1. Common features in gut microbiome correlate with the treatment response and AEs of ICIs across various solid tumors
Using search keywords “gut microbiome” AND “cancer” AND “immunotherapy”, a total of 240 articles were retrieved from PubMed. With stringent manual screening and inclusion of 2 additional studies from the references, a total of 12 clinical studies were identified that meet our criteria to study the role of gut microbiome in cancer immunotherapy (Figure 1). The vast majority are prospective studies. Among them, 10 studies[6, 9, 12-19] had response/efficacy data and 3 studies[17, 20, 21] had AEs data using ICI therapy, and 1 study had both[17]. Of note, the documented AEs in that 3 studies[17, 20, 21] were virtually all irAEs. The types of solid tumor involved include melanoma, non-small cell lung cancer (NSCLC), small cell lung cancer (SCLC), renal cell carcinoma (RCC) and hepatocellular carcinoma (HCC). There were 433 cancer patients (age range, 21-92 years-old) from four countries: USA (4 studies), France (3 studies), China (2 studies) and Japan (1 study) included in studies relevant to therapeutic response/efficacy; and 86 subjects (age range, 28-85 years-old) from the countries of USA, France and China included in studies relevant to the AEs of ICI treatment.
We extracted the taxa data of gut microbiome and plotted on phyloT. As shown in Figure 2, at the phylum level, it is clear that the enrichment of Firmicutes and Verrucomicrobia are correlated with better clinical outcome (labeled in green; related to better treatment response and/or longer survival), whereas increased abundance in Proteobacteria was clearly associated with poor response (labeled in red). Although enrichment of Bacteroidetes correlated to poor response in some studies, opposite association and contradictory findings (labeled in grey) were also noticed in some other studies. Similarly, a mixed association of Actinobacteria to ICI treatment response was noticed.
However, regarding the potential link of gut microbiome to the AEs from ICIs, we noticed that the enrichment of Firmicutes interestingly correlated to higher incidence of AEs (essentially all irAEs, colored in red). This is reminiscent of clinical observations that patients who develop ICI AEs seem to have better treatment response[22]. In contrast, Bacteroidetes, which is believed to be associated with less response, also correlated to less AEs (labeled in green, Figure 3).
2. The potential impact of antibiotics on the therapeutic effect of ICIs
Noticing the association of gut microbiome with ICI treatment response, we questioned if antibiotics, as potent modifiers of gut microbiota, could potentially affect the treatment response from ICIs. Using search keywords “antibiotics” AND “immunotherapy” AND “microbiome” AND “cancer”, we identified 17 eligible studies (Supplemental Figure 1) including 2 prospective[23, 24] and 15 retrospective studies[9-11, 25-36]. There were in total 2,593 participants with various solid tumors including lung cancer, melanoma, RCC, HCC, colorectal cancer, head and neck cancer, bladder cancer, gastric cancer, esophageal cancer, cervical cancer and others. Among them, 29.9% (775) of them received antibiotics treatment, 15 out of 17 received broad-spectrum antibiotics while 2 did not report the types of antibiotics.
As shown in Figure 4a, majority of these studies supported the hypothesis that the use of antibiotics has negative impact on the clinical outcome in patients receiving ICI treatment. However, there were also a few studies that suggested no obvious association or impact. Interestingly, 2 prospective studies[23, 24] and 1 retrospective study[25] provided seemingly different results (negative vs. no impact) when different timing of antibiotic exposure was put into consideration, suggesting that the timing and possibly the duration of antibiotics during ICI treatment are potentially important and will need further studies to clarify its impact..
In order to validate this hypothesis, we isolated the effect of the timing and duration of antibiotic exposure from all studies. Supplemental Figure 3 showed individual studies that exhibited either negative (labeled with black bars ) or no association (labeled in grey bars) with ICI treatment. Among them, two studies[23, 24] were prospective (labeled with *). Across all studies, it clearly demonstrated that only antibiotic exposure within 2 months prior to the initiation of ICIs universally exhibited negative impact on treatment response of ICIs (Figure 4b and c), except one study[10] (Supplemental Figure 3).
3. Diet could potentially affect the efficacy of cancer immunotherapy
Using search keywords “diet” or “nutrition”, “microbiome”, “cancer” and “immunotherapy”, and their combinations, we were not able to extract sufficient number of clinical studies that directly link diet to cancer immunotherapy, including those published in abstract format[37], which is suggestive of an unmet need in this area. Since gut microbiome impacts cancer immunotherapy, we then investigated whether diet will have effect on gut microbiome that could potentially affect cancer immunotherapy. Based on Figure 2 and published data, here we define “ICI-favoring” diet as those that enrich Firmicutes or Verrucomicrobia, or reduce the abundance of Proteobacteria, or increase α diversity in gut microbiota, and the “ICI-unfavoring” diet as those that have the opposite effects.
To minimize confounding factors (especially various disease status), we used search terms “diet” AND “gut microbiome” AND “healthy adult” and included only clinical studies in healthy participants that have detailed diet and gut microbiome information (Supplemental Figure 2). We identified 16 eligible clinical studies[38-53] that included in total 771 subjects. Among them, 428 were females and 343 were males. Their age ranged 18 -72.4 years and BMI ranged 19 - 36.6 kg/m2. These clinical studies were conducted in five countries including USA, China, Germany, UK and Belgium. We broadly categorized diet into plant-based diet which mainly contained whole grain, brassica vegetables, walnut and almond, etc; and animal-based diet which used red meat, animal fat and cheese, etc. There are only 4 studies using animal-based diet[40, 41, 47, 53]. Although they also contained non-animal-based diet component, we were able to precisely derive data that are only relevant to animal-based diet.
Figure 5a in each category, depicts increase or decrease in relative abundance of Firmicutes or Verrucomicrobia or Proteobacteria or α diversity with demonstration of corresponding plant-based diet (labeled as solid dot) and animal-based diet (labeled as hollow circle), respectively. Using above defined “ICI-favoring” and “ICI-unfavoring” criteria, we found that 3 animal-based diet studies were “ICI-unfavoring” and none were “ICI-favoring”. Among the 12 plant-based diet studies, we found 5 were “ICI-favoring” and 1 was “ICI-unfavoring”. In summary, plant-based diet is found to be significantly associated with “ICI-favoring” gut microbiome, whereas animal-based diet is the opposite (Figure 5b, p=0.0476). Diet studies that have mixed association, for example a reduced abundance in both the Firmicutes and Proteobacteria as shown in study n (Henning et al. 2017, reference) in Figure 5a, were not included in the statistical analysis. We have also looked into various dietary patterns such as Mediterranean diet, Western diet, high-fiber diet, etc., however we were able to identify only very few relevant studies for us to draw meaningful conclusions.