The role of PDGFRA as a therapeutic target in young colorectal cancer patients

This study was approved by the Institutional Review Board (IRB) of the Samsung Medical Center (IRB Approval No. SMC 2013-11-007-001). Written informed consent was obtained from each patient. The study subjects were 49 patients diagnosed with CRC at the Samsung Medical Center, Seoul, Korea. Tumor and matched adjacent normal tissues were obtained from surgical specimens. All patients were not treated before surgery.

The gene expression profile of 594 mRNASeq colorectal adenocarcinoma (COAD) samples from cBioPortal (https://​www.​cbioportal.​org/​, TCGA, PanCancer Atlas) was utilized for validation.

Genomic DNA and RNA in tissues were purified using the AllPrep DNA/RNA Mini Kit (Qiagen, Valencia, CA, USA). Genomic DNA from peripheral blood was extracted using the QIAamp DNA Blood Mini Kit (Qiagen). The genomic DNA concentration and purity were measured using a NanoDrop 8000 UV–Vis Spectrometer (Thermo Scientific Inc., Wilmington, DE, USA) and a Qubit 2.0 Fluorometer (Life Technologies Inc., Grand Island, NY, USA). DNA degradation was estimated by measuring median DNA size and ΔCt values with a 2200 TapeStation Instrument (Agilent Technologies, Santa Clara, CA, USA) and real-time PCR (Agilent Technologies), respectively. For RNA, the concentration and purity were measured with the NanoDrop and Bioanalyzer (Agilent Technologies).

Genomic DNA (1 µg) from each sample was sheared using the Covaris S220 (Covaris, Woburn, MA, USA), and a library was constructed using SureSelect XT Human All Exon V5 and a SureSelect XT Reagent Kit, HSQ (Agilent Technologies). The kit captures 335,756 exons of 21,058 genes, covering ~ 71 Mb of the human genome. After enriched exome libraries were multiplexed, the libraries were sequenced on a HiSeq 2500 platform (Illumina, San Diego, CA, USA). Briefly, a paired-end DNA sequencing library was prepared by gDNA shearing, end-repair, A-tailing, paired-end adaptor ligation, and amplification. After hybridization of the library with bait sequences for 16 h, the captured library was purified and amplified with an index barcode tag, and the library quality and quantity were measured. Sequencing was performed in the 100-bp paired-end mode of the TruSeq Rapid PE Cluster Kit and TruSeq Rapid SBS Kit (Illumina).

Raw sequencing reads were aligned to the UCSC hg19 reference genome (downloaded from http://​genome.​ucsc.​edu), using the Burrows–Wheeler Aligner (BWA) [24] version 0.7.5a with default settings. PCR duplicates were marked using Picard-tools-1.93 (http://​picard.​sourceforge.​net/​), data cleanup was performed with GATK, and variants were identified with GATK-3.5 [25]. Point mutations were identified using MuTect [26] with paired samples. ANNOVAR [27] was used to annotate variants. Somatic copy number alterations were identified using the Excavator tool [28]. Regions of the genome that were significantly amplified or deleted across samples were detected by the GISTIC tool [29].

Library construction for RNA sequencing was carried out with a Truseq RNA Sample Preparation v2 Kit (Illumina). Isolated total RNA was used in a reverse transcription reaction with poly (dT) primers, using SuperScriptTM II Reverse Transcriptase (Invitrogen/Life Technologies). Briefly, an RNA sequencing library was prepared by cDNA amplification, end-repair, 3′ end adenylation, adapter ligation, and amplification. Library quality and quantity were measured using the Bioanalyzer and Qubit. Sequencing of the RNA library was performed on the 100-bp paired-end mode of the TruSeq Rapid PE Cluster Kit and the TruSeq Rapid SBS Kit (Illumina).

Reads from the FASTQ files were mapped to the hg19 human reference genome using STAR aligner version 2.5.0a [30] and read counts for each gene were normalized as transcripts per million (TPM) using the RSEM program [31]. Differentially expressed genes (DEGs) were identified using the DESeq R package [32] with a cut-off (|log2 fold-change|> 2 and false discovery rate (FDR) < 0.001) [33]. DEGs were mapped to the pathway using the ReactomeFI Cytoscape [34]. CMS calls in CRC cohorts were predicted based on the gene expression profile using the R package CMSclassifier [19].

The organoid culture medium was refreshed every 2 days. For the organoid passage, BME was broken up by pipetting, and organoids were collected in a tube. The organoids were centrifuged at 1000 rpm for 3 min, and the medium was removed. Next, 5 mL of Triple Express (Invitrogen) was added, and the organoids were incubated at 37 °C for approximately 5 min. Every minute, a visual check was performed to verify the size of organoids. Care was taken not to treat the organoids too long with the Triple Express. FCS and medium were added, and the cells were spun down at 1500 rpm for 3 min. The pellet was taken up in BME, and the cells were plated in droplets of 5–10 μL each. After allowing the BME to solidify, HICS minus Wnt, both supplemented with 10-µM LY27632, was added to the plates, and organoids were incubated at 37 °C [35].

For transient transfection assays, PDGFRA siRNAs (#1: 5′ CCUCUAUCCUUCCAAAUGAUU 3′, 5′ UCAUUUGGAAGGAUAGAGGUU 3’, #2: 5′ CCGUCAAAGGAAAGAAGUUUU 3′, 5′ AACUUCUUUCCUUUGACGGUU 3′, Genolution, Seoul, Korea) were transfected into PDCs using Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol.

Cell proliferation was measured in triplicate using a colorimetric assay that determines cellular viability by evaluating the metabolic conversion based on the quantification of the ATP present using CellTiter-Glo (Promega, Madison, WI, USA). The viability of the colon cancer cells was assessed at various times, and assays were performed by adding CellTiter-Glo directly to the culture wells and incubating for 30 min at room temperature. Luminescence was measured using a Mithras microplate reader (Berthold-bio, Bad Wildbad, Germany). Three different experiments were performed for each experimental condition.

To prepare the whole-cell extract, cells were lysed using a Pro-prep buffer (Intron Biotechnology, Seoul, Korea) containing protease inhibitors. A total of 10–40 μg of protein extract was resolved by SDS-PAGE and transferred to polyvinylidene fluoride (PVDF) membranes. The membranes were probed with primary antibodies against PDGFRA (ab65258, Abcam) and β-actin (#3700, Cell Signaling Technology) followed by incubation with secondary antibodies conjugated to horseradish peroxidase (Santa Cruz Biotechnology, CA, USA). β-actin was used as a loading control in the western blot analysis.

The chip layout was designed for drug screening in a single micropillar chip. In the micropillar chip, ~ 80–100 cells were immociliized with 0.75% alginate. We tested 67 drugs in CRC PDCs. Among them, we revealed regorafenib data in this study. A 50-nL droplet of a 1:1 cell mixture of 1.5% alginate and 950-nL droplet of 3D culture media was dispensed with the ASFA Spotter ST (MBD). After overnight incubation, a 950-nL droplet of the drugs was dispensed with the ASFA Spotter ST and stamped with the micropillar chip containing the cells. The combined chips were incubated for 5 days at 37 °C and 5% CO₂ in an incubator for the cell viability assay. After incubation, the micropillar chips were stained with a staining buffer (MBD-STA50, Medical and Bio Device) in a Calcein-AM (Invitrogen, live-cell starting dye) for 1 h in the dark at room temperature. The stained micropillar chips were washed with a staining buffer for 30 min and dried in the dark.

Statistical comparisons were performed using GraphPad Prism. Data are expressed as the mean ± standard deviation (SD). P-values < 0.05 were considered statistically significant (*P < 0.05, **P < 0.01, or ***P < 0.001).

To clarify the difference in the molecular characteristics between young and old patients, we divided the samples into three groups according to the age of 49 CRC patients (Table 1). The young (n = 13), middle-aged (n = 23), and old age (n = 13) groups included patients < 40 years old, between 41 and 60 years old, and > 60 years old, respectively. After categorizing the patient groups, CRC patients were classified according to CMS using gene expression profiles [19]. In particular, CMS4 (53.85%) was the most frequent in young CRC, followed by CMS3 (38.46%) and CMS2 (7.69%). CMS1 was absent in the young age group (Fig. 1a). In the middle CRC group, the order of the frequencies was the same as in the young age group, but the frequencies of CMS2 (21.74%) and CMS1 (13.04%) were higher (Fig. 1b). In the old CRC group, the frequency of CMS4 (23.08%) was the lowest, while CMS2 (38.46%) and CMS3 (38.46%) showed the same frequency (Fig. 1c). Thus, CMS4 frequency gradually decreased with age, while CMS2 was more frequent in the old age group (Fig. 1 a-c). Taken together, these results indicate that young CRC patients are closely associated with CMS4, which has been considered as the worst prognosis type [19]. On the other hand, young CRC patients are less related to CMS2, which has a better prognosis than the other CMSs [19].

To systematically identify specific gene expression patterns related to young CRC, we detected differentially expressed genes (DEGs) from RNA sequencing data of 49 CRC tissues and their paired adjacent normal tissues. We first identified age-related genes showing a significant difference between normal samples of the young and old age groups (Fig. 2a right panel, log2 fold change (FC) ≥  ± 1, FDR ≤ 0.05). In addition, DEGs in the tumor tissues were obtained by comparing both groups using tumor samples (Fig. 2a left panel, log2 FC ≥  ± 1, FDR ≤ 0.05). Therefore, 155 genes from the DEGs in normal samples and 204 genes in the tumor were obtained. Finally, by eliminating 155 age-related genes in the normal samples, we considered a total of 180 genes as age-related DEGs specific to cancer (Fig. 2b). Among them, 105 genes were significantly upregulated and the other 75 were downregulated (Fig. 2c, log2 fold change ≥  ± 1, p-value ≤ 0.05, FDR ≤ 0.05). To identify cancer-related genes to be considered biomarkers of young CRC patients, we extracted only oncogenes [36] and tumor suppressor genes[37]. A number of oncogenes and tumor suppressor genes were highly upregulated or downregulated in young CRC patients (eight and four upregulated genes, and six and five downregulated genes, respectively) (Tables 2 and 3). Next, we identified the significantly changed pathways (34 upregulated and 26 downregulated) from DEGs using the network-based pathway analysis tool [38] (Additional file 1: Fig. S1). Many CMS4-related pathways [19] were enriched in young CRC patients, including PDGFRA signaling, positive regulation of cell proliferation by VEGF-activated platelet-derived growth factor (PDGF) receptor signaling, cell chemotaxis, regulation of chemotaxis, positive regulation of phospholipase C (PLC) activity, and phosphatidylinositol phosphorylation (Fig. 2d). Notably, these pathways are closely related to metastasis, angiogenesis, proliferation, and chemoresistance, which are related to CMS4 [19] and are associated with PDGFRA.

We also identified numerous somatic mutations, including missense, nonsense, and splicing mutations in primary colon tumors of CRC patients (Fig. 2e). The mutation profile showed that a TP53 mutation was more frequent (p = 0.0029) in young CRC (76.9%, 10/13) than in old CRC (23%, 3/13) (Fig. 2e). A comparison of the copy number amplified and deleted genomic regions between both groups (Fig. 2f) showed that the copy number at chromosome 5q22.2 was significantly deleted only in young CRC samples (q-value < 0.01).

To investigate whether PDGFRA is correlated with the age of CRC patients, its correlation pattern with age was examined. PDGFRA expression was negatively correlated with the age of CRC patients in our cohort (Fig. 3a, Spearman’s correlation coefficient; P = 0.0001, R = − 0.5264). Additionally, TCGA data consistently showed an inverse correlation between its expression and the age of CRC patients (Additional file 2: Fig. S2a, P = 0.0113, R = − 0.1697). Figure 3b shows that the expression of PDGFRA in the young CRC group was significantly higher than that in the old CRC patient group (P = 0.0001). The TCGA data revealed a similar pattern (Additional file 2: Fig. S2b, P = 0.2796), although this data had a limitation due to the small number of young CRC patients. To confirm whether PDGFRA was a young CRC factor rather than a CMS4 factor, its expression was compared between both groups within samples of the CMS4 type. The expression of PDGFRA in the young CRC group was higher than that in the old CRC group, even when considering only the CMS4 type (Fig. 3c, P = 0.0708). There was no significant difference due to the small sample size. To confirm our results, we also used a publicly available gene expression data set (GSE14333) derived from the primary CRC tissue (n = 188). To set similar conditions as our data, we divided it into two groups according to percentile (young CRC group: less than approximately 25th percentile, < 59 years old (n = 49) and old CRC group: greater than approximately 75th percentile, > 76 years old (n = 49)). With the CMS classification according to age, CMS4 significantly increased in the younger group (Fig. 3d, upper panel). The expression of PDGFRA in the younger group was significantly higher than that in the older group (P < 0.001) (Fig. 3d, lower panel).

CMS4 is a mesenchymal subtype of CRC and is activated in stromal invasion and angiogenesis [19]. We examined whether PDGFRA was closely associated with epithelial-mesenchymal transition (EMT) markers [39] and angiogenesis markers [40]. Among the EMT markers, six markers, namely Vim, ZEB1, ZEB2, Twist1, Twist2, and MMP2, showed positive co-expression with PDGFRA in our data set (Fig. 4a). Their expression levels were also different than the two groups, showing separate distributions. As with PDGFRA, the samples identified as CMS4 had a high expression. Among the angiogenesis markers, five markers, including FLT1, FLT4, ICAM1, TIE1, and KDR, showed a positive correlation in their gene expression (Fig. 4b). These markers also showed the same pattern as EMT markers. In the TCGA data, these EMT and angiogenesis markers also showed the same results as our sequencing data (Additional file 3: Fig. S3a, b). Each marker positively correlated with PDGFRA and showed a negative correlation with age (Additional file 3: Fig. S3c, d) and a co-expression pattern among each other (Additional file 3: Fig. S3e).

To investigate the key roles of PDGFRA in young CRC, we used PDCs derived from young and old age CRC patients that were not treated before surgery. (Table 4). The RT-qPCR analysis showed that PDGFRA expression was higher in young PDCs than in old PDCs (Fig. 5a, Additional file 4: Fig. S4a). Western blot analysis revealed that the expression of PDGFRA was higher in young PDCs than in old PDCs (Fig. 5b). To evaluate the possible roles of PDGFRA in young CRC, we transfected siRNAs in PDCs with two different siRNAs against PDGFRA to compare their proliferation ability using a Cell Titer GLO proliferation assay. Knockdown of PDGFRA in PDCs from young patients significantly reduced viability (Fig. 5c). However, PDCs from old patients did not alter viability compared to the control (Fig. 5d). To determine the PDGFRA-targeted drug for young CRC, we analyzed the correlation between drug sensitivity and gene expression [41]. The Cancer Cell Line Encyclopedia (CCLE) data indicate that sorafenib targeting PDGFRA only had a potential drug response in colorectal cancer (Additional file 4: Fig. S4b). However, our results showed that sorafenib did not show a distinguished drug response according to PDGFRA expression or age in PDCs (Additional file 4: Fig. S4c). Regorafenib is a small molecule multi-kinase inhibitor, including PDGFRA, and was recently approved by the FDA to treat patients with metastatic CRC [42‐44]. To investigate the treatment strategy for young CRC, we used regorafenib in PDCs. PDCs from young patients were sensitive to regorafenib treatment, whereas PDCs from old patients were relatively resistant to regorafenib (Fig. 5e). When conducting HTS, the young age group was set to younger than 50 years old due to the small number of samples, and the results showed that the IC50 of regorafenib was significantly lower in the young group than in the old group (Additional file 4: Fig. S4d).