Prior Antibiotic Exposure Correlates with Unfavorable Efficacy and Prognosis of B-NHL after CAR-T Cell Therapy

Background: The chimeric antigen receptor (CAR) -T cell has greatly changed the therapeutic landscape of B-cell malignancies, but treatment failure and cytokine release syndrome (CRS) remain its obstacles. Previous studies demonstrated that gut microbiota could affect the effectiveness and toxicities of CAR-T cells. In this study, we hypothesized that the previous antibiotics exposure would impact the efficacy and safety of CAR-T cells by disturbing the gut microbiota. Method: We collected data from 89 patients treated with the humanized CD19 CAR-T cells in our center (ChiCTR-OIC-17011180), and retrospectively analyzed the relations between antibiotic exposure (AE) before CAR-T infusion and safety/efficacy with appropriate statistical methods. We also obtained 35 fecal samples from 13 B-NHL patients along the treatment of CAR-T cells. The 16S sequencing was administrated to profile the microbial characteristics of these samples. The linear discriminant analysis effect size (LEFSe) was used to identify different bacteria abundant in patients who achieved complete remission (CR) or not complete remission (NCR) with a linear discriminant analysis score threshold >2.5. Result: The rate of AE was 52.8% among our patients with B-cell malignancies, including B-cell leukemia (B-ALL) and Non-Hodgkin lymphoma (B-NHL). The B-ALL patients used more kinds of antibiotics (mean 3.7 vs 2.5 types) a with longer exposure time (median 11.3 vs 10.8 days) before the infusion. After dividing the whole cohort into two groups, non-AE (NAE) and AE, we compared the baseline characteristics between the two groups. Patients in the AE group had higher ECOG score, higher levels of LDH and CRP, as well as the lower levels of hematological counts, reflecting their worse condition and heavier tumor burden. Patients who experienced severe infections were in an inflammatory status characterized by abnormally elevated LDH and CRP. Therefore, we excluded 25 patients who had to receive the treatment of antibiotics for the confirmed infections to further investigate the potential effects of antibiotics. After the exclusion, the confounder LDH level was balanced between two groups. However, the AE group still showed higher CRP level and lower absolute neutrophil count (ANC) compared to the NAE group at baseline, which provided a rationale for the use of antibiotics. We identified 5 variables that were related to CRS occurrence, including ECOG, LDH, disease, ANC before lymphodepletion and carbapenems by univariate Logistic analysis. However, carbapenems lost their significance after the multivariate Logistic analysis. For the prognosis, the AE group also had an unfavorable median progression-free survival (mPFS 4.3 vs 17.9 months, P=0.014). For B-NHL, the best overall response rate of the patients in the AE group was lower than that in the NAE group (50.0% vs 86.2 %, P=0.041 ). The B-NHL patients in the AE group also showed a poorer PFS (mPFS 3.0 vs 17.9 months, P=0.028) than those in the NAE group. Using univariate Cox regression, we were able to identify three high-risk antibiotics: carbapenems, glycopeptides, and fluoroquinolone. Then, we analyzed the data of 16S sequencing and found that there were five bacteria taxa among those abundant in NCR patients closely correlating with unfavorable PFS, including f_Enterobacteriaceae, g_Lachnoclostridium, s_Anaerotignum_lactatifermentans, s_Klebsiella_pneumoniae, and s_Phocaeicola_massiliensis with univariate Cox analysis. The last three of them are significantly higher in the AE group, which means the antibiotics may lead to an enrichment of some adverse bacterias. Conclusion: Priorexposure to antibiotics is associated with poor prognosis and may increases unfavorable gut bacterias like s_Anaerotignum_lactatifermentans, s_Klebsiella_pneumoniae, and s_Phocaeicola_massiliensis, which are correlated with unsatisfactory outcomes.