Can Consideration of the Microbiome Improve Antimicrobial Utilization and Treatment Outcomes in the Oncology Patient?

Antimicrobial therapy has markedly improved the outcome of cancer patients over the past 50 years. The potential dramatic impact of antimicrobials in oncology became clear when infections replaced hemorrhage and leukemia itself as the leading cause of death among acute leukemia patients in the 1960s (1). By the early 1970s, the development of methicillin for penicillin-resistant Staphylococcus aureus and carbenicillin for Pseudomonas aeruginosa (P. aeruginosa) meant serious infections could be effectively treated, even amid persistent neutropenia (2). As a result, neutropenic fever became an oncologic emergency demanding the rapid administration of broad-spectrum antibiotics that markedly improved the outcomes of neutropenic patients with proven infections, particularly due to P. aeruginosa (3, 4). Eventually, the high rates of morbidity and mortality associated with infections in patients with hematologic malignancies led to large randomized controlled trials that demonstrated that prophylactic administration of a fluoroquinolone (i.e., levofloxacin) reduced rates of neutropenic fever and confirmed infections (5). Thus, current oncology dogma primarily considers bacteria as a threat to patient health, with a low threshold for initiation of broad-spectrum antimicrobials in the preventive or therapeutic setting.

Feasible and affordable genetic means to comprehensively assay the bacteria present in a variety of sample types has paved the way for large-scale investigations such as the Human Microbiome Project (6). In addition to 16S rRNA gene sequencing, microbiome characterization methodologies have expanded to other “-omics” approaches to include whole-genome shotgun sequencing, RNA-seq, and metabolomics, which more precisely delineate bacterial community structure, gene presence/expression, and metabolic activity (7). Use of these methodologies has illuminated that the microflora have profound effects on human health, such as altering cytokine profiles, influencing inflammatory immune responses, and altering metabolites (8–11).

Although it is recognized that systemically administered antimicrobials can have a dramatic impact on the composition and function of the gastrointestinal microbiome (12), recent advances have also demonstrated that antibiotic effects on the microbiome influence the response to cancer immunotherapy. Specifically, Iida and colleagues described tumor necrosis and immune responses to be significantly reduced in antibiotic-treated colon carcinoma and melanoma tumor–bearing mice receiving immunostimulatory CpG-oligodeoxynucleotide treatment (13). Similarly, Vetizou and colleagues demonstrated that melanoma tumors in antibiotic-treated mice failed to respond to CTLA-4 blockade immunotherapy and that the presence of Bacteroides fragilis was critical to the antitumor effect (14). Recently, it has also been discovered that specific microbiota shape innate and adaptive immune system influencing the PD-1–PD-L1 axis (15, 16), although no studies have specifically shown the effects of antibiotic treatment on the microbiota and anti–PD-L1 treatment response.

In addition to influencing immunotherapy response, antibiotic-treated animals also display significantly reduced tumor regression and survival in cytotoxic therapy scenarios, such as oxaliplatin-treated lymphoma-bearing mice (13). Likewise, Viaud and colleagues observed that receipt of antibiotics with activity against Gram-positive bacteria reduced Th lymphocyte and lymphoma responses in mice treated with cyclophosphamide (17). Beyond animal models, recently, it was shown that patients being treated with cyclophosphamide for chronic lymphocytic leukemia and cisplatin for relapsed lymphoma who also received anti–Gram-positive antibiotics had significantly lower overall response rate and survival (18).

Furthermore, it is becoming increasingly clear that antimicrobial-induced microbiome disruption is also a key factor in cancer treatment–related toxicities. For example, administration of antibiotics to mice undergoing hematopoietic stem cell transplantation (HSCT) significantly increased the severity of GHVD and mortality (19, 20). Consistent with murine data, investigators found receipt of antibiotics with potent antianaerobic activity was associated with increased GVHD risk and GVHD-related mortality following allogeneic HSCT in patients (19, 21). In addition, fluoroquinolone receipt, low microbial diversity, and Gammaproteobacteria domination of fecal microbiota were predictive of pulmonary complications among HSCT recipients (22).

Antibiotic-induced microbial dysbiosis is also a crucial aspect in the cancer patient's risk for infectious toxicities. A prime example of the “Catch-22” relationship between antimicrobial therapy and cancer care is the hematologic malignancy patient. In these patients, depletion of native commensals by antibacterial prophylaxis and empirical treatment of neutropenic fever is compounded by mucosal barrier injury from cytotoxic chemotherapy, leading to proliferation of pathogenic bacteria, translocation across disrupted intestinal epithelium, and subsequent infection (Fig. 1; ref. 23). In leukemia and HSCT patients, receipt of particular antibiotics, such as metronidazole, was associated with decreased microbial diversity and, consequently, increased intestinal domination by pathogens that commonly cause hospital-acquired and bloodstream infections (24–27). Similarly, it was also found that broad-spectrum antibiotic receipt, specifically carbapenems, was associated with loss of bacterial diversity in both the oral and stool microbiomes of patients with acute myeloid leukemia during induction chemotherapy, which was, in turn, correlated with higher subsequent infectious risk in the 90 days after neutrophil recovery (28). Moreover, a considerable body of evidence exists that clearly links high rates of Clostridium difficile (C. difficile) infections in cancer patients, particularly those undergoing HSCT, to disruption of the normal intestinal flora due to a combination of repeated use of antibiotics, immunosuppression, and cancer therapy (24).

The studies outlined above indicate that the ability to comprehensively assess or alter human microbiota composition may be a valuable tool in improving cancer outcomes. Indeed, from a prognostic standpoint, the composition of the microbiome prior to chemotherapy has been demonstrated to be predictive of infectious outcomes for patients with acute myelogenous leukemia and lymphoma (28, 29). Similarly, the diversity of the gastrointestinal microbiome at the time of engraftment following HSCT is associated with risk of development of and mortality from GVHD (20, 30). Finally, the abundance of the Bacteroidetes phylum was recently correlated with resistance to the development of immune-mediated colitis in patients with melanoma treated with the immune checkpoint inhibitor ipilimumab (31). Thus, these data suggest the importance of developing probability indices that risk stratify cancer patients with respect to microbiome measurements and other clinical factors, such as antimicrobial administration (32–34). Predictive risk scores that incorporate microbiome measurements would need to include factors such as diversity metrics, absence of beneficial microbes or microbial byproducts (i.e., those associated with pathogen colonization resistance, resistance against treatment complications, or antitumor effects), and domination by specific microbes related to infection. Through these types of examinations, one could envision the microbiome being incorporated as a baseline screening tool to predict which patients may respond better to cancer therapy, are at risk for treatment-related toxicity, or are at risk for infectious complications.

Microbiome composition measurements may also assist with optimizing the choice and duration of antimicrobials in the cancer patient with respect to maintaining beneficial commensal microorganisms. For instance, we advocate for trials to assess whether rapid de-escalation of broad-spectrum antimicrobials can be done safely in patients with negative cultures in the setting of asymptomatic febrile neutropenia. By merging such studies with longitudinal microbiome analyses, it could be determined whether such de-escalation helps preserve microbiome composition and whether particular microbiome characteristics are associated with a need to reinitiate antimicrobials. In addition, investigating the use of more narrow-spectrum antibiotics as well as shorter duration of therapy for infections are needed, as it has been suggested that the number of antibiotics and total antibiotic exposure is linked with recurrent infectious complications in leukemia patients (35). Moreover, pharmacokinetic studies that link concentrations of antimicrobials in the intestinal lumen to effects on the microbiome are needed. Along the same lines, it will be important to assess not only the impact of the antimicrobial itself but also its elimination (biliary vs. renal). Such data could be used to design interventions to minimize the off-target effects of systemically administered antimicrobials on the commensal microbiota.

Microbiome measurements also raise the possibility to extend antimicrobial administration in oncology patients into the arena of personalized medicine. As genomic methodologies advance in terms of decreasing price and rapid availability of results, the ability to use microbiome samples to rapidly determine the scope of pathogens and antibiotic resistance genes present within an individual is becoming a real possibility (36). By cataloging the antimicrobial resistome for each patient using metagenomic analyses, physicians could inform their therapeutic considerations for prophylaxis and infection. For example, if a particular patient were known to have intestinal domination by a pathogen resistant to standard empiric antimicrobials, oncologists could take a more individualized approach to antimicrobial initiation if that patient were to develop infectious symptoms. Moreover, using microbiome measurements, other DNA sequencing–based approaches, or biomarkers, such as procalcitonin, to separate infectious from noninfectious fevers in the oncology patient would also greatly facilitate antimicrobial targeting and microbiota preservation. In addition to differentiating infectious from noninfectious fevers, it is also crucial to understand and discern colonization versus infection. For example, the unmet need to better distinguish C. difficile–colonized patients from patients with C. difficile colitis is leading to a mass overdiagnosis and overtreatment (37). It is highly likely that knowledge of microbiome interactions with the host will play an essential role in answering these needs.

Direct manipulation of the microbiome also offers a possibility for improving cancer therapy, minimizing toxicities, and mitigating the impact of infectious diseases. For example, a recent study suggests that fecal transplantation may ameliorate steroid-resistant GVHD in HSCT recipients (38). Similarly, fecal transplantation in mice increased responses to immunotherapy, raising the possibility that optimizing the microbiome prior to immune-modulating treatment could improve response (15). However, randomized trials examining the efficacy of microbiome remediation are needed to fully evaluate therapeutic potential.

As antimicrobials are a dwindling resource, using the microbiome as a direct interventional tool could improve antimicrobial utilization by offering an alternative treatment strategy for infectious complications, alleviating antibiotic resistance and preserving drug efficacy. Such microbiome-based methods include fecal transplantation, targeted addition of a single or defined combination of bacterial species (probiotics), or prebiotics designed to stimulate the growth and retention of specific beneficial species in the form of dietary-based intervention. It is thought that autologous fecal transplant could prevent pathogen intestinal colonization, infection, and development of antibiotic resistance (39, 40). Consequently, if cancer hospitals begin to bank patient feces prior to cancer treatment, a patient's native fecal microbiota could be implanted either continuously throughout treatment or administered after broad-spectrum antimicrobial treatment to counteract the microbiome damage potentially caused by antimicrobial treatment or chemotherapy.

However, one concern with using the administration of specific bacterial cocktails or fecal transplant in the immunocompromised patient is the risk for infection, as there have been numerous reports of septicemia associated with use of probiotic therapy, such as Lactobacillus bacteremia or Saccharomyces fungemia (41, 42). Thus, prebiotic administration or dietary intervention may be more desirable toxicity mitigation strategies. Recent examinations have suggested the beneficial impact of fiber on the microbiome, as it relates to inflammation and mucosal barrier injury, particularly in that specific fibers increase the number of butyrate-producing bacteria (43, 44). The short-chain fatty acid butyrate is importantly involved in adaptive immune responses, such as colonic T-cell differentiation (45–47). These data indicate that bypassing the microbiota and providing bacterial metabolites, such as short-chain fatty acids, is an alternative possibility. In addition, these studies suggest the importance of performing microbiome examination in tandem with metabolic and immunology research to improve intervention strategies that specifically target the host microbiota.

It is important to remember the era of cancer–microbiome research is relatively nascent and, thus, fundamental questions remain unanswered. For example, how useful are single-microbiome measurements given the microbiome inter- and intrapatient variability, particularly when ill? Although the majority of oncology patients lose microbial diversity during chemotherapy, interpatient changes are highly variable, with some patients maintaining a relatively preserved microbiota, whereas others exhibit microbiome domination by one or two pathogens (25, 28). Gaining knowledge regarding the factors that drive such drastically different microbiota trajectories is essential to designing and targeting microbiota preservation strategies. In addition to differences among individuals, more information is needed regarding variance in local microbiota composition at the intestinal mucosa versus what is present in stool samples.

Moreover, the integration of more advanced approaches, such as whole-genome sequencing and metabolomics, is needed to potentially uncover mechanisms by which the microbiome can impact clinical outcomes. For example, significant progress is being made toward culturing the entire intestinal bacterial microbiome using methods such as “culturomics,”not only to improve upon the identification of viable species within the gut but also to capture the functional biodiversity (48). Moreover, elucidating the role that the mycobiome and virome play in immune responses, cancer therapy response, cancer treatment toxicities, and infectious complications will also need to be incorporated in future research, as these areas remain mostly unexplored. This effort, however, will need to include improving sequencing methods and databases for fungi and viruses.

It is also crucial to improve our statistical methodologies so that the complex nature of microbiome data, particularly with regard to longitudinal sampling, can be incorporated into clinical models. Statistical challenges include developing strategies to look for associations in high-dimensional data, a problem that is also being addressed by other types of big data (i.e., exome, proteomics, transcriptomics, etc.). Some challenges are unique to the microbiome, which features the additional layer of evolutionary relationships and potential interactions between bacteria, fungi, and viruses. The further development of biostatistical methods that can identify statistically meaningful relationships among networks integrating high-dimensional microbiome data with complex variables, such as gene function, metabolites, antibiotic administration, diet, and patient outcomes, is key to conceiving dependable interventions.

The dramatic impact of the commensal microbiota on the health of the cancer patient is increasing in appreciation. As profound effects of antibiotics on the human microbiome have been demonstrated, it is imperative that antibiotic administration and stewardship strategies in patients with malignancy be considered within the context of the microbiome. Many possible future avenues of investigation exist that could potentially aid physicians in treating cancer-related infections while limiting collateral damage to the microbiota (Table 1). As more exploratory work is done to understand the microbiome's role in cancer and cancer treatment–related toxicities, carefully designed animal models and interventional trials will be critical to moving beyond basic association or biomarker studies to determine the mechanisms by which the microbiome modulates patient outcomes. The integration of microbiome-based approaches into the clinical arena offers a tremendous new opportunity to improve outcomes across the cancer care continuum.

R.R. Jenq holds ownership interest (including patents) in Seres Therapeutics, and is a consultant/advisory board member for Evelo Biosciences, Seres Therapeutics, and Ziopharm Oncology. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.R. Galloway-Peña, S.A. Shelburne

Writing, review, and/or revision of the manuscript: J.R. Galloway-Peña, R.R. Jenq, S.A. Shelburne

The authors would like to thank Dr. Cathy Eng, Dr. Elizabeth Mittendorf, and Chelcy Brumlow, M.S. for their contribution of making suggestions for this article.

J.R. Galloway-Peña is supported by the Odyssey Program and CFP Foundation at The University of Texas MD Anderson Cancer Center. R.R. Jenq is funded through the NHLBI R01HL124112 and the Cancer Prevention and Research Institute of Texas Recruitment of Rising Stars Program RR160089. S.A. Shelburne is funded through The University of Texas MD Anderson Multi-Disciplinary Research Program and other funds through the MD Anderson Cancer Center.