Characteristics of the gut microbiota in patients with advanced non-small cell lung cancer who responded to immune checkpoint inhibitors

Abstract

Despite the introduction of immune checkpoint inhibitors (ICIs) in the treatment of lung cancer, the number of deaths from lung cancer remains high, and further improvements in response rates are necessary. Recently, the gut microbiota has been reported to be involved in the therapeutic effects of ICIs; however, only a few studies have examined patients with lung cancer in this context. In the current study, we aimed to explore the association between the gut microbiota before therapy and the efficacy of ICIs in patients with advanced non-small cell lung cancer (NSCLC). The a-diversity of the intestinal microbiota in patients who responded to ICI treatment (responders) was significantly higher than that in those who did not respond to ICIs (non-responders). Additionally, the abundance of Bifidobacteriaceae was significantly higher in the responders than in the non-responders. Furthermore, patients with a high abundance of Bifidobacteriaceae had significantly longer overall survival than those with a low abundance. Counts of Levilactobacillus brevis were significantly higher in responders than in non-responders. Our findings suggest that a higher diversity of the gut microbiota and an abundance of Bifidobacterium and/or L. brevis are distinctive features of the microbiota in patients with NSCLC who respond to ICI treatment.

Introduction

Approximately 2.2 million individuals are newly diagnosed with lung cancer annually worldwide¹, and approximately 75% of these patients die within five years after diagnosis². In Japan, approximately 120,000 individuals are diagnosed with lung cancer every year, which is also the leading cause of cancer-related deaths³. Despite efforts to improve early detection, many cases have been shown to have progressed to an advanced stage by the time of initial presentation and are not eligible for surgery. Therefore, further advances in detection are necessary for appropriate diagnosis and treatment. The emergence of immune checkpoint inhibitors (ICIs), such as those targeting programmed cell death 1 (PD-1) or its ligand (PD-L1), has transformed the treatment and management of advanced non-small cell lung cancer (NSCLC), which constitutes most lung cancers¹. Several studies have established the anticancer effects of anti-PD-1 antibodies^4–6, and PD-L1 expression in tumor tissues has been validated as a companion diagnostic test in clinical practice. Nevertheless, the efficacy of ICI treatment in NSCLC remains insufficient in patients with specific metastases or those requiring retreatment. These facts highlight the need for continued research to enhance the effectiveness of ICI treatment in NSCLC and identify patient subgroups most likely to benefit^7,8. Identifying biomarkers that can more accurately predict clinical responses and/or developing more effective means of sensitizing tumors to ICI therapy is a pressing need to improve the efficacy of ICI therapy in advanced NSCLC.

In recent years, it has become evident that the gut microbiota plays a crucial role in the response to cancer therapy, including response to ICI treatment, and can be targeted to enhance the efficacy of cancer immunotherapy^9–11. Studies have confirmed the relationship between specific microbial species and their effects on cancer immunotherapy for various bacterial species¹². In a clinical study investigating the therapeutic effect of anti-CTLA-4 antibodies on advanced malignant melanoma, patients with Faecalibacterium and other Bacillota (formerly Firmicutes) demonstrated longer progression-free survival (PFS) and overall survival (OS) than those with a predominance of Bacteroidota (formerly Bacteroidetes)¹³.

Several studies have revealed a significant disparity in the relative abundance of Akkermansia muciniphila, Alistipes putredinis, Bifidobacterium longum, Prevotella copri, Ruminococcaceae USG 13, and Agathobacer in the gut microbiota of patients with lung cancer who responded to treatment (responders) and those who did not (non-responders). These species are associated with a favorable clinical outcome in patients with NSCLC^14–16. Meanwhile, in terms of poor prognostic predictors, an increased abundance of Firmicutes and Actinobacteria phyla, as well as Bifidobacterium breve has been reported to correlate with short PFS¹⁷. Despite numerous studies conducted in this domain in recent years, the results are inconsistent. Several questions remain unanswered regarding the relationship between the gut microbiota and treatment response in patients with lung cancer. The observed inconsistencies in the results may be attributed to individual variations in intestinal flora, which are influenced by the host’s genetic makeup, history of antibiotic usage, and dietary habits. Additional factors contributing to these inconsistencies include geographical differences, racial diversity, variations in sample collection timeframes, reference databases, and study endpoints. Furthermore, technical discrepancies, such as those arising from 16S rRNA sequencing and metagenomic approaches, may also play a role. The functional diversity of the gut microbiota is considered a significant factor as well^17,18. In particular, we believe that the conventional wisdom that the human gut microbiome differs according to the race and residential area of the study subjects has a significant impact on the inconsistent results¹⁹. For example, Bifidobacterium spp. is common among Japanese individuals^20,21. It is important to identify unique and original species and the composition of microbiota depending on race and residential area. Focusing on regional characteristics makes it possible to determine a relationship between regional bacterial flora and therapeutic effects. Consequently, we posit that the human intestinal microbiota exhibits significant regional variation, primarily influenced by dietary habits. Common bacterial species beneficial to ICI treatment response are unlikely to be universally identified across different regions and ethnicities. Therefore, instead of pursuing a universal approach to predict intestinal microbiota profiles and ICI treatment response, it may be more pragmatic to identify biomarkers predictive of ICI treatment response by focusing on regions with similar dietary practices and intestinal microbiota profiles.

Therefore, in the current study, we aimed to investigate the association between the gut microbiota before therapy and the efficacy of ICI therapy in patients with advanced NSCLC in Osaka. Stool samples were collected from patients prior to initiation of ICI treatment, and the fecal bacterial content was analyzed. The gut microbiota of responders and non-responders was prospectively compared, and the relationships between specific bacterial groups and response rates to ICIs, antibiotic history, and survival time were examined.

Methods

Study design

This single-center study was designed as a prospective observational investigation to explore the relationship between the gut microbiome before therapeutic intervention and the efficacy of ICI therapy in patients with NSCLC.

Participants

We enrolled patients with a definitive diagnosis of NSCLC who underwent ICI therapy between June 2017 and June 2018 at the National Hospital Organization NHO Osaka Toneyama Medical Center. A total of 23 patients were included in the study, all of whom met the selection criteria after receiving an explanation and providing informed consent. Finally, analysis was performed on 19 cases that submitted analyzable samples. The samples were numbered with experimental ID and then sent to the Yakult Central Institute where microbiota analysis was carried out. Therapeutic effects were diagnosed using the Response Evaluation Criteria in Solid Tumors (RECIST) guidelines (version 1.1), and we defined responders (R) (partial response and stable disease after ICI treatment) and non-responders (NR) (progressive disease after ICI treatment). Information regarding the results of diagnosis and stratification (R or NR) was not disclosed until the data were fixed in the microbiota analysis. After fixing the data in microbiota analysis, the samples and therapeutic effects were connected according to the experimental ID.

Preparation of stool samples and DNA extraction

Fecal samples from 19 patients were obtained before ICI therapy commenced as per the study’s explanatory documentation. Stool samples were prepared as described by Kubota et al.²². Each sample was weighed and diluted tenfold using RNAlater (feces:RNAlater = 1:9). The mixture was then agitated for 10 min at 1048 rpm using a ShakeMaster Auto (Bio Medical Science Co., Ltd., Tokyo, Japan). Subsequently, 200 µL of the suspension prepared from the fecal diluent was added to a 2 mL screw cap tube containing 1 mL of phosphate-buffered saline. The contents were stirred using a vortex mixer and centrifuged at 13,000 × g for 5 min at 4 °C, after which 1 mL of the supernatant was discarded. This process was repeated twice. The resulting suspension (200 µL) was stored at –30 °C until DNA extraction, which was carried out as previously described by Matsuki et al.²³.

16S rRNA gene sequence for microbiota analysis

The V1-2 regions of the 16S rRNA gene in each sample were amplified with an ABI PRISM® 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA), using 27Fmod2 and 338R primers, as previously described²⁴. Amplified DNA was purified using the AMPure XP kit (Beckman Coulter, Brea, CA, USA). After quantitation using the Quant-iT Pico-Green double-stranded DNA assay kit (Life Technologies Japan, Ltd.), the gene was sequenced on MiSeq system (Illumina, Inc., San Diego, CA, USA) using MiSeq Reagent Kits v2 (Illumina, Inc.), as reported in our previous study²⁵. The sequences obtained from the MiSeq platform were analyzed using the open-source software package Quantitative Insights Into Microbial Ecology 2 (QIIME2) (2020.2)²⁶, and the SILVA 138 database (https://www.arb-silva.de/) was used to annotate taxonomic information. To account for differences in sampling effort between samples, the Shannon index, Faith’s phylogenetic diversity (PD), and the number of observed features was estimated for 13,000 randomly selected sequences to calculate a-diversities. The sequence count of 13,000 was determined based on the sample with the smallest number of reads, following confirmation of an adequate number of reads by consulting the rarefaction curve.

Determination of bacterial count by rRNA-targeted reverse transcription quantitative PCR

We used the Yakult Intestinal Flora-SCAN (YIF-SCAN®), a bacterial rRNA-targeted reverse transcription quantitative PCR (RT-qPCR) system, to analyze the composition of the gut microbiota^24–26. The threshold values of the standard curve in the linear range were used to obtain the corresponding bacterial cell count in each nucleic acid sample, which was used to determine the number of bacteria per sample. The specificity of the RT-qPCR assay was determined using group, genus, and species-specific primers as previously described^27–30. The primer sequences are listed in Supplementary Table S1^{23,27–29,31–36}. The total number of bacterial groups was considered as the total number of bacteria. In the present study, the bacterial counts of 17 of 19 patients were compared. Specimens from two participants were excluded because their feces were poorly preserved and were not suitable for measuring bacterial counts.

Statistical analysis

Statistical analyses were conducted using R software, v.3.5.2. Fisher’s exact test was used to compare the patient detection rates between the groups. For data with a skewed distribution, the median [interquartile range] was used and the Mann–Whitney U test was used for analysis. The number of gut bacteria was logarithmically transformed for analysis. When the microbiota counts were lower than the limit of detection, a value of half the detection limit was assigned for statistical calculation. Statistical significance was established at p < 0.05. Cox proportional hazards regression analysis was used to compare disease progression and death between two groups stratified based on body mass index (BMI) and diversity and abundance of Bifidobacteriaceae, independently. The patients were divided into the “Yes” group and the “No” group, or the positive group and the negative group based on antibiotic treatment and the detection of L. brevis, respectively. Patients with median values of BMI, diversity, and abundance of Bifidobacteriaceae were assigned to the “low BMI”, “low diversity,” and “low abundance of Bifidobacteriaceae” groups, respectively, because their values were below average. Progression-free survival (PFS) and overall survival (OS) were calculated using the Kaplan–Meier method, and the survival distributions between the two groups, high and low abundance of Bifidobacteriaceae, and between two groups, L. brevis-positive, and L. brevis-negative, were compared using the log-rank test.

Ethics statement

The Institutional Ethics Committee of Osaka Toneyama Medical Center approved this study (1706-8), and written informed consent was obtained from all patients. The study was conducted in accordance with the Ethical Guidelines for Medical and Health Research Involving Human Subjects and current legal regulations in Japan and the Declaration of Helsinki and its subsequent revisions.

Results

Baseline characteristics of the participants

Table 1 shows the pretreatment characteristics of the 19 patients before initiation of ICI therapy. A notable difference was observed in BMI between the R and NR groups, with the former exhibiting higher values. However, no significant differences were observed for any of the other parameters.

Table 1.

Patient characteristics at baseline.

Variables	R (n = 12)	NR (n = 7)	p-value
Sex, male/female	8/4	3/4	0.377
Age (years)§	74.0 [66.0, 76.5]	72.0 [68.5, 76.0]	0.832
Smoking status, ex/never	10/2	5/2	0.603
Histology, non-Sq/Sq	7/5	6/1	0.333
BMI (kg/m2)§	23.2 [22.0, 24.3]	19.3 [18.0, 21.1]	0.028
Staging, ≤ III/IV	4/8	3/4	1.000
ECOG performance status, ≤ 1/ ≥ 2	10/2	6/1	1.000
PD-L1 expression,  < 50%/ ≥ 50%	4/8	3/4	1.000
0%/ ≥ 1%	2/10	1/6	1.000
ICIs, nivolumab/pembrolizumab	3/9	2/5	1.000
ICI treatment line, first/second or later	8/4	3/4	0.377
PNI§	42.8 [38.0, 47.4]	39.5 [37.8, 45.3]	0.472
NLR§	3.44 [2.55, 4.15]	4.32 [3.56, 4.76]	0.196
Antibiotics prior to treatment, no/yes$	9/3	2/5	0.074
Steroids prior to treatment, no/yes	10/2	6/1	1.000

R responder, NR non-responder, Sq squamous, BMI body mass index, ECOG Eastern cooperative oncology group, PNI prognostic nutritional index, NLR neutrophile/lymphocyte ratio.

p-values were estimated using the Mann–Whitney U test for continuous variables or Fisher’s exact test for categorical variables. ^§Median [interquartile range].

^$Patients who received antibiotic treatment for more than four consecutive days within the last 1 month prior to ICI treatment. Patients who underwent bronchoscopy within one month prior to initiating ICI treatment were administered short-term antibiotics to prevent infection, with the maximum duration being four consecutive days. Consequently, cases were categorized into a high exposure intensity group and a relatively low exposure intensity group based on the criterion of receiving antibiotic treatment for four consecutive days. p-value entries in bold indicate p < 0.05.

Fecal microbiota measurement

The first step involved conducting a diversity analysis of the two groups of patients. The results indicated that the Shannon index and Faith’s PD were higher in the R group than in the NR group (Table 2). The relative abundance of intestinal bacteria was then compared in the R and NR groups (Table 3, Supplementary Table S2). At the family level, the percentage of Bifidobacteriaceae was significantly higher in the R group than in the NR group (Table 3). The PD-L1-positive rate was 16/19 in the participants of the present study (Table 1), and there was no significant difference in the PD-L1-positive rate/ PD-L1-expression rate between the high and low groups of a-diversities and relative abundance of Bifidobacteriaceae (Supplementary Table S3). Finally, a comparison was made between the bacterial counts in each group, as determined by RT-qPCR, between the R and NR groups (Table 4). Interestingly, L. brevis was detected only in patients in the R group (6/10 patients, p = 0.035) (Supplementary Table S4), and the L. brevis counts in the R group were significantly higher than those in the NR group (Table 4).

Table 2.

Differences in a-diversities in the gut microbiome of R and NR groups.

	R	NR	p-value
Shannon index	6.47 [6.11, 6.72]	5.67 [5.50, 5.93]	0.045
Faith’s PD	37.1 [33.5, 47.3]	30.3 [26.9, 30.7]	0.022
Observed features	405 [330, 519]	314 [247, 364]	0.100

R responder, NR non-responder, PD phylogenetic diversity.

Data are expressed as median [interquartile range].

p-values were estimated using the Mann–Whitney U test. p-value entries in bold indicate p < 0.05.

Table 3.

Differences in relative abundances at the family level by 16S rRNA sequencing in the gut microbiome of R and NR groups.

Family	Relative abundance (%)				p-value
	R		NR
Bifidobacteriaceae	2.96	[0.77, 9.83]	0.18	[0.09, 1.22]	0.045
Coriobacteriaceae	0.62	[0.26, 1.40]	0.42	[0.13, 2.26]	0.800
Eggerthellaceae	0.25	[0.15, 0.45]	0.08	[0.06, 0.43]	0.340
Bacteroidaceae	19.46	[11.04, 28.71]	30.17	[18.89, 37.98]	0.261
Barnesiellaceae	0.37	[0.03, 1.13]	0.00	[0.00, 1.00]	0.306
Marinifilaceae	0.33	[0.06, 0.69]	0.37	[0.16, 0.56]	1.000
Prevotellaceae	0.68	[0.01, 2.92]	6.28	[0.01, 12.60]	0.650
Rikenellaceae	1.31	[0.55, 3.63]	0.79	[0.45, 2.44]	0.592
Tannerellaceae	3.83	[1.43, 4.25]	3.06	[2.19, 3.71]	0.902
Desulfovibrionaceae	0.11	[0.03, 0.20]	0.11	[0.04, 0.14]	0.553
Erysipelatoclostridiaceae	0.35	[0.17, 1.09]	0.35	[0.23, 1.61]	0.536
Erysipelotrichaceae	0.57	[0.25, 1.20]	0.72	[0.19, 2.01]	0.773
Lactobacillaceae	0.16	[0.01, 0.52]	0.10	[0.01, 0.20]	0.650
Streptococcaceae	3.06	[0.78, 9.07]	1.37	[0.60, 4.39]	0.592
Christensenellaceae	0.14	[0.05, 0.93]	0.04	[0.01, 0.06]	0.090
Clostridiaceae	0.32	[0.02, 1.21]	0.03	[0.01, 0.46]	0.271
Lachnospiraceae	18.50	[13.04, 23.54]	24.27	[20.25, 28.43]	0.142
Monoglobaceae	0.10	[0.01, 0.12]	0.03	[0.00, 0.19]	0.865
Butyricicoccaceae	0.10	[0.05, 0.22]	0.14	[0.06, 0.24]	0.592
Oscillospiraceae	2.97	[0.87, 4.05]	0.86	[0.65, 1.11]	0.120
Ruminococcaceae	8.37	[4.61, 10.71]	1.94	[1.26, 4.87]	0.056
Anaerovoracaceae	0.21	[0.08, 0.36]	0.14	[0.05, 0.30]	0.612
Peptostreptococcaceae	0.98	[0.19, 1.64]	0.50	[0.09, 1.67]	0.902
Acidaminococcaceae	0.19	[0.01, 0.40]	1.03	[0.00, 1.22]	0.520
Veillonellaceae	0.15	[0.02, 0.49]	0.06	[0.01, 2.63]	0.773
Sutterellaceae	0.26	[0.00, 0.93]	0.03	[0.00, 0.64]	0.670
Enterobacteriaceae	0.37	[0.07, 0.76]	0.62	[0.21, 2.49]	0.711

R responder, NR non-responder.

Data are expressed as median [interquartile range]. Only families with median abundance above 0.1% are shown.

p-values were estimated using the Mann–Whitney U test. p-value entries in bold indicate p < 0.05.

Table 4.

Fecal counts of the gut microbiota determined by RT-qPCR.

	Fecal bacterial counts (log10 cells/g of feces)				p-value
	R		NR
Total bacteria	10.1	[9.70, 10.4]	9.97	[9.82, 10.2]	0.89
Clostridium coccoides group	8.84	[8.49, 9.34]	9.36	[9.19, 9.48]	0.32
C. leptum subgroup	9.13	[9.02, 9.33]	9.39	[9.08, 9.70]	0.48
Bacteroides fragilis group	7.70	[6.75, 8.59]	8.62	[7.99, 8.94]	0.32
Bifidobacterium	9.19	[8.48, 9.32]	8.22	[6.98, 9.18]	0.32
Atopobium cluster	9.37	[9.00, 9.54]	9.41	[8.41, 9.49]	0.89
Prevotella	5.16	[2.50, 6.93]	5.34	[2.50, 7.92]	0.61
C. perfringens	2.48	[1.15, 4.21]	1.15	[1.15, 4.19]	0.62
C. difficile	Not detected		Not detected		–
Total lactobacilli	7.17	[6.51, 8.14]	6.76	[6.09, 7.42]	0.32
Lactobacillus	5.70	[5.01, 7.05]	5.94	[4.60, 6.47]	0.85
(Formerly Lactobacillus gasseri subgroup)
Levilactobacillus brevis	2.82	[1.15, 3.74]	Not detected		0.017
(Formerly Lactobacillus brevis)
Lacticaseibacillus	5.75	[2.16, 5.87]	1.50	[1.50, 4.45]	0.19
(Formerly Lactobacillus casei subgroup)
Limosilactobacillus fermentum	6.77	[3.06, 7.89]	5.60	[3.78, 6.99]	0.37
(Formerly Lactobacillus fermentum)
Fructilactobacillus fructivorans	1.15	[1.15, 1.15]	Not detected		0.40
(Formerly Lactobacillus fructiborans)
Lactiplantibacillus	4.70	[3.29, 5.42]	3.29	[1.96, 4.51]	0.20
(Formerly Lactobacillus plantarum subgroup)
Limosilactobacillus (except L. fermentum)	5.35	[4.82, 6.46]	4.98	[4.12, 5.61]	0.26
(Formerly Lactobacillus reuteri subgroup)
Ligilactobacillus and Liquorilactobacillus	1.15	[1.15, 1.15]	1.15	[1.15, 1.84]	0.42
(Formerly Lactobacillus ruminis subgroup)
Latilactobacillus	3.55	[1.65, 4.09]	4.04	[1.15, 4.71]	0.88
(Formerly Lactobacillus sakei subgroup)
Enterobacteriaceae	6.39	[5.21, 7.55]	5.93	[5.63, 7.47]	0.85
Enterococcus	5.52	[2.42, 6.23]	4.58	[3.01, 6.89]	0.92
Streptococcus	9.03	[8.50, 9.93]	8.75	[7.47, 9.32]	0.23
Staphylococcus	4.33	[3.24, 5.02]	4.13	[2.32, 4.68]	0.56
Pseudomonas	Not detected		1.50	[1.50, 2.98]	0.082

R responder, NR non-responder.

Data are expressed as median [interquartile range]. p-values were estimated using the Mann–Whitney U test. p-value entries in bold indicate p < 0.05.

Influence of antibiotics on the gut microbiota

We examined the impact of antibiotics on the gut microbiota. Specifically, we compared the relative abundance of bacteria between antibiotics prior to treatment (“Yes” group) and non-treatment (“No” group) at the family level. Our findings revealed that the percentage of Bifidobacteriaceae was significantly lower in patients in the antibiotic “Yes” group than in the antibiotic “No” group (Supplementary Table S5). Additionally, the count of L. brevis was significantly lower in in patients the antibiotic “Yes” group than in the antibiotic “No” group (Supplementary Table S6).

Analysis of relationships between clinical and microenvironmental parameters of the gut and prognosis

Table 5 illustrates the results of univariate analysis of the characteristics of the patients and gut microbiota at baseline for disease progression and death. In terms of disease progression, there was a significant difference in BMI (p = 0.044) and counts of L. brevis (p = 0.024). In the low-BMI group, the risk was higher, and in the L. brevis-positive group, the risk was lower. In terms of death, the risk was significantly increased in the antibiotic “Yes” group (p = 0.046). However, high diversity indicated by the Shannon index, high abundance of Bifidobacteriaceae, and the L. brevis-positive group showed lower risk (p = 0.022, p = 0.015, and p = 0.048, respectively). Considering these findings, further analysis was conducted to examine the correlation between the two characteristics of the gut microbiota and prognosis using the log-rank test. The results revealed that the “high group” had longer PFS and OS than the “low group “ in terms of both the abundance of Bifidobacteriaceae (Fig. 1a) and counts of L. brevis (Fig. 1b).

Table 5.

Univariate analysis of characteristics of patients and the gut microbiota at baseline for disease progression and death.

Prognostic factors	For progression		p-value	For death		p-value
	HR	95% CI		HR	95% CI
BMI, < 22/ ≥ 22	3.19	1.03–9.85	0.044	1.48	0.39–5.52	0.563
Antibiotics prior to treatment, yes/no	1.79	0.55–5.81	0.331	4.25	1.03–17.6	0.046
Shannon index, high/low	0.59	0.20–1.78	0.351	0.16	0.03–0.77	0.022
Faith’s PD, high/low	0.75	0.25–2.24	0.602	0.51	0.14–1.85	0.308
Bifidobacteriaceae, high/low	0.36	0.11–1.19	0.094	0.07	0.01–0.61	0.015
L. brevis, positive/negative	0.20	0.05–0.81	0.024	0.12	0.02–0.98	0.048

HR hazard ratio, CI confidence interval, BMI body mass index, PD phylogenetic diversity.

p-values were estimated using Cox proportional hazards regression analysis. p-value entries in bold indicate p < 0.05.

Fig. 1.

Association between the gut microbiome and response to ICI therapy. Comparison of Kaplan–Meier plot survival curves by a log-rank test. PFS progression-free survival, OS overall survival, NA not available, HR hazard ratio, CI confidential interval. (a) Patients with high and low abundance of Bifidobacteriaceae. (b) Patients in L. brevis-positive and L. brevis-negative groups.

Discussion

The objective of our study was to establish a connection between the gut microbiota prior to therapy and the efficacy of ICIs in patients with advanced NSCLC. Nineteen individuals were enrolled and stratified into R and NR groups based on their baseline characteristics. A significant difference in BMI was observed between the two groups (Table 1), which indicated that BMI can affect prognosis after ICI treatment. BMI may influence the effectiveness of ICI therapy in patients with advanced NSCLC, and patients with a high BMI may experience longer PFS and OS than those with a low BMI^37,38. Consequently, patients with advanced NSCLC and a low BMI may have a poorer prognosis when undergoing ICI therapy.

Initially, our study revealed that the patients in the R group exhibited elevated a-diversity in their fecal microbiome, as determined by both the Shannon index and Faith’s PD (Table 2). In particular, individuals who responded positively to ICI treatment possess a diverse gut microbiota, which is consistent with our findings³⁹.

The analysis of the intestinal microbiota via NGS allowed to stratify patients according to R and NR groups and revealed that the abundance of Bifidobacteriaceae was significantly higher in the R group. In other words, the results suggest that ICI treatment may be more effective in patients with more indigenous bifidobacteria (Table 3). Previous studies have suggested that Bifidobacterium is more abundant in ICI treatment responders than in non-responders^14,40. This implies that Bifidobacterium plays a crucial role in enhancing the antitumor efficacy of ICIs. There was no significant difference in PD-L1-positive rate/ PD-L1-expression rate between the high and low groups of a-diversities and relative abundance of Bifidobacteriaceae (Supplementary Table S3). Thus, the possibility of a relationship between the intestinal microbiota and PD-L1 expression was ruled out.

In addition to the NGS results, the results of the gut microbiota analysis by RT-qPCR for the R and NR groups showed that the abundance of L. brevis was significantly higher in the R group (Table 4). Univariate analysis demonstrated that a high abundance of Bifidobacteriaceae and/or the presence of L. brevis in the gut microbiota significantly reduced the risk of death during ICI therapy. Log-rank analyses of PFS and OS suggested that the presence of Bifidobacteriaceae and/or L. brevis in the gut microbiota before administration of ICI may contribute to prolonged OS (Fig. 1). Bifidobacterium improves the effectiveness of anti-PD-L1 antibodies^15,40, leading to prolonged OS in patients with high Bifidobacteriaceae abundance. In fact, the relative abundance of Bifidobacteriaceae in patients in the “Yes” group, who received antibiotics prior to treatment, was significantly lower than that in patients in the “No” group (Supplementary Table S5). The effect of antibiotic dysbiosis was also observed in terms of L. brevis counts (Supplementary Table S6). Notably, the risk of death was significantly higher in the “Yes” group than in the “No” group (Table 5), and fewer antibiotic users were in the R group than in the NR group (Table 1).

In the analysis of the relationship between clinical and microenvironmental parameters of the gut and prognosis, the risk of death was significantly increased in the antibiotic “Yes” group. However, the risk of disease progression was not significant (Table 5). Since PFS is the duration until NSCLC progression, and OS is the duration until death caused by not only NSCLC but also all other reasons, they do not always fluctuate in the same manner. In the case of antibiotics especially, the risk was significantly increased in the antibiotic “Yes” group, because patients’ condition was bad and they might also have other disease, NSCLC progression itself might not be the cause of death.

Although antibiotics are essential for the treatment of tract infections and pneumonia, their use before starting ICI treatment likely disrupts the intestinal environment, affects the abundance of Bifidobacteriaceae, and reduces the efficacy of ICI treatments. Consistent with our results, Derosa et al. reported that antibiotic use reduced the therapeutic efficacy of ICIs⁴¹. The antibiotic usage prior to at least 6 months before the sample collection would be more appropriate to document. However, it was thought to be difficult to collect a sufficient number of patients who did not use antibiotics during the 6 months before sample collection in the NCSCL. Although the study did not evaluate changes in the gut microbiota due to antibiotic use, it suggested that the recovery and alleviation of gut microbiota dysbiosis, particularly the abundance of Bifidobacteriaceae, is important.

This study demonstrated that a high occupancy rate of bifidobacteria may be advantageous in ICI treatment. Nonetheless, a review of prior reports has indicated that bifidobacteria may exert a negative impact on ICI treatment^17,18. We achieved high accuracy in detecting bifidobacteria using 16S rRNA V1-2 sequencing; however, as noted in the limitations, we did not investigate the underlying mechanism. Short-chain fatty acids, the primary metabolites of bifidobacteria, have been reported to exert diverse physiological effects ranging from the intestinal tract to the entire body; however, their contribution to the therapeutic effects of ICI remains unknown^42,43. Although we did not examine bifidobacteria metabolites in this study, the discrepancy in occupancy rate comparisons may be elucidated by examining the mutual interactions with other coexisting microorganisms through an analysis of the functional diversity of the intestinal microbiota, including the effects of metabolites¹⁸.

In the present study, we identified L. brevis as a novel bacterium that may enhance the efficacy of ICIs. Patients who tested positive for L. brevis exhibited significantly longer PFS and OS than those who tested negative for L. brevis (Fig. 1b). Previous studies have reported anti-inflammatory^44,45 and immunostimulatory⁴⁶ effects of L. brevis; however, this study is the first to establish a connection between L. brevis and the efficacy of ICI treatment. The number of L. brevis in this species is typically low in the gut microbiota, with a low detection rate^47–49. However, in this study, the detection rate for L. brevis was high (6/17 patients). We did not investigate the history of food consumption of the participants; therefore, we cannot explain the high detection rate of L. brevis, especially in the R group (6/10 patients). Notably, Osaka, where this study was conducted, is situated near Kyoto, where L. brevis-rich pickles are traditionally consumed. Therefore, further investigation in a larger cohort is necessary to clarify the relationship between L. brevis and ICI treatment, considering the regional food culture.

Our study has certain limitations in terms of scope. Therefore, a larger sample size is necessary for a more comprehensive examination. Thus, additional large-scale studies are required to corroborate the role of Bifidobacteriaceae and L. brevis in the context of ICI therapy. We have no information that makes it possible to speculate on the mechanism of the role of the microbiota in the efficacy of ICI.

In conclusion, the key findings of this observational study revolved around the significance of the gut microbiome in patients with advanced NSCLC receiving ICI therapy. Specifically, the diversity and abundance of Bifidobacteriaceae in the fecal microbiomes of responders were higher than those of non-responders. Furthermore, our study highlights the potential adverse effects of antibiotics and the positive influence of L. brevis on therapy. Based on our findings, it is crucial to develop an effective strategy for maintaining a healthy gut microenvironment, particularly before and during ICI treatment, even when antibiotics are used. For instance, supplementation with Bifidobacterium and L. brevis prior to antibiotic treatment may mitigate the decline in indigenous Bifidobacterium and L. brevis populations and may enhance the effects of ICI treatment. It is essential to validate the efficacy of Bifidobacterium and L. brevis in large-scale studies, including controlled clinical trials.

Author contributions

Y.S., H.N., and M.K. proposed the study concept and design. Y.Y. and M.M. contributed to the acquisition of patients and data. H.K., T.S., Y.O., T.M., and M.A. conducted microbiota analyses. H.K. performed the data analysis, interpreted the data, and wrote the original draft. H. K. and Y. S. reviewed the manuscript. All the authors have read and agreed to the published version of this manuscript.

Data availability

The fecal 16S rRNA gene amplicon sequencing datasets are available in the Sequence Read Archive of the National Center for Biotechnology Information under accession number PRJNA1203384, https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1203384. The author declares that all other datasets generated during and/or analyzed during the current study are available in this article and its Supplementary file, or from the corresponding author on reasonable request.

Declarations

Competing interests

H. K., T. S., Y. O., T. M., M. A., H. N., Y. S., and M. K. are full-time employees of Yakult Honsha Co., Ltd., Tokyo, Japan. Y. Y. and M. M. received lecture fees from ONO PHARMACEUTICAL Co., Ltd. and MSD K.K

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-08049-4.