups. (H), PCoA based on Bray–Curtis distances of functional module compositions. I-J, Boxplots of functional diversity showing the Shannon index (I) and Simpson index (J) between the two groups at the species level. (K), The top 10 differential functional modules between IIM patients and healthy controls.

At the phylum level, MaAsLin2 analysis identified several phyla with significantly different relative abundances between IIM patients and healthy controls (q < 0.25, FC > 1.2). Actinomycetota and Verrucomicrobiota were enriched in the IIM group, whereas Bacteroidota was more abundant in healthy subjects (Table S2). At the genus level, MaAsLin2 detected 41 genera with significant differences after multiple testing correction. Genera enriched in IIM included Rothia, Limosilactobacillus, Lancefieldella, Trueperella, and Mogibacterium, along with several other taxa. In contrast, genera such as Prevotella, Megamonas, Lachnospira, Faecalibacterium, and Faecalibacillus were significantly enriched in healthy individuals (Figure 1F; Table S3). At the species level, MaAsLin2 identified 85 species whose relative abundances differed significantly between the two groups. Among these, 49 species were enriched in IIM patients, including Rothia mucilaginosa, Streptococcus parasanguinis, Trueperella pyogenes, Mogibacterium diversum, and Akkermansia muciniphila. Conversely, 36 species were enriched in healthy controls, including Megamonas funiformis, Eubacterium rectale, Faecalibacterium prausnitzii, Lachnospira pectinoschiza, and Phocaeicola plebeius (Figure 1G; Table S4).

The Shannon and Simpson diversity indexes for taxonomic composition were calculated based on the relative abundance profile at the species level using the vegan package in the R platform, whereas those of the functional composition were calculated at the modules level. The Shannon and Simpson indexes were significantly higher in the gut functional composition of IIM patients compared to controls (Wilcoxon rank-sum test, p < 0.01; Figures 1I-J). Consistent with the observations of phylogenetic composition, the two groups on the PCoA plot showed significant differences (PERMANOVA, p = 0.007; Figure 1H).

Using MaAsLin2, we further identified 60 functional modules that differed significantly between the two groups after multiple testing correction (q < 0.05). Among the 729 annotated KEGG modules, 38 were enriched in IIM patients, while 22 were enriched in healthy controls (Table S5). Representative modules enriched in IIM included the Aspartate/glutamate/glutamine transport system (M00228), Cholesterol biosynthesis (M00101), and dTDP-D-forosamine biosynthesis (M00802). In contrast, healthy controls showed enrichment of modules involved in NAD(P)H: quinone oxidoreductase (M00145) and Fluoroquinolone transport system (M00224) (Figure 1K).

Gut mycobiome: diversity, taxonomic composition, and function

Next, we conducted an analysis on the gut fungal composition in all fecal samples utilizing the available fungal genome database sourced from the refined database. This refined database includes 317 distinct reference species, which are categorized based on a criterion of 95% average nucleotide identity (ANI), derived from a collection of 1,490 genomes associated with the human microbiome. Subsequently, high-quality reads from each sample were aligned to the genomes of the 317 distinct species present in our database. Furthermore, our examination revealed the presence of 177 high-level taxa across the samples, including 76 genera and 5 subphyla. Rarefaction analysis indicated that, under the condition of equal sample size between the two groups, the number of observed species in the control group was significantly higher than that in the IIM group (Figure 2A). Meanwhile, we further supported this point by analyzing the number of observed species (Wilcoxon rank-sum test, p = 0.0015; Figure 2B). However, the comparison of fungal biodiversity within the samples revealed no statistically significant differences in the Shannon index and Simpson index of the intestinal fungal communities between the two groups (p > 0.05; Figures 2C-D). In contrast, the PCoA analysis of the gut mycobiome group revealed significant differences between the patient group and the control group (PERMANOVA, p = 0.001; Figure 2E). These findings highlight the alterations in the structural composition of the gut fungal community observed in patients with IIM.

Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.
Full size image

Difference in the gut mycobiome between IIM patients and healthy controls. (A) Rarefaction curves showing the number of observed fungal species in the two groups. For each sequencing depth, 30 random resamplings were performed, and the median and interquartile range are plotted. B-D, Boxplots showing species-level α-diversity of the gut mycobiome, including Shannon index (B), Simpson index (C), and the number of observed fungal species (D) in IIM patients and healthy controls. (E), PCoA analysis of Bray–Curtis distances based on species-level fungal community composition, illustrating significant separation between the two groups. Each point represents one sample. Lines connect samples within the same group, and ellipses indicate variation around group centroids. (F), Pie plot shows the subphylum-level taxonomic annotation of the gut mycobiome. (G), Stacked bar plots show the genus-level taxonomic composition of the gut mycobiome in IIM patients and healthy controls. (H), Boxplot shows the representative differential gut mycobiome species when compared between patient and control groups.

At the genus level, MaAsLin2 identified several fungal genera that differed significantly between IIM patients and healthy controls (q < 0.25, FC > 1.2). Genera enriched in patients with IIM primarily included Exophiala, Ogataea, and Nakaseomyces, whereas healthy controls were characterized by higher abundances of genera such as Malassezia, Blastomyces, and Meyerozyma (Figures 2F–G; Table S6). At the species level, MaAsLin2 detected 33 fungal species with significantly different relative abundances between the two groups (q < 0.05; Figure 2H; Table S7). Among these species, 9 were enriched in IIM patients, including several opportunistic fungal taxa such as Aspergillus sp. c56, Trichophyton rubrum c61, and Candida boidinii c94. Conversely, 20 species were enriched in healthy controls, such as Malassezia restricta c191, Malassezia furfur c28, and Blastomyces emzantsi c231.

Gut virome: diversity, taxonomic composition, and function

The rarefaction analysis revealed that, when considering an equal number of individuals, the observed species richness in IIM patients was comparable to that in the healthy control group (Figure 3A). The results of the analysis on the number of observed species, which indicate no statistically significant difference between the two groups (Wilcox rank-sum test, p > 0.05), further validate this point (Figure 3B). The Shannon and Simpson indexes indicated no statistically significant difference in the gut virome diversity between IIM patients and the control group (p > 0.05; Figures 3C-D). However, the PCoA analysis consistently demonstrated a significant difference in viral composition at the vOTU level between the two groups (PERMANOVA, p = 0.001; Figure 3E).

Fig. 3
Fig. 3The alternative text for this image may have been generated using AI.
Full size image

Characteristics of the gut virus catalog and gut virome. (A), Rarefaction curves based on the number of observed viral operational taxonomic units (vOTUs), showing the vOTU richness under equal sampling depth in both groups. B-D, Boxplots showing the α-diversity of the gut virome at the vOTU level, including Shannon diversity index (B), Simpson index (C), and the number of observed vOTUs (D). (E), PCoA analysis of Bray–Curtis distance based on the composition of gut virome, revealing the separations between two groups. (F), Stacked bar plots showing the family-level taxonomic composition of the gut virome. (G), Boxplots showing representative viral families that differ significantly in relative abundance between IIM patients and healthy controls. (H), Volcano plot showing fold change versus –log10(p-value) for all vOTUs. The x-axis represents the fold change in vOTU abundance between groups, and the y-axis represents the significance of group differences. vOTUs enriched in IIM patients or healthy controls are colored in red and blue, respectively. (I), Pie charts showing the family-level taxonomic distribution of IIM-enriched vOTUs and HC-enriched vOTUs. (J), The occurrence rates of KEGG Orthologs (KOs) whose frequencies differ significantly between IIM-enriched and control-enriched vOTUs.

At the family level, MaAsLin2 identified several viral families that differed significantly between IIM patients and healthy controls (q < 0.25, FC > 1.2). Salasmaviridae, Metaviridae and Siphoviridae was significantly enriched in patients with IIM, whereas Microviridae, Quimbyviridae, Myoviridae and Inoviridae were more abundant in the control group (Figure 3G; Table S8). Next, we compared the composition of the gut virome at the vOTU level between IIM patients and healthy controls. We observed significant differences in the relative abundances of a total of 938 vOTUs between these two groups (Wilcoxon rank-sum test, p < 0.05; Figure 3H; Table S9). Among them, there were 284 vOTUs that showed enrichment in patients with IIM (IIM-enriched), while for the remaining 654 vOTUs their abundance was decreased compared to controls (control-enriched). The IIM-enriched vOTU contained five known virus families, primarily Siphoviridae and Myoviridae (Figure 3I), while the control-enriched vOTU covered eight families, predominantly Siphoviridae, Myoviridae and Microviridae. Furthermore, we predicted a total of 51,773 protein-coding genes from 1,782 differential vOTUs and annotated 19.7% of them based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database28. These annotated genes were assigned to 2,026 KEGG orthologous genes (KOs) for further analysis. 71 KOs exhibited significant differences in occurrence frequencies between IIM-enriched and control-enriched vOTUs (Fisher’s exact test, p < 0.05; Table S10). Among these KOs, 39 were encoded at a higher frequency in IIM-enriched vOTUs, such as K14059 (integrase), K03497 (parB, spo0J; ParB family transcriptional regulator, chromosome partitioning protein) and K03466 (DNA segregation ATPase FtsK/SpoIIIE), while 44 KOs, including K04763 (xerD; integrase/recombinase XerD), K00682 (gamma-glutamylcyclotransferase), and K03169 (topB; DNA topoisomerase III), were more enriched in the control-enriched vOTUs (Figure 3J).

Relationships between the gut bacteriome, mycobiome, and virome

Using SparCC (|r| > 0.6, p < 0.05), we constructed multi-kingdom ecological networks for both IIM patients and healthy controls. Edges in Figures 4A–B represent significant SparCC associations that met the predefined thresholds. After filtering, no significant negative correlations were retained in the final networks, so all displayed edges correspond to positive associations. Because the number of significant edges is large, Figures 4A–B are designed to emphasize group-level differences in overall network size and topology rather than specific pairwise relationships. Detailed edge-level correlation coefficients and the corresponding taxon pairs are provided in Tables S11–S12. The network of the IIM group consisted of 71 nodes and 322 edges, whereas the HC network contained a comparable number of nodes (70) but only 174 edges (Figures 4A–B, Table S11 and S12). This marked increase in the number of edges in the IIM network suggests that microbial interactions became more intensive and complex under disease conditions, indicating a possible restructuring of ecological relationships within the gut ecosystem of IIM patients. In addition to edge density, the connectivity patterns differed substantially between groups. The IIM network exhibited pronounced shifts in node centrality: fungal taxa emerged as dominant hubs, exerting strong influence on cross-kingdom associations. Key fungal species—including Penicillium c215, Malassezia globosa c182, Aspergillus nomiae c227, and Meyerozyma guilliermondii c264—showed markedly high importance scores and occupied central positions within the IIM microbial network (Figure 4C). In contrast, the HC network exhibited a more balanced and bacteria/virus-centered structure, in which bacterial and viral taxa served as the primary hub nodes, whereas fungal species contributed only minimally to network connectivity (Figure 4D). This stands in sharp contrast to the IIM network, where fungal taxa emerged as central hubs. The shift from a bacteria- and virus-dominated network in healthy individuals to a fungi-centric interaction architecture in IIM patients highlights a pronounced reorganization of interkingdom relationships associated with the disease.

Fig. 4
Fig. 4The alternative text for this image may have been generated using AI.
Full size image

Interactions among gut bacteriome, mycobiome, and virome in IIM patients and healthy controls. (A, B), SparCC-based co-abundance interaction networks (|r| > 0.6, q < 0.05) showing cross-kingdom correlations among bacterial species, fungal species, and viral vOTUs in IIM patients (A) and healthy controls (B). C, D, Barplots showing the top 30 microbial species with the highest degree (number of connections) in the networks of IIM patients (C) and healthy controls (D). Species from the bacteriome, mycobiome, and virome are distinguished by different colors.

Together, these findings indicate that IIM is accompanied not only by taxonomic dysbiosis but also by profound alterations in microbial interaction architecture, particularly characterized by a strengthened role of fungal components in shaping gut ecological networks.

Classification of IIM using multi-kingdom signatures

Finally, we evaluated the classification potential of multi-kingdom microbial signatures (bacteriome, mycobiome, and virome) for distinguishing IIM patients from healthy controls. Using a random forest classifier with 10-fold cross-validation, the gut bacterial, fungal, and viral profiles achieved AUCs of 0.916, 0.986, and 0.996, respectively (Figure 5A). When integrating all three microbial kingdoms, the combined model further improved the discriminatory ability, yielding an AUC of 0.997 (Figure 5B). The top-ranked microbial features contributing to classification performance are shown in Figure 5C. Certain bacterial taxa (e.g., Rothia mucilaginosa) and fungal species (such as Malassezia restricta c191 and Penicillium c215) demonstrated notable importance in differentiating IIM patients from controls. However, in the combined model, viral signatures dominated the feature importance ranking, suggesting that virome alterations may play a major role in distinguishing disease status. This observation highlights the need for further investigation into the functional relevance of these viral markers in IIM and related immune-mediated conditions.

Fig. 5
Fig. 5The alternative text for this image may have been generated using AI.
Full size image

Classification performance of multi-kingdom microbial signatures in distinguishing IIM patients from healthy controls. (A) Receiver operating characteristic (ROC) curves showing the classification performance of the gut bacteriome, mycobiome, and virome using a random forest classifier with 10-fold cross-validation. (B), ROC curve of the combined multi-kingdom model integrating bacteriome, mycobiome, and virome features (C-F), Barplot showing the top microbial features ranked by their importance in the combined random forest model.

Discussion

Idiopathic inflammatory myopathies (IIM) represent an autoimmune disorder characterized by a diverse range of systemic manifestations. Current research highlights a limited understanding of the association between immune disorders and gut microbiota in patients with IIM. Therefore, this study employed whole-metagenome sequencing to analyze fecal samples obtained from 34 IIM patients and 37 healthy controls, aiming to investigate the microbial compositions within these two cohorts. Our results revealed significant variations in the abundance of 85 bacterial species, 33 fungal species, and 938 viral species between the two groups. The distinctive features of the gut microbiome offer novel insights into comprehending IIM and associated immunological disorders.

Our findings suggest that there were no significant alterations observed in the diversity of gut bacteria among individuals with IIM. Upon reviewing the limited number of published articles on gut microbiota in IIM, we have identified a controversy surrounding this topic. Consistent with our results, two studies conducted in China similarly reported no changes in intestinal diversity among IIM7,11. However, three other studies indicated that the gut bacteria diversity of IIM patients were lower compared to that of healthy individuals9,10,29. It should be noted that these three studies did not exclude patients who were using immunosuppressive or antibiotic medications, implying that the observed decrease in gut microbiota diversity may potentially be attributed to drug usage.

Although there is no alteration in the diversity of gut bacteriome in IIM, a significant disparity exists in bacterial composition between these two cohorts. At the species level, several taxa enriched in IIM patients have been previously associated with mucosal inflammation, immune dysregulation, or epithelial barrier impairment, which provides biological plausibility for their involvement in IIM-related immune disturbances. For example, Rothia mucilaginosa and Streptococcus parasanguinis are well-recognized oral-associated opportunistic species frequently detected in dysbiotic gut environments30,31. Their enrichment may reflect an “oralization” of the gut microbiome, a phenomenon linked to heightened inflammatory tone and increased Th17 activation in autoimmune diseases. In addition, Trueperella pyogenes and Mogibacterium diversum are inflammation-associated taxa capable of producing proinflammatory metabolites, which may further enhance immune activation and contribute to the altered inflammatory milieu observed in IIM32,33. The increased abundance of Akkermansia muciniphila in IIM is also notable, as this mucin-degrading bacterium has been implicated in epithelial barrier turnover and immune modulation. Excessive mucin degradation could potentially alter intestinal permeability and increase antigen exposure in susceptible individuals34.

In addition, a decreased abundance of Faecalibacterium prausnitzii, Megamonas funiformis, and Eubacterium rectale were observed in IIM. It is notable that the majority of them possess anti-inflammatory properties and could promote intestinal health by facilitating the production of short-chain fatty acids (SCFAs). Faecalibacterium prausnitzii is a prominent producer of butyrate in the gut, and butyrate plays a pivotal role in modulating the activity of intestinal T-cells, thereby mitigating intestinal inflammation35,36. A consistent reduction in Faecalibacterium prausnitzii has been observed within the gut microbiotas of various autoimmune diseases including Crohn’s disease37, rheumatoid arthritis14, and osteoarthritis12. The administration of Faecalibacterium prausnitzii was observed to attenuate the population of IL-17-producing immune cells and modulate the microbial composition in a mouse model of rheumatoid arthritis38. It is suggested that Faecalibacterium prausnitzii exhibits potential therapeutic properties and may serve as a novel approach for the treatment of IIM. The abundance of another significant butyrate producer, Eubacterium rectale, has been observed to be diminished in individuals with inflammatory bowel disease (IBD) and adult-onset type 1 diabetes39,40. All these findings indicate that these butyrate-producing microorganisms play a pivotal yet inadequately explored role in the pathogenesis of IIM. These bacterial species are expected to serve as valuable indicators for assessing intestinal health.

In the gut mycobiome, the abundance of Aspergillus c56 is observed to increase in the intestines of IIM patients. Aspergillus fungi are opportunistic pathogens exists extensively in the nature. It can invade the skin, mucosa, or internal organs of immunocompromised individuals, leading to infections. Patients with myositis may develop Aspergillus infection as a consequence of immunosuppressive agent usage41. However, further investigation is required to determine whether the increase in intestinal Aspergillus is associated with bacterial transfer within the lung-gut axis. On the contrary, several taxa including the Pichia genus were depleted in the mycobiome of IIM patients. Pichia fermentans c90 can exert its metabolic activities to enhance the growth of beneficial microorganisms while suppressing pathogens in the intestinal environment, thereby contributing to the maintenance of intestinal health42. The decrease in Pichia abundance may suggest a potential attenuation of the fungal community’s functional role on the host; however, further investigations are needed. Several Malassezia species—including M. restricta c191, M. furfur c28, and M. arunalokei c192—were significantly reduced in IIM patients compared with healthy controls. The biological significance of this decrease remains unclear, as the roles of these skin-associated yeasts in systemic autoimmune diseases are still poorly understood. Further mechanistic studies are needed to determine whether the loss of these commensal fungi influences mucosal immunity or contributes to dysregulated host–microbe interactions in IIM.

At the vOTU level, the prevalence of Siphoviridae viruses was found to be higher in the intestines of IIM patients, whereas certain viruses belonging to Microviridae and Inoviridae exhibited relatively lower abundance. The Siphoviridae family typically possesses a long tail structure, facilitating its attachment to the bacterial surface and subsequent delivery of genetic material into the host cell43. A recent study has demonstrated an increase prevalence of Siphoviridae in patients with rheumatoid arthritis and rheumatoid arthritis-associated interstitial lung disease14. However, our current understanding of intestinal viruses and their involvement in autoimmune diseases remains limited. Considering the well-established role of the bacteriome in the etiology and progression of autoimmune diseases, alterations in bacteriophage abundance suggest their potential involvement in pathogenesis through modulation of bacterial taxa or direct interactions with the host immune system. Elucidating the composition of the gut virome in IIM may uncover novel disease mechanisms or potential therapeutic targets. In addition to taxonomic differences, the functional profiles of gut viruses showed marked shifts between IIM patients and healthy controls. Many of the KOs enriched in IIM-associated vOTUs—including K14059 (integrase), K03497 (ParB family chromosome-partitioning protein), and K03466 (FtsK/SpoIIIE DNA segregation ATPase)—are key components of phage-mediated genome integration, DNA recombination, and chromosome partitioning34. These functions are typically associated with enhanced horizontal gene transfer and stress adaptation in bacterial hosts. Their increased frequency in the IIM-enriched virome suggests that phage–bacteria interactions may be more dynamic under disease conditions, potentially promoting bacterial genomic rearrangement, metabolic plasticity, or antigenic variation that could influence host–microbe immune interactions. Conversely, several functions enriched in the control-associated vOTUs, such as K04763 (xerD recombinase), K00682 (gamma-glutamylcyclotransferase), and K03169 (topB; DNA topoisomerase III), are commonly linked to chromosomal maintenance and metabolic regulation in stable bacterial populations44. The depletion of these functions in IIM may indicate a loss of virome-mediated regulatory processes that help maintain microbial homeostasis in healthy individuals. Together, these functional alterations suggest that IIM is accompanied not only by changes in viral community composition but also by a shift toward phage activities that may destabilize bacterial genomes, reshape microbial ecology, and amplify immune-relevant microbial dynamics.

The SparCC-based ecological network analysis revealed profound restructuring of interkingdom microbial interactions in IIM. Although the number of microbial taxa included in the networks was similar between groups, the IIM network contained nearly twice as many edges as the HC network, indicating a more densely connected and potentially unstable microbial community under disease conditions. Importantly, network centrality shifted markedly toward fungal taxa in IIM patients, with species such as Penicillium c215, Malassezia globosa c182, Aspergillus nomiae c227, and Meyerozyma guilliermondii c264 emerging as dominant hubs. This contrasts sharply with healthy controls, where bacterial and viral taxa represented the major hub nodes and fungi played only a minimal structural role. Such a transition toward a fungi-centric network architecture suggests substantial rewiring of cross-kingdom relationships and may reflect a breakdown of ecological balance driven by inflammation, impaired gut barrier function, or altered immune–microbial crosstalk. These findings align with the growing recognition that the mycobiome can influence immune responses and raise the possibility that fungal components may actively contribute to the immune landscape of IIM. Consistent with the network results, the combined multi-kingdom classification model showed that viral signatures dominated the top predictive features, indicating that virome alterations hold strong discriminative power for distinguishing IIM from healthy controls. Taken together, the ecological and machine-learning evidence suggests that, beyond classical bacterial dysbiosis, both the mycobiome and virome play underappreciated roles in IIM-associated microbial disruption. Further mechanistic studies are needed to determine whether these fungal- and viral-centered alterations contribute directly to disease pathogenesis or arise as secondary consequences of immune dysregulation.

Our study has several limitations. The sample size was relatively modest, and we were not able to fully account for potential population factors such as sex and age. In addition, detailed clinical indicators of disease activity were not collected, which prevented us from examining how disease severity may relate to changes in the gut microbiome. As a result, although disease status was the primary focus of our analyses, we cannot completely rule out the possibility that population-related factors such as age, sex, or disease activity may have contributed to the observed microbiome differences and classification performance. Future studies will aim to incorporate more comprehensive and balanced clinical metadata to better account for these factors and to further disentangle disease-specific effects from population-related variation. Dietary habits are known to influence microbial composition. Although none of the participants reported extreme dietary patterns, we did not conduct a comprehensive assessment of dietary effects. IIM is a heterogeneous condition, and a more refined clinical classification of patients will be needed in future work. Finally, our findings are based on correlative analyses of the gut microbiota without experimental validation. Despite these limitations, the present study provides an important foundation for future mechanistic and clinical investigations.

Conclusions

In this multi-kingdom metagenomic study, we demonstrate that IIM is accompanied by substantial alterations in the gut bacteriome, mycobiome, and virome, despite largely unchanged microbial diversity. Disease-associated taxa included numerous pro-inflammatory bacteria and opportunistic fungi, whereas beneficial short-chain fatty acid–producing bacteria and commensal fungi were markedly reduced. The virome exhibited distinct taxonomic and functional signatures, including enrichment of phage genes involved in genome integration and DNA recombination, suggesting active viral modulation of bacterial dynamics in IIM. Interkingdom ecological network analysis further revealed a striking shift toward a fungi-dominated interaction architecture in IIM, contrasting with the bacteria- and virus-centered networks observed in healthy individuals. These changes indicate fundamental restructuring of microbial interactions within the gut ecosystem under disease conditions. Machine-learning analyses supported the biological findings, with viral signatures emerging as the strongest predictors of IIM status, underscoring the diagnostic potential of virome-informed biomarkers. Together, our findings reveal that IIM is characterized by multi-layered microbial dysbiosis and substantial remodeling of cross-kingdom interactions. These insights expand current understanding of host–microbe relationships in IIM and provide a foundation for future mechanistic and translational studies aimed at identifying microbial drivers and therapeutic targets in this autoimmune disorder.