MicroRNA expression signature in gastrointestinal stromal tumour & their molecular & histological features

Sample processing

Formalin-fixed and paraffin-embedded (FFPE) blocks of the initial 58 cases were obtained from the department of Histopathology, PGIMER; however, complete repeated immunocytochemical analysis and miRNA profiling were possible in 45 cases; 13 cases could not pass the quality metrics for miRNA amplification, so they were excluded from the study.

Clinically and histologically diagnosed GIST affected individuals over eight-year period (2012 to 2020) were included for miRNA expression analyses. The cases were from department of surgery, PGIMER, who had presented with abdominal mass. Fifty GIST samples including 45 FFPE and five fresh tissues were included. Peripheral non-tumour tissues were used as controls. Among these, FFPE specimens also were previously used for molecular analysis in KIT (exons 9, 11, 13 and 17) and PDGFRA (exons 12, 14 and 18) for diagnosis. However, the mutational data were available in 42 GIST cases (Supplementary Table). The pathological information of each case was reviewed including cell morphology, primary tumour site, tumour size, mitotic index and risk grade.

Diagnosis and screening

After establishing GIST diagnosis from the surgically resected specimens, frozen samples were subjected to study the composition of the tissue. All surgically resected specimens were processed for routine diagnosis purposes; hematoxylin and eosin staining and immunohistochemistry. Light microscopic review and immunohistochemical analysis for CD117 [Clone, Rabbit polyclonal anti-CD117 antibodies (CD 117, Dako, Denmark) at dilution 1:200] were performed and reviewed independently without knowledge of previous clinical information. One of the limitations of this study was that we did not test PDGFRA and DOG-1 immunostaining in any GIST case. We based our diagnosis on CD117 staining.

RNA extraction from FFPE and fresh tissues

Total RNA was isolated from FFPE and fresh tissue using miRNA-specific kit (Qiagen, Germany) Adjacent control tissues, displaying no macroscopic abnormalities, were meticulously chosen for comparison with the tumour samples. Surgical margins, ranging from 5 to 7 cm away from the primary tumour sites, were excised and meticulously evaluated through histopathological examination to ensure the absence of bias in the control samples. The frozen samples were ground to fine particles with the help of a metal spatula, and the powdered tissue was scraped and mixed with Lysis Buffer. The FFPE blocks having sufficient tumours were selected for RNA isolation. Using a standard microtome, 2-3 sections of FFPE tissue with a thickness of 5 µm were cut and immediately deparaffinized using 300 μl of deparaffinization solution (Qiagen, Germany). The tissue was digested in 150 μl of PKD buffer and10μl Proteinase K (Qiagen) and incubated at 56°C for overnight digestion followed by 80°C incubation for 15 min. Manufacturer’s protocol was followed for miRNA extraction. The RNA quantity and quality were checked by measuring absorbance in a nanodrop (NanoDrop 2000, ThermoScientific, USA) at 260 nm (1 O.D. = 40 μg/ml); A260/A280 ratio of 2 was considered purified RNA and A260/A230 ratio >1.7 considered as highly purified RNA.

cDNA synthesis

Following the protocol outlined by the manufacturer of the miScript II RT cDNA synthesis kit (Qiagen, Germany), 100 ng of RNA was subjected to reverse transcription in a total volume of 20 μl. The concentrated RNA was diluted in RNase-free water to make the final concentration of 1μg of RNA used for cDNA synthesis. Master mix for reverse transcription was prepared on ice in accordance with the stated instructions. The final incubation at 37°C for 60 min was followed by the termination of the reaction by incubating for 5 min at 95°C. The products of the reverse transcription reaction were either utilized immediately for qPCR or kept at -20°C for further use.

miRNA amplification using real-time qPCR

The expression levels of microRNAs were assayed by Quantitative RT-PCR analysis performed on AriaMX Real-time PCR system (Agilent, US). About 50 ng of cDNA was used in each sample with the SYBRGreen (Agilent, US). The qPCR master mix was prepared in a final volume of 10 μl using 5 μl of SYBRGreen, 0.5 μl of forward and reverse primer, 50 ng/μl of cDNA and nuclease-free water in 96 well plates on ice. The plate was covered with sealing film and centrifuged at 1000 g/1 min at 4°C to eliminate the bubble that could interfere with PCR reactions. All samples were evaluated in triplicates using the following cycling conditions: 40 cycles of initial denaturation for 10 sec at 95°C, 15 sec of annealing at 58°C, and extension at 72°C for 20 seconds were performed after the initial denaturation for 5 min at 95°C. The melting curve conditions were 95°C for 10 sec, 5°C (+annealing temp) for 1 min and 95°C for 20 sec. The details of miRNA primer sequences are depicted in Supplementary Table.

Analysis of real-time qPCR data

Following the generation of a melting curve for each test in each run to determine the precise amplification efficiency, miRNAs expressions in non-tumour and tumour tissue samples were statistically evaluated and adjusted against RNU6 levels for all samples. Positive PCR control and negative control were also used. AriaMx software version 1.5 (Agilent, US) was used to evaluate the data obtained as a threshold cycle (CT value). The baseline was automatically assigned for all wells in the software to calculate threshold values above background noise. A negative call was deemed to be a CT value of less than 35. Each miRNAs expression profiling data was analysedusing a relative quantification method, delta-delta Ct method (ΔΔCT)¹³. The miRNA of interest (MOI) and the U6 RNA were amplified in each sample to determine the Ct value for each.

Statistical Analysis

miRNAs expression levels were calculated and compared with clinicopathological characteristics i.e., immunostaining, tumour site, and cell morphology as a factor using the Kruskal-Wallis rank sum test and Wilcoxon rank sum exact test. The P≤0.05 was considered as a significant difference and the pair-wise Wilcoxon rank sum exact test was applied if the Kruskal-Wallis rank sum test was found significant. The box plot was chosen to create a distributional (i.e., variance, interquartile range and outliers) overview for each miRNA expression level. A post-hoc test of Tukey’s was applied as a correction to calculate the final P value. The analysis was performed on the R Studio base package and plotted using the R package “ggpubr”.

Sample characteristics

Deregulation of miRNA expression is considered to be one of the most significant factors contributing to carcinogenesis. We tried to focus on the previously described four miRNAs including miR-222 and miR-494, miR-221 and miR-34a in 50 GISTs specimens. Four potential miRNAs were selected because of their statistically significant correlation with KIT expression. MiR-222 and miR-221, miR-494 have been suggested to be targeting the KIT gene while miR-34a targets PDGFRA. We investigated the miRNAs’ expression in GIST and contrasted the variations with adjacent peripheral non-tumour tissue, clinicopathological characteristics and molecular alterations.

The study participant’s clinicopathological characteristics are depicted in Table I. All the study participants were diagnosed based on c-KIT immunostaining. Among these, 45 cases showed positivity for CD117, whereas five cases were negative. Of the 50 cases, eight study participants developed metastasis, most commonly in liver (n=6) and one case each in lymph node and omentum, respectively. Seven study participants presented with recurrent tumours. In c-KIT negative cases, one patient with jejunal involvement exhibited regional lymph node metastasis.

Based on the morphological analyses, 39 cases showed spindle cell morphology followed by epithelioid (n=8) and the remaining three were mixed type. The overall tumour size ranged from 3 to 19 cm. The large tumour size (>10 cm) were found in the majority of the cases (n=16), while intermediate tumour size (>5-<10 cm) was reported in 12 cases. However, small size tumours (<5 cm, n=5) were fewer. The presentation of GISTs varies depending on the primary tumour location with differing prognostic and predictive relevance. Most primary GISTs had gastric localization (n=30) followed by the small bowel (n=15) and the remaining were found in the Colon (n=2), rectum (n=1), omentum (n=1) and mesentery (n=1).

Mitotic count is among the most crucial aspects of pathological examination for determining the likelihood of tumour risk. The mitotic rate ranged between 2-5/50HPF and was reported in the majority of the cases (n=16) followed by 5-10/50HPF in six cases, and the higher mitotic rate (>10/50HPF) was found only in one case. The patients were classified according to the NIH standard criteria for determining the probability of aggressive clinical progression of primary GISTs. The findings of our study revealed that a total of 16 cases came under the high risk category, nine cases as intermediate, and six cases were categorized under the low-risk category. Molecular data were available in 41 cases in which 25 cases had KIT exon 11 mutations followed by exon 9 (n=5), Exon 13 (n=1) and 17 (n=4) while PDGFRA exon 12 and 18 mutations were reported in one and four cases, respectively.

MiRNA expression in GIST tumour and normal tissue

In our study, we performed real-time PCR (RT-qPCR) analyses to assess the expression levels of miR-221, miR-222, miR-494 and miR-34a between tumours and peripheral non-tumour tissue. We observed that in the majority of the cases, there was a significantly reduced expression in miR-221 (P=0.009), miR-222 (P=0.012), miR-494 (P=0.015), miR-34a (P=0.024), in tumour tissues relative to non-tumour (P=0.48) (Fig. 1).

Close

Fig. 1.

miRNA expression level in tumour and non-tumour tissue. miR-221 and miR-222, miR-494 and miR-34a showing considerable difference in expression level between GIST tumour and pheripheral non-tumour tissue (median expression level, 0.015).

Export to PPT

MicroRNAs expression and its correlation with histological features

miRNA expression levels in KIT-negative and KIT-positive tumours exhibited a substantial difference (Fig. 2A). We found a significant correlation (P=0.00000008) between c-KIT and miR-221, miR-494, miR-222, and miR-34a expression pattern, especially between KIT positive and negative tumours. The miR-221, miR-494, miR-34a and miR-222 expression levels, were significantly down-regulated in KIT-positive tumours whereas, KIT-negative tumours displayed the exact opposite expression pattern. However, there were only five KIT-negative cases in our study. Based on our study, this inverse relationship demonstrated that these miRNAs directly or indirectly regulated KIT expression and might be the modulators of KIT in GISTs. Further, we explored the clinical importance of miRNA expressions in histological subtypes of GIST. MiR-221, miR-494, miR-222 and miR-34a were found to be down-regulated in the majority of the spindle subtype; however, up-regulation was reported in a few cases. These miRNAs did not show a significant correlation with epithelioid and mixed sub-types; these cases were very few.

Close

Fig. 2.

miRNA expression level and its association with histological features. (A) KIT immunostaining: miRNAs expression levels in KIT-negative and KIT-positive tumours exhibited a substantial difference (P value: 0.00000008). miRNAs were significantly down-regulated in KIT-positive tumours whereas, KIT-negative tumours displayed the exact opposite expression pattern as KIT-positive tumours. (B) Cellular morphology: miRNAs were found down-regulated in the majority of the spindle subtype (P value=0.16). However, these did not showed any significant correlation with epithelioid and mixed sub-types. (C) Primary tumour sites: majority of the gastric and small intestinal GIST showed increased expression level (P value=0.19) however; there were no significant correlation with other primary tumour.

Export to PPT

Moreover, all these miRNAs showed up-regulation in a few cases, irrespective of various histological parameters. Based on the overall statistical evaluation, the Kruskal-Wallis rank sum test did not find a significant association between miRNA and cellular morphology (P=0.19; Fig. 2B).

Further, we attempted to identify the correlation between miRNAs and the site of tumour occurrence. The miRNA-data were analyzedamong gastric, small intestinal, colon, rectum, omentum and mesentery. MiR-221, miR-494, miR-34a and miR-222 were found down-regulated in the majority of the gastric and small bowel GIST; however, there was no positive correlation in the colon, rectum, omentum and mesentery. Based on the Kruskal-Wallis rank sum test, overall expression showed a non-significant association with tumour sites (P=0.16; Fig. 2C).

MiRNA and its correlation with molecular subtypes

To categorize different subtypes of GISTs based on miRNA expression profiles, molecular analysis was carried out using relevant gene sets and exonal mutation with miRNA expression. The miRNA-data correlated with KIT exon 9, 11 and 17 and PDGFRA exon 18. We found a significant correlation between KIT exons 11 and 17 mutant subtypes and miRNA expression. Further analysis revealed that miR-221 and miR-222, miR-494 expression level were greatly reduced in KIT Exon 11 and 17 mutated GIST, most commonly reported in Gastric GIST(n=15) followed by small intestine (n=6).Similarly, miRNA-34a was found down-regulated in PDGFRA exon 18 mutant subtypes most frequently in small bowel GIST (n=3).However, there was no significant correlation found in KIT exon 9 and PDGFRA exon 18 and miR-222, miR-221, miR-494 expression in the small intestine and Gastric GIST. MiRNAs and its correlations with mutational subtypes and primary tumour sites are depicted in Table II.