Molecular characterization of genomic breakpoints of ALK rearrangements in non-small cell lung cancer

Abstract

ALK rearrangement is called the ‘diamond mutation’ in non-small cell lung cancer (NSCLC). Accurately identifying patients who are candidates for ALK inhibitors is a key step in making clinical treatment decisions. In this study, a total of 783 ALK rearrangement-positive NSCLC cases were identified by DNA-based next-generation sequencing (NGS), including 731 patients with EML4-ALK and 52 patients with other ALK rearrangements. Diverse genomic breakpoints of ALK rearrangements were identified. Approximately 94.4% (739/783) of the cases carried ALK rearrangements with genomic breakpoints in the introns of ALK and its partner genes, and 2.8% (21/739) of these cases resulted in frameshift transcripts of ALK. Meanwhile, 5.6% (44/783) of the ALK rearrangement-positive cases had breakpoints in the exons that would be expected to result in abnormal transcripts. RNA-based NGS was performed to analyse the aberrant fusions at the transcript level. Some of these rearranged DNAs were not transcribed, and the others were fixed by some mechanisms so that the fusion kinase proteins could be expressed. Altogether, these findings emphasize that, when using DNA-based NGS, functional RNA fusions should be confirmed in cases with uncommon/frameshift rearrangement by RNA-based assays.

Abbreviations
• ALK
• anaplastic lymphoma kinase
• EML4
• echinoderm microtubule-associated protein-like 4
• IGV
• integrated genomics viewer
• IHC
• immunohistochemistry
• KIF5B
• kinesin family member 5B
• KLC1
• kinesin light chain 1
• NGS
• next-generation sequence
• NSCLC
• non-small-cell lung cancer
• QC
• quality control
• TKI
• tyrosine kinase inhibitor
• TPR
• translocated promoter region

Introduction

Lung cancer is a major malignancy that threatens human life and health worldwide, with a high incidence and mortality in both male and female patients [1]. Non-small cell lung cancer (NSCLC) is the main type of lung cancer, accounting for approximately 85% of cases, and it includes lung squamous cell carcinoma, lung adenocarcinoma and large cell lung carcinoma [2]. As a molecularly heterogeneous disease, multiple genetic alterations can drive the occurrence of NSCLC [3]. Approximately 3–7% of NSCLC patients harbour anaplastic lymphoma kinase gene (ALK) rearrangements [4,5]. The wild-type ALK gene encodes a transmembrane protein that is a classic receptor tyrosine kinase located on the cell membrane [6–8]. When the tyrosine kinase domain (exon 20 to exon 28) of ALK is retained in ALK-containing fusion proteins, it results in oncogenic tyrosine kinases capable of driving oncogenesis through several downstream signalling pathways, including the RAS/MEK/ERK, PI3K and JAK/STAT pathways [6,9]. Tyrosine kinase inhibitors (TKIs) represent a major milestone in the treatment of ALK rearrangement-positive NSCLC patients, playing a crucial role in combating these oncogenic alterations [10–12].

Multiple methods have been developed to detect gene rearrangements/fusions in various clinical diagnostic settings [13,14]. An assay utilizing DNA-based next-generation sequencing (NGS) has been applied frequently in recent years. Abundant types of ALK rearrangements are identified by DNA-based NGS [15]. The most common partner gene for ALK is echinoderm microtubule-associated protein-like 4 (EML4) [16], and other noncanonical partner genes have been identified, such as kinesin family member 5B (KIF5B), kinesin light chain 1 (KLC1) and translocated promoter region (TPR) [17]. Previous studies have reported diverse genomic breakpoints of ALK rearrangements that occur in different regions (introns or exons) in NSCLC, and intronic breakpoint fusions usually result in in-frame chimeric fusion transcripts/proteins [15,18]. Multiple ALK-fusion variants caused by variable genomic breakpoints have been reported with different sensitivities to ALK TKIs, especially in canonical EML4-ALK fusions [19].

In theory, the potential pathogenicity of fusion variants ensures that the component of the kinase domain is in frame in the transcripts [20]. However, the predicted transcripts of some rearrangement types may be imprecise based on the coding sequence of the DNA. The potential unreliability of genomic breakpoints identified by DNA-based NGS in predicting fusion transcripts has been proposed [15]. Therefore, the validation of ALK rearrangements detected at the DNA level, especially the uncommon genomic breakpoints of rearranged genes, needs to be constantly supplemented [21].

In this study, we retrospectively analysed the DNA molecular characteristics of ALK rearrangements in a local NSCLC database, and ALK rearrangements with noncanonical partner genes and uncommon genomic breakpoints were identified. To explore the actual transcripts of these rearrangements, which may result in abnormal transcripts, an RNA-based NGS assay was performed. This study aimed to effectively and accurately determine the actual fusion status of the ALK gene in the context of specific ALK rearrangements.

Materials and methods Patients and samples

From February 2018 to November 2021, a total of 783 lung cancer patient samples (718 tissues and 65 plasma fractions) were recruited from the Affiliated Hospital of Qingdao University, the Zhejiang Provincial People's Hospital and the Second Hospital of Shandong University, and these cases were detected as ALK rearrangement-positive by DNA-based NGS. In this study, ALK rearrangements retaining the 3′ ALK kinase domain were included and divided into canonical (EML4-ALK) and noncanonical (other partner genes-ALK) types. Their clinical characteristics were collected from their medical records and analysed. This study was conducted in accordance with the Declaration of Helsinki (as revised in 2013), and it was approved by the ethics committee of Zhejiang Provincial People's Hospital (No. QT2022218). The experiments were undertaken with the understanding and written consent of each subject.

DNA sample extraction and library construction

The sequencing methods have been described in earlier papers [22]. DNA samples from NSCLC patients were analysed using targeted deep sequencing with NGS technology. Genomic DNA was extracted from FFPE samples using a QIAamp DNA FFPE Tissue Kit (Qiagen, Valencia, CA, USA) following the manufacturer's instructions. Plasma cfDNA was extracted using a MagMAX Cell Free DNA Isolation Kit (Thermo Fisher Scientific, Waltham, MA, USA). DNA samples were quantified with the Qubit 2.0 Fluorometer using a Qubit dsDNA HS Assay kit (Life Technologies, Carlsbad, CA, USA) following the manufacturer's instructions. Genomic DNA from each FFPE sample was sheared into 150- to 200-bp fragments using the M220 Focused-ultrasonicator (Covaris, Woburn, MA, USA). Fragmented genomic DNA and cfDNA libraries were constructed with the KAPA HTP Library Preparation Kit (KAPA Biosystems, Wilmington, MA, USA) following the manufacturer's protocol. The concentration of DNA in the library was determined using the Qubit dsDNA HS assay kit.

Sequencing data analysis

DNA libraries were analysed using an OncoFocus panel (Genetron Health, Beijing, China), which includes 63 major lung cancer-related genes. Quality control was undertaken on the raw sequencing data to remove the adapters and low-quality regions using trimmomatic version 0.36 (Max Planck Institute of Molecular Plant Physiology, Potsdam, Germany). Local alignments of reads to the hg19 genome (GRch37) were carried out using the burrows–wheeler aligner tool (version 0.7.10) [23]. Somatic single nucleotide variants were retrieved using mutect (https://software.broadinstitute.org/cancer/cga/mutect) [24], somatic insertions and deletions were retrieved using strelka (https://github.com/Illumina/strelka) [25], and structural variations were determined using genefuse version 0.6.1 (https://github.com/OpenGene/GeneFuse) [26]. The variants were filtered and excluded with a population frequency over 0.1% based on guidelines by the Exome Aggregation Consortium. The remaining variants were annotated with Oncotator and Vep.

RNA-based NGS

A Fusioncapture panel (Genetron Health, Beijing, China), which is a 395-gene RNA panel, was used to identify gene fusions at the transcript level. Total RNA was isolated using the AllPrep DNA/RNA Mini Kit (Qiagen) and then reverse transcribed to cDNA using SuperScript III Reverse Transcriptase (Thermo Fisher Scientific, Waltham, MA, USA). The libraries were constructed with the KAPA HTP Library Preparation Kit (KAPA Biosystems) and subjected to Illumina HiSeq X-Ten for paired-end sequencing. Sequencing reads were mapped to a human reference genome (hg19) using hisat2-2.0.5 (Johns Hopkins University School of Medicine, Baltimore, MD, USA). Gene fusions were identified using fusionmap [27].

Immunohistochemistry

The immunohistochemistry (IHC) assay has been described in the earlier studies [14,28]. The IHC for ALK protein expression was performed on FFPE sections using a VENTANA ALK (Clone D5F3) CDx Kit and benchmark Ultra Immunostainer (Ventana Medical Systems, Inc., Tucson, AZ, USA, Cell Signaling Technology, Danvers, MA, USA) according to the manufacturer's instructions. The presence of granular cytoplasmic staining in the tumour cells (any percentage of positive tumour cells) was considered positive for ALK, while the absence of granular cytoplasmic staining in the tumour cells was considered negative for ALK.

Statistical analyses

The clinical characteristics of the study population were statistically analysed by the chi-square test and Student's t-test. A P value < 0.05 indicated statistical significance. Analyses and the data presentation were undertaken using ibm spss statistics 26.0 (IBM, Armonk, NY, USA) and graphpad prism 8.0.1 (GraphPad, La Jolla, CA, USA). The rearrangements and fusions were illustrated using Integrative Genomics Viewer, igv 2.11.4 (Broad Institute, Cambridge, MA, USA).

Results Identification of ALK rearrangements in NSCLC patients

In this study, a total of 783 ALK rearrangement-positive cases were identified by DNA-based NGS, including 731 cases with canonical EML4-ALK rearrangements and 52 cases with noncanonical ALK rearrangements (Table 1). In these cases, with EML4-ALK, 95.2% (696/731) of the cases carried rearrangements of EML4 introns to incorporate ALK introns at the DNA level, 2.5% (18/731) of the cases were EML4 exons rearranged to incorporate ALK introns, and 2.3% (17/731) of the cases were EML4 introns rearranged to incorporate ALK exons (Fig. 1A). For noncanonical ALK rearrangements, the ‘intron–intron’, ‘exon–intron’, ‘intron–exon’ and ‘exon–exon’ types were identified in 82.7% (43/52), 9.6% (5/52), 5.8% (3/52) and 1.9% (1/52) of cases, respectively (Fig. 1B). In order for the ALK-related fusion to be pathogenic, the ALK components have to remain in frame within the structure of the detected transcripts (ALK components that are out of frame would not be expected to be oncogenic because of the deletion of the kinase domain). We applied this logic to assess the DNA-based NGS data and predicted chimeric transcripts of these fusion patterns. Thus, 2.0% (14/696) of the ‘intron to intron’ EML4-ALK and 16.3% (7/43) of the ‘intron to intron’ noncanonical ALK rearrangements were predicted to be frameshifts with respect to the 3′ gene ALK (Fig. 1A,B). This prediction was based on the termination codon that appeared early due to the frameshift of the fusion transcript. The frameshift collection included four types of EML4-ALK (intron 3 : intron 19, intron 19 : intron 19, intron 14 : intron 19 and intron 17 : intron 19) (Fig. 1A) and seven types of noncanonical ALK rearrangements (TPM3 intron 7 : ALK intron 19, TOGARAM2 intron 8 : ALK intron 19, ARHGEF33 intron 8 : ALK intron 19, SAMD12 intron 2 : ALK intron 19, AFF1 intron 2 : ALK intron 19, CCDC9 intron 10 : ALK intron 19 and C11orf63 intron 2 : ALK intron 18) (Fig. 1B). Thus, we separated the ALK rearrangements into three categories: in-frame, frameshift and exon breakpoints (‘exon to intron’, ‘intron to exon’ or ‘exon to exon’) (Table S1).

Table 1 Clinicopathologic feature in lung cancer patients with ALK rearrangements by DNA-based NGS. ADC, adenocarcinoma; ex : int, exon to intron; int : ex, intron to exon; int : int, intron to intron.

[Table omitted. See PDF]

Fig. 1. Categories of genomic breakpoints of ALK rearrangements by DNA-based NGS (n = 783). (A) The distribution of canonical ALK rearrangements and the frequency of predicted transcript types in NSCLC patients with ‘intron–intron’ rearrangements. The table list the EML4-ALK rearrangements whose predicted transcripts are frameshift. (B) The distribution of noncanonical ALK rearrangements and the frequency of the predicted transcript types with ‘intron–intron’ rearrangements. The table list the noncanonical ALK rearrangements whose predicted transcripts are frameshift.

Characterization of the genomic breakpoints in the ALK rearrangements in the NSCLC patients

Based on our DNA NGS panel (covering all exons of ALK, as well as introns 16–20 of ALK), abundant information on the breakpoints was found. By analysing the genomic breakpoints of our EML4-ALK-positive samples, the breakpoints of ALK were found to be relatively concentrated and distributed in six of the regions. A total of 96.7% (707/731) of ALK breakpoint regions were located at intron 19, five cases were located at intron 18, and two cases were rearranged at intron 16 and intron 17. In addition, the ALK genomic breakpoints of 10 cases occurred in exon 19, and seven cases occurred in exon 20 (Fig. 2A). In contrast, the breakpoints of EML4 were variable and distributed in 18 of the regions, most of which were in 10 intronic regions (713/731, 97.5%), the most common regions being intron 6 (294/731, 40.2%), intron 13 (271/731, 37.1%) and intron 20 (84/731, 11.5%). In addition, 2.5% (18/731) of the fusion breakpoints occurred in the exonic regions of EML4 infrequently, including exon 3, exon 7, exon 13, exon 14, exon 16, exon 18, exon 19 and exon 21 (Fig. 2B). The distribution of common genomic breakpoints of EML4 and ALK suggested common types of EML4-ALK variants, including V1 (E13; A20), V2 (E20; A20) and V3 (E6; A20) (Table 2).

Fig. 2. Protein domain structure and functional motifs of the EML4 and ALK genes, as well as the distribution of the breakpoints of the EML4-ALK rearrangements. (A) Protein structures of ALK and the distribution of the genomic breakpoints of ALK. Wild-type ALK is a transmembrane protein and is a classical receptor tyrosine kinase located on the cell membrane. Its extracellular structures include two MAM (Meprin, A5 protein, and protein tyrosine phosphatase Mu) domains and one LDL, G-rich, transmembrane (TM) kinase domain. The chart on the right shows the proportion of genomic breakpoints of ALK. Blue represents ALK introns, and yellow represents ALK exons. (B) Protein structures of EML4 and the distribution of genomic breakpoints of EML4. The N-terminal coiled-coil region forms a trimer structure (TD) that can combine with microtubules through the basic region, and the β-propeller (TAPE) domain (composed of the hydrophobin motif (HELP) and the repeated WD40 domains) can bind to soluble tubulin. The chart on the right shows the proportion of genomic breakpoints of EML4. Green represents EML4 introns, and pink represents EML4 exons.

Table 2 Frequency of EML4-ALK rearrangements based on DNA NGS. ex, exon; int, intron.

EML4-ALK (n = 731)	Predicted RNA type (%)	DNA-based NGS	Counts (%)
Intron : intron (n = 696)	Inframea (n = 682, 93.3%)	int13 : int19	267 (36.5)
		int20 : int19	82 (11.2)
		int6 : int19	282 (38.6)
		int2 : int19	5 (0.7)
		int18 : int19	31 (4.2)
		int6 : int18	4 (0.5)
		int6 : int16	1 (0.1)
		int21 : int19	8 (1.1)
		int6 : int17	1 (0.1)
		int13 : int18	1 (0.1)
	Frameshiftb (n = 14, 1.9%)	int14 : int19	4 (0.5)
		int17 : int19	4 (0.5)
		int3 : int19	3 (0.4)
		int19 : int19	3 (0.4)
Exon : intron (n = 18)	Exon breakpointsb (n = 18, 2.5%)	ex7 : int19	1 (0.1)
		ex3 : int19	1 (0.1)
		ex21 : int19	8 (1.1)
		ex19 : int19	1 (0.1)
		ex18 : int19	2 (0.3)
		ex16 : int19	1 (0.1)
		ex14 : int19	3 (0.4)
		ex13 : int19	1 (0.1)
Intron : exon (n = 17)	Exon breakpointsb (n = 17, 2.3%)	int6 : ex19	4 (0.5)
		int6 : ex20	2 (0.3)
		int2 : ex19	1 (0.1)
		int19 : ex19	1 (0.1)
		int19 : ex20	2 (0.3)
		int14 : ex20	1 (0.1)
		int13 : ex19	3 (0.4)
		int20 : ex19	1 (0.1)
		int17 : ex20	1 (0.1)
		int20 : ex20	1 (0.1)

^aCommon rearrangements.

^bUncommon rearrangements.

For noncanonical ALK rearrangements, KIF5B, H1P1, DTCN1, KLC1, STRN, CLTC, CCDC9 and PRKAR1A were detected multiple times as partner genes, and the distribution of the genomic breakpoints of these partner genes was diverse (Table 3). Meanwhile, the genomic breakpoints of ALK were mainly distributed in intron 19 (43/52, 82.7%), which was similar to the canonical EML4-ALK rearrangements, followed by intron 18 (5/52, 9.6%). Breakpoints in exons 17, 18, 19 and 20 of ALK accounted for one case each (Table 3).

Table 3 Catalog of noncanonical partner genes of ALK rearrangements based on DNA NGS. ex, exon; int, intron.

[Table omitted. See PDF]

^aCommon rearrangements.

^bUncommon rearrangements.

Validation of the ALK frameshift rearrangement pattern by RNA-based NGS

RNA-based NGS was performed on nine stored samples as a frameshift cohort. Limited by the low-quality RNA samples, two of the FFPE samples were not tested due to failure during the RNA quality control (QC) process. We finally detected seven qualified samples from available tissue, including four cases (#P2106140203, #P2011280013, #P2008100038 and #P1902170006) with EML4-ALK and three cases (#P2007070051, #P2005010014 and #L-2018-00005429) with noncanonical ALK rearrangements (Table 4; Table S2). Furthermore, ALK-IHC was performed on several available samples as a supplement validation of the NGS results, although there might be several mechanisms, such as ALK fusions, amplification and alternative transcription initiation of ALK, that can drive the overexpression of ALK and result in a positive IHC result [29,30].

Table 4 Frameshift cohort of ALK rearrangements/fusions at DNA, RNA, protein levels. ex, exon; int, intron.

Patient ID	DNA-based NGS	RNA-based NGS	IHC
P2106140203	EML4 int17 : ALK int19	EML4 int17 : ALK ex20	Positive
P2103020235	EML4 int14 : ALK int19	Fail	N/A
P2011280013	EML4 int19 : ALK int19	EML4 ex19 : ALK ex20	N/A
P2008100038	EML4 int19 : ALK int19	Negative	Negative
P1902170006	EML4 int3 : ALK int19	Negative	N/A
L-2018-00000747	EML4 int14 : ALK int19	Fail	N/A
P2007070051	TOGARAM2 int8 : ALK int19	EML4 ex13 : ALK ex20	N/A
P2005010014	SAMD12 int2 : ALK int19	EML4 ex20 : ALK ex20	N/A
L-2018-00005429	AFF1 int2 : ALK int19	EML4 ex13 : ALK ex20	N/A

In two cases (#P2008100038 and #P1902170006), the EML4-ALK fusions were negative at the transcript level despite positivity in DNA-based NGS, and these results were confirmed by IGV (Fig. S1). We assumed that the rearranged genetic material of these two cases is not transcribed, and the ALK-IHC results of case #P2008100038 showed that ALK protein expression was negative (Fig. S2C).

Another 2 EML4-ALK-positive cases (#P2106140203 and #P2011280013) were positive for a fusion in both the RNA-based and DNA-based NGS assays. The predicted transcript of case #P2106140203 was exon 17 of EML4 fused to exon 19 of ALK and it would not have been in frame (Fig. 3A). However, the actual transcript detected by the RNA-based NGS assay of case #P2106140203 did not match the predicted transcript. The IGV showed a novel variant composed of a sequence derived from ALK intron 19 (42 adjacent nucleotides, 5′-CCAGGCTGCCAGGCCATGTTGCAGCTGACCACCCACCTGCAG-3′) and a sequence derived from EML4 intron 17 (26 nonadjacent nucleotides, 5′-GAGACAAAAACATGAAGTCAATTTTC-3′) inserted between exon 17 of EML4 and exon 20 of ALK (E17ins26; ins42A20) (Fig. 3B). In addition, the fusion type of case #P2011280013 was intron 19 of EML4 fused to intron 19 of ALK at the genomic level, and the predicted transcript was not in frame (Fig. 3C). However, a novel variant (E19ins1; A20) with a nucleotide inserted between exon 19 of EML4 and exon 20 of ALK was detected by RNA-based NGS and it did not match the predicted transcript (Fig. 3D). The RNA NGS results of #P2106140203 and #P2011280013 suggested that these fusion types did not follow the conventional splicing signal in the exon–intron boundary but instead formed novel fusion variants. We performed IHC in the case (#P2106140203) with some remaining tissue and verified ALK protein expression positivity (Fig. S2D).

Fig. 3. Integrative genomics viewer (IGV) screenshots, predicted or actual transcript and amino acid sequences of EML4-ALK detected using DNA-based NGS and RNA-based NGS. Blue, green, red and orange blocks represent the ‘C’, ‘A’, ‘T’ and ‘G’ bases, respectively. (A, C) DNA-NGS IGV screenshots and the predicted transcript and amino acid sequences of cases #P2106140203 and #P2011280013, respectively. (B, D) RNA-NGS IGV screenshots and the actual transcript and amino acid sequences of cases #P2106140203 and #P2011280013, respectively.

Moreover, three cases (#P2007070051: TOGARAM2 intron 8 : ALK intron 19; #P2005010014: SAMD12 intron 2 : ALK intron 19; #L-2018-00005429: AFF1 intron 2 : ALK intron 19) with noncanonical ALK rearrangements were detected by RNA-based NGS, and the results showed EML4-ALK fusions at the level of transcription (Table 4; Fig. S3A–C), indicating that the splicing alteration of the ALK frameshift rearrangements was different between EML4-ALK and the noncanonical ALK rearrangements.

Validation of the ALK rearrangement patterns with genomic breakpoints at the exons

In contrast to conventional genomic breakpoints of gene fusions that are located at introns, some partial breakpoints that led to gene fusions observed in this study occurred at exons. We detected the sequence of the fusion transcripts by RNA-based NGS from 13 available samples to explore the actual transcripts produced by the ALK rearrangements with genomic breakpoints located in the exons of ALK or its partner genes. Nine of the 13 samples were qualified by the RNA QC metric, including six cases (#P2006160041, #P2107150201, #P1911070042, #P2010230040, #P2009100057 and # P2008280124) with EML4-ALK and three cases (#P2009120038, #P2004270003 and #P2003200071) with noncanonical ALK rearrangements (Table 5; Table S2).

Table 5 Genomic breakpoints in exons cohort of ALK rearrangements/fusions at DNA, RNA, protein levels. ex, exon; int, intron.

Patient ID	DNA-based NGS	RNA-based NGS	IHC
P2006160041	EML4 ex3 : ALK int19	EML4 ex2 : ALK ex20	N/A
P2107150201	EML4 ex14 : ALK int19	EML4 ex13 : ALK ex20	Positive
L-2018-00002291	EML4 ex14 : ALK int19	Fail	Positive
P1901260022	EML4 ex14 : ALK int19	Fail	N/A
P1911070042	EML4 ex21 : ALK int19	EML4 ex20 : ALK ex20	N/A
L-2018-00010529	EML4 ex21 : ALK int19	Fail	Positive
P2010230040	EML4 int6 : ALK ex19	EML4 int6 : ALK ex20	N/A
P2009100057	EML4 int13 : ALK ex19	EML4 ex13 : ALK ex20	N/A
P2008280124	EML4 int19 : ALK ex20	EML4 ex19 : ALK ex20 (partial)	Positive
P2011280079	EML4 int20 : ALK ex19	Fail	N/A
P2009120038	PRR23C ex1 : ALK int19	KIF5B ex17 : ALK ex20	N/A
P2004270003	SQSTM1 ex5 : ALK int19	SQSTM1 ex4 : ALK ex20	N/A
P2003200071	SFTPB ex2 : ALK ex17	SFTPB ex1 : ALK ex18	N/A

For canonical EML4-ALK types, three cases (#P2006160041, #P2107150201 and #P1911070042) showed genomic breakpoints located in exons of EML4 (exon 3, exon 14 and exon 21), yet the actual transcript fusion sites were altered to exon 2, exon 13, and exon 20, respectively (Fig. S4A–C). The residual exons of these three cases were not retained, and the fusion sites were skipped to adjacent exons. Thus, these three variants were classified into the known variants V1 (E13; A20), V2 (E20; A20) and V5a (E2; A20). In two additional cases (#P2010230040 and #P2009100057), the genomic breakpoints were located at exon 19 of ALK, but the residual exon 19 was not retained in the actual transcripts and it skipped, fusing to the boundary of exon 20 (Fig. S4D,E). These two variants were classified into the known V3b (E6ins33; A20) and V1 (E13; A20) variants. In contrast, the genomic breakpoints of another case (#P2008280124) occurred in the region of exon 20 (Fig. 4A). The actual transcript revealed that partial exon 20 was not removed, thereby retaining 172 nucleotides (14 nucleotides deleted) or 166 nucleotides (20 nucleotides deleted) of exon 20 to form multiple isoform variants with an intact kinase sequence (E19; del14A20 and E19; del20A20) (Fig. 4B,C). In addition, some of the cases with remaining tissue samples were confirmed to be positive for ALK protein expression by IHC (Fig. S2E–H). Moreover, the transcript fusion sites of two noncanonical ALK rearrangement-positive cases (#P2004270003: SQSTM1 exon 5 : ALK intron 19; #P2003200071: SFTPB exon 2 : ALK exon 17), whose genomic breakpoints were located in exons, were skipped to adjacent exons (SQSTM1 exon 4 : ALK exon 20; SFTPB exon 1 : ALK exon 18), similar to the common exon skipping mode of EML4-ALK (Table 4; Fig. S3D,E). In addition, the genomic breakpoint of PRR23C was out of the coding sequence (located at the 5′UTR) in case #P2009120038 (PRR23C exon 1 : ALK intron 19), and the transcript changed to KIF5B exon 17 : ALK exon 20 (Table 5; Fig. S3F).

Fig. 4. IGV screenshots, predicted or actual transcript and the amino acid sequences of EML4-ALK (case #P2008280124) detected using DNA-based NGS and RNA-based NGS. Blue, green, red and orange blocks represent the ‘C’, ‘A’, ‘T’ and ‘G’ bases, respectively. (A) DNA-NGS IGV screenshot of the predicted transcript and amino acid sequences. (B, C) RNA-NGS IGV screenshots and the actual transcript and amino acid sequences of two different EML4-ALK fusion types.

Discussion

In this study, a total of 731 NSCLC cases with canonical EML4-ALK rearrangements and 52 NSCLC cases with noncanonical ALK rearrangements were identified. Among them, complex genomic breakpoints of ALK rearrangements were detected in the exons or introns of ALK and its partner genes. For rearrangements whose genomic breakpoints are located in exons, their transcripts cannot be inferred from conventional splicing signals. There are also some rearrangements that result in a frameshift transcript that cannot be translated into a fusion protein containing the amino acid sequence of ALK. Therefore, the actual transcripts of these ALK rearrangement types were verified by RNA-based NGS.

Frameshift of the fusion gene caused by chromosomal rearrangement is uncommon, especially in common carcinogenic-driven fusion mutations [31]. In this study, partial canonical and noncanonical ALK rearrangement-positive cases were speculated to exhibit frameshift possibilities based on DNA-based NGS data. For the canonical ALK rearrangements, the results of the actual transcripts were negative in two cases (#P2008100038 and #P1902170006) and positive in the other two cases (#P2106140203 and #P2011280013). To our knowledge, reports on ALK fusion frameshifts are rare, and only one case has been reported in detail. In this case, CMTR1-ALK (intron 2 : intron 19) was determined to be positive by DNA-based NGS, yet the patient did not respond to crizotinib treatment, and the expression of the ALK protein was negative by IHC [32]. Presumably, the two cases in this study with genomic-positive and transcript-negative EML4-ALK rearrangements will also not show a clinical response to ALK-targeted inhibitors. In addition, the insertion of diverse nucleotide sequences between the nearest fusion exons (#P2106140203 and #P2011280013) prevents frame shifts and maintains the functional transcription of EML4-ALK fusions, which is similar to the EML4-ALK variants reported in previous studies [33–39]. Alternative splicing caused by translocation can explain inserted or deleted nucleotide sequences to maintain a multiple of 3 required for a codon in frame to produce a functional protein. In contrast, transcripts of the canonical EML4-ALK fusions were detected in frame-shift cases with noncanonical ALK rearrangements (#P2007070051, #P2005010014 and #L-2018-00005429), which are associated with a complex mechanism of chromothripsis, resulting in posttranscriptional removal of other gene sequences that joined between ALK and EML4 [15,40]. Similarly, the transformation of PRR23C to KIF5B in case #P2009120038 may also be related to chromothripsis. Therefore, the results of complex genomic rearrangement events detected by DNA-based NGS may inaccurately reflect clinically actionable fusions [40]. Although our results showed the rarity of the predicted frameshift transcript of the ALK rearrangement pattern, further verification of these samples by RNA or protein assays is necessary to accurately diagnose patients at the molecular level who are candidates for targeted drug treatments.

Most genomic breakpoints of the rearranged gene occur in intronic sequences rather than in coding sequences [41]. According to conventional splicing principles, 5.6% of the rearrangement breakpoints were located in exonic regions of ALK or its partner genes in this study, and their predicted transcripts may be inaccurate or out-of-frame. Comparing the results from DNA-based NGS and RNA-based NGS, we found that exon skipping existed in ‘exon breakpoints’ cases carrying canonical or noncanonical ALK fusions. It may be reasoned that the lack of classical 3′ or 5′ accepter splice sites in the ‘exon–intron’, ‘intron–exon’ or ‘exon–exon’ structures resulted in the removal of the broken exon together with the previous intron to restore the reading frame [18,42]. Notably, although lacking the 5′ acceptor splice site of ALK exon 20, the actual transcript of case #P2008280124 excluded partial nucleotides and retained a portion of exon 20 through an alternative splicing signal at the RNA level rather than implementing exon skipping splicing and it resulted in two different variants (E19; del14A20 and E19; del20A20). Further comparative analysis of the transcripts and the amino acid sequences showed that partial retention of exon 20 could ensure the in-frame sequence with integrity of the ALK kinase domain [43,44]. Patients harbouring multiple EML4-ALK variants implied a poor prognosis due to the high heterogeneity in the tumour tissue [45]. Therefore, RNA-based NGS showed an advantage in detecting fusion patterns in which multiple variants coexist, and more precise splicing results at the transcription level were illustrated.

However, there are some limitations of this study. Due to the retrospective nature of this study, only a small number of tissue samples were available and met the quality control necessary for RNA sequencing, so only a few samples were verified by RNA sequencing. And for some patients with advanced lung cancer, their tissue samples could not be obtained, so blood samples were taken for DNA sequencing. Furthermore, the response of some patients with these uncommon ALK rearrangements to ALK inhibitors is unknown, large-scale validation of relationship between uncommon ALK rearrangements and treatments is necessary. In addition, the DNA NGS panel covers intronic regions where ALK rearrangements frequently occur and it may miss some rare intronic breakpoints. In the future, we will conduct more comprehensive clinical trials to explore the clinical benefits of these patients with specific ALK rearrangements from ALK inhibitor therapy.

Conclusions

In conclusion, by systematically analysing the DNA-based NGS data of ALK rearrangements in lung cancer patients, we identified variable and uncommon genomic breakpoints of ALK and its 5′partner genes. We further verified this finding by RNA-based NGS and found that the genomic breakpoints at the transcript level did not match those predicted by the genomic breakpoints; furthermore, we found that some of the fusions identified at the DNA level may be a false-positive. The ALK fusion results at the transcript level were better able to explain their functional significance. Therefore, the identification of ALK fusion status in NSCLC patients may need to use orthogonal assays based on multiomics for fusion detection to achieve an accurate molecular diagnosis and ensure the reliability of the targeted drug use indicators.

Acknowledgements

This work was supported by Scientific research project of Zhejiang province Department of Education (Y202044574); and the Natural Science Foundation of Zhejiang Province (LY18H160037), the Zhejiang Province Medical and Health Research Projects (grant numbers 2016KYB027, 2017KY010).

Conflict of interest

The authors declare no conflict of interest.

Author contributions

TM, WW, ZW and YH were involved in conception and design; HT, WL and JM were involved in administrative support; Zhifang Liu, Zhengchuang Liu, HT, JM and WL were involved in the provision of study materials or patients; Zhifang Liu, YH, MX, WW, WL and HT were involved in the collection and assembly of data; ZW, YH and JZ were involved in data analysis and interpretation; all authors were involved in manuscript writing and revising and final approval of manuscript.

Peer Review

The peer review history for this article is available at https://publons.com/publon/10.1002/1878-0261.13348.

Data accessibility

Data are available upon request.