Molecular characteristics and clinical outcomes of complex ALK rearrangements identified by next-generation sequencing in non-small cell lung cancers.

BACKGROUND: Complex kinase rearrangement, a mutational process involving one or two chromosomes with clustered rearrangement breakpoints, interferes with the accurate detection of kinase fusions by DNA-based next-generation sequencing (NGS). We investigated the characteristics of complex ALK rearrangements in non-small cell lung cancers using multiple molecular tests.
METHODS: Samples of non-small cell lung cancer patients were analyzed by targeted-capture DNA-based NGS with probes tilling the selected intronic regions of fusion partner genes, RNA-based NGS, RT-PCR, immunohistochemistry (IHC) and fluorescence in situ hybridization (FISH).
RESULTS: In a large cohort of 6576 non-small cell lung cancer patients, 343 (5.2%) cases harboring ALK rearrangements were identified. Fourteen cases with complex ALK rearrangements were identified by DNA-based NGS and classified into three types by integrating various genomic features, including intergenic (n = 3), intragenic (n = 5) and "bridge joint" rearrangements (n = 6). All thirteen cases with sufficient samples actually expressed canonical EML4-ALK fusion transcripts confirmed by RNA-based NGS. Besides, positive ALK IHC was detected in 13 of 13 cases, and 9 of 11 cases were positive in FISH testing. Patients with complex ALK rearrangements who received ALK inhibitors treatment (n = 6), showed no difference in progression-free survival (PFS) compared with patients with canonical ALK fusions n = 36, P = 0.9291).
CONCLUSIONS: This study firstly reveals the molecular characteristics and clinical outcomes of complex ALK rearrangements in NSCLC, sensitive to ALK inhibitors treatment, and highlights the importance of utilizing probes tilling the selected intronic regions of fusion partner genes in DNA-based NGS for accurate fusion detection. RNA and protein level assay may be critical in validating the function of complex ALK rearrangements in clinical practice for optimal treatment decision.

Background

Rearrangements of the anaplastic lymphoma kinase (ALK) gene have been identified in approximately 3–7% of non-small cell lung cancer (NSCLC) patients, with echinoderm microtubule-associated protein like-4 (EML4) representing the most common fusion partner [1, 2]. ALK-rearranged NSCLC define a distinct molecular subset with high sensitivity to ALK tyrosine kinase inhibitors (TKIs). Crizotinib, a well-tolerated first generation ALK inhibitor [3, 4], has been approved by Food and Drug Administration in US for the treatment of ALK-rearranged NSCLC in 2011. Second generation ALK inhibitors, such as alectinib and ceritinib, are effective not only in crizotinib-naive patients [5], but also in patients with acquired resistance to crizotinib [6-9]. The identification of ALK rearrangements and the approval of a number of ALK TKIs have revolutionized the treatment of patients harboring ALK fusions. Therefore, accurate detection for ALK rearrangements is crucial.

A challenge for precision oncology is identifying novel or complex translocations. Traditional methods, including fluorescence in situ hybridization (FISH) and immunohistochemistry (IHC), have limitations, such as FISH does not permit identification of ALK partner genes or non-canonical breakpoints, and ALK IHC could be confounded in principle by overexpression of ALK driver rather than a true fusion protein [10]. While next-generation sequencing (NGS) techniques provide an effective and accurate detection for known and novel oncogenic fusions, and have been widely applied in clinical diagnostics.

Complex kinase rearrangements, herein referred to a mutational process involving one or two chromosomes with clustered rearrangement breakpoints. Recent studies have revealed that complex genomic rearrangements generated 74% of known fusion oncogenes in human lung adenocarcinoma of non-smokers, including EML4-ALK, CD74-ROS1, and KIF5B-RET [11]. However, complex genomic rearrangements frequently hindered proper capture in DNA-based NGS assay [12]. Accumulating evidences have suggested that genomic breakpoints identified by DNA sequencing are an unreliable predictor of breakpoint at the transcript level owing to genomic complexities [13, 14]. The identification and clinically functional validation of complex kinase rearrangements remain elusive, which makes the oncologists confused to choose the appropriate treatments. A combination methodology of DNA-based NGS technique followed by RNA-based NGS provides a unique opportunity to explore the mutational processes in cancer genomes. Although there has been landmark study characterizing the complex intergenic-breakpoint fusions [14], it was largely based on exome and selected introns in ALK gene. It is lack of the study using DNA-based NGS designed for intronic regions from fusion partner genes known to likely harbor the genomic breakpoint.

Herein, we utilized DNA-based NGS panel specifically designed with multiple probes tilling selected intronic regions of fusion partner genes, and identified three types of complex ALK rearrangements in 14 cases from a large cohort of NSCLC patients. Further functional validation performed by RNA or protein assay elucidated the importance of DNA and RNA-based NGS for the comprehensive detection of kinase fusions and guiding optimal treatment decision.

Methods

Patients and samples

Samples from a cohort of 6576 patients with NSCLC from January 2018 to July 2020 were collected for molecular testing. Pathological and clinical information was obtained from clinical records. The study was approved by the Institutional Review Board of the First Affiliated Hospital of Zhengzhou University. All patients provided informed written consent for these genomic analyses.

DNA/RNA extraction

The pathological diagnosis of each case was confirmed on routine hematoxylin and eosin stained slides, and the corresponding optimal blocks containing a minimum of 20% tumor cells were forwarded for DNA/RNA extraction. Genomic DNA (gDNA) and total RNA were extracted from the formalin-fixed paraffin-embedded (FFPE) tumor tissue samples using AllPrep DNA/RNA FFPE Kit (Qiagen, USA) according to the manufacturer’s instructions. As a control, gDNA from the white blood cell samples was extracted using MagPure Blood DNA DA Kit (Magen, China) according to the manufacturer’s instructions. The quality of purified DNA/RNA were assayed by gel electrophoresis and quantified by Qubit® 4.0 Fluorometer (Life Technologies, USA). The amounts of extracted DNA more than 30 ng were considered sufficient for analysis. In the extracted FFPE RNA samples, the 28S and 18S rRNA bands were degraded, and ≥ 200 ng RNA were optimal for high analytical sensitivity.

DNA-based NGS

The purified gDNA was first fragmented into DNA pieces about 300-bp using enzymatic method (5X WGS Fragmentation Mix, Qiagen, USA), followed by end repairing, T-adaptors ligation, and PCR amplification, resulting in pre-library. An in-house designed panel targeting most exons and selected introns in 86 cancer-related genes was used to capture DNA fragments to detect SNV/Indel, copy number variation and gene fusions (Additional file 1: Table S1). Particularly, hybrid capture-probes tilling the intronic regions of ALK (intron 18–19), EML4 (intron 6, 13, 20) and KIF5B (intron 15–16, 24) were designed for the detection of ALK rearrangements event. Sequencing libraries were generated after PCR amplification and then sequenced on NovaSeq 6000 platform (Illumina, San Diego, USA) with 150PE mode.

Initial read mapping against the human reference genome hg19 and alignment processing was performed using BWA [15]. SAMtools [16] and Genome Analysis Toolkit GATK 3.8 [17] were used to call SNVs and small indel variants. Large indels and chromosomal rearrangements (including ALK rearrangements) were analysed using Fusionmap [18]. The nonsynonymous SNVs with VAF > 0.5% or with VAF > 0.1% in cancer hotspots collected from patient database were kept for the further analysis. Fusions with coverage ≥ 300 and supported mutation reads number ≥ 3 were identified and reported. For breakpoints in intergenic regions, the nearest gene in each direction was reported as the predicted fusion partner.

RNA-based NGS

An in-house designed RNA fusion panel based on hybrid capture sequencing (Berry Oncology Corporation) was performed to detect gene fusions, which tilling all coding exons of common fusion genes in cancer and allowing for detection of known and novel fusions without a limitation for fusion partner or breakpoint. Briefly, the purified total RNA was first converted to complimentary DNA (cDNA) through reverse transcription reaction. The pre-libraries construction consisted of end repairing, adaptor ligation and PCR amplification, which of the total amounts was optimized for a desired value of ≥ 600 ng. The follow-up hybridization-captured libraries were sequenced on NovaSeq 6000 platform (Illumina, San Diego, USA) with paired-end 150-bp reads. Gene fusions were called based on Fusionmap software [18]. Bioinformatically identified fusions were verified by manual inspection of the breakpoints.

FISH

In brief, FFPE tumor tissue samples was analyzed by FISH using the Vysis LSI ALK Dual color, Break Apart Rearrangement Probe (Abbott/Vysis, Abbott Park, IL, USA). In 50 scored tumour cells of every sample, if more than 15% of the scored tumour cells had split one or both ALK 5′ and 3′ probe signals or had isolated 3′ signals, the sample was considered to be FISH positive. Every FISH slide was evaluated by two pathologists independently.

IHC

Immunohistochemistry of ALK protein was performed on a fully automated Ventana Benchmark XT stainer (Ventana Medical Systems, Roche Group, Tucson, AZ). FFPE tumor samples were stained using the pre-diluted Ventana anti-ALK (D5F3) Rabbit monoclonal primary antibody and a matched Rabbit Monoclonal Negative Control Ig antibody, together with the Optiview DAB IHC detection kit and Optiview Amplification kit. Every IHC slide was evaluated by two pathologists independently. Neoplastic cells labeled with the ALK IHC assay are evaluated for presence or absence of the DAB signal according to the method previously described [19]. If strong granular cytoplasmic staining was observed in any tumor cells at any percentage, the sample was considered to be ALK positive, while the sample without strong granular cytoplasmic staining in tumor cells was considered to be ALK negative.

Clinical response evaluation and statistical analysis

For a subset of patients who received targeted ALK inhibitors treatment, clinical responses were assessed based on computed tomography (CT) imaging, following the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. The association of patient characteristics and clinicopathological factors was investigated by the chi-square test. Progression-free survival (PFS) was calculated using the Kaplan–Meier method and differences in variables using the log-rank test. A two-sided P < 0.05 was considered to be statistically significant. Statistics were analyzed using GraphPad Prism (version 7.04).

Results

Characteristics of patients and ALK rearrangements

All of 6576 samples from NSCLC patients were profiled with DNA-based NGS between January 2018 and July 2020. The clinical characteristics of the patients are described in Table 1. ALK fusions were identified in 343 (5.2%) cases with higher incidences in female, age < 60 or adenocarcinoma patients. Canonical EML4-ALK fusions occurred most frequently accounting for 78.4% (269/343). Most of the genomic breakpoints of the ALK gene were detected within intron19, while the EML4 potential breakpoints differ and may generate various fusion protein variants at the genomic level. As shown in Fig. 1, EML4-ALK variant 3 (E6:A20, 109/269, 40.5%) was the most predominant type, followed by variant 1 (E13:A20, 77/269, 28.6%) and variant 2 (E20:A20, 35/269, 13.0%).

Table 1

Characteristics of NSCLC patients subjected to DNA-based NGS

Patient	Total no	ALK Gene fusions	P
Gender
Male	3699	166 (4.5%)	0.002601
Female	2877	177 (6.2%)
Age
≥ 60	3704	107 (2.9%)	< 0.001
< 60	2872	236 (8.2%)
Histology
Adenocarcinoma	4897	311 (6.4%)	< 0.001
Squamous	982	7 (0.7%)
Other*	697	25 (3.6%)
Sample type
Surgical	1689	98 (5.8%)	0.3983
Biopsy/cell block	4571	239 (5.2%)
Liquid biopsy	316	6 (1.9%)

*Other carcinomas included adenosquamous carcinoma (n = 76), large cell cancer (n = 17), large cell neuroendocrine carcinoma (n = 25), neuroendocrine carcinoma (n = 22), sarcomatoid carcinoma (n = 36), non-small cell lung cancer-not otherwise specified (n = 287) and unknown (n = 234)

Fig. 1

Schematic diagrams and distribution frequency of EML4-ALK fusion variants in the study cohort (N = 269). Abbreviation: E: EML4 exon; A: ALK exon

Identification and validation of complex ALK rearrangements

Among the 343 ALK fusion cases, complex ALK rearrangements in 14 cases were identified using targeted DNA-based NGS across 86 cancer-related genes panel with multiple probes tilling selected intronic regions of fusion partner genes (Table 2). These cases could be divided into three types by integrating various genomic features, including intergenic (n = 3), intragenic (n = 5) and “bridge joint” rearrangements (n = 6). A subset of 13 cases retained enough specimens were validated for additional RNA-based NGS tilling all coding exons of common fusion genes. Surprisingly, we found that the fusion genes and breakpoint positions had significant discrepancies between DNA and RNA sequencing. All thirteen cases actually expressed canonical EML4-ALK fusion transcripts. Besides, positive ALK IHC was detected in 13 of 13 cases, and 9 of 11 cases were positive in FISH testing.

Table 2

Comparison of DNA-based NGS, RNA-based NGS, IHC and FISH results among 14 lung cancer cases

Case	Sex	Age	Diagnosis	DNA-based NGS	RNA-based NGS	IHC	FISH
1	M	54	ADC	EML4-intergenic (intron13: intergenic) WDR43-ALK (3′ intron1: 3′ intron19)	EML4-ALK (exon13: exon20)	ALK+	N/A
2	F	32	ADC	intergenic-EML4 (intergenic: intron6) ALS2CR11-ALK (3′ intron9: 3’intron19) ALK-RBKS (intron19: intron5)	EML4-ALK (exon6: exon20)	ALK+	+
3	M	45	ADC	EML4-intergenic (intron6: intergenic) PLXNA4-ALK (3′ intron3: 3′ intron19) intergenic-EML4 (intergenic: intron6)	EML4-ALK (exon6: exon20)	ALK+	+
4	M	28	ADC	EML4-ALK (intron13: intron3) ALK-ALK (intron3ins9: intron19) GALM-EML4 (intron4: intron13) ALK-intergenic (intron19: intergenic)	EML4-ALK (exon13: exon20)	ALK+	+
5	M	63	ADC	EML4-ALK (5′ intron6: 5′ intron19) ALK-ALK (3′ intron18: 3′ intron19) IL1RAPL2-ALK (5′ intron6: 5′ intron18) ALK-PTCHD1 (3′ intron18: 3′ intron1)	EML4-ALK (exon6: exon20)	ALK+	+
6	F	54	ADC	EML4-ALK (intron6ins39: intron4) ALK-ALK (intron4ins57: intron19)	N/A	N/A	N/A
7	F	54	ADC	EML4-EML4 (5′ intron6: 5′ intron6) EML4-ALK (3′ intron6: 3′ intron19)	EML4-ALK (exon6: exon20) EML4-ALK (exon6ins33: exon20)	ALK+	+
8	M	50	ADC	EML4-EML4 (5′ intron6: 5′ intron6) EML4-ALK (3′ intron6ins4: 3′ intron19)	EML4-ALK (exon6: exon20)	ALK+	+
9	M	75	ADC	EML4-LCLAT1 (5′ intron13: 5′ intron1) LCLAT1-ALK (3′ intron1: 3′ intron19)	EML4-ALK (exon13: exon20)	ALK+	−
10	M	59	ADC	EML4-NEB (intron13: intron24) NEB-ALK (intron24: intron19)	EML4-ALK (exon13: exon20)	ALK+	+
11	F	44	ADC	EML4-KCNQ3 (intron6: intron1) KCNQ3-ALK (intron1: intron19) ALK-EML4 (intron19: intron6)	EML4-ALK (exon6: exon20)	ALK+	+
12	F	43	ADC	EML4-RUNX1 (intron5: intron5) RUNX1-ALK (intron6: intron19)	EML4-ALK (exon5: exon20) RUNX1-ALK (exon6: exon20)	ALK+	−
13	M	56	ADC	EML4-ZNF362 (intron13: intron1) ZNF362-ALK (intron1: intron19)	EML4-ALK (exon13: exon20)	ALK+	N/A
14	M	31	ADC	EML4-RPIA (intron13: intron3) MAP4K3-ALK (intron2: intron19)	EML4-ALK (exon13: exon20)	ALK+	+

F: female; M: male; ADC: adenocarcinoma; N/A: not available; +: positive results; −: negative results

Case 1, a representative intergenic complex rearrangement case, harbored WDR43-ALK (3′ intron1: 3′ intron19) and EML4-intergenic fusions identified by DNA-based NGS (Fig. 2A), with positive results detected using ALK IHC assay (Fig. 2D), but RNA-based NGS detected the canonical EML4-ALK fusion transcript joining EML4 exon 13 to ALK exon 20 (Fig. 2B). Sequencing data indicated that the intergenic complex rearrangement involved multiple fusion junctions, comprising EML4, LINC01913 upstream intergenic region, WDR43 and ALK (Fig. 2C). Evidences of such intergenic complex rearrangements were also detected in case 2 and case 3 by DNA-based NGS, which harbored a canonical EML4-ALK variant 3 (E6:A20) transcript identified by RNA sequencing, and were positive by IHC and FISH assays (Table 2).

Fig. 2

Representative example of intergenic complex rearrangements generating EML4-ALK fusion transcripts from case 1. A DNA sequencing reads indicating WDR43-ALK and EML4-intergenic fusion regions were visualized by the Integrative Genomic Viewer (IGV) software. B RNA sequencing reads indicating EML4-ALK fusion regions were visualized by the IGV software. C Possible schematic diagram of WDR43-ALK and EML4-intergenic fusions that were detected by DNA-based NGS, but a EML4-ALK fusion transcript was identified by RNA-based NGS. Red dashed lines indicate fusion breakpoints and red arrows indicate direction of transcription. D Positive ALK expression detected by IHC assay

The more remarkable observation, shared by cases 4–8, was the rare and complicated intragenic rearrangement of ALK or EML4 gene identified at DNA level. Case 4 typically harbored multiple distinct rearrangements involving ALK locus, consisting of 5′ EML4 (intron 13) and 3′ ALK (intron 3) fusion, ALK-ALK fusion in which intron 3 of ALK was jointed to intron 19 of ALK with a 9-bp insertion, GALM-3′ EML4 fusion, and 5′ ALK-intergenic fusion (Fig. 3A). Only the first two connecting fusion-oncogene-associated rearrangements appeared capable of producing a functional pathogenic fusion transcript joining EML4 exon 13 to ALK exon 20 detected by RNA-based NGS data (Fig. 3B and C). The other two fusions without transcription product may be the reciprocal fusions. Meanwhile, clear split signals of ALK gene were detected by FISH using a break-apart probe kit (Fig. 3D), and IHC test of the surgically resected sample revealed a positive result (Fig. 3E). Similarly, case 6 harbored 5′ EML4 (intron 6) and 3′ ALK (intron 4) fusion with a 39-bp insertion and ALK-ALK fusion in which intron 4 of ALK was jointed to intron 19 of ALK with a 57-bp insertion, indicating to product the canonical EML4-ALK variant 3 (E6:A20) transcript without enough specimens for validation assays. In case 5, a special inversion of ALK gene from intron 18 to intron 19 was detected, in which 3′ intron 18 of ALK was jointed to 3′ intron 19 of ALK and 5′ intron 19 of ALK was jointed to 5′ intron 6 of EML4 and RNA-based NGS detected the canonical EML4-ALK variant 3 (E6:A20) transcript. Similarly, inversions of EML4 intron 6 were identified in case 7 and 8, which also harbored the canonical EML4-ALK variant 3 (E6:A20) transcript and were IHC and FISH positive (Table 2).

Fig. 3

Representative example of intragenic complex rearrangements generating EML4-ALK fusion transcripts from case 4. A DNA sequencing reads indicating EML4 (intron13)-ALK (intron3) fusion and ALK (intron19)-ALK (intron3) fusion regions were visualized by the IGV software. B RNA sequencing reads indicating EML4-ALK fusion regions were visualized by the IGV software. C Possible schematic diagram of genomic rearrangements involving the fusion breakpoints in DNA and RNA level. DNA-based NGS detected that intron 13 of EML4 fused with intron 3 of ALK, and intron 3 of ALK jointed to intron 19 of ALK with a 9-bp insertion. RNA-based NGS detected the canonical fusion transcript joining EML4 exon 13 to ALK exon 20. Red dashed lines indicate fusion breakpoints and red arrows indicate direction of transcription. D Positive ALK FISH pattern with separate red and green signals. E Positive ALK expression detected by IHC assay

In cases 9–13, multiple gene fusions were identified, herein defined as “bridge joint” rearrangements owing to that both EML4 and ALK jointed with an identical gene at the genomic level, respectively. Taking case 9 for example, DNA-based NGS detected that the intron 13 of EML4 fused with the downstream region of intron 1 of LCLAT1, and the upstream region of intron 1 of LCLAT1 joined to the intron 19 of ALK (Fig. 4A and C). Due to intronic splicing, it is reasonable that RNA-based NGS identified the canonical EML4-ALK variant 1 (E13:A20) transcript without intron1 of LCLAT1 (Fig. 4B and C). IHC assays showed clearly positive ALK protein expression, but FISH revealed negative results, perhaps due to break-apart probe design or technical aspects yielding a risk of false-negative result (Fig. 4D and E) [20]. Similarly, in cases 10–13, DNA-based NGS revealed that the intron of EML4 fusion partner gene firstly joined to the intronic region of a novel “bridge” gene and then to the intron of ALK kinase gene, as “bridge joint” complex rearrangements. Most of the intronic regions of the novel “bridge” genes were removed by splicing, leading to canonical EML4-ALK transcripts. In particular, the exon6 of “bridge” gene (RUNX1) was involved in the complex rearrangement and the RUNX1-ALK (exon6: exon20) transcript was detected in case 12, which may be part of the EML4-RUNX1-ALK (exon5: exon6: exon20) transcript hardly to be identified. Moreover, the EML4- ALK (exon5: exon20) transcript was also detected in case 12, perhaps due to the alternative splicing. Interestingly, case 14, harboring EML4-RPIA and MAP4K3-ALK fusions, was identified as the canonical EML4-ALK variant 1 (E13:A20) transcript, suggesting that RPIA and MAP4K3 were both the “bridge” genes and their intronic regions were connected together (Table 2).

Fig. 4

Representative example of “bridge joint” rearrangements generating EML4-ALK fusion transcripts from case 9. A DNA sequencing reads indicating EML4-LCLAT1 and LCLAT1-ALK fusion regions were visualized by the IGV software. B RNA sequencing reads indicating EML4 and ALK fusion regions were visualized by the IGV software. C Possible schematic diagram of EML4-LCLAT1 and LCLAT1-ALK fusions that were detected by DNA-based NGS, but a EML4-ALK fusion transcript was identified by RNA-based NGS. Red dashed lines indicate fusion breakpoints and red arrows indicate direction of transcription. D Negative ALK FISH pattern with combined red and green signals. E Positive ALK expression detected by IHC assay

Targeted therapies and clinical outcomes of complex ALK rearrangements

Among the 14 cases with complex ALK rearrangements, only 8 patients received targeted ALK inhibitors (crizotinib or alectinib) treatment, including 2 intergenic complex rearrangements, 3 intragenic complex rearrangements and 3 “bridge joint” rearrangements. Treatment and response to therapy, as defined by RECIST v1.1, were outlined in Table 3, which showed that 6 patients (75%) achieved clinical objective response, including 5 partial responses (PR) and 1 complete response (CR).

Table 3

Clinical outcomes in patients with complex ALK rearrangements who received targeted therapy

Case	DNA-based NGS	RNA-based NGS	IHC	FISH	Other variants	Matched treatment	Optimal response
1	EML4-intergenic (intron13: intergenic) WDR43-ALK (3′ intron1: 3′ intron19)	EML4-ALK (exon13: exon20)	ALK+	N/A	None	Crizotinib	PR
3	EML4-intergenic (intron6: intergenic) PLXNA4-ALK (3′ intron3: 3′ intron19) intergenic-EML4 (intergenic: intron6)	EML4-ALK (exon6: exon20)	ALK+	+	ALK p.T1012M ROS1 p.E1902K TP53 p.R337C TSC2 p.A678T	Crizotinib	PR
5	EML4-ALK (5′ intron6: 5′ intron19) ALK-ALK (3′ intron18: 3′ intron19) IL1RAPL2-ALK (5′ intron6: 5′ intron18) ALK-PTCHD1 (3′ intron18: 3′ intron1)	EML4-ALK (exon6: exon20)	ALK+	+	GNAS p.A175T TP53 splicing pathogenic	Alectinib	CR
6	EML4-ALK (intron6ins39: intron4) ALK-ALK (intron4ins57: intron19)	N/A	N/A	N/A	ROS1 p.E1902K SMAD4 p.N129H	Crizotinib	PR
8	EML4-EML4 (5′ intron6: 5′ intron6) EML4-ALK (3′ intron6ins4: 3′ intron19)	EML4-ALK (exon6: exon20)	ALK+	+	TP53 p.R273C NOTCH3 p.A38Lfs	Crizotinib	PD
10	EML4-NEB (intron13: intron24) NEB-ALK (intron24: intron19)	EML4-ALK (exon13: exon20)	ALK+	+	None	Crizotinib	PR
13	EML4-ZNF362 (intron13: intron1) ZNF362-ALK (intron1: intron19)	EML4-ALK (exon13: exon20)	ALK+	N/A	None	Alectinib	PR
14	EML4-RPIA (intron13: intron3) MAP4K3-ALK (intron2: intron19)	EML4-ALK (exon13: exon20)	ALK+	+	None	Crizotinib	SD

N/A: not available; +: positive results; −: negative results; CR: complete response; PR: partial response; SD: stable disease; PD: progression disease

Case 1 and case 3 with intergenic complex rearrangements both had positive response to crizotinib, and the endpoint of progression-free survival (PFS) was still not reached, lasted for at least 8 and 7 months, respectively (Table 3). Differential ALK inhibitor responses were observed among intragenic rearrangements variants in ALK-positive lung adenocarcinoma (case 5, case 6 and case 8). The identical EML4-ALK fusion cases 5 and 8, both got transcript joining EML4 exon 6 to ALK exon 20 and positive results in FISH and IHC, achieved quite discrepant clinical outcome to crizotinib, CR for case 5 and progressive disease (PD) after 5 months treatment for case 8. We speculated that the other variant occurred in case 8, TP53 p.R273C mutation, enhanced cancer cell proliferation, invasion and drug resistance [21]. As to the “bridge joint” rearrangements, one of the three cases, case 14, exhibited stable disease (SD) 4 weeks after crizotinib treatment (Table 3), different with the PR states of the other two cases, which implied that the poor clinical outcomes for ALK inhibitor in some patients could be caused by primary drug resistance to targeted therapies [22].

Patients with complex ALK fusions (n = 5) received crizotinib treatment exhibited comparable median progression-free survival (mPFS) with patients harboring canonical ALK fusions (n = 34), which displayed in Fig. 5A with the values of 7.0 months (95% CI 0.3–2.0) versus 9.0 months (95% CI 0.5–3.3) and the P value of 0.7616. Similarly, no significant difference in mPFS was observed between complex and canonical ALK fusions when patients with alectinib and crizotinib treatment were analyzed together (8.0 months [95% CI 0.4–2.1] versus 9.0 months [95% CI 0.5–2.7], P = 0.9291, Fig. 5B).

Fig. 5

Survival curves for complex ALK fusion subtype and canonical ALK fusion subtype detected by DNA-based NGS. A Survival curves for patients with crizotinib treatment with complex ALK fusions (n = 5) and canonical ALK fusions (n = 34). B Survival curves for patients with crizotinib or alectinib treatment with complex ALK fusions (n = 6) and canonical ALK fusions (n = 36)

Discussion

In this study, we identified three types of complex ALK rearrangements, intergenic complex rearrangements, intragenic complex rearrangements and “bridge joint” rearrangements. The complex ALK rearrangements could be attributed to a distinct mechanism, termed chromothripsis, happened at least 2 to 3% of all cancers and often promoted tumorigenesis in a wide variety of tumors [23]. Figures 2, 3 and 4 showed a possible mutational process mediated by inversion and chromothripsis. It caused a one-off chromosome breakage and subsequent random reassembly of the chromosome fragments, resulting in ALK and EML4 joining to the intergenic/intragenic/ “bridge joint” regions respectively which were removed during transcription and generating the canonical EML4-ALK oncogenic fusions. ALK break-apart FISH analysis showed that there was more aberrant chromosome 2 fragmented and scattered in tumor cells, probably because chromosome structure was damaged severely by chromothripsis (Fig. 2D). Besides chromothripsis, translocation was reported as a novel mechanism of intragenic ALK rearrangements in neuroblastoma tumors in 2014 [24]. Another recent study revealed that intragenic complex rearrangements were related to RB1 inactivation in EGFR-mutant lung cancer cell [25]. Here, we detected 5 cases intragenic complex rearrangement in lung adenocarcinoma, including 3 intragenic ALK rearrangements and 2 intragenic EML4 rearrangements, which all generated canonical EML4-ALK fusion transcripts except for one not performed RNA-based NGS testing owing to insufficient specimen.

NGS technology has been widely used in rearrangements detection first in DNA-based method, which becomes a first-line pathological methodology in high-income country, such as US. Sampling requirement and quality metrics of DNA-based NGS is not as strict as RNA-based approach. One case in our study was not successfully performed RNA-based NGS because specimen cannot meet RNA stricter quality standards. DNA-based NGS can identify genomics rearrangements not limited to fusion, such as amplification of the ALK locus, which reveals a novel truncated form and activates drivers but not lead to fusion transcripts and proteins [26]. The targeted-capture DNA-based NGS panel are usually designed to target exonic and selected intronic regions of kinase genes, which has high probability to harbor the genomic breakpoints and could effectively identify kinase fusions. However, DNA-based NGS has some inherent limitations when targeted-capture introns are too long, or contain repetitive elements or involve complex genomic events [27]. The genomic rearrangement couldn’t be fully captured by DNA-based NGS panel when oncogenic fusion is caused by one or more complex DNA rearrangements. In contrast, RNA-based NGS offers a more direct approach to detect clinically actionable fusions, as RNA sequencing focused on exons post-splicing which may bypass genomic complexities [27]. As currently the most comprehensive and efficient strategy for exact fusion transcripts detection, RNA-based NGS testing is widely applied in the molecular diagnosis of gene fusion [28]. In our cohort, complex ALK rearrangements expressing canonical EML4-ALK fusion transcripts had been detected in 13 cases by DNA and RNA-based NGS. Genomic breakpoints within intronic regions of EML4 were involved in the complex ALK rearrangements, which hardly detected by common NGS panel without probes capturing EML4 introns. Using optimized probes tilling the selected intronic regions of EML4, genomic breakpoints within intronic regions of EML4 were detected clearly by DNA-based NGS and illuminated the whole possible structures of the complex ALK rearrangements. Our finding suggested that it may be critical to utilize DNA-based NGS with optimized probes tilling the selected intronic regions of fusion partners followed by RNA-based NGS, which could effectively identify accurate oncogenic rearrangements and comprehensively guide optimal treatment decision not just in lung cancer but also across different types of tumor. Furthermore, there were 2 patient samples (case 9 and case 12) with discordant results between FISH and other assays. The discrepancy between multiple molecular testing could be considered as a ‘wake-up call’ for oncologists to ensure more accurate molecular diagnosis by identifying and functionally validating the clinically relevant complex genomic rearrangements.

Crizotinib, an oral small-molecule tyrosine kinase inhibitor (TKI) targeting ALK, MET, and ROS1 tyrosine kinases, has been approved for ALK-rearranged NSCLC in USA, European Union, China and other countries, with objective response rate (ORR) of proximately 60.8% and median progression-free survival (mPFS) of 9.7 months [29]. Besides crizotinib, multiple second-generation (such as alectinib and ceritinib) ALK-TKIs have been developed for patients with ALK-positive NSCLC, all with higher potency than crizotinib [30-33]. Although ALK-TKI has dramatically expanded the therapeutic landscape of ALK-positive NSCLC, the substantial question, whether patients harboring complex genomic rearrangements could benefit from this target therapy, is not fully defined. Kodama et al. confirmed that alectinib and crizotinib were both effective against EML4-ALK-positive tumors mediated by chromothripsis in a patient-derived cell line, and the potency of alectinib was approximately 13-fold higher than crizotinib [23]. In our follow-up clinical data, 8 cases harboring complex ALK rearrangements showed the optimal responses with 1 CR (alectinib treatment), 5 PR (1 alectinib treatment and 4 crizotinib treatment), 1 SD (crizotinib treatment), and 1 PD (crizotinib treatment). It seemed that alectinib had a more remarkable response to complex ALK rearrangements than crizotinib in this “real world” data set. However, more studies should be performed in the future to verify the results with larger cohorts.

It was interested that case 8 achieved quite discrepant clinical outcome to crizotinib compared with case 5, PD after 5 months treatment for case 8 vs CR for case 5. Both patients had the identical rare and complicated intragenic EML4-ALK rearrangement detected in DNA and RNA-based NGS, and positive in FISH and IHC. The possible reason for the poor outcome might be the TP53 p.R273C mutation detected in DNA-based NGS, which have been reported to enhance cancer cell proliferation, invasion and drug resistance [21]. It was reasonable that no significant difference was found in mPFS between the patients carrying complex and canonical ALK fusions regardless of their first-line treatment, crizotinib or alectinib, as all of them generated canonical EML4-ALK transcripts in RNA level. Limitation of the survival analysis in this study includes that the number of cases with complex ALK fusions receiving targeted therapy is relatively small (Additional file 1).

Conclusions

This study firstly reveals the molecular characteristics and clinical outcomes of complex ALK rearrangements in NSCLC, sensitive to ALK inhibitors treatment, and highlights the importance of optimizing probe design of NGS panel for tilling the selected intronic regions of fusion partner genes. The discordant results of complex ALK rearrangements between DNA and RNA-based NGS indicate that DNA and RNA-based NGS assay should be both warranted in fusion detection. RNA and protein level assay may be critical in validating the function of complex ALK rearrangements in clinical practice for optimal treatment decision.

Additional file 1: Table S1. The list of genes in DNA-based NGS panel.