The Association Between the Gut Microbiota and Systemic Lupus Erythematosus, a Meta-Analysis

Evaluate the changes of gut Microbiota in patients with Systemic Lupus Erythematosus (SLE) and healthy people by meta-analysis. We searched the case-control studies of SLE and healthy controls (HCs) for detecting the diversity of gut Microbiota and the abundance level of some microbiota in the two groups. StataMP16 software was applied for this meta-analysis. The Newcastle-Ottawa quality assessment scale (NOS) was used to assess the quality of the included studies.


Introduction
Systemic lupus erythematosus (SLE) is a chronic systemic autoimmune disease with complex clinical manifestations, which is characterized by excessive activation of B cells, T cells and the production of autoantibodies, which can cause injury in multiple organs and tissues [1] . Clinical manifestations include butter y erythema, serous effusion, proteinuria, arthritis, oral ulcer and so on. At present, there is no radical cure, only immunosuppression and immunoregulation. Although the medical community is constantly improving diagnostic methods and treatment strategies, the morbidity and mortality of SLE are still increasing [2][3] .
The pathogenesis of SLE is not clear. It has been proposed in recent years that gut microbiota may play an important role in the pathogenesis of SLE as gut microbiota can help the development of immune system and has close relationship with innate and adaptive immunity [4][5][6] . Recently, studies mainly focus on the mechanism of gut microbiota triggering SLE. Some investigators speculate that the occurrence of SLE may be associated with impaired gut microbes and impaired gut barrier [7] . It may be the pathogenesis of SLE that disrupted gut barrier leads to 'leaky gut', activating immune factors and causing systemic autoimmunity [8] . Another point of view is that molecular simulation is the key to the occurrence of SLE induced by gut microbiota. The orthologs of Ro60 in gut microbiota binds to the B cells and T cells, resulting in cross-reaction, so that it stimulates autoimmune diseases [9] . The incidence of lupus is gender-speci c. Woman have a higher incidence of SLE compared to man [10] . Through the experimental study in mice, it was found that sex hormones may cause the difference of gut microbiota [11][12] .
From that, the disorder of gut microbiota is closely related to the systemic lupus erythematosus, which can be re ected in the changes of diversity of bacteriome and the related microbial abundance. However, the changes of gut microbiota are affected by genetic genes, diet, BMI, region, immunosuppressants and other causes [13][14][15][16] , which may affect the experiments results about the changes of gut microbiota in patients with SLE and healthy people. For example, Marc JanBonder [17] proposed that gene-diet interaction can regulate the abundance of Bi dobacterium. The experimental results of Chiara Bellocchi [18] and Jingquan He [19] showed there was no signi cant difference in Shannon-wiener diversity index and Chao1 richness estimator between SLE and healthy controls (HCs). Weifang Zhu [20] drew a different experimental result that chao1 richness estimator of SLE was higher and the Shannon-wiener diversity index was lower compared with HCs.
Thus, in order to verify the changes of gut microbiota in patients with SLE and the factors that may affect the changes of gut microbiota, we performed a meta-analysis to analyzed the relevant research results around the world, providing an evidence-based medicine basis for follow-up research.
This meta-analysis referred to the Observational Studies in Epidemiology (MOOSE) group, which provide meta guidance for observational studies [21] . This meta-analysis was registered in the International Prospective Record of Systematic Reviews (PROSPERO) on May 17, 2021

Study Selection
The study inclusion criteria for meta-analysis were as follows: (1) population as the research object; (2) the research contents included the changes of gut microbiota diversity or the relative abundance of microbiota between SLE patients and HCs; (3) fecal samples; (4) the experimental results were described by using median, quartile range, average, standard deviation or P value, ensuring the data can be converted. The exclusion criteria were: (1) animal experiment; (2) the research contents were irrelevant; (3) review articles; (4) failure to get complete experimental data.

Data Extraction
Two researchers (STX and YQQ) extract relevant data from the included study and cross-verify it. If any disagreement was raised, it was resolved through discussion and consultation with a third researcher (SHQ). The contents include: (1) the rst author, the year of publication, the country and region in which the study was published; (2) the characteristics of the subjects (sex, average age, population, drugs currently taken and related doses) (3) Experimental methods (diagnostic criteria, sample size, gut microbiota assessment technique); (4) data on alpha diversity and Beta diversity between patients with SLE and HCs. (5) data of the relative abundance of gut microbiota. Since we cannot get in touch with the author of studies, the data we used were all public.

Data conversion
Some studies do not provide relevant experimental data directly, GetDataGraphDigitizer2.25 software (http://getdata-graph-digitizer.com/) was applied to extract the su cient data. Refer to the operation method proposed by Xiang Wan [22] and carry out data conversion.

Quality Assessment
The two researchers (STX and YQQ) independently rated the quality of the included studies and crosschecked it. If encounter disagreement, we sought third-party adjudication. The nine-star Newcastle-Ottawa Quality Assessment Scale (NOS) for case-control studies was used to assess the quality of the included studies [23] . The NOS scale includes seven items: (1) Adequate De nition of Cases; (2) Representativeness of the Cases; (3) Selection of Controls; (4) De nition of Controls; (5) Ascertainment of Exposure; (6) Same method of ascertainment for cases and controls; (7) Non-Response Rate. With a total score of 9. We selected studies with a score of 6-9 points, as high-quality articles, and include them in our research.

Statistical Analysis
The collected or converted data (including Shannon-wiener diversity index, Chao1 richness estimator and the relative abundance of gut microbiota) were sorted out. The number of studies for data were four or more data would be statistically analyzed by Stata software version 16.0. The results were displayed by forest plot. When the data included are continuous variables and the measurement methods are the same, we choose weighted mean differences (WMD) as the effect scale. When there is a large difference in the mean or standard deviation between the included studies, we choose standardized mean differences (SMD) as the effect scale [24] .
Statistical heterogeneity was evaluated using the chi-square-based Q statistic test. I 2 was used as an index to evaluate heterogeneity, of which 25%, 50% and 75% were the boundaries of moderate, large and extreme heterogeneity, respectively. When I 2 > 50%, random effect model was selected for data analysis, while I 2 < 50%, xed effect model was selected for data analysis [25] . In our meta-analysis, I 2 > 25% was taken as the level of statistical signi cance of heterogeneity. The level of signi cance was set at p < 0.05. Subgroup analysis were performed by year of publication, sample size, population, percentage of women in the sample size, and whether or not taking drugs. By deleting one study at a time and combining the effect values of the remaining studies, sensitivity analysis was performed to evaluate the effect of a particular study on the overall results.
Publication bias was evaluated by Egger test [27] , Begg test [28] and funnel chart. Among them, Egger test and Begg test showed that there was no publication bias when P > 0.05, and there was publication bias when P < 0.05. We take the result of Egger test as the reference value when the result of Egger test is inconsistent with that of Begg test. However, due to the limited number of studies included, the published bias assessment results may be not su ciently reliable.

Study selection and Characteristics
A total of 720 studies were retrieved. Deleting duplicate studies by Endnote, 459 articles were remained.
After manual review of the title and abstract, a total of 425 articles which were not related to the research content and whose experimental subjects were animals or reviews were deleted. To further read the full text of the literature, we deleted a total of 3 studies whose experimental results were incomplete, the experimental procedure was unclear, or the sample was non-human feces. In the remaining 31 studies, the data of Shannon-wiener diversity index, Chao1 richness estimator and abundance level of some gut microbiota were analyzed. There was no difference in the experiment results of Beta diversity between patients with SLE and HCs in various studies, so it was not included in the meta-analysis. Among them, twenty-four studies left after deleted eight studies in which the data could not be collected or gut microbiota was described as log10 bacteria per gram of feces or LDA SCORE. After quality evaluation, eleven studies were included in meta-analysis. The systematic search process is summarized in Fig. 1.
A total of eleven studies were included in this study (nine in English and two in Chinese). This metaanalysis involved 373 patients with SLE and 1288 HCs. The experimental samples used in the eleven studies were human feces and all available studies described the gut microbiome by 16S rRNA-gene amplicon sequencing. The diagnostic criteria of SLE patients refer to the criteria set by American College of Rheumatology (ACR). Detailed studies characteristics are showed in Table 1. All the eleven studies were observational case-control studies. According to the NOS scale, we evaluated the quality of each study to ensure that the included studies was of high quality. The evaluation results are shown in Table 2.  [29] .  [34] .  [20] .  [19] .

Meta-Analysis of Standardized Mean Difference
We extracted continuous data from studies for meta-analysis, including the relative abundance levels of Ruminococcaceae, Enterobacteriaceae, Lachnospiraceae, Enterococcaceae, Bacteroides, and alpha diversity of gut microbes (Shannon-wiener diversity index and Chao1 richness estimator) in SLE patients

Subgroup analysis
Removing some of the included studies could not reduce the statistical heterogeneity of Ruminococcaceae and Enterobacteriaceae. In addition, after sensitivity analysis, the heterogeneity of Enterococcaceae still higher than 25%. Subgroup analysis was conducted strati ed by year of publication, the size of the sample, the population, the percentage of women in the sample size, and whether or not to take drugs.
When the sample size (n) was greater than 20 or less than 20 for subgroup analysis, the source of heterogeneity of Enterobacteriaceae could be excluded (Fig. 4A). In the subgroup with n < 20, the relative level of Enterobacteriaceae in patients with SLE was higher than that in the HCs when in the subgroup with n > 20, the relative abundance of Enterobacteriaceae was not different.
The population identi ed subgroups with different results in Ruminococcaceae and Enterococcaceae. The relative abundance of Ruminococcaceae in patients with SLE was lower than that in HCs (SMD=-0.89; 95% CI =-1.34 to -0.45; P = 0.000), but there was no signi cant difference in non-Chinese subgroup (SMD=-0.22; 95% CI=--0.58 to 0.13; P = 0.216). In the subgroup classi ed as Chinese, the abundance level of Enterococcacea in patients with SLE was higher than that in healthy people (SMD = 0.77; 95% CI = 0.34 to 1.21; P = 0.001), while there was no difference between the subgroups classi ed as non-Chinese (SMD=-0.08; 95% CI=-0.49 to 0.32; P = 0.682).

Discussion
At present, the research on the pathogenesis and treatment of SLE is a di cult point in the medical eld. A growing number of studies have demonstrated that human gut microbiota is one of the important factors affecting the development of autoimmune diseases [37] . In patients with SLE, the increase of some gut microbiota can inhibit the production of IL-12p70 and enhance the response of IL-8, IL-6, IL-10 and TNF-α, such as Streptococcus and Veillonella. They lead to pro-in ammatory response [35][36][37][38] . In addition, abnormally enriched, antigen mimicry and metabolic response of gut microbiota can cause the disorder of immune response [19,39] . Thus, we tried to include as many studies as possible to obtain and extracted experiment results of alpha diversity and some gut microbiota abundance between SLE patients and healthy people, analyzing the changes of gut microbiota in SLE patients by meta-analysis.
Our meta-analysis contained eleven case-control studies, comprising three hundred and seventy three SLE patients and one thousand two hundred and eighty eight healthy people, involving ve countries and nine different cities.

Meta-analysis results
Our meta-analysis showed that both Shannon wiener diversity index and Chao1 richness estimator, which are used to measure diversity in speci c areas or ecosystems of the intestine, decreased in patients with SLE. Chao1 richness estimator is an indicator of species richness [40] . Shannon-wiener diversity index is an index to measure the uniformity of intestinal microorganisms [41] . This means the abundance and uniformity of the gut microbiota of patients with SLE are impaired. As a result, the stability of the microcommunity is reduced, the structure and function of the micro-community are weakened, and the ability to resist changes is reduced.
Compared with healthy people, the relative abundance of Ruminococcaceae in patients with SLE decreased. It is an bene cial ora. Ruminococcaceae is belongs to Firmicutes, one of the cellulosedegrading bacteria (CDB). It can produce short-chain fatty acids (SCFA), which considered as potential orchestrators of the cross talk between gut microbiota and the host metabolism [43] . The lower relative abundance of Ruminococcaceae may lead to the imbalance of SCFA production, affecting the metabolism in human body. By studying lupus-prone mice, SCFA can inhibit B cell AID and Blimp1 expression, plasma cell differentiation, autoantibodies' class switching, and prevent IgG1/IgG2 deposition in the kidney and prevent lupus skin lesions [44] . Therefore, the decrease in the relative abundance of Ruminococcaceae may cause some complications in patients with lupus. The study also found that SCFA can protect the integrity of the small intestinal epithelial cell membrane. The decrease of the relative abundance of Ruminococcaceae may also lead to 'leaky gut'.
The results of Meta-analysis showed that the relative abundance of Enterobacteriaceae increased after suffering from SLE. Enterobacteriaceae is one of the families of Proteobacteria, in which most of the ora have pathogenicity and can produce in ammatory reaction. For example, Enterobacteriaceae is the most common pathogen causing abdominal infection, and the extended-spectrum beta-lactamases (ESBLs) production is the main mechanism of its pathogenesis [47] . Studies have shown that high levels of fS100A8-A9 in the intestine of infants can reduce the abundance level of Enterobacteriaceae in the intestine through the expansion of Tregs, and promote the good development of the intestinal micro ora [48] . It can be seen that the abundance level of Enterobacteriaceae is associated with T cells, which may be the mechanism by which changes in the abundance level of Enterobacteriaceae affect the occurrence and development of SLE. Enterobacteriaceae is also one of the main pathogens causing pulmonary infection and lung infection [49] . Increasing relative abundance of Enterobacteriacea with SLE may be one of the factors leading to multiple infections such as lung and abdominal cavity. Enterococcaceae, one of the strains of Firmicutes, is a bene cial micro ora and plays a certain role in alleviating gastrointestinal damage. According to the analysis, the relative abundance of this ora in patients with SLE is higher than that in healthy people. The result might indicate that there may be bacteria can reconstruct gut homeostasis by potential compensatory regulation [42,51] . Furthermore, there was no signi cant difference in Lachnospiraceae and Bacteroides between SLE patients and healthy controls in the included study.

Results of sensitivity analysis and subgroup analysis
Of the eleven studies included, six provided data on the Chao1 index. Weifang Zhu (2018) [20] was a source of heterogeneity through sensitivity analysis. Some non-antibiotic drugs can have an impact on gut microbiota taken by SLE patients, such as glucocorticoids. We found that the study includes SLE patients with newly diagnosed patients (who had not used any drugs) and revisited patients (who had used hormone or immunosuppressive therapy). But it had not recorded the amount of immunosuppressants or hormones used in every SLE patients, while the other studies had recorded that the maximum dosage of hormones used should not exceed 10 mg. It has been reported by Mengchen Guo [34] that the gut microbial of glucocorticoid treatment was similar to that of the HC group. Among the included SLE patients, the dose of prednisone was up to 20 mg, which indicated that glucocorticoid could recovered gut microbiota stability in patients with SLE. In addition, some researchers have found that when the dose of glucocorticoids in SLE patients is greater than 15mg per day with longer duration, the possibility of osteonecrosis of the femoral head will increase [52] . Hence, we suspect that the dose of corticosteroids in patients with SLE may have an impact on the results of the experiment.
After remove the article Chiara Bellocchi (2019) [18] , the heterogeneity of Enterococcaceae meta was eliminated. In this study, 70.37% of the patients were taking hydroxychloroquine (HCQ), while the other three studies did not record that. A short-term treatment with high-dose or long-term use of HCQ can change gut microbiota, resulting in the decrease of relative abundance of Firmicutes [53][54] . Hence, the use of HCQ in patients with SLE may have an effect on the abundance of gut microbiota.
After omitting the study of Taco A. van der Meulen (2019) [29] , the heterogeneity between Lachnospiraceae and Bacteroides was completely eliminated, which shows that this study was the source of heterogeneity of the two analysis results. We found among the thirty SLE patients included in the study, twenty-eight were White/European ethnic background, but none of the healthy controls had White/European ethnic background. In addition, none of the SLE patients were born in the Netherlands, but nine hundred twenty-nine of the nine hundred and sixty ve healthy people in the control group were born in the Netherlands. Except for the in uence of disease on gut microbiota, diet, environment and race also have certain effects on it [55][56] . Therefore, the environmental and ethnic differences between the experimental group and the control group will affect the results. This suggests that the conditional differences between the included controlled studies should be controlled in follow-up clinical studies.
Our subgroup analysis showed that the relative abundance of Ruminococcaceae and Enterococcaceae in patients with SLE may be affected by population. In the Chinese subgroup, Ruminococcaceae showed a signi cant decrease, but in the non-Chinese subgroup, this phenomenon was not prominent [33] . Several investigators analyzed gut microbiota and human genome data, and showed the in uence of lineage on microbial composition [57][58] . Quantitative trait loci (QTLs) analysis was performed on Six hundred and forty-ve cross-line mice, and thirteen genetic loci that were signi cantly related to microbial abundance were detected, further con rmed the close relationship between heredity and intestinal microorganisms. However, some research suggested selective in uence of genes on gut microbiota. For example, JuliaK. Goodrich (2016) [60] . analyzed one thousand one hundred and twenty-six pairs of gut microbiota from British twins, found that the Christensenellaceae was the most highly heritable taxon. It is speculated that Ruminococcaceae and Enterococcaceae families may also be one of the strains with heredity. Therefore, the different abundance level of gut microbiota varies with different populations when suffering from SLE. Meta-analysis showed the prevalence rate of a disease increases with the increase of sample size, which indicates that the sample size has an effect on the experimental results [61] . Subgroup analysis showed there is a certain difference in the meta-analysis results of Enterobacteriaceae when the sample size was less than twenty or it was greater than twenty. This means that the impact of clinical sample size on the results of the study should be taken into account in followup clinical studies. When the sample size is greater than twenty, there was still some heterogeneity.
Increasing the grouping may get satisfactory results. However, due to the lack of studies included in this meta-analysis, it is impossible to carry out further research.

Limitation
Whether the dose range of glucocorticoid therapy can signi cantly change the diversity or abundance of intestinal microorganisms has not been studied. Weifang Zhu [20] studies have shown that there is no signi cant statistical difference in intestinal dominant ora between SLE patients who have not been treated with drugs and SLE patients who have been treated with drugs. However, Mengchen Guo [34] pointed out that glucocorticoid therapy can regulated the balance of gut microbiota. In clinical treatment of systemic lupus erythematosus, there is a large difference in hormone dose between active stage and remission stage, so our meta-analysis lacks the changes of intestinal microorganisms in different stages of the disease, as well as the changes of gut microbiota in different periods.

Conclusion
This study systematically summarized the changes of gut microbiota between patients with SLE and HCs, including the changes of diversity and the relative abundance of some gut microbiota. Through the study, we found that there was a disorder of gut microbiota in patients with SLE, including the decrease of some probiotics and the increase of harmful bacteria, but there may be compensatory micro ora regulating intestinal stability at the same time. This suggests that it can be used as one of the effective methods for the treatment of SLE by modulating the abundance of gut microbiota. In addition, we also studied the sources of heterogeneity through sensitivity or subgroup analysis, so as to provide reference for future clinical research. All data generated or analysed during this study are included in this published article (and its supplementary information les).

Competing interests
The authors declare that they have no competing interests YQ consulted and screened out the literature needed for this research, then evaluated the quality of the literature. YQ also participated in the writing of the paper.
SH acts as the third party to mediate disputes in this study, and was responsible for completing the processing of pictures and tables in the paper.
WY, YB, LJ were responsible for the translation of the paper language.
XH put forward the idea of the paper and relieved the di culties encountered in the process of writing the paper. Figure 1 Flow chart of selection and inclusion of studies in this meta-analysis