Targeted RNA-sequencing for the evaluation of gene fusions in lung tumors: current status and future prospects

1. Introduction

Routinely, patients with locally advanced or metastatic non-small cell lung cancer (NSCLC) are screened for mutations and fusions to direct therapy decision. The majority of molecular laboratories perform a multitude of different tests for the work up of all biomarkers. Especially with the fast emerging development of new drugs, testing for specific fusion genes has gained much attention within the last years.

RNA-Seq can be achieved by either comprehensive whole transcriptome- or targeted approaches [1]. Both allow the expression analysis of a defined gene set. Targeted RNA-sequencing, as a method for enrichment-sequencing of specific transcripts, is of growing importance in molecular diagnostics. Depending on the technology used, enrichment-based approaches further enable the detection of known and novel gene fusion events. In comparison with whole-transcriptome RNA-Seq, targeted RNA-Seq, focuses only on selected transcripts, thereby allowing increased sensitivity, reduced costs as well as increased throughput opportunities [2].

According to the statement of the ESMO Precision Medicine Working Group, multigene sequencing on DNA or RNA level is highly recommended to increase patient´s access to approved or innovative drugs and speeding up clinical research [3].

2. Current status of fusion detection in NSCLC

Since the first discovery of chromosomal rearrangements as oncogenic drivers, a large number of fusions in various tumor types could be detected through technological advancements [4]. The method, which is currently most widely used, for formalin-fixed, paraffin-embedded (FFPE) tissue is in situ hybridization (ISH). Due to its high sensitivity, specificity and short turnaround time, this assay is still the gold standard in most laboratories. However, ISH analysis suffers from resolution limits and complex rearrangements are not easily detectable and can lead to false-negative results [5].

Regarding fusions leading to an overexpression of the target gene, immunohistochemistry (IHC) can be a time saving and inexpensive tool to perform screening for rearrangements by protein localization. However, there are no antibodies for all clinically relevant fusions. Additionally, this approach has a high rate of false-positive events and a second method is needed to verify fusions found by IHC [1].

Both approaches, ISH and immunohistochemistry, have the major disadvantage to lack multiplexing opportunities. Another drawback of ISH and IHC is that at least one fusion partner has to be already known or assumed. Nonetheless, the identification of unknown fusion partners is becoming increasingly important because the partners can influence treatment choices and can be of prognostic value [6].

Next-Generation Sequencing (NGS) can identify known and unknown fusions on DNA and RNA level including both fusion partners in a single multiplex approach and therefore seems to be more promising.

3. Comprehensive technologies: from WGS to WTS

Among NGS strategies, whole-genome sequencing (WGS) can be used to find a greater number of rearrangements and discover novel fusions. It is possible to characterize exact breakpoints, including those in non-coding regions. However, WGS is expensive and time-consuming due to its enormous data output and intensive analysis [7]. An alternative to WGS can be whole-exome sequencing (WES). Nevertheless, only rearrangements with breakpoints near or in exons can be detected. Since that is true for only a few events, WES has no meaning in fusion detection [8].

As with WGS, whole-transcriptome sequencing (WTS) is sufficient to discover several gene fusions but is at the same time more efficient and sensitive [9]. Even with these advantages, WTS is still too expensive and data analyses and interpretation requires specialized personnel.

4. DNA vs RNA- based targeted sequencing

Besides approaches analyzing large areas of the genome, targeted sequencing can be used to isolate and sequence a subset of genes and detect fusions in a more distinct manner. By enrichment, targeted sequencing results in a higher coverage of the included genomic areas and the sensitivity for detecting fusions increases simultaneously. Further, data output is moderate compared to WGS or WTS and therefore more suitable for daily routine testing [10].

It is possible to use either DNA or RNA as input material for targeted sequencing. DNA-based fusion detection is limited due to the possibility that fusion breakpoints may occur in intronic areas. Thus, large regions have to be covered. RNA-based approaches can overcome this problem by using transcripts for the detection of fusions. Nevertheless, their sensitivity is dependent on fusion gene expression levels [11]. Currently, there are different approaches available to detect fusions in a targeted setup: hybrid capture-based methods, amplicon-based methods and anchored multiplex PCR (see Figure 1). Various off-the-shelf panels are available for all approaches as summarized by De Luca et al. and compared by Heydt et al. [7,12]. Hybrid capture-based approaches have the advantage to cover larger areas including flanking regions of the target area. Thereby the possibility to miss a fusion is reduced. However, the input of nucleic acids needed for hybrid capture is relatively high. In contrast, amplicon-based approaches using fusion-specific primers need only a low input, making it applicable for routine diagnostics especially with FFPE material. New fusions or unknown fusion partners cannot be detected though. A more recent way to find fusions is the anchored multiplex PCR (AMP-PCR). Since this approach uses gene-specific primers anchored to exon-intron boundaries and universal reverse primers, it can be used to enrich unknown genomic regions of interest. Even unknown fusion partners can be detected by this approach. A comparative analysis for targetable fusions revealed that assays using AMP-PCR or hybrid capture enrichment detect a broader spectrum of fusions than other RNA-based assays [5].

Figure 1. Schematic principles of library preparation workflows for amplicon-based (a), anchored multiplex PCR-based (b) and hybrid capture-based (c) targeted RNA-seq methods

5. Major challenges and optimization opportunities

The major challenge for many laboratories is to find a way to optimally navigate diagnostic assays without exhausting the limited tumor tissue of small biopsies or even cytological specimen. To meet this, the diversity of diagnostic procedures has to be reduced. A method to definitely save material is to combine DNA and RNA extraction for further NGS on both [12–14]. This can be achieved by using RNA specified extraction kits without the DNA digestion step and adjusting the extraction volume to an optimum. This allows simultaneous analysis of DNA and RNA for NGS instead of sequential analysis. Extraction kits for total nucleic acid extraction (tNA) from FFPE material with following AMP-based RNA-seq have been described and compared extensively by Kresse et al., 2020 (AllPrep DNA/RNA FFPE Kit(Qiagen Inc.); truXTRAC FFPE RNA Kit (Covaris Inc.)) and Heydt et al., 2020 (Maxwell RSC RNA FFPE Purification Kit (Promega). For DNA-free RNA, the genomic DNA was removed from the tNA extracts using the TURBO DNA-free Kit (Thermo Fisher Scientific)) [12,14]. Once established, it can be a powerful tool to save tissue and shorten turnaround time.

n addition, sequencing of RNA from FFPE tissue seems to be a challenge itself, since it is often highly degraded and fixation processes often lead to chemical modifications of the nucleic acid. Nevertheless, various studies have successfully proven the utility of RNA derived from FFPE for different RNA-seq approaches [15–17].

Another important point is the rapidly growing knowledge on fusion events in certain tumor entities. Due to new insights on the significance of fusion partners and the development of new therapeutics, the number of diagnostically relevant genes is growing constantly. Thus, NGS panels that allow a flexible design with the possibility to boost areas of interest should be preferred. Their content can be easily adapted to new requirements. An alternative to find unknown fusions are large comprehensive panels or WTS. Their advantage is diminished by the fact that these panels are to date too expensive for use in routine diagnostics, need high sequencing capacities and complex bioinformatic algorithms for evaluation.

6. Detection of rare structural variants by RNA-Seq: examples from clinical practice

Employing RNA-Seq in routine diagnostics revealed that it is also a helpful tool in detecting activating aberrations, like the MET exon 14 skipping event or duplications of the EGFR kinase domain. Especially for the characterization whether mutations and deletions in MET within or in close proximity to splice regions lead to skipping of exon 14, RNA-Seq has proven reliability. A large variety of so-called activating MET exon 14 skipping mutations has been described by Awad et al and Frampton et al. [18–20]. However, for some deletions and mutations within the splice region the effect on splicing is only assumed or even unknown. Indeed, for such cases showing unknown deletions and/or mutations within the splice region of MET exon 14 skipping can be proven by subsequent targeted RNA sequencing. Noteworthy also silent mutations within the splice site could lead to exon 14 skipping. Carrying out NGS on DNA level only might miss the correct interpretation of such mutations as they could easily be overlooked. Silent mutations within the splice site do not alter the amino acid encoded, but the alteration on nucleic acid level could lead to the inactivation of splice sites resulting in exon skipping [21].

Additionally, interesting and putative targetable rare alterations of the EGFR gene can be detected. Such include VSTM2A-EGFR fusions, EGFR kinase domain duplications and the EGFR variant III. They all affect the activity of the EGF receptor, but therapy options vary strongly. The duplication of the EGFR kinase domain region results from the so-called EGFR ITD (internal tandem duplication), which is a fusion of EGFR exon 18 and exon 25 [22]. This fusion consequently leads to constitutively activated signaling of the receptor. It has been shown to be sensitive to second-generation EGFR TKI offering a strong treatment opportunity [23]. Additionally, the fusion of EGFR exon 1 with EGFR exon 8 is known to result in a structural variant of EGFR (known as EGFR vIII). The kinase domain itself is not affected but a constitutive downstream signaling of the EGF receptor has been shown [24]. Unfortunately, EGFR TKI are not effective against this variant, but promising efficacy of EGFR antibodies has been shown in glioblastoma patients carrying this aberration [25]. Finally, the VSTM2A–EGFR fusion has not been characterized so far and it remains to be ambiguous whether it has a driving effect via the EGF receptor and which therapeutics might present a treatment opportunity.

As already mentioned, RNA-Seq allows to determine the exact fusion partner and the fused exons. By now it is known that the fusion partner of certain drivers can influence therapy response and patient outcome. For the EML4–ALK fusion different variants are described. The most frequent ALK variants in NSCLC patients are ALK-EML4 variant 1 (EML4 (exon 13)-ALK (exon 20)) and 3 (EML4 (exon 6)-ALK (exon 20)). Although the fusion partner gene is the same, the fused regions vary and lead to obvious differences in resistance acquisition under therapy. Especially the well-described resistance mutation under ALK inhibitor therapy p.G1202R is only acquired in variant 3 patients [26].

Another example for the importance to determine the fusion partner has been described by Li and coworkers. They analyzed the survival data of ROS1 translocated patients. Patient cohorts were separated into CD74-ROS1 and non-CD74–ROS1 patients. Progression-free survival (PFS) as well as overall survival (OS) was significantly higher within the non-CD74–ROS1 group [27].

7. Conclusion and expert opinion

With regard to our experience, we recommend targeted RNA sequencing as a reliable method to evaluate known and novel gene-fusions. Nevertheless, the question still remains, whether RNA-seq will replace currently used molecular methods such as FISH for the detection of fusion events in NSCLC. To date, molecular diagnostic labs that apply RNA-seq still use IHC and FISH. The current clinical setting relies on the time- and cost-effective screening of targetable fusions via IHC (e.g ALK, ROS rearrangements) and FISH (e.g. RET rearrangements). Further validation of the screened fusion is then performed with NGS-based approaches, which also allows the determination of the fusion partner. In addition, patients without any targetable driver mutation are often subjected to RNA-seq fusion analysis.

Therefore, the applied NGS panels should be flexible to be easily adapted to clinically emerging requirements. In order to adequately save tissue material multigene testing as well as simultaneous RNA- and DNA extraction can be implemented into diagnostic procedures. Examples from clinical practice strongly underline the importance to correctly identify the exact fusion partner. Moreover, the detection of even new but probably targetable fusions and structural variants opens treatment options and drives clinical research toward innovative drugs.

In conclusion, the exploration of fusions plays an important role in the understanding of cancer. Thus, the development and improvement of NGS-based molecular approaches such as RNA-sequencing are growing and will lead perspectively to a better treatment of NSCLC patients.

Declaration of interest

S Wagener-Ryczek has received honoraria from Invitae. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This paper was not funded.