Genomic and immune characteristics of HER2‐mutated non‐small‐cell lung cancer and response to immune checkpoint inhibitor‐based therapy

The efficacy of immunotherapy in advanced HER2‐mutated non‐small‐cell lung cancer (NSCLC) remains incomprehensively studied. A total of 107 NSCLC patients with de novo HER2 mutations were retrospectively studied at Guangdong Lung Cancer Institute [GLCI cohort, exon 20 insertions (ex20ins): 71.0%] to compare clinical/molecular features and immune checkpoint inhibitor (ICI)‐based therapy efficacy between patients with ex20ins and non‐ex20ins. Two external cohorts (TCGA, n = 21; META‐ICI, n = 30) were used for validation. In the GLCI cohort, 68.2% of patients displayed programmed death‐ligand 1 (PD‐L1) expression < 1%. Compared with ex20ins patients, non‐ex20ins patients had more concurrent mutations in the GLCI cohort (P < 0.01) and a higher tumour mutation burden in the TCGA cohort (P = 0.03). Under ICI‐based therapy, advanced NSCLC patients with non‐ex20ins had potentially superior progression‐free survival [median: 13.0 vs. 3.6 months, adjusted hazard ratio (HR): 0.31, 95% confidence interval (CI): 0.11–0.83] and overall survival (median: 27.5 vs. 8.1 months, adjusted HR: 0.39, 95% CI: 0.13–1.18) to ex20ins patients, consistent with findings in the META‐ICI cohort. ICI‐based therapy may serve as an option for advanced HER2‐mutated NSCLC, with potentially better efficacy in non‐ex20ins patients. Further investigations are warranted in clinical practice.


Introduction
Human epidermal growth factor receptor 2 (HER2/ ERBB2) is an oncogenic driver of tumour cell proliferation and metastasis [1]. Mutated HER2 genes are rare (prevalence: 2-4%) in non-small-cell lung cancer (NSCLC) patients, and are enriched in patients who are females, non-smokers and younger, and in adenocarcinoma patients [2,3]. Over 20 types of HER2 exon 20 insertion (ex20ins) mutations have been identified in NSCLC, accounting for 25-50% of NSCLC patients exhibiting HER2 mutations [4][5][6]. However, the genomic and immune characteristics of patients with different HER2 mutations have not been comprehensively investigated.
Because neither TKIs nor ADCs have been approved as first-line treatments for HER2-mutated NSCLC patients with advanced disease [18], immune checkpoint inhibitors (ICIs) with/without platinum doublet chemotherapy serve as the standard first-line therapy [19]. Treatment-naive patients receiving ICI combination treatment achieved ORR of 52% and mPFS of 6 months [20], whereas other studies have reported ORRs of 7.0-27.0% and mPFS of 2.2-4.0 months under ICI-based therapy in second-line or subsequent treatment [21,22]. Notably, it has been revealed that none of the ICI responders harboured HER2 YVMA mutation [23], suggesting that ICI efficacy in patients with HER2 ex20ins mutations might differ from patients harbouring other HER2 mutations. Moreover, the efficacy of immunotherapy has not been systematically investigated in HER2 non-ex20ins patients, owing to the low prevalence of HER2 mutations in NSCLC and the focus on HER2 ex20ins. Thus, the association between HER2 mutation subtypes and immunotherapy efficacy remains controversial, and the potential underlying mechanisms have not been comprehensively investigated.
This study aimed to compare the molecular features and tumour microenvironment (TME) characteristics of NSCLC patients harbouring HER2 ex20ins and HER2 non-ex20ins mutations. The efficacy of ICIbased therapy was analysed in patients with advanced disease. Two external datasets of HER2-mutated NSCLC patients were used to validate the results.

Patients
This retrospective study was performed at Guangdong Lung Cancer Institute (GLCI), Guangdong Provincial People's Hospital; participants in the GLCI cohort were consecutively enrolled between January 2016 and December 2020. The main inclusion criteria were follows: (1) adults having at least 18 years of age; (2) pathologically confirmed NSCLC, according to the 2016 World Health Organization classification; and (3) identified with somatic de novo HER2 mutations in tissue or plasma samples. Patients with other oncogenic drivers, including sensitizing EGFR mutations, BRAF V600E mutation, KRAS G12X/Q61H mutation, ALK/ NTRK/RET/ROS1 fusion, and MET exon 14 skipping, were excluded. Insertion mutations in HER2 exon 20 (residues 770-831) were identified as HER2 ex20ins; HER2 mutations outside exon 20 (residues before 770 or after 831) were identified as HER2 non-ex20ins mutations. Patients harbouring both HER2 exon 20 and non-ex20ins mutations were grouped into the non-ex20ins subgroup. The clinical stages and histological subtypes of NSCLC were determined according to the '8th edition of the American Joint Committee on Cancer classification system'. Demographics and clinical characteristics of the participants, including age, sex, Eastern Cooperative Oncology Group performance status, and smoking history were obtained from the electronic medical record system of Guangdong Provincial People's Hospital.
ICI-based therapy was defined as a regimen that included inhibitors of programmed cell death-1 receptor or its ligand (PD-1/PD-L1). Advanced NSCLC patients in the GLCI cohort treated with ICI monotherapy or ICI combination therapy were grouped as the GLCI-ICI cohort (Fig. 1), and they were followed up until November 2021 or until death. The clinical response to ICI-based therapy was evaluated using computed tomography according to the 'Response Evaluation Criteria in Solid Tumours version 1.1'. The study procedures were approved by the Ethics Committee of Guangdong Provincial People's Hospital (2013185H), and written informed consent was obtained from each patient. The study methodologies also conformed to the standards set by the Declaration of Helsinki.

External cohorts
From The Cancer Genome Atlas Program (TCGA), NSCLC patients with de novo HER2 mutations and without other NSCLC oncogenic drivers were included in the TCGA cohort for the validation of genomic features and immune characteristics (Fig. 1). Tumour mutational burden (TMB) of each patient in this cohort was recalculated using a standardized method [24]. The abundance of tumour-infiltrating leukocytes was estimated by CIBERSORT using RNA-Seq data, and significantly differentially expressed genes between patients with HER2 ex20ins and HER2 non-ex20ins mutations were investigated using Gene Set Enrichment Analysis.
From six external data sets, advanced HER2-mutated NSCLC patients who did not harbour other NSCLC oncogenic drivers and received ICI-based therapy were grouped into the META-ICI cohort (Fig. 1), including 10 patients from a study by Rizvi et al. [25], nine patients from a study by Gandara et al. [26], two patients from a study by Miao et al. [27], five patients from a study by Anagnostou et al. [28], and four patients from a study by Samstein et al [29]. Clinicopathological and prognostic data were analysed.  2.3. DNA extraction, library preparation, and next-generation sequencing data processing Genomic profiling of tumour tissue or plasma samples before systemic treatment was performed using multiple targeted next-generation sequencing (NGS) panels, according to the protocol approved by the Ethics Committee of Guangdong Provincial People's Hospital. Tumour genomic DNA was extracted from formalinfixed paraffin-embedded (FFPE) samples using the QIAamp DNA FFPE Tissue Kit (QIAGEN, Dusseldorf, Germany). Genomic DNA was extracted from leukocyte (normal blood controls) using the QIAamp Circulating Nucleic Acid Kit (QIAGEN). Peripheral blood was collected and centrifuged (at 1800 g for 10 min at room temperature) within 2 h to separate the plasma and leukocytes. Cell-free DNA was extracted from the plasma using a QIAamp Circulating Nucleic Acid Kit (QIA-GEN). Sequencing libraries were prepared using the KAPA Hyper Prep Kit (KAPA Biosystems, Wilmington, MA, USA). Briefly, fragmented genomic DNA was subjected to end-repair, A-tailing, adapter ligation, size selection, polymerase chain reaction amplification, and purification sequentially. Target enrichment was performed using a customized xGen Lockdown Probes Panel (Integrated DNA Technologies, Coralville, IA, USA), human cot-1 DNA (Life Technologies) and xGen Universal Blocking Oligos (Integrated DNA Technologies). All procedures were performed according to the manufacturer's instructions. The enriched libraries were sequenced using the Illumina Hiseq4000 NGS platforms (Illumina, San Diego, CA, USA). TRIMMOMATIC was used for quality control of FASTQ files by removing leading/trailing low quality (reading < 15) or N bases [30]. Sequencing data were then aligned to the reference human genome (build hg19) and processed using the PICARD suite and the GENOME ANALYSIS TOOLKIT (GATK) [31,32]. A somatic mutation, filtered for common single nucleotide polymorphisms and germline mutations, was retained when it had at least 1% mutant allele frequency and at least three unique reads on different strands with good quality scores. Gene fusions and copy number variations were analysed using FACTERA and ADTEx [33,34], respectively, and manually reviewed using INTEGRATIVE GENOMICS VIEWER Software (IGV; Broad Institute, Cambridge, MA, USA). A total of 72 overlapping cancer-relevant genes from multiple NGS panels were included in the data analysis (Table S1).

Immunohistochemistry (IHC) for PD-L1 expression
Formalin-fixed paraffin-embedded tumour tissue specimens were stained using PD-L1 IHC 22C3 pharmDx (Agilent, Santa Clara, CA, USA), and a PD-L1positive cell was defined as complete circumferential or partial cell membrane staining of viable cells with 1+ to 3+ intensity. PD-L1 protein expression was evaluated using the tumour proportion score (TPS), which was calculated as the percentage of PD-L1-positive tumour cells divided by total tumour cells. Patients were categorized into three subgroups according to PD-L1 TPS: < 1%, 1-49%, and ≥ 50%.

Statistical analysis
Progression-free survival was defined as the time from the initiation of ICI-based therapy to disease progression or death from any cause; overall survival (OS) was defined as the time from the initiation of ICI-based therapy to death from any cause. The median follow-up time for the GLCI-ICI cohort was estimated using the observation time method. Fisher's exact test and two-sample t-test were performed to compare the frequencies and means of patients with HER2 ex20ins and HER2 non-ex20ins mutations, respectively. For survival data, Kaplan-Meier curves for PFS and OS were generated, and log-rank tests were used to compare differences. Hazard ratios (HR) with 95% confidence intervals (CI) were estimated using Cox proportional hazards models. Multivariable Cox proportional hazards models included clinical and molecular features that were identified as having a potentially strong influence on PFS or OS in univariate analyses or with significantly unbalanced distribution between HER2 ex20ins and non-ex20ins patients. The proportionality of hazards was assessed using log(Àlog) survival plots. Individuals with missing data were excluded from analysis. All quoted Pvalues were two-tailed, and P-values < 0.05 were considered to be statistically significant. Data were analysed using R software (version 4.0.3, Vienna, Austria) and the survival package.

Patient characteristics
A total of 107 eligible patients (76 with HER2 ex20ins and 31 with HER2 non-ex20ins mutations) were retrospectively enrolled in the GLCI cohort, with only one patient identified to have ex20ins and non-ex20ins mutations simultaneously. Thirty-five of patients were identified in the GLCI-ICI cohort, all of whom received ICI-based therapy (Fig. 1). The median age of the GLCI cohort was 59 years (range:

Immune microenvironment features
To gain more insight into the TME of patients with HER2 mutations, transcriptome data from the TCGA cohort were analysed using the CIBERSORT algorithm, and the degree of immune cell infiltration was estimated. Among the 21 patients in the TCGA cohort, HER2 non-ex20ins patients might have a potentially higher density of resting CD4 + memory T cells than HER2 ex20ins patients, while the difference was not statistically significant (15.6% vs. 12.3%, P = 0.54, Fig. S1A).

Clinical efficacy of ICI-based therapy
The genomic and TME features suggested that the efficacy of ICIs in advanced NSCLC patients with HER2 mutations was possibly associated with the subtype of HER2 mutation. Next, we analysed the efficacy of ICIs in the GLCI-ICI cohort (23 with HER2 ex20ins, Not otherwise specified    (Table 2) and genomic profiles of the GLCI-ICI cohort were similar to those of the entire study cohort (Fig. S1B). Fifteen patients in the GLCI-ICI cohort were treated with ICI monotherapy (Fig. 3A), and the remaining 20 patients received ICI combination therapy (Fig. 3B). The overall ORR was 20.0% (95% CI: 8.0-41.2%); the mPFS and median OS (mOS) of the GLCI-ICI cohort were 5.2 (95% CI: 3.5-9.9) months and 14.8 (95% CI: 8.1-not reached) months, respectively. Patients carrying HER2 non-ex20ins appeared to achieve relatively high disease control rates compared with those carrying HER2 ex20ins mutations (monotherapy, 85.7% vs. 28.6%, P = 0.10; ICI combination therapy, 100% vs. 66.7%, P = 0.27). Also, patients with HER2 non-ex20ins mutations exhibited a relatively high durable clinical benefit (complete/partial response or stable disease lasting over 6 months) compared with those with HER2 ex20ins (66.7% vs. 34.8%, P = 0.07, Fig. 3C); however, they did not have a significantly higher ORR (25.0% vs. 17.4%, P = 0.67). Details of treatment and ICI efficacy in the GLCI-ICI cohort are shown in Table 3.
To validate the findings in the GLCI-ICI cohort, 30 eligible patients from external data sets of ICI-based therapy studies were grouped into the META-ICI cohort (13 with HER2 ex20ins and 17 with HER2 non-ex20ins mutations, four with unavailable PFS data, and one with unavailable OS data, Table S3). Notably, the majority of patients in the META-ICI cohort were treated with ICI monotherapy. The mPFS and mOS of the META-ICI cohort were 4.0 (95% CI: 1.9-not reached) months and 18.9 (95% CI: 8.0-24.9) months, respectively. The demographic and clinical characteristics of the META-ICI cohort are summarized in Table S3. The mPFS and mOS of patients harbouring HER2 ex20ins were 1.9 and 8.0 months, respectively, and patients with HER2 non-ex20ins mutations presented a longer mPFS of 13.2 months and a longer mOS of 23.3 months. HER2 non-ex20ins

Discussion
In this study, we analysed the molecular and TME characteristics of NSCLC patients with de novo HER2 mutations, as well as their responses to immunotherapy. HER2 non-ex20ins patients had higher mutation numbers than HER2 ex20ins patients, whereas similar low-level PD-L1 expression was detected. HER2 non-ex20ins mutations were potentially associated with superior PFS and OS under ICI-based therapy, which was consistent with the findings in external cohorts. Based on the GLCI cohort, we provided a landscape view of the genomic and clinical features of 107 HER2-mutated NSCLC patients without common driver mutations. The diversity of HER2 mutations suggested potentially uniform responses to the same treatment. In this study, HER2 ex20ins mutations were detected in 76 (71.0%) patients, similar to the results of another East Asian cohort study by Tan et al.
(72.7%) [35]. In contrast, a study based on a western cohort (n = 84) revealed that HER2 ex20ins accounted for 34.4% HER2 mutations [6]. For concurrent TP53 mutations, the prevalence was similar between patients harbouring HER2 ex20ins and non-ex20ins mutations in our study (67.6% vs. 59.3%), whereas the prevalence in the TCGA cohort appeared to be relatively high in the non-ex20ins subgroup (28.6% vs. 71.4%). Consistent with the TMB results reported by Tan et al. [35], HER2 non-ex20ins patients in our study had higher mutation numbers than patients with HER2 ex20ins mutations.
Our data provide preliminary insights into the TME heterogeneity in HER2-mutated NSCLC. Similar to the results of a previous study [36], a generally immunosuppressed TME was observed, whereas a heterogeneous TME was observed in HER2-mutated NSCLC. In our study, patients with HER2 ex20ins and non-ex20ins mutations showed similar PD-L1 TPS, with over 60% of the patients having negative expression levels (< 1%), which aligned with the findings of previous studies [5,20,35]. In the TCGA cohort, non-ex20ins patients might be enriched with resting CD4 + memory T cells,   which should be further investigated in a large cohort of HER2-mutated NSCLC patients. Our findings suggest that ICI-based therapy might serve as a treatment option for advanced NSCLC patients with HER2 mutations, and that those harbouring HER2 non-ex20ins mutations could potentially benefit more. In the GLCI-ICI cohort, the mPFS and ORR of ICI-based therapy were 5.2 months and 20.0%, respectively, consistent with the result of previous studies [20][21][22][23]. In the German nNGM lung cancer cohort, a 26.2% ORR was reported in 61 HER2-mutated NSCLC patients receiving ICI-based therapy as the first-line and later lines of treatment [20]. In the French Lung Cancer Group 01-2018 and MSKCC research, ICI-based therapy showed ORRs of 27.3% and 11.5%, respectively [21,22]. In our study, a potentially better response to ICI was observed in patients harbouring HER2 non-ex20ins mutations than in those harbouring HER2 ex20ins, which was similar to the findings of previous studies [5,[37][38][39][40]. For instance, as reported in Fan's study, none of the six HER2 ex20ins patients treated with PD-1 inhibitors achieved an objective response [38]. In a study by Tan et al. [35], none of the four ex20ins patients achieved a response after receiving pembrolizumab monotherapy. Our study also revealed that non-ex20ins patients were more likely to achieve a better response to ICI-based therapy than ex20ins patients; however, further research with a larger sample size is warranted. The relatively poor prognosis of ex20ins patients could be partially explained by the low TMB levels in patients harbouring HER2 ex20ins [29,41], as advanced NSCLC patients with higher TMB (above 50% percentile) could achieve better PFS than patients with lower TMB (below 50% percentile) [25]. Mutated SETD2, enriched in non-ex20ins NSCLC, was identified as a favourable predictive biomarker for immunotherapy, associated with higher TMBs and better OS (HR: 0.55, 95% CI: 0.46-0.65) [42]. LRP1B mutations, being relatively frequent in HER2 non-ex20ins patients, were also associated with prolonged survival (HR: 0.63, 95% CI: 0.40-0.97) [43]. In our study, we did not detect a strong association between PD-L1 expression and PFS/OS, which might have resulted from the similar proportions of patients with at least 1% PD-L1 expression between the ex20ins and non-ex20ins subgroups.
In the era of HER2-ADC, comprehensive studies including considerable HER2 non-ex20ins patients were warranted to investigate how HER2 mutation subtype affects ADC efficacy and whether immunotherapy or HER2-ADC would be the optimal first-line regimen for NSCLC patients with specific subtypes of HER2 mutations. In DESTINY-Lung01, HER2 non-ex20ins patients did not achieve as good response to  T-DXd as ex20ins patients (ORR: 33% vs. 73%, P = 0.02) [15], suggesting that the efficacy of T-DXd might depend on HER2 mutation subtype. For HER2mutated patients in East Asia, Tan et al. reported that three of four (75.0%) ex20ins patients responded to T-DXd; however, the efficacy of ADCs in non-ex20ins patients remained unclear. In our study, HER2 non-ex20ins patients achieved ORR of 25.0% and mPFS of 13.0 months. Further studies comparing the treatment efficacy of HER2-ADC and immunotherapy might be useful for the treatment of advanced NSCLC patients with HER2 non-ex20ins mutations. Our study has some limitations. As this was a retrospective study, the timing of the disease progression assessment was not standardized. The potentially superior PFS in non-ex20ins patients might be influenced by the line of therapy, which could not be well controlled, owing to the limited sample size of the GLCI-ICI cohort. In addition, multiple NGS panels were performed on participants' tumour tissue or plasma samples, and only overlapping cancer-relevant genes were included for data analyses, resulting in the failure to accurately calculate TMB in the GLCI cohort and compare key ICI treatment-related signalling pathways. We presented the number of mutations per person instead of the TMB value in the GLCI cohort, even though TMB data were available for patients in the GLCI-ICI cohort. Another limitation is the limited sample size of the GLCI-ICI cohort, resulting in an inability to comprehensively analyse the prognostic data for patients treated with ICI monotherapy or combination therapy, separately. Additionally, the results of the TME comparison were not comprehensive, owing to the lack of RNA-Seq data in the GLCI-ICI cohort.

Conclusion
Our study revealed that HER2-mutated lung cancers present with a high-level molecular heterogeneity. Patients harbouring HER2 non-ex20ins mutations had higher mutation numbers than patients harbouring HER2 ex20ins mutations, whereas similar PD-L1 expression levels were detected. HER2 non-ex20ins mutations could potentially be considered positive predictors of ICI efficacy in advanced HER2-mutated NSCLC patients, and ICI-based therapy might be a good option for patients with HER2 non-ex20ins mutations.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Transcriptomic data of The Cancer Genome Atlas Program (TCGA) cohort and the genomic profile of the Guangdong Lung Cancer Institute-immune checkpoint inhibitor (GLCI-ICI) cohort. Table S1. 72 Overlapping cancer-relevant genes. Table S2. Demographics and clinical characteristics of TCGA cohort. Table S3. Demographics and clinical characteristics of META-ICI cohort.