Purpose:

To provide a better understanding of the interplay between the immune system and brain metastases to advance therapeutic options for this life-threatening disease.

Experimental Design:

Tumor-infiltrating lymphocytes (TIL) were quantified by semiautomated whole-slide analysis in brain metastases from 81 lung adenocarcinomas. Multi-color staining enabled phenotyping of TILs (CD3, CD8, and FOXP3) on a single-cell resolution. Molecular determinants of the extent of TILs in brain metastases were analyzed by transcriptomics in a subset of 63 patients. Findings in lung adenocarcinoma brain metastases were related to published multi-omic primary lung adenocarcinoma The Cancer Genome Atlas data (n = 230) and single-cell RNA-sequencing (scRNA-seq) data (n = 52,698).

Results:

TIL numbers within tumor islands was an independent prognostic marker in patients with lung adenocarcinoma brain metastases. Comparative transcriptomics revealed that expression of three surfactant metabolism-related genes (SFTPA1, SFTPB, and NAPSA) was closely associated with TIL numbers. Their expression was not only prognostic in brain metastasis but also in primary lung adenocarcinoma. Correlation with scRNA-seq data revealed that brain metastases with high expression of surfactant genes might originate from tumor cells resembling alveolar type 2 cells. Methylome-based estimation of immune cell fractions in primary lung adenocarcinoma confirmed a positive association between lymphocyte infiltration and surfactant expression. Tumors with a high surfactant expression displayed a transcriptomic profile of an inflammatory microenvironment.

Conclusions:

The expression of surfactant metabolism-related genes (SFTPA1, SFTPB, and NAPSA) defines an inflamed subtype of lung adenocarcinoma brain metastases characterized by high abundance of TILs in close vicinity to tumor cells, a prolonged survival, and a tumor microenvironment which might be more accessible to immunotherapeutic approaches.

Translational Relevance

Immunotherapies have become a powerful addition to the established therapies in metastatic non–small cell lung cancer. Given the dismal prognosis of brain metastases in these patients it is essential to identify subsets of patients that have a tumor microenvironment that might be more accessible to immunotherapeutic approaches. We provide evidence that a fraction of lung adenocarcinoma brain metastases is indeed characterized by higher tumor-infiltrating lymphocyte numbers and is in close association with an inflamed microenvironment and high expression of surfactant genes.

Lung cancer is the most common type of cancer worldwide and the most common cause of cancer-related death (1, 2). Lung cancer can histologically be divided into two groups: small-cell lung cancer and non–small cell lung cancer (NSCLC). NSCLC can be further subdivided into subtypes, the most frequent being adenocarcinoma (2, 3). Lung adenocarcinoma also accounts for the largest histologic subset of brain metastases (4). Of note, brain metastases are even more frequent than primary brain tumors (5). Patients with brain metastases have a poor median overall survival of 7–13 months (6). Treatment is usually multimodal and consists of systemic chemotherapy combined with microsurgery, stereotactic radiosurgery, and/or radiotherapy (7).

There is increasing evidence for the role of the host immune system in cancer development, suppression, and recurrence (8). In colorectal cancer for instance, the quantification of tumor-infiltrating lymphocytes (TIL) has become a valid prognostic marker for patient survival and is believed to be superior to the tumor–node–metastasis classification (9). Many studies have since confirmed the prognostic power of immune cell infiltrates in a large number of cancer types. The positive effect of cytotoxic T cells and Th1 T cells on survival has been shown; however, the influence of intratumoral Th2, Th17, and regulatory T cells (Tregs) on survival is less clear (10). In patients with lung adenocarcinoma with high overall T-cell counts and high cytotoxic T-cell counts, prolonged survival has been observed (11). Tregs, on the other hand, seem to impair prognosis (12).

However, in addition to quantity, the spatial distribution of immune cells and their vicinity to tumor cells in terms of localization within tumor islands or retention in the tumor stroma (13, 14) and the differences between primary tumors and subsequent metastases must be taken into account. It has been shown that infiltration with cytotoxic T cells decreases from primary lung cancer to metastases (15, 16). Nevertheless, data on the impact of T-cell infiltration on survival in brain metastasis remain controversial, with some studies suggesting a favorable role for CD3+, CD8+, and CD45R0+ T cells (17) and other studies showing no benefit (18, 19).

Understanding the interplay between the immune system and cancer cells is becoming even more critical in patients with brain metastases as checkpoint inhibitor therapies are on the rise for treating lung adenocarcinoma (20). Knowledge about the immune microenvironment in primary lung cancer cannot be extrapolated to include brain metastases, because of specialized resident immune and stromal cells, including microglia and astrocytes (21–23), and physical restriction by the blood–brain barrier (24). Genomic alterations between primary lung cancer and brain metastasis, as well as their connection to the immune system are the subject of ongoing research (25). Other data suggest that the amount of T-cell infiltration in brain metastasis might be related to different DNA methylation patterns (26). Thus, a thorough examination of brain metastasis biology, including similarities and differences to the primary tumor, is needed to enable us to understand the role of the immune system in brain metastasis.

Our study aims to provide a comprehensive analysis of the lung adenocarcinoma brain metastasis microenvironment. We assessed the number of TILs able to enter tumor islands and get into contact with tumor cells and showed a positive effect of overall TIL infiltration, as well as helper and cytotoxic T-cell infiltration on patient survival. Moreover, based on comparative transcriptomic analyses, we related the extent of infiltration in brain metastases with the expression of surfactant pathway–related genes and showed that this was also the case in primary lung adenocarcinoma. Multi-omic analysis suggested a surfactant pathway–associated change in the immune microenvironment that might render lung adenocarcinoma susceptible to immune checkpoint inhibition.

Patients

Brain metastasis samples were obtained from 81 patients who underwent surgery between November 2002 and December 2014 at the Department of Neurosurgery at Heidelberg University Hospital (Heidelberg, Germany). All patients were diagnosed with lung adenocarcinoma. Histology and a tumor cell content ≥ 60% were confirmed by board-certified neuropathologists (A. von Deimling and D. Reuss). Clinical follow-up and survival information was obtained from clinical records and the respective citizens' registration offices (between November 2003 and January 2016). The study was approved by the ethics committee of the Medical Faculty, University of Heidelberg (Heidelberg, Germany; reference number: 005/2003) and written informed consent was obtained from all patients in accordance with the Declaration of Helsinki and its later amendments.

Immunofluorescence staining

Tumor tissue was snap-frozen in liquid nitrogen–cooled isopentane immediately after resection. Frozen tissues were cut into 5–7 μm slices, acetone-fixed, and stored at −80°C until staining. Immunofluorescent staining was performed using anti-CD3 (Dako, A0452), anti-CD8 (Clone YTC182.20, Abcam, Ab60076), and anti-FOXP3 (Clone 236A/E7, Abcam, Ab20034) as described previously (27). DAPI (Invitrogen, D1306) was used to counterstain the nuclei. The incubation time was 1 hour, and this was followed by three washing steps. Secondary antibodies, including anti-rabbit Alexa Fluor 647 (Life Technologies, A21245), anti-rat Alexa Fluor 488 (Life Technologies, A11006), and anti-mouse Alexa Fluor 555 (Life Technologies, A31570) were then applied for 1 hour, and this was followed by three washing steps. Tumor cells were labeled using anti-pan cytokeratin (Progen, 61835; Dako, M0630) staining in combination with an antibody labeling kit (Thermo Fisher Scientific, Z25002). Isotype controls were rabbit IgG (Dako, X0936), IgG2b (eBioscience, 14-4031), and IgG1 (Abcam, ab91353). The antibody characteristics are listed in Supplementary Table S1.

Semi-automated quantification of T-cell infiltration

Following immunofluorescent staining, whole-slide images were generated for all 81 tumor samples using an automated microscope (Olympus IX51 equipped with a F-View II camera, both Olympus) at 20-fold magnification with the cellSens Software (Olympus). TissueQuest Software (version 4.0.1.0137, TissueGnostics GmbH) was then used for a semiautomated, objective, and quantitative analysis of stained cells in the whole-tumor section. Regions of stroma and necrosis were manually excluded to enable exclusive analysis of vital tumor islands. The following gating strategy was used. First, DAPI staining was used to identify nuclei and thus intact cells. CD3-expressing T cells were further characterized into cytotoxic and Tregs through coexpression of CD8 and FOXP3, respectively. CD4 cells were indirectly assessed by the CD3+/CD8 population. Parameters measured included staining intensity, range and variance of intensity, and nuclei size. Manual backward gating was performed for quality control. T-cell infiltration was quantified (cells/mm2), analyzed, and grouped into tissues with high and low infiltration based on the median. Further detail including definition of different cell types is provided in Supplementary Fig. S1A.

RNA and DNA isolation and further analysis

RNA (n = 63) and DNA (n = 20) were extracted from the 81 tumor samples using the AllPrep DNA/RNA/Protein Mini Kit (Qiagen). Analyte quality and concentration were monitored using the NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific) and Bioanalyzer 2100 (Agilent). Microarray and methylation analysis were performed at the DKFZ Genomics and Proteomics Core Facility (DKFZ) using HumanHT-12 v4 Expression BeadChip and Infinium MethylationEPIC BeadChips Kits (Illumina). Microarray data obtained were normalized, log2-transformed, and median-centered (28). Differences in genes expressed between tumor tissues with high- and low–T-cell infiltration were identified by an independent two-sample t test and the log2-fold change (P < 0.05, fold change > |1.5|). The transcriptomic classification of brain metastases into the surfactanthigh and surfactantlow group was done using k-means clustering with k = 2. Methylation data were further processed using the Minfi package (version 1.28.4) in the R Software environment and split by the mRNA surfactant groups. Datasets are accessible at ArrayExpress (accession numbers E-MTAB-8659 and E-MTAB-8660).

Multi-omic The Cancer Genome Atlas data analysis

Normalized RNA sequencing (RNA-seq) data [The Cancer Genome Atlas (TCGA) RNA-Seq v2 Level 3, n = 230] for lung adenocarcinomas were obtained through the FireBrowse database (www.firebrowse.org, accessed on January 28, 2016). Mutation (n = 230), SNP-array (SNP6 level 3, n = 357), and methylation data (450k methylation data level 1 and 3, n = 206) and associated clinical data were obtained from the supplementary data of the primary TCGA publication (29).

The surfactant class (surfactantlow vs. surfactanthigh) was defined by k-means clustering (k = 2) of all lung adenocarcinoma samples (n = 230) according to the RNA expression of surfactant metabolism–related genes SFTPB, STFPA1, and NAPSA.

Only protein-changing variants were considered for mutation data. Analyses of copy number data were performed using the GenomicRanges package (version 1.34.0) in the R Software environment. Methylation of the target genes was assessed and compared with their level of RNA expression using an independent two-sample t test. Differentially methylated genes were identified between the matched gene expression profiles by t test, log2-fold change CpG sites were mapped to the genome using the Gviz package (version 1.26.5), and immune cell counts were estimated using TCGA data level I methylation data and a deconvolution algorithm within the Minfi package, as mentioned above.

The immune subtype of TCGA samples (C1–C6) was derived from supplementary data recently published by Thorsson and colleagues (30).

Correlation of bulk transcriptomics to single-cell sequencing analysis in NSCLC and normal lung tissues

To assess the cell type–specific expression of the surfactant-related genes SFTPA1, SFTPB, and NAPSA, we downloaded preprocessed single-cell RNA-sequencing data (scRNA-seq; log2 counts per million, log2cpm), tSNE map, as well as the categorization into the different cell types from ArrayExpress (accession numbers E-MTAB-6149 and E-MTAB-6653). The dataset published by Lambrechts and colleagues (31) comprises a catalog of 52,698 cells from both normal lung tissue as well as non–small cell lung carcinomas. Alveolar cells type 2 (AT2) cell fractions were estimated in the bulk transcriptomic brain metastasis microarray data using the standard least-squares methodology (32) in the CellMix R package (33).

Statistical analysis

All data and statistical analyses were carried out using the R software (version 3.5.1). Log-rank tests and the Cox proportional hazard model were used for univariate and multivariate survival comparisons, respectively. Correlation analysis was done using the Pearson product-moment correlation coefficient. A P < 0.05 was considered statistically significant. Significance levels were labelled as follows *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

Clinicopathologic parameters of study set

Samples from 81 patients with lung adenocarcinoma brain metastasis were included in the semiautomated whole-slide analysis of immune infiltrates (Table 1). Univariate analysis of clinicopathologic parameters revealed a significant survival benefit for younger patients (P = 0.003, age ≤ 61 years), patients with a Karnofsky score above 70 (P < 0.001), and patients without extracranial metastases (P = 0.048; Supplementary Fig. S1B–S1D).

Table 1.

Descriptive statistics and univariate survival analysis of clinical parameters in the lung adenocarcinoma brain metastases study set (n = 81). The presented P value is based on the log-rank test.

Descriptive statisticsSurvival analysis
Variablen patients (%)median (range)PHR [95% CI (months)]
Gender Male 40 (49.4)  0.606 1.16 (6.5–30.5) 
 Female 41 (50.6)    
Age at diagnosis [years] < Median 41 (50.6) 61 (40–80) 0.003** 0.43 (12.5–33.9) 
 ≥ Median 40 (49.4)    
Number of BMa Single 49 (60.5) 1 (1–19) 0.667 1.13 (10.6–28.3) 
 Multiple 32 (39.5)    
BM occurrence Synchronous 37 (45.7)  0.936 1.02 (9.1–30.9) 
 Metachronous 42 (51.9)    
 Unknown 2 (2.5)    
Preoperative Yes 28 (34.6)  0.492 0.60 (10.6–NA) 
Chemotherapy No 40 (49.4)    
 Unknown 13 (16)    
Preoperative WBRT Yes 2 (2.5)  NA NA 
 No 63 (77.8)    
 Unknown 16 (19.8)    
Postoperative WBRT Yes 60 (74.1)  0.230 0.64 (15.4–28.3) 
 No 15 (18.5)    
Karnofsky scorea > 70 60 (74.1) 90 (40–100) < 0.001*** 0.35 (3.3–17.6) 
 ≤ 70 17 (21.0)    
 Unknown 4 (4.9)    
Smoker Yesb 69 (85.2)  0.646 0.84 (10.6–25.9) 
 Never 11 (13.6)    
 Unknown 1 (1.2)    
Tumor stagec Stage I–III 33 (40.7)  0.756 0.91 (10.1–30.5) 
 Stage IV 46 (56.8)    
 Unknown 2 (2.5)    
Size of resected BM ≥ 3 cm 37 (45.7) 3.0 (0.9–6.0) 0.786 1.08 (8.9–25.9) 
 < 3 cm 37 (45.7)    
 Unknown 7 (8.6)    
Extent of resection Totald 45 (55.5)  0.868 0.95 (11.2–28.3) 
 Partiale 34 (42.0)    
 Unknown 2 (2.5)    
Extracranial metastasesa Yes 19 (23.4)  0.048* 1.89 (2.8–17.6) 
 No 60 (74.1)    
 Unknown 2 (2.5)    
Descriptive statisticsSurvival analysis
Variablen patients (%)median (range)PHR [95% CI (months)]
Gender Male 40 (49.4)  0.606 1.16 (6.5–30.5) 
 Female 41 (50.6)    
Age at diagnosis [years] < Median 41 (50.6) 61 (40–80) 0.003** 0.43 (12.5–33.9) 
 ≥ Median 40 (49.4)    
Number of BMa Single 49 (60.5) 1 (1–19) 0.667 1.13 (10.6–28.3) 
 Multiple 32 (39.5)    
BM occurrence Synchronous 37 (45.7)  0.936 1.02 (9.1–30.9) 
 Metachronous 42 (51.9)    
 Unknown 2 (2.5)    
Preoperative Yes 28 (34.6)  0.492 0.60 (10.6–NA) 
Chemotherapy No 40 (49.4)    
 Unknown 13 (16)    
Preoperative WBRT Yes 2 (2.5)  NA NA 
 No 63 (77.8)    
 Unknown 16 (19.8)    
Postoperative WBRT Yes 60 (74.1)  0.230 0.64 (15.4–28.3) 
 No 15 (18.5)    
Karnofsky scorea > 70 60 (74.1) 90 (40–100) < 0.001*** 0.35 (3.3–17.6) 
 ≤ 70 17 (21.0)    
 Unknown 4 (4.9)    
Smoker Yesb 69 (85.2)  0.646 0.84 (10.6–25.9) 
 Never 11 (13.6)    
 Unknown 1 (1.2)    
Tumor stagec Stage I–III 33 (40.7)  0.756 0.91 (10.1–30.5) 
 Stage IV 46 (56.8)    
 Unknown 2 (2.5)    
Size of resected BM ≥ 3 cm 37 (45.7) 3.0 (0.9–6.0) 0.786 1.08 (8.9–25.9) 
 < 3 cm 37 (45.7)    
 Unknown 7 (8.6)    
Extent of resection Totald 45 (55.5)  0.868 0.95 (11.2–28.3) 
 Partiale 34 (42.0)    
 Unknown 2 (2.5)    
Extracranial metastasesa Yes 19 (23.4)  0.048* 1.89 (2.8–17.6) 
 No 60 (74.1)    
 Unknown 2 (2.5)    

Note: Significant values were displayed in bold. P values: *, < 0.05; **, < 0.01; ***, < 0.001.

Abbreviations: BM, brain metastases; HR, hazard ratio; NA, not available; n, number; WBRT, whole-brain radiotherapy.

aAt the time of neurosurgical resection.

bCurrent and past.

cAt the time of first diagnosis.

dMacroscopically no tumor tissue left in the brain.

eIncomplete resection of singular metastasis or other metastases left in the brain.

Semiautomated whole-slide analysis of brain metastases reveals vast difference in TIL density

Multicolor immunofluorescent staining was performed (Fig. 1A and B) to quantify TILs in general (CD3+), helper T cells (CD3+CD8FOXP3), and cytotoxic T cells (CD3+CD8+FOXP3). FOXP3 staining was used to define Tregs (CD3+CD8FOXP3+ for classical Tregs and CD3+CD8+FOXP3+ for CD8+ Tregs; Supplementary Fig. S2). The majority of stained T cells, independent of their subtype, were retained in the tumor stroma while few cells were found to infiltrate tumor islands and thus be in direct contact with tumor cells (Fig. 1B). The term TILs is usually imprecisely used for both stromal and epithelial T cells (34). To obtain more accurate data on T cells that can reach and kill tumor cells, we excluded the stromal areas from our analysis and focused only on the actual tumor-infiltrating T cells. Marked differences were observed in overall TIL density (range, 10–1,700 CD3+ cells/mm2; Fig. 1C). Median intratumoral T-cell infiltration was 162 cells/mm2. Th cells were the most common subtype (156/mm2) followed by cytotoxic T cells (87/mm2), Tregs (19/mm2), and CD8+ Tregs (2/mm2). The overall frequency of FOXP3+ T cells was 8/mm2 (ranging from 0 to 170/mm2).

Figure 1.

Study design and T-cell infiltration in lung adenocarcinoma brain metastases. A, Graphical abstract of the conducted experiments. In the discovery cohort of brain metastases, immunofluorescent stainings (IF, n = 81) and RNA microarray analysis (n = 63) were performed. Validation was done using RNA-seq, methylation, mutation, and alteration of copy number data from the TCGA primary lung adenocarcinoma (LUAD) dataset. In addition, a scRNA-seq dataset of NSCLC was used. Finally, an integrative data analysis was conducted. B, Immunofluorescent stainings of DAPI, CD3, and cytokeratins. Pictures show a representative sample in multicolor view and single-color channels. TILs, defined as infiltrating into the tumor, are marked with a red arrow. C, Barplots and boxplots of the different T-cell subtypes in the cohort. Th cells (CD3+CD8FOXP3, median: 156/mm2), cytotoxic T cells (CD3+CD8+FOXP3, median: 87/mm2), classical Tregs (CD3+CD8FOXP3+, median: 19/mm2), and CD8+ Tregs (CD3+CD8+FOXP3+, median: 2/mm2). Clinical parameters are indicated below the barplot.

Figure 1.

Study design and T-cell infiltration in lung adenocarcinoma brain metastases. A, Graphical abstract of the conducted experiments. In the discovery cohort of brain metastases, immunofluorescent stainings (IF, n = 81) and RNA microarray analysis (n = 63) were performed. Validation was done using RNA-seq, methylation, mutation, and alteration of copy number data from the TCGA primary lung adenocarcinoma (LUAD) dataset. In addition, a scRNA-seq dataset of NSCLC was used. Finally, an integrative data analysis was conducted. B, Immunofluorescent stainings of DAPI, CD3, and cytokeratins. Pictures show a representative sample in multicolor view and single-color channels. TILs, defined as infiltrating into the tumor, are marked with a red arrow. C, Barplots and boxplots of the different T-cell subtypes in the cohort. Th cells (CD3+CD8FOXP3, median: 156/mm2), cytotoxic T cells (CD3+CD8+FOXP3, median: 87/mm2), classical Tregs (CD3+CD8FOXP3+, median: 19/mm2), and CD8+ Tregs (CD3+CD8+FOXP3+, median: 2/mm2). Clinical parameters are indicated below the barplot.

Close modal

A correlation analysis was performed to assess the extent to which T-cell subtypes infiltrate tumor islands simultaneously or independently from one another (Supplementary Fig. S3). A Pearson correlation coefficient greater than 0.6 was found for all except one test, indicating a positive correlation between all T-cell subtypes except Th and CD8+ Tregs. We further assessed the impact of synchronous versus metachronous metastases, as well as preoperative therapeutic regimens on the TIL measurements. No correlation was found (Fig. 1C; Supplementary Fig. S4).

We observed a substantial but highly variable intratumoral T-cell infiltration of brain metastasis. This mainly consisted of cytotoxic and Th cells, with a dominating majority of Th cells, while the frequency of FOXP3+ T cells was more than 19-fold lower. Correlation analysis revealed that infiltration of T-cell subsets correlated with the total T-cell count and were independent from time between diagnosis of cancer and the occurrence of brain metastases.

Tumor-entering TILs are an independent prognostic parameter for overall survival in patients with brain metastases

We used log-rank tests to assess whether the observed infiltration differences had an impact on overall survival. As previously mentioned, we divided our study sample into high (n = 40) and low (n = 41) infiltration for overall TIL density and each of the T-cell subtypes.

A significant survival benefit was observed in tumors with high overall TIL density (P = 0.007), and for Th cell (P = 0.021) and cytotoxic T-cell (P = 0.011) subtypes (Fig. 2AG). There was, however, no significant difference in survival for FOXP3+ cells (Tregs, P = 0.415; CD8+ Tregs, P = 0.767). In the multivariate model, age at diagnosis and Karnofsky score maintained significance. High overall tumor-entering TIL density was an independent prognostic marker (P = 0.016; HR, 2.06). A trend was observed for Th cells (P = 0.057; HR, 1.78), cytotoxic T cells (P = 0.110; HR, 1.61). Classical Tregs as well as CD8+ Tregs did not convey a survival benefit (P = 0.592, HR: 1.17; P = 0.472, HR: 1.24, respectively).

Figure 2.

Survival association with density of different tumor-entering T-cell subsets and multivariate analysis. Infiltration was grouped by median separation into high (solid) and low (dashed) infiltration density. Log-rank test was used to calculate survival differences, and Kaplan–Meier plots were drawn for display. AC, High overall TIL density, high TH-cell density, and high-cytotoxic T-cell density improved survival. DF, No survival difference was observed for Tregs. G, Age at diagnosis, Karnofsky score, and overall TIL density remained independent prognostic factors in multivariate analysis (Cox proportional hazard ratio). Significant values were displayed in bold; P values: *, < 0.05; **, < 0.01; ***, < 0.001.

Figure 2.

Survival association with density of different tumor-entering T-cell subsets and multivariate analysis. Infiltration was grouped by median separation into high (solid) and low (dashed) infiltration density. Log-rank test was used to calculate survival differences, and Kaplan–Meier plots were drawn for display. AC, High overall TIL density, high TH-cell density, and high-cytotoxic T-cell density improved survival. DF, No survival difference was observed for Tregs. G, Age at diagnosis, Karnofsky score, and overall TIL density remained independent prognostic factors in multivariate analysis (Cox proportional hazard ratio). Significant values were displayed in bold; P values: *, < 0.05; **, < 0.01; ***, < 0.001.

Close modal

In conclusion, high overall TIL density and high helper and cytotoxic T-cell density were found to improve survival in univariate analyses. Overall TIL density remained significant in multivariate analysis. The density of FOXP3+ T cells had no impact on survival. CD3+ T cells were identified as the most reliable prognostic immune parameter in both the univariate and multivariate analysis.

TIL density was associated with overexpression of surfactant pathway–related genes in lung adenocarcinoma brain metastases

As a result of the observed survival association, we next aimed to identify molecular determinants related to high- and low-TIL infiltration and performed comparative transcriptomic analyses in 63 of the 81 brain metastasis samples. Ten genes were found to be differentially expressed between the overall TIL and/or Th/cytotoxic cell high and low groups, including FOLR1, CLIC6, CEACAM5, SFTPA1, NAPSA, MUC16, C9orf152, TMPRSS2, SFTPB, and CLDN10 (Fig. 3A; Supplementary Table S2). Of these, SFTPB and NAPSA showed the most distinct differences in expression (Supplementary Fig. S5A and S5B). These genes also showed significantly higher expression in the highly infiltrated groups (Supplementary Fig. S5B). Pathway and gene-function analysis revealed these genes, together with SFTPA1, to be part of the surfactant metabolism pathway.

Figure 3.

Differentially expressed genes and further analysis of surfactant genes in our cohort (n = 63) and a TCGA dataset (n = 230). A, Scatterplot of differentially expressed genes between the groups of high- and low-overall TIL density, Th-cell density, and cytotoxic T-cell density (Heidelberg brain metastasis dataset). Cutoff for differential expression: log-rank P < 0.05 and log2 fold-change > 1.5. B, Heatmap of SFTPA1, SFTPB, and NAPSA expression (Heidelberg brain metastasis dataset). Two groups can be identified by hierarchical clustering. C, Survival analysis of the two clusters using log-rank test and Kaplan–Meier plots (Heidelberg brain metastasis dataset). Cluster 2 (surfactanthigh class) is positively associated with survival. D, Heatmap of RNA-seq expression of SFTPA1, SFTPB, and NAPSA (TCGA dataset). Clustering was done as described above and revealed a surfactanthigh and surfactantlow group. E, Multivariate survival analysis using Cox-proportional hazard ratio (TCGA dataset). Parameters included patient age at diagnosis and surfactant expression clustering. Significant differences were observed, especially in the group of older patients. LUAD, lung adenocarcinoma.

Figure 3.

Differentially expressed genes and further analysis of surfactant genes in our cohort (n = 63) and a TCGA dataset (n = 230). A, Scatterplot of differentially expressed genes between the groups of high- and low-overall TIL density, Th-cell density, and cytotoxic T-cell density (Heidelberg brain metastasis dataset). Cutoff for differential expression: log-rank P < 0.05 and log2 fold-change > 1.5. B, Heatmap of SFTPA1, SFTPB, and NAPSA expression (Heidelberg brain metastasis dataset). Two groups can be identified by hierarchical clustering. C, Survival analysis of the two clusters using log-rank test and Kaplan–Meier plots (Heidelberg brain metastasis dataset). Cluster 2 (surfactanthigh class) is positively associated with survival. D, Heatmap of RNA-seq expression of SFTPA1, SFTPB, and NAPSA (TCGA dataset). Clustering was done as described above and revealed a surfactanthigh and surfactantlow group. E, Multivariate survival analysis using Cox-proportional hazard ratio (TCGA dataset). Parameters included patient age at diagnosis and surfactant expression clustering. Significant differences were observed, especially in the group of older patients. LUAD, lung adenocarcinoma.

Close modal

Hierarchical clustering of the brain metastasis samples, on the basis of the expression of the three surfactant metabolism pathway genes SFTPA1, SFTPB, and NAPSA (referred to as surfactant genes from now on), revealed two clusters and coexpression of the three genes. The clusters are further referred to as the classes surfactanthigh and surfactantlow (Fig. 3B; surfactanthigh, n = 22; surfactantlow, n = 41). We tested for survival differences between the two groups, and observed a clear survival benefit in the surfactanthigh group (Fig. 3C; P = 0.002). When analyzing the influence of each of the three genes on survival separately, we found that the prognostic performance of the SFTPA1 expression was even more pronounced than the combined model of all three genes (P < 0.0001; Supplementary Fig. S6A). NAPSA also maintained significance when divided by its median (P = 0.003; Supplementary Fig. S6B). The effect on survival for SFTPB was most pronounced when divided into a smaller low-expression (n = 22) and a larger high-expression group (n = 41), which was more similar to its distribution in the heatmap (P = 0.007; Supplementary Fig. S6C). We further analyzed possible associations between clinical parameters and surfactant gene expression. Only patient age was identified to differ between the two surfactant groups (older age in surfactanthigh group; Supplementary Table S3). As done for the full study cohort of 81 patients, we again performed univariate survival analysis for the patient subset with mRNA microarray information (n = 63; Supplementary Table S4). Here, age and Karnofsky score but not extracranial metastases were prognostic in this subcohort. In a multivariate model, the surfactant grouping remained an independent prognostic parameter (Supplementary Table S5).

Altogether, we found 10 differentially expressed genes, 3 of which belonged to the surfactant metabolism pathway (surfactant genes). The two classes, surfactanthigh and surfactantlow revealed considerable survival differences in favor of patients with high-surfactant gene expression, independent of clinical parameters.

Surfactant classes are present in primary lung adenocarcinomas

After identifying two surfactant classes of prognostic relevance, we tried to confirm this using an independent dataset. We used the publicly available TCGA primary lung adenocarcinoma datasets for this purpose, which contain RNA-seq, mutation, copy number, and methylation data.

When the 230 lung adenocarcinoma samples with RNA-seq data available were clustered according to expression of SFTPB, SFTPA1, and NPASA, 196 samples were found to belong to the surfactanthigh class and 24 to the surfactantlow class (Fig. 3D).

In contrast to brain metastases, the fraction of primary lung adenocarcinoma harboring a surfactanthigh class was greater. As before, we queried the prognostic impact regarding overall survival of the surfactant class. Among the prognostic clinical parameters of our brain metastasis study sample, only age at diagnosis was available for the TCGA dataset. Univariately, the prognostic performance of surfactant class and age at diagnosis just failed to reach significance (P = 0.052 and P = 0.053, respectively). When combined in a multivariate model, however, they had a significant effect on patient survival (P < 0.0001; Fig. 3E). The subset with the most dismal prognosis was older patients with a surfactantlow tumor.

In primary lung adenocarcinoma, a subset of tumors shows coordinated downregulation of the surfactant genes (surfactantlow class) although in a lower proportion of patients than in brain metastasis. This, in combination with age at diagnosis, was associated with poor survival in a multivariate model.

A subset of primary lung adenocarcinoma tumor cells resemble surfactant-expressing AT2 cells

In the normal lung, expression of surfactant genes is characteristic for AT2 cells. To understand the expression of surfactant genes in lung adenocarcinoma brain metastases, we investigated the intratumoral heterogeneity of primary lung adenocarcinomas by mining the scRNA-seq atlas published by Lambrechts and colleagues (31). The atlas is comprised of 52,689 cells of NSCLC and normal lung samples. A subset of tumor cells showed a transcriptomic profile clustering together with AT2 cells from normal lung samples (Fig. 4A). Visualizing the expression landscape separately for AT2 cells and tumor cells confirms an AT2-like high expression of surfactant genes in a considerable subset of tumor cells (Fig. 4B). Notably, while SFTPA1 and NAPSA were expressed almost exclusively in AT2 cells, SFTPB was also found in AT1 and epithelial cells and quite prominently in club cells (Supplementary Fig. S6D–S6F). Furthermore, applying the expression signature of AT2 cells to our mRNA study sample of brain metastasis using a deconvolution method (standard least-squares), we found that the estimated fraction of AT2 cells was significantly higher in the surfactanthigh group (P = 0.024; Fig. 4C). In conclusion, these findings suggest that in lung adenocarcinoma brain metastases, a higher TIL density is associated with a higher fraction of AT2-like tumor cells.

Figure 4.

ScRNA-seq analysis in NSCLC and normal lung atlas (52,698 cells). A, tSNE plot of all tumor cells and AT2 cells only. B, tSNE plot substratified for AT2 cells and tumor cells. Cells are colored according to the expression of the surfactant genes (log2cpm). C, Deconvolution analysis in lung adenocarcinoma brain metastasis cohort (n = 63) to estimate the fraction of AT2 cells in the bulk expression data.

Figure 4.

ScRNA-seq analysis in NSCLC and normal lung atlas (52,698 cells). A, tSNE plot of all tumor cells and AT2 cells only. B, tSNE plot substratified for AT2 cells and tumor cells. Cells are colored according to the expression of the surfactant genes (log2cpm). C, Deconvolution analysis in lung adenocarcinoma brain metastasis cohort (n = 63) to estimate the fraction of AT2 cells in the bulk expression data.

Close modal

Expression of surfactant genes was not driven by mutations or copy number changes

We performed multi-omic analysis using the TCGA data to shed light on the underlying mechanism leading to differential expression of the surfactant genes. We preprocessed and matched the available mutation data (total number of mutations across the study cohort = 68,270) to our preexisting surfactant classes. We first had a look at the number of mutations found within each of the three surfactant genes (Supplementary Table. S6). We found 1 patient with a missense mutation (c.230G>T) in SFTPA1; all other patients had no mutation in the gene. Two patients had a mutation in SFTPB (one nonsense mutation c.133C>T; one splice site, c.393_splice), and 1 patient had a missense mutation (c.25C>T) in NAPSA. There was no defined difference between the most recurrent genes in the two surfactant groups (Supplementary Table S7).

We subsequently assessed whether alterations in copy number affected differential surfactant gene expression. The SNP6 data were matched to the RNA-seq data in the aforementioned way. We plotted the chromosomes on which our target genes are located (chromosomes 2, 10, and 19) and colored the region of interest according to the respective surfactant group (Supplementary Fig. S6G). None of the genes showed any remarkable difference in copy numbers.

In summary, mutations and alterations in copy number did not play a role in the differential expression of the surfactant genes.

DNA methylation analysis of surfactant genes

DNA methylation was investigated after mutations and alterations in copy number were ruled out as reasons for differential expression in the surfactant genes. Thus, we took the 450k methylation dataset available from the TCGA study sample and calculated differential beta values between the two surfactant groups. While we could not observe a global difference in the methylome (Supplementary Fig. S7A) between surfactanthigh and surfactantlow tumors, significant hypermethylation of CpG sites could be observed for all 3 surfactant genes in the surfactantlow class (SFTPA1, P = 0.001; SFTPB, P = 0.005; NAPSA, P < 0.001; Supplementary Fig. S7B–S7E; Supplementary Table S8).

As this observation was only made in the primary lung adenocarcinomas of the TCGA, we aimed to validate this finding in our lung adenocarcinoma brain metastasis cohort and performed methylation analysis of 10 surfactanthigh and surfactantlow tumors. Likely due to the small sample size, we could only validate a differential methylation for SFTPB (Supplementary Fig. S7F and S7G).

In conclusion, we observed a hypermethylation of the surfactant genes in surfactantlow primary lung adenocarcinoma. However, this finding could only be validated for SFTPB.

Estimation of immune cell fractions identified higher abundance of CD4+ T cells and natural killer cells in surfactanthigh primary lung adenocarcinomas

Immune cell fractions in primary lung adenocarcinoma were estimated from the methylome data using a deconvolution algorithm (35). This enabled comparison with our initial T-cell count experiment (Fig. 5A). CD4+ T cells were the largest group with an estimated median abundance of over 30%. This was followed by B cells, monocytes, natural killer (NK) cells, and granulocytes with an abundance of around 20% each. CD4+ T cells were significantly more prevalent in the surfactanthigh group, whereas NK cells were significantly more abundant in the surfactantlow group (Fig. 5B). There was no difference in CD8+ T cells between the two groups. However, reliable estimates for CD8+ T cells were difficult to achieve due to a smaller proportion of this cell type.

Figure 5.

Dissecting the tumor microenvironment of surfactantlow and surfactanthigh primary lung adenocarcinomas. A, Abundance distributions of CD4+ T cells, B cells, monocytes, NK cells, granulocytes, and CD8+ T cells quantified by means of a methylation-based deconvolution algorithm in the full study set. B, Cell estimation differences between the surfactanthigh and surfactantlow classes (CD4, CD8, and NK cells). t Tests revealed significantly higher CD4+ T cells and significantly lower NK cells in the surfactanthigh class. C, Categorization of surfactant groups according to the immune subtypes (C1–C6) proposed by Thorsson and colleagues (30). The inflammatory C3 subtype is overrepresented in the surfactanthigh group. D, Prediction of responsiveness to immunotherapies (ITs) based on immune types in light of recent literature (38). Grouping into hot, altered, and cold tumor immune status (TIS) was based on the recently reviewed ImmunoScore (36). E, Expression levels of surfactant metabolism genes according to immune type.

Figure 5.

Dissecting the tumor microenvironment of surfactantlow and surfactanthigh primary lung adenocarcinomas. A, Abundance distributions of CD4+ T cells, B cells, monocytes, NK cells, granulocytes, and CD8+ T cells quantified by means of a methylation-based deconvolution algorithm in the full study set. B, Cell estimation differences between the surfactanthigh and surfactantlow classes (CD4, CD8, and NK cells). t Tests revealed significantly higher CD4+ T cells and significantly lower NK cells in the surfactanthigh class. C, Categorization of surfactant groups according to the immune subtypes (C1–C6) proposed by Thorsson and colleagues (30). The inflammatory C3 subtype is overrepresented in the surfactanthigh group. D, Prediction of responsiveness to immunotherapies (ITs) based on immune types in light of recent literature (38). Grouping into hot, altered, and cold tumor immune status (TIS) was based on the recently reviewed ImmunoScore (36). E, Expression levels of surfactant metabolism genes according to immune type.

Close modal

Surfactanthigh lung adenocarcinomas display a transcriptomic profile resembling an inflammatory microenvironment

To explore differences in the microenvironment of surfactantlow and surfactanthigh tumors, we obtained the transcriptome-based immune subtypes (C1–C6) recently published by the TCGA working group (30). Types C1 (wound healing) and C2 (INFγ dominant) were most prevalent in the surfactantlow group (38.2% and 41.2%, respectively; Fig. 5C). The other types, including C3 (inflammatory, 2.9%), C4 (lymphocyte depleted, 5.9%), C5 (immunologically quiet, 0.0%), and C6 (TGF-β dominant, 11.8%), were less frequent. In the surfactanthigh group, the largest group was C3 (39.4 %), followed by C2 (35.8%), C1 (12.4%), C6 (7.8%), C5 (4.7%), and C4 (0.0%). Most strikingly, there was a clear difference in the proportions of immune subtype C3 (inflammatory). Likewise, expression for the surfactant genes was highest in the subtype C3 (Fig. 5E).

As the proposed immune subtypes have not been tested for their predictive performance in prospective clinical trials, we related the TCGA classification to the ImmunoScore developed by Galon and colleagues (ref. 36; Fig. 5D). Galon and colleagues mainly differentiate the tumor immune status (TIS) into cold, altered, and hot. C1 and C2 are most closely related to an altered TIS and respond less well to immunotherapies due to higher proliferation rates (37). In contrast, the inflammatory C3 type corresponds to a hot TIS, which is an indicator of good response to immunotherapy (36). The immune types C4 and C5 correspond to cold TIS, for which immune checkpoint inhibition is believed to be unsuccessful. It remains unclear how the TGF-β dominant group (C6) corresponds to TIS (36). We further evaluated the expression of surfactant genes stratified by immune subtype. Expression of all 3 genes was found to be highest in the C3 type.

In summary, our data were matched to preexisting nomenclatures to characterize the different immune subtypes of surfactanthigh and surfactantlow tumors. We found the inflammatory C3 type to be the largest group in surfactanthigh tumors, but it was rarely found in surfactantlow tumors. Interestingly, the C3 type is believed to be most susceptible to immunotherapies (30, 36–38).

The role of TILs as a prognostic marker has been increasingly recognized in a number of tumor entities (10). In NSCLC, CD3 and CD8 TILs in particular have been associated with a favorable prognosis (34). Few studies on TILs in brain metastasis have been published and despite a robust design, their results are controversial (17, 39). Data regarding the relevance of TILs that are able to get in close contact with tumor cells in brain metastasis and are not retained in the tumor stroma, and the molecular determinants involved in this process, are missing.

In this study, we found lymphocyte density to be highly variable in lung adenocarcinoma brain metastases. Some tissues were barely infiltrated, as expected in primary brain tumors (27), but as previously described some had abundant infiltrates (17). With regards to the association between infiltration of lymphocytes in direct contact with tumor cells and patient outcome, we showed that high overall TIL density is an independent positive prognostic marker for patients with lung adenocarcinoma brain metastasis. Helper and cytotoxic T cells appear to play the most important role within the overall TILs, despite not being significant in the multivariate model. These findings are in accordance with recent studies that showed the beneficial effect of high–T-cell infiltration (measured in the entire brain metastasis tissue, derived from lung adenocarcinoma or otherwise) on patient survival (40, 41). In contrast, Castaneda and colleagues found no significant difference between high and low CD3 TILs in 64 cases of brain metastasis (42). Tregs, which have been shown to impair the immune response to primary brain malignancies (43) and primary lung adenocarcinoma (44), did not impact patient survival in our study.

Our study is the most comprehensive of this sort to date. First, we used an innovative, semiautomated, objective, and quantitative, whole-tumor slide multicolor immunofluorescent approach and separated tumor islands from tumor stroma, as emphasized by Geng and colleagues (34). Second, our sample size of 81 patients with lung adenocarcinoma brain metastasis was relatively large. Third, we assessed the survival benefit of intraepithelial lymphocyte infiltration and thus of T cells with the potential to kill tumor cells.

To gain insight into the tumor microenvironment as well as differences in gene expression associated with T-cell infiltration, we performed mRNA microarray analysis of 63 lung adenocarcinoma brain metastasis. Investigating differential expression between high- and low–T-cell infiltration, we found that 3 surfactant genes, SFTPA1, SFTPB, and NAPSA, were significantly overexpressed in patients with high intraepithelial T-cell infiltration. Multivariate survival analysis revealed that surfactanthigh gene expression was a strong positive indicator of survival independent of the clinical prognostic parameters age and Karnofsky score. To the best of our knowledge, this has never previously been shown for lung adenocarcinoma brain metastasis. There are data, however, which suggest a favorable role of surfactant in primary lung cancer. Recent studies suggest that surfactant proteins A and B are able to suppress the progression in NSCLC (45, 46), possibly through interaction with immune cells (46, 47) or by reducing the activity of EGFR and thereby acting in a similar manner to tyrosine-kinase inhibitors (48). The role of NAPSA in lung cancer is much less established, but recent studies suggest that NAPSA could have a supportive function with regards to the effect of EGFR tyrosine kinase inhibitors (49). Nevertheless, a limitation of the prognostic performance of the surfactant groups remains the lack of information regarding the postoperative palliative systemic therapies.

Next, we analyzed published scRNA-seq data to interrogate the cell type–specific expression of the surfactant genes. In normal lung tissues, surfactant is produced by AT2 cells. This could be confirmed in the scRNA-seq landscape. Intriguingly, the surfactant genes were also expressed by a subset of tumor cells that transcriptionally resemble AT2 cells. While it has been proposed that AT2 cells are the cell of origin of lung adenocarcinomas (Lambrechts and colleagues), our data, for the first time, shows that AT2-like tumor cells are also present in brain metastases and seem to be linked to a high-TIL density (31).

We subsequently performed an integrative multi-omic analysis in the primary lung adenocarcinoma TCGA cohort to question potential drivers of surfactant gene expression. This approach seemed most feasible because multi-omics datasets for lung adenocarcinoma brain metastases are not available.

Analysis of RNA-seq data revealed the presence of surfactanthigh and surfactantlow groups in both brain metastasis and the primary tumor; however, the proportions of high and low expression differed significantly between the groups. In brain metastasis, approximately two-thirds of patients presented with low expression of the surfactant genes. This is in sharp contrast to healthy lung tissue (www.proteinatlas.org), where the surfactant genes are highly expressed (50), and also in contrast to our findings in primary lung cancer (especially lung adenocarcinoma) where only 17.3% of cases presented with downregulation of the surfactant genes.

Mutation, methylation, and alteration in copy number were investigated in patients with primary lung adenocarcinoma to further understand the cause of differential expression. While mutation analysis did reveal some differences between the surfactanthigh and surfactantlow groups, this could not be linked to the differential expression of surfactant genes. This indicates that mutational burden is not the driver for overexpression. Methylation analysis, however, revealed significant differences in the methylation across the CpG sites of the 3 surfactant genes. To validate this finding in brain metastases, we analyzed a small brain metastasis subset of 20 cases. Mostly due to the small study sample we could only validate the differential expression for SFTPB. Larger studies are warranted to investigate this change in DNA methylation.

After having identified no clear driver for the overexpression of the surfactant genes, we next aimed to compare the tumor microenvironment between the surfactanthigh and surfactantlow groups. Methylation-based estimation of immune cell fractions in primary lung adenocarcinoma could validate the association between surfactant gene expression and a higher T-cell count (35). Yet, the deconvolution algorithms are not on par with the quantitative cell-based approach we used for brain metastasis.

It has been shown that numerous factors may play a role in the effectiveness of immunotherapies, including mutational burden, T-cell infiltration, and the tumor microenvironment with its various components (30, 51). Different classification systems are emerging to accommodate the overwhelming and yet-to-be integrated new information. Thorsson and colleagues provide one of the most comprehensive, pan-cancer classifications using TCGA data from more than 10,000 patients, and divide the tumor microenvironment into six distinct immune subtypes (30). Further efforts by Galon and colleagues have been directed to classify tumors into hot, altered, and cold by the amount of infiltrating lymphocytes (36). By using these findings, we predicted which immune types surfactanthigh and surfactantlow tumors correspond to and tried to understand what impact this could have on the effectiveness of immunotherapy. Interestingly, we found an enrichment of an inflammatory tumor subtype (C3) in surfactanthigh tumors, likely to correspond to the best response due to its inflammatory nature and intermediate proliferation (30, 36–38). Nevertheless, this putative biomarker warrants further preclinical and clinical confirmation. The association with the C3 subtype was also true for the expression of surfactant genes NAPSA, SFTPA1, and SFTPB.

Altogether, we identified the expression of the surfactant pathway–related genes SFTPA1, STFPB, and NAPSA to be a promising molecular determinant of the lung adenocarcinoma microenvironment. Intriguingly, this feature was shared between primary lung adenocarcinomas and lung adenocarcinoma brain metastases. Tumors with a high-surfactant gene expression (surfactanthigh class) harbored a high-intraepithelial T-cell infiltration and were characterized by an inflammatory and less immunosuppressive tumor environment.

J. Debus reports receiving commercial research grants from Merck Serono, Accuray, and Raysearch, and is an advisory board member/unpaid consultant for Accuray, Merck Serono, and ViewRay. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Pocha, A. Mock, C. Rapp, S. Dettling, J. Debus, C.C. Herold-Mende

Development of methodology: K. Pocha, A. Mock, N. Grabe

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Pocha, C. Jungk, D. Reuss, J. Debus, A. von Deimling, A. Abdollahi, C.C. Herold-Mende

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Pocha, A. Mock, S. Dettling, R. Warta, C. Geisenberger, L.R. Martins, A. von Deimling, C.C. Herold-Mende

Writing, review, and/or revision of the manuscript: K. Pocha, A. Mock, C. Rapp, S. Dettling, C. Geisenberger, C. Jungk, J. Debus, A. Abdollahi, A. Unterberg, C.C. Herold-Mende

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Warta, J. Debus, A. von Deimling

Study supervision: S. Dettling, R. Warta, J. Debus, C.C. Herold-Mende

We thank Mandy Barthel, Frederik Enders, and Anja Metzner for review of patient data. Furthermore, we thank Farzaneh Kashfi, Hildegard Göltzer, Ilka Hearn, and Melanie Greibich for their excellent technical assistance.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Parkin
DM
,
Bray
F
,
Ferlay
J
,
Pisani
P
. 
Global cancer statistics, 2002
.
CA Cancer J Clin
2005
;
55
:
74
108
.
2.
Ginsberg
MS
,
Grewal
RK
,
Heelan
RT
. 
Lung cancer
.
Radiol Clin North Am
2007
;
45
:
21
43
.
3.
Youlden
DR
,
Cramb
SM
,
Baade
PD
. 
The International Epidemiology of Lung Cancer: geographical distribution and secular trends
.
J Thorac Oncol
2008
;
3
:
819
31
.
4.
Nayak
L
,
Lee
EQ
,
Wen
PY
. 
Epidemiology of brain metastases
.
Curr Oncol Rep
2012
;
14
:
48
54
.
5.
Gavrilovic
IT
,
Posner
JB
. 
Brain metastases: epidemiology and pathophysiology
.
J Neurooncol
2005
;
75
:
5
14
.
6.
Kondziolka
D
,
Martin
JJ
,
Flickinger
JC
,
Friedland
DM
,
Brufsky
AM
,
Baar
J
, et al
Long-term survivors after gamma knife radiosurgery for brain metastases
.
Cancer
2005
;
104
:
2784
91
.
7.
Kim
YH
,
Nagai
H
,
Ozasa
H
,
Sakamori
Y
,
Mishima
M
. 
Therapeutic strategy for non-small-cell lung cancer patients with brain metastases (review)
.
Biomed Rep
2013
;
1
:
691
6
.
8.
Galon
J
,
Pagès
F
,
Marincola
FM
,
Angell
HK
,
Thurin
M
,
Lugli
A
, et al
Cancer classification using the Immunoscore: a worldwide task force
.
J Transl Med
2012
;
10
:
205
10
.
9.
Galon
J
,
Mlecnik
B
,
Bindea
G
,
Angell
HK
,
Berger
A
,
Lagorce
C
, et al
Towards the introduction of the “immunoscore” in the classification of malignant tumours
.
J Pathol
2014
;
232
:
199
209
.
10.
Fridman
WH
,
Pagès
F
,
Sautès-Fridman
C
,
Galon
J
. 
The immune contexture in human tumours: impact on clinical outcome
.
Nat Rev Cancer
2012
;
12
:
298
306
.
11.
Bremnes
RM
,
Busund
L-T
,
Kilvær
TL
,
Andersen
S
,
Richardsen
E
,
Paulsen
EE
, et al
The role of tumor-infiltrating lymphocytes in development, progression, and prognosis of non-small cell lung cancer
.
J Thorac Oncol
2016
;
11
:
789
800
.
12.
Tao
H
,
Mimura
Y
,
Aoe
K
,
Kobayashi
S
,
Yamamoto
H
,
Matsuda
E
, et al
Prognostic potential of FOXP3 expression in non-small cell lung cancer cells combined with tumor-infiltrating regulatory T cells
.
Lung Cancer
2012
;
75
:
95
101
.
13.
Al-Shibli
KI
,
Donnem
T
,
Al-Saad
S
,
Persson
M
,
Bremnes
RM
,
Busund
L-T
. 
Prognostic effect of epithelial and stromal lymphocyte infiltration in non-small cell lung cancer
.
Clin Cancer Res
2008
;
14
:
5220
7
.
14.
Zhou
J
,
Gong
Z
,
Jia
Q
,
Wu
Y
,
Yang
Z-Z
,
Zhu
B
. 
Programmed death ligand 1 expression and CD8+ tumor-infiltrating lymphocyte density differences between paired primary and brain metastatic lesions in non-small cell lung cancer
.
Biochem Biophys Res Commun
2018
;
498
:
751
7
.
15.
Müller
P
,
Rothschild
SI
,
Arnold
W
,
Hirschmann
P
,
Horvath
L
,
Bubendorf
L
, et al
Metastatic spread in patients with non-small cell lung cancer is associated with a reduced density of tumor-infiltrating T cells
.
Cancer Immunol Immunother
2016
;
65
:
1
11
.
16.
Lee
M
,
Heo
S-H
,
Song
IH
,
Rajayi
H
,
Park
HS
,
Park
IA
, et al
Presence of tertiary lymphoid structures determines the level of tumor-infiltrating lymphocytes in primary breast cancer and metastasis
.
Mod Pathol
2019
;
32
:
70
80
.
17.
Berghoff
AS
,
Fuchs
E
,
Ricken
G
,
Mlecnik
B
,
Bindea
G
,
Spanberger
T
, et al
Density of tumor-infiltrating lymphocytes correlates with extent of brain edema and overall survival time in patients with brain metastases
.
Oncoimmunology
2016
;
5
:
e1057388
9
.
18.
Berghoff
AS
,
Ricken
G
,
Widhalm
G
,
Rajky
O
,
Dieckmann
K
,
Birner
P
, et al
Tumour-infiltrating lymphocytes and expression of programmed death ligand 1 (PD-L1) in melanoma brain metastases
.
Histopathology
2015
;
66
:
289
99
.
19.
Harter
PN
,
Bernatz
S
,
Scholz
A
,
Zeiner
PS
,
Zinke
J
,
Kiyose
M
, et al
Distribution and prognostic relevance of tumor-infiltrating lymphocytes (TILs) and PD-1/PD-L1 immune checkpoints in human brain metastases
.
Oncotarget
2015
;
6
:
40836
49
.
20.
Jindal
V
,
Gupta
S
. 
Expected paradigm shift in brain metastases therapy—immune checkpoint inhibitors
.
Mol Neurobiol
2018
;
55
:
7072
8
.
21.
Galea
I
,
Bechmann
I
,
Perry
VH
. 
What is immune privilege (not)?
Trends Immunol
2007
;
28
:
12
8
.
22.
Michell-Robinson
MA
,
Touil
H
,
Healy
LM
,
Owen
DR
,
Durafourt
BA
,
Bar-Or
A
, et al
Roles of microglia in brain development, tissue maintenance and repair
.
Brain
2015
;
138
:
1138
59
.
23.
Gimsa
U
,
Mitchison
NA
,
Brunner-Weinzierl
MC
. 
Immune privilege as an intrinsic CNS property: astrocytes protect the CNS against T-cell-mediated neuroinflammation
.
Mediators Inflamm
2013
;
2013
:
320519
.
24.
Farber
SH
,
Tsvankin
V
,
Narloch
JL
,
Kim
GJ
,
Salama
AKS
,
Vlahovic
G
, et al
Embracing rejection: immunologic trends in brain metastasis
.
Oncoimmunology
2016
;
5
:
e1172153
.
25.
Popper
HH
. 
Progression and metastasis of lung cancer
.
Cancer Metastasis Rev
2016
;
35
:
75
91
.
26.
Zeiner
PS
,
Zinke
J
,
Kowalewski
DJ
,
Bernatz
S
,
Tichy
J
,
Ronellenfitsch
MW
, et al
CD74 regulates complexity of tumor cell HLA class II peptidome in brain metastasis and is a positive prognostic marker for patient survival
.
Acta Neuropathol Commun
2018
;
6
:
18
.
27.
Lohr
J
,
Ratliff
T
,
Huppertz
A
,
Ge
Y
,
Dictus
C
,
Ahmadi
R
, et al
Effector T-cell infiltration positively impacts survival of glioblastoma patients and is impaired by tumor-derived TGF-β
.
Clin Cancer Res
2011
;
17
:
4296
308
.
28.
Geisenberger
C
,
Mock
A
,
Warta
R
,
Rapp
C
,
Schwager
C
,
Korshunov
A
, et al
Molecular profiling of long-term survivors identifies a subgroup of glioblastoma characterized by chromosome 19/20 co-gain
.
Acta Neuropathol
2015
;
130
:
419
34
.
29.
Cancer Genome Atlas Research Network
. 
Comprehensive molecular profiling of lung adenocarcinoma
.
Nature
2014
;
511
:
543
50
.
30.
Thorsson
V
,
Gibbs
DL
,
Brown
SD
,
Wolf
D
,
Bortone
DS
,
Ou Yang
T-H
, et al
The immune landscape of cancer
.
Immunity
2018
;
48
:
812
4
.
31.
Lambrechts
D
,
Wauters
E
,
Boeckx
B
,
Aibar
S
,
Nittner
D
,
Burton
O
, et al
Phenotype molding of stromal cells in the lung tumor microenvironment
.
Nat Med
2018
;
24
:
1277
89
.
32.
Abbas
AR
,
Wolslegel
K
,
Seshasayee
D
,
Modrusan
Z
,
Clark
HF
. 
Deconvolution of blood microarray data identifies cellular activation patterns in systemic lupus erythematosus
.
PLoS One
2009
;
4
:
e6098
.
33.
Gaujoux
R
,
Seoighe
C
. 
CellMix: a comprehensive toolbox for gene expression deconvolution
.
Bioinformatics
2013
;
29
:
2211
2
.
34.
Geng
Y
,
Shao
Y
,
He
W
,
Hu
W
,
Xu
Y
,
Chen
J
, et al
Prognostic role of tumor-infiltrating lymphocytes in lung cancer: a meta-analysis
.
Cell Physiol Biochem
2015
;
37
:
1560
71
.
35.
Houseman
EA
,
Accomando
WP
,
Koestler
DC
,
Christensen
BC
,
Marsit
CJ
,
Nelson
HH
, et al
DNA methylation arrays as surrogate measures of cell mixture distribution
.
BMC Bioinformatics
2012
;
13
:
12
.
36.
Galon
J
,
Bruni
D
. 
Approaches to treat immune hot, altered and cold tumours with combination immunotherapies
.
Nat Rev Drug Discov
2019
;
410
:
1107
.
37.
Pabla
S
,
Conroy
JM
,
Nesline
MK
,
Glenn
ST
,
Papanicolau-Sengos
A
,
Burgher
B
, et al
Proliferative potential and resistance to immune checkpoint blockade in lung cancer patients
.
J Immunother Cancer
2019
;
7
:
27
.
38.
Punt
S
,
Langenhoff
JM
,
Putter
H
,
Fleuren
GJ
,
Gorter
A
,
Jordanova
ES
. 
The correlations between IL-17 vs. Th17 cells and cancer patient survival: a systematic review
.
Oncoimmunology
2015
;
4
:
e984547
.
39.
Mustafa
DAM
,
Pedrosa
RMSM
,
Smid
M
,
van der Weiden
M
,
de Weerd
V
,
Nigg
AL
, et al
T lymphocytes facilitate brain metastasis of breast cancer by inducing Guanylate-Binding Protein 1 expression
.
Acta Neuropathol
2018
;
135
:
581
99
.
40.
Zakaria
R
,
Platt-Higgins
A
,
Rathi
N
,
Radon
M
,
Das
S
,
Das
K
, et al
T-Cell densities in brain metastases are associated with patient survival times and diffusion tensor MRI changes
.
Cancer Res
2018
;
78
:
610
6
.
41.
Téglási
V
,
Reiniger
L
,
Fábián
K
,
Pipek
O
,
Csala
I
,
Bagó
AG
, et al
Evaluating the significance of density, localization, and PD-1/PD-L1 immunopositivity of mononuclear cells in the clinical course of lung adenocarcinoma patients with brain metastasis
.
Neuro Oncol
2017
;
19
:
1058
67
.
42.
Castaneda
CA
,
Castillo
M
,
Bernabe
LA
,
Sanchez
J
,
Casavilca
S
,
García-Corrochano
P
, et al
Impact of pathological features of brain metastases in prognosis
.
Biomark Med
2018
;
12
:
475
85
.
43.
Fecci
PE
,
Mitchell
DA
,
Whitesides
JF
,
Xie
W
,
Friedman
AH
,
Archer
GE
, et al
Increased regulatory T-cell fraction amidst a diminished CD4 compartment explains cellular immune defects in patients with malignant glioma
.
Cancer Res
2006
;
66
:
3294
302
.
44.
Shimizu
K
,
Nakata
M
,
Hirami
Y
,
Yukawa
T
,
Maeda
A
,
Tanemoto
K
. 
Tumor-infiltrating Foxp3+ regulatory T cells are correlated with cyclooxygenase-2 expression and are associated with recurrence in resected non-small cell lung cancer
.
J Thorac Oncol
2010
;
5
:
585
90
.
45.
Lee
S
,
Kim
D
,
Kang
J
,
Kim
E
,
Kim
W
,
Youn
H
, et al
Surfactant protein B suppresses lung cancer progression by inhibiting secretory phospholipase A2 activity and arachidonic acid production
.
Cell Physiol Biochem
2017
;
42
:
1684
700
.
46.
Mitsuhashi
A
,
Goto
H
,
Kuramoto
T
,
Tabata
S
,
Yukishige
S
,
Abe
S
, et al
Surfactant protein A suppresses lung cancer progression by regulating the polarization of tumor-associated macrophages
.
Am J Pathol
2013
;
182
:
1843
53
.
47.
De Masson
A
,
Giustiniani
J
,
Marie-Cardine
A
,
Bouaziz
JD
,
Dulphy
N
,
Gossot
D
, et al
Identification of CD245 as myosin 18A, a receptor for surfactant A: a novel pathway for activating human NK lymphocytes
.
Oncoimmunology
2016
;
5
:
e1127493
.
48.
Hasegawa
Y
,
Takahashi
M
,
Ariki
S
,
Saito
A
,
Uehara
Y
,
Takamiya
R
, et al
Surfactant protein A downregulates epidermal growth factor receptor by mechanisms different from those of surfactant protein D
.
J Biol Chem
2017
;
292
:
18565
76
.
49.
Zhou
L
,
Lv
X
,
Yang
J
,
Zhu
Y
,
Wang
Z
,
Xu
T
. 
Napsin A is negatively associated with EMT-mediated EGFR-TKI resistance in lung cancer cells
.
Mol Med Rep
2018
;
18
:
1247
52
.
50.
Peng
L
,
Bian
XW
,
Li
DK
,
Xu
C
,
Wang
GM
,
Xia
QY
, et al
Large-scale RNA-Seq transcriptome analysis of 4043 cancers and 548 normal tissue controls across 12 TCGA cancer types
.
Sci Rep
2015
;
5
:
13413
.
51.
Chan
TA
,
Yarchoan
M
,
Jaffee
E
,
Swanton
C
,
Quezada
SA
,
Stenzinger
A
, et al
Development of tumor mutation burden as an immunotherapy biomarker: utility for the oncology clinic
.
Ann Oncol
2019
;
30
:
44
56
.