Morphological Correlates of KIT and PDGFRA Genotypes in Gastrointestinal Stromal Tumour

Objective: The aim of the study was to study the clinicopathological and immunohistochemical features of gastrointestinal stromal tumours and correlation with KIT/PDGFRA mutations. Material and Method: Eighty consecutive resected cases were genotyped for KIT exons 11, 9, 13, 17 and PDGFRA exons 18, 14, 12 and correlated with histomorphology by nonparametric tests. Results: Forty-seven cases (58.8%) were in the high-risk group. Males had higher rates of KIT exon 11 and PDGFRA exon 18 mutations than females (p=0.03). KIT and PDGFRA mutation frequencies were lower (58.8%) than western data showing KIT exon 11 mutation in 63.8%, KIT exon 9 mutation in 19% and PDGFRA exon 18 mutation in 17% of the cases. Extragastrointestinal stromal tumours (n=6) showed 100% mutation. KIT exon 11 deletion was associated with gastric location (60%) (p=0.04), spindle cells (63.3%), and high-risk stratification (66.6%) (p=0.01) while KIT exon 9 mutation was common in small intestine (66.7%) (p=0.04), in higher risk groups (66.7%) (p=0.01) and 75% of codon 502-503 duplications (p=0.03). PDGFRA 18 mutation was common in males (p=0.03), in gastric location (62.5%) (p=0.04), in cases showing mild to moderate atypia (62.5%) (p=0.01) and lower risk stratification (62.5%) (p=0.01). KIT/PDGFRA mutations were significantly associated with gender (p=0.03), location (p=0.04), nuclear atypia (p=0.01) and risk stratification (p=0.01). Conclusion: Morphological features and anatomic location may be useful in deciding molecular testing strategy, particularly in resource-limited settings, when a plethora of targetable mutations are present. An algorithm may be derived for genotyping with KIT exon 11 and PDGFRA exon 18 heading the list of targetable mutations. This approach may reduce financial burden on patients as well as workload on hospital staff.


INTRODUCTION
Gastrointestinal stromal tumour (GIST) is the most common mesenchymal neoplasms of the gastrointestinal tract (GIT) that arises from interstitial cells of Cajal (1).Morphologically, GISTs can be of spindle cell, epithelioid, or mixed phenotype that expresses CD117 (KIT) on immunohistochemistry.These tumours also express DOG1 and CD34 (2).Mutational analysis of GISTs have shown mutations in KIT mostly occurring in exon 11 followed by exon 9 and in PDGFRA exon 18 (3).KIT exon 11 mutations include deletions, point mutations and insertions.Certain morphological features are associated with KIT 11 deletions such as gastric location, spindle cell morphology and large tumour size (4,5).KIT 11 insertions are also associated with spindle cell morphology but are generally seen in non-gastric locations (6).GISTs harbouring PDGFRA mutations are common in gastric, omental and mesenteric tumours showing nuclear pleomorphism, epithelioid morphology, plasmacytoid cells and multinucleated giant cells (7).Understanding the molecular and morphological correlations is important and may help to prioritize the sequential testing for GIST mutations, particularly in a resource-limited setting in developing countries.This study aims to compare the histomorphological and immunohistochemical features with various genotypes.

MATERIALS and METHODS
Eighty consecutive resected GISTs received within a period of 8 years were included in the study.Clinical findings and follow-up data were retrieved from the hospital records.All cases were reviewed for tumour size, location, histological type, cellularity, presence of atypia, necrosis, mitosis, secondary changes, lymph node and distant metastasis.Risk stratification was done according to Miettinen's classification (8).Extraintestinal GIST (EGIST) was defined by a combined approach of radiological, operative, gross and microscopic examination where the bulk of the tumour was outside the gastrointestinal tract and did not show a clear cut connection with the bowel or stomach wall.Immunohistochemistry was performed for CD117 (DAKO, Denmark), DOG1 (Novocastra, United States), CD34 (DAKO, Denmark), SMA (DAKO, Denmark), S100 (DAKO, Denmark), desmin (DAKO, Denmark) and vimentin (DAKO, Denmark).

Molecular Analysis
DNA extraction was done from formalin-fixed paraffinembedded tissue using the QiaAmp FFPE kit.Polymerase chain reaction (PCR) was performed for KIT (exons 11, 9, 13, 17) and PDGFRA (exons 18,14,12).Briefly, each PCR reaction was done in 25 μl volume using 250 ng of DNA on the ABI SimplyAmp TM thermal cycler followed by Sanger sequencing on the ABI 3130 Genetic Analyzer.

Statistical Analysis
Categorical variables were correlated using the Chi square test and Fisher's exact test where appropriate.Recurrencefree survival was analysed using Kaplan-Meier log-rank analysis.SPSS version 16 was used.A p value of < 0.05 was considered significant.The study was approved by the institute's ethics committee.
The extent of tumour infiltration was limited to the submucosa in one case and muscularis propria in 9 cases.Exophytic GISTs (n=15) involving the serosa alone were seen in 5 and serosa to muscularis propria in 10 cases.Twenty-nine cases (36.2%) had entire thickness involvement with ulceration of overlying mucosa.
Overall CD117 and DOG-1 immunoreactivity were 93.8% (n=75) and 91.3%, respectively, with high concordance between the two markers; excluding five cases (6.25%) with the CD117 (+)/DOG1(-); three (4%) cases with CD117(-)/DOG1(+) and two cases (2.5%) with CD117 (-)/DOG1(-) immune profile.The latter two cases were also KIT/PDGFRA wild type and the diagnosis was based on characteristic histomorphology, anatomical association with the gastrointestinal tract, presence of CD34 and absence of diffuse and strong expression of other immunohistochemical markers.The detailed gross, histological and immunohistochemical features are mentioned in Table I.

Mutational Analysis
Overall KIT and PDGFRA mutations were present in 47 cases (58.8%).
KIT exon 9 mutations were found in 9 cases (19%) which showed duplications Tyr502-503Asp in 4 cases and point mutations in 5 cases.These mutations were common in the small intestine (66.7%) followed by EGIST (22.2%) and stomach (11.1%).The spectrum of mutations in KIT exon 11 and 9 are summarized in Table II.
PDGFRA exon 18 mutations were present in 8 cases (17%), commonly located in the stomach (62.5%), and had epithelioid/mixed morphology.D842V mutation was seen in 1 case whereas D842E was seen in 5 cases (Figure 1B).The spectrum of PDGFRA exon 18 mutations are summarised in Table III.No mutations were found in PDGFRA exons 12 and 14.

Genotype and Clinicopathological Features
Males had higher KIT exon 11 and PDGFRA mutations than females (p=0.03).GISTs with PDGFRA exon 18 mutations had smaller tumour size than cases with KIT Vol.36, No. 2, 2020; Page 116-125 exon 9 and exon 11 mutations (p=0.01).Nearly 67% of cases with KIT exon 11 mutations were present in the stomach that included 12 simple deletions, 7 point mutations and one complex mutation (Table III).Overall 66.7% of KIT exon 9 mutations were found in the small intestine (p=0.04) of which codon 502-503 duplication was present in 75% of the small intestine (p=0.03).PDGFRA exon 18 mutations were mostly present in the stomach (62.5%) and all D842 point mutations were found in the gastric location as well (p=0.04).All colonic GISTs were wild type whereas all EGISTs were mutant (KIT exon 11 deletion in 2 cases, KIT exon 11 point mutation in one case, KIT exon 9 duplication/point mutation and PDGFRA exon 18 point mutation in one case each).The correlation of the GIST genotype with clinicopathological variables and risk stratification is shown in Table IV.
A diffuse sheet-like pattern was seen in 10 cases; 6 of which harbored KIT exon 11 deletions and 4 were wild type.All cases with substitutions and insertions had fascicles and/ or a palisaded pattern.Paranuclear vacuolization was commonly observed in cases with KIT exon 9 mutations (Figure 2D).It may be assumed that cases with both KIT exon11 deletions and wild type have an aggressive course with high cellularity and hence have a diffuse sheet-like pattern.

DISCUSSION
GISTs are mesenchymal tumours with malignant risk potential ranging from very low to high risk.The diagnosis of GIST relies on a panel of immunohistochemical markers whereas genotyping of GIST is essential for dosage and predicting the response of tyrosine kinase inhibitors.Since GISTs harbour various mutations having a wide range of frequency with their available targets, it is prudent to prioritize genetic mutation in this disease and use the resources judiciously, especially in developing countries.
In the present study, tumour size ranged from 1.5-30 cm with a mean of 10.3 cm, with larger tumours in large intestinal (mean: 17 cm) and extraintestinal location (mean: 12.9 cm) compared to the stomach and small intestine (mean-9.8cm).In the series by Miettinen et al., the median size of gastric GISTs was 6 cm (9) and several European series have reported a median size of 5-5.6 cm (2-10 cm) (10,11).On the contrary, median tumour size from other Indian series varies from 10.6 to 10.9 cm (12, 13), similar to our series which could be related to later admission to the hospital at an advanced symptomatic stage.
Mixed and epithelioid cell types are usually encountered in a gastric location (14).Interestingly, in the present study 50% of our EGISTs also showed a mixed cell type and this has not been observed in the EGIST of other series (9, 10).
High concordance (87.5%) was observed between CD117 and DOG1 immunoreactivity; however, four of five (80%) CD117 negative cases stained for DOG1.Foo et al. have reported one third of CD117 negative cases to be positive for DOG1 (15).We also observed similar findings where 5 of 8 PDGFRA mutated cases had epithelioid morphology and two cases had plasmacytoid cells and multinucleated cells.Unfortunately, SDH, KRAS and BRAF mutations could not be performed in KIT-PDGFRA wild type cases due to financial constraints, which is one of the major limitations of this study.

Suggested Algorithm for Molecular Testing in GIST Based on Morphological Features
Considering all histomorphological or genotypical features observed in the present study, we can argue that certain morphological features can be associated with a particular genotype.Therefore, we can suggest that the sequence of testing can be modified especially for patients/centres where financial affordability for multiple investigations is a concern (Figure 4).As most of the GIST mutations are mutually exclusive, the order of mutation testing can be devised for each set of morphological features.Tumours in the gastric location with spindle cell morphology and high mitosis may be tested for KIT exon 11 while gastric tumours with low mitosis, epithelioid morphology and low cellular atypia may be tested initially for PDGFRA exon 18 in the first phase, or if feasible, gastric GISTs should be first tested for KIT exon 11 followed by PDGFRA exon 18.If both are wild type then one may proceed with KIT exon 9 testing.About 20-40% of double negative GISTs are positive for SDH mutations and another 15% of triple negative (KIT, PDGFRA, SDH negative) cases harbour mutations in RAS/BRAF (25,26), which may be checked if the above mutations are negative.SDH immunohistochemistry can be done either following or in conjunction with KIT exon 11 and 9 mutation testing.Later, the BRAF, KIT exons 13, 17 and PDGFRA exon 12 mutations may be tested, which account for <5% of all mutations.Most SDH deficient GISTs have characteristic dumb bell/lobulated shape with thick fibrous bands, epithelioid morphology and frequent lymph node metastasis; SDH testing may precede KIT exon 11 mutation testing for suspected cases.It should be noted that this suggested algorithm is suited mostly for resource limited situations, and does not disagree with studying multiple mutations if laboratory resources and the patient's status permits.

C
Median recurrence free survival in all mutant and wild GIST Median recurrence free survival in KIT exon 11 codon 557-558 deletion and all other missense mutations Median recurrence free survival in KIT exon 11 codon 557

Table I :
Gross, histological and immunohistochemical features of GIST at different locations.

Table II :
Spectrum of KIT mutations with location and immunohistochemistry.

Table III :
Spectrum of PDGFRA exon 18 mutations with locations and immunohistochemistry.

Table IV :
Correlation of histopathological variables with mutation.
*Pearson Chi square and Fisher Exact tests were used wherever applicable