Gut Faecalibacterium abundance in patients with plasma cell disorders is associated with survival after autologous HSCT Open Access

Multiple myeloma (MM) is a malignant hematologic disorder in the family of plasma cell neoplasms (PCNs) that requires long-term therapy with median overall survival (OS) of ∼6 years.¹ The gut microbiota (GM) has been shown to affect the immune system and hematopoiesis,² and, as a result, has been studied extensively in the setting of allogeneic hematopoietic stem cell transplantation (HSCT) in patients with hematologic malignancies.³ Recent evidence has associated the GM and factors that affect it, such as diet, antibiotics, and steroids, to MM development and progression from precursor states, response to systemic treatment, and adverse events to therapy.⁴ The composition and role of the GM on outcomes in patients with MM undergoing autologous HSCT remain less known and require further exploration.

Our group previously reported on the loss of diversity in GM composition that occurred early in 33 autologous and allogeneic HSCT recipients (15 of whom had MM), and the importance of a baseline protective GM profile, including expansion of Faecalibacterium, against major infection during transplant.⁵ The genus Faecalibacterium includes the most abundant anaerobic commensal bacterium in the GM, Faecalibacterium prausnitzii, that accounts for >5% of the total bacterial population in healthy adults.⁶ F prausnitzii has been shown to play an important role in regulating the health of the gut by producing anti-inflammatory metabolites.⁶ 

Here, we present results in 83 patients undergoing autologous HSCT for PCN who comprised subset of patients participating in an ongoing prospective phase 3, randomized, single institution clinical trial investigating different types of diets and changes in the microbiome during treatment of patients with various hematologic malignancies. In this substudy, we hypothesized, given the interaction between the GM, the adaptive and innate immune system, and the bone marrow microenvironment,⁷ that preintervention GM and changes in GM early after transplant may be associated with progression-free survival (PFS) after autologous HSCT in patients with PCN. Secondary objectives included evaluating the association of preintervention GM with transplant-associated toxicities and nonprophylactic antibiotic exposure early after transplant. Within the GM, we focused our attention on the genus Faecalibacterium, the most common genus in healthy adults.

The stool samples for this study were collected prospectively between 2017 and 2021 from a subset of patients with MM who were enrolled in the clinical trial (ClinicalTrials.gov identifier: NCT03016130). This interventional trial conducted at the University of Florida comparing a liberalized hospital diet with fresh fruits and/or vegetables with a neutropenic diet in patients undergoing induction or reinduction for hematologic malignancies (including acute myeloid leukemia, myelodysplastic syndrome, and acute lymphoblastic leukemia) or allogeneic or autologous HSCT for any indication. Per protocol, the diet intervention started when the absolute neutrophil count (ANC) was <500 cells per mm³ or at enrollment in patients receiving induction chemotherapy who were not undergoing HSCT if the patient was already experiencing neutropenia. The diet intervention continued until an ANC of ≥500 cells per mm³, discharge from the hospital, transfer to the intensive care unit, or after 30 days of an ANC of <500 cells per mm³ while on the study diet. The study was approved by the institutional review board of the University of Florida.

All patients were aged at least 18 years, with an expected duration of neutropenia (ANC of <500 cells per mm³) of at least 7 days that would be managed in an inpatient setting. Excluded were patients with untreated major infection, uncontrolled invasive fungal infection, human immunodeficiency virus, hepatitis B or C, anti-infective exposure within 7 days of start of the study diet, or unwillingness to eat fresh produce. Patients received levofloxacin prophylaxis starting on the day of transplant until neutrophil recovery (ANC >500 cells per mm³). The development of neutropenic fever prompted initiation of empiric cefepime (preferred antibiotic per institutional standard) and discontinuation of levofloxacin. Major infections were defined as 1 of the following during the time from onset of neutropenia to the end of neutropenia: bloodstream infection, pneumonia, invasive fungal infection, Clostridium difficile infection, or typhlitis.

Most patients in this substudy received standard-of-care therapy for PCN with 4 to 6 cycles of a triplet regimen, with the most common regimens being either lenalidomide/bortezomib/dexamethasone or cyclophosphamide/bortezomib/dexamethasone, followed by high-dose melphalan (200 mg/m²) or reduced-dose melphalan (140 mg/m²), if they were undergoing second transplant or had limiting comorbidities, on day −1. Patients received maintenance therapy, most frequently with lenalidomide, beginning +3 to +4 months after autologous HSCT.

The study stool samples were collected at least at 3 predetermined time points specified by the study protocol: preintervention, mid intervention, and end intervention (intervention referring to the type of diet). Preintervention samples were collected 2 days before or within 5 days of initiation of chemotherapy and/or at the initiation of intervention ± 2 days; mid-intervention samples were collected during neutropenia; and end-intervention samples were collected up to 2 days before or within 5 days after ANC recovery (ANC of >500 cells per mm³), discharge from the hospital, transfer from the transplant unit, or after 30 days of ANC of <500 cells per mm³. Samples were not always available for each patient at each time point. For those patients with multiple samples available at preintervention, the sample collected closest to the time of transplant was selected.

Stool samples were collected per institutional standards using a collection device that separated urine from stool. Samples were placed in a clean specimen collection container (no medium or additive) and stored in the Blood and Transplant/Malignant Hematology Unit in a refrigerator at 37°F for up to 48 hours. Approximately 0.5 to 1.0 mL of sample was then scooped into a 2-mL cryovial (no medium or additive) and transported to the laboratory within 48 hours of collection for processing and storage in a −80°C freezer until analysis.

Fecal DNA was extracted using PowerLyzer PowerSoil DNA isolation kit (Qiagen) via bead beater disruption option as per manufacturer’s recommendation. The V3-V4 hypervariable region of the 16S ribosomal RNA gene was amplified using primer pair 341F (5′-CCTACGGGNGGCWGCAG-3′) and 785R (5′-GACTACHVGGGTATCTAATCC-3′). Both the forward and the reverse primers contained universal Illumina paired-end adapter sequences, as well as unique individual 4- to 8-nucleotide barcodes between the polymerase chain reaction (PCR) primer sequence and the Illumina adapter sequence to allow multiplex sequencing. PCR products were visualized on an agarose gel before and after samples were purified using the Agencourt AMPure XP kit (Beckman Coulter) and quantified by quantitative PCR with the Kapa Library Quantification kit (Kapa Biosystem). Equimolar amounts of samples were then pooled and sequenced (amplicon size of 444 bases) using Illumina MiSeq (paired-end, 2 × 300−base pair reads). For extraction control, ZymoBIOMICS Microbial Community Standard (catalog no. D6300; Zymo) and multiple water controls mixed in between sample extraction were performed. For amplification control, ZymoBIOMICS Microbial Community DNA Standard (catalog no. D6305; Zymo), Escherichia coli genomic DNA along with samples from extraction controls (positive and negative control) were used. Samples were then processed using DADA2 (version 1.26) pipeline.⁸ See the supplemental Methods, which include supplemental Figures 1 and 2, for detailed information.

Response to therapy (partial response, very good partial response, complete response, stringent complete response, or progression of disease) was determined for all patients using the International Myeloma Working Group criteria. Adverse events to therapy were graded (grade 1-5) by the Common Terminology Criteria for Adverse Events (version 4.0), as applicable.⁸ High-risk cytogenetics were defined as ≥1 of the following, t(4;14), t(14;16), or deletion 17p, identified on fluorescence in situ hybridization testing.

When calculating the median time of follow-up for survivors, patients who were alive at 1 month after HSCT were considered survivors. PFS was calculated from the start of the transplant conditioning regimen to progression of disease or last follow-up. OS was calculated from the start of conditioning until death or last follow-up. Kaplan-Meier estimates for PFS and OS were computed.

Data were analyzed by applying different types of statistical models depending on the response. Simple linear regression was used for continuous data and logistic regression for binary response data. To study ordinal responses, we used ordinal logistic regressions. For analyzing the right censored survival outcomes, univariable or multivariable Cox proportional hazard models were applied as pertinent, and 95% confidence intervals (CIs) were reported. The Wilcoxon signed-rank test was used to compare changes in GM diversities from preintervention to mid intervention, and from mid intervention to end intervention. For each patient, Bray-Curtis distances were calculated between the preintervention and end intervention GM samples. To estimate the effect of major infection on β-diversity, we ran permutational multivariate analysis of variance using the “adonis2” function from the R package “vegan.”⁹ Because major infections occurred primarily between the mid-intervention and end-intervention time points, we considered β-diversity among the end-intervention samples. Finally, the variables deemed significant in the univariable analyses were integrated into a multivariable model to facilitate subsequent analysis.

Pearson product moment correlation was used to evaluate the correlation between binary and continuous outcomes.

Analyses were conducted using R version 4.2.3 and GraphPad Prism version 9.5.1 (733) for Windows (GraphPad Software, San Diego, CA; www.graphpad.com). Throughout the analyses, a P value <.05 was defined as statistically significant. Because of the exploratory nature of the analysis of Faecalibacterium abundance and α-diversity with respect to survival, the focus was to limit false negatives, and thus, raw P values are reported. Otherwise, when applicable, the Benjamini-Hochberge corrected P values were used for interpretation and the raw P values are also provided for clarity.

This cohort of patients was derived at the cutoff time point of the first planned interim analysis of the prospective clinical trial; 160 patients were randomized, and 140 patients were evaluable. Of the 140 patients, 84 had PCN. All patients underwent autologous HSCT, except for 1 who underwent allogeneic transplant and was excluded from this study. Table 1 provides the characteristics and transplant details of the 83 patients.

The median time of follow-up for survivors (n = 82) was 32 (range, 0.7-61) months. The median PFS was 40 months, whereas the median OS was not reached (supplemental Figure 3). Diet (liberalized vs neutropenic) was not significantly associated with PFS (hazard ratio [HR], 0.90; 95% CI, 0.45-1.79; P = .8) or OS (HR, 0.96; 95% CI, 0.39-2.38; P > .9) of this group of patients. Whether patients received first vs salvage transplant was also not associated with PFS (HR, 1.12; 95% CI, 0.53-2.35; P = .8) or OS (HR, 0.59; 95% CI, 0.24-1.44; P = .2).

Transplant-related toxicities experienced by patients were low grade (grade 1-2) with the exception of a 6% (n = 5) incidence of grade 3 mucositis and 1.2% (n = 1) incidence of grade 3 diarrhea (Table 1).

The preintervention fecal samples were collected before day 0 of transplant for 26 patients, with median day −1 (range, −3 to −1); on day 0 for 9 patients; and after day 0 for 48 patients with median day +1 (range, +1 to +7; Figure 1). Of note, microbiome samples were not available for 4 patients at the preintervention, 43 patients at mid intervention, and 17 patients at end intervention due to stool not being collected in all but 4 of cases.

Changes in GM α-diversity were assessed at preintervention, mid intervention, and end intervention using the Chao1 index with a steady decrease noted during transplant (P < .01; Figure 2A; supplemental Figure 4A). Faecalibacterium comprised 3.5%, 1.7%, and 2% of the GM at each time point, respectively, with a downward trend in abundance (P < .01; Figure 2B; supplemental Figure 4B).

As detailed in Table 2, analysis of the preintervention, mid-intervention, and end-intervention time points, and the changes in GM α-diversity between the time points with respect to PFS revealed a lack of association. The α-diversity of the samples at each time point and between the 3 time points vs OS also did not reveal any associations. In contrast, the preintervention Faecalibacterium abundance showed a significant association with PFS (HR, 0.92; 95% CI, 0.86-0.99; P = .02). Similarly, there was a significant association of the preintervention Faecalibacterium abundance with OS (HR, 0.88; 95% CI, 0.78-0.98; P = .02). We illustrate the estimated survival for those with high, medium, and low preintervention Faecalibacterium abundance, with higher abundance being associated with superior PFS (Figure 3; supplemental Figure 5). The low, medium, and high Faecalibacterium abundances were quantified as less than 33rd, between 33rd and 67th, and more than 67th percentiles for preintervention Faecalibacterium abundances, respectively.

On univariable analysis, age (<65 vs ≥65 years), sex (male vs female), race (Black vs White vs other), Durie-Salmon stage (I vs II vs III), high-risk cytogenetics (present vs not present), transplant performed as part of frontline therapy vs salvage transplant, dose of melphalan conditioning (200 vs 140 mg/m²), and nonprophylactic antibiotic exposure (yes vs no) were not associated with PFS and receipt of high-dose of melphalan was associated with PFS (HR, 0.38; 95% CI, 0.19-0.75; P = .006; Figure 4A). On multivariable analysis, high-dose melphalan (HR, 0.42; 95% CI, 0.20-0.87; P = .02) and preintervention Faecalibacterium abundance were both significantly associated with PFS (HR, 0.93; 95% CI, 0.86-0.99; P = .04; Figure 4B).

To determine whether the observed association of Faecalibacterium abundance with PFS and OS was merely because of its abundance, we also performed analyses of genera in the GM with propensity or log-normalized preintervention abundance equal to or greater than that of Faecalibacterium (supplemental Table 1). Fourteen other genera were identified, including Blautia, Bacteroides, Lachnoclostridium, Streptococcus, Parabacteroides, Anaerostipes, Ruminococcus, Alistipes, Incertae Sedis, Roseburia, Lachnospira, Eubacterium, Oscillospira, and Rikenella. Supplemental Table 2 lists the associations of the abundance for each genus at preintervention and between time points with PFS and OS. In contrast to Faecalibacterium, none of the other genera at preintervention were associated with PFS or OS.

We investigated whether preintervention α-diversity and Faecalibacterium abundance were associated with common side effects, including mucositis, nausea, vomiting, diarrhea, neutropenic fever, major infection, and engraftment syndrome. As shown in Table 3, the preintervention GM α-diversity was statistically significantly associated with the duration of nausea (P = .009), duration of diarrhea (P = .01), as well as duration of steroid use for the treatment of engraftment syndrome (P = .009). Preintervention Faecalibacterium abundance was significantly associated with the duration of steroid administration for management of engraftment syndrome (P < .001). The remaining associations of the preintervention GM and toxicities were nonsignificant and are presented in Table 3.

We evaluated the relationship between empiric and/or treatment antibiotic exposure during transplant and GM variables. Although all patients received prophylactic antibiotics, exposure to antibiotics for the treatment of suspected or proven infection occurred in 65% of patients (n = 54). Based on clinical indication, in 5% of patients (n = 4), piperacillin/tazobactam was prescribed instead of cefepime, and in 17% of patients (n = 14), vancomycin was added to cefepime. The timing of nonprophylactic antibiotic exposure is noted for each patient in Figure 1. Nonprophylactic antibiotic exposure showed a significant negative correlation with GM α-diversity at mid intervention (r = −0.37; P = .02) as well as at end intervention (r = −0.47; P < .001; Figure 5A-B). The correlation was also statistically significant for change from preintervention to mid intervention (r = −0.40; P = .02) and preintervention to end intervention (r = −0.35; P = .01; Figure 5C-D).

We also computed the correlations between Faecalibacterium abundances and nonprophylactic antibiotic exposure, and they were all nonsignificant. Specifically, the correlations were at mid intervention (r = 0.05; P = .7) and end intervention (r = −0.04; P = .7; Figure 5E-F), as well as from preintervention to mid intervention (r = 0.07; P = .7) and preintervention to end intervention (r = −0.12; P = .3; Figure 5G-H).

The correlation between nonprophylactic antibiotics from preintervention to end intervention and β-diversity was not statistically significant (r = −0.20; P = .1). Here, β-diversities were calculated between preintervention and end-intervention samples for patients with data available at both time points.

Lower preintervention Faecalibacterium abundance was associated with worse PFS and OS. We hypothesized that this possibly occurred because of the influence of the gut microbiome on the bone marrow microenvironment. Various potential mechanisms including bacterial synthesis of short chain fatty acids and use of cytokines have been proposed to explain the interaction between the gut and marrow.^10,11 

Intestinal microbiota dysbiosis caused by factors such as diet, toxic materials exposure, and antibiotics, may lead to different intestinal and metabolic diseases.¹² In this study, members of the genus Faecalibacterium constituted 3.53% of the preintervention GM, which is lower than reported normal levels in a healthy gut likely because all patients had already received myeloma therapy before transplant. The relative abundance decreased by 51% by mid intervention, which was ∼1 week from the start of the conditioning regimen and prophylactic antibiotics, but it started recovering (increased by 9% from trough) by the time of engraftment. A low abundance of members of the Faecalibacterium genus at preintervention was associated with lower survival after transplant. We examined all 14 other genera with abundance at the same or higher level as Faecalibacterium to see whether abundance was responsible for its association with OS and PFS. Although we did not make the same observation in our analysis of the other genera in the GM, members of Lachnospira and Eubacterium are part of the core human GM, and the Lachnospiraceae family has previously been associated with decreased mortality from graft-versus-host disease after allogeneic HSCT.^13,14 The interplay between these organisms and their overall effect on clinical outcomes require further examination in coabundance and co-occurrence networks that capture the dynamic and interactive nature of the human GM, especially during HSCT.^15,16 

In our study, twice as many patients received high-dose melphalan and, as expected, the higher dose conditioning chemotherapy was associated with higher PFS. Higher dose melphalan (200 mg/m²) was associated with better PFS than lower dose melphalan (140 mg/m²) on both univariable and multivariable analyses. The design of this trial does not permit knowing whether the lower PFS in patients with reduced-dose melphalan is because of less antitumor effect with the lower dose or because of the conditions that led to the clinician’s decision for reducing the dose (the need to undergo a second transplant due to recurrent or refractory disease or the presence of limiting comorbidities). On multivariable analysis, after taking into account the dose of melphalan, preintervention Faecalibacterium abundance maintained its independent association with PFS. The decrease in Faecalibacterium levels occurred in the context of overall GM α-diversity decrease early on in the transplant process, similar to that observed in the GM of patients in other studies.^17,18 

Different gut bacteria have been linked to response to various therapies, including chemotherapy, HSCT, and chimeric antigen receptor T-cell therapy in patients with PCN.^19,F prausnitzii has been shown to be 1 of the bacteria associated with measurable residual disease negativity after induction therapy for MM.²⁰ Notably, in a large, prospective study including >300 patients with MM and amyloidosis, F prausnitzii was found in 83 of 99 fecal samples collected on day +9 to day +16 after melphalan followed by autologous HSCT, but its relative abundance was not significantly associated with PFS.²¹ That is consistent with the findings in this study in which the association of Faecalibacterium abundance and survival was observed at preintervention only. Our findings differed, however, in that postintervention α-diversity was not associated with survival, which could be because of difference in exposures, such as type of antibiotic exposure (ciprofloxacin was the preferred antibiotic for prophylaxis on the aforementioned study) or type of diet (some patients received liberal diet on this study).

We found that the preintervention GM α-diversity was strongly associated with transplant-related adverse events, including duration of nausea and length of diarrhea. It also was associated with the length of steroid administration for engraftment syndrome. The preintervention abundance of Faecalibacterium also showed an association with the duration of steroid administration for engraftment syndrome. These findings are in agreement with those in other studies on adverse events after exposure to melphalan administered as part of autologous HSCT in which bacterial abundance was closely related to certain toxicities, for example, high preintervention Bacteroides was associated with reduction in severe diarrhea but high abundance of Blautia and Ruminococcus was associated with increases in severe diarrhea, nausea, and vomiting.¹⁹ 

In this study, neither preintervention GM α-diversity nor preintervention Faecalibacterium abundance predicted the development of neutropenic fever or major infection. Furthermore, nonprophylactic antibiotic exposure administered for suspected or confirmed infection during transplant correlated with reduced mid-intervention and end-intervention GM α-diversity but not with mid-intervention or end-intervention Faecalibacterium abundance or change in preintervention or end-intervention β-diversity. Given that infections typically occur after development of neutropenia at mid intervention, and in our patients, major infection occurred before neutrophil recovery for all but 2 patients who experienced it in the follow-up period, the exposure to empiric and/or treatment antibiotics later during neutropenia is expected to lead to alterations in the GM at the end of transplant. Although antibiotics have been implicated in disturbing the GM composition in autologous and allogeneic HSCT recipients,^3,22 the type of antibiotic matters. In 1 prospective study, piperacillin/tazobactam and vancomycin, but not cefepime, was shown to lead to significant reduction in GM α-diversity 28 days after autologous transplant and vancomycin only in GM β-diversity.²² In our population, although most patients (65%) received cefepime for treatment of neutropenic fever, only a limited number of patients received piperacillin/tazobactam (5%) and/or vancomycin (17%). This might have resulted in nonprophylactic antibiotics having no major impact on the preintervention to end-intervention β-diversity.

Our results are mostly consistent with our previous findings that the shift in GM β-diversity at several clinical time points throughout transplant (baseline, conditioning, neutropenia, fever, and neutrophil recovery) remained independent of the effect of antibiotics.⁵ The difference in the GM α-diversity at mid intervention and end intervention in this study can likely be accounted for by the larger, uniform patient population and type of treatment. In this analysis vs previous analysis, there were 100% vs 45.5% patients with PCN, none vs 24.2% allogeneic HSCT recipients, and a 22.9% vs 33% rate of major infection, respectively. Although patients had fewer major infections overall, melphalan conditioning, received by all patients included here, can increase the risk for infection, especially by enteric bacteria, by affecting the gut mucosal integrity as well as altering not only the diversity and abundance of species in the GM but the function of the GM as well. Melphalan is known to decrease cecal short chain fatty acid levels, resulting in mucosal disequilibrium, alkalization of the luminal environment, proliferation of pathogens such as Enterobacteriaceae, and ultimately leading to acute gut infections and bacteremia requiring empiric or treatment antibiotics.²³ 

Strengths of this study include its prospective nature, and most patients undergoing melphalan plus autologous HSCT as part of frontline therapy, which is the most common clinical scenario in practice and 1 that follows the current National Comprehensive Cancer Network guidelines.²⁴ Limitations include the single-center design of the study, unavailable fecal samples at certain time points, absence of a greater variety of PCNs, lack of measurable residual disease testing on the bone marrow aspirate of patients before and after intervention, and need for validation of the results in a separate cohort. Longer follow-up of patients is required to draw definitive conclusions about the association between the GM and OS. A more in-depth analysis of the impact of diet type on the microbiome along with calorie counts will be reported in detail in a separate publication focusing on the entire study population.

In conclusion, our study showed that GM α-diversity and abundance of the genus Faecalibacterium decreased during autologous HSCT in patients with PCN. Lower preintervention Faecalibacterium abundance was associated with worse PFS and OS. Higher preintervention GM α-diversity and Faecalibacterium abundance were associated with fewer HSCT-induced toxicities. The results point to the potential value of also investigating externally altering the GM before or during transplant to improve clinical outcomes in patients with PCN undergoing HSCT.

The authors thank Precious Williams and Nadine Hill for study coordination and collection of clinical data and samples. In addition, the authors are grateful to XiangYang Lou of the University of Florida (UF) Biostatistics Department for providing guidance with the statistical analysis of this article.

R.Z.G. is supported by UF Health Cancer Center funds.

Contribution: J.R.W., R.Z.G., C.J., and E.A.D. designed the study; E.A.D. and R.Y.L. collected clinical data; R.Z.G., J.G., and C.J. performed the analysis of stool specimens; R.Z.G., A.R., D.M.L., and E.A.D. analyzed the data; E.A.D., R.Z.G., and A.R. wrote the manuscript; and all authors approved the final version of the manuscript.

Conflict-of-interest disclosure: J.R.W. has financial interests (consultancy) in Cidara, Celgene, F2G, Orca, and Takeda, none are related to the subject of this article. The remaining authors declare no competing financial interests.

Correspondence: Erin A. Dean, Division of Hematology and Oncology, Department of Medicine, University of Florida, 1515 SW Archer Rd, Gainesville, FL 32610; email: erin.dean@medicine.ufl.edu.