Introduction
Intrahepatic cholangiocarcinoma (ICC) is an aggressive malignancy with low incidence and poor prognosis. The 5-year survival for ICC is only 14–40%, whereas the rate of recurrence and metastasis is as high as 60–70%.1,2 Radical surgical resection remains the sole curative method, which is seldom performed because of the lack of early diagnostic markers. The therapeutic effectiveness of traditional radiotherapy and chemotherapy are unsatisfactory for the treatment of ICCs. The median survival after first-line chemotherapy with gemcitabine plus platinum is less than 12 months in advanced ICC,3 which presents considerable challenges.
The fibroblast growth factor (FGF) family plays an important role in tumorigenesis and comprises 18 secreted signaling proteins. In the human body, extracellular FGFs activate four receptor tyrosine kinases (FGFR 1-4) and participate in the early stages of embryonic development, including organogenesis, glucose and lipid metabolism, tissue repair, and regeneration.4 Given its essential role, any abnormality in FGF-FGFR signal transduction may cause tumor formation by affecting cell survival, apoptosis, and migration. In ICC, the FGFR2 fusion protein promotes tumor generation, proliferation and angiogenesis by up-regulating the RAS, JAK and PI3K/mTOR pathways.5,6
Recently, study of molecular targets, such as FGFR2 rearrangement, NTRK fusion, and IDH1/2 mutations, has emerged as a focus in ICC. Inhibitors of FGFR2 in particular have facilitated a recent breakthrough in ICC-targeted therapy.7,8 Based on recent phase II clinical trials on patients with FGFR2 rearrangement, FGFR targeted therapy has an objective remission rate of 14–35% and a disease control rate of 75–83% for advanced ICCs, along with a prolonged progression-free survival of 5.8–6.9 months.9–11 Therefore, the USA’s Food and Drug Administration (FDA) has recently approved pemigatinib, a targeted FGFR inhibitor, to treat ICCs with FGFR2 rearrangement that have failed preceding chemotherapy.12 Targeted therapy for FGFR2 rearranged ICCs has shown promise to date, and several phase III clinical trials that are expected to bring radical changes to ICC treatment have been conducted.
In clinical diagnosis, molecular testing methods for FGFR2 rearrangement are diverse, and include fluorescence in situ hybridization (FISH), next-generation sequencing (NGS), and immunohistochemistry (IHC). Currently, most large-scale clinical trials that target FGFR2 rearrangements have established DNA-based NGS, which is costly in clinical practice, as one of the inclusion criteria. To date, there is a lack of well-established guidelines for FGFR2 detection in targeted therapy of ICC. In addition, the incidence and clinicopathological features of ICCs with FGFR2 rearrangement have not been fully assessed. FGFR2 molecular detection is expected to provide evidence-based recommendations for better screening of potential candidates. This study investigated the rearrangement of FGFR2 in ICCs using multiple molecular detection methods. We sought to determine 1) the ideal method for FGFR2 screening in ICCs in clinical applications and 2) the clinicopathological and molecular features of ICCs with FGFR2 rearrangements.
Methods
Case selection and sample preparation
In this study, 167 consecutive ICC cases archived at Zhongshan Hospital, Fudan University between January 2014 and December 2017 were retrospectively analyzed. All ICCs were surgically resected with formalin-fixed paraffin-embedded (FFPE) tissues and hematoxylin and eosin (H&E)-stained slides. The inclusion and exclusion criteria were pathologically confirmed ICC after surgical resection, and no evidence of preoperative distant metastasis, and having received no anti-cancer treatment. Tissue microarrays (TMAs) were constructed from all FFPE specimens using a TMA instrument (Bonan Co. Ltd., Shanghai, China). According to the H&E slides marked by a pathologist, two 0.2 mm cores from tumor areas and one from the normal liver were punched out of each FFPE tissue and arrayed into recipient TMA blocks. FISH and IHC analyses were performed for all 167 cases using TMA blocks. Four unstained FFPE tumor sections (4 µm thick) and paired normal liver sections were prepared for targeted DNA sequencing. All study procedures were approved by the Zhongshan Hospital Research Ethics Committee (B2020-194) and performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from each patient.
Pathology and clinical data
Clinical data, including age, sex, symptoms of jaundice, serum hepatitis B surface antibody positivity, liver cirrhosis, primary biliary cholangitis, primary sclerosing cholangitis, hepatolithiasis, liver fluke infection, and presurgery serum CA199, AFP, and CEA levels, were retrieved from the hospital medical records. Pathological data, including gross features, tumor grade, histological subtype, microvascular invasion, perineural invasion, perihilar invasion, portal tract invasion, liver capsule invasion, mucus production, and major necrosis, were interpreted carefully by two pathologists (YJ and YZ) according to the literature.13 The histological subtype was determined according to the 2019 World Health Organization diagnostic criteria for ICC classification.14 Pathological tumor-node-metastasis stage (pTNM) was determined according to the eighth edition of the American Joint Committee on Cancer/Union for ICC.15
FISH analysis
FISH for FGFR2 rearrangement was performed on the 4 µm thick TMA sections using a commercially available probe kit in accordance with the manufacturer’s recommendations (10q26 gene break-apart probe set, Linked Biotech Pathology Co. LTD, Guangzhou, China). In brief, xylene was used to deparaffinize the sections. Sections were rehydrated in a 100%, 85%, 70% alcohol gradient and then boiled for 20 m and air dried, followed by digestion in a proteinase K solution (0.05 mg/mL, pH 7.0). After that, the FGFR2 testing probe was added, and hybridization was carried out at 80°C for 5 m and 37°C overnight. Before hydration and air-drying, sections were immersed in 2× saline sodium citrate (2 × SSC) buffer for 5 m and washed in NP40/2×SSC for 7 m. Sections were counterstained with 4′,6-diamidino-2-phenylindole, DAPI. The dual-labeled probe hybridized with the neighboring 5′-telomere (STSG-72444, labeled with spectrum red) and 3′-centromere (SHGC-85446, labeled with spectrum green) FGFR2c sequences.
FISH slides were interpreted by two pathologists using a fluorescence microscope (Olympus BX43, Olympus Optical Co. LTD, Tokyo, Japan). Owing to the colocalization of the green and red spectra, the unsplit 5′/3′ spectrum was shown in yellow. We defined the 5′ and 3′ probe signals located at a distance greater than 1× the single signal size as split positive signals. Thus, the rearrangement-positive cell contained any split signals above (YGR type) and an isolated red signal (YR type). Cells with an isolated green signal (YG type), were considered negative, as that denoted deletion of the 5′ sequence of the FGFR2 gene.16 Using those criteria, at least 100 non-overlapping nuclei of tumor cells were examined for each case, and the percentage of rearrangement-positive cells was calculated. Based on the results of a previous study,5 a 15% positive cell ratio was adopted as the cutoff value for FISH analysis.
IHC analysis
IHC staining was performed on 4 μm thick TMA sections using an Ascend Aliya autostainer (Ascend Microsystems, Guangzhou, China). A commercially available rabbit monoclonal FGFR2 antibody (#23328, clone D4L2V, 1:500; Cell Signaling Technology, Danvers, MA, USA) was used for the FGFR2 IHC staining. FGFR2 protein expression was evaluated by two pathologists blindly and separately. According to the manufacturer’s protocol, cytoplasmic and/or membrane staining was considered positive and nuclear staining was considered negative. In accordance with the IHC grading in a previous study,17 IHC-FGFR2 staining was classified as 0 (negative staining), 1+ (poor to moderate staining), and 2+ (strong staining). Cases with 1-2+ staining were defined as IHC-FGFR2 positive, and as negative otherwise.
An anti-programmed cell death-ligand 1 (PD-L1) 22C3 antibody (M3653; Dako, Glostrup, Denmark) was used for the PD-L1 IHC test. Two standardized counting methods, the tumor proportion score (TPS) and combined positive score (CPS),18 were adopted for PD-L1 interpretation. The optimal cutoff values of PD-L1 positivity, estimated using X-tile software,19 were 5 for CPS and 5% for TPS interpretation.
Targeted DNA sequencing
Genomic DNA was extracted from three or four 4 µm FFPE tumor tissue slides and paired peritumoral normal liver tissue. FFPE samples containing at least 20% tumor cells were used for NGS detection. At least 50 ng of cancer tissue DNA was extracted from each 40 mm3 FFPE tumor sample. Targeted DNA sequencing was performed at OrigiMed Co. Ltd. (Shanghai, China). SU-450 panel and Illumina platforms were used for sequencing. Hybridization capture was used in the NGS platform. The detection targets included all known and unknown FGFR2 rearrangements, and 450 cancer-related genes. Gene variants, such as single nucleotide variation (SNV), long indels, and substitutions, were detected with an effective sequencing depth of 800–1,000×. Genomic DNA was isolated by using QIAamp DNA FFPE Tissue Kit and QIAamp DNA Blood Midi Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. The concentration of DNA was measured by Qubit and normalized to 20–50 ng/µL. All the sequencing data were obtained by using Illumina NextSeq 500 (Illumina, Inc., San Diego, CA, USA) in the laboratory certified by College of American Pathologists (CAP) and Clinical Laboratory Improvement Amendments (CLIA).
Statistical analysis
The chi-square and Fisher’s exact tests were used to analyze the association between FGFR2 rearrangement and clinicopathological features. Univariate survival analysis for ICC with rearranged FGFR2 was performed using the Kaplan-Meier curve and log-rank test. The McNemar paired chi-square and kappa consistency tests were used to compare the concordance of results detected using IHC, FISH, and NGS. Statistical significance was set at p<0.05.
Results
Baseline characteristics of 167 ICC patients are shown in Table 1.
Table 1Patient and tumor characteristics of the cohort
| Characteristics | Frequencies, n=167 |
|---|
| Sex | |
| Male | 104 (62) |
| Female | 63 (38) |
| Age in years | |
| >65 | 56 (34) |
| ≤65 | 111 (66) |
| pTNM stage | |
| I | 49 (29) |
| II | 52 (31) |
| III | 47 (28) |
| IV | 19 (11) |
| Cirrhosis | |
| Yes | 16 (10) |
| No | 151 (90) |
| PBC | |
| Yes | 4 (2) |
| No | 163 (98) |
| PSC | |
| Yes | 3 (2) |
| No | 164 (98) |
| Hepatolithiasis | |
| Yes | 8 (5) |
| No | 159 (95) |
| Serum HBsAb positivity | |
| Positive | 63 (38) |
| Negative | 104 (62) |
| Tumor grade | |
| Poorly differentiated | 56 (34) |
| Moderate/well differentiated | 111 (66) |
| Histological subtype | |
| Large duct | 55 (33) |
| Small duct | 93 (56) |
| Unclassified | 19 (11) |
| Median overall survival in months (range) | 34 (0–67) |
FGFR2 gene rearrangement detected by FISH
Of the 167 cases, 21 were deemed FGFR2 rearrangement-positive (FISH-FGFR2 positive). The vast majority of FISH patterns showed typical positive signals, as previously defined. More specifically, 18/21 exhibited the YGR pattern (Fig. 1A, B), whereas the other two had the YR pattern (Fig. 1C, D). The remaining one case, was atypical, with separate red and green signals as described, accompanied by clusters of green signals (Fig. 1E, F). The percentage of positive cells in FISH-FGFR2 positive cases ranged from 17–80%, with an average of 53%. FISH-FGFR2 negative cases had background cell signal ratios that ranged from 0–14%, with an average of 7.8%. Among the 21 FISH-FGFR2 positive cases, men had a slight advantage over women (12 vs. 9). Histologically, the small duct subtype accounted for the majority of cases (18/21, 86%; Table 2).
Table 2Cases with FGFR2 rearrangements by FISH and NGS
| Case | Sex | Age | Stage | HBV/HCV | Cirrhosis | Histological Subtype | FISH+ cell rate, % | FISH signal type | IHC | NGS
|
|---|
| Type | Partner |
|---|
| 1 | M | 61 | IV | (+) | N | Large | 68 | YGR | Weak | Fusion | FGFR2 | GPHN |
| 2 | F | 67 | IIIB | (−) | N | Small | 76 | YGR | Strong | Fusion | FGFR2 | NRBF2 |
| 3 | M | 30 | IB | (+) | N | Small | 32 | YGR | Strong | Fusion | FGFR2 | BICC1 |
| 4 | F | 66 | IA | (+) | Y | Small | 60 | YGR | Negative | Rearrangement | MAPK1IP1L | FGFR2 |
| 5 | F | 73 | IA | (+) | N | Small | 72 | YGR | Negative | Fusion+Substitution | FGFR2 | BICC1 |
| 6 | M | 53 | IA | (−) | N | Small | 47 | YGR | Negative | Fusion | FGFR2 | NOL4 |
| 7 | F | 42 | IIIB | (−) | N | Small | 74 | YGR | Nuclear | Rearrangement | SYT1 | FGFR2 |
| 8 | M | 50 | II | (−) | N | Unclassified | 49 | YGR | Weak | Fusion | FGFR2 | AHCYL1 |
| 9 | F | 41 | IA | (+) | N | Small | 61 | YGR | Negative | Fusion | FGFR2 | AHCYL1 |
| 10 | M | 69 | IA | (−) | N | Small | 71 | YGR | Negative | Rearrangement | FGFR2 | intergenic |
| 11 | F | 52 | II | (−) | N | Unclassified | 80 | YR | Negative | Fusion | FGFR2 | SHTN1 |
| 12 | M | 41 | IA | (+) | N | Small | 40 | YGR | Negative | Fusion | FGFR2 | ETV6 |
| 13 | F | 53 | II | (−) | N | Small | 60 | YGR | Weak | Fusion | FGFR2 | INA |
| 14 | M | 60 | IA | (+) | Y | Small | 36 | YR | Weak | Rearrangement | FGFR2 | SORBS1 |
| 15 | F | 67 | IIIB | (−) | N | Small | 56 | YGR | Weak | Fusion | FGFR2 | BICC1 |
| 16 | M | 74 | II | (−) | N | Small | 56 | YGR | Weak | Fusion | FGFR2 | BICC1 |
| 17 | M | 57 | II | (−) | N | Small | 28 | YGR | Weak | Substitution | – | – |
| 18 | M | 53 | II | (−) | N | Small | 17 | YGR | Weak | Not detected | – | – |
| 19 | M | 62 | II | (+) | N | Small | 22 | YGR | Negative | Not detected | – | – |
| 20 | M | 60 | II | (+) | Y | Small | 56 | Atypical | Negative | Not detected | – | – |
| 21 | F | 71 | IA | (−) | N | Small | 62 | YGR | Weak | Not tested | – | – |
FGFR2 gene rearrangement detected by targeted NGS
Targeted NGS was successfully performed on all 20 FISH-FGFR2 positive cases and eight matching FISH-FGFR2-negative cases. One FISH-positive sample (case 21) failed the NGS test because of DNA extraction. DNA sequencing was used to target 450 cancer-related genes, and identified 191 somatic mutations, including 102 single nucleotide variants, 59 copy number variations, 26 fusions/rearrangements, and four long indels (Fig. 2). The average number of somatic mutations in each case was 3.8 (range, 0–45), and the average mutation Mb was 2.23 (range: 0–7.3). G>A, G>T, A>G, and C>T transitions were primarily detected in abnormal FGFR2 gene samples. The mutation spectrum of the 20 most common genes in ICCs is shown in Figure 3. Mutations of the BAP1 gene were the most common, second only to FGFR2. TP53 is the most commonly mutated gene in ICCs with normal FGFR2 genes.
NGS-FGFR2 alteration was detected in 18 cases, of which 15 were rearrangements, two were substitutions, and one had both FGFR2 rearrangements and mutations. Of the 16 FGFR2 rearranged cases, 11 fused FGFR2 partner genes were identified, including GPHN, NRBF2, BICC1, MAPK1IP1L, NOL4, SYT1, AHCYL1, SHTN1, ETV6, INA, and SORBS1. Among those, FGFR2-BICC1 was the most common fusion type (6/16, 36%; Table 2). KRAS and IDH1 mutations and FGFR2 rearrangements were mutually exclusive; BAP1 (28%), CDKN2A (28%), and CDKN2B (17%) were common concomitant genetic alterations of FGFR2 (Fig. 3).
FGFR2 protein expression detected by IHC
Cases were divided by IHC into FGFR2-negative (Fig. 4A) and positive groups (+/++; Fig. 4C, D). Of the 167 patients, 80 (48%) were positive for IHC-FGFR2,and 87 (52%) were negative. One case had nuclear staining (Fig. 4B). As stated above, nuclear staining was classified as negative. However, this case with nuclear staining was detected SYT1:FGFR2 rearrangement by NGS, which is worthy of further study.
Concordance of FGFR2 status by FISH, NGS, and IHC
Based on our data, the results of FISH consistence with those obtained by NGS using the consistency test (Kappa value=0.696, p<0.001; Table 3). The four cases with discordance in FISH and NGS are shown in Table 2, including case 17 (FISH+/NGS− with substitution), cases 18 and 19 (FISH+/NGS−), and case 20 (FISH+ with atypical pattern/NGS−). However, the IHC results failed the consistency test for both FISH (Kappa value=0.048, p=0.365; Table 3) and NGS (Kappa value=0.125, p=0.508; Table 3). IHC results were discordant with those of FISH and NGS.
Table 3Consistency test for FISH, NGS, and IHC
| Test | FISH | Positive | | Negative |
|---|
| positive | 12 | | 68 |
| IHC | | | | |
| negative | 9 | | 78 |
| McNemar’s chi-squared=43.688, df=1, p-value=3.851e-11 |
| Kappa value | 0.048 | p=0.365 | |
| FISH | positive | | negative |
| positive | 16 | | 0 |
| NGS | | | | |
| negative | 4 | | 8 |
| McNemar’s chi-squared=2.25, df=1, p-value=0.1336 |
| Kappa value | 0.696 | p<0.0001 | |
| IHC | positive | | negative |
| positive | 10 | | 6 |
| NGS | | | | |
| negative | 6 | | 6 |
| McNemar’s chi-squared=0, df=1, p-value=1 |
| Kappa value | 0.125 | p=0.508 | |
Clinicopathological characteristics and prognosis related to FGFR2 positivity in ICCs
The clinicopathological characteristics of cases with FGFR2 positivity were evaluated. Histologically, the small duct subtype was significantly related to FGFR2 positivity (p=0.003; Table 4) with markedly reduced mucus secretion (p=0.025; Table 4). In addition, FGFR2 rearrangements were more common in cases of early-stage disease (stage I–II, p=0.041; Table 4). Macroscopic mass-forming type, no history of hepatolithiasis or liver fluke, and low preoperative serum AFP level (<20 ng/mL) were found in all 21 cases of FGFR2 rearrangement, although statistical significance was not achieved when compared with cases of normal FGFR2. Possible factors evaluated in other studies, including younger age and serum HBsAb positivity, were unrelated to the FGFR2 status in our cohort (p>0.05; Table 4). Kaplan-Meier survival analysis showed that FGFR2-positive cases had better overall survival (p=0.013; Fig. 5).
Table 4Univariate analysis of FGFR2-positive patients determined by FISH (n=167)
| Parameter | Positive, n=21 | Negative, n=146 | p-value |
|---|
| Sex | | | |
| Female | 9 | 54 | 0.604 |
| Male | 12 | 92 | |
| Age | | | |
| >65 | 7 | 49 | 0.984 |
| ≤65 | 14 | 97 | |
| Jaundice | | | |
| Yes | 1 | 3 | 0.419 |
| No | 20 | 143 | |
| Serum HBsAb | | | |
| Positive | 9 | 54 | 0.604 |
| Negative | 12 | 92 | |
| Hepatolithiasis | | | |
| Yes | 0 | 8 | 0.598 |
| No | 21 | 138 | |
| Liver fluke infection | | | |
| Yes | 0 | 13 | 0.375 |
| No | 21 | 133 | |
| Cirrhosis | | | |
| Yes | 2 | 14 | 0.992 |
| No | 19 | 132 | |
| PBC | | | |
| Yes | 1 | 3 | 0.419 |
| No | 20 | 143 | |
| PSC | | | |
| Yes | 0 | 3 | 0.507 |
| No | 21 | 143 | |
| pTNM stage | | | |
| I–II | 17 | 84 | 0.041* |
| III–IV | 4 | 62 | |
| T Stage | | | |
| T1–2 | 20 | 125 | 0.315 |
| T3–4 | 1 | 21 | |
| N Stage | | | |
| N0 | 18 | 110 | 0.411 |
| N1 | 3 | 36 | |
| M Stage | | | |
| M0 | 20 | 128 | 0.473 |
| M1 | 1 | 18 | |
| Serum CA199 in U/mL | | | |
| <37 | 11 | 70 | 0.703 |
| ≥37 | 10 | 76 | |
| Serum AFP in ng/mL | | | |
| <20 | 21 | 141 | 0.861 |
| ≥20 | 0 | 5 | |
| Serum CEA in ng/mL | | | |
| <5 | 17 | 109 | 0.531 |
| ≥5 | 4 | 37 | |
| Gross feature | | | |
| Mass forming | 21 | 125 | 0.349 |
| Periductal infiltrating | 0 | 6 | |
| Mixed | 0 | 15 | |
| Tumor grade | | | |
| Poorly differentiated | 4 | 52 | 0.133 |
| Moderate/well differentiated | 17 | 94 | |
| Histological subtype | | | |
| Large duct | 1 | 54 | 0.003* |
| Small duct | 18 | 75 | |
| Unclassified | 2 | 17 | |
| Margin positivity | | | |
| Positive | 0 | 5 | 0.901 |
| Negative | 21 | 141 | |
| Major necrosis | | | |
| Yes | 10 | 59 | 0.531 |
| No | 11 | 87 | |
| Microvascular invasion | | |
| Positive | 10 | 47 | 0.163 |
| Negative | 11 | 99 | |
| Perineural invasion | | | |
| Positive | 5 | 49 | 0.372 |
| Negative | 16 | 97 | |
| Portal tract invasion | | | |
| Positive | 3 | 24 | 0.802 |
| Negative | 18 | 122 | |
| Perihilar invasion | | | |
| Positive | 1 | 19 | 0.474 |
| Negative | 20 | 127 | |
| Liver capsule invasion | | | |
| Positive | 13 | 99 | 0.591 |
| Negative | 8 | 47 | |
| Mucin production | | | |
| Yes | 2 | 49 | 0.025* |
| No | 19 | 97 | |
| PD-L1TPS | | | |
| Positive | 3 | 44 | 0.131 |
| Negative | 18 | 102 | |
| PD-L1CPS | | | |
| Positive | 8 | 76 | 0.336 |
| Negative | 13 | 70 | |
According to the PD-L1 interpretation criteria, 84 cases (50%) were estimated as PD-L1-CPS positive in this study, and 47 cases (28%) were estimated as PD-L1-TPS positive. PD-L1 positivity, interpreted using both TPS and CPS, was not correlated with FGFR2 rearrangement in our cohort (p=0.131 and p=0.336, respectively; Table 4).
Discussion
ICC is a rare malignancy of the biliary tract associated with poor prognosis. As mentioned, long-term survival of ICC can be improved by surgical resection or first-line chemotherapy with gemcitabine and cisplatin in rare cases. With the recent emergence of sequencing technology, molecular therapies targeting biliary malignancies have developed significantly.20FGFR2 rearrangement is considered an important oncogenic change in ICC. Therapies targeting the FGFR2 gene have achieved satisfactory results in several clinical trials and are expected to be applied in clinical practice soon.21–23
At present, molecular testing of tumor-related fusions, such as FGFR2 and ALK, are diverse and include targeted DNA sequencing, RNA sequencing, multiplex RT-PCR, FISH, and IHC. Compared to the well-established guidelines for ALK detection in targeted therapy of non-small cell lung cancer,24FGFR2 targeted therapy lacks clear evidence-based recommendations for molecular testing to screen potential candidates. Drawing on experiences using FISH in ALK detection, FISH analysis enables visual observation of the cell morphology and direct detection of the FGFR2 gene break status at the DNA level. This offers high sensitivity and specificity, regardless of the limitation caused by specimen preservation. Although FISH cannot identify specific fusion partners and breakpoints when compared with NGS, FISH is more feasible in large groups owing to short turnaround time and lower cost. Therefore, this study adopted dual-color break-apart probes for FISH-FGFR2 detection throughout the cohort. DNA-based targeted second-generation sequencing was used to verify the FISH results. Consequently, 12.6% of our ICC cohort was deemed FISH-FGFR2 positive, which was in accordance with the 9–14% reported previously.8 Compared with the NGS results, four FISH false-positive cases emerged, which made us focus on FISH interpretation, including both the cutoff percentage of tumor nuclei with FGFR2 positivity and the discrimination of atypical FISH patterns. Three of the FISH false-positive cases had reduced positive cell ratios of <30%, but the ratios in positive cases confirmed by NGS were >30%. In fact, most were >50%. A similar situation was observed in a recent study by Maruki et al.16 Based on the results of a small preliminary test, Maruki et al.16 defined a cutoff value of 7%, but RNA sequencing indicated that two FISH false-positive cases had reduced FISH ratios that were close to the cutoff value, with other positive cases all having distinct values of >15%. This study found that false positivity in FISH may be derived from setting the cutoff value. Based on our results, the indicated cutoff value was 30%. The remaining false positive FISH case had an atypical FISH pattern that was characterized by a pair of separated red and green signals and accompanied by clusters of 3′-green signals. In contrast to typical break-apart signals (i.e. YGR or YR patterns), the interpretation of atypical FISH signals requires extra clinical attention and practice.25,26 Eventually, experience with targeted therapy may set the standard for interpreting these atypical signals. Overall, FISH is an ideal method to achieve preferable consistency with NGS when detecting FGFR2 rearrangements. In the future, we expect more evidence-based clinical trials to establish FISH as a recommendation for FGFR2 screening. Meanwhile, the results of our study suggest that when the positive cell ratio is <30% or an atypical FISH pattern appears, the ultimate discrimination of the results should be comprehensively considered with multiple methods.
Most large-scale clinical trials that targeted FGFR2 rearrangements established DNA-based NGS rather than FISH as a standardized screening method.12 NGS accurately identifies genomic breakpoints and precise fusion partners. FGFR2 alteration can be classified as a known fusion partner or in-frame with FGFR2, and the second with no identifiable partner or a partner out-of-frame with FGFR2 in the intergenic region.10 In this study, 14 of 16 NGS-FGFR2 positive cases had the first type of fusion/rearrangement. One case had the second type mentioned above, and one case had both FGFR2-BICC1 fusion and a long indel mutation. Current studies indicate that different types of rearrangement partners are expected to benefit comparably from FGFR2 targeted therapy. Whether there is a difference in therapeutic effect between the two rearrangement types remains to be studied.10 Cases of simultaneous FGFR2 fusion and mutation are rarely reported. A case study we performed illustrates the case of a 73-year-old woman with stage IA ICC and a history of hepatitis B but no liver cirrhosis, whose tumor was a histologically small duct subtype. No chemotherapy or targeted therapy was administered after surgery, and no recurrence was found during the 41-month follow-up. The clinical value of this case awaits further study through a long-term follow-up. Approximately 40 fusion partners have been identified.27 In our current study, we identified 11 fusion partners. FGFR2-BICC1, was the most common fusion type in our study, in concordance with the literature. A phase II clinical trial12 showed that FGFR2-BICC1 fusion cases exhibited no significant difference in response to FGFR2 inhibitors compared to other FGFR2 fusion types to date. FGFR2 mutations, however, are not considered genomic aberration (GA) candidates for targeted therapy,12 which were also found in two cases in our cohort. Considering that the diversity of aberration types may lead to different therapeutic effects of molecular inhibitors, we believe that in-depth research on specific GA classifications will be of great value.
FGFR2 aberrations are currently believed to occur mainly in ICC rather than other segments of the biliary tree.16,28 Several studies have shown that FGFR2 gene abnormalities are mutually exclusive to KRAS, IDH, and BRAF mutations.5,29,30 Our results found that one case of FGFR2 substitution was concomitant with a KRAS mutation, whereas no KRAS gene abnormality was found in FGFR2 rearrangement cases. IDH1 mutation was found in 2 cases, none of which was accompanied by FGFR2 aberration. In this study, we observed that BAP1 (28%), CDKN2A (28%), and CDKN2B (17%) were common concomitant genetic alterations of FGFR2. Among these, BAP1 is considered the most common concomitant alteration of FGFR2, which indicates a favorable prognosis and relatively indolent disease.27,31 Our FGFR2-impaired cases concomitant with BAP1 mutation had an average survival of 36 months, which was higher than that of patients with normal BAP1 (32 months). Nevertheless, the concomitant presence of FGFR2 in ICCs and its prognostic value need to be confirmed further using larger samples.
The IHC results in this present study were inconsistent that of both NGS and FISH. The low expression of FGFR2 fusion protein and deviation caused by tumor heterogeneity on TMA could be reasons for the blame. Therefore, this study suggests that IHC analysis using this antibody is not recommended as an alternative screening method before NGS or FISH.
Clinicopathologically, this study showed that the molecular subtype of FGFR2 positivity was associated with early clinical stage and predominately histological small-duct subtype with diminished mucus secretion, and is associated with favorable overall survival. The findings are consistent with those in previous reports.5,32 In addition, FGFR2 positivity is reportedly correlated to age, serum HBsAb positivity, and ductal growth pattern,5,13,32–34 were not confirmed in this study. However, the propensity for FGFR2-positive cholangiocarcinoma patients with HBV infection was found in a cohort with both intra- and extrahepatic tumors.5 Our cohort included only ICCs, which may have caused the difference in HBV infection propensity between the studies. Sequential application of FGFR2 and PD-L1 inhibitors can reportedly enhance the response to immune checkpoint inhibitors in patients with advanced urothelial cancer.35 However, PD-L1 positivity in ICCs with FGFR2 rearrangements has rarely been studied. Zhu et al.36 reported that PD-L1 positivity is enriched in ICCs with FGFR2 rearrangements, in contrast to our findings. Considering that the percentage of FGFR2 rearrangement in Zhu’s study was lower than reported, and their interpretation of PD-L1 positivity differed from ours, the association between PD-L1 expression and FGFR2 rearrangement in ICCs remains to be studied further. Overall, this study demonstrates that FGFR2-positivity is related to the unique clinicopathological features of ICCs. Patients with an early clinical stage, histologically small duct subtype, and diminished mucus secretion should be prioritized for FGFR2 screening.
Our study has some limitations. First, the sample size was relatively small, and multicenter studies including larger ICC cohorts need to be conducted. Second, none of the ICC cases with FGFR2 rearrangement in our study received FGFR inhibitor treatment. The corresponding response to FGFR inhibitors remains the decisive factor when choosing detection methods in targeted therapy, and data on ICCs treated with FGFR inhibitors should be collected further.
Conclusions
FISH achieved satisfactory concordance with NGS and has potential value for FGFR2 detection in targeted therapies. FGFR2 detection should be prioritized for unique clinical subgroups in ICC, which features a histological small duct subtype, early clinical stage, and reduced mucus production.
Abbreviations
- FGFR2:
fibroblast growth factor receptor 2
- FISH:
fluorescence in situ hybridization
- ICC:
intrahepatic cholangiocarcinoma
- IHC:
immunohistochemistry
- NGS:
next-generation sequencing
- PD-L1:
anti-programmed cell death-ligand 1
- TMAs:
tissue microarrays
Declarations
Ethical statement
The study procedures were approved by the Zhongshan Hospital Research Ethics Committee (B2020-194) and performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from each patient.
Data sharing statement
The original contributions presented in the study are included in the article/Supplementary Material. Further data and calculation tools are available by request via [email protected].
Funding
This study was supported by Shanghai Municipal Key Clinical Specialty (shslczdzk01302).
Conflict of interest
The authors have no conflict of interests related to this publication.
Authors’ contributions
Conceived the idea and designed the study (YJ), performed the study and drafted the article (YZ), revised the article (YJ), and conducted the data acquisition, analysis and interpretation (YZ, KZ, YP, JH, XZ, ZJ, YH, WG). All authors discussed the results and agreed to be accountable for all aspects of the work. All authors read and approved the final manuscript.