FGFR2 alteration as a potential therapeutic target in poorly cohesive gastric carcinoma

Background Poorly cohesive (PC) is a unique histologic subtype of gastric cancer (GC), with an increasing incidence in recent years. However, the molecular characteristics and therapeutic targets of PC GC are not yet well studied and there are no effective therapies for these patients. Methods Formalin fixed paraffin embedded (FFPE) samples of 556 GC patients, including 64 PC GC, were collected for next-generation sequencing (NGS). Clinical characteristics and genomic profiling were analyzed. FGFR2 expression was detected by quantitative real time polymerase chain reaction (qRT-PCR) and immunohistochemistry (IHC). FGFR2 inhibitors response was studied in vitro. Results Among 556 GC patients, PC GC patients were younger (P = 0.004), had lower tumor mutation burden (TMB-L) (P = 0.001) than non-PC GC. The top 10 most frequently mutated genes in PC GC were TP53 (48%), CDH1 (31%), ARID1A (14%), FGFR2 (14%), ERBB2 (9%), CDKN2A (9%), FGF3 (8%), LRP1B (9%), FGF19 (8%) and FGF4 (8%). Noticeably, FGFR2 is more frequently mutated than non-PC GC (14% vs. 6%, P = 0.037), including copy number variants (CNVs, 12.5%) and gene rearrangements (3.1%, FGFR2/VTI1A and FGFR2/TACC2). Former studies have confirmed that gain of copy number could increase FGFR2 expression and sensitivity to FGFR2 inhibitors in GC. However, no research has verified the function of FGFR2 rearrangements in GC. Our results showed that cell lines of GC transfected with TACC2-FGFR2 fusion had increased mRNA and protein expression of FGFR2, and were more sensitive to FGFR2 inhibitors. FGFR2 inhibitors might be a new therapeutic target for PC GC. In addition, we found patients of PC GC harboring gene rearrangements (n = 9) had poorer overall survival (OS) in comparison with patients without any gene rearrangement (n = 19) (16.0 months vs 21.0 months, P = 0.043). Gene rearrangement might be an adverse prognostic factor for PC GC patients. Conclusions FGFR2 alterations were recurrent in PC GC and FGFR2 inhibitors might be a new therapeutic target for PC GC. Supplementary Information The online version contains supplementary material available at 10.1186/s12967-021-03079-8.

overall incidence of GC, however, the proportion of GC with WHO PC histology is increasing [3]. PC GC is defined as tumor composed of isolated or small groups of tumor cells by the WHO classification [3]. Compared to other GC subtypes, PC GC is poorly differentiated and easy to metastasized. None of current treatments, including chemotherapy, radiotherapy, targeted therapy and immunotherapy, has shown good results [4,5].
Although the histopathological classifications currently remain the most commonly used for therapy decision making in the clinical setting, molecular classifications have been developed to guide future treatment development [5]. Several different molecular GC classification systems have been proposed these years, attempting to relate molecular features to histological phenotypes and clinical features [6,7]. In 2014, The Cancer Genome Atlas (TCGA) research network proposed four molecularly distinct GC subtypes: Epstein-Barr virus infected (EBV), microsatellite instability (MSI), genomically stable (GS), and chromosomal instability (CIN) [8]. Also, in 2015, the Asian Cancer Research Group classified GC into four molecular subtypes: mesenchymal-like, microsatelliteunstable, the tumor protein 53 (TP53)-active and TP53inactive types [9]. However, there are few detailed studies on the mutational spectrum of PC GC. Further investigation of the genetic alterations may provide useful information to explore new therapies for PC GC patients.
In this study, we compared the clinical characteristics and genomic profiling of PC GC patients and non-PC GC patients. Furthermore, we investigated whether genetic alterations are associated with patient prognosis in PC GC. Finally, we tried to seek new therapies according to specific genetic alterations of PC GC.

Patients and samples
We collected 556 patient samples of gastric adenocarcinoma. All histopathologic diagnoses were reviewed by at least two senior pathologists independently. Clinical information was retrospectively collected and the overall survival (OS) was measured from the date of surgery to the date of death or the last follow-up visit. The last follow-up date was August 8th, 2016. This study was conducted in accordance with the code of ethics of the World Medical Association (Declaration of Helsinki) and approved by the Ethics Committee of Nanjing Drum Tower Hospital (No. 2016-196-01).

Nucleic acid preparations and next-generation sequencing
Formalin fixed paraffin-embedded (FFPE) tissues obtained from patients with GC were collected for NGS in a 450-gene panel assay, whose accuracy has been verified before (Additional file 1: Figure S1). DNA was extracted from the unstained FFPE sections with tumor content of no less than 20% and was fragmented to ~ 250 bp by sonication. A library was constructed using the KAPA Hyper Prep Kit (KAPA Biosystems) and hybridization capture was performed with a custom panel containing individually synthesized 5ʹ-biotinylated DNA probes. The capture probes target exons of 450 cancer-related genes and frequently rearranged introns of 38 genes. Paired-end sequencing was performed according to the manufacturer's protocols. Genomic alterations, including single base substitutions, short and long insertions/deletions (INDELs), copy number variations, and gene rearrangements, were assessed using the OrigiMedpipeline. Genomic alterations relevant for cancer immunotherapy, which included TMB levels and MSI, were also evaluated [10,11]. Mutation signature was predicted using a public software deconstuctSigs. All substitution mutations were classified into 96 kinds of trinucleotides and the frequency of each was precisely calculated as the characteristic signature of these samples. The signature was then compared with the typical 30 signatures from COSMIC to identify the most similar combination and the percentage of each contributed [12].

Cell lines
To verify the function of TACC2-FGFR2 fusion, we generated gastric cancer cell lines with stable expression of TACC2-FGFR2. The Lenti-EF1a-EGFP-pGK-Puro retroviral vectors containing the particular fusion genes were transfected into 293 T cells to produce virus. MKN45 and NUGC4 cells were then infected with the viral supernatant containing expression constructs. Stable transfectants were obtained and maintained under selection pressure by puromycin dihydrochloride (2 µg/ml, Beyotime, ST551-10 mg). Under selection conditions, clones were picked and maintained.

Quantitative real time polymerase chain reaction (qRT-PCR) analysis
Total RNAs were extracted by TRIzol ™ Reagent (Thermo Fisher Scientific, 15596018), then dissolved in RNasefree water. cDNA synthesis is performed in the first step using total RNA, Random Primers (Thermo Fisher Scientific, 48190011) and dNTP Mix (10 mM ea, 18427013), 5 min at 65 °C. In the second step, PCR is performed in a separate tube using probes specific for the gene of interest, M-MLV Reverse Transcription (200 U/ul, 28025013) and RNase OUT ™ Recombinant Ribonuclease Inhibitor (Thermo Fisher Scientific, 10777019), following the procedures of 10 min at 25 °C, 45 min at 37 °C, 10 min at 70 °C, and then 4 °C. TaqMan ™ Fast Advanced Master Mix (Thermo Fisher Scientific, 444556) was used with the following PCR parameters, 1 cycle of 20 s at 95 °C, 40 cycles of 1 s at 95 °C and 20 s at 60 °C using QuantStudio ™ 7 Flex (Applied Biosystems). Probes used in this study are FGFR2 (Thermo Fisher Scientific, Hs01552918_m1) and β-actin (Thermo Fisher Scientific, Hs99999903_m1).

Immunohistochemical staining of tumor tissues
The expression of PD-1, PD-L1, CD3, FGFR2 in tumors was evaluated via immunohistochemical analysis (anti-PD-1, 1:100, NAT105, Cell Marque; anti-PD-L1, 1:100, SP142, Spring Bio; anti-CD3, 1:500, CD3-565-L-CE, Leica/Novocastra; anti-FGFR2, 1:200, ab58201, Abcam). The PD-1, PD-L1 is observable in the cytoplasm or on the membrane of the tumor cell or the TILs. The immunoreactivity of PD-1, PD-L1 was evaluated semiquantitatively according to the percentage and intensity of positive cells. Specimens in which PD-1 or PD-L1 were observed in more than 1% of tumor cells or immune cells were considered PD-1 or PD-L1 positive. CD3 was detected in the nuclei of the TILs. The distribution of CD3+ TILs was observed in the areas with the highest density of TILs first at low magnification. The amount of positive TILs was then recorded at high magnification (HPF 400× magnification). The number of CD3+ TILs was determined in 30 random high power fields in each section.

Fluorescence in situ hybridization (FISH)
Two-color FISH was performed on 2 µm thick sections from formalin-fixed, paraffin-embedded tumor tissues with FGFR2 rearrangements and paired normal tissues. Before hybridization, sections were deparaffinized, dehydrated in 100% ethanol, and air-dried. Commercially available, locus-specific FGFR2 probe (anbiping, F.01197-01) were used according to the manufacturer's recommendations. 377-kb Spectrum Green directly labeled fluorescent DNA in the 3ʹ of FGFR2 and 446-kb Spectrum Red directly labeled fluorescent DNA in the 5ʹ of FGFR2.

Statistical analysis
Statistical analysis was performed by SPSS statistics software, version 25.0 (SPSS, Chicago, IL, USA). The continuous variables were tested for normal distribution before analyzing by t test. The categorical variables were taken with the Pearson's Chi-square test (or Fisher's exact test). Impact of clinical characteristics and genetic alterations on survival outcomes were estimated by using Kaplan-Meier method, Cox proportional hazard modeling. A two-tailed P value of less than 0.05 was regarded as statistically significant.

Clinical characteristics and genomic profiling of PC GC and non-PC GC patients
In this study, a total of 556 GC samples were included. Among all GC patients, 64 (11.5%) are diagnosed as PC GC and 492 (88.5%) are non-PC GC. We compared the clinical characteristics between PC GC and non-PC GC patients (Table 1). It showed that PC GC patients were younger than non-PC GC (proportion of patients younger than 60 years old was 65.6% vs. 46.3%, P = 0.004) ( Table 1). Also, PC GC patients are more likely to be female, although there was no significant difference (proportion of female patients was 40.6% vs. 32.7%, P = 0.208) ( Table 1).

The value of gene rearrangements in predicting prognosis for PC GC
We detected gene rearrangements in 20.3% (13/64) PC GC patients (Fig. 1b). Specially, FGFR2 rearrangements (FGFR2/VTI1A and FGFR2/TACC2) were recurrently detected in 3.1% (2/64) PC GC tumor samples. We collected clinical characteristics and prognostic information of 28 PC GC patients from Drum Tower Hospital. It demonstrated that PC GC patients with gene rearrangements had a higher N stage (P = 0.001) and a higher tumor stage (P = 0.010) defined by 8th American Joint Committee on Cancer (AICC) criterion (Table 2). Moreover, patients harboring gene rearrangements (n = 9) had a shorter overall survival (OS) in comparison with patients without any gene rearrangement (n = 19) (16.0 months vs 21.0 months, P = 0.043) (Fig. 2a). To further explore the risk factors related to survival outcome of PC GC, we employed univariate and multivariate Cox regression analyses to identify protective or adverse prognostic factors. As shown in Additional file 1:  (Fig. 2b). Among IIIA-B stage (HR = 2.64, 95% CI 1.03-6.78, P = 0.043), PD-L1 positive (HR = 4.74, 95% CI 1.04-21.63, P = 0.045) PC GC patients, gene rearrangement was a negative factor for overall survival (Fig. 2b).

Function of TACC2-FGFR2 fusion in PC GC
Fluorescence in situ hybridization (FISH) and RNA sequencing (RNA-seq) was performed in two samples with FGFR2 rearrangements (FGFR2/VTI1A and FGFR2/TACC2). FGFR2/VTI1A and FGFR2/TACC2 rearrangements could both be detected by FISH in corresponding patient's samples (Additional file 1: Figure   Figure S3). However, no RNA product of FGFR2-VTI1A rearrangement was detected by RNA-seq. We presumed that it might be due to no RNA product was transcribed by FGFR2-VTI1A rearrangement indeed. Next, we evaluated the expression level of FGFR2 protein in patient's surgical samples with TACC2-FGFR2 using immunohistochemistry (IHC), finding increased FGFR2 expression in tumor area compared to adjacent normal surface epithelial area (Fig. 3b).
However, these molecular characteristics of PC GC patients are still unclear for treatment and prognosis. To better understand the molecular characteristics of PC GC, we respectively analyzed SNVs and INDELs, CNVs and gene fusions and rearrangements. We found that 20% (13/64) PC GC patients harboring somatic gene rearrangements. And, patients with rearrangements (n = 9) had a shorter overall survival (OS) in comparison with patients without any gene rearrangement (n = 19) (16.0 months vs 21.0 months, P = 0.043) (Fig. 3b). It may owe to gene rearrangements could form severe cancer driving structural variants (SV), including insertions, deletions, tandem duplications, inversions, translocations, and more complex rearrangements [19]. This is an indication that gene rearrangements play an important role in PC GC development and prognosis. Gene rearrangement might be an adverse prognostic factor for PC GC patients and more effective therapies targeting gene rearrangements should be studied for PC GC. FGFR2, known as fibroblast growth factor receptor-2, is a transmembrane tyrosine kinase receptor, regulating cell proliferation, survival, migration and angiogenesis [20]. Genetic alterations in FGFR2, including gene amplification, mutations or rearrangements may dysregulate the FGF signaling pathway, influence the development and progression of various cancers by activating the downstream PI3K-AKT and MAPK-ERK pathways [21]. Previous study demonstrated that FGFR2 overexpression, mainly due to FGFR2 gene amplification, is associated with poor pathological features, including deeper tumor invasion, more LN metastasis, advanced tumor stage, and worse survival in GC [22]. And, many researches indicated that the incidence of FGFR2 overexpression might differ between histological subtypes according to Lauren's classification [21]. Hattori found FGFR2 overexpression in 52.6% (20/38) diffuse-type GCs, but in none of 11 intestinal-type GCs [23]. PC GC tend to be diffuse-type according to Lauren's classification, however, no study verified the situation of FGFR2 genomic alterations in PC GC. Our results firstly showed that FGFR2 is more frequently mutated in PC GC than non-PC GC (14% vs. 6%, P = 0.037), including 12.5% CNVs. Several small-molecule inhibitors and antibodies for FGFR2 are under clinical trials [24,25]. Recently, Catenacci [25] verified bemarituzumab, an IgG1 antibody specific for the FGFR2b receptor, seemed to be well tolerated and demonstrated single-agent activity as lateline therapy in patients with advanced-stage gastric and gastroesophageal junction adenocarcinoma (GEA). Moreover, bemarituzumab is currently being evaluated in combination with chemotherapy in a phase III trial as front-line therapy for patients with high FGFR2b-overexpressing advanced-stage GEA. Preliminary results showed that bemarituzumab, added to mFOLFOX6 chemotherapy, led to clinically meaningful and statistically significant improvements in PFS (9.5 m vs. 7.4 m, P = 0.0727), OS (not reach vs. 12.9 m, P = 0.0268) and ORR (47% vs. 33%). And, the higher the FGFR2 expression, the better the prognosis [24]. FGFR2 inhibitors may be a good choice for PC GC with FGFR2 CNVs. In addition, PC GC patients not only have more FGFR2 CNVs, but also have recurrent FGFR2 rearrangement (FGFR2/VTI1A and FGFR2/TACC2) in our study. FGFR rearrangements present in 13% to 17% of intrahepatic cholangiocarcinoma and represent driver mutations [26]. Several pre-clinical studies and clinical trials have demonstrated that FGFR2 rearrangements in cholangiocarcinoma can predict tumor sensitivity to FGFR2 inhibitors and become an important therapy in these highly selected patients [27][28][29]. To verify whether FGFR2 rearrangements can be a marker for targeted therapy in PC GC patients, we transfected TACC2-FGFR2 construct into GC cell lines (MKN45 and NUGC4), which has not been studied in PC GC before. We chose FGFR2 inhibitors, including AZD4547, Erdafitinib, and BGJ398 to treat GC cell lines with or without TACC2-FGFR2 fusion, and got similar results. Recently, many FGFR2 inhibitors are in the pipeline, these three drugs belonged to the first FGFR2 inhibitors studied. AZD4547 has been used in clinical trials of gastric adenocarcinoma [30], and Erdafitinib (Balversa ™ , Janssen Pharmaceutical Companies) [31] and BGJ398 (Infigratinib) [27,32] was or will be approved by FDA for treatment of tumor with FGFR2 alterations. Our results firstly verified that the FGFR2 mRNA and protein expression level (Fig. 4b, c) were increased in GC cells with FGFR2 rearrangement and cells became more sensitive to FGFR2 inhibitors.

Conclusion
In conclusion, we firstly identified FGFR2 alteration was more frequently among PC GC than non-PC GC, including CNVs and rearrangements. Moreover, we verified TACC2-FGFR2 fusion could increase FGFR2 expression in mRNA and protein level, and GC cell lines with TACC2-FGFR2 fusion were more sensitive to FGFR2 inhibitors. All these results suggested that FGFR2 may be a potential therapeutic target for PC GC.