Patient clinical characteristics
The tumor tissue samples of 155 NSCLC patients were collected retrospectively. The detailed baseline characteristics are shown in Table 1. Among all NSCLC patients, 128 patients (82.6%) had LUAD, two patients (1.3%) had LUSC, and 25 patients (16.1%) had other types of NSCLC. All patients had gene translocation events in this study. There were 84 patients (54.2%) with ALK rearrangement, 38 patients (24.5%) with RET rearrangement, and 33 patients (21.3%) with ROS1 rearrangement. The median age of this cohort was 55.0 years, ranging from 29.0 to 86.0 years. A total of 104 patients (67.1%) were females, and 51 patients (32.9%) were males. A total of 122 patients (78.7%) had no smoking history, these proportions were higher in the RET fusion-positive subgroup, which were 81.6% (Table 1). The patients in this cohort were characterized by being younger, being female, having adenocarcinoma, and having no history of smoking, consistent with the classic features of patients with ALK fusions. These data were all consistent with previous studies.3 The clinical features of RET and ROS1 rearrangements are largely consistent with those of ALK rearrangement, so those three fusion types were subsequently combined into one category for further analysis.
TP53 and CDKN2A/B were the most common covariation genes in ALK/RET/ROS1-rearrangement samples
In this NSCLC cohort with ALK/RET/ROS1 rearrangement mutations in all samples, the most frequently observed covariations were TP53 (31%), CDKN2A/B (15%), ARID1A (7%), PTPRD (6%), SETD2 (5%), and MDM2 (5%). EGFR mutations were mutually exclusive with ROS1 fusions and had a low mutation frequency (3%) in ALK/RET fusion patients (Fig. 1A). The mutation rates of TP53 in different ALK fusion variants (V1, V2, and V3a/b) were similar. However, CDKN2A/B had the highest and most similar mutation frequency in patients with ALK V1 and V3a/b variants (Fig. 1B).
When the genomic alterations were split into SNVs and CNVs, two parts were identified. TP53 was the most frequent co-occurring gene in the SNV part (Fig. 1C). In comparison, CDKN2A/B was the most frequent co-occurring gene in the CNV part, while all variants showed loss of copy number (Fig. 1D). In detail, the top 10 co-occurring SNV alterations at all sample levels were TP53 (31.6%), ARID1A (8.4%), BRIP1 (7.7%), SETD2 (5.8%), HMCN1 (5.8%), SMAD4 (4.5%), EP300 (4.5%), ATRX (4.5%), PTEN (3.8%), and KMT2C (3.8%) (Fig. 1C). The top 10 CNVs were CDKN2A/B (loss, 13.5%), PAK7 (loss, 3.8%), PTPRD (loss, 3.8%), MDM2 (gain, 3.2%), RAC1 (gain, 2.6%), PIK3C3 (loss, 2.6%), PAX5 (loss, 2.6%), TERT (gain, 1.9%), SMAD2 (loss, 1.9%), and SDHC (gain, 1.9%) (Fig. 1D). The genes with the highest mutation frequencies of SNVs and CNVs in the cohort data were TP53 and CDKN2A/B, respectively. According to the mutation statuses of TP53 and CDKN2A/B, the patients were divided into 3 subgroups: TP53 mutation and fusion co-occurring group, CDKN2A/B variation and fusion co-occurring group, TP53/CDKN2A/B variation and fusion co-occurring group and their corresponding wild-type (WT) group.
Figure 1E and 1F show the differences in the clinicopathological characteristics of patients among the groups. The stages of cancer in the CDKN2A/B co-occurring group were significantly different from those in the WT group (p = 0.0271). The CDKN2A/B co-occurring group had more patients with stage IV disease (58.33% [n = 14/24] vs. 33.59% [n = 14/131]) than the CDKN2A/B WT group. In contrast, the incidence of stage I disease was significantly lower in the CDKN2A/B co-occurring group than in the CDKN2A/B WT group (0% vs. 20.61% [n = 27/131]). The same trends were also found in the TP53 or CDKN2A/B co-occurring group and TP53/CDKN2A/B WT group (p = 0.0066). Compared with all other stages, the proportion of patients with stage IV disease was much higher in the TP53 or CDKN2A/B co-occurring group (52.54%, n = 31/59), followed by stage III, accounting for 25.42% (n = 15/59), while the proportion of patients with stage IV disease was only 28.12% (n = 27/96) in the TP53/CDKN2A/B WT group (Fig. 1E). The sex of the patients was significantly different between the co-occurring and WT groups. The TP53 co-occurring group had more males than the TP53 WT group (56.25% [n = 27/48] vs. 22.43% [n = 24/107], p < 0.001), similar to the TP53 or CDKN2A/B co-occurring group compared with the control group (53.54% [n = 31/59] vs. 20.83% [n = 20/96], p < 0.001) (Fig. 1F).
Other clinical characteristics, including smoking, age, and cancer type, differed between the groups, as shown in Supplementary Fig. 1. These data showed that the clinical characteristics of NSCLC patients with covariation were predominantly male sex and stage IV compared with wild-type NSCLC patients. For the following reason, we combined ALK/RET/ROS1 fusions in the next analysis. The reason was that although the fusions occurred in different genes, such as RET, and ROS1, the clinical characteristics of the included patients were well distributed and consistent with the typical clinical features of ALK fusions.
TMB and the number of neoantigens were higher in samples with TP53 or CDKN2A/B covariations
As shown in Fig. 2A, the samples in the 3 co-occurring groups had higher TMB than the samples in their corresponding WT groups (p < 0.001, p = 0.017, and p < 0.001). Further analysis revealed that the number of neoantigens was higher in the TP53 co-occurring group, as well as in the TP53/CDKN2A/B co-occurring group (p = 0.014 and p = 0.0032). The CDKN2A/B co-occurring group also had more neoantigens, although the difference was not statistically significant (p = 0.31) (Fig. 2B). Even in the HLA-A and HLA-B subgroups, there were no differences between the co-occurring and WT groups, and HLA-A*03 was a prevalent HLA type present in those groups.
TP53 or CDKN2A/B covariations were associated with an immune microenvironment characterized by high PD-L1 expression and low CD8 + TIL infiltration
A previous study reported that TP53/ALK comutation led to a poorer prognosis.8 To investigate the possible reasons for this phenomenon, fluorescent mIHC with classic immune markers (CD8, CD68, PD-1, and PD-L1) was performed on 50 samples from 155 NSCLC patients harboring ALK/RET/ROS1 rearrangements, and then the differences in TIL infiltration levels were analyzed. Representative images of classic markers on tumor tissues from patient A (co-occurring group) and patient B (WT group) are shown in Fig. 3A. The lower image shows the fluorescence map of the double marker (CD8&PD-1 and CD8&PD-L1) field of view on the mIHC technology platform. The percentage of all markers was calculated separately from the tumor area, stromal area and all areas of tumor tissue. The results of these three kinds of areas were obtained separately, and the p values are presented in Supplementary Table 1.
We compared the expression levels of classic markers of the immune microenvironment in 3 regions (tumor, stroma, and all regions) in the TP53/CDKN2A/B co-occurring and WT groups. The co-occurring group had significantly less infiltration of CD8+ (p = 0.043), CD8+PD-1− (p = 0.029), and CD8+PD-L1− (p = 0.025) TILs than the WT group (Fig. 3B). In stromal areas, the level of CD8+PD-L1+ cells was significantly increased in the covariation group compared with the WT group (p = 0.021). There were slightly higher trends of CD8+, CD68+, CD8+PD1−, and CD68+PD-L1− infiltration in the covariation group (Fig. 3C). In the total areas, none of the markers except PD-L1 showed statistically significant differences between the two groups, and only some markers, such as CD8+ and CD8+PD-L1−, showed a decreasing trend in the covariation group (Fig. 3D). For the comparison of the TP53 and CDKN2A/B co-occurring groups with the WT group, similar trends were also found (Supplementary Fig. 2).
Notably, the level of PD-L1+ cells was higher in the covariation group than in the WT group in tumor areas (p = 0.00038), stromal areas (p = 0.0016), and total areas (p = 0.00064) (Fig. 3E). To reverse verify PD-L1+ expression in the two groups, the expression level of PD-L1 in three different regions of the two groups was divided into high expression and low expression groups according to a 5% threshold. The results confirmed that the percentage of samples with TP53/CDKN2A/B covariations was significantly higher in the PD-L1 high expression group than in the PD-L1 low expression group in tumor areas (p < 0.001), stromal areas (p = 0.04), and total areas (p = 0.003) (Fig. 3F).
Overall, these results indicated that patients with co-occurring TP53/CDKN2A/B and ALK/RET/ROS1 rearrangements had an immunosuppressive microenvironment, including elevated PD-L1 expression levels and reduced overall CD8 + TILs, compared to those without co-occurring TP53/CDKN2A/B and ALK/RET/ROS1 rearrangements.
Validation of poor prognosis in patients with covariation in the TCGA NSCLC cohorts
In NSCLC patients harboring ALK/RET/ROS1 rearrangements, the above results provide important insights into the differences in prognosis between the co-occurring groups and the WT groups. Since the covariation group had higher TMB and PD-L1 expression but less CD8+ TIL infiltration, we predicted a poorer prognosis in this group of patients. To validate our hypothesis, we explored the genomic co-occurring status in the LUSC and LUAD cohorts from the TCGA database. The clinical baseline characteristics included the baseline age, sex, type of cancer, stage, and smoking history. Among all 340 patients, 64.6% (n = 219) were female, and 35.6% (n = 121) were male. The median age at baseline was 60.9 years, and the age range was 16–90 years (Supplementary Table 2). The patients with co-occurrence in TCGA were characterized as young and female with adenocarcinoma and no history of smoking, which is consistent with the characteristics of our study cohort.
In NSCLC patients with ALK/RET/ROS1 rearrangements from the TCGA cohort (Fig. 4A), the top 5 mutated genes were CDKN2A/B (24%), TP53 (16%), MDM2 (8%), MYC (7%), and SETD2 (7%). As with the results of our cohort, in patients with covariation, the TMBs were significantly higher than in patients without covariation (p < 0.001, p = 0.0018, and p < 0.001) (Fig. 4B). The prognostic analysis showed that, regardless of whether it was the TP53 comutation group, the CDKN2A/B covariation group, or the TP53/CDKN2A/B covariation group, the OS of these three groups was shorter than that of their control groups (WT groups) (p < 0.001, HR = 2.53, 95% CI = 1.59–4.04; p = 0.0625, HR = 1.54, 95% CI = 0.97–2.42; p < 0.001, HR = 2.21, 95% CI = 1.47–3.33).
In summary, the trends of the validation cohort were generally consistent with our datasets in terms of clinical characteristics, the genomic alteration landscape, and TMB. Furthermore, the findings in the validation cohort suggested that the group of patients harboring the co-occurring TP53/CDKN2A/B covariations and ALK/RET/ROS1 rearrangements had a worse prognosis than the group of patients without the co-occurring TP53/CDKN2A/B covariations and ALK/RET/ROS1 rearrangements.