Gut microbiota diversity and composition in predicting immunotherapy response and immunotherapy-related colitis in melanoma patients: A systematic review

Abstract

BACKGROUND

Gut microbiome (GM) composition and diversity have recently been studied as a biomarker of response to immune checkpoint blockade therapy (ICB) and of ICB-related colitis.

AIM

To conduct a systematic review on the role of GM composition and diversity in predicting response and colitis in patients with melanoma treated with ICB.

METHODS

The review protocol was registered in PROSPERO: CRD42021228018. From a total of 300 studies, nine studies met inclusion criteria. Two studies were phase I clinical trials, while the remainder were prospective observational studies. All but one study has moderate risk of bias. In addition, we conducted a relevant search by Reference Citation Analysis (RCA) (https://www.referencecitationanalysis.com).

RESULTS

Fecal samples enriched in Firmicutes phylum were associated with good response to ICB, whereas the Bacteroidales family was associated with poor response to ICB. Samples with greater GM diversity were associated with more favorable response to ICB [hazard ratio (HR) = 3.57, 95% confidence interval = 1.02-12.52, P < 0.05]. Fecal samples with a higher abundance in Firmicutes were more susceptible to ICB-related colitis (P < 0.01) whereas samples enriched in Bacteroidetes were more resistant to ICB-related colitis (P < 0.05). Overall, there was limited concordance in the organisms in the GM identified to be associated with response to ICB, and studies evaluating GM diversity showed conflicting results.

CONCLUSION

This highlights the need for further prospective studies to confirm whether the GM could be used as a biomarker and potential intervention to modulate ICB response in melanoma patients.

Key Words: Melanoma, Gut microbiome, Microbiota, Immunotherapy, Biomarker, Immune checkpoint blockade therapy

Core Tip: Since the introduction of immune checkpoint inhibitors as part of standard of care for melanoma patients, there has been a growing interest in identifying biomarkers of response and immune related adverse events. Amongst these biomarkers, the composition of the gut microbiome has been one of the most intriguing discoveries. Our aim was to ascertain the current published evidence on the gut microbiome diversity and composition as a biomarker of response to immunotherapy. We demonstrated high variability in the results and limited concordance on the organisms identified. We highlight the conflicting aspects of these reports as well as their few commonalities.

INTRODUCTION

Melanoma is the most lethal form of skin cancer accounting for 73% of skin-cancer related mortality and over 50000 deaths worldwide annually[1,2]. Survival for metastatic melanoma has significantly improved since the introduction of immunotherapy and targeted therapy with a 5-year survival rate of up to 50%[3-5]. Currently, the standard first-line therapy for metastatic melanoma include BRAF-targeted therapies and immune checkpoint blockade (ICB) consisting either anti-programmed death (PD)-1 monotherapy or combination of anti-PD-1 as well as anti-cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) therapy[6]. Despite the considerable benefit of ICB, 40%-60% of melanoma patients do not experience objective responses to the therapy[7-9]. Thus, tremendous efforts are now focused on identifying novel biomarkers which could accurately predict the subset of patients who would benefit from ICB[10-14]. These biomarkers include tumor mutational burden, cytokines, circulating tumor DNA, human leukocyte antigen, gut microbiota (GM) diversity and composition, among many others[15].

The GM is a community of 100 trillion microorganisms of more than 1000 species mainly bacteria but also, archaea, viruses and fungi which colonize the human intestines[16]. The relationship which exists between GM and the host is a mutualistic relationship where one benefits the other[16]. In return for the nutrients derived from the host, the GM performs numerous critical functions such as fermentation of dietary fiber into short-chain fatty acids; synthesis of vitamins; protection against pathologic gut microbes; and induction and regulation of the immune system[17,18]. The gut microbial balance is pivotal in the optimal functioning of all of these roles and thus any discrepancy in this delicate equilibrium could produce a state of dysbiosis which has been associated with many pathologies including cancer[19]. In the context of cancer, preclinical studies have demonstrated that some GM subpopulations have pro-tumorigenic effects, whereas others have tumor-suppressive effects[20-22]. Additionally, the GM has also been shown to modulate response to chemotherapy and immunotherapy[23-25]. This could be linked to the role of GM in metabolizing anti-cancer compounds and regulating the host’s immune response[16]. Thus, GM has been studied intensely as a potential biomarker of response to ICB[12,26-31]. This is particularly relevant for melanoma, where ICB has become standard of care given its demonstrated pronounced effectiveness.

Studies investigating GM composition and/or diversity in patients with melanoma have identified distinct GM composition in responders to ICB compared to non-responders, offering hope of a novel biomarker for predicting response to ICB[12,26-32]. Additionally, studies exploring whether certain GM composition and diversity could be predictive of ICB-related colitis - one of the major factors of ICB treatment cessation and thus failure to derive full benefit of ICB - have also been conducted[27,33]. This systematic review will be the first to compile the existing data regarding the role of GM composition and diversity in predicting response to ICB and ICB-related colitis specifically in patients with melanoma. Notably previous reviews have combined multiple cancers.

MATERIALS AND METHODS

Literature search strategies

This review was conducted following the preferred reporting Items for systematic reviews and meta-analyses guidelines[34]. The review protocol was submitted to the international prospective register of systemic reviews (PROSPERO Registration number: CRD42021228018).

In this comprehensive literature search, original studies exploring the variation in GM community in fecal samples of melanoma patients who responded and did not respond to immunotherapy, experienced colitis and did not experience colitis were identified. Medline and Embase were searched for eligible papers published prior to December 2021 using the following search terms: (fecal OR gut) AND (microbiota OR microbiome) AND (melanoma) AND (immunotherapy OR checkpoint OR nivolumab OR ipilimumab OR pembrolizumab). OpenGrey and the Grey Literature Report were also searched for eligible unpublished papers and grey literature. The following keywords and its synonyms will be used for our search strategy: “fecal microbiota”, “melanoma”, “immunotherapy”.

Duplicate and irrelevant publication types such as symposium agendas were removed from the initial search results. Titles and abstracts of relevant publications were screened independently by Oliver O and Simadibrata DM based on inclusion and exclusion criteria stated below. Subsequently, reference lists within each relevant publication were examined for further pertinent studies. The full texts of these publications were then reviewed.

Inclusion criteria

Inclusion criteria for the systematic review included randomized controlled trials (RCTs), original cohort, case-control studies published in a peer-reviewed journal exploring GM diversity and composition in fecal samples from melanoma patients treated with ICB which can be anti-PD-1 and/or anti-PD-L1 and/or anti-CTLA-4. Studies included should assess treatment outcome and/or ICB-related colitis incidence following treatment with ICB. Treatment outcomes should be determined by RECIST criteria and/or progression free survival (PFS) and/or overall survival (OS) and ICB-related colitis confirmed by colonoscopy.

Only studies which utilized fecal samples obtained from human subjects receiving ICB were included. Studies which assessed treatment response to immunotherapy in animal models were excluded. Two reviewers (Oliver O, Liu YY) independently screened and read the full text of the included articles for eligibility.

Data extraction

Two investigators (Oliver O, Liu YY) independently reviewed the eligible studies and extracted data from each study. Extracted variables included title, first author, year of publication, number of participants, type of immunotherapy received, GM analysis method, and study outcomes (GM composition and diversity in responders/non-responders and ICB-related colitis/non-ICB-related colitis patients). Extracted GM composition data included a list of the GM at the level of phyla, class, order, family, genus and species, whereas extracted GM diversity extracted included alpha diversity or the Shannon index. Any discrepancies found by the investigators on data extraction were resolved by consensus. In addition, we conducted a relevant search by Reference Citation Analysis (RCA) (https://www.referencecitationanalysis.com).

Quality assessment

Non-randomized studies, including cohort studies, case-control studies and single-arm clinical trial that were included in this systematic review were independently evaluated by Oliver O and Simadibrata DM for any risk of bias using the Risk of Bias in Non-randomized Studies of interventions (ROBINS-I) assessment tool, a tool which assesses seven items: confounding, selection, intervention classification, deviation from intervention, missing data, measurement of outcome and selection of reported result. Each item was assessed according to the ROBINS-I guideline, where each bias domain can be classified as either low, moderate, serious or critical risk of bias, or no information mentioned.

RESULTS

Study selection and risk of bias assessment

The initial search from Medline, Embase, OpenGrey and Grey Literature retrieved 300 studies. After deduplication, the studies were screened by reviewing their abstracts and 10 articles selected for full assessment (Figure 1). One study by Vétizou et al[35] was excluded because while the fecal samples were obtained from patients treated with anti-CTLA-4, treatment response to ICB was assessed in an in-vivo mice model of melanoma following fecal transplantation rather than humans.

Figure 1 Prisma flow diagram of study selection.

From the nine included studies, two studies were phase I clinical trials, while the remainder were prospective observational studies[26,28]. Unfortunately, no RCTs were available to date. According to the ROBINS-I assessment tool, all but one study was shown to have moderate risk of bias (Table 1). The study by Matson et al[29] had a serious risk of bias as there was a lack of clarity regarding the definition of intervention used.

Table 1 Risk of bias assessment with Risk of Bias In Non-randomised Studies - of Interventions.

Ref.	Confounding	Selection	Intervention classification	Deviation from intervention	Missing data	Measurement of outcome	Selection of reported result	Overall
Dubin et al[33], 2016	Moderate	Low	Low	Low	Low	Low	Low	Moderate
Chaput et al[27], 2017	Moderate	Low	Low	Moderate	Low	Low	Moderate	Moderate
Gopalakrishnan et al[12], 2018	Moderate	Low	No information	Low	Low	Low	Low	Moderate
Matson et al[29], 2018	Moderate	No information	Serious	No information	No information	Low	Moderate	Serious
Peters et al[30], 2019	Moderate	Low	No information	No information	Low	Low	Low	Moderate
Baruch et al[26], 2020	Moderate	Low	Low	Low	Moderate	Low	Low	Moderate
Wind et al[31], 2020	Moderate	No information	No information	No information	No information	Low	Low	Moderate
Davar et al[28], 2021	Moderate	Moderate	Low	Low	No information	Low	Low	Moderate
Andrews et al[36], 2021	Moderate	Low	Low	Low	Low	Low	Low	Moderate

GM composition and diversity in predicting immunotherapy response

Eight studies assessed the role of GM composition and/or diversity and response to ICB in melanoma patients (Table 2). Seven studies compared the GM between responders and non-responders to ICB, and two studies analyzed the GM in patients undergoing fecal microbiota transplant (FMT).

Table 2 Characteristics of included studies exploring link between gut microbiome composition and diversity and response to immune-checkpoint blockade therapy in metastatic melanoma patients treated with immune-checkpoint blockade therapy.

Ref.	Year	Therapy	Method	Sample size/ time point	Dominant microbes	Microbial diversity
Chaput et al[27], 2017	2017	Anti-CTLA-4	16S rRNA gene sequencing of fecal samples	26 before tx	Responders: Faecalibacterium and Firmicutes	N/A
Matson et al[29], 2018	2018	Anti-PD-1 or anti-CTLA-4	16S rRNA gene and shotgun metagenome sequencing of fecal samples; qPCR on selected bacteria	42 before tx	Responders: Bifidobacterium longum, Collinsella aerofaciens, and Enterococcus faecium Non-responders: Ruminococcus obeum and Roseburia intestinalis	N/A
Gopalakrishnan et al[12], 2018	2018	Anti-PD-1	16S rRNA gene and shotgun metagenome sequencing of fecal samples	43 before tx	Responders: Clostridiales, in particular Faecalibacterium Non-responders: Bacteroidales, in particular Bacteroides thetaiotaomicron; as well as Escherichia coli, and Anaerotruncus colihominis	Higher alpha diversity in patients with longer PFS
Peters et al[30], 2019	2019	Anti-PD-1 or anti-CTLA-4	16S rRNA gene and shotgun metagenome sequencing of fecal samples	27 before tx	Responders: Faecalibacterium, Parabacteroides, and Faecalibacterium prausnitzii Non-responders: Bacteroides and Biophilia	Higher microbial community richness and diversity was associated with longer PFS
Wind et al[31], 2020	2020	Anti-PD-1 or anti-CTLA-4	Shotgun metagenome sequencing of fecal samples	25 before tx	Responders: Ruminococcus gnavus, Streptococcus parasanguinis, and Bacteroides massiliensis. Non-responders: Bifidobacterium longum and Peptostreptococcaceae	No significant difference in alpha-diversity between responder and non responders
Baruch et al[26], 2020	2020	Anti-PD-1 refractory	16S rRNA gene and shotgun metagenome sequencing of fecal samples	10 anti-PD-1 refractory patients	FMT donors (responders): Lachnospiraceae, Veillonellaceae, and Ruminococcaceae Post FMT Responders: Enterococcaceae, Enterococcus, and Streptococcus australis Non-responders: Veillonella atypica	No significant difference in GM composition prior to FMT, but significant difference post-FMT between responders and non-responders Lower microbial richness in the donor of responding recipients
Davar et al[28].2021	2021	Anti-PD-1 refractory	Shotgun metagenomic sequencing of fecal samples	15 anti-PD-1 refractory patients, before FMT	Responders: Firmicutes (Lachnospiraceae and Ruminococcaceae families) and Actinobacteria (Bifidobacteriaceae and Coriobacteriaceae families)	Higher GM diversity of donors who were complete responders compared to donors who were partial responders No significant difference in GM diversity between donors and recipients prior to FMT
Andrews et al[36], 2021	2021	Combined ICB - Anti-PD-1 and anti-CTLA-4	16S rRNA gene and shotgun metagenome sequencing of fecal samples	38	Responders: Bacteroides stercoris, Parabacteroides distasonis, Fournierella massiliensis. Non-responders: Klebsiella aerogenes and Lactobacillus rogosae	No significant difference in GM diversity between responders and non-responders

Anti-CTLA-4: Anti-cytotoxic T lymphocyte-associated antigen-4; N/A: Not applicable; Anti-PD-1: Anti-programmed death-1; qPCR: Quantitative real-time polymerase chain reaction; FMT: Fecal microbiota transplant; GM: Gut microbiome; ICB: Immune checkpoint blockade therapy.

The study by Chaput et al[27] assessing fecal GM composition of 26 metastatic melanoma patients prior to and post commencing anti-CTLA-4 therapy revealed that GM composition varied according to response. Patients showing long term response to therapy (nine out of 26 patients) were found with fecal samples with significantly higher Faecalibacterium percentages (P = 0.0092) while patients with poor clinical benefit had higher proportions of Bacteroides (P = 0.034). When patients were grouped based on their microbiota composition, those with high prevalence of Faecalibacterium and other Firmicutes had a longer PFS (P = 0.0039) and to a lesser extent longer OS (P = 0.051) relative to patients whose fecal samples were abundant with Bacteroides. Additionally, these patient groups were noted to derive long-term clinical benefit compared to the latter (67% vs 0%; P = 0.0017)[28].

In an analysis of stool samples from 42 metastatic melanoma patients prior to treatment with anti-PD-1 (n = 38) and anti-CTLA-4 (n = 4) therapy, Matson et al[29] showed a significant difference in GM composition between responders (16 patients) and non-responders (26 patients) (P < 0.01). In responders, eight microbial species namely, Enterococcus faecium, Collinsella aerofaciens, Bifidobacterium adolescentis, Klebsiella pneumoniae, Veillonella parvula, Parabacteroides merdae, Lactobacillus sp. and Bifidobacterium longum were found to be more abundant in responders than in non-responders[29]. In non-responders, two microbial species, specifically, Ruminococcus obeum and Roseburia intestinalis were more abundant[29]. To further assess the applicability of GM composition as a biomarker of response to ICB, they explored the correlation between the ratio of total numbers of potentially “beneficial” and “nonbeneficial” operational taxonomic units (OTUs), and change in tumor size, as assessed by the RECIST[29]. Patients with an OTU ratio of greater than 1.5 demonstrated clinical response to ICB[29].

In another study by Gopalakrishnan et al[12], fecal samples of 43 metastatic melanoma patients prior to treatment with anti-PD-1 therapy were analyzed. In responders (30 patients), analysis of fecal samples revealed abundance of GM from Ruminococcaceae family of the Clostridiales order, whereas in non-responders (13 patients), abundance of GM from the Bacteroidales order was noted[12]. Further analyses demonstrated that Faecalibacterium genus was notably enriched in fecal samples from responders and Bacteroides thetaiotaomicron, Escherichia coli, and Anaerotruncus colihominis were enriched in non-responders[12]. In addition, to investigate durability of response, patients were stratified based on their fecal composition of Faecalibacterium genus and Bacteroidales order and correlated to their PFS[12]. Results demonstrated that patients with Faecalibacterium-enriched fecal samples have longer PFS than those with low abundance (P = 0.03) and patients with Bacteroidales-enriched fecal samples have shorter PFS than those with low abundance (P = 0.05). Beyond specific microbial taxa, GM diversity, as assessed by Simpson's reciprocal index, was higher in responders compared to non-responders (P < 0.01)[12]. Moreover, high GM diversity was significantly associated with anti-PD-1 therapy response, when compared to patient groups of intermediate diversity [hazard ratio (HR) = 3.60, 95% confidence interval (CI): 1.02-12.74, P < 0.05) and low diversity (HR = 3.57, 95%CI: 1.02-12.52, P < 0.05). Other important predictors of therapy response include abundance of Faecalibacterium (HR = 2.92, 95%CI: 1.08-7.89) and Bacteroidales (HR = 0.39, 95%CI: 0.15-1.03) in the fecal microbiome[12].

Peters et al[30] examined the correlation between GM taxa and PFS in pre-treatment fecal samples of 27 metastatic melanoma patients receiving anti-PD-L1 and/or anti-CTLA-4. GM which was associated with shorter PFS included genera Bacteroides and Bilophila, and species Bacteroides ovatus, Blautia producta, and Ruminococcus gnavus, whereas those which correlated with longer PFS included genera Faecalibacterium and Parabacteroides and species Faecalibacterium prausnitzii[12]. With regards to GM richness the authors compared the β-diversity or between-sample microbiome diversity relative to survival. Multivariate analysis adjusting for age, sex, BMI, stage, number of sites of metastases, and antibiotic use in the last 6 mo revealed that higher GM richness was correlated with longer PFS (number of 16S sub - OTUs: HR [95%CI] = 0.97 [0.95, 1.00], P = 0.02; number of shotgun subspecies: HR [95%CI] = 0.89 [0.79, 0.99], P = 0.03)[30]. Furthermore, analysis of the 16S but not shotgun dataset showed that higher diversity of GM, as assessed by the Shannon index, was associated with longer PFS (P = 0.02)[12].

Similarly, Wind et al[31] analyzed fecal samples from 25 metastatic melanoma patients - 12 responders, 13 non-responders - prior to start of treatment with anti-PD-1 or anti-CTLA-4. Analysis revealed that the fecal samples of responders were mainly enriched in Ruminococcus gnavus, Escherichia coli, Eubacterium biforme, Phascolarctobacterium succinatutens and Streptococcus salivarius, whereas samples from non-responders were abundant in Bifidobacterium longum, Prevotella copri, Coprococus sp, Eggerthella unclassified and Eubacterium ramulus[31]. When correlated with survival, fecal samples of participants enriched in Bacteroides massiliensis and Streptococcus parasanguinis were associated with longer PFS (HR: 3.79, 95%CI: 1.06-13.52 P = 0.04) and OS (HR: 5.05, 95%CI: 1.33-19.21, P = 0.017) respectively, whereas those who were carriers of Peptostreptococcaceae were associated with shorter PFS (HR: 0.18, 95%CI: 0.05-0.62, P = 0.007) and OS (HR: 0.12, 95%CI: 0.01-0.96, P = 0.046)[31]. In terms of GM diversity, as assessed by Shannon index, no significant difference between responders and non-responders was noted[31].

The study by Andrews et al[36] analyzed gut microbiome samples from a subset of 77 metastatic melanoma patients - 27 responders, 11 non-responders - who underwent combined ICB. There was no significant association in Firmicutes phyla and Clostridiales order and response to ICB (P = 0.39 and P = 0.38, respectively) and no significant difference in alpha diversity between responders and non-responders to ICB[36]. Fecal samples from responders were mainly enriched with Bacteroides stercoris, Parabacteroides distasonis and Fournierella massiliensis (P = 0.03, P = 0.04 and P = 0.008, respectively) while fecal samples from non-responders were abundant in Klebsiella aerogenes and Lactobacillus rogosae (P = 0.04 and P = 0.02, respectively)[36].

In a first clinical trial of its kind (phase 1), Baruch et al[26] demonstrated that FMT from anti-PD-1 treated metastatic melanoma patients who were complete responders (2 donors), triggered response to anti-PD-1 therapy in metastatic melanoma patients who were refractory to at least one line of anti-PD-1 therapy. Out of 10 patients included in the trial, 3 patients demonstrated objective responses with 1 achieving complete response and 2 patients achieving partial response[26]. Notably, the PFS milestone of 6 mo was reached in all responders[26]. Upon analysis of pre-treatment fecal samples of donors, donor of the responding recipients had a lower microbial richness than the other donor of the non-responding patients[26]. There was no significant difference on the GM composition prior to FMT of recipients who responded compared to those who did not respond[26]. Metagenome sequencing found that recipients post FMT have higher proportions of Veillonellaceae family and a lower relative abundance of Bifidobacterium bifidum. Donors were found with high amounts of Lachnospiraceae, Veillonellaceae, and Ruminococcaceae. Comparison of a small subset of non-responders with responders, found statistically significant higher abundance of Enterococcaceae, Enterococcus, and Streptococcus australis, and a lower relative abundance of Veillonella atypica. However clear deductions on specific GM taxa cannot be made, as there were non-responders and pre-treatment fecal samples with similar dynamics. it is crucial to note that this trial was primarily designed to assess safety of FMT and not statistically powered to assess efficacy[26].

In a separate trial, Davar et al[28] showed that fecal microbial transplant (FMT) from metastatic melanoma patients (7 donors) who had complete (4 donors) or partial response (3 donors) to anti-PD-1 therapy helped overcome resistance in anti-PD-1 treatment-refractory metastatic melanoma patients (15 patients). Following FMT and anti-PD-1 therapy, 6 out of 15 patients achieved clinical benefit, with 3 patients achieving objective responses and 3 patients experiencing stable disease lasting more than 12 mo[28]. Analysis of stools after FMT revealed that samples from responders were abundant in the phyla, Firmicutes (Lachnospiraceae and Ruminococcaceae) and Actinobacteria (Bifidobacteriaceae and Coriobacteriaceae) and had decreased proportions in phylum Bacteroidetes[28]. In terms of GM diversity assessed with inverse Simpson index, GM diversity of donors who were complete responders were more diverse than donors who were partial responders. There was no significant difference in GM diversity between donors and recipients prior to FMT[28].

Gut microbiota composition and diversity in predicting ICB-related colitis

To date only three studies have reported on the correlation between pre-treatment GM composition and/or diversity and ICB-related colitis (Table 3).

Table 3 Characteristics studies exploring link between gut microbiome composition and diversity and immune-checkpoint blockade therapy-related colitis in metastatic melanoma patients treated with immune-checkpoint blockade therapy.

Ref.	Year	Therapy	Method	Sample size/ timepoint	Dominant microbes	Microbial diversity
Dubin et al[33], 2015	2015	Anti-CTLA-4 immunotherapy	16S rRNA gene and shotgun metagenome sequencing of fecal samples	34	Colitis-resistant: Bacteroidetes (Bacteroidaceae, Rikenellaceae and Barnesiellaceae)	No significant difference in microbial richness and diversity
Chaput et al[27], 2017	2017	Anti-CTLA-4 immunotherapy	16S rRNA gene sequencing of fecal samples	26	Colitis-resistant: Bacteroidetes; Colitis-prone: Firmicutes	Decreased bacterial diversity was associated with colitis
Andrews et al[36], 2021		Combined ICB - Anti-PD-1 and anti-CTLA-4	16S rRNA gene and shotgun metagenome sequencing of fecal samples	38	Colitis resistant: Firmicutes; Colitis prone: Bacteriodetes	No significant difference in alpha diversity

Anti-CTLA-4: Anti-cytotoxic T lymphocyte-associated antigen-4; N/A: Not applicable; Anti-PD-1: Anti-programmed death-1; qPCR: Quantitative real-time polymerase chain reaction; ICB: Immune checkpoint blockade therapy.

Firstly, a prospective study by Dubin et al[33] explored the link between GM composition, and subsequent colitis development in 34 metastatic melanoma patients treated with ipilimumab, showed that the Bacteroidetes phylum was more abundant (P = < 0.05) in fecal samples of the 24 patients who did not develop ipilimumab-induced colitis compared to those who did. Further analysis revealed that within the Bacteroidetes phylum, the population of Bacteroidaceae, Rikenellaceae and Barnesiellaceae was significantly more abundant in the former than the latter (P < 0.01, P < 0.05 and P < 0.05 respectively)[33]. However, there was no significant difference in microbial richness and diversity, as assessed by Shannon and inverse Simpson indices, between those who developed ipilimumab-induced colitis relative to those who did not[33].

In a similar study by Chaput et al[27], analysis of fecal samples of metastatic melanoma patients receiving ipilimumab demonstrated high proportions of Firmicutes in patients who developed ipilimumab-induced colitis (P = 0.009). In contrast, fecal samples of those that did not develop colitis were mainly enriched with Bacteroidetes (P = 0.011)[27]. Accordingly, patients with the former GM composition also tend to have a shorter colitis-free cumulative incidence compared with patients with the latter composition[27]. Several OTUs known to be predictive to colitis such as F. prausnitzii L2-6, butyrate producing bacterium L2-21 and G. formicilis ATCC 27749 were associated with longer OS[27].

Finally, Andrews et al[36], analyzed gut microbiome samples in metastatic melanoma patients undergoing combined ICB and their link to ileitis and colitis events. No significant difference in alpha diversity was observed between those that did and did not develop colitis[36]. Fecal samples of patients developing colitis were enriched in Bacteroides intestinalis and Intestinibacter bartlettii (P = 0.009 and P = 0.009, respectively) while those that did not were abundant in Anaerotignum lactatifermentans and Dorea formicigenerans (P = 0.016 and P = 0.06, respectively)[36]. For both B. intestinalis and D. formicigenerans, associations with their risk of colitis were still maintained after adjustment using a logistic regression model [OR = 4.54 (95%CI = 1.06-24.7) and OR = 0.35 (95%CI = 0.082-1.35), respectively][36].

DISCUSSION

Our review of current reports assessing the GM composition relative to response to ICB, indicated high variability in the results and limited concordance on the organisms identified (Figure 2). Amongst the few commonalities, we found that fecal samples enriched in organisms from the Firmicutes phylum (Lachnospiraceae and Ruminococcaceae family) especially the Faecalibacterium genus were associated with ICB responders in 4 of 9 studies[12,27,28,31], while Bacteroidetes phylum was found in higher proportions in non-responders in 2 of the studies[12,30]. However, other than these two findings, there was no clear correlation between specific GM composition and response to ICB.

Figure 2 Phylogenetic tree showing family and species of gut microbiome abundant in responders and non-responders to immune-checkpoint blockade therapy in all included studies. Gut microbiome species highlighted in red: Abundant in non-responders to immune-checkpoint blockade therapy; blue: Abundant in responders to immune-checkpoint blockade therapy; purple: Abundant in both responders and non-responders.

In fact, our analysis mainly identified inconsistencies in the GM composition reported to be associated with response to ICB. For instance, Bifidobacterium longum was found to be abundant in responders in the study by Wind et al[31], but found to be enriched in non-responders in the study by Matson et al[29]. Some species from the Firmicutes family were found in both responders and non-responders such as Roseburia intestinalis[29]. Similarly, species from the Bacteroidales order were found in both responders and non-responders, such as Bacteroides massiliensis[31]. The overlap in GM composition in responders and non-responders may suggest that the functional capacity of the GM may be more important than individual GM family/order/species in determining response to ICB[30].

In contrast to individual species or taxas, GM diversity have been heralded to a marker of good health[37]. Here four studies - Gopalakrishnan et al[12], Peters et al[30], Wind et al[31] and, Andrews et al[36] - assessed its potential to predict ICB responsiveness. The two first studies demonstrated that higher GM diversity in the responder group compared to non-responder arm[12,30]. However, the other two studies, Wind et al[31] and Andrews et al[36], found no differences in GM diversity between both groups. Nevertheless, in other cancer types such as renal cell carcinoma and non-small cell lung cancer, greater GM diversity has also been associated with improved responses to anti-PD-1 therapy[37-39].

Study results showing associations with GM diversity were consistent with previous studies which showed that greater GM diversity is prevalent in healthy state across multiple diseases, plausibly suggesting that a greater GM diversity produces the optimal immune environment needed for normal physiological functioning[40-42]. One major reason is the promotion of a favorable immune phenotype, as evidenced by the positive correlation between Shannon diversity index and several CD8+ T cell and NK cell signatures, required to produce a robust anti-tumoral response[38].

Previous studies have demonstrated that GM from Firmicutes family and Bacteroidales order play a significant role in mediating the response to immunotherapy in melanoma patients[12,27,29]. For instance, abundance of Firmicutes was associated with increased frequencies of CD4+, CD8+ T cells, CD 45+ myeloid and lymphoid tumor-infiltrating cells and preserved cytokine response to anti-PD-1 therapy[12]. Additionally, abundance of Firmicutes was linked with decreased frequency of intestinal and systemic regulatory T cells (Tregs) and B7⁺ T cells, cells responsible for limiting immune response robustness[27]. This resulted in increased antigen presentation and effector T cell function in both the periphery and tumor microenvironment[12,27,29]. However, other GM such as Bacteroidales were unfavorable in terms of anti-tumoral response in that its abundance was associated with higher frequencies of Tregs and myeloid-derived suppressor cells and a blunted cytokine response[12]. These findings combined demonstrated that certain GM play a crucial role in mediating systemic and antitumor immune responses which have clear implications on efficacy on ICB therapy in metastatic melanoma patients.

Notably, GM has also been shown to potentially serve as not just a predictor of ICB therapy response, but also for boosting response to ICB therapy. FMT on anti-PD-1 treatment-refractory metastatic melanoma patients produced a complete response to anti-PD-1 therapy in one-third (9 out of 25 patients) of the otherwise therapy refractory patients[26,28].

Another aspect of the GM analyzed here, was its association with ICB-related colitis. The three studies included in this review demonstrated that GM which was abundant in ICB-related colitis-prone patients was enriched in responders to ICB (Firmicutes) while GM which was abundant in ICB-related colitis-resistant patients was enriched in non-responders to ICB (Bacteroidetes). This is consistent with the understanding that a more effective anti-tumoral response will produce greater off-target effects. The Bacteroidetes phyla has been linked with low-grade systemic inflammation, which could explain the observation that Bacteroidetes phyla was abundant in ICB-related colitis-resistant patients[27]. In line with this observation is the finding that level of Bacteroidetes is lower in inflammatory bowel disease - an autoimmune condition which produces chronic inflammation of the digestive tract - patients relative to healthy patients[43]. Conversely, Firmicutes phyla, especially F. prausnitzii has been associated with induction of Tregs which express high levels of CTLA-4, fueling speculation that it may cause sequestration of Tregs within the intestine[44]. Since Tregs express high levels of CTLA-4, their actions are inhibited, thereby limiting self-tolerance and promoting the development of colitis. These findings reiterate that GM has an immunomodulatory role, giving them the potential to be utilized as biomarkers of ICB-related colitis, in addition to response to ICB.

Our systematic review has several strengths. Firstly, unlike previous reviews which combined studies in various cancer types, this review focused solely on the effect of GM composition and diversity only in patients with melanoma. Secondly, we conducted a comprehensive search for RCTs and observational studies, performed a risk-of-bas assessment and studied clinically important outcomes - clinical response and ICB-related colitis - an adverse event reported in up to 25% of patients treated with ICB[45]. Thirdly, we only included studies which assessed response to immunotherapy in humans, not animals. However, several limitations exist in our systematic review. Studies which we included used distinct approaches when segregating patients into the responder and non-responder groups, using different response criteria to evaluate treatment response in patients. Additionally, there were differences in methods of stool collection and analysis of GM composition and diversity. For example, Chaput et al[46] collected multiple stool samples every 3 wk of ICB, while other studies such as Dubin et al[33] and Matson et al[29] collected stool samples only prior to initiation of ICB. Furthermore, only 4 studies considered confounding factors such as variation in diet and antibiotic use[27,29-31]. Therefore, inter-study comparison of the GM composition and diversity in responders vs non-responders and those who experienced colitis vs non-colitis should be addressed with caution. Furthermore, included studies only enrolled a small number of patients, which could explain inconsistent results between studies.

CONCLUSION

In conclusion, GM composition and diversity holds some potential as a biomarker of response and toxicity to ICB in melanoma. Larger prospective studies with standardized experimental protocol ought to be conducted to elucidate whether distinct GM signatures are required for robust response to different ICB regimens. Additionally, more studies correlating metagenomic and metatranscriptomic data of GM to outcomes of melanoma patients on immunotherapy ought to be performed as the functional capacity may be more important rather than individual GM family/order/species. In addition, we eagerly await the outcome of multiple large-scale RCTs involving FMT in the context of ICB-refractory melanoma such as NCT04577729 and NCT04988841 (PICASSO) (ClinicalTrails.gov).We foresee that together with other promising biomarkers, GM composition and diversity will be integrated into a multiparameter model to accurately predict which subset of melanoma patients are likely to respond to ICB[10,11,47].

ARTICLE HIGHLIGHTS

Research background

Survival for metastatic melanoma has significantly improved since the introduction of immune checkpoint blockade (ICB) therapy. However, despite their considerable efficacy, 40%-60% of melanoma patients do not experience objective responses to the therapy. Additionally, some patients experience ICB-related colitis as a consequence of ICB therapy, preventing them from deriving the full benefit of ICB therapy. Recent studies have demonstrated that the gut microbiome (GM) may affect tumor immunity by regulating the host immune system and tumor micro-environment, thus suggesting that GM may affect response to ICB therapy and susceptibility of ICB-related colitis.

Research motivation

The GM has shown great potential as a biomarker of response to ICB therapy in melanoma patients. Previous studies investigating GM composition and/or diversity in patients with melanoma have identified distinct GM composition and diversity in responders to ICB compared to non-responders, as well as those more susceptible to ICB-related colitis than those who are not.

Research objectives

To be the first to compile the existing data regarding the role of GM composition and diversity in predicting response to ICB and ICB-related colitis specifically in patients with melanoma.

Research methods

Comprehensive literature search was done in various platforms using the following search terms: (fecal OR gut) AND (microbiota OR microbiome) AND (melanoma) AND (immunotherapy OR checkpoint OR nivolumab OR ipilimumab OR pembrolizumab). From a total of 300 studies, nine studies met inclusion criteria. Two studies were phase I clinical trials, while the remainder were prospective observational studies. All but one study has moderate risk of bias. Data from these studies including but not limited to, number of participants, type of immunotherapy received, GM analysis method, and GM composition and diversity were collected and interpreted.

Research results

Fecal samples enriched in Firmicutes phylum were associated with good response to ICB therapy, however they were associated with increased susceptibility to ICB-related colitis. Fecal samples enriched in Bacteroidales family were associated with poor response to ICB. Samples with greater GM diversity were associated with more favorable response to ICB. Fecal samples enriched in Bacteroidetes were associated with decreased incidence of ICB-related colitis. Overall, there was limited concordance in the organisms in the GM identified to be associated with response to ICB, and studies evaluating GM diversity showed conflicting results.

Research conclusions

GM composition and diversity holds some potential as a biomarker of response and toxicity to ICB in melanoma. Further prospective studies, including several RCTs that are underway, are needed to confirm whether the GM could be used as a biomarker and potential intervention to modulate ICB response in melanoma patients.

Research perspectives

With other promising biomarkers, GM composition and diversity holds potential to be integrated into a multiparameter model to accurately predict which subset of melanoma patients are likely to respond to ICB.