Identifying Microbiota and Immune Host Factors Associated With Bleeding Risk in Children With Immune Thrombocytopenia

The hallmark of immune thrombocytopenia (ITP) is a decrease in the number of platelets in the blood, leading to excessive bruising and bleeding. While most clinically significant bleeding events in ITP occur at platelet counts < 30,000/uL, patients with comparable platelet counts may have variable bleeding phenotypes. Inflammation has been shown to promote thrombosis through increasing pro-coagulant factors and inhibiting anticoagulant pathways. The gut microbiome is critical in host immune system education and thrombotic pathways, but its role in bleeding risk in ITP has not been studied. For example, lipopolysaccharide (LPS), the outer membrane glycoprotein found in gram-negative bacteria, increases coagulability by activating the toll-like receptor-4 (TLR4)-induced coagulation pathway [1]. In contrast, the gut microbiota-derived metabolite butyrate induces immune tolerance, mitigates inflammation, and minimizes LPS translocation in the intestines [1]. Systemic cytokine production correlates with specific gut microbiota, as evidenced by lower TNF- levels in humans with high levels of Bifidobacterium adolescentis bacteria [1]. We aimed to investigate the role of the microbiome and cytokine production in ITP patients with mild versus moderate–severe bleeding phenotype. We hypothesized that children with ITP with a moderate–severe bleeding phenotype would have increased levels of gut microbiota associated with mitigating local and systemic inflammation and increased anti-inflammatory systemic cytokines, as inflammation has been shown to promote thrombosis.

Our prospective IRB-approved cohort study included patients < 18 years of age with acute ITP within 3 months of diagnosis and with platelet counts ≤ 30,000/uL at diagnosis at the University of Texas Southwestern (UTSW) Medical Center and Children's Health. Fecal and blood samples were obtained in inpatient and outpatient settings and stored de-identified with study-specific sample IDs. Stool samples were obtained within 36 h of inpatient IVIG or steroid initiation and prior to outpatient steroid treatment. 16S rRNA genes (variable region 4, V4) were amplified, sequenced, and analyzed from each sample. Alpha diversity metrics (Simpson diversity, Shannon diversity, Chao1 and Faith richness) were calculated. Output matrices were further analyzed by principal coordinate analysis (PCoA) using weighted UniFrac and Bray–Curtis, along with linear discriminant analysis (LDA) effect size (LEfSe) to identify differences in relative abundance at taxonomic levels. Blood samples were loaded into the MAGPIX system (Luminex corporation, Austin, Texas, USA) for cytokine analysis in the UTSW Genomics and Microarray Core Facility.

We assigned study participants to two groups, mild or moderate–severe, using the Buchanan-Adix bleeding score, with mild participants having a score ≤ 3a and moderate–severe a score ≥ 3b [2]. Score cutoffs were decided based on the most recent American Society of Hematology guidelines for the treatment of ITP [2]. Sample sizes for blood and stool samples were estimated based on previous research comparing cytokine and microbiota data between healthy controls and ITP participants to achieve a power of 80% [3, 4]. Clinical information obtained includes age, gender, ethnicity, antibiotic use within 1 month of diagnosis, ITP treatment, preceding viral infection, history of autoimmune disease, and bleed location.

Thirty-eight patients with ITP were evaluated for inclusion in this study. Eight patients were excluded, two of whom had past oncologic diagnoses, one with ITP > 6 months past diagnosis; five refused to participate. Thirty participants with a median age of 5.5 years (IQR 2–9.5) were included, and blood samples were collected from each participant. 11 (36.7%) were diagnosed with mild phenotype and 19 (63%) with moderate–severe phenotype. The median platelet count in the moderate–severe group was 4 (IQR 4–6), and 6 (IQR 4–15) in the mild group (p = 0.32, Table S1). The median hematocrit in the moderate–severe group was 35.20 (IQR 29.8–37.9) and 33.6 (IQR 30.7–35.8) in the mild group (p = 0.99). The most common presenting bleeding location was mucosal and cutaneous in the moderate–severe group and cutaneous in the mild group. 19 (100%) participants in the moderate–severe group required acute treatment with steroids and/or IVIG, while 2 (18.2%) in the mild group did. Antibiotic use with amoxicillin was seen in 2 (10.5%) participants in the moderate–severe group and 1 (9.1%) in the mild group. No patients had a history of or were diagnosed with an autoimmune disorder during the study.

Stool samples were received for 15 participants, 3 in the mild group and 12 in the severe group. To assess the composition of the gut microbiome of ITP patients, we performed 16S ribosomal RNA (rRNA) gene sequencing analysis (V4 region) on collected stool samples. Alpha diversity (Chao1, Faith, Shannon, and Simpson), the microbial diversity within an individual patient, was not significantly different when comparing mild versus moderate–severe patients (Figure S1A). Beta diversity (Bray-Curtis and weighted Unifrac), that is, the gut microbiome diversity between the two groups, was also not significantly different (Figure S1B). Bacteroidota (particularly the family Prevotellaceae) were enriched in mild patients (Figure 1A), and in contrast, Proteobacteria and the family Enterococcaceae were enriched in the moderate–severe group (Figure 1B). Gut microbiota families Erysipelotrichaceae, Ruminococaceae, and Oscillospiraceae were significantly enriched (≥ two-fold increase in LDA score, p < 0.05 Kruskal–Wallis) in moderate–severe versus mild patients. Mild phenotype patients had increased Megamonas and Fusobacterium taxa compared to moderate–severe counterparts (Figure 1C).

Systemic cytokine profiling revealed a significant increase in IL-1ar in the moderate–severe group compared to those with mild phenotype (p = 0.045, Mann Whitney). In contrast, IL-1β, G-CSF, and IFN-γ were significantly enriched in the moderate–severe group (Figure 1D). Further, a positive correlation between Ruminococcus gut microbiota and IL-1ar cytokine levels was identified (p = 0.086, Spearman correlation).

Our study revealed alterations in microbial composition and cytokine profiles between participants with mild and moderate–severe bleeding phenotypes. We found increased abundance of Ruminococcus microbiota in moderate–severe bleeders, while mild bleeders had increased Megamonas and Fusobacterium taxa. Increased IL-1ra was seen in moderate–severe bleeders and was positively correlated with Ruminococcus taxa. These differences may play a role in modulating bleeding risk in ITP patients.

Interestingly, Ruminococcus gut microbiota were enriched in moderate–severe bleeders. In a recent study, Ruminococcus was found to be decreased in radiation proctitis patients without hematochezia versus with hematochezia, showing a possible role in increasing bleeding risk [5]. Ruminococcus species are known to produce short-chain fatty acid butyrate, which is shown to inhibit the growth of pro-inflammatory gut microbiota and enhance intestinal barrier integrity [1]. In contrast, increased platelet activation and adhesion have been associated with acute inflammatory conditions leading to a higher incidence of thrombosis during inflammation. As such, increased Ruminococcus in moderate–severe bleeders may be associated with a dampened inflammatory response and thus increased bleeding risk at platelet counts similar to those of mild bleeders.

Mild bleeders had increased Megamonas and Fusobacterium taxa in their gut microbiome, all of which are Gram-negative bacteria that contain LPS on their outer membrane. In ulcerative colitis patients, those with mild UC had more Megamonas taxa than those with moderate to severe ulcerative colitis, indicating a possible role in modulating bleeding and disease severity [6]. Neither bacteria produce butyrate, which minimizes LPS translocation in the intestines by improving intestinal barrier integrity [1]. An increase in the Megamonas and Fusobacterium taxa in mild bleeders may be linked to increased clot-forming potential from LPS-induced coagulability and heightened inflammation due to lower butyrate levels leading to decreased bleeding risk compared to moderate–severe bleeders who have higher Gram-positive, butyrate-forming Ruminococcus.

Moderate–severe bleeders also exhibited increased levels of IL-1ra. Cytokines IL-1ra, IL-4, IL-10, IL-11, and IL-13 have been shown to be anti-inflammatory. We also observed a significant increase in IL-1β in moderate–severe bleeders, which is considered a pro-inflammatory cytokine, and significant increases in G-CSF and IFN-γ, the latter of which can exhibit pro-inflammatory or anti-inflammatory properties in a context-dependent fashion. While we did not see a significant increase in other anti-inflammatory cytokines, we noted a positive correlation between IL-1ra and the butyrate-producing Ruminococcus, suggesting a potential role in reducing inflammation in moderate–severe bleeders and concomitantly increasing bleeding risk.

Our study has many strengths, including our exclusive focus on bleeding phenotypes in the pediatric ITP population, a largely under-studied area. Viral infections and antibiotic use have been shown to alter the gut microbiome and affect inflammation; however, only a small group of participants had antibiotic use within a month of sample collection, and antibiotic use appears at similar percentages in both cohorts, making it a non-differential classification or bias. IVIG is also known to affect cytokine levels; thus, treatment in the moderate–severe group prior to sample collection could have influenced cytokine values. While our study is overall powered, the limited availability of stool samples in mild bleeders is a limitation. A larger sample size would be beneficial to increase power in future studies and validate the observations of this pilot study.

In summary, patients with moderate–severe bleeding phenotypes are enriched in anti-inflammatory bacteria and cytokines, such as Ruminococcus spp. and IL-1ra. These bacteria can produce metabolites, specifically short-chain fatty acids such as butyrate, which mitigate inflammation in preclinical models. Mild bleeders, however, have increased pro-inflammatory bacteria. An increase in other anti-inflammatory cytokines in moderate–severe bleeders or any pro-inflammatory cytokines in mild bleeders was not observed, setting the stage for a future study to validate our findings and establish causality. These results show the potential use of the microbiome composition as a biomarker for predicting bleeding severity in ITP and may provide a means for disease prognostication.

Shelly Saini, Andrew Y. Koh, and Ayesha Zia conceptualized, designed, and performed research. Parastoo Sabaeifard, Laura Coughlin, Nicole Poulides, Shuheng Gan, and Xiaowei Zhan analyzed the data. Shelly Saini, Andrew Y. Koh, and Ayesha Zia wrote, and all others critically edited the manuscript.

A.Y.K. received research funding from Novartis. A.Y.K. is a co-founder of Aumenta Biosciences.

IRB approval was obtained from the University of Texas Southwestern Medical Center IRB.

Informed consent was obtained for all patients.