Potential Immune Microenvironment Biomarkers in SCLC: J-TAIL-2 Observational Study.

Introduction
SCLC accounts for up to 15% of all lung cancers. It is often detected at an advanced stage and is characterized by poor prognosis.1,2 Approximately 60% to 70% of patients with SCLC have extensive-stage SCLC (ES-SCLC), for which the historical standard of care has been the combination of cisplatin or carboplatin and etoposide chemotherapy.2
Results of the phase 3 IMpower133 (NCT02763579) and CASPIAN (NCT03043872) studies established the combination of anti-programmed death-ligand 1 (PD-L1) antibodies with platinum-based chemotherapy as the new standard of care for ES-SCLC, with significant improvement in survival rates compared with chemotherapy alone.3, 4, 5, 6 Based on these results, the combinations of atezolizumab plus carboplatin and etoposide (CE) and durvalumab plus cisplatin or carboplatin with etoposide were approved as first-line treatment regimens for ES-SCLC.7
However, effective predictors of response remain limited in ES-SCLC, with an urgent need for the identification of predictive factors of response to anti–PD-L1 and CE combination therapy. Patients with various cancers have derived benefit from anti–PD-(L)1 therapy, which inhibits the programmed cell death protein 1 (PD-1) or PD-L1 pathway and exerts antitumor effects through T-cell activation.8, 9, 10 Therefore, intratumoral infiltration of CD8+ T cells (CD8+ tumor-infiltrating lymphocytes [TILs]) was found to be a predictor of response to anti–PD-1 or anti–PD-L1 therapy in NSCLC and melanoma.11, 12, 13
Previous studies have also identified four subtypes of SCLC defined by the RNA expression of the transcriptional regulators achaete-scute homolog 1 (ASCL1), neuronal differentiation 1 (NEUROD1), POU class 2 homeobox 3 (POU2F3), and yes-associated protein 1 (YAP1). These subtypes were SCLC-A (ASCL1 dominant), SCLC-N (NEUROD1 dominant), SCLC-P (ASCL1/NEUROD1 double-negative with POU2F3 expression), and SCLC-Y (ASCL1/NEUROD1 double-negative with YAP1 expression).14 Analysis of these markers using immunohistochemistry (IHC) in SCLC samples indicated that YAP1 expression was low and not exclusive of other subtypes. Therefore, based on IHC results, the fourth subtype was reclassified as SCLC-O (ASCL1/NEUROD1 double negative not otherwise specified).15 IHC results further indicated that the distribution of these subtypes in SCLC samples was 69% SCLC-A, 17% SCLC-N, 7% SCLC-P, and 7% SCLC-O.15
A recent study used non-negative matrix factorization to analyze publicly available RNA sequencing data sets to identify additional transcriptional subtypes of SCLC that can predict responses to chemotherapies or immunotherapies.16 The identified subtypes were SCLC-A (ASCL1 dominant), SCLC-N (NEUROD1 dominant), SCLC-P (POU2F3-dominant), and SCLC-I (inflamed phenotype, with high expression of immune checkpoint and human leukocyte antigen genes, and interferon-γ activation).16 When this classification was applied to the IMpower133 data set in an exploratory analysis, the trend toward overall survival (OS) benefit of atezolizumab plus CE therapy was maintained across all four subtypes, with a numerically higher magnitude of benefit in the SCLC-I subtype compared with other subtypes.16
The J-TAIL-2 study is a multicenter prospective observational study evaluating atezolizumab combination therapy in patients with ES-SCLC or NSCLC in Japan.17,18 The availability of this large J-TAIL-2 data set of patients with ES-SCLC treated with a combination of immunotherapy and chemotherapy may allow for the improved identification of markers of survival benefit with these treatments. This exploratory analysis of the J-TAIL-2 study aims to identify biomarkers in ES-SCLC tissue samples that are associated with survival benefit and may enable appropriate patient selection for atezolizumab combination therapy.

Materials and Methods

Study Design and Patient Population
The prospective observational J-TAIL-2 study enrolled two cohorts of patients (those with NSCLC and those with ES-SCLC) across 150 institutions in Japan.17,18 In the ES-SCLC cohort, eligible patients were above or equal to 20 years old and scheduled to begin atezolizumab plus CE combination therapy to treat ES-SCLC in clinical practice. Each drug was administered according to its local label (dosage and schedule).
J-TAIL-2 was conducted in accordance with the Declaration of Helsinki, the Act on the Protection of Personal Information, and the Ethical Guidelines for Medical and Health Research Involving Human Subjects. Enrolled patients received a full explanation of the clinical research and provided written consent to participate. The Ethics Review Committee of each study site approved the study protocol and informed consent form before the site could take part in the study. Tissue samples from a subset of patients (from 96 participating institutions) in the ES-SCLC cohort of J-TAIL-2 were used for this exploratory analysis after consent was obtained from patients.

Assessment of Biomarker Expression and CD8+ TILs
The expression of ASCL1, NEUROD1, POU2F3, YAP1, delta-like ligand 3 (DLL3), tumor protein 53 (TP53), and retinoblastoma 1 (Rb1) was assessed using IHC in tissue samples from the J-TAIL-2 ES-SCLC cohort.
ASCL1, NEUROD1, POU2F3, and YAP1 expression were evaluated as positive if any (>0%) tumor cells demonstrated moderate to strong nuclear staining based on morphologic examination by two board-certified pathologists. DLL3 expression was evaluated based on the percentage of positive cells and categorized as negative (<1% of cells), moderate (1%–49%), or high (≥50%). TP53 expression status was classified as wild type, null, or abnormal. Rb1 expression was classified as Rb1 retained or Rb1 inactivated.
SCLC subtypes were defined as follows: SCLC-A (ASCL1 dominant), SCLC-N (NEUROD1 dominant), SCLC-P (ASCL1/NEUROD1 double-negative with POU2F3 expression), and SCLC-O (ASCL1/NEUROD1 double-negative not otherwise specified). The tumors co-expressing ASCL1 and NEUROD1 were classified into either SCLC-A or SCLC-N based on the relative predominance of positive staining. The intratumoral density of CD8+ TILs was analyzed by measuring lymphocyte counts in the tumor (cells/mm2). Two hotspot regions with the highest and second highest density of CD8+ cells within a tumor were identified by pathologists using digital images scanned at ×40 magnification (NanoZoomer 2.0-HT, Hamamatsu Photonics). CD8+ TIL density was calculated as the average number of positive cells in each hotspot (per 0.1 mm2, Supplementary Fig. 1). In cases of disagreement, the slides were reviewed using a multihead microscope to reach consensus.
IHC was performed using antibodies against ASCL1 (24B72D11.1, 1:200; BD Biosciences, Franklin Lakes, NJ), NeuroD1 (EPR17084, 1:400; Abcam, Cambridge, United Kingdom), POU2F3 (6D1, 1:200; Santa Cruz Biotechnology, Dallas, TX), YAP1 (D8H1X, 1:400; Cell Signaling), DLL3 (78110, 1:100; Cell Signaling Technology, Danvers, MA), CD8 (4B11, 1:50; Leica Biosystems, Nussloch, Germany), p53 (DO-7, 1:100; Dako), and RB1 (G3-245, 1:400; BD Biosciences). Heat-induced epitope retrieval was performed, and the EnVision system (Agilent Dako, Santa Clara, CA) was used for detection. Hematoxylin and eosin counterstaining was also performed.

Statistical Analysis
Median (m) OS and progression-free survival (PFS) were estimated using Kaplan-Meier methodology, and 95% confidence intervals (CIs) for medians were calculated using the Brookmeyer-Crowley method. Categorical data and differences in clinical data between the two groups were analyzed using the Fisher’s exact test. Numerical variables were compared using the Fisher’s test to detect statistically significant differences between the two groups. The Kruskal-Wallis test was used for group comparisons.
Receiver operating characteristic (ROC) curve analysis was used to investigate whether the density of tumoral CD8+ TILs was capable of distinguishing between a tumor response and nonresponse. Responders were defined as patients with a PFS of 6 months or longer after chemoimmunotherapy, whereas nonresponders were defined as those with a PFS of less than 6 months. ROC curve analysis was used to investigate the cutoff value for CD8+ TIL density. A density of 54 cells/mm2 or greater was defined as CD8+ TIL high, and a density of less than 54 cells/mm2 was defined as CD8+ TIL low.
The Cox proportional hazards model was used for univariate and multivariate analyses to identify mPFS. The association of OS and PFS with SCLC subtype and CD8+ TIL density was evaluated in all patients assessable for the respective biomarker who received atezolizumab combination therapy in the first line.
In this exploratory analysis, the significance level was set at 5%, with no adjustment for multiplicity. All analyses were performed using R version 4.0.5 (The R Foundation for Statistical Computing, Vienna, Austria; www.r-project.org) and GraphPad Prism 8 (GraphPad Software, Inc., Boston, MA).

Results

Biomarker Expression, CD8+ TIL Density, and SCLC Subtypes
From April 7, 2021, to February 3, 2022 (data cutoff: February 3, 2023), 403 patients were enrolled in the J-TAIL-2 SCLC cohort and 111 tissue samples were collected for this exploratory substudy. Of these, 11 were excluded due to inadequate quality or content (six crushed, three not SCLC, two not submitted), resulting in 100 assessable samples (Supplementary Fig. 2). The samples of 100 patients were evaluated for biomarker expression. The association between SCLC markers is found in Supplementary Table 1. Of the 10 cases of POU2F3-positive SCLC, two were positive for ASCL1 and one was positive for NEUROD1. Of the seven cases of YAP1-positive SCLC, four were positive for ASCL1 and two were positive for NEUROD1. ASCL1 positivity was found in 82% of samples, NEUROD1 positivity in 55%, POU2F3 positivity in 10%, and YAP1 positivity in 7% (Fig. 1A–D). Based on these expression data, SCLC samples were classified as SCLC-A (73%), SCLC-N (16%), SCLC-P (8%), and SCLC-O (3%; Fig. 1E). Analysis of baseline characteristics by SCLC subtype revealed that liver metastases were more common in the SCLC-P subgroup (Supplementary Table 2). All patients (100%) with the SCLC-O subtype had high CD8+ TIL density, followed by 86% of patients with SCLC-P, 56% with SCLC-A, and 50% with SCLC-N. CD8+ TIL infiltration was higher in ASCL1- and DLL3-negative versus -positive tumors and in POU2F3-positive versus -negative tumors (p < 0.05; Fig. 1F). Representative images of ASCL1, NEUROD1, POU2F3, YAP1, and hematoxylin and eosin staining by IHC are found in Figure 1G.
Among the 100 patients whose tumor samples were analyzed, four were excluded from the survival analysis because they received atezolizumab plus CE as second- or later-line therapy. The samples of 96 patients who received atezolizumab plus CE in the first line were assessed for CD8+ TIL density (Supplementary Fig. 1; Fig. 2A). A density of 54 cells/mm2 or greater was defined as CD8+ TIL high, and a density of less than 54 cells/mm2 was defined as CD8+ TIL low (Fig. 2B). Among patients who received atezolizumab combination treatment in the first line, 56 patients were classified into the CD8+ TIL-high subgroup and 40 patients into the CD8+ TIL-low subgroup. Baseline characteristics in the CD8+ TIL-high and CD8+ TIL-low subgroups are found in Table 1. The CD8+ TIL-high subgroup had more patients aged 75 years or older (37% versus 12%, p = 0.01) and a lower proportion of patients with liver metastases (16% versus 38%, p = 0.03).
CD8+ TIL density was associated with PFS (p = 0.03, Fig. 2C), and CD8+ TIL density was higher in patients with PFS of 6 months or longer (median ± SD: 100 ± 260 cells/mm2) versus patients with PFS of 0 to less than 6 months (median ± SD: 36 ± 163 cells/mm2) (p = 0.004, Fig. 2C and D). Cox regression analysis revealed that CD8+ TIL densities of more than 25 cells/mm2 and more than 50 cells/mm2 were associated with improved PFS outcomes (Fig. 2D). In addition, we analyzed treatment efficacy according to CD8+ TIL status. The proportion of patients with PD was lower in the CD8+ TIL-high group (12%) than in the CD8+ TIL-low group (22%, Supplementary Table 3). As of the data cutoff in February 2023, the median follow-up duration was 19.0 months and the proportion of censored cases in the OS analysis was 39.6%.

Association Between SCLC Subtypes, CD8+ TIL Density, and Clinical Outcomes
Among all patients analyzed (n = 96), mPFS was 5.1 months (95% CI: 4.3–6.0) and mOS was 14.2 months (95% CI: 10.6–21.8; Supplementary Fig. 3). mPFS was 6.1 months (95% CI: 4.5–7.5) in the CD8+ TIL-high subgroup versus 4.4 months (95% CI: 4.0–5.1) in the CD8+ TIL-low subgroup (p = 0.01; Fig. 2E), and mOS was 18.4 months (95% CI: 11.8–not estimable [NE]) versus 10.8 months (95% CI: 7.7–16.2) (p = 0.04; Fig. 2F).
CD8+ TIL infiltration was highest in patients with SCLC-P (median ± SD: 360 ± 325 cells/mm2) subtype tumors, followed by SCLC-O (median ± SD: 151 ± 89 cells/mm2), SCLC-A (median ± SD: 60 ± 208 cells/mm2), and SCLC-N (median ± SD: 56 ± 110 cells/mm2) (Fig. 3A). mPFS was 5.2 months (95% CI: 4.5–6.0) in the SCLC-A subgroup, 4.1 months (95% CI: 2.9–5.1) in SCLC-N, 6.2 months (95% CI: 1.4–20.3) in SCLC-P, and 7.5 months (95% CI: 0.3–NE) in SCLC-O (p = 0.1; Fig. 3B). mOS was also not significantly different between SCLC subtypes (p = 0.11; Fig. 3C). In SCLC-A, the PFS of patients with high density of CD8+ TILs was significantly longer than those with low density of CD8+ TILs (high versus low: 7.0 mo [95% CI: 4.5–8.1] versus 5.1 mo [95% CI: 4.2–5.3], p = 0.03; Fig. 3D; Supplementary Fig. 4A). The OS also tended to be longer in patients with high density of CD8+ TILs (high versus low: 25.1 mo [95% CI: 13.4–NE] versus 12.1 mo [95% CI: 9.6–20.5], p = 0.05; Fig. 3E; Supplementary Fig. 4B). Conversely, in SCLC-N, there was no significant difference in the PFS and the OS between the high and low densities of CD8+ TILs (PFS: high versus low: 4.3 mo [95% CI: 2.4–6.3] versus 3.8 mo [95% CI: 0.7–5.1], p = 0.26; OS: high versus low: 8.3 mo [95% CI: 2.4–NE] versus 8.8 mo [95% CI: 0.7–NE], p = 0.76; Fig. 3D and E; Supplementary Fig. 4C and D). However, the limited sample size of this subgroup (n = 16) may have reduced the statistical power to detect meaningful differences.

Association Between Neuroendocrine Marker Expression, CD8+ TIL Density, and Clinical Outcomes
Among the SCLC tissue samples analyzed (n = 100), 48% were ASCL1 and NEUROD1 double positive, 34% were positive for ASCL1 alone, and 7% were positive for NEUROD1 alone. The remaining 11% were ASCL1/NEUROD1 double negative, and thus non-neuroendocrine (Supplementary Fig. 5A). CD8+ TIL density was highest in ASCL1 and NEUROD1 double-negative tumors (median ± SD: 230 ± 288 cells/mm2) and lowest in NEUROD1-positive–only tumors (median ± SD: 45 ± 119 cells/mm2) (Supplementary Fig. 5B). Analysis of PFS and OS by neuroendocrine marker expression revealed that mPFS and mOS were numerically shorter in the NEUROD1-positive group compared with the non-neuroendocrine (ASCL1/NEUROD1-double negative), ASCL1/NEUROD1-double positive, and ASCL1-positive groups (Supplementary Fig. 5C and D).

Rb1, TP53, and DLL3 Status and Association With Clinical Outcomes
Representative images of Rb1 staining by IHC are found in Figure 4A. Rb1 was inactivated in 88% of tumors, and wild-type Rb1 was retained in 12% of tumors (Fig. 4B). mPFS and mOS were numerically longer in patients with Rb1-inactivated tumors than in those who retained wild-type Rb1 (inactivated [n = 83] versus retained [n = 12]; PFS: 5.2 mo [95% CI: 4.5–6.0] versus 3.4 mo [95% CI: 0.7–7.1], p = 0.08, OS: 16.1 mo [95% CI: 10.8–23.6] versus 6.5 mo [95% CI: 11.5–NE], p = 0.01, Fig. 4C and D). Furthermore, among patients with Rb1 inactivation (n = 83), those with high density of CD8+ TIL had significantly longer PFS compared with those with low density of CD8+ TIL (high versus low: 6.3 mo [95% CI: 4.6–7.6] versus 4.6 mo [95% CI: 4.1-5.2], p = 0.02). In contrast, among those with retained Rb1 (n = 12), no significant difference in PFS was observed between high and low CD8+ TIL groups (high versus low: 2.7 mo [95% CI: 0.3–NE] versus 3.7 mo [95% CI: 1.4–NE], p = 0.67, Supplementary Table 4).
PFS and OS did not differ significantly based on TP53 expression status (abnormal [n = 62], versus null [n = 26], versus wild type [n = 7], PFS: 4.8 mo [95% CI 4.2–6.4] versus 5.2 mo [95% CI: 3.5–6.0] versus 5.1 mo [95% CI: 4.3–6.8], OS: 15.3 mo [95% CI: 9.9–23.6] versus 14.1 mo [95% CI: 5.9–NE] versus 13.4 mo [95% CI: 7.9–NE], Supplementary Fig. 6). In addition, in the TP53-null subgroup, high CD8+ TIL density was associated with significantly prolonged PFS compared with low TIL density (high versus low: 6.2 mo [95% CI: 2.8–10.4] versus 4.7 mo [95% CI: 1.8–5.3], p = 0.01), whereas no significant difference was observed in the abnormal or wild-type subgroups (high versus low, abnormal: 6.2 mo [95% CI: 4.2–7.6] versus 4.2 mo [95% CI: 3.5–5.1], p = 0.18, wild type: 6.0 mo [95% CI: 4.5–NE] versus 5.1 mo [95% CI: 4.3–NE], Supplementary Table 4).
PFS was similar between the DLL3-negative (n = 29), DLL3-moderate (n = 32), and DLL3-high groups (n = 34), but OS was numerically longer in the DLL3-moderate group (DLL3-negative versus DLL3-moderate versus DLL3-high: PFS: 4.8 mo [95% CI: 3.5–6.0] versus 5.3 mo [95% CI: 4.3–7.2] versus 5.1 mo [95% CI: 3.5–6.2], OS: 13.4 mo [95% CI: 7.2–NE] versus 20.5 mo [95% CI: 9.9–NE], versus 11.8 mo [95% CI: 5.9–15.2], Supplementary Fig. 7). In addition, in DLL3-negative and DLL3-high tumors, patients with high density of CD8+ TIL had significantly longer PFS than those with low density of CD8+ TIL (high versus low: DLL3-negative: 6.0 mo [95% CI: 3.9–7.9] versus 3.4 mo [95% CI: 0.9–5.1], p = 0.003, DLL3-high: 6.5 mo [95% CI: 2.9–8.1] versus 4.3 mo [95% CI: 1.8–5.1], p = 0.01), whereas no significant difference was observed in DLL3-moderate tumors (high versus low: 6.3 mo [95% CI: 2.8–11.5] versus 5.3 mo [95% CI: 4.2–6.8], p = 0.52, Supplementary Table 4).

Analysis of Predictors of PFS and OS Benefit
Univariate analysis of the association of baseline factors with PFS benefit identified several characteristics as the independent predictor of PFS. Higher Eastern Cooperative Oncology Group performance status (ECOG PS) (HR 1.78, 95% CI: 1.04–3.05, p = 0.03), a higher number of metastatic organ sites (HR 1.33, 95% CI: 1.14–1.54, p < 0.001), and presence of liver metastases (HR 2.17, 95% CI: 1.32–3.56, p = 0.002) were associated with poorer survival, but a lower number of chemotherapy cycles (HR 0.24, 95% CI: 0.14–0.39, p < 0.001) and high CD8+ TIL density (HR 0.57, 0.37–0.88, p = 0.01) were associated with improved PFS (Supplementary Table 5). In a multivariate analysis, a higher number of metastatic organ sites (HR 1.79, 95% CI: 1.05–3.05, p = 0.03) were also associated with worse PFS, whereas fewer chemotherapy cycles (HR 0.12, 95% CI: 0.06–0.24, p < 0.001) and high CD8+ TIL density (HR 0.48, 95% CI: 0.28–0.81, p = 0.01) were associated with improved PFS.
In a univariate analysis of OS, higher ECOG PS (HR 2.15, 95% CI: 1.15–4.01, p = 0.02), a higher number of metastatic organ sites (HR 2.01, 95% CI: 1.19–3.40, p = 0.01), and the presence of liver metastases (HR 2.65, 95% CI: 1.52–4.62, p < 0.001) were associated with worse OS, whereas fewer chemotherapy cycles (HR 0.19, 95% CI: 0.11–0.34, p < 0.001) and higher CD8+ TIL density (HR 0.58, 95% CI: 0.35–0.98, p = 0.04) were associated with better OS. In the multivariate model, higher ECOG PS (HR 2.75, 95% CI: 1.33–5.69, p = 0.01), a higher number of metastatic sites (HR 2.41, 95% CI: 1.27–4.58, p = 0.01), and the presence of liver metastases (HR 2.46, 95% CI: 1.19–5.07, p = 0.01) were associated with worse OS, whereas fewer chemotherapy cycles (HR 0.09, 95% CI: 0.04–0.19, p < 0.001) were identified as an independent predictor of improved OS. CD8+ TIL density did not retain independent significance in this model (Supplementary Table 6).
Patient-level analysis of baseline characteristics and clinical features among the 13 patients with long-term survival benefit (PFS ≥ 12 mo) in this analysis revealed that 11 of 13 (85%) long-term responders had tumors of SCLC-A subtype and 10 of 13 (77%) had high CD8+ TIL density in their tumors (Supplementary Table 7). Of the 13 long-term responders, five were aged 75 years or older. Notably, three of the four patients with mPFS of 2 or more years were aged 75 years or older. Notably, none of the long-term responders had tumors of SCLC-N subtype and only one tumor with liver metastasis.

Discussion
This exploratory analysis of the observational J-TAIL-2 study evaluated the association of CD8+ TIL infiltration and SCLC subtypes with clinical outcomes in patients with SCLC. Overall, high CD8+ TIL infiltration was associated with improved PFS and OS outcomes, and poor clinical outcomes were observed in patients with SCLC-N subtype tumors.
In this analysis, IHC staining allowed subtype classification of SCLC tumors as previously reported into SCLC-A (73%), SCLC-N (16%), SCLC-P (8%), and SCLC-O (3%) subtypes. The distribution of these subtypes was similar to that reported previously based on IHC (39%–78% SCLC-A, 6%–32% SCLC-N, 7%–28% SCLC-P, and 1%–10% SCLC-O).15,19, 20, 21, 22 SCLC tumors were also classified by neuroendocrine status (ASCL1 and NEUROD1 double positive or double negative). Low infiltration of CD8+ TILs was found in tumors of the neuroendocrine subtype and high infiltration of CD8+ TILs in the non-neuroendocrine subtype.
Results of this analysis indicated that OS and PFS with atezolizumab plus CE combination therapy were significantly longer in patients with high CD8+ TIL density than in patients with low CD8+ TIL density. In particular, in SCLC-A subtype tumors, PFS and OS were longer in the CD8+ TIL-high versus CD8+ TIL-low subgroups. These data indicate that the density of CD8+ TILs may be a potential predictor of survival benefit with atezolizumab combination therapy in patients with SCLC, especially in those with the SCLC-A subtype. Tumors with high infiltration of CD8+ T cells have an active immune microenvironment, are immunogenic, and benefit from immunotherapy in a variety of cancers.11,12 Another exploratory analysis applied non-negative matrix factorization to tumor samples from IMpower133 and further refined previously described SCLC subtypes using the larger data set. In this analysis, the SCLC subtypes were defined as SCLC-A (ASCL1 dominant), SCLC-N (NEUROD1 dominant), SCLC-I-NE (SCLC-I and SCLC-A enriched, neuroendocrine inflamed), and SCLC-I-non-NE (SCLC-P and SCLC-I enriched, non-neuroendocrine inflamed).23 The SCLC-I-NE subgroup had the best clinical outcomes, with near doubling of mOS with atezolizumab plus CE compared with placebo plus CE.23
In this study, patients with low CD8+ TIL infiltration had a significantly higher rate of liver metastases, and the presence of liver metastases was the only independent negative predictor of survival benefit with immunochemotherapy. Liver metastases are resistant to immunotherapy in a variety of cancers, including SCLC.24, 25, 26 Previous studies have revealed that liver tumors can induce the loss of systemic tumor-specific effector T cells by activation of regulatory T cells and/or hepatic macrophages, leading to a reduction in the number of CD8+ T cells in intrahepatic tumors, and lowered expression of PD-1 in these cells.27 This is a possible mechanism for the impaired response found in patients with liver metastases in this analysis, regardless of CD8+ TIL infiltration status.
Conversely, in the SCLC-N group, PFS and OS did not differ between the CD8+ TIL-high and CD8+ TIL-low subgroups. PFS and OS were also numerically shorter in the SCLC-N group versus other SCLC subtypes, indicating that SCLC-N tumors have poor response to atezolizumab combination therapy, regardless of tumor infiltration status. Recent transcriptomic analyses have demonstrated increased infiltration of FOXP3+ regulatory T cells in the SCLC-N subtype, supporting the notion of an immunosuppressive microenvironment in this subtype.28,29 Regulatory T cells have been found to contribute to immunotherapy resistance by suppressing effector T cells, such as CD8+ T cells, which play an important role in cancer cell death in the host.30, 31, 32 Therefore, in the SCLC-N tumors, despite the appearance of high CD8+ TIL infiltration by IHC, improvement in clinical outcomes such as PFS may not be observed.
PFS was shorter and OS was significantly shorter in patients who retained wild-type Rb1 expression than in patients with inactivated Rb1 status. CDKN2A mutations and CCND1 amplifications are known to be enriched in SCLC tumors that express wild-type Rb1.33 It has been reported that clinical outcomes with immune checkpoint inhibitor therapy are poor in patients with NSCLC who have CDKN2A or CCND1 gene mutations.34, 35, 36
Limitations of this analysis include its observational and exploratory nature. Results should therefore be interpreted with caution. The possibility of survival bias due to incomplete follow-up and data censoring should be considered when interpreting the OS results. Moreover, our classification of SCLC subtypes was based on IHC protein expression, whereas other studies, such as the IMpower133 trial, used transcriptomic profiling, which may yield different subtype distributions.16 Notably, prior research has revealed limited concordance between mRNA and protein levels, with only 32% of genes having statistically significant correlation, underscoring methodological differences between IHC and transcriptomic approaches.37 As a retrospective observational study, this analysis may also be susceptible to selection bias. Future prospective trials are needed to validate the prognostic utility of the density of CD8+ TIL in randomized controlled trials. Standardized IHC protocols for CD8+ TIL assessment, including hotspot definition and quantification methods, are needed to ensure reproducibility and support clinical application. In addition, the relatively small sample size may have limited the statistical power to detect subtype-specific associations. Tumor specimen heterogeneity could also have affected the accuracy of immunohistochemical evaluation. As this was not a randomized trial, it remains challenging to definitively distinguish whether CD8+ TILs serve as predictive markers of response to immunotherapy or as general prognostic indicators of outcome.
In conclusion, the results of this exploratory analysis of the observational J-TAIL-2 study evaluating atezolizumab combination therapy in patients with ES-SCLC in Japan indicate that CD8+ TIL density is a potential marker of survival benefit with anti–PD-1/PD-L1 therapy. CD8+ TIL density is associated with improved outcomes in patients with the SCLC-A tumor subtype, but not those with the SCLC-N subtype.

CRediT Authorship Contribution Statement
Masayuki Shirasawa: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review and editing, Visualization, Funding acquisition.
Makoto Nishio: Investigation, Writing - review and editing, Supervision, Project administration.
Kadoaki Ohashi: Investigation, Writing - review and editing.
Atsushi Osoegawa: Investigation, Writing - review and editing.
Eiki Kikuchi: Investigation, Writing - review and editing.
Hideharu Kimura: Investigation, Writing - review and editing.
Yasushi Goto: Investigation, Writing - review and editing.
Junichi Shimizu: Investigation, Writing - review and editing.
Eisaku Miyauchi: Investigation, Writing - review and editing.
Hiroshige Yoshioka: Investigation, Writing - review and editing.
Ichiro Yoshino: Investigation, Writing - review and editing.
Toshihiro Misumi: Formal analysis, Investigation, Writing - review and editing.
Yasushi Yatabe: Conceptualization, Investigation, Resources, Data curation, Writing - review and editing, Supervision, Funding acquisition.
Tatsuya Yoshida: Conceptualization, Investigation, Data curation, Writing - review and editing.
Jumpei Kashima: Investigation, Data curation, Writing - review and editing.
Masahide Oki: Investigation, Writing - review and editing.
Hisao Ashimura: Writing - review and editing.
Yuki Kobayashi: Writing - review and editing.
Misa Tanaka: Writing - review and editing.
Akihiko Gemma: Investigation, Writing - review and editing, Supervision, Project administration.

Disclosure
Dr. Shirasawa reports receiving honoraria from Asahi Kasei, AstraZeneca K.K., and Merck Sharp & Dohme. Nishio declares receiving lecture fees from AbbVie, AstraZeneca, Boehringer Ingelheim, Bristol Myers Squibb, Chugai Pharmaceutical Co., Ltd., Daiichi Sankyo, Eli Lilly, Janssen Pharmaceutical, Merck, Merck Sharp & Dohme, Ono Pharmaceuticals, Nippon Kayaku, Novartis, Taiho Pharmaceutical, Takeda, and Pfizer. Dr. Ohashi declares receiving funding, grants, and honoraria from Chugai Pharmaceutical Co., Ltd. Osoegawa declares receiving honoraria from AstraZeneca, Bristol Myers Squibb, Chugai Pharmaceutical Co., Ltd., Merck Sharp & Dohme, and Ono Pharmaceutical; and grants from 10.13039/100010795Chugai Pharmaceutical Co., Ltd., and 10.13039/100009954Taiho Pharmaceutical. Dr. Kikuchi declares receiving honoraria from Chugai Pharmaceutical Co., Ltd. Kimura has received honoraria and grants from Chugai Pharmaceutical Co., Ltd. Goto declares receiving grants (to the institution) from AbbVie, Bristol Myers Squibb, Daiichi Sankyo, Eli Lilly, Kyowa Kirin, Novartis, Ono Pharmaceutical, Pfizer, and Preferred Networks; receiving honoraria from Boehringer Ingelheim, Bristol Myers Squibb, Chugai Pharmaceutical Co., Ltd., Eli Lilly, Merck, Merck Sharp & Dohme, Novartis, Ono Pharmaceutical, Pfizer, Taiho Pharmaceutical, and Thermo Fisher Scientific; having advisory board participation for AstraZeneca, Boehringer Ingelheim, Bristol Myers Squibb, Chugai Pharmaceutical Co., Ltd., Daiichi Sankyo, Eli Lilly, Guardant Health Inc., Illumina, Merck Sharp & Dohme, Novartis, Ono Pharmaceutical, Pfizer, and Taiho Pharmaceutical; and having leadership or fiduciary roles with Cancer Net Japan and JAMT. Dr. Shimizu declares receiving speakers bureau fees from Amgen, AstraZeneca, Chugai Pharmaceutical Co., Ltd., Merck Sharp & Dohme, Novartis, Pfizer, Taiho Pharmaceutical, and Takeda. Dr. Miyauchi declares receiving grants from Chugai Pharmaceutical Co., Ltd; receiving honoraria from Amgen, AstraZeneca, Boehringer Ingelheim, Bristol Myers Squibb, Chugai Pharmaceutical Co., Ltd., Daiichi Sankyo, Eli Lilly, Kyowa Kirin, Merck, Merck Sharp & Dohme, Nippon Kayaku, Novartis, Ono Pharmaceutical, Pfizer, Sysmex, Taiho Pharmaceutical, Takeda Pharmaceutical, and Thermo Fisher Scientific; and having advisory board participation for Boehringer Ingelheim, Chugai Pharmaceutical Co., Ltd., Daiichi Sankyo, Eli Lilly, Merck, and Ono Pharmaceutical. Dr. Yoshioka declares receiving funding from Chugai Pharmaceutical Co., Ltd; receiving research funding from 10.13039/100004325AstraZeneca, Boehringer Ingelheim, 10.13039/501100022274Daiichi Sankyo, Delta Fly Pharmaceutical, Janssen Pharmaceutical, Merck Sharp & Dohme, and Novartis; receiving consulting fees from Delta Fly Pharmaceutical; and receiving lecture fees from Amgen, AstraZeneca, Boehringer Ingelheim, Bristol Myers Squibb, Chugai Pharmaceutical Co., Ltd., Daiichi Sankyo, Eli Lilly, Kyowa Kirin, Merck, Merck Sharp & Dohme, Nippon Kayaku, Nipro Pharma, Novartis, Ono Pharmaceutical, Otsuka Pharmaceutical, Pfizer, and Taiho Pharmaceutical. Yoshino declares receiving consulting fees from AstraZeneca, Chugai Pharmaceutical Co., Ltd., Covidien Japan, Johnson and Johnson, and Medicaroid; and honoraria from AstraZeneca, Chugai Pharmaceutical Co., Ltd., Covidien Japan, Daiichi Sankyo, Johnson and Johnson, Merck Sharp & Dohme, and Takeda. Misumi has received payment or honoraria for lectures and speakers bureaus or educational events from Chugai Pharmaceutical Co., Ltd; and has received payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from AstraZeneca and Miyarisan. Dr. Yatabe has received grants from AstraZeneca, Optieum Biotechnologies, and NEC; has received contracted support/research from Merck Biopharma, Chugai Pharmaceutical Co., Ltd., and Konica-Minolta REALM; has received honoraria for lectures from Merck Sharp & Dohme, AstraZeneca, Merck Biopharma, Novartis, Amgen, Daiichi Sankyo, and Thermo Fisher Scientific; has received honoraria for lectures and speakers bureaus from Chugai Pharmaceutical Co., Ltd.; and has advisory board participation for AstraZeneca, Merck Sharp & Dohme, AbbVie, Novartis, Amgen, Daiichi Sankyo, Janssen Pharma, and Konica-Minolta REALM. Dr. Yoshida has received grants or contracts and honoraria from Chugai Pharmaceutical Co., Ltd.; has received grants or contracts from Novartis, AbbVie, Amgen, Daiichi Sankyo, AstraZeneca, Merck Sharp & Dohme, Astellas, Medpace, Boehringer Ingelheim, Bristol Myers Squibb, Ono Pharmaceutical, and Merck; has received payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from AstraZeneca, Bristol Myers Squibb, Eli Lilly, Daiichi Sankyo, Merck, Merck Sharp & Dohme, Novartis, Ono, Pfizer, and Takeda; and has participation on a data safety monitoring board or advisory board for Amgen, Boehringer Ingelheim, Chugai Pharmaceutical Co., Ltd., Merck Sharp & Dohme, Novartis, and Pfizer. Dr. Kashima has nothing to declare. Dr. Oki has received grants (to institution) and honoraria from Chugai Pharmaceutical Co., Ltd. Mr. Ashimura reports having employment by Chugai Pharmaceutical Co. Ltd. Mr. Kobayashi and Mr. Tanaka declare having employment by and stock ownership in Chugai Pharmaceutical Co., Ltd. Gemma declares having study participation as an investigator for the J-TAIL-2 study; has received honoraria for educational lectures from Nihon Kayaku; and has participated on an interstitial lung disease board for Merck Sharp & Dohme, AstraZeneca, Daiichi Sankyo, and Chugai Pharmaceutical Co., Ltd.