Introduction
Lung squamous cell carcinoma (LUSC) is one of the most common histological subtypes of non-small-cell lung cancer (NSCLC), the remaining leading cause of cancer-related death worldwide for a long period [
1‐
3]. Unlike lung adenocarcinoma (LUAD) with oncogenic driver alterations, therapeutic progress for LUSC is limited and conventional platinum-based chemotherapy remains the standard-of-care for many years [
4‐
6]. Recently, immune checkpoint inhibitors targeted programmed cell death 1 (PD-1) and its ligand (PD-L1) has a shift the paradigm in both LUAD and LUSC. To date, several anti-PD-1/PD-L1 antibodies including nivolumab, pembrolizumab, and atezolizumab have been approved as second-line settings for patients with advanced NSCLC [
7‐
11]. Moreover, pembrolizumab monotherapy showed superiority as a first-line setting when compared with chemotherapy in patients with PD-L1 tumor proportion score (TPS) > 1%, and pembrolizumab plus platinum-based chemotherapy become the standard-of-care for first-line setting patients with LUSC [
11,
12]. Emerging evidence indicated that positive PD-L1 expression and high tumor mutation burden (TMB) could predict the response to anti-PD-1/PD-L1 therapies in NSCLC [
12,
13]. However, the ideal predictive biomarkers for immunotherapy are still an unmet need in clinical practice. Inspiringly, latest reports found gene profiling showed the potent to predict response to immune checkpoint inhibitors [
14‐
16].
For patients with early-stage LUSC, curative surgery with adjuvant chemotherapy is the main therapeutic option even though adjuvant chemotherapy just brought limited survival benefit [
17]. Recently, neoadjuvant and adjuvant immune checkpoint inhibitors showed promising results in this setting. A pilot study found that neoadjuvant nivolumab resulted in a major pathological response of 45% [
18]. Further biomarker analysis found that TMB was predictive of the pathological response to neoadjuvant nivolumab, while in LCMC3 study, TMB did not associate with pathological response for neoadjuvant atezolizumab and response was found in patients with PD-L1 expression negative [
19]. Therefore, comprehensively depicting the genomic and immune landscape, their correlations are needed to elucidate the ideal phenotype to benefit neo-adjuvant immunotherapy in patients with early-stage NSCLC. Although several studies have reported the relevant data on LUAD [
20,
21], the situation in LUSC is still largely unknown.
Herein, we performed this study in 189 Chinese patients with early-stage LUSC to evaluate (1) the genomic landscape of LUSC and its correlation with PD-L1 expression, TMB, and six immune infiltrates and (2) their associations with clinical parameters, disease-free survival (DFS), and overall survival (OS). Additionally, we also compared our data to other ethnicities and tumor types such as adenocarcinoma.
Methods
Sample collection
We retrospectively identified patients who underwent surgical resection of the lung (lobectomy or pulmonectomy) and histologically confirmed LUSC at Shanghai Pulmonary Hospital from 2012 to 2015. We firstly checked the histological types of each case using medical electronic records. Then all primary diagnoses were further independently evaluated by two experienced pathologists (Z.W.D and L.K.H) according to the WHO nomenclature for squamous carcinoma. The specimens of eligible case must have a confirmed diagnosis of LUSC and had at least 50% tumor cellularity. Major exclusion criteria were inadequate or poor quality samples, missing baseline clinicopathological features, mixed histology, and incomplete follow-up data. Corresponding formalin-fixed and paraffin-embedded (FFPE) tissues were used for immunohistochemistry (IHC) staining and whole-exome sequencing. The major baseline features including age, sex, smoking history, Eastern Cooperative Oncology Group performance status (ECOG PS), tumor size, node status and stage, vascular invasion, differentiation degree, tumor stage, DFS, and OS. A never-smoker was defined as a patient who had smoked less than 100 cigarettes during his/her lifetime. DFS was defined as the time from the initial surgical resection until recurrence. OS was calculated from the date of LUSC diagnosis to death from any cause or was censored at the last follow-up date. This study was conducted in accordance with the provisions of the Declaration of Helsinki and was approved by the ethics committee of Shanghai Pulmonary Hospital.
PD-L1 expression
PD-L1 expression was firstly tested by using anti-human PD-L1 (#13684, clone E1L3N, Cell Signaling Technology, Danvers, MA, diluted 1:200) according to the manufacturer’s recommendations and previous publications using 4–5 μm FFPE sections [
22,
23]. Then all of the samples were re-evaluated by using another antibody assay (clone SP142, Spring Bioscience, Ventana, Tucson, AZ, diluted 1:100) [
24]. For E1L3N staining, PD-L1 expression was defined as the percentage of tumor cells showing membranous immunoreactivity (central or marginal tumor region). The cutoff value was 5% for PD-L1 positivity or negativity (PD-L1
+/−). PD-L1 > 50% was defined as PD-L1 strong positivity. For SP142 staining, positive cells were defined as cancer cells displaying membranous staining for PD-L1, and the proportion of PD-L1
+ cells was evaluated as the percentage of total cancer cells in whole sections. The cutoff values of 1%, 5%, 10%, and 50% were set and 1% was determined for PD-L1
+/− [
25,
26]. Sections from human placenta tissues were used as the positive controls of PD-L1 IHC staining. Breast cancer cell line (MCF-7) was utilized as the negative control. All of the stained sections were independently reviewed by two pathologists (Z.W.D and L.K.H). Any discrepancies were discussed together and a consensus was achieved under the guidance of another experienced pathologist (C.Y.W).
CD8+ tumor-infiltrating lymphocyte (TIL) density
CD8
+ TILs density was assessed by using a mouse anti-CD8 monoclonal antibody (M7103, clone C8144B, DAKO). Lymphocytes with immunostained CD8 infiltrating within a tumor region (central or marginal) were defined as CD8
+ TILs. On the basis of the percentage of CD8
+ TILs displayed within a tumor region, we determined high/low CD8
+ TIL density (CD8
+ TIL
+/−) with a cutoff of 5%, which was analogous to the previous studies [
23,
27].
TMB calculation
The details of whole-exome sequencing and data processing were listed in Additional file
1. TMB was defined as the number of somatic, coding, base substitution, and indel mutations per megabase of genome examined by using nonsynonymous and frameshift indels at 5% limit of detection. Variants in low confidence regions and repeat regions, driver mutations and germline mutations were removed via using population online databases (The Exome Aggregation Consortium v.03, Genome Aggregation Database, and 1000 Genome) followed by variant allelic frequency cutoff of 0.2%. The tumor mutation calculation formula was as follows:
$$ \mathrm{TMB}=\frac{S_n\times 1000000}{N} $$
Sn represents the absolute number of somatic mutations, and N represents the number of exonic bases coverage depth ≥ 100 ×.
Estimation of immune cells infiltration
The abundances of six immune cell infiltrations including B cells, CD4
+ T cells, CD8
+ T cells, macrophages, neutrophils, and dendritic cells (DC) in specific groups of LUSC were estimated by using online database, named Tumor Immune Estimation Resource (TIMER). TIMER is a comprehensive resource for systematical analysis of immune cell infiltrations across diverse cancer types, which is validated using pathological estimations. The details and statistical methods were listed in this website (
https://cistrome.shinyapps.io/timer/) and their previous publications [
28,
29].
Statistical analysis
The Pearson correlation coefficient was used to determine the correlation of PD-L1 expression level between two antibody assays. Spearman’s rank correlation was utilized to assess the correlations among PD-L1 expression, TMB, and CD8+ TIL density. Correlations between PD-L1 expression/CD8+ TIL density and clinical parameters were analyzed using the chi-squared or Fisher’s exact test for categorical variables. The continuous variable was analyzed by ANOVA and Tukey’s multiple comparison tests. Mann-Whitney U tests or Kruskal-Wallis rank sum tests were used for comparisons of continuous variables across multiple groups. The Kaplan-Meier curve was leveraged to assess the patients’ survival curves. The log-rank test was used to test the significance of differences between two or four groups. Cox proportional hazards model was utilized for uni- and multivariate survival analyses to calculate the hazard ratios (HR) and related 95% confidence intervals (CI). P < 0.05 was considered significant. All statistical analyses were performed using GraphPad PRISM 6.0 and the SPSS statistical software, version 22.0 (SPSS Inc., Chicago, IL, USA).
Discussion
Whole-exome sequencing of our cohort identified significant somatic mutations and CNVs that was consistent with previous publications on the genomic profile of early-stage LUSC [
30,
31,
34,
35]. However,
SOX2 amplification was only found in 6.0% of all cases, which was significantly lower than other studies [
30,
34]. Importantly, we identified a high frequency of
EML4-ALK fusion (3.2%), one of the most common driver alterations in LUAD, which might mainly due to a high rate of never-smoker in our cohort, and infer the necessity to detect the common driver gene alterations in never-smokers with LUSC.
LUSC was previously found to have higher TMB than other solid tumors due to the close correlation to tobacco exposure [
31,
36,
37]. However, the present study did not observe the association between smoking history and TMB level. In line with our results, Tatsuro et al. also did not find a correlation between the amount of smoking and TMB [
35]. Furthermore, although the current cohort involved a large proportion of patients with never smoking, the mean and median TMB was similar to the results of TCGA and CHOICE study [
30,
31]. These findings suggested that the relationship between tobacco exposure and TMB still remains further investigation. Intriguingly, patients with CNVs had significantly higher TMB than those without. This were reminiscent of an elegant study that examined the data from 5255 tumor/normal samples representing 12 tumor types from TCGA and found a positive correlation between somatic CNVs level and the total number of mutations [
38], suggesting the potential value of CNVs for predicting the TMB level and its application for predicting who are most likely to benefit from immunotherapy. Of note, some kinds of CNVs, such as
FGFR1,
EGFR, and
PIK3CA amplifications and loss of
CDKN2A, were associated with significantly lower six immune infiltrates. This finding could partly explain that the fraction of copy number altered genome was highest in NSCLC patients treated with anti-PD-1/PD-L1 therapy but lack of durable benefit due to the importance of these immune infiltrates in antitumor immune response [
14]. It also indicated that distinct kinds of CNVs would have a different effect on the immune response.
Understanding the interplay between molecular underpinnings and immune landscape may help improve strategies for precise immunotherapy [
14,
39]. Towards this aim, we investigated the associations between frequent genomic alterations and PD-L1 expression, TMB, or CD8
+ TIL density. The results showed that only
KEAP1 mutation was significantly associated with lower CD8
+ TIL density, and
NFE2L2 mutation was associated with marginally higher TMB. Moreover, a recent study reported that there was a higher rate of
KEAP1/NFE2L2 >mutations in Chinese LUSC than those in Western populations [
31]. KEAP1-NFE2L2 interaction plays a significant role in the dysregulation of oxidative stress pathway in lung cancer [
40]. Genetic alterations of
KEAP1 or
NFE2L2 would destroy this process and lead to oncogenesis and drug and radio resistance in different types of solid tumors. Considering these findings, we could infer that tumor with
KEAP1 or
NFE2L2 mutation would have a higher level of oxidative stress, which could lead to the destruction of immune cells including CD8
+ TILs and increased DNA damage level, resulting in the increase of somatic mutations of tumor cells. We also used the online database to calculate the abundance of tumor-infiltrating immune cells [
28,
29]. Consistent with our IHC results, we observed that
KEAP1 mutation had the negative effect on CD8
+ T cell abundance.
KEAP1 mutation was reported to be associated with poor response to adjuvant chemotherapy in both LUSC and LUAD [
19,
34,
41]. Whether the relationship between
KEAP1 mutation and lower CD8
+ T cell infiltration could explain the negative predictive value on adjuvant chemotherapy warrants further examinations.
Currently, TMB and PD-L1 expression are the two developed predictive biomarkers for anti-PD-1/PD-L1 therapy. We found no association between TMB and PD-L1 expression, which was consistent with previous findings. Yu et al. reported that PD-L1 protein expression was not correlated with TMB in both tumor cells and tumor-infiltrating immune cells of early-stage LUSC [
32]. Rizvi et al. also found that TMB did not correlate with PD-L1 expression in patients with NSCLC treated with anti-PD-1/PD-L1 therapy [
14]. Moreover, our study indicated that TMB did not correlate with CD8
+ TIL density but a significant positive correlation was found between CD8
+ TIL density and PD-L1 expression. Similar to previous reports [
42,
43], positive CD8
+ TIL density and lower TMB were independently associated with significantly longer DFS even though none of TMB, PD-L1 expression, or CD8
+ TIL density were associated with OS. The combination of TMB and PD-L1 expression or CD8
+ TIL density showed an improved yield in stratifying patients with discrepant DFS and OS, suggesting that the incorporation of these biomarkers into multivariable predictive and prognostic models worth further investigation in the future.
Furthermore, we classified the TIME into four types based on PD-L1 expression and CD8
+ TIL density. Our data indicated that type IV TIME was dominant while type III was the least common in Chinese early-stage LUSC, which is in contrast to the previous studies [
21,
44]. The potential reasons may include the different populations, testing methods and platforms, together with the discrepancy between the online RNA-seq and IHC staining data. As we know, type I TIME is most likely to benefit from anti-PD-1/PD-L1 monotherapy for the reason that these tumors have evidence of pre-existing intratumor T cells that are silenced by PD-L1 engagement. Constantly, the proportion of type I TIME in our cohort was similar to the response rate of patients with LUSC received single-agent anti-PD-1/PD-L1 checkpoint blockade [
7‐
9]. We also found that different types of TIME had distinct genetic alterations but similar TMB, clinical variables, and prognosis, suggesting the unique shaping function of genomic landscape on immune phenotypes.
Additionally, we observed that high TMB was correlated with significantly longer DFS in never-smoker but not associated with DFS in former/current smoker. CD8+ TIL+ was not associated with DFS in never-smoker but associated with markedly longer DFS in former/current smoker. Both TMB level and CD8+ TIL+ were not associated with OS. PD-L1+ was not correlated with both DFS and OS in two groups. NSCLC of never-smokers are entirely different from these of former/current smokers. In never-smoking group, high TMB in tumors with never-smoking would mainly come from the intrinsic mechanism, which means the high possibility to be recognized and eliminated by the immune system. In this process, more and more CD8+ TILs would be induced and translated into the exhausted phenotype. Hence, high TMB was correlated with significantly longer DFS but CD8+ TIL+ was not associated with DFS in never-smoker. Conversely, high TMB in smokers caused by tobacco exposure could not predict the clinical outcome. However, CD8+ TILs in smokers could recognize and eliminate the malignant cells after exposure to the different carcinogens. Hence, high CD8+ TIL density was found to associate with markedly longer DFS in the current study. We must mention that the relatively small sample size and multiple different subsets based on distinct cutoffs of TMB, PD-L1 expression, CD8+ TIL density, and/or smoking status for these exploratory analyses would also be the potential reason for these discrepancies. Future investigations with a larger number of patients with LUSC are still needed.