Background
Brain metastases have become the leading cause of death in patients with non-small cell lung cancer (NSCLC), occurring approximately in 30–50% of patients with NSCLC [
1]. Furthermore, with the ageing of the population, rise of people’s living standards, and improvements in imaging technology, an increasing number of individuals are diagnosed with brain metastases. The prognosis of patients with NSCLC brain metastases is dismal, with a median survival of 7 months [
2]. Therapeutic methods for patients with NSCLC brain metastases are limited. Standard treatments include surgical resection, whole-brain radiotherapy, and stereotactic radiosurgery. However, despite active treatment, the prognosis remains poor [
3].
Cancer immunotherapy has advanced significantly in recent years. A reciprocal effect is observed between the human immune system and cancer cells, and regulating immune responses at tumor sites is a crucial mechanism for tumor immune evasion [
4,
5]. The cytotoxic T-lymphocyte antigen 4 (CTLA-4) and programmed cell death protein 1 (PD-1)/programmed death ligand 1 (PD-L1) pathways, also known as an immune checkpoint, are recognized as an essential immunosuppressive mechanism in various tumors [
6,
7]. Moreover, results from multiple clinical trials have confirmed the significant therapeutic effects of anti-CTLA-4 and anti-PD-1/PD-L1 inhibitors on NSCLC [
8‐
12]. The discovery and clinical application of immune checkpoint inhibitors is a new direction in the treatment of NSCLC.
However, using immune checkpoint inhibitors have several issues. First, the use of immune checkpoint inhibitors is only effective for particular patients. Not all patients express CTLA-4 and PD-1/PD-L1 in their lesions, rendering immunotherapy less effective [
13,
14]. In addition, for patients with NSCLC brain metastases, owing to the existence of the blood-brain barrier (BBB) and the distinctive immune environment of the brain, heterogeneity may exist in the tumor immune microenvironment between the primary tumor and metastases [
15]. These reasons may lead to different therapeutic effects of immune checkpoint inhibitors on primary tumors and metastases.
To address these issues, we turned our attention to several other immune checkpoints than CTLA-4 and PD-1/PD-L1 that may be useful for treatment, and we compared the differences in the tumor immune microenvironment between primary and metastatic lesions in patients with NSCLC brain metastases to find therapeutic methods that benefit the patients.
B7-H3 and B7-H4 in the B7 family, indoleamine 2,3-dioxygenase 1 (IDO1), and EphA2 play important roles in immune suppression and immune escape, preventing the proliferation and activation of T cells. EphA2 also promotes the proliferation and migration of malignant tumor cells [
16‐
23]. Therefore, B7-H3, B7-H4, IDO1, and EphA2 may be potential targets for immunotherapy. However, only a few studies are available on these molecules in patients with brain metastases from NSCLC. Knowledge about their expression levels in brain metastases lesions and whether their interaction with tumor immune microenvironment is related to the prognosis of NSCLC brain metastases patients is inadequate.
Herein, immunohistochemical investigation was performed on resected primary NSCLC and its brain metastases to analyze the differences in the tumor immune microenvironment. We also studied whether these differences were affected by the time interval of metastasis. In addition, survival analyzes were performed in lung adenocarcinoma brain metastases patients.
Materials and methods
Patients and tissue microarrays
Samples were obtained from the Department of Pathology, Zhongnan Hospital of Wuhan University, involving the patients with NSCLC brain metastases who underwent tumor resection at the Department of Neurosurgery and Department of Thoracic Surgery, Zhongnan Hospital of Wuhan University, between January 2016 and March 2021. The inclusion criteria for cases are as follows: (1) Aged between 18 and 80 years old; (2) Brain lesions confirmed by histopathology to be brain metastases from NSCLC (including lung adenocarcinoma, lung squamous cell carcinoma and lung large cell carcinoma); (3) Available formalin-fixed and paraffin-embedded (FFPE) tumor tissue. Patients who had received other treatments for brain metastasis before surgery and those with a history of other malignancies were excluded. The clinicopathological information of patients was collected from electronic records and pathology reports. Finally, we collected FFPE primary NSCLC tumor samples (n = 17), their paired brain metastases tissue (n = 17), and unpaired brain metastases tissue (n = 45) from 62 NSCLC patients with brain metastases. Among them, lung cancer samples and brain metastasis samples taken from the same patient are defined as paired sample, seventeen paired samples are from 17 different patients, and 45 unpaired brain metastasis tissues are from another 45 patients with NSCLC brain metastases. Waived consent was obtained from all patients. The study was approved by the Ethics Committee, Zhongnan Hospital of Wuhan University (ethics No. 2,019,048).
Tissue microarrays were performed following standard methods [
24]. In brief, tumor core regions were marked on slides stained with hematoxylin and eosin (H&E) and prepared for tissue microarrays construction. A 1-mm tumor core region was selected with a needle from surgically resected FFPE tumor samples as representative tumor regions, and the removed tumor core regions were arrayed in blank recipient paraffin blocks.
Immunohistochemistry staining (IHC)
Four-micrometer-thick FFPE serial sections were obtained from the tissue microarrays blocks. IHC staining of B7-H3, B7-H4, IDO1, and EphA2 was conducted as follows: The sections were dewaxed in xylene and rehydrated in alcohol. Thereafter, these were put in a boiling pressure cooker with EDTA antigen repair buffer for 150 s for antigen retrieval. Next, sections were incubated with 3.0% hydrogen peroxide solution for 15 min to block endogenous peroxidase activity. Then, sections were washed and blocked for 30 min with 5% goat serum. The primary antibody was added dropwise to the sections, and the sections were placed at 4 °C and incubated overnight. The sections were then incubated with the secondary antibody at room 37 °C for 30 min. Diaminobenzidine (DAB) was used for the chromogenic reaction. Finally, sections were counterstained with hematoxylin, and blue reaction was performed with 0.02% ammonia water, followed by rinsing with water, dehydrating with graded alcohol, and fixing with neutral gum sealing slides. In addition, immunohistochemical staining of CTLA-4, PD-1, PD-L1, CD3, CD4, CD8, CD20, CD68, CD163 and Ki-67 was performed on a Leica Bond Max automated stainer (Leica) according to the manufacturer’s protocol. Human tonsil tissue was included as a positive control. Use PBS rather than primary antibody as negative control. All samples were stained in one run.
The primary antibodies used for immunohistochemistry are as follows: B7-H3 (1:200 dilution, D9M2L, Cell Signaling Technology), B7-H4 (1:150 dilution, D1M8I, Cell Signaling Technology), IDO1 (1:200 dilution, SP260, Abcam), EphA2 (1:100 dilution, SC-398,832, Santa Cruz Biotechnology), CTLA-4 (UMAB249, ready-to-use, Zhongshan Golden Bridge Biotechnology), PD-1 (UMAB199, ready-to-use, Zhongshan Golden Bridge Biotechnology), PD-L1 (1:200 dilution, E1L3N, Cell Signaling Technology), Ki-67 (MIB1, ready-to-use, Zhongshan Golden Bridge Biotechnology), CD3 (F7.2.38, prediluted, Dako), CD4 (4B12, prediluted, Dako), CD8 (C8/144B, prediluted, Dako), CD20 (L26, prediluted, Leica), CD68 (KP1, prediluted, Dako) and CD163 (MRQ-26, prediluted, Dako). Anti-rabbit and anti-mouse secondary antibodies were acquired from Zhongshan Golden Bridge Biotechnology (Beijing, China).
Multiplex immunofluorescence staining
As described in a previous report [
25], multiplex immunofluorescence staining was performed using the Opal 7-Color IHC Kit (PerkinElmer, Waltham, MA, USA) in the FFPE tissue sections of a typically paired NSCLC and its brain metastases. The stained slides were scanned by a Vectra 3.0 multispectral imaging system (PerkinElmer, Waltham, MA, USA). The immunofluorescence markers consisted of B7-H4 (1:150 dilution, clone D1M8I, Cell Signaling Technology), CD3 (F7.2.38, prediluted, Dako), CD8 (C8/144B, prediluted, Dako), CD20 (L26, ready-to-use, Leica), CD68 (KP1, prediluted, Dako), and CK (AE1/AE3, prediluted, Dako). DAPI was used for nuclei highlighting. After dewaxing and rehydration, antigen retrievals were performed using a Meidi microwave (Meidi, China). For each primary antibody, tyramide signal amplification linked to specific fluorochrome from the multiplex immunofluorescence staining Kit was used for incubation and visualization. We performed the complete multiplex immunofluorescence staining procedure following the manufacturer’s instructions. In addition, human tonsil FFPE tissues were analyzed with and without primary antibodies following the same multiplex immunofluorescence staining procedure to establish positive and negative (autofluorescence) controls.
IHC expression scoring
The immunohistochemical staining results were assessed using semiquantitative methods by two trained pathologists blinded to the patients’ clinical data. As no standard scoring system is available for the immunohistochemical staining results of these antibodies, we assessed each sample in terms of the degree of staining and proportion of positive tumor cells. Based on the degree of staining, they were classified into four semiquantitative groups: no positive staining (score 0), weak (score 1), moderate (score 2), and strong (score 3). The proportion of tumor cells stained was also classified into four categories: ≤25% (score 1), 26-50% (score 2), 51–75% (score 3), and > 75% (score 4). Next, the two scores were multiplied to obtain the final immunohistochemical staining score. The median of the final score was considered as the cutoff value to decide high and low expression classification. PD-L1 expression was assessed by tumor proportion score (TPS), and PD-L1 TPS ≥ 1% was considered positive, consistent with other studies [
26,
27]. The number of TILs and TAMs was counted and averaged over three high-powered fields [
28], and the median of the counts was considered as the cutoff value to decide high and low expression classification (since CTLA-4 and PD-1 are also expressed in lymphocytes, we also used this method for counting them). We selected 50% as the cutoff value for the Ki-67 labeling index (Supplementary Fig.
1 shows the typical immunohistochemical images of Ki-67). After careful review and discussion of all the slides by two pathologists, each slide was scored consistently. Specific results regarding scoring can be found in Supplementary Table
1.
Statistical analysis
As described in previous studies, the expression of categorical variables between matched lesions was evaluated using agreement statistics (κ coefficient) [
28,
29]. Regarding continuous variables, the normality was evaluated using the Shapiro–Wilk normality test. Paired Student’s t-test or Wilcoxon matched-pairs signed rank test was used to assess the significance of differences between paired lesions. Mann-Whitney U Test, Student’s T-Test, Fisher Test, or Chi-Square test were used to analyze the correlation between biomarkers and clinicopathological features. The Kaplan–Meier method was used for survival analysis. Cox proportional hazards model was used to determine independent prognostic variables. Overall survival (OS) was measured from the date of brain metastases lesion excision until death or the last follow-up.
P-value < 0.05 was considered statistically significant. All statistical analyses were performed using GraphPad Prism version 8.0.1, SPSS version 26 and R version 3.6.3.
Discussion
Herein, using immunohistochemistry, we discovered that the tumor microenvironment of primary NSCLC and brain metastases has both spatial and temporal heterogeneity. Spatially, the expression of some immune checkpoints was inconsistent. Besides, the infiltration of lymphocytes in brain metastases was significantly reduced compared with primary tumors, suggesting a more immunosuppressed microenvironment. Temporally, in paired lesions, CD8+ lymphocytes were more significantly suppressed in synchronous metastases. In brain metastases lesions, the positive expression of PD-L1 was significantly associated with synchronous metastases. In addition, we also explored the factors related to the prognosis of patients with lung adenocarcinoma brain metastases, and the results revealed the age, expression of B7-H4 and IDO1, and infiltration of CD68+ TAM were related to the prognosis. The expression of B7-H4 and the infiltration of CD68+ TAM are independent prognostic factors for lung adenocarcinoma brain metastasis patients.
The brain has been considered immune-privileged for an extended period because of the BBB. In recent years, however, many researchers have mentioned that the brain is immunologically unique instead of privileged since the discovery of functional lymphatic vessels within the central nervous system, and the immune cell infiltration of brain tumors is rare, but it occurs [
15]. This echoes our findings on the spatial heterogeneity of primary NSCLC and brain metastases. We found that the TILs density in brain metastases was significantly lower than in primary tumors, which is consistent with the findings of some previous groups [
37,
38]. These results partly explain why some immune checkpoint inhibitors exhibit different therapeutic responses in the intracranial and extracranial regions in patients with NSCLC brain metastases [
39]. Clinicians should fully consider this before applying immunotherapy.
Only a few studies have used the time interval between lung cancer and brain metastasis as a subgroup to analyze differences in the tumor microenvironment. We found that in brain metastasis tissue, PD-L1 positivity was more common in synchronous metastases than in metachronous metastases. This is similar to the findings of Lee et al., who performed IHC of PD-L1 on primary lung cancer tissues of 270 patients with NSCLC brain metastases, and the results showed that compared with PD-L1-negative patients, synchronous brain metastases were more frequently observed in PD-L1-positive patients [
40]. This indicates that PD-L1 expression is strongly correlated with synchronous metastasis whether detected in primary lung tumors or brain metastases. This finding is interesting. Is PD-L1 related to the activation of metastasis of NSCLC to the brain? The reason is still unclear. The mechanism of lung cancer brain metastasis has been reported before, including the “hemodynamic hypotheses” and “seed-and-soil hypotheses” [
41,
42]. In our clinical observations, the time interval for lung cancer to metastasize to the brain considerably varies greatly. This may be related to different initiation times of metastasis, impediments in the metastasis process, or the length of latency of tumor cells in the brain. Whether these processes are affected by PD-L1 expression remains to be confirmed by further research. We also noticed that previous literature has studied the relationship between the time interval of metastasis and the tumor microenvironment (including the expression of PD-L1) in paired NSCLC and brain metastases resected lesions. Most research results show that compared with metachronous metastasis, the tumor microenvironment of the primary tumor and brain metastasis tumor of synchronous brain metastasis is more consistent and less difference [
28,
43‐
45]. Only one literature reported different results [
46]. This finding has important clinical significance. For synchronous NSCLC brain metastases, the more common treatment is surgical resection of the brain metastases without surgical intervention of the primary tumor. Because for advanced NSCLC with distant metastasis, surgical resection of the primary tumor is not considered an effective treatment method [
47], which leads to a result: we can only perform pathological analysis on brain metastasis tissue, such as the expression of CTLA-4, PD-1, PD-L1, etc., rather than lung. Determining the consistency of tumor microenvironment expression between primary tumors and brain metastases can help clinicians decided whether patients are suitable for subsequent immunotherapy.
Previous studies have shown that B7-H4 and IDO1 have predictive significance in NSCLC [
30,
33]. However, we know little about their relationship with the prognosis of NSCLC brain metastasis patients. We performed survival analyzes in lung adenocarcinoma brain metastases patients. The result showed patients with high B7-H4 and IDO1 expression in brain metastases had a shorter survival time (log-rank test), and B7-H4 was an independent prognostic factor in lung adenocarcinoma brain metastases patients. We noticed that a previous study also reported the effect of B7-H4 on overall survival in NSCLC brain metastasis [
48]. Although the classification criteria for high B7-H4 expression in this study were different, a consistent conclusion was reached, which indicates that B7-H4 is still associated with disease prognosis even at different cutoff values and is a meaningful predictor of prognosis in NSCLC brain metastases patients. As far as we know, this study is the first to investigate the prognostic value of IDO1 in lung adenocarcinoma brain metastases. Although the final results show that IDO1 is not an independent prognostic factor, the widespread expression of IDO1 in brain metastases and its inhibitory effect on immune cells may make it a potential therapeutic target. Phase I/II clinical trials of epacadostat, an IDO1 inhibitor, have been reported positively in many advanced solid tumors [
49,
50]. Despite a recent phase III ECHO-301 trial (NCT02752074) of PD-1 inhibitor in combination with an IDO1 inhibitor in metastatic melanoma showed no significant clinical benefit in the treatment group (pembrolizumab + epacadostat) compared with the control group (pembrolizumab + placebo) [
51]. Some immuno-oncologists still believe that IDO1 is not a “bad target,” and the IDO1-targeting therapy is still meaningful. Vernon K. Sondak et al. explained this phenomenon to be dose-related [
52]. However, the mechanism of IDO1 immunosuppression and specific biomarkers that can respond to IDO inhibitors need to be explored in future research.
The findings of the current study revealed that CD68
+ TAM infiltration have prognostic significance in lung adenocarcinoma brain metastases. Recently, TAMs have become a hotspot in tumor immunotherapy research. According to the traditional classification, TAMs are usually divided into M1 and M2 phenotypes. M1 phenotype is considered to have pro-inflammatory and anti-tumor effects, and M2 phenotype is considered to have angiogenic and pro-tumor effects [
53]. Previous research has indicated that most immune cells within primary brain tumors are macrophages, comprising approximately 30% of the tumor mass. Compared with extracranial organs, the brain as an immunologically unique organ has two sources of TAMs: brain resident microglia and bone marrow-derived macrophages, both of which can be labeled by CD68, and the latter are risen by circulating monocytes recruited into the brain in pathological conditions such as tumors [
54‐
56]. Up to now, there is still no unified standard for how to distinguish brain resident microglia and bone marrow-derived macrophages. In this study, the high density of CD68
+ TAM infiltration in brain lesions was confirmed to be a poor prognostic factor in patients with lung adenocarcinoma brain metastases, whereas CD163-labeled M2 macrophage infiltration and CD163/CD68 ratio were not associated with prognosis. This may be because brain resident microglia are contained within CD68-positive TAM. Previous study has shown that brain resident microglia, and not peripheral macrophages, are the main source of brain tumor mononuclear cells [
57]. Interactions between microglia and T cells can promote brain cancer heterogeneity and immunosuppression [
58]. This may partially explain our result. In addition, the co-expression of M1 and M2 markers and M1 and M2 phenotype switching is also existing in brain tumors [
59,
60]. The immune microenvironment in brain tumors is a dynamic process. In our study, we can only confirm the status of TAM at the moment of surgical resection, and whether there is a phenotype switching between M1 and M2 after surgery is unknown. In summary, targeting intracranial microglia or M2 macrophages may be a potential treatment modality for lung adenocarcinoma brain metastases patients and more studies are needed to confirm this.
Although we did not find a significant correlation between the prognosis of patients with brain metastatic NSCLC and the expression of PD-1, PD-L1 and B7-H3 in the current study, previous studies have confirmed that PD-1, PD-L1 and B7-H3 play an important role in primary brain tumors, including glioblastoma and primary central nervous system lymphoma, and are associated with patient prognosis [
61‐
63]. This shows that these tumor immune checkpoints still interact with the surrounding immune microenvironment in the brain, a special immune-privileged organ [
15], and thus exert corresponding functions. Although brain metastatic NSCLC is a tumor of extracranial origin, it eventually colonizes and proliferates in the brain, and also has the expression of PD-1, PD-L1 and B7-H3. The specific mechanism remains to be explored. In the future, it may be meaningful to include primary brain tumors and metastatic brain tumors in one study and analyze the expression of these markers and their relationship with the immune microenvironment.
Despite providing some valuable insights, this study has several limitations. First, the small number of paired samples is a major limitation of our study. It is a single-center study and not all patients have undergone primary lesion resection at Zhongnan Hospital of Wuhan University. In addition, some patients only received resection of brain metastases. For the primary lung cancer, they did not receive surgical intervention but chose radiotherapy and\or chemotherapy. This resulted in our inability to obtain primary tumor tissue of these patients. Second, the clinical data of some patients were incomplete, affecting subsequent analysis. Third, the tumor core tissue used to make the tissue microarrays may not accurately represent the expression of markers in the panorama of tumor tissue. However, the results from the tissue microarrays measuring the expression of immune markers in several studies were consistent with each other, which justifies this approach.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.