Differences in tumor microenvironments between primary lung tumors and brain metastases in lung cancer patients: therapeutic implications for immune checkpoint inhibitors

Two independent cohorts of lung cancer patients were retrospectively analyzed to investigate the differences between primary tumors and BM in two respects: cohort 1 for comparing extra/intracranial responses to anti–PD-1 antibody; cohort 2 for comparing differences in the immunological TME. Cohort 1 included advanced lung cancer patients who were treated with intravenous nivolumab or pembrolizumab as part of an expanded access program (EAP), Keynote 010 trial (NCT01905657), or in routine practice from February 2014 to November 2016 at Seoul National University Hospital (SNUH) and Seoul National University Bundang Hospital. Treatment continued until disease progression, unacceptable toxicity that precluded continuing drug treatment, or death. Patients were allowed to continue treatment despite disease progression if they were deriving a clinical benefit according to an investigator’s assessment. Patients were eligible for treatment if they had at least one BM, with the longest diameter being ≥5 mm, before or during treatment with anti–PD-1 antibody. Patients who received local treatment for previously known BM were not excluded. Cohort 2 included lung cancer patients who underwent both brain metastasectomies for their BM, and primary lung surgery. Therefore, patients who were available for paired lung cancer and BM surgical specimens were enrolled. The Tissue Registry at SNUH was searched between June 2011 and November 2016. Two expert pathologists (S.H.K. and Y.K.J) reviewed tissue sections for adequacy. This study was approved by the SNUH Institutional Review Board (IRB approval number: H-1702-158-836) and was conducted in accordance with Declaration of Helsinki provisions.

A systemic response to anti–PD-1 antibody was measured by standard Response Evaluation Criteria in Solid Tumors (RECIST; version 1.1). An intracranial response was assessed by brain gadolinium-enhanced magnetic resonance imaging (MRI), using RECIST modified to allow target central nerve system lesions, 5 mm or larger in maximum diameter, and with up to five BMs permitted (modified RECIST) [24]. Representative MRI images for response evaluations are available in Additional file 1: Figure S1.

Formalin-fixed paraffin-embedded specimens were examined by immuno-staining tumor cells and TILs. Immunohistochemistry (IHC) was performed using the following antibodies: rabbit anti–PD-L1 (E1L3N) XP® mAb (Cell Signaling Technology, Danvers, MA, USA), mouse anti-CD3 mAb (DAKO, Santa Clara, CA, USA), mouse anti-CD4 mAb (Cell Marque, Rocklin, CA, USA), rabbit anti-CD8 mAb (Thermo Scientific Fischer, Rockford, IL, USA), mouse anti–PD-1 mAb (Cell Marque, Rocklin, CA, USA) and mouse anti-FOXP3 mAb (Abcam, Cambridge, UK). IHC for TILs, except for CD8⁺ TILs, was performed using a Benchmark XT autostainer (Ventana Medical Systems, Basel, Switzerland). In the case of CD8⁺ IHC, a Bond-Max automated immunostainer (Leica Microsystems, Melbourne, Australia) was used. PD-L1 IHC was evaluated based on the intensity and proportion of membranous staining, with or without cytoplasmic staining, in tumor cells and was scored as follows: 0+ (no appreciable staining above background), 1+ (weak membranous staining and/or cytoplasmic staining), 2+ (moderate membranous staining and/or cytoplasmic staining), and 3+ (strong membranous staining and/or cytoplasmic staining) [25]. Each score was multiplied by the percentage of cells (0–100%) to achieve a H-score; H-score = (% of cells 3+)×3 + (% of cells 2+)×2 + (% of cells 1+). All slides were blinded with respect to clinical characteristics and outcomes, and were reviewed and scored by two experienced pathologists (S.H.K. and, Y.K.J.).

To determine the amount of CD3⁺, CD4⁺, CD8⁺, PD-1⁺ and FOXP3⁺ TILs, representative slides were immunostained for CD3, CD4, CD8, PD-1 and FOXP3 and scanned for virtual microscopy using an Aperio ScanScope (Aperio Technologies, Vista, CA, USA). The number of TILs was automatically counted in intact tumor areas lacking necrosis using modified nuclear IHC algorithms in Aperio ImageScope software, as previously described [25]. The mean value of TILs per unit area (mm²) was calculated.

Extra- and intracranial responses to anti–PD-1 antibody were descriptively analyzed. Differences in IHC markers between primary and BM were assessed using the paired Wilcoxon rank-sum test. Correlations between IHC markers were evaluated by Spearman’s rank correlation analysis. Survival analyses were performed for cohort 2 using the Kaplan–Meier method and were compared using a log-rank test. Overall survival (OS) was measured from the date of brain metastasectomy until either death due to any cause or the last follow-up date. A Cox proportional hazard regression model was applied to determine the hazard ratio (HR) for specific variables with respect to OS. For all statistical analyses, two-sided P-values < 0.05 were considered statistically significant. All statistical analyses were carried out using R version 3.3.2 (http://​www.​r-project.​org).

The clinicodemographic characteristics of patients in the two study cohorts are summarized in Table 1. Cohort 1 included 18 lung cancer patients who had BM before or during anti–PD-1 treatment. Most patients had an adenocarcinoma histology (66.7%), and stages IIIA–IV disease (83.3%). Three patients (16.7%) received cerebral irradiation by WBRT during treatment with an ICI. Cohort 2 included 20 lung cancer patients who underwent brain metastasectomy, and therefore paired primary and BM specimens were available. Seven pairs (35.0%) were obtained from cases with synchronous disease at their initial diagnosis, and 13 pairs (65.0%) were obtained from patients with metachronous disease who underwent metastasectomy 5 or more months after the initial diagnosis. The median interval between acquisitions of paired lesions was 17 months (range 0–81 months). No patient had been exposed to anti–PD-1 antibody in cohort 2, and five patients (25.0%) underwent cerebral irradiation before the acquisition of specimens.

Of 18 patients in cohort 1, the primary lung lesion had progressed in seven patients (38.9%). All such patients showed progressive disease in intracranial lesions (Fig. 1). The primary lesion showed a stable disease (n = 9, 50.0%) and partial response (n = 2, 11.1%) in 11 patients, with eight of these exhibiting progressive disease in BM. Although two patients showed a partial response in brain tumors, they were found to have received WBRT with nivolumab. Collectively, these results implied that BM was poorly responsive to anti–PD-1 antibody compared with the primary lung lesion.

On the hypothesis that the distinctive immunological TME of the brain determines the response to the anti–PD-1 antibody, the amounts of TILs and PD-L1 expression on tumor cells in the brain TME were compared with those of primary lung cancer specimens (Fig. 2). Only twelve (60.0%) out of 20 patients in cohort 2 were eligible for a paired comparison because lung biopsy specimens obtained from the other eight patients were too small to evaluate the TME; Three of them had synchronous BM, while the others had metachronous BM. The median interval between acquisitions of paired lesions in patients metachronous BM was 23.4 months (range 4.7–81.5 months). The eight patients were not excluded from cohort 2 because they were eligible for survival analysis. The amounts of CD3⁺, CD4⁺, and CD8⁺ TILs, and the PD-L1 H-score of tumor cells did not show statistically significant differences between primary lung cancers and BM tissues, with P values of 0.148, 0.123, 0.622 and 0.675, respectively. On the other hand, FOXP3⁺ TILs showed a decreasing tendency in the intracranial lesion (P = 0.077), and, more interestingly, PD-1⁺ TILs decreased in the brain in a statistically significant manner (P = 0.034). For fifteen patients with adenocarcinoma (75.0%), more distinctive patterns were observed in CD3⁺ (P = 0.078), CD8⁺ (P = 0.055), FOXP⁺ (P = 0.016), and PD1⁺ (P = 0.016) TILs (Additional file 2: Figure S2). Figure 3 shows representative images of a male patient (ID 17 of cohort 2) with adenocarcinoma histology. In the BM specimen of the patient, PD-1⁺ and CD8⁺ TILs were dramatically decreased, while PD-L1 expression on tumor cells was relatively unchanged. The infiltration of CD3⁺ TILs correlated positively with the PD-L1 H-score of tumor cells in the primary lung cancer (Spearman ρ = 0.545, P = 0.083; Additional file 3: Figure S3A), and BM (Spearman ρ = 0.444, P = 0.049; Additional file 3: Figure S3B). No correlation between PD1+ TILs and PD-L1 expression was noted in the primary lung cancer Spearman ρ = 0.116, P = 0.720; Additional file 3: Figure S3C). In contrast, the infiltration of PD1+ TILs positively correlated with PD-L1 expression of tumor cells in BM (Spearman ρ = 0.550, P = 0.012; Additional file 3: Figure S3D). For patients with metachronous disease, there was no significant correlation between PD-1⁺ TIL infiltration on brain TME and the interval of development of BM. (Spearman ρ = 0.133, P = 0.744; Additional file 4: Figure S4). Altogether, these results revealed that the brain TME differed from that of the primary cancer in terms of an immunological environment, with a markedly decreased amount of PD-1⁺ TILs. The infiltration of CD3⁺ and CD8⁺ TILs was also decreased, especially for lung adenocarcinoma.

To investigate the prognostic significance of immunological TME of brain metastasis, the OS rate of cohort 2 after brain surgery was studied in terms of the amount of CD3⁺ and PD1⁺ TILs, and PD-L1 expression on tumor cells. Since there were no clear-cut off points for these values, the median of each variable was arbitrarily used as a cut-off point. All patients in cohort 2 did not exhibit any survival differences (data not shown). For adenocarcinoma patients in cohort 2, however, a higher infiltration of CD3⁺ TILs tended was associated with a worse prognosis (HR 2.47, 95% CI 0.77–7.96; P = 0.130, log-rank P = 0.119; Fig. 4a) Interestingly, patients showing a higher infiltration of PD1⁺ TILs exhibited a dismal prognosis that was statistically significant (HR, 1.20, 95% CI, 1.01–11.04; P = 0.049; log-rank P = 0.039; Fig. 4b). In terms of PD-L1 expression on tumor cells, there was no prognostic implication according to the PD-L1 H-score (HR 1.24, 95% CI 0.39–3.91; P = 0.711, log-rank P = 0.710; Fig. 4c). Collectively, this analysis showed that the immunological TME in the brain had a prognostic implication after brain metastasectomy, especially for patients with adenocarcinoma histology.