Intestinal microbiome and myelodysplastic syndromes: current state of knowledge and perspectives for future

Abstract

The intestinal microbiome has been mechanistically linked with health and many disease processes. Cancer is no exception. Both in solid tumors and hematologic malignancies, there is increasing evidence supporting the involvement of the intestinal microbiome in tumor development, disease progression, response to treatment, and treatment toxicity. Consistent with microbiome mediation of the immune system and the potent effect of the immune system on cancer, the most compelling evidence has been obtained in the setting of cancer immunotherapy. Here, we review the current state of knowledge about microbiome effects in myelodysplastic syndromes, identify gaps and challenges in related research, and provide insights for future work.

Introduction

Myelodysplastic syndromes (MDS) are a heterogeneous group of clonal hematopoietic dyscrasias, characterized by dysplasia, ineffective hematopoiesis and a potential for transformation into acute myeloid leukemia (AML). Cytopenias, bone marrow dysplasia and a blast percentage of <20% distinguish MDS from other myeloid malignancies. Characterization of the underlying genetic aberrations has not entirely elucidated the pathophysiology of MDS phenotype(s) and their impact on outcomes.

A large number of microorganisms (10¹⁰-10¹¹ per wet gram of stool) live in the colon [1]. This community of microbes is referred to as the intestinal microbiome and their disruptions are generally known as dysbiosis. Intestinal dysbiosis has been linked with various aspects of cancer development, progression, and response to treatment. Here, we review the current state of knowledge about the connection between the intestinal microbiome and MDS.

Intestinal microbiome and cancer

Tumor microenvironment and microbiome

Whether or not tumors have an intrinsic microbiome is controversial. A large analysis suggested distinct intratumor microbiomes in melanoma, breast, lung, pancreas, ovary, bone cancer, and glioblastoma [2]. Tumor-resident bacteria may have relevant functions for the tumor, including its behavior in the face of therapy. For example, Escherichia coli isolates from CRC tumors degraded 5-fluorouracil ex vivo [3], suggesting an impact on local drug efficacy and toxicity. In patients with pancreatic cancer, survival was associated with intra-tumoral CD8+ T-cells directed at neoantigens homologous to microbial epitopes [4]. Even when not within the tumor, the microbiome may influence tumor development. Formation of bacterial biofilms in the colon may promote inflammation and colorectal cancer (CRC) development [5]. Compared to age-matched healthy controls, Flavonifractor plauti and microbial metabolites of tryptophan, bile acid and choline were increased in patients with early-onset CRC. In contrast, late-onset CRC was characterized by Fusobacterium nucleatum enrichment and short-chain fatty acid (SCFA) depletion [6]. SCFAs are microbial metabolic products of dietary fiber fermentation with immunomodulatory functions and regulatory effect on intestinal epithelial cell proliferation, differentiation and barrier function.

The intestinal microbiome can also regulate anti-tumor immune surveillance through augmentation of intra-tumoral lymphocyte populations. In melanoma patients treated with programmed cell death protein 1 (PD-1) blockade, fecal abundance of Faecalibacterium was positively associated with the number of tumor-infiltrating CD8+ and CD4+ effector T-cells, circulating CD8+ T-cells, and progression-free survival (PFS); and negatively associated with circulating regulatory T cells, myeloid-derived suppressor cells, and proinflammatory cytokines [7].

Intestinal microbiome and response to therapy

The intestinal microbiome can influence tumor response to treatment. This has been best demonstrated in the setting of cancer immunotherapy. In large retrospective studies of patients treated with immune checkpoint inhibitors (ICI), antibiotic use before initiating ICI has been associated with worse survival [8, 9]. The strongest direct evidence in solid tumors for intestinal microbiome modulation of immunotherapy response comes from non-small cell lung cancer (NSCLC) and melanoma. However, taxa associated with response (or lack thereof) have not been consistent. In lung cancer, ex vivo cytotoxic and helper T-cell response against Akkermansia muciniphila [8], ex vivo memory T-cell response to Enterococcus hirae [10], higher abundance of this species [8], and higher microbiome diversity [11] correlated with clinical response and survival. Some of the same microbiome characteristics may generalize to other tumors. In one study, for example, a score based on two clusters of intestinal microbial species predicted response to PD-1 blockade in NSCLC patients. The score was then validated in patients with genitourinary cancer treated with ICI [12].

A meta-analysis of five cohorts of melanoma patients treated with PD-1 blockade found that the abundance of members of Actinobacteria phylum and Lachnospiraceae and Ruminococcaceae families was associated with better response, while Gram-negative bacteria and Streptococcaceae were associated with worse response [13]. The SCFA synthesis pathway correlated with better response and survival in these patients [14]. However, the benefit of SCFAs has been equivocal. In one mechanistic murine study combined with patient data, butyrate limited anti-CTLA-4-induced immune effects in mice with melanoma such as dendritic cell maturation, accumulation of tumor-specific T cells and memory T cells. High blood butyrate and propionate levels were associated with resistance to CTLA-4 blockade and higher proportion of Treg cells. In patients with melanoma, low baseline butyrate and propionate levels were associated with better survival [15]. In two small cohorts of patients with PD-1 blockade-refractory melanoma, some patients responded to PD-1 blockade treatment after receiving FMT from responders to PD-1 blockade [16, 17].

Response to chimeric antigen receptor T (CAR-T) therapy has also been associated with antibiotic use, suggesting an effect potentially mediated by the intestinal microbiome. In a large retrospective cohort of patients with B-cell lymphoma and acute lymphoblastic leukemia (ALL) treated with CAR-T, intestinal microbiome diversity before CAR-T infusion was lower than in healthy subjects [18]. In addition, several members of Clostridia and Bacteroidetes were more abundant in responders. In contrast, Veillonellales and Veillonellaceae were associated with worse responses. Use of antibiotics, particularly those with strong anti-anaerobic activity, within 4 weeks prior to cell infusion predicted worse survival. In a multi-center study of B-cell lymphoma patients, broad-spectrum antibiotics given within 3 weeks prior to CAR-T infusion was associated with disease progression and inferior survival [19]. In patients not previously exposed to antibiotics, pre-treatment Bifidobacterium longum abundance was associated with improved survival, while poorer survival was associated with microbiome-derived peptidoglycan and D-galactose synthesis pathways. Bacteroides, Ruminococcus, Eubacterium, and Akkermansia were more abundant in responders. A question often asked is whether exposure to more or broader-spectrum antibiotics might be a sole surrogate of worse baseline health or more aggressive disease, which may independently worsen outcomes. In such a scenario, antibiotic exposures and their associated microbiome disruptions may not be the cause of worse outcomes. One of the studies referenced above examined this question and found that even after adjustment for various markers of patients’ overall health status and disease aggressiveness, the relationship between antibiotic exposures and worse outcomes persisted [18]. The question was evaluated in a different way by another study, where even after excluding patients exposed to high-risk antibiotics (i.e., those associated with worse outcomes), pre-CAR-T intestinal microbiome correlated with outcomes [19]. In another retrospective study [20], patients with B-ALL who received CAR-T experienced higher peak expansion of the product in peripheral blood if they were exposed to oral vancomycin during a month before infusion. In two murine tumor models, mice receiving vancomycin in combination with CD19 CAR-T had better tumor control and cross-presentation of tumor-associated antigens compared with CAR-T alone. Collectively, these findings suggest the effect of antibiotic exposures before CAR-T depends on antibiotic type and its specific spectrum of activity. The results with oral vancomycin suggest that some Gram-positive bacteria may inhibit CAR-T expansion and their elimination before infusion might be beneficial, while the results with other broad-spectrum antibiotics suggest other groups of bacteria stimulate CAR-T expansion and thus their preservation might be beneficial. More controlled murine studies may be needed to identify the specific bacteria involved in each mode of action.

In recipients of allogeneic hematopoietic cell transplantation (alloHCT), one of the most potent immunotherapeutic approaches in hematologic malignancies and the only potentially curative treatment for MDS, higher baseline intestinal microbiome diversity has been associated with lower mortality [21] and Eubacterium limosum abundance in the peri-engraftment period with less relapse [22].

Compared to immunotherapy, less is known about the potential role of the intestinal microbiome in response to chemotherapy. In multiple myeloma patients, one study showed an association between Eubacterium hallii abundance and minimal residual disease negativity after frontline treatment [23]. In a large multicenter study of patients with myeloma, lymphoma or amyloidosis treated with autologous hematopoietic stem cell transplant, peri-engraftment microbial diversity predicted better survival and less progression [24].

Intestinal microbiome and treatment toxicity

Microbiome composition can influence treatment toxicity. The best evidence in the immunotherapy setting comes from patients with melanoma treated with ICIs, patients with hematologic malignancies treated with CAR-T cells, and alloHCT recipients. In the setting of chemotherapy, the best evidence comes from AML patients receiving intensive chemotherapy.

In a meta-analysis of five cohorts of melanoma patients treated with PD-1 blockade, Streptococcus and Lachnospiraceae abundances were associated with immune-related adverse events [13]. In melanoma patients treated with a CTLA-4 blockade, Firmicutes and Faecalibacterium were associated with a higher risk of immune colitis, while Bacteroidetes phylum and its families Bacteroidaceae, Rikenellaceae, and Barnesiellaceae were associated with a lower risk [25, 26].

In patients treated with CAR-T cells for hematologic malignancies, a higher incidence of immune effector cell-associated neurotoxicity syndrome (ICANS) was observed in patients exposed to wide-spectrum antibiotics (particularly those with potent anti-anaerobic activity) pre-CAR infusion [18, 19]. Blautia, Bacteroides, Ruminococcus, Faecalibacterium and Faecalibacterium prausnitzii appeared protective against cytokine release syndrome and ICANS occurrence [18]. The potential mechanisms for these associations, particularly for ICANS, are unknown. However, the gut-brain axis through which the intestinal microbiome can modulate the blood-brain barrier [27] may be implicated.

In alloHCT recipients, reduced diversity and enterococcal/proteobacterial domination were associated with bacteremia risk [28]. In non-Hodgkin lymphoma patients with bacteremia after alloHCT, baseline intestinal microbiome was depleted in Barnesiellaceae, Coriobacteriaceae, Faecalibacterium, Christensenella, Dehalobacterium, Desulfovibrio, and Sutterella [29]. Vancomycin-resistant Enterococcus (VRE) colonization early post-alloHCT was associated with VRE bacteremia during hospitalization [30]. Acute GVHD (aGVHD) and its severity have been associated with lower microbiome diversity [21, 31–33], depletion of obligate anaerobic and butyrogenic bacteria [31, 32, 34–36], domination by individual taxa [31], expansion of Enterococcus and mucolytic bacteria [21, 31, 37, 38], and higher degree of microbiota fluctuations in the peri-engraftment period [33]. In the 20 days prior to aGVHD onset, aGVHD patients had decreased Clostridia, decreased butyrate-producing bacteria, and decreased ratio of strict-to-facultative anaerobic bacteria compared to patients who did not develop GVHD [34]. Fecal bile acids of microbial origin were decreased at the time of aGVHD development [39]. In patients who received mycophenolate mofetil (MMF) for GVHD prophylaxis, the abundance of β-glucuronidase producing bacteria in the first week post-transplant correlated with parameters of MMF pharmacokinetics [40]. β-glucuronidase enhances enterohepatic recirculation of mycophenolic acid by transforming its metabolites back to MPA.

Disruption of the intestinal barrier promotes bacterial translocation and bloodstream infection during intensive chemotherapy. Loss of tonic signaling from the intestinal microbiome to the intestinal epithelium aggravates the injury. In patients with acute leukemia treated with chemotherapy, circulating markers of bacteremia were positively correlated with fecal abundance of Akkermansia and Lactobacillus, and negatively with Clostridium cluster XIV [41]. Loss of intestinal microbiome diversity during anti-leukemia chemotherapy, greater microbiome compositional variably, and higher baseline abundance of Stenotrophomonas, Akkermansia, Pseudobutyrivibrio, Staphylococcus, Streptococcus, and Subdoligranulum predicted a higher risk for infections [42–44]. CPX-351, a liposomal formulation of cytarabine and daunorubicin seems to be an exception in that it fortified mucosal barrier, prevented dysbiosis and improved immune homeostasis in a murine AML model [45]. Mice treated with CPX-351 had fewer signs of intestinal pathology and less bacterial translocation. In addition, CPX-351 ameliorated epithelial damage caused by lipopolysaccharides. FMT from mice treated with CPX-351 restored the microbiome composition and activated the aryl hydrocarbon receptor pathway. This pathway improves gut barrier integrity, decreases local inflammation, and regulates the local microbial community [46–49]. Interleukin-10 (IL-10) increased in dysbiotic mice treated with FMT from CPX-351-treated mice. IL-10 is an anti-inflammatory cytokine, critical in maintaining tolerance to self antigens and the commensal intestinal microbiome [50, 51].

Intestinal microbiome in myeloid malignancies

The intestinal microbiome seems to be altered in patients with AML, even before starting therapy. However, the specifics of a disease-specific signature have not been consistently characterized. Compared to healthy controls, the baseline intestinal microbiome in AML patients had a lower diversity, with decreased Firmicutes, Faecalibacterium, Roseburia, Subdoligranulum and Bifidobacterium, and expansion of Bacteroides, Bacteroidetes, Blautia, Parabacteroides, and typical oral taxa including Actinomyces and Parvimonas micra [52, 53]. The presence of oral taxa in the colon is intriguing as it does not occur in healthy individuals [54, 55]. The clinical significance of this finding and the mechanisms involved are to be examined further. Theoretically, decreased gut colonization resistance due to reduced diversity of the intestinal microbiome and loss of key taxa may facilitate the survival of oral taxa that arrive in the gut via saliva. However, our recent analysis suggested that this mechanism is dispensable for the segregation of oral and gut microbiota [56]. SCFA levels were lower in the colon of patients with AML than healthy adults [52]. In one study, Faecalibacterium abundance was associated with lower disease burden, favorable-risk disease, and better response to treatment [52]. A recent study found evidence for intestinal microbiome involvement in the pathogenesis of AML-related cachexia (inc. anorexia, muscle wasting, inflammation, altered redox status, intestinal dysfunction, and insulin resistance) [53]. In this study, Eubacterium eligens (decreased in patients with AML) and Blautia were positively and negatively associated with muscle strength, respectively. In addition, Eubacterium eligens correlated negatively with systemic inflammatory markers and positively with blood citrulline levels. Citrulline is a marker of enterocyte mass and its decrease indicates intestinal barrier injury. AML patients had increased insulin resistance markers in blood and decreased levels of hippurate. Hippurate is a conjugate of benzoate, a bacterial metabolite, and has been associated with glucose tolerance and enhanced insulin secretion [57, 58].

The intestinal microbiome may contribute to MDS pathogenesis. In treatment-naïve MDS patients, sterile inflammation can perpetuate pyroptosis and cause hematopoietic niche dysfunction [59]. NLRP3 inflammasome can induce gene expression changes, leading to chronic inflammation and contributing to MDS pathogenesis [59–61]. In the MDS bone marrow, multiple overexpressed genes are regulated by Toll-like receptors (TLRs) [62]. TLRs are key receptors of innate immunity and their ligands include antigens from the intestinal microbiome. Intestinal dysbiosis can drive aberrant T-cell differentiation and cancer immune tolerance [63]. A large recent GWAS-based study evaluated the causal relationships of intestinal microbiome alterations with risk and outcome in MDS [64]. Taxa associated with increased risk for MDS included Blautia, Intestinibacter, and Ruminococcaceae UCG003. Those associated with decreased risk were Clostridia, Veillonellaceae, Coprococcus, Lachnospiraceae NK4A136 group, and Clostridiales. Five immunophenotypically characterized circulating cell types were identified as potential mediators between the intestinal microbiome and MDS. In a large study of patients with myeloid malignancies [65], bone marrow and peripheral blood microbiomes at baseline had a different composition from healthy controls. Microbiome composition in these patients was associated with disease subtype (AML, MDS, myeloproliferative neoplasms [MPN], MDS/MPN) and cytogenetic features (normal, complex karyotype, trisomy 8). Patients had a lower microbiome diversity compared to controls, with AML patients showing the lowest microbiome diversity.

Microbiota therapeutics in patients with hematological malignancies

Several approaches have been used to repair intestinal microbiome injury in hematologic malignancies. This includes early antibiotic de-escalation/discontinuation, non-absorbable antibiotics, non-absorbable antibiotic degraders, enteral nutrition, prebiotics, probiotics, and fecal microbiota transplantation (FMT).

During intensive chemotherapy, antibiotics cause intestinal dysbiosis. In a randomized trial of pediatric patients with hematological malignancies receiving chemotherapy or alloHCT recipients with neutropenic fever (NF), empiric antibiotics were stopped in hemodynamically stable patients after 72 hours of defervescence. Clinical outcomes did not differ significantly between the experimental group and controls [66]. A multicenter non-inferiority randomized trial compared shorter duration to the standard duration of carbapenem treatment in patients with hematological malignancies who developed NF after intensive chemotherapy or alloHCT. In the short-term treatment arm, the carbapenem was discontinued after 3 days, regardless of fever. In the extended treatment arm, the carbapenem was continued until neutrophil recovery or a total treatment duration of 9 days, whichever occurred first. If the participant was not afebrile for 5 consecutive days at day 9, treatment duration was extended until this criterion was met, up to a maximum of 14 days. Treatment failure (defined as recurrent fever or a carbapenem-sensitive infection between days 4-9, or septic shock/respiratory failure/death between day 4 and neutrophil recovery) was similar between the two arms, confirming non-inferiority of shorter carbapenem treatment [67].

A retrospective study of alloHCT patients compared two antibacterial regimens used for prophylactic gut decontamination: metronidazole/ciprofloxacin vs. rifaximin (non-absorbable) [68]. Rifaximin was associated with lower Enterococcus abundance and elevated urinary 3-indoxyl sulfate (3-IS), a metabolite of intestinal commensals. 3-IS was negatively associated with transplant-related mortality. An ongoing randomized trial in alloHCT recipients (NCT04692181) is evaluating the addition of oral ribaxamase (SYN-004) to standard NF treatment using beta-lactam antibiotics (meropenem, cefepime, or piperacillin-tazobactam). Ribaxamase is a non-absorbable beta-lactamase which may protect the intestinal microbiome against the fraction of intravenous beta-lactam antibiotics that reach the gut lumen without detracting from the desired systemic effects of antibiotics. Luminal antibiotic degraders, if proven effective and safe, would represent a unique “anti-antibiotic” approach in which antibiotic-related gut decontamination is actively opposed.

Parenteral nutrition promotes the expansion of species that do not depend on dietary fiber as their main nutrient source and can use mucin as an alternative energy resource [69]. In mice, treatment with meropenem aggravated colonic GVHD via the expansion of mucus-degrading Bacteroides thetaiotaomicron [38]. Although enteral intake is expected to preserve the interstitial microbiome better than parenteral nutrition, it remains a controversial topic, with inconsistent findings among studies. The results of two randomized trials have been published. In the first trial, alloHCT recipients receiving total parenteral nutrition during conditioning and 4 weeks post-HCT had better overall survival compared to patients who received hydration with a 5% dextrose solution containing electrolytes, minerals, trace elements, and vitamins [70]. A more recent randomized trial compared outcomes of alloHCT in patients who received predominantly enteral vs. predominantly parenteral nutrition [71]. The clinical outcomes included overall survival, aGVHD, hospitalization length, and need for additional beta-lactam antibiotics. There were no differences in these outcomes between the two arms. Patients who received predominantly enteral nutrition had higher intestinal Faecalibacterium and Ruminococcus bromii abundance.

Prebiotics are substrates such as fermentable dietary fiber used by the microbiome to confer a health benefit to the host. Probiotics, on the other hand, are live organisms with similar potential benefits for the host. In a randomized trial in children with acute leukemia, Lactobacillus rhamnosus GG administration reduced gastrointestinal toxicity of intensive chemotherapy [72]. Probiotic strain Enterococcus faecium M-74 enriched with selenium was a safe probiotic in patients with AML or chronic myeloid leukemia with severe neutropenia, but did not prevent NF [73]. In a feasibility trial, alloHCT recipients were treated with resistant potato starch from day −7 to day +100. Fecal butyrate levels increased during treatment. Treatment was associated with a distinct plasma metabolomic signature, including changes in SCFA levels [74]. Based on these results, an ongoing phase II randomized trial (NCT02763033) will compare incidence of grade II-IV GVHD at day +100 in patients receiving resistant potato starch (experimental arm) between days −7 and day +100 vs. those receiving corn starch as placebo.

FMT has been used in HCT recipients to restore the commensal microbiota, decolonize the gut from drug-resistant organisms, prevent infections, and prevent/treat aGVHD [75]. Patients with multidrug-resistant organisms (MDRO) who received allogeneic FMT before alloHCT had shorter hospitalization, less carbapenem exposure and fewer bloodstream and urinary tract infections than patients with MDRO but without FMT treatment [76]. In another study, FMT recipients with MDRO colonization had similar survival as the non-colonized controls who did not receive FMT, suggesting that FMT mitigated the negative clinical impact of MDRO colonization [77]. A randomized double-blind placebo-controlled trial in AML patients and alloHCT recipients showed amelioration of dysbiosis after FMT, though without a significant impact on infection rate, the primary endpoint of the trial [78]. FMT in this trial improved restoration of alpha diversity and commensal genera such as Collinsella, and reduced the abundance of potential pathogens such as Enterococcus. In another trial, alloHCT recipients with low intestinal abundance of Bacteroidetes received autologous FMT. Compared to alloHCT recipients who did not receive FMT, FMT recipients had an improved alpha diversity and expansion of Lachnospiraceae, Ruminococcaceae, and Bacteroidetes. The stool samples from the control group were enriched in pathways associated with microbial virulence, biofilm formation, and bacterial flagella assembly [79]. In 15 patients with steroid-refractory and steroid-dependent aGVHD, treatment with allogeneic FMT resulted in a complete clinical response in 10 patients. This response was associated with higher microbiome diversity, engraftment of donor taxa, and expansion of some butyrate-producing bacteria [80].

In a phase II, single-arm, multi-center study, 25 AML patients treated with intensive chemotherapy received autologous FMT after neutrophil recovery [81]. The co-primary endpoints were microbiome diversity and abundance of multidrug-resistant bacteria. Microbiome diversity declined from baseline until FMT (~1 month after initiation of chemotherapy). After FMT, microbiome diversity increased and reached baseline levels. The same pattern was seen with Lachnospiraceae and Ruminococcacea, two families enriched in obligate anaerobic butyrogenic commensals. Antibiotic resistance genes expanded during chemotherapy, but declined and reached baseline levels after FMT. Fecal neopterin, a marker of intestinal inflammation, and fecal immunoglobulin A, a marker of local immune activation, increased during induction chemotherapy. Both markers decreased after FMT. Lack of a placebo group made it difficult to distinguish FMT effects from spontaneous microbiome recovery after the initial insults were eliminated.

Concluding remarks

The last decade has witnessed a bloom of microbiome studies in various aspects of cancer ranging from mechanistic murine studies to randomized clinical trials (Fig. 1). A large level of inconsistency in findings from different studies has limited the development of effective microbiota therapeutics in patients with cancer. The largest clinical progress has been in patients with solid tumors receiving immunotherapy, where the intestinal microbiome has been mechanistically linked to both treatment response and toxicity. Even in these settings, the precise mechanisms are not yet fully clear, and some inconsistencies (e.g. SCFAs promoting or inhibiting response to CTLA-4 blockade in melanoma) remain. Tumor microbiome and circulating microbiome remain highly controversial topics. Similarly, the presence of a tumor type-specific signature in the intestinal microbiome is debatable and needs substantially more research. A point of attention when considering the role of microbiome in cancer is tumors such as cervical cancer which are caused by a specific microbiome such as the human papilloma virus. Microbiome refers to the community of microbes rather than a single species. The complex interactions within the microbiome (e.g. cross feeding, mutual antagonism, metabolic cross-talk) may promote single-species effects. Such effects are different from the occasional 1:1 relationship between a microbe and a cancer, where introduction of the microbe in isolation in the lab or in vivo can cause the tumor.

Figure 1:

Connections between intestinal microbiome and cancer

Arrow thickness indicates strength of evidence.

alloHCT: allogeneic hematopoietic cell transplantation; CAR-T: chimeric antigen receptor T cells; CRS: cytokine release syndrome; GVHD: graft-versus-host disease; ICANS: immune effector cell-associated neurotoxicity syndrome, ICI – immune checkpoint inhibitors

Several factors make microbiome research in hematologic malignancies challenging. First, patients often receive therapeutic agents that can affect microbiome composition, including antibiotics, chemotherapy, radiotherapy, and immunotherapy. A true baseline microbiome sample, before exposure to any of these potential sources of microbiome injury is critical to identify a disease signature or to demonstrate a role for the microbiome in disease development. Samples collected while on treatment or post-treatment are still valuable for questions regarding tumor progression, treatment response, and treatment toxicity, but proper statistical adjustments are needed to control for the confounding effects of antibiotics, anti-tumor treatments, and dietary changes that frequently occur, particularly during intensive chemotherapy. Second, the small effect size generally seen with high-quality microbiome studies suggests a need for large sample sizes if a compelling clinical effect is to be observed. Smaller studies are valuable and can be highly informative, but they carry a high risk of false discovery due to the often inevitable multiple testing in microbiome research. However, combining such studies with mechanistic murine studies in controlled settings can improve confidence in the results. Third, while several prospective studies have evaluated microbiome therapeutics in patients with cancer, only a few have been randomized with a placebo arm. To demonstrate the clinical success of microbiome therapeutics in a compelling way, randomization is indispensable. Once safety and feasibility of microbiome therapeutics have been demonstrated, we strongly encourage using a randomized design. Randomization will reduce confounding and help distinguish spontaneous microbiome recovery from treatment-assisted recovery (and their clinical effect).

To date, few studies have examined the potential role of the intestinal microbiome in MDS pathogenesis, progression, and response to treatment. Although not directly evaluated in MDS, the intestinal microbiome influences hematopoiesis [82, 83]. As hematopoietic defects are a hallmark finding in MDS, the potential involvement of intestinal dysbiosis in MDS-related cytopenias needs to be investigated. Importantly, cytopenias are major elements in MDS risk stratification systems. Treatment decisions in MDS are highly dependent on this risk, ranging from observation and transfusion support to alloHCT. Therefore, if a mechanistic link can be established between intestinal dysbiosis and cytopenias in MDS, microbiome therapeutics may “downgrade” MDS risk, potentially leading to less toxic treatments. An important feature of MDS pathogenesis is chronic pro-inflammatory signaling due partly to an altered bone marrow microenvironment and inflammasome activation which in turn induces pyroptosis (pro-inflammatory cell death) and clonal proliferation [61]. Considering the regulatory effect of the intestinal microbiome on systemic immunity and inflammation [84], dysbiosis may be an etiologic or contributory factor in MDS. Several other pathways connecting the intestinal microbiome to the bone marrow microenvironment – the so-called gut-marrow axis – can also be envisioned [85].

As more microbiome research has been done in AML than MDS, it would be helpful to know whether findings in AML can be generalized to MDS. Specifically, significant findings from exploratory AML studies may be used as hypothesis-driven questions in MDS. Stool is one of the easiest biospecimens to collect in humans, and there are abundant data on the intestinal microbiome in health and various disease states. Prospective treatment or observational trials in patients with MDS should collect longitudinal stool samples, from the time of initial diagnosis through transfusion dependence and AML transformation. Pairing microbiome profiles in these samples with concurrent bone marrow and blood samples could identify novel molecules and pathways in MDS pathogenesis and progression that may be targeted by modulating the intestinal microbiome. Finally, efforts to study the intestinal microbiome in MDS can be combined or aligned with those in aplastic anemia. As a bone marrow failure syndrome with great potential for transformation to MDS, aplastic anemia represents a unique disease where microbiome findings may be generalizable between disciplines.