Reliability analysis of exonic-breakpoint fusions identified by DNA sequencing for predicting the efficacy of targeted therapy in non-small cell lung cancer

A total of 7158 samples from 7148 patients with NSCLC who underwent molecular testing in our laboratory between January 2017 and December 2021 were retrospectively included in this study. Relevant clinical data, such as clinicopathological features and treatment history, were obtained from clinical records. The study was approved by the Institutional Review Board of the Cancer Hospital, Chinese Academy of Medical Sciences, and Peking Union Medical College. Written informed consent was obtained from each patient, and the methods were carried out in accordance with approved guidelines.

Tumor cellularity was evaluated through hematoxylin and eosin (HE)-stained slides by two or more pathologists, as previously described [14]. Tissues with ≥20% tumor cellularity were selected. Genomic DNA was extracted using a QIAamp DNA FFPE Tissue Kit, and RNA was extracted using a Novogene RNA Extraction Kit (Novogene, Tianjin, China). DNA and RNA quantities were measured using a Qubit 3.0 Fluorometer (Thermo Fisher Scientific, Carlsbad, CA, USA) and a NanoDrop ND-1000 Spectrophotometer (NanoDrop, Waltham, MA, USA), and the qualities were determined using an Agilent 2100 Bioanalyzer system (Agilent Technologies Inc. CA, USA).

Hybrid capture-based targeted DNA NGS was performed using a panel designed against 56 genes, as previously described [10]. Briefly, at least 50 ng of genomic DNA was used to generate sequencing libraries through sequential steps, including DNA fragmentation, PCR amplification, hybridization, and probe capture. Indexed sequencing libraries were sequenced using the Illumina Nextseq N500 platform (Illumina, San Diego, CA, USA), and sequencing data were analyzed with an in-house Molecular Diagnostics Management System to determine alterations such as mutations, copy number variants, and fusions. Fusions were identified by a variant allele frequency (VAF) ≥2% and coverage ≥1000×. All fusions were further manually analyzed by a geneticist to exclude artifactual fusion events owing to misalignment and mispriming, among other factors.

Hybrid capture-based targeted RNA NGS was performed using the TruSight RNA fusion panel (Novogene, Tianjin, China), consisting of the 95 genes listed in Additional file 1: Table S1. Briefly, 100 ng of total RNA was reverse transcribed using a random primer mix. Synthesized cDNA was end-repaired and then used to generate sequencing libraries according to the manufacturer’s instructions. The enriched libraries were sequenced with the Novaseq 6000 platform (Illumina, San Diego, CA, USA), and sequencing data were further analyzed using an in-house analysis system to identify fusion transcripts, including fusion genes and breakpoints, at the transcript level.

ALK, ROS1, and RET expression was evaluated by an IHC assay, as previously described [10]. In brief, tissue samples were incubated with a Ventana anti-ALK (D5F3) rabbit monoclonal antibody (Ventana Benchmark XT stainer, Ventana Medical Systems Inc., AZ, USA), an anti-ROS1 (OTI1A1) mouse monoclonal antibody (Zhongshan Golden Bridge Biotechnology, Beijing, China), or an anti-RET (EPR2871) rabbit monoclonal antibody (Abcam, Cambridge, MA, USA) according to the manufacturer’s instructions. Binary scoring system (−, positive; +, negative) was used for evaluating the ALK staining results, and ALK positivity (ALK+) was identified when tumor cells showed strong granular cytoplasmic staining [15]. ROS1 and RET expression was evaluated using the following scoring scheme: -, no staining; 1+, weak staining, 2+, moderate staining, and 3+, strong staining in > 10% of tumor cells.

FISH assays were carried out to detect ALK and ROS1 fusions at the DNA level. As previously described [15], tumor tissues were sectioned, mounted onto microscope slides, and incubated with Vysis LSI ALK/ROS1 Dual color, Break Apart Rearrangement Probes (Abbott/Vysis, Abbott Park, IL, USA). The slides were evaluated under a fluorescence microscope by two expert pathologists. Positive results were identified as more than 15% of tumor cells with splitting of one or both 5′ and 3′ probe signals or with isolated 3′ probe signals.

The clinical responses of advanced NSCLC patients who harbored ALK, ROS1, or RET fusions and had received matched targeted agents were assessed by oncologists based on radiographic imaging. The optimal response, including complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD), was evaluated following the guidelines of Response Evaluation Criteria in Solid Tumors version 1.1 [16]. The cutoff date was December 31, 2021, and data were censored if patients were still alive and showed no progression at the latest follow-up date. The objective response rate (ORR) was defined as the percentage of patients with CR or PR. Progression-free survival (PFS) was calculated from the date of treatment to the date of PD or death from any cause.

Tumor tissues from 7148 NSCLC patients were interrogated using DNA NGS. Most of these patients (6615/7148, 89.7%) were diagnosed with adenocarcinoma. The patient characteristics are listed in Additional file 1: Table S2. Kinase fusions, including ALK, ROS1, RET, NRG1, NTRK1, NTRK3, FGFR1, FGFR2, FGFR3, EGFR, and MET, were identified in 685 (9.6%) cases. ALK was the most commonly detected fusion gene, followed by RET and ROS1 (Fig. 1A). According to the genome base position of the breakpoint, fusions can be divided into four subtypes: intronic-breakpoint (“intron-intron” fusions), intergenic-breakpoint (“intron-intergenic,” “intergenic-intron,” or “intergenic-intergenic” fusions), exonic-breakpoint (“intron-exon,” “exon-intron,” or “exon-exon” fusions), and mixed-breakpoint fusions (“exon-intergenic” fusions, “intergenic-exon” fusions, or multiple fusions with two or more breakpoint subtypes). Overall, intronic-, intergenic-, exonic-, and mixed-breakpoint fusions were identified in 520 (75.9%), 51 (7.4%), 74 (10.8%), and 40 (5.8%) cases, respectively. The ratio of exonic-breakpoint fusion events was higher for uncommon fusions than for ALK, ROS1, and RET fusions (21.8% vs. 9.8%, P = 0.006, Fig. 1B), although no marked differences in clinicopathological characteristics were found between exonic-breakpoint fusions and other fusion subtypes (Additional file 1: Table S3). Paired samples from the same patient, including paired primary and metastatic tumors (n = 4), paired different metastatic tumors (n = 1), paired tissue and plasma samples (n = 2), and paired pre- and post-TKI samples (n=3), showed the same rearrangement involving exonic breakpoint, further supporting the presence of exonic-breakpoint fusions at the DNA level (Additional file 1: Table S4).

The exonic-breakpoint fusions could be further classified into three categories on the basis of relative genomic locations: (i) “exon-intron” fusions, 39.2% (29/74); (ii) “exon-exon” fusions, 4.1% (3/74); and (iii) “intron-exon” fusions, 56.7% (42/74) (Fig. 1C and Additional file 2: Fig. S1A-C). To verify the function and activity of exonic-breakpoint fusions, RNA NGS and/or IHC were successfully performed in 55 cases with sufficient tissue. Chimeric transcripts/proteins were identified in 44 cases (80%), whereas no expressed fusion transcript/protein was detected in 11 cases (20%, Fig. 1D).

Among the 11 cases with genomic-positive but RNA/protein-negative fusions, 4 (P1-P4) harbored “exon-intron” fusions involving ALK (n=2), NTRK1 (n=1), or FGFR1 (n=1) and 1 (P5) harbored an “exon-exon” ROS1 fusion (Table 1). Open reading frame disruption in the exonic regions of the 5′ portion may be responsible for these nonproductive rearrangements, as an exon disrupting breakpoint may introduce a premature stop codon that leads to the production of a truncated transcript/protein (Fig. 2A, B). Moreover, the remaining 6 cases (P6-P11) harbored “intron-exon” fusions involving ALK (n=1), ROS1 (n=1), RET (n=1), FGFR2 (n=1), MET (n=1), and EGFR (n=1). All of these specific fusions had a fusion partner that was rare or was never reported (Table 1). Upon manual review in Integrative Genomics Viewer (IGV) [17], we found that the 5′ portion of the kinase genes and the 5′ portion of the partners merged to form “5’-5’ fusions” in 4 cases (P6–P9; Table 1 and Fig. 2C). Similarly, the 3′ portion of the kinase genes and the 3′ portion of the partners merged to form “3’-3’ fusions” in 2 cases (P10–P11; Table 1 and Fig. 2E). The discrepancy between the DNA NGS and RNA NGS/IHC results was likely a result of the formation of antisense rearrangements in which the kinase genes and partners were transcribed in different orientations, resulting in no aberrant transcript/protein (Fig. 2D, F).

Among the 44 cases with consistent positive results identified by DNA NGS and RNA NGS/IHC, 19 harbored “exon-intron” fusions involving ALK (n=13), ROS1 (n=3), RET (n=1), NRG1 (n=1), and EGFR (n=1) and 25 harbored “intron-exon” fusions involving ALK (n=16), ROS1 (n=5), and RET (n=4) (Table 2). A manual review of matched DNA and RNA NGS results in IGV was conducted to reveal the possible mechanisms. For P12, with an “exon-intron” fusion (EML4 exon 14-ALK intron 19), DNA NGS indicated the genomic breakpoint in EML4 to be located in the middle region of exon 14, but RNA NGS revealed the breakpoint at the transcript level to be located in exon 13 (Fig. 3A), possibly due to the lack of the 3′ acceptor splice site of the EML4 exon 14 that led to exon skipping (Fig. 3B). Indeed, we found 11 such cases in our cohort (Table 2). Similarly, DNA NGS showed the genomic breakpoint in ALK to be located in the middle region of exon 19 in P23, which harbored an “intron-exon” fusion (EML4 intron 20-ALK exon 19). Nevertheless, exon 19 of ALK was spliced out of the fusion transcript, as revealed by RNA NGS (Fig. 3C). This phenomenon was observed in 12 cases in our cohort, possibly due to exon skipping (Table 2), as the disrupting exon of the 3′ portion lacked a 5′ donor splice site (Fig. 3D). In particular, primary/reciprocal ROS1 fusions were identified by DNA NGS in P26; the reciprocal ROS1 fusion (ROS1-CLK1) was also detected by RNA NGS, although its biological impact on cellular function remained to be confirmed (Table 2). Moreover, uncommon ALK fusions were detected in both P35 and P36 (P35: C2orf91 exon 4-ALK intron 19; P36: CLHC1 intron 4-ALK exon 19; Table 2) by DNA NGS, but the canonical EML4-ALK fusion transcript was identified in these two cases (Additional file 2: Fig. S2A and C). This indicates the involvement of complex rearrangements constituting multiple fusion junctions and rare genes at the DNA level. Nevertheless, these rare genes, even though they showed different transcriptional orientations from the kinase genes (P36), were spliced out during transcription (Additional file 2: Fig. S2B and D).

In addition, a rearrangement of SLC34A2 exon 13-ROS1 intron 33 was detected at the DNA level in P37, and RNA NGS revealed that the breakpoint at the transcript level was in SLC34A2 exon 13, upstream of the predicted breakpoint detected by DNA NGS (Fig. 3E and Table 2). Alternative donor site selection involving a cryptic splice site located within exon 13 of SLC34A2 might have contributed to the formation of an in-frame fusion transcript (Fig. 3F). In P38, the EML4 intron 6-ALK exon 20 rearrangement detected at the DNA level also produced an EML4 intron 6-ALK exon 20 fusion transcript (Fig. 3G and Table 2). The same breakpoint location was observed between matched DNA NGS and RNA NGS results, probably due to intron retention that the intron sequence was integrated into the transcript (Fig. 3H).

Four patients with genomic-positive but RNA/protein-negative ALK or ROS1 fusions received targeted agents (crizotinib or alectinib) as first-line treatment, but showed PD within 2 months because no expressed fusion transcript/protein was produced (Fig. 4A). Among those with genomic- and RNA/protein-positive fusions, 19 received a matched targeted therapy, including agents against ALK (n=15), ROS1 (n=2), and RET (n=2), and almost all of them benefited from this treatment (Additional file 1: Table S5). Among them, 11 patients harboring exonic-breakpoint ALK fusions who received crizotinib as a first-line treatment were further analyzed, together with 56 harboring intronic-breakpoint ALK fusions. All these cases were confirmed to be ALK positive by RNA NGS or IHC, and there were no marked differences in the baseline features of the two groups (Additional file 1: Table S6). Compared with patients harboring intronic-breakpoint ALK fusions, those harboring exonic-breakpoint ALK fusions exhibited no difference in the ORR (76.8%, 95% CI: 64.2–85.9% vs. 81.8%, 95% CI: 52.3–94.9%, P = 0.714; Additional file 1: Table S7). Moreover, no difference in median PFS was observed between patients with intronic-breakpoint ALK fusions (13.1 months, 95% CI: 11.5–14.8) and those with exonic-breakpoint ALK fusions (15.0 months, 95% CI: 11.6–18.4, P=0.500, Fig. 4B). Additionally, although no validation assays were available for the 19 patients harboring exonic-breakpoint fusions, 4 with ALK or ROS1 fusions treated with crizotinib or alectinib showed a durable response (10.1–23.8 months, Additional file 1: Table S5), strongly suggesting the presence of active fusion events in these 4 cases. However, two patients with exonic-breakpoint ROS1 fusions failed to respond to crizotinib therapy, indicating that the fusion events in these 2 cases may have been silenced with no detectable transcript or clinical relevance.