Introduction
Universal screening for mismatch repair deficiency (dMMR) amongst colorectal cancer (CRC) patients has been recommended to facilitate identification of Lynch syndrome (LS) and to direct optimal oncological treatment of those cases presenting a sporadic microsatellite unstable tumour [
1]. For the treatment of adult patients with unresectable or metastatic CRC with dMMR or microsatellite instability (MSI), European Medicines Agency (EMA) has approved first-line monotherapy treatment of immuno-oncological drug pembrolizumab in 2020. Besides DNA-repair deficiency phenotype in CRC, gene fusions that act as oncogenic drivers offer targets for cancer therapy. To this end, larotrectinib became the first and entrectinib the second tumour agnostic, i.e. ‘histology-independent’, cancer treatment approved by EMA (2019 and 2020, respectively) in patients whose solid tumours display a
neurotrophic tyrosine receptor kinase (
NTRK) gene fusion and are advanced, have spread to other parts of the body or are not amenable to surgery, and who have no satisfactory alternative treatments [
2]. The family of
NTRK genes consists of
NTRK1-3 encoding TRKA, TRKB and TRKC proteins that play a role in development and functioning of the nervous system, and act as drivers of oncogenesis in various cancers [
3]. Only 0.2–0.3% of CRCs harbour
NTRK gene fusions, which makes universal screening of CRC patients for this gene rearrangement impractical. However, recent studies have recognised an enrichment of
NTRK fusions in a subset of CRCs presenting dMMR due to loss of
MLH1 gene expression,
BRAFV600E wild-type (wt),
MLH1 promoter hypermethylation (
MLH1ph) and
RASwt [
4‐
7].
Gene fusions can be studied using DNA–, RNA– or combined DNA/RNA–based next-generation sequencing (NGS), as well as with fluorescence in situ hybridization (FISH), reverse transcription-polymerase chain reaction (RT-PCR) and immunohistochemistry (IHC). ESMO recommendations propose that IHC can be used as a screening method (if no smooth muscle or neuronal differentiation is present) to enrich patients with
NTRK fusions in an unselected population [
2,
8]. FISH and RT-PCR are recommended to be used in tumour types that harbour high frequency of a specific
NTRK fusion, such as
ETV6::NTRK3, which is relatively infrequently found in CRC. DNA– and/or RNA–based NGS panels can be used either upfront or to confirm the presence of
NTRK fusion in TRK immunopositive tumours or those devoid of other driver mutations, such as those in
BRAF and
RAS genes. Of the NGS platforms, RNA–based panels are favoured due to their ability to detect both known and novel fusions and higher sensitivity when compared to the DNA–based ones. Recently, a novel fully automated quantitative RT-PCR option has been introduced, the research use only (RUO) Idylla gene fusion assay, which can detect several oncogenic gene fusions.
The aim of this study was to investigate enrichment of NTRK and other oncogenic gene fusions in a cohort of dMLH1 and BRAFV600Ewt CRC cases (n = 62), which originated from universal dMMR screen of over two thousand consecutive CRC patients in a real-life diagnostic setting. First, gene fusions were analysed using a novel RNA–based FusionPlex Lung v2 NGS panel, and these results were then compared to a novel RNA–based Idylla GeneFusion assay and pan-TRK immunohistochemistry (IHC). In addition to MMR and BRAF mutation status, all 62 cases were analysed for MSI, MLH1ph and RAS mutation status.
Discussion
We screened 2079 CRC resection specimens for dMLH1 and BRAFV600Ewt using IHC and identified 64 such cases of which 62 were available for this study. The BRAFV600E IHC result was confirmed in this study by MS-MLPA and Idylla NRAS-BRAF test, which both were 100% concordant. The MSI status was confirmed with Idylla MSI test. Since
NTRK gene fusions have been shown to be enriched in this subgroup of CRC [
4,
7] and solid tumours with
NTRK fusion can be treated with larotrectinib or entrectinib [
2], we were especially interested to investigate this targetable gene rearrangement using several techniques. First, RNA–based Lung v2 NGS panel was used as the gold standard, and seven (7/62, 11.3%)
NTRK1 fusions were found in our CRC cohort with dMLH1/
BRAFV600Ewt. Westphalen et al. recently showed the prevalence of
NTRK fusions in CRC to be 0.22% in a large real-world cohort of adult CRC cases (
n = 34 590) using NGS–based database approach [
7]. This figure is comparable to previous screening studies for
NTRK fusions in CRC that have found the prevalence to be 0.14–0.35% using upfront NGS or IHC with NGS confirmation [
4‐
6,
13‐
16]. Thus, our
NTRK fusion frequency of 0.34% (7/2079) is at the higher end.
NTRK1 fusions with
TPM3,
LMNA and
TPR partners have been shown to be the most common ones in CRC [
6,
7] and
PLEKHA6::NTRK1 fusions have also previously been described in this disease [
17,
18]. Of these, we found
NTRK1 to partner with
TPM3,
LMNA and
PLEKHA6.
IRF2BP2::NTRK1 fusion was found in a single case of our series, and to our knowledge, this is the first time it has been reported in CRC. Interestingly, the
IRF2BP2::NTRK1 containing tumour was the only one located to the left colon (sigma) amongst our
NTRK1 fusion–positive cases. We were, however, unable to find any
NTRK3 fusions, although Lung v2 NGS can detect a wide range of
NTRK3 fusions. This may relate to the fact that the estimated proportion of
NTRK3 fusions of all
NTRK fusions is only 11% in CRC [
7]. Strengths of our study included the use of real-life diagnostic tissue material originating from a cohort of over two thousand consecutive surgically treated primary CRC patients and the use of multiple techniques, including detection of the
NTRK1 fusions using three independent RNA–based NGS platforms. Weakness of the study was that we investigated gene fusions only in a subgroup of CRC. To this end, we may have missed a few
NTRK fusion–positive CRCs, since a small proportion (11–19%) of
NTRK fusions has been found in microsatellite-stable CRCs [
4,
6]. Four samples failed with Lung v2 NGS, three of which were found to be fusion-negative by the FusionPlex CTL NGS analysis, whereas the fourth one failed with CTL panel as well.
Westphalen et al. found CRC to be the only cancer type in which
NTRK fusions are associated with sporadic MSI [
7]. Interestingly, Kim et al. have recently described that
NTRK fusions in CRC develop along the serrated pathway, in which sporadic d
MLH1 is a major molecular event, and these fusions can already be present in premalignant sessile serrated lesions [
19]. Additionally, several other NGS–based studies have suggested 2.6–7.3% occurrence of
NTRK fusions in dMMR/MSI CRCs [
15,
16,
18,
20,
21]. In the dMLH1
/BRAFV600Ewt subgroup, frequency of
NTRK fusions has been reported to be 5–28% [
4,
16,
17] and in the subgroup of dMLH1
/MLH1ph 14–19% [
18,
19], which are in line with our frequencies of 11% (7/62) and 16% (7/43), respectively. Yet other studies have reported
NTRK fusions to occur in dMLH1
/BRAFV600Ewt/
MLH1ph/
RASwt subgroup of CRC with as high frequency as 17–44% [
15‐
18,
22] being comparable to our prevalence of 23.3% (7/30).
We also evaluated the performance of the novel fully automated Idylla gene fusion assay and pan-TRK IHC. Idylla gene fusion test was 100% specific and sensitive to detect
NTRK1 expression imbalance in the seven
NTRK1 fusion CRC cases. However, Idylla reported initially invalid result in seven cases (7/62, 11%), of which three remained invalid after re-analysis, and additional 12 isolated invalid results for
NTRK3 expression imbalance. Interestingly, the high frequency of
NTRK3 invalids might originate from promoter methylation causing loss of
NTRK3 expression, which has been reported in over 11% of CRC cases [
23]. Furthermore, our study shows that specificity and sensitivity of pan-TRK IHC are optimal in CRC samples. Although specificity of this method varies between different tumour types, it has previously been reported to be 100% for pan-TRK in CRC [
4,
13,
17,
24]. Pan-TRK IHC positive CRCs are characterised by cytoplasmic staining with additional positivity in other subcellular compartments in a fusion partner-dependent manner. Membranous staining has been linked to
TPM3,
TPR and
PLEKHA6; perinuclear staining to
LMNA and
MUC2 and nuclear staining to
ETV6 [
6,
24‐
26]. We also detected moderate membranous staining along with variable intensity of cytoplasmic staining with
TPM3 and
PLEKHA6 fusion partners and perinuclear staining with
LMNA.
IRF2BP2::NTRK1 fusion represents only 2% of all
NTRK fusions [
7], but it has been reported in lung, thyroid and prostate cancers [
3]. In lung cancer, it shows cytoplasmic immunostaining [
25], which was the case with our CRC sample as well.
In addition to
NTRK1 fusions, Lung v2 NGS detected two
ALK fusions (2/62, 3.2%) in our CRC cohort, which both partnered with the most common
ALK fusion partner
EML4 in CRC [
27]. ALK IHC showed strong cytoplasmic staining in both cases, whilst the rest of the samples were completely negative. Our results thus suggest a prevalence of 0.10% (2/2079) for
ALK fusions in CRC, which is comparable to previously published
ALK fusion prevalence of 0.05–0.6% in CRC [
27‐
30]. The Idylla fusion assay detected the
EML4(e6)::
ALK(e20) fusion as a specific fusion and the
EML4(e21)::
ALK(e20) fusion as an
ALK expression imbalance. The
EML4(e21)::
ALK(e20) fusion with non-canonical breakpoint of
EML4 gene at exon 21 has been reported to constitute only about 2% of the
ALK-rearrangements in non-small cell lung cancer, where
EML4(e6)::
ALK(e20) is the most common
EML4::
ALK variant [
31]. The less frequent
EML4(e21)::
ALK(e20) variant is not covered by the Idylla’s fusion-specific detection, which is designed to catch the most relevant gene fusions in lung cancer. However,
EML4(e21)::
ALK(e20) fusion variant seems to be more frequent molecular event in CRC [
22,
27,
29,
30]. In addition to the NGS–identified
ALK fusions, Idylla detected seven false positive
ALK expression imbalances. Based on our study, detection of specific
ALK fusion seems to be a valid result, whereas all expression imbalance results need to be validated by a more specific method.
Besides
NTRK1 and
ALK fusions, Lung v2 NGS detected four
RET fusions (4/62, 6.5%) and seven
BRAF fusions (7/62, 11.3%) in our CRC cohort. We found
RET to partner with
CCDC6 in three cases and
NCOA4 in one case, both of which seem to be quite common
RET fusion partners in CRC [
32,
33].
BRAF fusions with partners
AGAP3,
TRIM24 and
MKRN1 found in our study have also been previously reported to occur in CRC [
15,
16,
34].
STARD3NL::BRAF fusion has earlier been described at least in one paediatric sarcoma [
35], whereas to our best knowledge,
LMTK2::BRAF has not been reported before in any tumour type. Two of the four Lung v2 NGS–detected
RET fusions were detected by the Idylla assay as both
RET–specific fusion and expression imbalance. The Idylla gene fusion test does not include
BRAF fusions. Taken together, the Idylla platform identified two specific
RET fusions that were in line with the NGS results, but did not report expression imbalance of two NGS–detected
RET fusions (detection of these specific fusions is not included in the Idylla assay).
Upfront RNA–based NGS analysis is the most comprehensive, sensitive and specific method to identify gene fusions, but it is also time-consuming and requires more labour, expertise and financial resources when compared to other methods. The Idylla platform offers the fastest turnaround time with moderate cost, whilst IHC is the most affordable option. It is however clear that both Idylla expression imbalance and pan-TRK IHC results need to be validated using an alternative method, preferably an RNA–based NGS [
2,
8]. To our knowledge, this is the first publication where the RUO FusionPlex Lung v2 NGS and the Idylla GeneFusion assay have been used in detecting gene fusions. As there is a tumour agnostic oncological treatment for cancer patients with an
NTRK fusion, we would like to propose that one should screen for this gene rearrangement in CRC patients with dMLH1/
BRAFV600Ewt/
MLH1ph using Idylla gene fusion test or pan-TRK IHC, followed by an RNA–based NGS confirmation of the positive cases, or alternatively using upfront RNA–based NGS depending on local resources.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.