FormalPara Key Points

Although rare, oncogenic RET aberrations can be found in various digestive tract tumours, and could be a potential therapeutic target.

RET aberrations alone share no prognostic impact in patients with digestive tract cancers.

Patients with RET-aberrant digestive tract tumours have a blunted treatment response to immune checkpoint inhibitors.

1 Introduction

The rearranged during transfection (RET) gene encodes a receptor tyrosine kinase and is a potent oncogenic driver in certain types of cancers [1, 2]. The putative oncogenic activation of RET derives mainly from structural rearrangements contributing to constitutive activity [3]. RET was initially identified from sporadic mutations in papillary thyroid cancer and germline mutations in multiple endocrine neoplasia syndromes 2A/B [4, 5]. RET aberrations, notably fusions/translocations, account for 1 to 2% in non-small cell lung cancer (NSCLC) and the therapeutic efficacy of selective RET inhibitors has been largely validated [6,7,8,9,10,11]. Albeit rare, growing evidence suggests that RET aberrations play a nonnegligible role in solid tumours other than thyroid or lung cancers, in which RET inhibitors also demonstrate promising therapeutic efficacy [12,13,14,15]. Indeed, RET aberrations might function as an oncogenic driver and could be harnessed as a potential therapeutic target in a broader spectrum of malignancies beyond thyroid or lung cancers.

In patients with hepatobiliary or gastrointestinal tract tumours, targeted therapeutic options are relatively limited, prompting a comprehensive elucidation of any potentially actionable targets. Adashek et al. reported in the GENIE database that scattered cases with oncogenic fusions involving RET can be found in colorectal, oesophagogastric, and hepatobiliary tumours, along with missense mutations in 1.5 to 15.0% of patients [12, 16]. Another study incorporating 4871 sequenced tumours revealed RET aberrations in 0.7 to 5.0% of digestive tract tumours, yet some of these were considered passenger mutations instead of oncogenic ones [17]. Zhao et al. reviewed a large-scale sequencing result in various types of tumours and concluded several potential oncogenic point mutations beyond known fusions that responded to selective RET inhibitors [18]. In addition to selective RET inhibitors, emerging reports also indicate a blunted treatment efficacy to immune checkpoint inhibitors (ICPi) in patients with RET-aberrant NSCLC or thyroid tumours [19, 20]. However, such investigations have not been performed in patients with digestive tract tumours, who constitute another patient population prone to receiving ICPi therapies in clinical practice.

Unlike the revelations of RET in thyroid cancer and NSCLC, the clinical, genomic, and prognostic features of RET-aberrant digestive tract tumours are not well known. In addition, since most of the studies published to date were derived from Western patient populations, confirmatory data on Asian patients could be informative with regard to regional or ethnic heterogeneity. Furthermore, the therapeutic effects of RET inhibitors or ICPi have not been explored in these patients, notably through real-world clinical information.

In the present cohort study, we retrospectively reviewed patients with hepatobiliary, oesophagogastric, pancreatic, and colorectal tumours. Panelized selected genome sequencing (SGS) of the tumours was applied to identify patients with RET-aberrant versus non-aberrant tumours. The aim of the study was to investigate the clinical and genomic characteristics of RET-aberrant digestive tract tumours and their potential responses to RET inhibitors or ICPi. The results may provide essential information on RET aberrations and help prioritize potential treatments in digestive tract malignancies.

2 Materials and Methods

2.1 Patients

We retrospectively reviewed patients with digestive tract carcinoma (primary malignancies of the oesophagus, stomach, liver, biliary tract including the ampulla of Vater, pancreas, and colorectum) detected from Jan 2016 to Jan 2021 at the National Cheng Kung University Hospital, Taiwan. We analysed patients with newly diagnosed or relapsed/recurrent disease who had received SGS encompassing cancer-related genes (inclusive of RET) from available formalin-fixed paraffin-embedded (FFPE) specimens. Patients meeting any of the following criteria were excluded: (1) the patient had a malignancy of the digestive tract that was not a carcinomatous neoplasm, such as lymphoma or sarcoma; (2) the neoplasms were not confirmed based on pathological specimens or were not used for tumour-based sequencing; (3) the patient did not receive any antineoplastic treatments in the study institute; and (4) the medical records were not adequate for evaluating the clinical characteristics or results of SGS. Patients with more than one malignant disease were not excluded and were analysed according to the neoplasm tested for SGS. For patients who received multiple serial tests for one malignant disease, only the most upfront or qualified sequencing result was considered. Enrolled patients could be treated based on sequencing results upon the physicians’ reasonable judgement and clinical availability, including referral to indicated clinical trials. Patients who did not receive sequencing-tailored therapies were still included in the study.

2.2 Clinical Characteristics and Treatment

We recorded the clinical and demographic characteristics from all available electronic medical records in the study institute. RET-aberrant tumours were defined by a positive sequencing result indicative of a putatively oncogenic RET aberration such as fusion/translocation, nonsynonymous mutation or indel, which was confirmed by public domain human genomic databases, including Catalogue Of Somatic Mutations in Cancer (COSMIC), Clinically Relevant Variations (ClinVar), Human Gene Mutation Database (HGMD), and cBioPortal for its pathogenicity with a latest access and acquisition date on 10 Nov 2022 [21,22,23,24,25]. Patients with a RET aberration that was deemed as a passenger variation or variant of uncertain significance (VUS) were excluded from the study. Overall survival (OS) was defined by the time interval between the initial diagnosis of the incited solid tumour to death of the patient from any cause. Exposure to selective RET inhibitors was defined as a treatment with the drugs selpercatinib or pralsetinib. Patients who received multitargeted tyrosine kinase inhibitors were not evaluated in the study owing to a potentially biased interpretation of the therapeutic effect which was not limited to RET. Exposure to ICPi was defined as administrations of programmed death ligand-1 (PD-L1), PD-1 or cytotoxic T-lymphocyte-associated protein 4 inhibitors at a reasonable therapeutic dose. The treatment responses were determined as the best evaluable response per Response Evaluation Criteria in Solid Tumours 1.1 (RECIST 1.1) by an independent reviewer (HY Huang) [26]. Therapeutic responses were analysed collectively or based on specific tumour type. PD-L1 expression was determined by either tumour proportion (TPS) or combined positive scores (CPS) by Dako 22C3 pharmDx in the available samples. Time to treatment discontinuation (TTD) was defined as the time interval between initiation of the defined treatment to discontinuation for any reason, including disease progression, adverse events, patient’s or physician’s decision or patient death.

2.3 Tumour-Based Selective Genome Sequencing

We conducted SGS by FoundationOne CDx from tumoural DNA. The source of materials was derived from either fresh or archival tumour samples with a time interval between sampling and SGS testing <180 days. Tumour specimens were required to be at least 1 mm3 in volume and 20% in tumoural content. The aforementioned assay was conducted on the Illumina HiSeq 4000 platform with a uniform read depth of at least 500 times as a median coverage and at least 100 times in over 99% of the exons sequenced. Genomic aberrations with a variant allele frequency (VAF) of 10% or less in the assayed tumour were denoted as subclonal changes and were not considered in the study. Microsatellite status was categorized as microsatellite instability-high (MSI-H) or stable (MSS). Tumour mutation burden (tMB) was reported as mutations per mega base pair (mut/106 bps). The sequenced data were mapped to the human genome 19 assembly (hg19). Detailed methodological information and the tested genes are provided in the electronic supplementary material (ESM). All detected genomic aberrations were verified by the COSMIC, ClinVar, HGMD, and cBioPortal databases [21,22,23,24,25].

2.4 Statistical Considerations

We presented clinical and genomic features by descriptive analyses. We used Student’s t test for comparing continuous variables and the chi-squared test for categorical variables. Nonparametric tests were used if the parametric assumptions were not met. Survival analysis was conducted by the Kaplan‒Meier method, and survival was compared with the log-rank test. Univariate and multivariable regression analyses were conducted for independent prognostic factors under the Cox proportional hazard model. Multiple pairwise comparisons of the frequency of differentially altered genes were conducted between RET-aberrant and non-aberrant tumours and adjusted for false discovery rates (FDR) by the Benjamini–Hochberg method [27]. Considering the baseline genomic heterogeneity in different cancers, we compared patients with colorectal cancer only as a supplemental analysis to enhance the robustness of the comparison. We presumed statistical significance at a two-tailed α of 0.1. We did not perform imputations in any of the missing values in the analysed data. Since the present work is a retrospective cohort study in nature, we have provided the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) assessment in Supplemental Table 1 (see ESM). R 4.0.4 and GraphPad Prism 9.0 were used for data management and graphic production.

3 Results

3.1 Patients and Clinical Characteristics

A total of 980 patients were initially screened. After the pre-screening process (n = 565), another 100 patients were excluded due to sequencing failure. A total of 465 patients with digestive tract tumours were further assessed, of whom 20 patients had putatively oncogenic RET-aberrant tumours excluding those with passenger aberrations or VUS (n = 12; Supplemental Table 2, see ESM), and another 433 patients had RET non-aberrant tumours. The flow diagram is shown in Fig. 1. In the studied population, the median age was 58 years. Eighty percent of patients had stage III–IV disease and 70% of them presented with lymph node metastasis. Two-thirds of the sequenced samples were primary site tumours. The most common tumours were derived from colorectum (51.0%), stomach/oesophagogastric junction (EGJ) (18.8%), and biliary tract (14.2%). In the studied population, 20 patients had RET-aberrant tumours, and the most common primary tumour locations were colorectum (n = 10), stomach/EGJ (n = 4), biliary tract (n = 2), and liver (n = 2). The clinical characteristics were similar between patients with RET-aberrant and RET-wild type tumours, except for less lymph node involvement in the former (p = 0.0462). We did not find any patients with RET-aberrant tumours designated as MSI-high. The median tMB was also slightly lower as compared with those with non-aberrant tumours, despite that the difference was not statistically significant (median tMB: 2.5 vs 3.8 mut/106 bps; p = 0.9195 by Mann‒Whitney U test). The detailed patient characteristics are shown in Table 1.

Fig. 1
figure 1

Flow diagram of the study. a. Multiple reports with inconsistent results (n = 7), technical failure (n = 5) and incorrect specimen for testing (n = 2). SGS selected genome sequencing

Table 1 Patient characteristics

3.2 RET and its Co-Occurring Aberrations

The oncoplot of RET and its co-occurring genomic aberrations are shown in Fig. 2. The detection incidence of RET aberrations in different tumours is provided in Supplemental Fig. 1 (see ESM), with the highest in oesophageal squamous cell carcinoma (7.7%) and the lowest in pancreatic adenocarcinoma (3.1%). We identified five cases with RET fusions, one case with frameshift indel and the remaining 14 cases with putatively oncogenic missense mutations. Among patients with fusions, two had a RET-CCDC6 fusion (gastric cancer), one had a NCOA4-RET fusion (colorectal cancer), one had a RET-NCOA4 fusion (pancreatic cancer), and one had a KIF5B-RET fusion (hepatocellular carcinoma). In the missense mutations, the most frequent sites were p.V804M (n = 3) and followed by p.E511K (n = 2), p.S904Y (n = 2) and p.M848V (n = 2). In total, RET-aberrant tumours accounted for 4.4% in the studied population, of which 1.1% had an oncogenic fusion and 3.3% potentially had a driver mutation beyond fusion. The detailed sequencing results, actionability and characteristics are provided in Supplemental Table 3 (see ESM). Other co-occurring aberrations were found, such as ERBB2 amplification (n = 3), PIK3CA mutation (n = 1), KRAS exon 2 to 4 mutation (n = 1), and BRCA2 pathogenic indel (n = 1) (Supplemental Fig. 2, see ESM).

Fig. 2
figure 2

RET and co-occurring aberrations. The right column denotes the recurrent frequencies (%). Only genes with recurrent frequencies ≥ 5% are shown. CHOL cholangiocarcinoma, COAD colorectal adenocarcinoma, ESCA oesophageal squamous cell carcinoma, F female, LIHC hepatocellular carcinoma, M male, MSI microsatellite instability, NA not available, PAAD pancreatic adenocarcinoma, STAD gastric adenocarcinoma, tMB tumour mutation burden

3.3 Frequently Altered Genes in RET-Aberrant Versus Non-Aberrant Tumours

We compared the frequently altered genes in RET-aberrant versus non-aberrant tumours collectively (n = 453). The results showed that APC, KRAS, TP53, MSH6, and STK11 were significantly different between the two groups, with APC, KRAS, and TP53 more frequently altered in non-aberrant tumours (all FDRs <0.001). To the contrary, MSH6 and STK11 (FDR = 0.0474) were more frequently altered in RET-aberrant tumours (Supplemental Table 4, see ESM).

Three of twenty patients with RET-aberrant tumours harboured a MSH6 alteration but none of them were designated as MSI-H or high tMB. In addition, we compared patients with colorectal tumours only (n = 231) and found a similar result in TP53, APC, and KRAS (all FDRs <0.05). STK11 also showed a decreased frequency of alterations in non-aberrant colorectal tumours, but the difference was not statistically significant (FDR = 0.1104) (Supplemental Table 5, see ESM).

3.4 Prognosis in Patients with RET-Aberrant Versus Non-Aberrant Tumours

In a median follow-up time of 51 months, we did not observe a significant prognostic difference between patients with RET-aberrant tumours and those with RET wild-type tumours (median OS: 53.5 vs 54.2 months, HR 0.79, 95% CI 0.44–1.44; p = 0.5307; Fig. 3A). The survival differences were not significant in patients with either early or late stages of the disease (Stage I–II, median OS: 78.2 vs 68.6 months, HR 0.78, 95% CI 0.27–2.26; p = 0.6414; Fig. 3B) (Stage III–IV, median OS: 45.1 vs 48.3 months, HR 1.32, 95% CI 0.59–2.92; p = 0.8839; Fig. 3C). Using Cox regression analysis, we found that tumour type of pancreatic adenocarcinoma or cholangiocarcinoma, stage IV disease, increased age, and poor performance status were significantly associated with a poor prognosis. Presence of RET oncogenic aberration was not an independent prognostic factor in the studied population (Table 2).

Fig. 3
figure 3

Overall survival. A All patients, B stage I–II patients, and C stage III–IV patients. Compared with log-rank test. CI confidence interval, HR hazard ratio

Table 2 Prognostic factors for patients with digestive tract tumours

3.5 Treatment with RET or Immune Checkpoint Inhibitors

Only two patients had been exposed to RET inhibitors in the study. One patient was a 65-year-old female who had metastatic colorectal cancer with NCOA4-RET fusion, and the other was a 34-year-old male who had hepatocellular carcinoma with a KIF5B-RET fusion. Both received selpercatinib with a duration of 4.6 and 7.5 months, respectively. The first patient had progressive disease, while the second had a best response of stable disease. On the other hand, a total of 58 patients were exposed to ICPi (RET-aberrant, n = 11; non-aberrant, n = 47). The illustrated swimmer plot for ICPi treatment duration is shown in Fig. 4. TTD was significantly shorter in RET-aberrant tumours as compared with those that were not (median TTD: 2.8 vs 6.2 months, p = 0.0008 by log-rank test). Best evaluable treatment responses of ICPi are provided in Fig. 5. Among patients with RET-aberrant tumours, we did not observe any responders, and the best response was stable disease (n = 3). In contrast, an objective response rate of 27.7% and a disease control rate of 80.1% were found in patients with non-aberrant tumours, including two achieving a complete response. The details of treatment responses, as stratified by ICPi monotherapy or combinations, are provided in Table 3.

Fig. 4
figure 4

Time to treatment discontinuation with immune checkpoint inhibitors. Swimmer plot for time to treatment discontinuation with ICPi in patients with RET-aberrant vs non-aberrant tumours. CHOL cholangiocarcinoma, COAD colorectal adenocarcinoma, CR complete response, ICPi immune checkpoint inhibitor, LIHC hepatocellular carcinoma, ms months, PAAD pancreatic adenocarcinoma, PD progressive disease, PR partial response, SD stable disease, STAD gastric adenocarcinoma

Fig. 5
figure 5

Treatment responses with immune checkpoint inhibitors. Waterfall plot for treatment responses with ICPi in patients with RET-aberrant vs non-aberrant tumours. CR complete response, PD progressive disease, PR partial response, SD stable disease. *Denotes patients with RET-aberrant tumours

Table 3 Treatment responses with immune checkpoint inhibitors

We found two of the 11 patients with RET-aberrant tumours had a PD-L1 CPS ≥5 and achieved a best response of stable disease and relatively durable TTD. However, another five patients with CPS <1 did not respond to ICPi treatment (Supplemental Table 6, see ESM). Conversely, the OS of patients who had been exposed to ICPi was similar between patients with RET-aberrant versus non-aberrant tumours (p = 0.410 by log-rank test) (Supplemental Fig. 3, see ESM). A subgroup analysis was conducted as stratified by cancer type. Despite the similar finding of reduced response rates across all cancer types, the results were statistically under-powered due to limited case numbers (Supplemental Table 7, see ESM). In addition, a cross-organ analysis revealed no significant differences in terms of TTD when the sequencing of tumours was conducted at a primary site as compared with metastatic sites (lung, liver or peritoneum) (Supplemental Table 8, see ESM).

4 Discussion

The current retrospective study focuses on RET-aberrant digestive tract tumours derived from real-world clinical practice. Tumours with putatively oncogenic RET aberrations accounted for 4.4% in the studied population, with 1.1% harbouring an oncogenic fusion. The result was comparable with large-scale studies incorporating tumour-based next-generation sequencing (NGS). Kato et al. indicated that RET aberrations were found in 1.8% of miscellaneous tumours, mostly composed of mutations, fusions, and amplifications, but reported scarcely in digestive tract cancers [17]. In the Memorial Sloan Kettering Cancer Center-IMPACT (MSK-IMPACT) cohort, 2.4% of the tumours had RET aberrations and were mainly derived from thyroid cancer, cancers of unknown primary site, and NSCLC, but not digestive tract tumours [1, 28]. Wu et al. reported a Chinese cohort in which 2% of patients with biliary tract, colorectal, gastric, and hepatocellular cancers harboured oncogenic rearrangements involving RET [29]. Yang et al. further indicated that 2.7% of patients with metastatic colorectal cancer had RET mutations, excluding those with fusions or rearrangements [30]. Our findings were consistent with published results showing that RET aberrations could be found in various types of digestive tract tumours. Although it represents only a small proportion of the patient population, we revealed the therapeutic promise as potentially actionable targets requiring detection by NGS in clinical practice.

In a myriad of digestive tract cancers, we disclosed that some patients with RET-aberrant tumours had co-occurring potentially actionable targets, such as PIK3CA, KRAS, ERBB2, and BRCA2. A median of two coexisting actionable aberrations was found in a cohort of 88 patients with RET-aberrant tumours [17]. Zhao et al. also demonstrated that patients with colorectal or pancreatic cancer had more RET passenger mutations than drivers, outweighing other cancer types [18]. Consistent with the published reports, we observed that RET passenger but not driver mutations were enriched in colorectal tumours and should not be counted as actionable targets [17, 31]. On the contrary, the oncogenic aberrations often involved the cadherin-like domain 4 and the tyrosine kinase catalytic domain of RET, which might be associated with oncogenic activation and could potentially respond to selective RET inhibitors, as predicted by collaborative genomic databases [18, 22, 23]. In our study, two out of five patients with RET fusions had received selpercatinib, but both were non-responders. Although we could not demonstrate its clinical efficacy owing to the limited number of cases, emerging studies have shown the promise beyond NSCLC and thyroid cancer [13,14,15, 32]. Extending the patient pool from RET fusions to other driver indels or mutations sheds new light on the treatment options for RET-aberrant tumours.

Several studies have proposed a correlation of antitumour immunogenicity with RET aberrations. Retrospective studies revealed that patients with RET-aberrant NSCLC or thyroid cancer had lower tMB and PD-L1 expression and a reduced responsiveness to ICPi therapies [19, 20, 33,34,35]. Mazieres et al. reported eight oncogenic driver alterations and concluded that patients with RET-aberrant lung cancer had a low ICPi treatment response (ORR 6%) and short progression-free survival (median progression-free survival 2.1 months) as compared with other driver alterations [36]. However, none of these results were validated in RET-aberrant digestive tract tumours. Scattered reports indicated that patients with RET-mutated colorectal cancer had a higher likelihood of MSI-H status and a higher tMB [30, 37]. However, the efficacy of ICPi had not been fully discussed. In our study, we did not find any MSI-H designation in patients with RET-aberrant tumours and the tMB was slightly lower in those with non-aberrant tumours. Nearly half of the RET-aberrant tumours had a low PD-L1 expression in our cohort. It is possible that patients with an MSI-H and/or high tMB tumour had a higher probability of RET aberrations owing to their genomically unstable condition but the aberrations were mostly nonspecific passengers or VUS, which were not putatively oncogenic and not accounted for in the study. Conversely, we did not observe any responders with RET-aberrant tumours treated with ICPi, while those with non-aberrant tumours showed ICPi responsiveness comparable to that of general patient population. In addition to the relatively immune-evasive condition with MS-stable and a lower tMB, we also discovered a higher propensity of co-occurring aberrations in STK11, which was known to reduce the efficacy of ICPi [38,39,40,41]. In summary, our study shared a coherent theme with previous studies in that RET-aberrant digestive tract tumours showed a blunted treatment response to ICPi. Applying panelized or thorough sequencing of tumours might increase the predictability of ICPi efficacy in conjunction with other known biomarkers. Prioritizing RET over immune checkpoint inhibition would provide a possible solution to overcome this therapeutic hurdle.

The prognostic role of RET remains largely undetermined. Hess et al. reported that RET fusion was not a significant prognostic factor in patients with NSCLC treated with standard therapies exclusive of selective RET inhibitors [42]. Another report by Tsai et al. indicated that patients harbouring RET-rearranged lung cancer had comparable survival to those with epidermal growth factor receptor-mutant disease [43]. In contrast, studies of colorectal cancer showed that patients with RET-aberrant tumours had poor survival and shorter treatment duration than those without [19, 37]. Whether selective RET inhibitors prolong survival in patients with RET-aberrant cancers as compared with other treatments remains unknown since the updated results are derived from single-arm early phase trials. We disclosed that presence of RET aberrations alone shared no independent prognostic influences in patients with digestive tract tumours. However, our survival result was largely not affected by selective RET inhibitor therapies. More clinical studies are required to delineate the prognostic role and correlation with RET inhibitor therapy.

The merits of the current study are its use of a valid and approved SGS method and focus on an extremely rare but actionable alteration in digestive tract tumours, in which therapeutic target options are usually scarce. In addition, we describe real-world clinical characteristics, and the results can be extrapolated to future prospective clinical studies. Furthermore, the treatment responses to ICPi are discussed in the study, which are seldom reported beyond their well-known use in NSCLC and thyroid cancer. However, there are some limitations in the study. First, the study is limited by its insufficient case numbers and cannot provide detailed information on particularly rare aberrations. Second, since we utilised FoundationOne CDx as a sole sequencing source in the study, we might underestimate the incidence due to a DNA-based rather than RNA-based detection for target fusions. Third, we did not have an adequate number of patients who had received selective RET inhibitors and thus could not assess claims of the efficacy. Lastly, despite all our efforts to exclude potential passenger RET mutations, we still could not totally eradicate the risk of including non-driver aberrations in the study. Nevertheless, the present study provides information on rare but potentially druggable targets in lesser known non-thyroid or lung malignancies.

5 Conclusion

Although rare, RET aberrations can be found in various digestive tract tumours and encompass oncogenic fusions, pathogenic indels, and mutations. Several potentially actionable mutations may accompany RET aberrations. Patients with RET-aberrant tumours have a reduced responsiveness to ICPi as compared with those with RET-wild type tumours. Together, these results provide insights into this rare but potentially actionable target in digestive tract tumours.