A total of 176 treatment-naïve NSCLC patients with HER2 alterations receiving commercial NGS of 1,021 cancer-related gene panel (Geneplus-Beijing, Beijing, China) between August 2016 and March 2022 were retrospectively enrolled. Twenty-one patients with HER2 alterations who received targeted therapy at Peking Union Medical College Hospital between March 2018 and April 2022 were retrospectively enrolled to investigate the efficacy and resistance mechanism to anti-HER2 TKIs. Written informed consent was obtained from each patient. All procedures conformed to the principles of the Helsinki Declaration with approval by institutional review boards of Peking Union Medical College Hospital.

The clinicopathological data, including age, sex, tumor histology, smoking status, and treatment information, were collected from the patient medical record. The outcome measures are PFS, ORR and disease control rate (DCR) assessed by standard radiologic criteria. PFS was defined as the interval from initial anti-HER2 targeted therapy to progression or death from any cause. ORR was defined as the proportion of patients with a partial response (PR) or complete response (CR). DCR was defined as the proportion of patients with a stable disease or PR or CR.

DNA was isolated from tissue samples, peripheral blood, cerebrospinal fluid and hydrothorax using commercial kits (Qiagen, Hilden, Germany). The KAPA Library Preparation Kit (Kapa Biosystems, Wilmington, MA, USA) was used to prepare indexed Illumina NGS libraries. Custom-designed 1,021 or 59 cancer-related gene panels were used to hybridize the DNA libraries, and their selected regions and genes are listed in Supplementary Tables 1 and 2 . The hybridized libraries were sequenced using a 100-bp paired-end configuration on a DNBSEQ-T7RS sequencer (MGI Tech, Shenzhen, China).

After the removal of terminal adaptor sequences and low-quality reads with FASTP (18), the remaining reads were mapped to the reference human genome (hg19) and aligned using the Burrows-Wheel Aligner (version 0.7.12-r1039) with default parameters. GATK (3.4–46-gbc02625) and MuTect2 (1.1.4) were used to call somatic single nucleotide variants and small insertions and deletions. Contra (2.0.8) was used to identify copy number variations (19). NCsv (in-house algorithm 0.2.3) was employed to detect structural variants (20, 21). All candidate variants were manually confirmed by using the integrative genomics viewer browser. Variants were filtered to exclude clonal hematopoietic mutations with an inhouse database of clonal hematopoiesis variants of >10000 pan-cancer patients and healthy individuals (22), germline mutations in dbSNP, as well as variants that occur at a population frequency of >1% in the Exome Sequencing Project.

The TMB was determined as the number of somatic nonsynonymous single nucleotide variants and small insertions/deletions per mega-base in the coding region (with VAF ≥ 0.03) (23). Immunohistochemistry with the PD-L1 IHC 22C3 pharmDx assay (Agilent Technologies, Santa Clara, CA, USA) was performed to evaluate PD-L1 expression of tumor tissues.

Statistical analyses were performed using SPSS version 19.0 (SPSS Company, Chicago, IL). The Mann-Whiney U test and Student’s t-test were used for non-normally and normally distributed continuous variables, respectively. The comparison of categorical variables was conducted with Pearson’s χ2 test or Fisher’s exact tests. All statistical tests were performed with two-sided methods, and P < 0.05 was considered to indicate statistical significance.

A total of 176 treatment-naïve NSCLC patients with HER2 alterations were enrolled to analyze the co-alteration profile of HER2 alterations. The median age at diagnosis was 59.5 years old (range, 25-80), with 56.8% (100/176) males and 30.1% (53/176) smokers. Most patients had stage IV disease (54.0%, 95/176), followed by stage I (21.6%, 38/176), stage III (16.4%, 29/176), and stage II (8.0%, 14/176). Via clinicopathological examination, 86.4% (152/176) patients had adenocarcinoma, with 7.4% (13/176) squamous cell carcinoma, 1.7% (3/176) adenosquamous carcinoma and 4.5% (8/176) unclear tumor histology. Of all 176 tumor tissues used for 1,021 gene panel sequencing, 123 (69.9%) samples were from primary lesions, 33 (18.7%) from metastases lesions and 20 (11.4%) were of unknown origin ( Table 1 ).

According to the sequencing data, a total of 1,945 somatic alterations were identified, with a mean of 11.1 alterations per sample. Among these alterations, 1,361 were substitution/small insertions and deletions, with 514 gene amplifications, 65 truncations, and 5 fusion/rearrangements. HER2 mutations were detected in 114 (64.8%) patients, including 17 patients with HER2 co-amplification. HER2 amplification only was found in 62 (35.2%) patients ( Figure 1A ). A total of 32 types of HER2 mutations were identified, including 22 (68.6%) oncogenic types, which are documented as pathogenic in the COSMIC database ( Figure 1B ). The tyrosine kinase domain (TKD) of HER2 was detected as the most frequently mutated region (75.4%), followed by the furin-like domain (19.3%) and others (13.1%). Across all HER2 oncogenic mutations, Y772_A775dup was the most common mutation (46.5%), followed by S310F/Y (15.8%) and G776delinsVC/IC/LC/SVC/VV (14.9%) ( Figure 1B ). A comparison of the clinical features between oncogenic TKD and non-TKD mutations indicated that patients with oncogenic non-TKD mutations were older than those with oncogenic TKD mutations ( Table 2 ). Furthermore, patients with oncogenic non-TKD mutations had a significantly higher prevalence of EGFR, KEAP1, and RBM10 mutations and EGFR amplification than those with oncogenic TKD mutations (EGFR mutation: 51.9% vs. 2.4%, P < 0.001; KEAP1 mutation: 14.8% vs. 1.2%, P = 0.01; RBM10 mutation: 18.5% vs. 0.0%, P = 0.001; EGFR amplification: 11.1 vs. 1.2%, P = 0.04).

Given the evidence that HER2 mutations and amplification exhibit distinct entities in carcinogenesis (1, 8), we elucidated the co-alteration profiles of HER2 mutations and amplification respectively. The most frequent concomitant alterations of HER2 mutations with/without amplification were TP53 (50.9%), EGFR (17.5%), LRP1B (9.6%), SMAD4 (8.7%), PIK3CA (7.9%), CDKN2A (7.9%) and SMARCA4 (7.0%) ( Figure 1C ). The most common concomitant alterations of HER2 amplification only were TP53 (79.0%), EGFR (43.6%), CDK12 (32.3%), MYC (32.3%), CDKN2A (27.4%), CCND1 (22.6%) and RB1 (19.4%) ( Figure 1D ). Of note, patients with HER2 amplification only had a significantly higher prevalence of EGFR mutation (40.3% vs. 15.8%, P = 0.005) and EGFR amplification (19.4% vs. 5.2%, P = 0.007), as well as lower prevalence of EGFR wild-type (56.5% vs. 82.5%, P < 0.001) than those with HER2 mutations ( Figure 1E ). We further compared TMB between patients with HER2 mutations and with HER2 amplification only in early- and late-stage disease respectively. TMB data was available for 106 patients with HER2 mutations and 60 patients with HER2 amplification only. Our results showed that patients with HER2 amplification only had a significantly higher TMB than those with HER2 mutation in early-stage but not in late-stage disease ( Figure 1F ). Collectively, the molecular characteristics of HER2 mutations and amplification differed.

Among 114 patients with HER2 mutations, 60 of them (52.6%) were early stage (stage I-III). Compared with late NSCLC, patients with early NSCLC were younger ( Supplementary Table 3 ). To compare the mutational profiles between early- and late-stage NSCLC with HER2 mutation and to avoid age as a potentially confounding variable, we conducted a propensity-matched comparison manually. As a result, a total of 100 matched patients were obtained, including 50 out of the 60 patients from the early-stage group and 50 out of 54 patients from the late-stage group. Their baseline characteristics are depicted in Table 3 . The matched early- and late-stage patients had similar HER2 mutation types ( Figure 2A ). Remarkably, the result of the co-alteration comparison between matched early- and late-stage patients indicated that TP53 mutation was more frequently observed in patients with late-stage disease (78.0% vs. 30.0%, P < 0.001) ( Figure 2B ). Moreover, patients with late-stage NSCLC had significantly higher TMB than those with early-stage (P = 0.03) ( Table 3 ). No significant correlations were found between PD-L1 expression and tumor stage in patients with HER2 mutations.

Across 62 patients with HER2 amplification only, 21 (33.9%) and 41 (66.1%) patients were at early and late stage respectively. Patients with late-stage NSCLC had a higher proportion of adenocarcinoma than those with early stage ( Supplementary Table 3 ). Similarly, a manual propensity-matched comparison was performed to compare the difference in genomic alterations between early- and late-stage NSCLC with HER2 amplification only. We identified 21 out of the 41 patients from the late-stage group that matched with 21 patients with early-stage NSCLC, and their clinical characteristics are displayed in Table 3 . No statistical differences were detected when comparing the frequency of co-alterations, TMB value, and PD-L1 expression level between the two matched groups. Therefore, molecular characteristics and PD-L1 expression of HER2-amplified NSCLC tumors were independent of tumor stages.

To assess the efficacy of targeted therapy in NSCLC patients with HER2 alterations, we retrospectively enrolled patients with HER2 alterations who received anti-HER2 targeted therapy in our center. Twenty-one patients were eligible, and their clinical and molecular characteristics are presented in Table 4 . The median age was 56 years (range, 39-76), and 52.4% (11/21) were female. Most patients had stage IV disease (81.0%, 17/21) and adenocarcinoma (95.2%, 20/21). Via genetic testing, most patients carried HER2 exon 20 insertion (76.2%, 16/21), one carried HER2 V659E mutation, one carried HER2 S310X (exact sequence unknown) mutation, two carried HER2 amplification only, and one carried HER2 alteration with unclear sequence.

Among these 21 patients, 14 (66.7%) patients received pyrotinib and 7 (33.3%) patients received afatinib. Two patients received pyrotinib as both second- and third-line treatment, of whom the clinical efficacy was only assessed on the second-line treatment. At the last follow-up date of October 2022, two patients were still on pyrotinib treatment and did not reach PFS, 18 patients developed disease progression, and one had died from the disease. The median line of afatinib and pyrotinib treatment was 2 (range, 1-5) and 2 (range, 1-4) respectively. The clinical response to targeted therapy was evaluated according to RECIST v1.1. As a result, pyrotinib yielded an ORR of 33.3% (4/12) and a DCR of 91.67% (11/12), and afatinib produced an ORR of 20% (1/5) and a DCR of 20% (1/5) ( Figure 3A ). The total mPFS was 5.0 months (95% CI, 3.8-10.1). Of note, patients who received pyrotinb treatment had a mPFS of 5.9 months (95% CI, 3.8-13.0), which is longer than those who received afatinib (4.0 months [95% CI, 1.9-6.3]), although statistical significance was not reached (P = 0.06) ( Figure 3B ). However, due to the limited number of patients enrolled in this analysis, this result is preliminary.

Pre- and post-treatment genomic landscapes were dissected to uncover the potential resistance to afatinib and pyrotinib in ten eligible patients. Among them, nine patients carried HER2 mutations, and one carried HER2 amplification ( Figure 4 ). All patients still harbored HER2 alterations after afatinib or pyrotinib treatment. We first assessed the HER2-dependent resistance to afatinib or pyrotinib. For patient P14 harboring HER2 G778_p780dup mutations and received afatinib as second-line treatment for 2 months, HER2 amplification was identified at disease progression. For patient P18, carrying HER2 amplification and received fifth-line afatinib treatment for one month, HER2 G518W oncogenic mutation and PIK3CA mutation were found upon progression disease (PD) ( Figure 4 ).

To explore HER2-independent resistance to afatinib and pyrotinib, we analyzed the acquired alterations from eight patients who did not have HER2 secondary alterations at PD. Mutations reported by COSMIC datasets as oncogenic mutations were elaborately analyzed. Ten oncogenic mutations were identified in five patients. TP53 participating in DNA damage repair (DDR) pathway, its oncogenic mutation was identified in two patients (P17 and P21) who received second-line afatinib and fourth-line pyrotinib respectively. Other alterations of DDR pathway mediated by MSH2 oncogenic mutation was found in patient P20, who received second-line afatinib. Of note, two of these three patients harbored high genomic instability, with acquired 12 copy number alterations in P17 and acquired 13 mutations in P21, likely due to the deficiency of DDR pathway. SMARCA4 gene, encoding a part of the SWI-SNF chromatin-remodeling complex, its loss of function mutation was detected in patient P12 treated with first-line pyrotinib, indicating that compromise of the SWI–SNF complex potentially confers resistance to pyrotinib. Additionally, mutations occurred in genes involved in epigenetic regulation (LRP1B, DNMT3A, MLL3, and TET2) were found in two patients (P12 and P15), suggesting epigenetic dysregulating might also play a role in resistance to HER2 targeted therapy.