From January 2019 to April 2021, 149 patients (119 men and 30 women; median age, 64; range, 38 to 75 years) with pathologically confirmed NSCLC who were scheduled to receive anti-PD-1-based chemo-immunotherapy at Shandong Provincial Hospital (Jinan, China) were screened. Certain patients who relapsed after complete resection were also included in the cohort, and their tumor-node-metastasis (TNM) staging information was determined at the beginning of anti-PD-1-based chemo-immunotherapy. Individuals who were aged 18-75 years, had pathologically confirmed stage III-IV NSCLC (according to the 8th edition of the AJCC/UICC classification for NSCLC) and were ineligible for radical surgery or radiotherapy, had at least one measurable lesion per the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1) (29), had an Eastern Cooperative Oncology Group (ECOG) performance status (PS) (30) no more than 2, and had archival tumor tissues obtained within 6 months before chemo-immunotherapy or fresh tumor samples, were included. Those who had symptomatic central nervous system metastasis, had used immunosuppressants within 2 weeks before chemo-immunotherapy, had received neoadjuvant chemotherapy, radiotherapy or immunotherapy, or had a history of other malignant tumors were excluded. Patients were administered intravenous anti-PD-1 agents, including camrelizumab, sintilimab, tislelizumab, pembrolizumab (at a dose of 200 mg, every 3 weeks) and toripalimab (at a dose of 240 mg, every 3 weeks), combined with chemotherapies such as cisplatin, carboplatin, pemetrexed, gemcitabine and paclitaxel according to the standards of relevant guidelines. Clinical and pathological information, including age, sex, smoking history, histologic type, TNM stage, ECOG PS and lines of therapy, were retrospectively obtained from the medical records.

The present study was approved (approval no. NSFC-2019-05) by the Institutional Review Board of Shandong Provincial Hospital affiliated to Shandong First Medical University (Jinan, China) and complied with all relevant ethical regulations of the Declaration of Helsinki.

For the evaluation of tumor response, CT scans were reviewed by specialized radiologists. Tumor assessment was performed at baseline and every 2 cycles according to the RECIST 1.1. On the basis of the best overall response, patients with complete or partial response were considered responders, while others with stable or progressive disease were considered non-responders. The progression-free survival (PFS) and OS were defined as the interval from treatment initiation to the date of clinical progression or death, and to the date of death from any cause, respectively (31). Survival status was obtained through medical records or telephone follow-up every 2 cycles of chemo-immunotherapy. Patients who had not progressed or succumbed to the disease were censored for PFS and OS at last follow-up (August 31, 2021).

All of the cases had available formalin-fixed paraffin-embedded (FFPE) specimens of primary tumors which were obtained from the most recent biopsy before treatment. Tissue sections (4-µm thick) from each FFPE block were used for FXR, PD-L1 and CD8 IHC staining as previously described (32). Slides were incubated overnight at 4°C with the following primary antibodies: Anti-bile acid receptor NR1H4 antibody (1:100; product code ab187735), anti-PD-L1 antibody (1:500; product code ab205921) and anti-CD8α antibody (1:100; product code ab101500; all from Abcam). Isotype controls were conducted simultaneously using concentration-matched non-specific mouse or rabbit IgG (product codes ab37355 and ab172730, respectively; both from Abcam). All stained slides were scanned using a high-resolution digital slide scanner (TissueFAXS plus; TissueGnostics Ltd.) up to a magnification of ×200.

For FXR and PD-L1, staining intensity was classified as negative (0), weak (1), moderate (2) and intense (3) according to the degree of dyeing. As for the percentage of stained cells, 0% (0), 1 to 25% (1), 26 to 50% (2), 51 to 75% (3) and 76 to 100% (4) was defined. The IHC score was generated by multiplying the staining intensity and percentage (27). If the staining intensity in a section was diverse, the highest was selected from the scores of different intensities and ratios. Since the median IHC score of FXR and PD-L1 was 6 and 4, respectively, an IHC score of 5 was used as the cut-off value to discriminate low and high expression of FXR and PD-L1, to ensure that the number of NSCLC patients with FXR or PD-L1 low-expression was generally equal to the number of NSCLC patients with FXR or PD-L1 high-expression. For CD8, 4-6 independent high-power fields (HPFs; ×200) which represented the densest lymphocytic infiltrates were selected to reflect the extent of CD8⁺ T-cell infiltration. The average CD8⁺ T-cell density (cells/HPF) was calculated as the mean value of the 4-6 areas (33). Two proficient pathologists who were blinded to the clinical data independently evaluated the IHC results and reached a final consensus.

Shapiro-Wilk method was used to assess the normality of the quantitative data. Comparisons between skewed distribution data were performed using Mann-Whitney U test. Categorical variables were compared using chi-square tests or Fisher's exact test. Spearman's rank correlation test was used to assess the correlation between FXR and PD-L1 and between FXR and CD8 in NSCLC. The comparison of infiltrating CD8⁺ T cells in four groups, according to FXR staining intensity, was performed using Kruskal-Wallis (K-W) rank sum test. Survival curves were estimated by Kaplan-Meier analysis and the log-rank test was utilized to examine the differences between groups. In addition, prognostic factors were evaluated based on univariate and multivariate Cox regression analyses. The variables with univariate regression P<0.1 were included in multivariate regression analysis. Statistical analyses were performed using the statistical software SPSS Statistics 26.0 (IBM Corp.) and GraphPad Prism 8.0 (GraphPad Software, Inc.). All tests were two-sided, and P<0.05 was considered to indicate a statistically significant difference.

A total of 149 patients treated with anti-PD-1-based chemo-immunotherapy were eventually enrolled in the present study. Their clinical and pathological characteristics are listed in Table I. The cohort included 119 men and 30 women with a higher proportion of elderly, and most of the patients were smokers. The majority of these NSCLC cases were non-squamous cell carcinoma (93/149, 62.4%), of which 90 were adenocarcinoma, 2 were sarcomatoid carcinoma and 1 was large cell neuroendocrine carcinoma. More than half of the patients (103/149, 69.1%) were in stage IV at the beginning of anti-PD-1-based chemo-immunotherapy. In the present cohort, all the patients received PD-1 inhibitors combined with chemotherapy as the first-line or higher lines with refractory progression after chemotherapy, radiation or targeted therapy. Among them, the most widely used PD-1 inhibitor was camrelizumab (111/149, 74.5%). There were 40 patients who had oncogene mutations among 67 patients with available gene analysis data. A total of 78 patients (52.3%) and 54 patients (36.2%) were classified as high FXR and PD-L1 expression, respectively (Table I).

As visualized using IHC, the expression of FXR was mainly localized in the nucleus and cytoplasm, while PD-L1 was expressed on the membrane of tumor cells (Fig. 1A). A total of 46 patients were classified as responders according to the RECIST 1.1, and the objective response rate (ORR) to chemo-immunotherapy in the present study was 30.9%. It was revealed that responsive tumors expressed higher levels of both FXR and PD-L1 compared with those irresponsive to chemo-immunotherapy (P=0.01 and 0.003, respectively; Fig. 1B and C). Meanwhile, the chi-square test revealed that high FXR and PD-L1 expression levels were associated with a higher ORR in all patients (P=0.036 and 0.002, respectively; Fig. 1D and E). These findings underlined the utility of high expression of FXR as a predictive biomarker for immunotherapy in addition to PD-L1.

As illustrated in Fig. 2A and B, the Kaplan-Meier and log-rank tests demonstrated that responders to anti-PD-1-based chemo-immunotherapy had both significantly longer PFS and OS than non-responders (P<0.001 and P=0.023, respectively). There was a non-significant trend towards improved PFS and OS in patients with high FXR expression as compared with their FXR low counterparts (Fig. 2C and D). In addition, PD-L1 high-expression patients were found to have a significantly longer PFS (P=0.031; Fig. 2E), as compared with PD-L1 low-expression patients; however, there was no statistical association between PD-L1 expression and OS (Fig. 2F).

To study the prognostic role of FXR and PD-L1 in all patients, a Cox regression model was applied including several clinical characteristics (age, sex, smoking history, histologic type, TNM stage, ECOG PS, lines of therapy and gene mutation state). Given that the number of NSCLC patients with an ECOG PS score of 2 was quite small (4/149, 2.7%), the ECOG PS data were analyzed as a binary variable (0 vs. ≥1). The univariate analysis revealed that TNM stage, line of therapy and PD-L1 expression were associated to PFS (P-values <0.1; Table SI). In multivariate analysis, TNM stage [P=0.011; hazard ratio (HR), 2.146; 95% confidence interval (CI), 1.188-3.874)] remained an independent prognostic indicator for PFS. Conversely, only age was defined as an independent prognostic factor for OS (P=0.021; HR, 4.117; 95% CI, 1.235-13.727; Table SII). Collectively, neither FXR nor PD-L1 expression could stratify PFS and OS in the present cohort.

It was previously reported that FXR^highPD-L1^low mouse LLC tumors exhibited an increased susceptibility to PD-1 blockade compared with mock LLC tumors (28). To further investigate whether FXR could be an effective predictor of clinical response to anti-PD-1-based chemo-immunotherapy among PD-L1^low patients with NSCLC, a subgroup analysis of tumor responses and prognosis based on the correlation between FXR and PD-L1 was conducted. Firstly, a significant increase of PD-L1 expression in 'FXR low' tumors was found (P=0.016; Fig. 3A), which was consistent with a previous study (28). The chi-square test revealed that FXR was inversely associated with PD-L1 expression in specimens with NSCLC (P=0.005; Fig. 3B). In addition, the result of Spearman's correlation analysis demonstrated that there was a significant inverse correlation between FXR and PD-L1 expression in the entire cohort (r=−0.236, P=0.004; Fig. 3C). Furthermore, it was investigated whether PD-L1 expression was different between squamous cell carcinoma and adenocarcinoma in the present study. There was no statistically significant difference in PD-L1 expression between squamous cell carcinoma and adenocarcinoma in 149 specimens with NSCLC enrolled in the present study (data not shown).

Then, four subgroups based on the IHC levels of FXR and PD-L1 were defined (Fig. 4A). Notably, patients with high expression of both PD-L1 and FXR exhibited the highest ORR (60%), followed by the FXR^lowPD-L1^high group (38.2%). In these PD-L1 high-expression patients, although responsive tumors expressed higher levels of FXR than that of non-responsive tumors (Fig. S1), no significant association was identified between FXR expression and tumor response (Fig. 4A). Of note, patients with high expression of FXR demonstrated a higher ORR as compared with FXR low-expression patients in the presence of PD-L1 low-expression (31.0 vs. 8.1%, P=0.009). Additionally, PD-L1^low-responsive tumors expressed significantly increased FXR compared with the PD-L1^low-non-responsive tumors (P<0.001; Fig. 4B). Collectively, these results suggested FXR as a promising predictive biomarker for clinical efficacy to anti-PD-1-based chemo-immunotherapy when PD-L1 is low or negative.

The Kaplan-Meier survival curves in Fig. 4C and D revealed that high FXR expression was associated with both longer PFS (P=0.013) and OS (P=0.03) among the PD-L1^low patients with NSCLC. However, consistent with the association with therapeutic responses, the extent of FXR expression in PD-L1^high patients was not associated with either PFS or OS (Fig. 4C and D).

Cox regression models were then applied, including clinical variables to verify the prognostic value of FXR in PD-L1^low patients. As presented in Table II, TNM stage and FXR expression were found to be significantly correlated with PFS in univariate Cox regression analysis. These two variables were then analyzed in a multivariate Cox regression model. Intriguingly, FXR expression was still identified as an independent predictor for PFS in PD-L1^low patients with NSCLC (P=0.038; HR, 0.552; 95% CI, 0.315-0.967).

As for the univariate Cox regression analysis for OS in PD-L1^low patients, age and FXR expression were found to stratify OS significantly at the level of P<0.1 (Table III). Additionally, multivariate analysis showed that FXR expression remained an independent prognostic indicator for OS in PD-L1^low patients with NSCLC (P=0.029; HR, 0.377; 95% CI, 0.157-0.905).

The predictive value of FXR on anti-PD-1-based chemo-immunotherapy in the perspective of TME was then sought to be explained. Since CD8⁺ T cells represent the most crucial tumoricidal effector cells and the main target of the PD-L1/PD-1 checkpoint pathway in the TME (34,35), the infiltration of CD8⁺ T cells in specimens with NSCLC was examined. Representative microphotographs of the infiltration levels of different CD8⁺ T cells are shown in Fig. 5A. The cells positively stained for CD8 were semi-quantified and low or high groups were defined according to the median value. IHC evaluation revealed a statistically significant decrease of infiltrating CD8⁺ T cells in FXR^high NSCLC specimens (P=0.014; Fig. 5B). Chi-square analysis demonstrated that FXR expression was inversely associated with the infiltration of CD8⁺ T cells in specimens with NSCLC (P=0.004; Fig. 5C). Importantly, the Spearman's correlation analysis revealed that there was a significant inverse correlation between FXR expression and infiltrating CD8⁺ T cells in the enrolled 149 specimens with NSCLC (r=−0.217, P=0.008; Fig. 5D). The CD8 expression data were analyzed in four groups, according to FXR staining intensity (negative, weak, moderate and intense). However, the K-W analysis showed that there was no difference in the degree of infiltration of CD8⁺ T cells among the four FXR staining groups in the present study (data not shown). Thus, it was considered that the immunosuppressive effects of FXR previously reported (28) in in vitro co-culture and mouse models could also be recapitulated in clinical patients with NSCLC.