Lenalidomide-related diarrhea is associated with gut microbiota dysbiosis and disruption of the bile acid pool

Abstract Background Lenalidomide is a cornerstone of multiple myeloma (MM) treatment, but frequently causes gastrointestinal (GI) toxicity, which often manifests as diarrhea. Lenalidomide-related diarrhea (LRD) affects quality-of-life and therapeutic tolerance. A case series of 12 patients suggested bile acid (BA) malabsorption is present in LRD (PMID: 25301337). Symptom responsiveness to BA binders (colesevelam) has been described (PMID: 39300066), but the underlying mechanism of LRD is not known. Because microbial metabolism of BAs is integral to GI homeostasis, disruption of the gut microbiota and BA pool can exacerbate many GI diseases. We hypothesized gut microbiota disruption affects BA processing capacity and contributes to LRD. Methods Two prospective trials evaluating single-agent lenalidomide maintenance after front-line MM therapy were performed at our institution (NCT02538198, NCT04497961). Stool and PRO-CTCAE questionaries were collected pre-treatment (baseline) and on lenalidomide (on-len) at regular intervals. Patients submitted additional stool and PRO-CTCAE at diarrhea onset. Dietary food frequency questionaries were also collected. On-len stool samples were overlayed with PRO-CTCAE diarrhea responses and classified as on-len control (Never, Rarely) or on-len diarrhea (Occasionally, Frequently, Almost Constantly). Patients reporting diarrhea at baseline were excluded. 16S rRNA sequencing was performed on stool samples. Alpha-diversity (Simpson's reciprocal index) was modelled with a multivariable generalized estimating equation (GEE). Between-group microbial differences (beta-diversity) were defined by Bray-Curtis distance and assessed with PERMANOVA. FLORAL GEE models were used to select predictive microbial taxa. Stool BAs were quantified by LC-MS/MS and compared using linear mixed-effects models with Tukey-adjusted pairwise comparisons. PICRUSt2 was used to predict metagenomic functions. Dietary nutrient densities were compared with Wilcoxon rank-sum tests, reported as raw p-values and Benjamini-Hochberg adjusted q-values. Analyses used R and Python. Results 72 patients with MM were included. 40/72 (56%) previously underwent autologous hematopoietic cell transplantation (AHCT). 129 stool samples from 72 patients were profiled and classified as follows: baseline (N=54), on-len control (N=58), and on-len diarrhea (N=17). A GEE model (adjusted for AHCT, sex, and antibiotic exposure) revealed significantly lower alpha-diversity in on-len diarrhea samples compared to both baseline (p<0.001) and on-len control (p=0.004). Beta-diversity was significantly shifted in the on-len diarrhea group from both the baseline and on-len control groups (F=3.9, p=0.001). In a FLORAL GEE model comparing baseline and on-len diarrhea groups, Faecalibacterium and Romboutsia genera were depleted in the on-len diarrhea group. Metabolomic profiling of BAs was performed in 103/129 samples. Primary BAs, which promote BA diarrhea, were significantly enriched in on-len diarrhea samples compared to baseline, including cholate (p=0.002) and chenodeoxycholate (p=0.010). Primary BAs are metabolized into secondary BAs by microbial enzymes; secondary BA levels were similar between baseline, on-len control and on-len diarrhea groups, including deoxycholate (overall p=0.679) and lithocholate (overall p=0.250). Despite the higher primary BA load, the predicted abundance of bile salt hydrolase (BSH), which mediates the first microbial step of BA metabolism, was significantly lower in the on-len diarrhea group compared to baseline (p=0.029) and on-len control groups (p=0.030). Predicted abundance of bai operon genes, which mediate downstream processing of secondary BAs, was similar between the three groups (overall p=0.890). Analysis of dietary macronutrient density from 39/72 patients with baseline data revealed those who developed diarrhea had a trend toward higher total fat intake (p=0.059). In an analysis of nutrient subclasses, patients who did not develop diarrhea had trends toward higher intake of soluble fiber (p=0.011; q=0.085) and insoluble fiber (p=0.067; q=0.178). Conclusion Our study provides the first evidence of gut microbiota dysbiosis in patients with LRD. We reveal primary BA enrichment in patients with LRD, which correlated with disrupted microbial processing capacity. These data suggest that beyond bile acid sequestration, strategies to improve gut microbial diversity and functional capacity may warrant further study for mitigation of LRD.