Gut Microbiome and Metabolome Were Altered and Strongly Associated With Platelet Count in Adult Patients With Primary Immune Thrombocytopenia

Gut microbiota has been implicated in the pathogenesis of many autoimmune diseases. This is still an area of active research given that the role of gut microbiota on the primary immune thrombocytopenia (ITP) remains unclear. In this study, fecal samples of 30 untreated adult primary ITP patients and 29 healthy controls (HCs) were used to investigate the gut microbial community and metabolite profiles. Our results show that fecal bacteria such as Blautia, Streptococcus, and Lactobacillus are enriched, whereas bacteria such as Bacteroides are depleted in ITP patients. Notably, fecal metabolites such as fatty acyls and glycerophospholipids are enriched and strongly correlate with discrepant gut microbiota. Furthermore, combinations of Weissella and Streptococcus anginosus, or Cer (t18:0/16:0), Cer (d18:1/17:0), and 13-hydroxyoctadecanoic acid could provide good diagnostic markers for ITP. Moreover, a strong negative correlation was found between platelet count and altered gut microbiota such as S. anginosus and gut metabolites such as Cer (t18:0/16:0) in ITP. In conclusion, dysbiosis of both gut microbiota and metabolome develops in ITP patients compared to HCs. Several ITP-altered gut bacteria and metabolites can be diagnostic biomarkers for ITP, and are highly correlated with platelet count, suggesting that they may also play a role in ITP pathogenesis.


INTRODUCTION
Primary immune thrombocytopenia (ITP), formerly known as idiopathic thrombocytopenic purpura, is an autoimmune disorder characterized by peripheral blood platelet count of less than 100 × 10 9 /L excluding conditions known to cause thrombocytopenia, such as infections, other autoimmune disorders, and drug effects (Rodeghiero et al., 2009). The incidence of ITP ranges from 16 to 27 new cases per million, while its prevalence ranges from 45 to 105 per million in adults and nearly 46 per million in children. At least two-thirds of children with ITP recover spontaneously within 6 months, but a relapse is common in adult patients following current standard first-line therapy with corticosteroids (Perera and Garrido, 2016).
Complex multifactorial alterations have been reported in the immune systems of patients with ITP, such as pathologic antiplatelet antibodies (McMillan, 2009;Webster et al., 2006), impaired platelet production (Khodadi et al., 2016), and T cell-mediated effects (Olsson et al., 2003). However, pathogenesis and etiology of ITP are still not fully understood. Genetic and environmental factors play an important role during this process, in a similar fashion to other autoimmune diseases. First, molecular mimicry that has been implicated in pathogenesis of many other autoimmune disorders may also contribute to occurrence of secondary ITP, such as between the glycoprotein 120 of human immunodeficiency virus or the core envelope 1 protein of hepatitis C virus and the platelet GPIIb/IIIa (Bettaieb et al., 1992;Zhang et al., 2009). Nevertheless, mechanisms other than molecular mimicry may dominantly contribute to secondary ITP. For instance, Helicobacter pylori decreases expression of the inhibitory receptor FcγRIIb on monocytes (Kodama et al., 2007). Secondly, ITP was reported to be improved by transplantation of fecal microbiota (Borody et al., 2011). Taken together, this suggests a potential role of gut microbiota in the pathogenesis of primary ITP.
The gut microbiota of many autoimmune diseases was altered and even was identified as one of the etiology. More than 1000 species of bacteria colonize the human gut, and play important roles in health and diseases (Tremaroli and Bäckhed, 2012). Nearly 60% of human immunity comes from the gut. The immune system of germ-free mice is underdeveloped, and their immune cells greatly decline. Recent studies have suggested a link between compositional and functional alterations of gut microbiota and disease manifestations, severity, and responsiveness to treatment (Proal and Marshall, 2018;Russell et al., 2019). Importantly, probiotic supplementation or fecal microbiome transplantation has been shown to be clinically beneficial to patients with autoimmune diseases (Xu et al., 2015;Uusitalo et al., 2016;Kouchaki et al., 2017). In addition, gut microbiota have also been associated with many extraintestinal autoimmune diseases and immune disorders, including rheumatoid arthritis, type 1 diabetes, multiple sclerosis, and system lupus erythematosus (Alkanani et al., 2015;Mu et al., 2015;Jangi et al., 2016;Maeda and Takeda, 2017). However, to the best of our knowledge, there has been no report on the role played by gut microbiota in primary ITP.
In the present study, we analyzed changes of 16S rRNA gene sequencing and liquid chromatography-mass spectrometry (LC-MS) based metabolomics in gut microbiota and fecal metabolic phenotype in patients to explore the gut microbiota characteristics and their role in adult primary ITP.

Study Participants
All primary ITP patients, used in the study, were diagnosed according to the established criteria (Lambert and Gernsheimer, 2017) and did not have other serious diseases or health problems. Stool samples were collected before therapy. Exclusion criteria for both ITP subjects and healthy controls (HCs) were as follows: no antibiotic use in the prior 6 months; no probiotic use; pregnancy or breast-feeding; history of gastrointestinal and cardiovascular diseases, diabetes, metabolic syndrome. Individuals who had a history of bowel surgery, irritable bowel syndrome, or other autoimmune disease were also excluded. Baseline characteristics of all participants are shown in Table 1. The study was approved by the ethics committee of the First Affiliated Hospital of Zhejiang University (reference number: 2018-42). Personal data and stool samples were collected according to the ethical guidelines of the Declaration of Helsinki. All participants signed an informed consent form before being enrolled in the study.

DNA Extraction and Sequencing of 16S rRNA Gene
Sequencing of the 16S rRNA was performed as previously described (Lv et al., 2016). Briefly, bacterial genomic DNA was extracted using the QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, Germany), according to the manufacturer's instructions. The 16S rRNA V3-V4 region was amplified via PCR, and then sequenced using the MiSeq platform (Illumina, San Diego, CA, United States). The raw sequences are available on the NCBI Sequence Read Archive (SRA) database (Accession No. SRP191565).

Sequence Analysis
Paired-end 16S rRNA sequencing data were joined and filtered using FLASH v1.2.10 software, and then chimeric sequences identified and removed using UCHIME (v4.2.40). All sequences were analyzed using QIIME (v1.9.1 1 ), then UPARSE 2 used to perform minor modifications and cluster high-quality sequences into operational taxonomic units (OTUs), using 97% similarity as a cutoff. The taxonomy of each 16S rRNA gene sequence was identified and classified by comparing them to those in the Silva (SSU123) 16S rRNA database with the Ribosomal Database Project (RDP) classifier algorithm 3 using a 70% confidence threshold. Bacterial richness and diversity among the samples were calculated using various indices, such as Chao 1 and Shannon. Differences in the composition of OTUs among different samples were analyzed using principal coordinates analysis (PCoA) plots implemented in R package 4 .

Sample Preparation for Metabolic Profiling
To prepare samples for profiling metabolites, 20 µL of 2-chlorol-phenylalanine (0.3 mg/mL) was first dissolved in methanol to prepare an internal standard. One milliliter of a methanol: water (7: 3 = v: v) mixture was then added to a 60 mg fecal sample and vortexed for 15 s, and left to stand for 5 min. Thereafter, the mixture was shaken for another 15 min and centrifuged for 15 min at 13,000 rpm. The supernatant was collected, filtered through a 0.22-µm microfilter, and pretreated for LC-MS analysis. Aliquots of all samples were pooled for quality control.

LC-MS Analysis
An ACQUITY UHPLC system (Waters Corp., Milford, MA, United States), coupled with an AB SCIEX Triple TOF 5600 System (AB SCIEX, Framingham, MA, United States) were used to analyze the metabolic profiles in both electrospray ionization (ESI) positive and negative ion modes with an ACQUITYTM UPLC BEH C18 column (2.1 × 100 mm, 1.7 µm). The mobile phase consisted of solvents A water (containing 0.1% formic acid, v/v) and B acetonitrile (containing 0.1% formic acid, v/v). Each 5 µL sample was injected for separation using an elution gradient as follows: 5% B for 0 min; 20% B for 2 min; 25% B for 4 min; 60% B for 9 min; 100% B for 17 min; 100% B for 19 min; 5% B for 19 min; and 5% B for 20 min. The process involved a flow rate of 0.4 ml/min, and a column temperature of 45 • C.

Multivariate Analysis of LC-MS Data
Raw LC-MS data, which are available in the "Baidu Netdisk 5 " with extraction code 2tw3, were captured with the progqenesis QI software (Waters Corporation) before analysis. Positive and negative datasets were integrated and analyzed using the SIMCA-P 14.0 software (Umetrics, Umea, Sweden). Metabolic alterations between the two groups were analyzed using principle component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA). Variable importance in the projection (VIP) was used to rank the overall contribution of each variable to the OPLS-DA model, and those variables with VIP > 1 were considered relevant for discriminating HC and ITP patients. Metabolites were identified by progqenesis QI Data Processing Software, based on public databases, including http: //www.hmdb.ca/, http://www.lipidmaps.org/ and self-built ones.

Statistical Analysis
The Mann-Whitney U test was used to compare any two data sets that were abnormally distributed for the clinical measurements [represented as mean ± standard deviation (SD)] as well as alpha diversity of the microbiota and metabolites; otherwise, a one-way analysis of variance (ANOVA) followed by the Student-Newman-Keuls method were performed. Beta diversity was assessed using analysis of similarities (ANOSIM), while the crucial bacterial taxa [represented as mean ± standard error of mean (SEM)] in the two groups were compared using the Wilcoxon rank sum test combined with the Benjamini-Hochberg method (Benjamini and Hochberg, 1995). The relationships between microbiota and metabolites were analyzed using the Spearman's rank correlation test. In addition, the receiver operating characteristic (ROC) curve and area under curve (AUC) were calculated, using the pROC package, to assess diagnostic performance of the model. All data analysis procedures were performed using the SPSS statistical package (version 22.0) and R software (version 3.4.4). A p-value or q-value (false discovery rate adjusted) of less than 0.05 (p ≤ 0.05) were considered statistically significant.

Characteristics of Participants
A total of 46 adult patients were diagnosed with primary ITP, with 16 meeting the exclusion criteria. The remaining 30 were included in the study. Twenty-nine HCs matched for age and sex were also enrolled. Platelet counts, as well as levels of hemoglobin, total protein, globulin, direct bilirubin and creatinine were lower, while those of alanine aminotransferase (ALT) and alanine aminotransferase (AST) were higher in ITP than HCs patients (Table 1).

Differences in Gut Microbiota Between ITP Patients and HCs
From 59 fecal samples, we obtained 4,071,185 sequences following metagenomic sequencing of the V3-V4 region of the 16S rRNA gene. The number of sequences for these samples ranged from 54,487 to 74,946. During identification and classification, about 17 phylum, and 624 species with relative abundances ranging from 10 −7 to 10 −1 , were observed. Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria were the four most predominant phyla, accounting for more than 96% of the total sequences. In the alpha diversity analysis, neither the bacterial community richness (estimated by the Chao1 index) nor the community diversity (characterized by the Shannon index) was significantly different between the groups ( Figure 1A). In beta diversity analysis, PCoA plots at the OTU level showed a separate clustering in microbiota profiles between the ITP and HCs (Figure 1A), with a significant difference (r = 0.14, p = 0.001) in composition of gut microbiota generated by ANOSIM.

Variations in Fecal Metabolomes in ITP Patients and HCs
Fecal metabolomes were analyzed using LC-MS under positive and negative ion modes. OPLS-DA resulted in separate clustering of the ITP and HC groups (Figure 2A), indicating that the fecal metabolomic profiles of these two groups were significantly different. The slope of the straight line was large and the intercept of Q2 was −0.352, indicating that OPLS-DA model was not overfitting ( Figure 2B). A total of 226 differential metabolites, with VIP values greater than 1.0 (p < 0.05) were identified, including 34 known classes and several unclassified compounds. The top ten classes ranked by mean relative abundances were fatty acyls (18.58%), prenol lipids (8.41%), steroids and steroid derivatives (7.94%), glycerophospholipids (7.52%), carboxylic acids (6.94%), sphingolipids (6.64%), steroids and steroid derivatives (6.19%), glycerolipids (5.75%), and organooxygen compounds (4.87%).

Certain ITP-Altered Gut Bacteria and Metabolites Show Good Diagnostic Potentials
The ITP-altered genera, species, and gut metabolites were evaluated for their potential diagnostic ability using ROC analysis. An AUC value higher than 0.85, which indicates good prediction, was used as the threshold during screening.

ITP-Altered Gut Microbiota and Metabolites Closely Correlate With ITP Indicators
Results from the Spearman's correlation showed a predominantly positive correlation between levels of ITP-altered gut bacterial taxa and gut metabolites, at p  (Figure 4B).
Then, we constructed a correlation network comprising ITP-altered bacterial taxa, metabolites and blood parameters using p < 10 −6 . Our results showed platelet count as the center of the network. Two important clusters of correlation networks for microbiota, metabolites and platelet count, except the positive correlation between Bacteroidetes and platelet count mentioned above were evident (Figure 4C). With regards to platelet count, one cluster comprised S. anginosus with Cer (t18:0/16:0) and/or Cer (d18:0/h17:0), while the other had PI [20:1(11Z)/22:4(7Z,10Z,13Z,16Z)] with Streptococcus and/or S. anginosus and/or S. salivarius and with platelet count.

Predicted Metagenomes of Gut Microbiota Vary Between ITP Patients to HCs
To assess potential differences in microbiota functions between HCs and ITP patients, we predicted functional gene compositions in the metagenomes based on Bacterial OTU tables at PICRUSt, a bioinformatics tool with evolutionary modeling and reference genomic databases. Figure 5A shows a total of 28 pathways related to bacterial functions, including xylene degradation, dioxin degradation, and signal transduction mechanisms, were found weakened in gut microbiota of ITP patients. However, 14 pathways, including those related to NOD-like receptor signaling, protein digestion and absorption, glycosphingolipid biosynthesis-ganglio series, glycosaminoglycan degradation and other glycan degradation were strengthened in gut microbiota of ITP patients (Figure 5C).

ITP-Altered Gut Metabolites Were Involved in Inflammation Pathways
To explore the potential contribution of known metabolites to ITP, 226 ITP-altered metabolites were submitted to the KEGG website for pathway analysis. Our results show that ITP-enriched metabolites were mainly related to pathways such as neurotrophin signaling pathway, Th1, Th2, and Th17 cell differentiation (Figure 5B), indicating that enrichment of these metabolites may be potentially linked to inflammation.

DISCUSSION
Increasing research efforts have focused on the relationship between gut microbiota and autoimmune diseases as well as mechanisms involved in their interactions. So far, alterations of gut microbiota in more than 20 types of autoimmune diseases, including type 1 diabetes, multiple sclerosis, rheumatoid arthritis, ankylosing spondylitis, autoimmune hepatitis, and Crohn's disease, have been reported relative to HCs (Alkanani et al., 2015;Zhang et al., 2015;Jangi et al., 2016;Mottawea et al., 2016;Wen et al., 2017;Wei et al., 2019). However, to the best of our knowledge, there has been no report on gut microbiota in patients with ITP. The findings of this study showed that the alterations of ITP gut microbiota were mainly due to enrichment of Lactobacillus, Streptococcus and their members, as well as depletion of Bacteroides and their members. In addition, fecal metabolomes of ITP patients comprised ITP-enriched Cer (t18:0/16:0) and the ITP-depleted lysoPE (14:0/0:0) and were significantly different from those of HCs. Several ITP-altered bacteria such as S. anginosus and ITP-altered metabolites such as Cer (t18:0/16:0) had good diagnosis potential and intensively correlated with platelet count. Moreover, predicted metagenomes of gut microbiota were significantly different from ITP patients relative to HCs. Finally, KEGG analysis showed that several altered metabolites were intensively involved in inflammatory pathways. Overall, these results may contribute to research in pathogenesis research, diagnosis and treatment of ITP.
Alterations of gut microbiota in ITP patients showed similarity to some but also varied from other known autoimmune diseases. Particularly, our results showed that Actinomycetaceae, Lactobacillus, and Streptococcus were enriched, while Bacteroidetes was depleted in ITP patients. In patients with Crohn's disease, gut Bacteroidales were depleted, while Proteobacteria and Veillonella were enriched (Mottawea et al., 2016). In patients with rheumatoid arthritis, Lactobacillus salivarius were enriched, with no observation of Bacteroidetes depletion (Zhang et al., 2015). Patients with primary biliary cirrhosis were found to show enrichment of Actinobacteria, Lactobacillales, Streptococcus, Klebsiella and Veillonella (Lv et al., 2016). Other studies have reported enrichment of Actinobacteria in patients with ankylosing spondylitis, and depletion of Fusobacteria and Verrucomicrobia (Wen et al., 2017) while those with type 1 diabetes, showed a combination of an increased relative abundance of Bacteroidetes together with decreased levels of Firmicutes regardless of geographical location, which is contrary to their alterations in ITP (Siljander et al., 2019). Overall, we infer that the alterations in gut microbiota of ITP patients possess several special characteristics that make them critical to the occurrence and development of the condition. This is despite not listing all the differences in gut microbiota between all known autoimmune diseases and ITP, and the fact that accuracy of the comparisons is limited by possible differences in technology and methods between different studies. Some bacteria, similar to ITP-enriched gut taxa, play important roles in immunity and autoimmunity. First, our results showed Lactobacillus and Streptococcus (including S. anginosus, Streptococcus sp., S. parasanguinis, and S. salivarius) were enriched in ITP patients. The Lactobacillaceae family and its most abundant genus Lactobacillus in ITP patients were significantly higher than in HCs, consistent with the results in patients with severe fever with thrombocytopenia syndrome (SFTS) (Xu et al., 2018). Recent studies revealed regulation of intestinal microbiota in autoimmune arthritis by promoting differentiation and migration of T follicular helper (Tfh) cells (Block et al., 2016;Teng et al., 2016). Lactobacillus was demonstrated to enhance antigen-specific IgA secretion and induce Tfh cells in mice (Satoshi et al., 2018). Most importantly, the involvement of Tfh cells in the pathogenesis of ITP has been well documented (Audia et al., 2017). Therefore, it is speculated that lactobacillus might indeed play a key role in the progression of ITP. However, further validation is needed. Second, S. anginosus is increasingly recognized as an opportunistic pathogen that can cause diseases such as bloodstream infections, bacteremia and invasive pyogenic infections. S. anginosus strain NCTC10713T, showed clear β-hemolysis on blood agar plates. In addition, Streptolysin S produced from this strain exhibited a cytotoxic potential similar to that produced by Streptococcus pyogenes. The result was a peptide hemolysin that exhibits hemolytic activity on erythrocytes as well as cytotoxic activity in cell culture lines in vitro and in vivo using a mouse model (Tabata et al., 2019). Autoimmune disorders, induced by some other members of Streptococcus except S. anginosus, have also been reported. For example, group G Streptococcus induced an autoimmune carditis mediated by IL-17A and interferon gamma in the Lewis rat model of rheumatic heart disease (Sikder et al., 2018). A small fraction of children with S. pyogenes infections was reported to develop Sydenham's chorea or pediatric autoimmune neuropsychiatric disorders (Cutforth et al., 2016). Results of the current study showed that S. anginosus not only increased 35-fold in its relative abundance from ITP patients to HCs, but was also highly negatively correlated to platelet count. Overall, these results demonstrate that ITP-enriched bacteria, such as S. anginosus, may play important roles in ITP occurrence.
Besides, our results showed depletion of Bacteroidetes and increase of Firmicutes/Bacteroidetes ratio in ITP patients relative to HCs. Members of Bacteroidetes, mainly Gram-negative bacteria, mostly colonize the distal gut. They can ferment indigestible polysaccharides, thus producing short-chain fatty acids thereby providing energy for the host (Gibiino et al., 2018). Moreover, they contribute play a role in development and maintenance of host immunity. Deletion of the TLR2 gene (the main receptor of lipopolysaccharide from cell wall of Gram-negative bacteria, including Bacteroidetes) decreased platelet production and numbers (Kovuru et al., 2019). Besides, Firmicutes/Bacteroidetes ratio is considered a representative of health status and may reflect eubiosis of the gastrointestinal tract (Qin et al., 2010).
Gut metabolites can be an intermediary of the crosstalk between gut microbiota and their host. Recently, analysis of fecal metabolomes has unraveled a correlation between metabolic phenotype changes and gut microbiota perturbations during the development of autoimmune conditions, including IBS and Crohn's disease (Jansson et al., 2009;Le Gall et al., 2011). Our results showed a significant difference in fecal metabolic profiles between ITP and those of normal controls. Then, we constructed a correlation network comprising ITPaltered bacterial taxa, metabolites and blood parameters. Our results showed a predominantly correlation between microbiota, metabolites, and platelet count, which suggested the intestinal flora and its metabolites may play a key role in the development of ITP. In addition, PICRUSt analysis revealed that the enriched KEGG pathways of gut microbiota in ITP mostly focused on lipid metabolism. Previous studies demonstrated important role of abnormal lipid metabolism in autoimmune diseases, such as diabetes and multiple sclerosis (Kim, 2017;Kurz et al., 2018). Therefore, we proposed possible involvement of lipids and lipidlike molecules in the onset and development of ITP. Meanwhile, our KEGG-based pathway analysis results showed that ITPaltered metabolites significantly located in Th1/Th2 and Th17 cell differentiation, which is closely related to ITP pathogenesis (Wang et al., 2005;Rocha et al., 2011).
The role of gut microbiota during disease diagnosis has attracted increasing attention. So far, no "gold standard" can be used as a reliable diagnosis for ITP. A presumptive test for this condition is mainly based on exclusion, which largely depends on clinical expertise and observations. A detailed history and examination are essential to ensure elimination of thrombocytopenia from other causes of thrombocytopenia, such as presence of lymphadenopathy, and help identify potential secondary causes (Ejaz and Radia, 2019). Our results showed that several gut bacteria and metabolites, such as S. anginosus and Cer (t18:0/16:0), had great potential as invasive diagnostic biomarkers for ITP. Ceramides have also been described as specific disease biomarkers in other autoimmune diseases, including multiple sclerosis and rheumatoid arthritis (Kosinska et al., 2014;Kurz et al., 2018). However, little is known regarding the potential of S. anginosus for diagnosis of other autoimmune diseases, besides ITP. Combinations of several ITP-altered gut bacteria and/or metabolites from our results, can not only improve ability, but also avoid repetition of a single biomarker for diagnosing other autoimmune diseases. However, the effectiveness of the diagnostic capabilities of these biomarkers will need to be verified using larger studies in future. Furthermore, the excellent diagnosis potential of gut bacteria and metabolites indicate that they are also critical to the occurrence or development of ITP.
Although our investigations attempt to provide a comprehensive insight into the potential contribution of the gut microbiome in ITP, this study has several limitations that must be addressed in future studies. Firstly, this was a single-center, crosssectional study involving a small number of samples. Secondly, although ITP patients were matched for age, sex, and BMI in the analysis, the results may be influenced by other confounding effects, such as dietary factors. Thirdly, the study does not include animal experiments or researches regarding the possible mechanism linking gut microbiota and metabolism to the development of ITP. Therefore, further studies are needed with larger data set of patients and animal experiments to explore the potential causal mechanisms between gut microbiota and ITP.

CONCLUSION
In summary, our study identified alterations of gut metabolome, particularly in lipids, elucidated dysbiosis of gut microbiota characterized by enrichment of Streptococcus, Lactobacillus, and depletion of Bacteroidetes in ITP patients. Moreover, the findings showed that some ITP-altered gut bacteria and metabolites have great potential for diagnosis of ITP. This is based on an evidently positive correlation of these parameters with platelet count, suggesting that they may also play a role in ITP pathogenesis. Identifying the key players from gut microbiota and illustrating their roles and mechanism of action in ITP occurrence and development are further avenues for future research.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/supplementary material.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the ethics committee of the First Affiliated Hospital of Zhejiang University (reference number: 2018-42). The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
WQ conceived and designed this study. XZ, SG, and LL performed the experiments, collected and analyzed the data, and wrote the manuscript. LY, YX, DZ, YC, RY, HJ, and YL collected and analyzed the data. All authors contributed to the article and approved the submitted version.