Characterization of the Fecal Metabolome in Patients With Early Systemic Sclerosis

Abstract

Objective

To evaluate the fecal metabolome in patients with early systemic sclerosis (SSc) compared with unaffected controls and to determine if altered metabolites are associated with specific bacterial genera in patients with early SSc.

Methods

Stool samples and clinical data were collected from 106 patients with early SSc and 79 unaffected control patients. Targeted metabolomics was performed on fecal samples using liquid chromatography mass spectrometry, and 16S ribosomal RNA gene sequencing was used to determine the microbial composition of fecal samples.

Results

Compared with unaffected controls, patients with early SSc had higher levels of nicotinamide, 5’‐methylthioadenosine, and several short‐chain fatty acids (SCFAs) including valeric acid, propionic acid, and caproic acid. Conversely, patients with early SSc had lower levels of xylonic acid, orotate, methionine sulfoxide, and sarcosine. SCFAs were associated with unique bacterial genera, several of which were more abundant in patients with SSc compared with unaffected controls.

Conclusion

The fecal metabolome is altered in patients with early SSc, with a shift toward increased SCFAs.

INTRODUCTION

Systemic sclerosis (SSc) is a rare autoimmune fibrosing disease that affects 24.4 per 100,000 people in the United States. ¹ The most common organ systems involved in SSc are the skin and the gastrointestinal (GI) tract, each affecting >90% of patients with SSc. ² GI involvement in SSc includes upper GI tract dysfunction (eg, esophageal reflux, esophagitis, gastritis, and delayed gastric emptying) as well as lower GI tract dysfunction (eg, small intestinal bacterial overgrowth [SIBO], malabsorption, constipation, diarrhea, and fecal incontinence). ³ Patients experience significant morbidity because of GI manifestations of the disease. In addition, malnutrition because of malabsorption also contributes to the high mortality associated with SSc. ⁴

Emerging research suggests that GI dysfunction is a feature of very early and early SSc. A recent study found that patients with a very early diagnosis of SSc (VEDOSS) ⁵ reported GI tract symptoms at rates that were similar to those of patients with longer‐standing disease. These symptoms included gastroesophageal reflux disease (GERD), malnutrition, diarrhea, constipation, and SIBO and had a significant impact on the patients’ quality of life. ⁶

Previous studies have also shown that alterations in the gut microbiome are a feature of both early and established SSc. ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ In a study by Andréasson and colleagues, investigators found that patients with a recent diagnosis of SSc (within 3 years) had lower abundance of commensal genera and higher abundance of pathobiont genera compared with controls. ⁷ Additionally, in a study conducted by Bellando‐Randone and colleagues in Italy, investigators found that dysbiosis (as defined by loss of anti‐inflammatory intestinal flora) was already present in a cohort of patients with VEDOSS. ¹⁶

A limitation of prior work in this area, however, is a lack of understanding of the role of metabolites in SSc‐GI disease pathogenesis. Microbiome‐derived metabolites may be the key to understanding how alterations in the gut microbiome affect disease both locally and in distant organs (ie, the gut‐lung axis or the gut‐brain axis). For example, short‐chain fatty acids (SCFAs) are microbial‐derived metabolites that have been shown to play a role in GI immune function through the activation of G‐coupled protein receptors and inhibition of histone deacetylases. ¹⁷ Several studies have shown that SCFAs are decreased in patients with inflammatory bowel disease (IBD) ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² and that supplementation with butyrate (an SCFA) in a mouse model of ulcerative colitis leads to decreased intestinal inflammation. ²³

In this study, we aimed to characterize the fecal metabolome of patients with early SSc (diagnosed within the past 3 years). We hypothesized that, similar to alterations in the microbiome, alterations in the fecal metabolome would be a feature of early SSc. Additionally, we hypothesized that specific fecal metabolites would correlate with specific bacterial genera. Through this work, we hope to improve understanding of the association between the fecal metabolome and the fecal microbiome in patients with early SSc.

PATIENTS AND METHODS

Study participants

Study participants and recruitment were the same as previously described in Andréasson et al. ⁷ Adult patients fulfilling the 2013 American College of Rheumatology/EULAR Classification Criteria for SSc ²⁴ were consecutively recruited from the rheumatology clinics of Lund University in Sweden. To determine whether alterations in the fecal metabolome were present early in the SSc disease course, patients with a disease duration of <3 years from the time of their SSc diagnosis were included. Unaffected control patients without a history of autoimmune diseases or GI disease were simultaneously recruited from Lund University. Exclusion criteria for patients with SSc and unaffected controls included IBD, current/prior GI malignancy, use of any antibiotics within 4 weeks of stool collection, chronic use of antibiotics (defined as the use of antibiotics for any indication more than three times in the preceding year), use of supplemental probiotics, any prior GI surgery (except for a remote history of appendectomy), and a history of fecal microbial transplantation. Patients were allowed to remain on proton pump inhibitors (PPIs) to minimize the risk of unnecessary morbidity during the study.

The University of California, Los Angeles (UCLA) Institutional Review Board (#13‐001089) and the Regional Ethics Review Board, Lund, Sweden (#2011‐596) approved the study protocol, and written informed consent was obtained from each participant. The study protocol was in compliance with the ethical guidelines of the 1975 Declaration of Helsinki. Representatives from the national patient organization for SSc in Sweden approved and encouraged patient participation in this study.

Clinical assessments

As previously described, ⁷ clinical features of the SSc participants were evaluated at the time of the stool collection (Table 1). Disease duration was defined based on the onset of the first non‐Raynaud symptom attributable to SSc. Immunomodulatory therapy use was based on any use of immunomodulatory medications up until the date of the stool collection. Interstitial lung disease was defined by the presence of interstitial lung disease on high‐resolution computed tomography. Right heart catheterization was used to diagnose pulmonary hypertension. The presence of all other disease features was defined based on a physician's clinical diagnosis identified through chart review and/or confirmed based on objective testing (eg, a diagnosis of SIBO was based primarily on lactulose breath testing).

Table 1.

Characteristics of patients with early SSc included in this study^*

Characteristic	N = 106
Age, mean ± SD, a y	55.31 ± 15.9
Female, n/N (%)	90/106 (84.9)
Diffuse disease, n/N (%)	20/106 (18.9)
Limited disease, n/N (%)	86/106 (81.1)
Disease duration, b y
Median (IQR)	2.0 (4)
Mean ± SD	4.5 ± 6.9
History of SIBO, n/N (%)	5/106 (4.7)
History of pseudo‐obstruction, n/N (%)	2/106 (1.9)
History of fecal incontinence, n/N (%)	7/106 (6.6)
Current BMI, mean ± SD, kg/m2	25.5 ± 4.8
Current BMI <18.5 kg/m2, n/N (%)	1/106 (0.9)
History of GAVE, n/N (%)	2/106 (1.9)
History of GERD, n/N (%)	65/106 (61.3)
HRCT‐defined ILD, n/N (%)	35/106 (33.0)
Pulmonary hypertension by RHC, n/N (%)	7/106 (6.6)
History of renal crisis, n/N (%)	1/106 (0.9)
History of SSc cardiac involvement, n/N (%) c	20/106 (18.9)
History of inflammatory myopathy, n/N (%)	9/106 (8.5)
Current mRSS (0–51), mean ± SD	5.5 ± 7.6
History of calcinosis, n/N (%)	4/106 (3.8)
History of digital ulcers, n/N (%)	12/106 (11.3)
Immunosuppression use ever, n/N (%) d	35/106 (33.0)
Current immunosuppression use, n/N (%) d	35/106 (33.0)
Ever smoker, n/N (%)	54/105 (51.4)
Vegetarian, n/N (%)	5/106 (4.7)
Current PPI use, n/N (%)	62/104 (59.6)
Current promotility agent use, n/N (%) e	2/106 (1.9)
Scl‐70 antibody, n/N (%)	19/106 (17.9)
Anticentromere antibody, n/N (%)	42/106 (39.6)
RNA polymerase III antibody, n/N (%)	3/106 (2.8)

^*BMI, body mass index; GAVE, gastric antral vascular ectasia; GERD, gastroesophageal reflux disease; HRCT, high‐resolution computed tomography; ILD, interstitial lung disease; IQR, interquartile range; mRSS, modified Rodnan skin score; PPI, proton pump inhibitor; RHC, right heart catheterization; SIBO, small intestinal bacterial overgrowth; SSc, systemic sclerosis.

^a Among 79 unaffected controls, the mean ± SD age was 55.75 ± 12.9, and the percentage of female patients was 70.9%.

^b Based on the onset of the first non‐Raynaud symptom attributable to SSc.

^c Type of cardiac involvement (patients could have more than one cardiac manifestation): right bundle branch block (n = 7), pericardial effusion (n = 4), left anterior fascicular block (n = 4), diastolic dysfunction (n = 3), cardiomyopathy (n = 2), atrial fibrillation (n = 1), atrioventricular block (n = 1), and left bundle branch block (n = 1).

^d Immunosuppression use was based on any consumption of immunosuppressive medications until the date of the stool collection. Please see Supplementary Table 1 for a summary of the immunosuppressive medications used.

^e Type of promotility agents used included metoclopramide prn (n = 1) and domperidone prn (n = 1).

Fecal metabolomics

Participants provided stool specimens, which were immediately frozen at −80°F, as previously described. ⁷ All samples were shipped overnight on dry ice to UCLA for further processing. Targeted metabolomics was performed on frozen stool samples as described previously. ²⁵ Briefly, frozen stool samples were lyophilized overnight. Lyophilized stool samples were then ground using a mortar and pestle, weighed, and added to methanol extraction buffer. Samples were homogenized using bead homogenization and a volume equivalent to 1.5 mg of each sample was removed to a new tube for drying. Dried metabolites were reconstituted in 100 μL of a 50% acetonitrile 50% distilled H₂O solution. Samples were vortexed and spun down for 10 min at 17,000g, and 70 μL of the supernatant was then transferred to high‐performance liquid chromatography glass vials. Subsequently, 10 μL of these metabolite solutions were injected per analysis. Samples were run on a Vanquish (Thermo Scientific) ultra high‐performance liquid chromatography system as previously described. ²⁵

The mzXML files were imported into the MZmine 2 software package. Ion chromatograms were generated from MS1 spectra via the built‐in Automated Data Analysis Pipeline (ADAP) chromatogram module, and peaks were detected via the ADAP wavelets algorithm. Peaks were aligned across all samples via the random sample consensus aligner module, gap‐filled, and assigned identities using an exact mass MS1 (± 15 parts per million) and retention time (RT; ± 0.5 min) search of our in‐house MS1‐RT database. Peak boundaries and identifications were then further refined by manual curation. Peaks were quantified by area under the curve (AUC) integration and exported as CSV files. ²⁵ Raw peak areas were normalized to four internal standards to account for any variability between runs on the mass spectrometer. Metabolite feature tables were filtered to retain metabolites that were detected in at least 25% of participants. This resulted in a final feature table of 122 named metabolites.

16S ribosomal RNA gene sequencing and microbial composition analysis

The fecal microbiome of the patients included in this study was previously reported in the study by Andréasson et al. ⁷ As described previously, ⁷ microbial DNA was extracted from stool specimens by bead beating, and the V4 region of the bacterial 16S ribosomal RNA gene was sequenced using an Illumina NovaSeq 6000 (Illumina, Inc.) to a mean sequence depth of 301,239 per sample. ⁹ All samples were analyzed simultaneously at UCLA to avoid any batch effects. DADA2 was used to perform quality filtering, merge paired end reads, remove chimeras, and cluster sequences into exact amplicon sequence variants (ASVs). ²⁶ Taxonomy was assigned for ASVs based on the SILVA v132 database down to the level of family, genus, or species, depending on the depth of reliable classifier assignments. Samples were filtered to retain genera with at least 10% nonzero counts and center log ratio transformation was applied.

Statistical analysis

Fecal metabolomics

Contrast analysis within the framework of the general linear model (GLM; latent class analysis GLM) tested for group differences between patients with SSc versus unaffected controls in log2 transformed metabolites controlling for age and sex. For this statistical analysis, we considered a P value of <0.05 as the threshold for reporting and provide false discovery rate corrected P values (Q) corrected for the number of features (ie, metabolites) tested.

To determine whether there was a multivariate metabolomic signature that could discriminate SSc from controls, we employed sparse partial least squares discriminant analysis (sPLS‐DA). Although the primary goal of this analysis was to generate a parsimonious model through feature selection and data reduction, ²⁷ we also examined the predictive accuracy of the derived brain signature. sPLS‐DA is a latent variable approach that employs a supervised framework forming linear combinations of the predictors (ie, metabolites) based on class membership (ie, SSC/control) and reduces the dimensionality of the data by finding a set of orthogonal components each made up of a selected set of features or variables. ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ These components were referred to as the metabolomic signatures. Each variable made up of a metabolomic signature had an associated “loading,” which was a measure of the relative importance of the variable for the discrimination into two groups. ²⁸ The model was tuned for the optimal number of components and species per component using 10‐fold cross validation repeated 50 times, and the model that minimized the balanced error rate (BER), the average proportion of wrongly classified samples in each class, was selected as best‐fitting. The scores for each participant on each component/signature were plotted in two‐dimensional space to visualize the discrimination in sample plots. An independent t‐test was used to quantify the group differences on the derived signature and effect size differences between groups on the signature were computed using Cohen's d. The data set was divided (80/20) into a training and test set. The predictive accuracy of the signatures derived from the training set used to build the model were assessed in the unseen test data by calculating the BER, the average proportion of wrongly classified samples in each class, weighted by the number of samples in each class. Balanced accuracy or AUC, the average of the sensitivity and the specificity, was also calculated.

Associations between SCFAs and the fecal microbiome

GLMs were applied to determine the associations between genus‐level gut bacteria and SCFAs after adjusting for age, sex, SIBO, body mass index (BMI), PPI use, and any immunosuppression. We used Q < 0.05 as a reporting threshold and forest plots to visualize the results. Latent class analysis GLM was applied to determine whether genera positively associated with SCFAs were more abundant in patients with SSc compared with unaffected controls controlling for age and sex based on a reporting threshold of P < 0.05.

For all statistical analyses we also provided effect size estimates, Cohen's d for mean differences (small 0.20, medium 0.50, and large 0.80) and standardized beta (small 0.10–0.29, medium 0.30–0.49, and large ≥0.50) for association analyses. ³⁴ All analyses were performed in R.

RESULTS

Patient characteristics

This study included 115 patients with early SSc and 79 unaffected controls. Of the 115 patients with SSc who provided a stool sample, 106 had a stool sample that underwent metabolite analysis and had paired clinical data at the time of the stool collection. The mean age of patients with SSc was 55.3 years, which was comparable to the mean age of unaffected controls (55.8 years) (Table 1). A greater proportion of patients in the SSc group were female (84.9%) compared with the unaffected control group (70.9%). Among the patients with SSc, most had limited cutaneous disease (81.1%) and a median disease duration (based on the onset of the first non‐Raynaud symptom attributable to SSc) of 2 years. Most patients with SSc had a history of GERD (61.3%) and were taking a PPI (59.6%) at the time of stool collection. Given that patients with SSc were early in their disease course, a minority (33%) was receiving immunosuppression or had ever received immunosuppression.

Alterations in the fecal metabolome in early SSc

Among 122 detected metabolites, the levels of 10 fecal metabolites were significantly altered in patients with SSc compared with unaffected controls after adjusting for age and sex (Table 2). Compared with unaffected controls, patients with SSc had higher levels of the following SCFAs: valeric acid isomers, propionic acid, and caproic acid (= hexanoic acid) (Figure 1), as well as nicotinamide (vitamin B₃) and 5’‐methylthioadenosine (a methionine salvage pathway intermediate). On the other hand, patients with SSc had lower levels of xylonic acid (a metabolite of xylose), orotate (pyrimidine biosynthesis), methionine sulfoxide (formed by the oxidation of methionine), and sarcosine (amino acid) compared with unaffected controls.

Table 2.

Significant alterations in fecal metabolites in patients with early SSc versus unaffected controls^*

Metabolite	Mean diff	P value	Q value	Effect size (95% CI)
Metabolites increased in patients with SSc compared with unaffected controls
Short‐chain fatty acids
Valeric acid isomers	0.55	0.001	0.080	0.52 (0.22 to 0.83)
Propionic acid	0.43	0.0107	0.216	0.39 (0.09 to 0.69)
Caproic acid	0.79	0.0187	0.246	0.36 (0.06 to 0.66)
Vitamins
Nicotinamide	0.75	0.002	0.117	0.48 (0.17 to 0.78)
SAM‐cycle intermediates
5'‐methylthioadenosine	1.58	0.005	0.188	0.43 (0.13 to 0.73)
Metabolites decreased in patients with SSc compared with unaffected controls
Nucleotides
Orotate	−0.75	0.012	0.216	−0.38 (−0.68 to 0.08)
Other
Xylonic acid	−1.06	0.008	0.216	−0.4 (−0.71 to 0.1)
Methionine sulfoxide	−0.74	0.016	0.237	−0.37 (−0.67 to 0.07)
Sarcosine	−0.42	0.020	0.246	−0.35 (−0.65 to 0.05)

^*CI, confidence interval; mean diff, mean difference; SAM, S‐adenosylmethionine; SSc, systemic sclerosis.

Figure 1.

Patients with early SSc (blue) have increased abundance of specific short‐chain fatty acids compared with unaffected controls (red). SSc, systemic sclerosis.

Fecal metabolite signature discriminating SSc and controls

sPLS‐DA resulted in a one‐component fecal metabolite signature made up of 20 metabolites (see the loadings plot in Figure 2). The metabolite with the highest influence in the discrimination of SSc and controls was valeric acid, an SCFA. Two other SCFAs (propionic acid and caproic acid) also made up this signature. Many of the metabolites making up this fecal metabolite signature were also found to be significant in univariate GLMs at uncorrected P values (see Table 2). Compared with controls, on average SSc scored higher on this fecal metabolite signature (t[147] = −5.22, P = 6.11e−07, Cohen's d = 0.86) (Figure 2). The BER for this signature was 0.42 (AUC 58%, sensitivity 76%, and specificity 40%). Given the finding that SCFAs were significantly altered in patients with SSc and, in combination with other metabolites, were able to discriminate between patients with SSc and unaffected controls, we decided to focus further analyses on SCFAs.

Figure 2.

Sparse partial least squares discriminant analysis for differentiating SSc from unaffected controls based on fecal metabolome. (A) Orange bars indicate metabolites that are more abundant in patients with early SSc compared with controls. Blue bars indicate metabolites that are more abundant in controls compared with early SSc. (B) The classification performance of sparse partial least squares discriminant analysis in part A when applied to patient samples. Unaffected controls are represented by blue circles, and SSc patients are represented by orange triangles. CMP, cytidine monophosphate; GMP, guanosine monophosphate; SSc, systemic sclerosis.

Integrative analysis of the fecal SCFAs and microbiome

We previously demonstrated that patients with SSc in this cohort had a lower abundance of commensal genera (eg, Faecalibacterium) and a higher abundance of pathobiont genera (eg, Desulfovibrio) compared with unaffected controls. ⁷ Based on the finding that levels of valeric acid and other SCFA discriminated between patients with SSc and unaffected controls, we subsequently investigated whether specific SCFAs were associated with specific bacterial genera in patients with SSc. Our analysis revealed significant associations among butyric acid, propionic acid, caproic acid, and total SCFAs and 18 bacterial genera in patients with SSc, even after adjusting for confounding factors, including age, sex, SIBO, BMI, PPI use, and immunosuppressive therapy (Figure 3; Supplementary Table 1). Notably, three of the bacterial genera that were positively associated with SCFAs were more abundant in patients with SSc compared with unaffected controls (Lachnospira t. ratio = 2.83, P = 0.005, Q = 0.046, effect size 0.43, 95% confidence interval [CI] 0.13–0.73; Lachnoclostridium t. ratio = 1.98, P = 0.049, Q = 0.197, effect size 0.3, 95% CI 0–0.6; and Oscillibacter t. ratio = 2.58, P = 0.011, Q = 0.063, effect size 0.39, 95% CI 0.09–0.69), whereas one of the bacterial genera that was negatively associated with SCFAs was decreased in patients with SSc compared with unaffected controls (Candidatus arthromitus t. ratio = −3.33, P = 0.001, Q = 0.019, effect size = −0.5, 95% CI −0.8 to −0.2) (Figure 4).

Figure 3.

Forest plot showing significant associations (Q < 0.05) between gut bacteria (genus‐level taxa) and SCFAs after adjusting for age, sex, small intestinal bacterial overgrowth, body mass index, proton pump inhibitor, and any immunosuppression. Red dots indicate negative associations, and blue dots indicate positive associations. Bacterial genera that are outlined in green are more abundant in patients with SSc. Bacterial genera that are outlined in orange are more abundant in unaffected controls. CI, confidence interval; SCFA, short‐chain fatty acid.

Figure 4.

Patients with early SSc (red) have an increased abundance of bacteria that are positively associated with short‐chain fatty acids and a decreased abundance of bacteria that are negatively associated with short‐chain fatty acids. SSc, systemic sclerosis.

DISCUSSION

There is a scarcity of disease‐modifying therapies for SSc‐GI involvement, in part because of our limited understanding of the pathogenesis of this clinical dimension of SSc. In the present study, we endeavored to understand the role of GI metabolites in patients with SSc by studying a cohort of patients with relatively early SSc (median disease duration 2 years). We found that patients with early SSc had a significantly altered fecal metabolome compared with that of unaffected controls. Specifically, we found that patients with SSc had higher levels of SCFAs in their stool compared with unaffected controls. In our integrative analysis combining gut microbiome data with fecal metabolome data, we found that SCFAs correlated with specific bacterial genera, several of which were higher in patients with SSc compared with unaffected controls.

To our knowledge, this is the first study to examine the fecal metabolome in patients with early SSc. Two recent studies, however, looked at the level of fatty acids in fecal samples collected from patients with SSc. In a study by Russo and colleagues, investigators observed a significant increase in fecal propionic acid in patients with SSc who were positive for antitopoisomerase antibody and a trend toward increased valeric acid in serum from patients with SSc who were positive for anticentromere antibody. ³⁵ These findings are in agreement with our results. In a study by Bellando‐Randone and colleagues, investigators observed decreased fecal butyrate in patients with VEDOSS and SSc compared with healthy controls. ¹⁶ Although no other studies have looked at the fecal metabolome, several studies have looked at the serum and/or plasma metabolome in patients with SSc. These studies have reported alterations in several metabolites including amino acids as well as metabolites of amino acids. ³⁶

Several studies have looked at the fecal metabolome in other disease states such as IBD and obesity. These studies have highlighted the role of SCFAs in these disease states although they have had conflicting results. In obesity, several studies have found higher levels of SCFAs in obese individuals compared with normal‐weight individuals. ³⁷ , ³⁸ Similarly, in a study by de la Cuesta‐Zuluaga and colleagues in 2019, investigators found that SCFAs were not only associated with obesity but also with lower gut microbiota diversity, dyslipidemia, hypertension, and gut permeability. ³⁹ In contrast, a study by Barczyńska et al found that concentrations of SCFAs were lower in obese children compared with normal‐weight children. ⁴⁰

In IBD, several studies have observed that fecal SCFAs are lower in patients with IBD compared with healthy controls, ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² and supplementation with the SCFA butyrate has been shown to improve inflammation and stool frequency in both mouse models of IBD and patients with IBD. ²³ , ⁴¹ , ⁴² , ⁴³ On the other hand, other groups have found that fecal SCFAs are higher in patients with IBD. ⁴⁴ , ⁴⁵

In the present study, we observed that fecal SCFA levels were higher in patients with early SSc compared with unaffected controls. We also observed that the abundance of several bacterial genera, which were correlated with SCFA levels, were higher in patients with SSc compared with unaffected controls. These findings are intriguing, as SCFAs are generally recognized for their anti‐inflammatory properties in both in vitro and in vivo studies. ¹⁷ There are several possible explanations for these findings. (1) In early SSc, a shift toward SCFA‐producing bacteria might occur as a compensatory response to gut inflammation. (2) Patients with SSc may experience decreased absorption of microbiome‐derived metabolites, leading to higher SCFA levels in stool. (3) The levels of SCFAs in the gut may be carefully maintained in healthy individuals and any increase or decrease in SCFAs may have negative downstream effects. (4) SSc‐induced intestinal dysmotility may result in an enrichment of SCFA‐producing bacteria. Future research is needed to determine which of these possibilities is most likely as well as the mechanisms underlying these findings.

Our study has several important limitations. First, this is a cross‐sectional study and thus does not allow for determining cause versus effect. Second, we were only able to control for sex and age in our multivariable contrast analysis because we did not have detailed clinical data for unaffected control patients. Thus, it is possible that other factors (including diet) may contribute to the differences observed between SSc and unaffected control patients. Third, the participants in the study were recruited from a single center in Sweden and thus the results may not be generalizable to other SSc populations. Lastly, all stool samples underwent two freeze‐thaw cycles (one for microbiome analysis and one for metabolome analysis) and it is possible that metabolites that were particularly sensitive to freeze‐thawing may not have been detected as a result.

Despite these limitations, our study has notable strengths. It included a relatively large sample size of patients with early SSc. We also performed the first integrative analysis of the GI metabolome and microbiome in SSc, which could help inspire future research efforts in this area. In addition, our integrative analysis adjusted for key variables known to influence the GI microbiome including age, sex, SIBO, BMI, PPI use, and any immunosuppression.

In conclusion, our study provides novel insights into the fecal metabolome in early SSc. We identified significant alterations in fecal metabolite profiles, including elevated SCFAs, and demonstrated an association between SCFAs and specific bacterial genera. Although our cross‐sectional design limits causal inferences, this work highlights the importance of studying metabolite‐microbiome interactions to uncover possible mechanisms driving disease progression. Future studies investigating the impact of SCFA supplementation or restriction on disease activity in a mouse model of SSc are warranted to clarify whether the alterations observed in the fecal metabolome contribute to disease pathogenesis or are a consequence of the disease itself.

AUTHOR CONTRIBUTIONS

All authors contributed to at least one of the following manuscript preparation roles: conceptualization AND/OR methodology, software, investigation, formal analysis, data curation, visualization, and validation AND drafting or reviewing/editing the final draft. As corresponding author, Dr Volkmann confirms that all authors have provided the final approval of the version to be published and takes responsibility for the affirmations regarding article submission (eg, not under consideration by another journal), the integrity of the data presented, and the statements regarding compliance with institutional review board/Declaration of Helsinki requirements.

Supporting information

Disclosure form.

Table S1: