This study revealed a range of protective and harmful microbial taxa. Some taxa, such as members of the order Clostridiales exhibited both positive and negative effects; genera such as Bilophila and Ruminococcus were strongly correlated with increased and decreased risks of anemia, respectively. To summarize these findings, DMP data showed that one class, one order, and three genera were associated with a decreased risk of IDA, whereas three genera were linked to an increased risk. With respect to AA, one genus was associated with a decreased risk; one phylum, one class, three orders, four families, and five genera were associated with an increased risk (Supplementary Table 3). In validation analysis using the MiBioGen consortium data, two genera were associated with an increased risk of IDA; one phylum, one class, one order, and four genera were associated with an increased risk of AA. These findings were consistent with research by other investigators; there were also some novel discoveries. The present study provided genetic evidence of relationships between the gut microbiota and two types of anemia. Below, we describe the findings for each type of anemia.
With respect to IDA, studies have shown a decrease in the phylum Firmicutes among patients who received hemoglobin therapy[15], consistent with our findings; thus, this phylum is commonly identified in studies of IDA. Furthermore, we found that the genus Desulfovibrio, a sulfate-reducing bacterium typically present in the gut, was associated with a decreased risk of IDA. Desulfovibrio can also serve as an opportunistic pathogen in primary bacteremia and abdominal infections, such as abscesses and cholecystitis[34]. One animal study showed that weaned pigs fed a diet containing iron near the upper tolerable limit exhibited a significant increase in Desulfovibrio abundance[35], consistent with our finding that Desulfovibrio is protective against IDA. The genera Pseudoflavonifractor and Ruminococcus, both belonging to the order Clostridiales, demonstrated remarkably opposite effects in the present study. A close relative of Pseudoflavonifractor, the species Flavonifractor plautii, has been linked to arterial function, suggesting that gut microbiota can be targeted for vascular health[36]; however, the specific mechanism remains unknown. In contrast, one study showed that Ruminococcus abundance was higher in naturally recovering IDA patients than in patients who were recovering with medication[37]; this result was consistent with our findings. Additionally, the abundance of Ruminococcus increased with treatment progression in renal anemia patients receiving medication for disease management[38].
Sutterella, isolated from Campylobacter gracilis, has been identified in patients with appendicitis, peritonitis, abdominal inflammation, and peritoneal infection[39], suggesting that Sutterella has a negative impact on health; however, its relationship with IDA remains unclear. In the present study, Streptococcus and its subspecies were associated with IDA, consistent with the proposition of 'Streptococcus therapy' for IDA patients by Chen et al.; their work indicated that oral Streptococcus intake before iron supplementation might help to regulate gut homeostasis[40]. In particular, their study indicated that IDA patients had biological pathways for iron-carrier non-ribosomal peptides, adenosine triphosphate-binding cassette (ABC) transport proteins, and gut microbiota membrane transport that differed from the pathways in healthy individuals, supporting a link between IDA and gut microecology. Moreover, their fecal metagenomics and metabolomics analyses showed a decrease in Bacteroides abundance among IDA patients, which contrasted with our findings linking Bacteroides to IDA; the discrepancy could be related to bias from the use of fecal samples. Additionally, Mandal et al. found a significant increase in Bacteroides abundance among children with severe malaria anemia, validating the findings in a mouse model that simulated a comparable microbial background, and suggesting that a Bacteroides-targeting treatment strategy could be feasible[41].
In recent years, the 'gut-spine axis' concept has been proposed, in which the gut microbiota are assumed to influence spinal health through nutrition, systemic inflammation-related immunoregulation[42]. This axis impacts the bone marrow-specific inflammation[43]. A metabolomic analysis of plasma samples revealed that patients with severe AA had higher abundances of Enterobacteria and Lactobacillaceae, whereas Clostridia, Bacilli, and Candida were less abundant[8]. Our analysis also supported conclusions about Enterobacteria, Lactobacillaceae, and Clostridia, providing new corroborating evidence regarding the relationships of gut microbiota with AA. The order Burkholderiales, which contains the family Pseudomonadaceae, has been observed in antibiotic-treated mice with IDA. Sustained iron supplementation led to an increased abundance of the genus Parasutterella, suggesting an improved prognosis. Considering that the genus Parasutterella was associated with a reduced risk of AA in our study[44], we suspect that there is a causal relationship worthy of exploration Similarly, the family Oxalobacteraceae id.2966 from the same order (Burkholderiales) was linked to a reduced risk of AA, although its involvement in the present study was limited. The order Coriobacteriales, within the phylum Actinomycetota, was associated with a decreased risk of AA in the present study. Surprisingly, the family Coriobacteriaceae from the same order (Coriobacteriales) was associated with an increased risk of AA. Muleviciene et al. found a significant decrease in the relative abundance of the family Coriobacteriaceae among children exhibiting IDA compared with healthy children, which makes these bacteria more intriguing[45]. The species Bilophila wadsworthia, a Gram-negative rod initially isolated from a patient with gangrenous appendicitis, has well-known harmful effects[46]. For example, it promotes cognitive dysfunction in mice on a ketogenic diet by disrupting hippocampal synaptic plasticity, neurogenesis, or gene expression[47]. Its associations with inflammation, such as inflammatory bowel disease, have been substantiated by multiple studies[48], but its associations with AA risk through similar mechanisms have not been explored.
Surprisingly, the order Clostridiales showed a strong association with both diseases, suggesting that it has a more fundamental connection with anemia pathogenesis, rather than the symptoms of disease. Shao et al. also found a lower abundance of the order Clostridiales in severe AA patients than in healthy individuals[8], potentially related to the immunodeficiency that is characteristic of severe AA. In a study by Phipps et al., colorectal cancer patients with concurrent IDA showed a significantly increased abundance of the order Clostridiales within non-tumor microbiota after oral iron therapy[49]. A common factor potentially linking various pathologies is butyrate, a substance collectively produced by Clostridiales. In studies focused on anemic patients with Inflammatory Bowel Disease, a disruption in the production pathways of butyrate was observed. Butyrate plays a crucial role in regulating iron metabolism. It upregulates the expression of the iron transporter protein in macrophages. This regulation occurs through the reduction of histone deacetylase enrichment, which in turn promotes iron efflux[50]. Research has shown that supplementing with butyrate can effectively restore iron metabolism in IDA models. Additionally, butyrate exhibits a significant inverse relationship with graft-versus-host disease in the context of Allogeneic Hematopoietic Stem Cell Transplantation treatments for AA. This effect is mediated through the activation of GPR43, which leads to the activation of inflammasomes in intestinal epithelial cells, culminating in the secretion of local cytokines, including IL-18, and subsequently influencing the prognosis of AA[51]. Although these studies suggest that this bacterial order is strongly correlated with both types of anemia, the detailed mechanistic links have not been described; they warrant further investigation.
For our MR analysis, we selected exposure data from both the DMP (i.e., Lopera-Maya et al.) and the MiBioGen consortium GWAS microbiota association studies. When the two databases were compared with respect to efficacy, the DMP data exhibited greater statistical significance and greater power. As the largest available species-level database, the DMP contains 412 taxa-related entries (including subspecies information), compared with the 211 taxa in the MiBioGen consortium GWAS. The species-specific information in the DMP helps to avoid heterogeneous effects involving the same genus but different species, enhancing the credibility of our statistical findings. Moreover, the acquisition of results at the phylum, class, order, family, and genus levels from published literature improved the associations identified in the present study.
Several limitations of the present study must be acknowledged. First, the participants predominantly had European ethnicities, which may reduce the generalizability of the findings to the global population. Second, anemia can be cytologically classified as macrocytic, normocytic, or microcytic[52]; because of missing GWAS data, the present study did not include macrocytic anemia. Nonetheless, the two types of anemia studied are sufficiently representative (IDA is microcytic, whereas AA is normocytic). Third, the associations identified by multiple comparisons in this study only suggest causality; they do not definitively establish the directions of causal relationships. Fourth, considering the completeness of the statistical data, we did not adjust P-values across statistical methods to control for Type I errors related to multiple testing. Finally, not all selected taxa were significant across all statistical methods. Although we utilized several methods to evaluate and exclude potential heterogeneity or pleiotropy, the presence of unidentified heterogeneity or pleiotropy cannot be conclusively ruled out.