Disordered intestinal microbes are associated with the activity of Systemic Lupus Erythematosus Running title: Intestinal microbes in systemic lupus erythematosus

Intestinal dysbiosis is implicated in Systemic Lupus Erythematosus (SLE). However, the evidence of gut microbiome changes in SLE is limited, and the association of changed gut microbiome with the activity of SLE, as well as its functional relevance with SLE still remains unknown. Here, we sequenced 16S rRNA amplicon on fecal samples from 40 SLE patients (19 active patients, 21 remissive patients), 20 disease controls (Rheumatoid Arthritis patients), and 22 healthy controls, and investigated the association of functional categories with taxonomic composition by Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt). We demonstrated SLE patients, particularly those active patients, had significant dysbiosis in gut microbiota with reduced bacterial diversity and biased community constitutions. Among the disordered microbiota, the genera Streptococcus, Campylobacter, Veillonella, the species anginosus and dispar, were positively correlated with lupus activity, while the genus Bifidobacterium was negatively associated with disease activity. PICRUSt analysis showed metabolic pathways were different between SLE and healthy controls, and also between active and remissive SLE patients. Moreover, we revealed that a random forest model could distinguish SLE from RA and healthy controls (AUC = 0.792), and another random forest model could well predict the activity of SLE patients (AUC = 0.811). In summary, SLE patients, especially the active patients, show an apparent dysbiosis in gut microbiota and its related metabolic pathways. Among the disordered microflora, 4 genera and 2 species are associated with lupus activity. Furthermore, the random forest models are able to diagnose SLE and predict disease activity.

College of Rheumatology (ACR) classification criteria for SLE or RA disease (18)(19)(20). All patients with acute intercurrent illnesses or infections and those who used probiotics or antibiotics within 1 month before admission were excluded (6). The gender-and age-matched healthy controls (HC) who had no known history of autoimmune diseases were also recruited from the Health Examination Centre of Nanfang Hospital. All the participants were female. Average age of SLE, RA and HC group was 37.46 ± 14.17, 44.00 ± 6.53, and 37.18 ± 14.67 respectively (P = 0.142).
Based on the systemic lupus erythematosus disease activity index (SLEDAI) (21), all the SLE patients were divided into the active SLE patients (A) (SLEDAI ≥ 8) (n = 19) and remissive SLE patients (R) (SLEDAI < 8) (n = 21). Exception of the age and gender distribution, patients in group A showed many significant differences from that of group R, having more severe symptoms, including anemia, hypocomplementemia, impaired renal functions and increased autoantibodies, all of which are consistent with the clinical characteristics of SLE (Table 1).
For all participants, the fresh fecal samples were frozen at -80 °C immediately after collection.
Ethics approval was granted by the Ethics Committee of Nanfang Hospital, and all of the methods used were in accordance with the approved guidelines. Written informed consent was required from all patients and healthy volunteers in the study.

Illumina Miseq sequencing of 16S rRNA gene-based amplicons and data processing
Total DNA was extracted from thawed fecal samples using the LONGSEE STOOL DNA KIT (Longsee med Bio Medicai., LTD., Guangdong, China) following the manufacturer's instructions. All the individually processed human fecal DNA extractions were amplified by polymerase chain reaction (PCR). The forward primer (5'-ACT CCT ACG GGA GGC AGC AG-3') and reverse primer (5'-GGA CTA CHV GGG TWT CTA AT-3') were used to amplify the 16S rRNA gene V3-V4 variable region from the bacteria by polymerase chain reaction (PCR) as described previously (22). Briefly, amplifications were performed using a step cycling protocol consisting of 98 °C for 30 s, 35  For the sequencing of 16S rRNA gene-based amplicons, the amplicon library was prepared using a TruSeq Nano DNA LT Library Prep Kit (Illumina Inc, CA, USA). The sequencing reaction was conducted using Illumina MiSeq platforms and the data were analyzed by the Quantitative Insights Into Microbial Ecology platform (QIIME, www.qiime.org) using the default parameters (23). The raw sequence data for 16S rRNA gene sequencing data sets was available from the Sequence ReadArchive (SRA) database (http://www.ncbi.nlm.nih.gov/sra) at accession number PRJNA493726.
Before assembly, sequence reads were first filtered to remove low-quality or ambiguous reads, including reads lacking exact matching with the primer, sequences with mismatch ratio sequences higher than 0.05 in the overlap region and raw reads shorter than 100 bp with Trimmomatic v.0.32 software (24). Paired-end clean reads were merged using FLASH (25) according to the relationship of the overlap between the paired-end reads when at least 10 of the reads overlapped the read generated from the opposite end of the same DNA fragment, the maximum allowable error ratio of an overlap region of 0.2, and the spliced sequences were called raw tags.
High-quality Sequences with a distance-based similarity of 97% or greater were grouped into operational taxonomic units (OTUs) using the Vsearch algorithm. Representative sequence was then extracted from each OTU. Next, the chimeric sequences were detected and removed. To assign taxonomy information to each clustered feature, extracted representative sequences were subjected to similarity search against Greengenes sequence and taxonomy database using RDP classifier algorithm (ucluster approach with default settings) and the classify-sklearn plugin within Qiime software (version 1.9.1). The phylogenetic relationships were determined based on a representative sequence alignment using Fast-Tree (26). Computation of α-diversity metrics and β-diversity metrics were performed on all samples within the feature table with Qiime diversity alpha/beta plugin. Rarefaction curve plots the number of individual's sample versus the number of species, which was done with Qiime diversity alpha-rarefaction plugin. Rank abundance curve portray relative abundance and species diversity within a community by plotting relative abundance of species (y-axis) against their rank in abundance (x-axis), which plotted using QIIME v.1.9.1 software.

Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) is a
bioinformatics software package designed to predict metagenome functional content from marker gene surveys and full genomes. PICRUSt analysis was performed to identify Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathways, and determine functional categories associated with taxonomic composition (27).
Comparisons of relative abundance of taxa between groups were performed using Linear discriminant analysis Effect Size (LEfSe), a non-parametric Mann-Whitney U test applied to detect features with significant differential abundance with respect to the groups compared, followed by a Linear Discriminant Analysis (LDA) to estimate the effect size of each differentially abundant feature in Linux platform (28).

Statistical analyses
We used the mean (±SD) to express measurement data that obeyed a normal distribution, the median (interquartile range) to express measurement data that obeyed a skewed distribution, and a percentage to express enumeration data. Mann-Whitney U test or Student t test was performed to compare the variables of 2 sample groups. Multiple group comparisons were made by the Kruskal-Wallis test or one-way analysis of variance. False discovery rate (FDR) correction for multiple comparisons was employed, and the statistical power was analyzed via power and sample size calculation in R software (29,30), then the False Discovery Rate q-value was calculated.
The α-diversity determines the species richness and evenness within bacterial populations. The α-diversity metrics include: Observed species and Chao1 (microbial richness), and Shannon index and Simpson index (microbial diversity) (31). The β-diversity determines the shared diversity between bacterial populations. Different distance metrics reveal distinctive views of community structure.
UniFrac distances measure the shared phylogenetic diversity between communities. A smaller UniFrac distance between two samples indicates a higher similarity among the two microbial communities (32).
Principal coordinates analysis (PCoA) was plotted using the package in R software (Version 3.4.4). The Wilcoxon rank sum test was used to determine significance in α-diversity and β-diversity.
We used spearman algorithm to analyze the relationship among microbiota, predicted pathways and SLE activity index. The Random Forest models were trained by "randomForest" package with default parameters in R, then the performance of the model was assessed with a ten-fold cross-validation approach and measured by area under the receiver-operating characteristic (ROC) (33). All tests were

Characteristics of 16S rRNA sequences.
A total of 82 samples were subjected to 16S rRNA sequencing. These samples were composed of three groups including 22 healthy individuals, 20 RA patients and 40 SLE patients. We obtained 2182143 16S rRNA sequencing reads from stool samples of SLE patients, 976140 reads from RA patients and 1277858 reads from HC, which belong to 714 kinds of operational taxonomic unit (OTUs). The parameters, including Chao1 rarefaction curves, Shannon rarefaction curve, and rank abundance of OTUs, were evaluated to confirm the reliability of the sequencing data (Supplementary Figure S1).

Difference of the gut microbiota in SLE patients from those of controls
The α-diversity between two groups was compared using Chao1, Observed species, Shannon index and Simpson diversity indices. Overall, the α-diversity metrics Chao1 and Observed species were significantly higher in healthy controls than in SLE patients (P = 0.038; P = 0.004, respectively), indicating that the gut microbiome in SLE patients exhibited a lower richness than healthy controls ( Figure 1A and B and Supplementary Table S1). However, no difference in Shannon and Simpson index (P = 0.089; P = 0.092, respectively) was observed between SLE patients and healthy individuals, suggesting that the evenness of the gut microbiome of the two groups had no significant difference (Supplementary Table S1). There were no associations between α-diversity and drug treatments, such as Hydroxychloroquine, Glucocorticoid, Cyclophosphamide, and Biological agent (Supplementary Figure   S2).
To measure the extent of the similarity of fecal microbial communities, β-diversity was calculated using unifrac distances. Principal coordinate analysis (PCoA) based on weighted and unweighted UniFrac distance matrix were used for visualizing sample relationships, and ADONIS analysis was used to test the homogeneity of dispersion among different groups. Our results suggested that there were no associations between β-diversity and medicine treatments, including Hydroxychloroquine, Glucocorticoid, Cyclophosphamide, and Biological agent (Supplementary Table S2 and Figure S3), however, the unweighted UniFrac distance analysis of β-diversity difference demonstrated that the structure of microbiota of SLE patients differed from healthy controls (ADONIS analysis, P < 0.001, R 2 = 0.054) ( Figure 1C and Supplementary Table S3). Thus, the microbial diversity was significantly different between SLE group and healthy controls.
We then analyzed the phylum-level profiles of feces between SLE patients and healthy controls. The phylum level profiles for gut microbiota of SLE patients and controls were fairly similar, except for reads from the phyla Fusobacteria (P = 0.027) and Tenericutes (P = 0.002) ( Figure 1D-F). A lower Firmicutes / Bacteroidetes (F/B) ratio was reported in the feces of remissive SLE patients compared to healthy controls (34). However, we showed the ratio of F/B the feces of SLE had a decreasing trend but no significant difference compared with HC group (Supplementary Table S4).
To further determine the phylogenetic clustering pattern between these two groups, the logarithm linear discriminant analysis (LDA) was performed ( Figure 2 Kruskal-Wallis test] with LDA score >2 were shown). Taken together, sequence profiling of the gut microbiota revealed an apparent dysbiosis of the gut microbiota in SLE patients, which was characterized by reduced bacterial α-diversity and biased community constitutions. These results demonstrated the gut microbiota of patients with SLE differed from those of healthy controls.
To investigate whether the disordered intestinal microbes were specific to SLE patients, we further compared the intestinal microflora distribution between SLE and RA patients. There were no significant difference in α-diversity (Supplementary Table S1) and β-diversity (Supplementary Table S3 and Figure   S4C-D) between two groups. LEfSe analysis showed the different microbiota between SLE group and RA group (Supplementary Figure S4) Collectively, these results demonstrated that the gut microbiota of SLE patients differed from healthy individuals, however, there was no significant difference in gut microflora diversities between SLE and RA patients. The genera Streptococcus and Megasphaera were specifically increased in the feces of SLE patients compared with healthy controls and RA patients.

Difference of microbiota profiling in active SLE patients from remissive SLE patients.
Given that the gut microbiota was significantly different between SLE patients and healthy controls, we next investigated whether the gut microbiota was associated with disease activity of SLE. Firstly, we compared 16S rRNA sequences of A group (active SLE patients) with R group (remissive SLE patients).
The unweighted UniFrac distance analysis of β-diversity difference demonstrated that the structure of the microbiota of A group differed from R group (ADONIS analysis, P =0.047, R 2 = 0.039) ( Figure 3A and Supplementary Table S3), while no obvious difference was observed in α-diversity (Supplementary Table S1), suggesting that the community constitutions in A group were distinctly different from R group, but no difference was found in microbial diversity.
As shown in Figure 3, LEfSe analysis further demonstrated that Actinomycetales and Bifidobacteriales from phylum Actinobacteria showed clustered differences, and the genus Bifidobacterium was increased in the feces of remissive SLE patients compared with active SLE patients.
In addition, the species Ruminococcus. gnavus was reduced in the feces of active SLE patients, whereas  Table S4). Altogether, these results indicated that the gut microbiota profiling of active SLE patients were markedly different from that of remissive SLE patients.

patients.
Another emphasis of our study was to disclose the functional variation in the SLE gut microbiota community. Therefore, we predicted the microbiota-derived pathways using the PICRUSt algorithm with the KEGG database and compared functional abundances among the SLE, RA, and HC groups. In total, we characterized six different pathway categories between SLE group and HC group ( Figure 4).
The pathways of Apoptosis and Purine metabolism were significantly increased in SLE patient group compared with HC group ( Figure 4A), while four pathways, including Pathways in cancer, Bacterial chemotaxis, Bacterial motility proteins, and Flagellar assembly, were decreased in SLE patients ( Figure   4B). In addition, nine different functional pathways were identified between A group and R group ( Figure 5). Five were related to Synthesis and degradation of ketone bodies, Apoptosis, Lipid metabolism, Secretion system, and Staphylococcus aureus infection, which were significantly higher in active SLE patients than remissive patients ( Figure 5A). Conversely, Alanine aspartate and glutamate metabolism, Carbohydrate metabolism, Primary bile acid biosynthesis, and Secondary bile acid biosynthesis, were obviously increased in remissive SLE patients compared with active patients ( Figure   5B). However, there was no different pathway between SLE and RA group (date not shown).
We further examined correlations among SLE/HC-associated taxa and disordered functional pathway to obtain an overview of how specific taxa act during metabolic dysfunction in patient gut. For SLE patients, we characterized a positive correlation between the enrichment of Streptococcus and increased Apoptosis pathway (r = 0.807, P < 0.000, FDR < 0.000) and a negative correlation between Streptococcus and Pathways in cancer (r = -0.550, P < 0.000, FDR < 0.000) ( Figure 6A). Further analysis also revealed the active SLE patient-enriched genus Streptococcus was negatively associated with pathways of Alanine aspartate and glutamate metabolism, Primary and secondary bile acid biosynthesis (r = -0.680; r = -0.437; r = -0.434, P < 0.01, FDR < 0.05, respectively) ( Figure 6B), but positively associated with five increased pathways, including Synthesis and degradation of ketone bodies, Apoptosis, Lipid metabolism, Secretion system, and Staphylococcus aureus infection (r = 0.574; r = 0.829; r = 0.406; r = 0.486; r = 0.903, P < 0.01, FDR < 0.05, respectively) ( Figure 6C).
Thus, several aberrant microbiome-associated gut metabolic pathways were associated with SLE using PICRUSt analysis. Interestingly, the SLE-enriched genus Streptococcus was positively associated with the pathways of Apoptosis, the metabolism of lipid, amino acid and bile acid, Secretion system, and pathogenic bacteria infection. The genus Streptococcus, which was specifically associated with the activity of SLE, was related to eight aberrant microbiome-associated pathways ( Figure 6). We further explored whether these eight disordered pathways were also related to the activity of SLE (Supplementary Table S5). Alanine aspartate and glutamate metabolism, Secondary bile acid biosynthesis, and Lipid metabolism were closely associated with SLEDAI (r = -0.376; r = -0.382; r = 0.318, FDR q <0.001, respectively) ( Figure   7G-I). As such we hypothesized that the genus Streptococcus might play an important role in the disease progression of SLE through these three pathways.

Potentials of gut microbiota for SLE diagnosis or disease activity monitoring
Given that the gut microbiota in SLE patients, especially in active SLE patients, had a distinct dysbiosis in microbiota, we next addressed the potential diagnostic value of the gut microbiota as potential biomarkers for SLE by ROC curve analyses. Due to its non-parametric assumptions, random forest was used to detect linear and nonlinear effects and potential taxon-taxon interactions, to identify taxa that could differentiate SLE subjects from control subjects (healthy controls and RA patients), and to discriminate active SLE patients from remissive patients. We used 10-fold cross-validation approach to evaluate the performance of model, and predictive power was scored in ROC analysis. We first made the mode to differentiate the SLE patients from healthy controls and RA patients based on the genus and species levels. We showed that the area under the curve (AUC) was 0.792 (95% CI: 0.750−0.835) Table S6 and Figure 8A), suggesting that the gut microbiota had the potential to diagnose SLE from healthy and disease controls (RA patients). We observed that in the model, out of the top 10 genera and species, 8 belonged to the phylum Firmicutes, 1 belonged to Fusobacteria, and 1 belonged to Actinobacteria. Of the 8 genera in the Firmicutes phylum, 5 were part of Clostridia class, and 3 were Bacilli (Supplementary Table S6). Furthermore, among the 10 genera and species, the mucosa, Lactobacillus, Megasphaera, and Streptococcus were significantly enriched, while  Table S5). Accordingly, most of the genera and species in the model were the disordered genera in the feces of SLE group compared with healthy controls and RA patients.

(Supplementary
We further built another model to distinguish active SLE patients from remissive patients based on the genus and species levels. In this model, the AUC was 0.811 (95% CI: 0.754-0.869) (Supplementary Table S7 and Figure 8B), suggesting that the gut microbiota had the potential to monitor the activity of SLE. Anti-dsDNA was reported to be reasonably sensitive and specific in the diagnosis of SLE, and raised titers of anti-dsDNA along with hypocomplementemia were associated with the activity of SLE (37). We showed that the AUC value for combination of Complement C3 and anti-dsDNA was only 0.773 (95% CI: 0.597-0.949) (Supplementary Figure S6). These results indicated that the combination of the gut microbiota might have a better surveillance value for SLE activity than the combination of Complement C3 and anti-dsDNA. Morever, as shown in the model, out of the top 10 genera and species, Among the 5 genera in the Firmicutes phylum, 3 were from the Clostridia class, 1 was Erysipelotrichi and 1 was Bacilli (Supplementary Table S7). In this case, the Campylobacter, Streptococcus, and Oribacterium were enriched, while the gnavus and Bifidobacterium were reduced in active SLE patients compared with remissive SLE patients ( Figure 3 and Supplementary Table S7). Altogether, a great part of the genera and species in the model were disordered genera in the feces of active SLE patients, suggesting that the disordered intestinal flora might have potential to diagnose SLE, even monitor disease activity.

Discussion
SLE is an autoimmune disease that affects multiple tissues, and causes joint pain, renal disease, muscle pain, fever, poor circulation, inflammation, fatigue, loss of appetite and other symptoms (38). Though the cause of SLE still remains unclear, it is thought to be involved with hormonal, genetic and environmental factors (39). The gut microbiome was believed to be a key factor in influencing predisposition to autoimmunity diseases (40). Recent studies further supported that gut microbiome dysbiosis could act as an important factor in promoting chronic inflammation into autoimmune diseases (2,41,42). However, there were only limited works in exploring the potential relationship of gut microbiome with SLE (8-10, 17, 39). In this study, we have provided new evidence about the gut microbiome dysbiosis in female SLE patients by fecal bacteria sequencing. Importantly, we for the first time explored whether the gut disordered microbes were associated with the activity of SLE.
We investigated the profiling of the gut microbiota and showed a distinct dysbiosis of the gut microbiota in SLE patients, which was characterized by reduced bacterial α-diversity and biased community constitutions. Most of the patients in our study were currently on various immunosuppressants and glucocorticoids treatments. Veena Taneja et al. have demonstrated that RA patients using methotrexate (MTX) and hydroxychloroquine exhibited an increase in species richness and diversity (43). However, our results showed no significant relationship between drug treatments and the abundance diversity of gut microbiota in the SLE patients, which might because the most of enrolled patients were treated with steroids or immunosuppressants, while only 4 patients did not use any drugs.
Phyla Firmicutes together with Bacteroidetes usually account for more than 90% of all phylogenetic species, were involved in host metabolism and immunity (44). In our study, the Phyla Firmicutes and Bacteroidetes occupied the most abundant microorganism, consistent with the typical human intestinal microbiome structures. It was reported that the Firmicutes / Bacteroidetes ratio was significantly lower in the feces of SLE patients in remission (8). However, no significant different for our cohort of remissive SLE patients and healthy controls (P > 0.05). Also, there was no significant difference in the Firmicutes / Bacteroidetes ratio between active SLE and healthy controls (P > 0.05), consistent with the available data (17). The changes of the genera in SLE patients of our study were only partly consistent with previous studies (8,10,17,45), which might partially due to the sample size and geographical locations of patients. It is well known that cohorts with different patient characteristics, including disease stage, geographical locations, diet and status, might exhibit different gut microbiota profiling (15,16,(46)(47)(48). Therefore, the alterations of gut microbiome associated with SLE should display differences among different geographical locations and disease status.
In this study, we found that the abundance of pathogenic genus Streptococcus, with its species anginosus, and genus Megasphaera were significantly enriched in the feces of SLE patients compared with healthy controls; genus Streptococcus and its species anginosus were positively correlated to the activity of SLE. In addition, the genus Veillonella and its species dispar were significantly increased in the gut microbiota of SLE patients compared with RA patients and had a positive association with the activity of SLE. The association of these disordered genera with the activity of SLE was most striking, and to our knowledge, this is the first study to describe such a significant relationship with SLE. The genera Streptococcus and Megasphaera were reported to be closely related to the intestinal disturbance of autoimmune disorders. For example, Streptococcus and Megasphaera were enriched in primary biliary cirrhosis (49) and Pediatric Autoimmune Neuropsychiatric Disorders (50). Also, Streptococcus was relatively increased in RA patients (43). It was demonstrated that S. anginosus rarely caused infections in healthy individuals, but caused infections in the immunodeficient individuals (51). As reported, genera Streptococcus and Veillonella had pro-inflammatory effects. For example, the combination of Streptococci with Veillonella appeared to negate IL-12p70 production, while augment IL-8, IL-6, IL-10, and tumor necrosis factor alpha (TNF-α) response (52).
The SLE patients, especially the active patients, had an increased population of oral bacteria, which is an interesting phenomenon that occurred in the intestinal flora of SLE. However, the gut microbiome of liver cirrhosis, colorectal cancer, RA, and ACVD patients also showed an increase in the abundance of oral bacteria in gut microbiota (53), and only RA and ACVD have been epidemiologically associated with periodontitis. Interestingly, our results suggested that the abundance of genus Streptococcus were enriched in active SLE patients, suggesting that the oral microbiota might be overrepresented in the lower gastrointestinal populations of patients with active SLE. Besides, more severe forms of periodontitis were found in SLE subjects that had higher bacterial loads (54), resulting in an increase in oral bacteria entering the intestine.
Furthermore, our data showed that many beneficial commensal microbes, such as Roseburia, Faecalibacterium and its species prausnitzii were depleted in SLE patients. Meanwhile, the genus Bifidobacterium was adversely correlated with activity of SLE. These microbes belongs to the phylofunctional core of the intestinal microbiota (55,56) , which can produce short chain fatty acids (SCFAs), especially butyrate-acid, to play multiple critical roles in the maintenance of human health, including producing energy components and intestinal epithelial nutrition (57), reducing the severity of inflammation (58), maintaining intestinal barrier functions (59) and enhancing colon motility functions (60).
Moreover, we observed an increased abundance of beneficial commensal genus Lactobacillus and its species Lactobacillus mucosae in the feces of SLE cohort compared to healthy controls.
Supportive of a role for Lactobacilli in the pathogenesis of lupus, taxa in this genus were found to be enriched in female NZB/W F1 mice, the model of systemic lupus. In this study, Lactobacillus spp. were associated with more severe disease, whereas they were reduced as disease is controlled with dexamethasone (17). As reported (61) Lactobacillus. reuteri increased over time in the feces of mice from both lupus models as their disease progress, in addition, Lactobacillus spp. were increased in a longitudinal cohort of SLE patients compared with healthy controls. In this study, the pDC/IFN-promoting properties of L. reuteri in the context of a lupus-prone host suggest a paradigm in which a bacterium that is normally considered a probiotic may become harmful under certain genetic or environmental conditions. We also observed that Lactobacillus were enriched in feces of SLE patients, suggesting a potential role for these taxa in SLE pathogenesis, which need further research in the future.
Our study has demonstrated that some pro-inflammatory bacteria in genera Streptococcus, and Campylobacter expanded, while some anti-inflammatory bacteria in genera Roseburia, Faecalibacterium, and Bifidobacterium reduced in the feces of SLE patients, especially the active patients, resulting in the release of inflammatory factors, then aggravating the systemic inflammation level. Some pro-inflammatory pathogens increased accompanied with the intestinal mucosal barrier compromised, which lead to more bacterial LPS transferring into lymph nodes and blood to stimulate the TOLL-like pathway of the host cells, and produce inflammatory cytokine (62). SLE patients generally used massive immunosuppressive agents and glucocorticoids during the active period, which could inhibit the immune system and might cause a large increase in opportunistic pathogens (63,64).
Notwithstanding, it is questionable whether such changes in gut bacterial profile are a cause or consequence of SLE. However, to posit further on this, is beyond the scope of this study, and we will focus on this in the future research.
In addition, several aberrant microbiome-associated gut metabolic pathways were revealed to be associated with SLE using PICRUSt analysis. We found that SLE patients were enriched in multiple metabolic pathways containing gene functions of Apoptosis, Purine metabolism, and the Apoptosis were positively associated with the genus Streptococcus that was highly enriched in SLE patients and especially in active patients. As reported, Apoptosis pathway played an important role in the pathogenesis of SLE (65). Besides, among these altered pathways, the alanine, aspartate and glutamate metabolism, Secondary bile acid biosynthesis, and Lipid metabolism were not only related to the disease activity, but also significantly associated with Streptococcus. The alanine, aspartate and glutamate metabolism, which was identified to be increased in remissive SLE patients in our study, had been previously reported to play a pivotal role in resting or activated T cells (66). Lipid metabolism participates in the regulation of many cellular processes such as cell growth, proliferation, differentiation, survival, apoptosis, inflammation, motility, etc (67). In active SLE patients, the dyslipidemia was more prevalent, suggesting that inflammation may be related to lipid metabolism (68). Thus, Streptococcus might play an important role in the pathogenesis of SLE through these pathways.
Due to the heterogeneous presentation of SLE patients and their unpredictable disease course, there is a great need for accurate assessment of disease activity. Several immunologic markers including anti-double stranded DNA (dsDNA) antibody and complement are common used in laboratory monitoring of disease activity, however, these traditional biomarkers are better related to certain clinical manifestations of the disease, especially nephritis, rather than to the activity of the disease itself (69).
Currently, disease activity in SLE can be assessed using composite disease activity indices, such as SLEDAI score and British Isles Lupus Assessment Group (BILAG) score (70). However, the composite disease activity indices depend on differential organ involvement and physical assessments (71,72).
Besides, they could be complex for use in routine clinical practice. Thus, there is a great urgent for the identification of new biomarkers that can quantify disease activity (73,74).
Moreover, due to the existence of a remarkable difference in microbiota between SLE status, the random forest models were built in this study to examine whether microbiota composition could identify their disease status. Of note, a random forests model was identified for diagnosing SLE from healthy controls and RA patients with a AUC value of 0.792. To be mentioned, another random forest predictive model showed to be a suitable model for the prediction of disease activity of SLE with the AUC of 0.811, which was higher than the combination of Complement C3 and anti-dsDNA (AUC=0.773). Accordingly, our results suggested that the gut microbiota might be potential biomarkers for diagnosis of SLE and even monitoring SLE activity in a non-invasive method. However, the sample size enrolled in our study was relatively small, therefore, more samples are needed to evaluate the performance of the disordered genera in the future.
In summary, these disordered bacteria and related metabolic pathways might provide clues in studying of the SLE pathogenesis, and in searching for suitable biomarkers for the diagnosis SLE or monitoring SLE activity in a non-invasive method. Specific microbial clades might be viable targets for the therapeutic manipulation by dietary interventions, prebiotics, probiotics and specifically tailored antibiotics. Determining the functions of the microbial clades that expand or contract in SLE will contribute to developing effective strategies to target them. However, the key role of microbiota in SLE pathogenesis and prospective mechanistic studies still need to be further investigated.

Conclusions
In this study, we reveal that intestinal dysbiosis and aberrant metabolism pathways are existed in SLE patients, especially in active SLE patients. Notably, there are 4 disordered genera and 2 species that are associated with the clinical disease activity in our patient cohort. Furthermore, there are two kinds of genera-panels can be the indicators for diagnosing or monitoring disease activity of SLE by random forest algorithm. However, we also recognize the limitations of our study. Since the results are deduced by a single-center study with a relatively small sample size, larger and prospective cohort studies will be required to verify and validate this predictive model.

Supplemental materials
.           A, the active SLE patients; R, the remissive SLE patients. The ADONIS analysis was used to determine significance in β-diversity. * P < 0.05; ** P < 0.01; *** P < 0.001. Wilcoxon rank sum test was used to determine significance. Table S5: Association of disordered genera and aberrant microbiome-associated pathway with activity of SLE. False discovery rate (FDR) correction for multiple comparisons was employed; the False Discovery Rate q-value was then calculated (r coefficient and FDR q-value were indicated for each parameter). * q < 0.05; ** q< 0.01; *** q < 0.001.

Genus and species levels
.