Overexpression of the cohesin-core subunit SMC1A contributes to colorectal cancer development

OncoScan analysis

Normal mucosa, early adenoma, and carcinoma from 16 subjects affected by CRC (Additional file 1: Table S1) were processed for identification of copy number imbalances through Molecular Inversion Probe (MIP) based OncoScan Array following protocols provided by the manufacturer (Affymetrix). DNA samples were normalized to 12 ng/μL, mixed with MIPs and incubated overnight to anneal (16–18 h). Next, each reaction was divided equally into A and B reactions and “Gap Fill” master mix was added with either AT dNTPs (A reaction) or GC dNTPs (B reaction) and incubated. Following the “Gap Fill” reaction, exonuclease was added to remove unligated probes and genomic DNA. Next, MIPs were linearized with a restriction enzyme and PCR amplified (PCR 1). Reactions were taken through a second round of amplification (PCR 2), and subsequently digested with HaeIII restriction enzyme. The digested products were hybridized to the OncoScan Array at 58 °C for 16–18 h. Arrays were stained and washed using the GeneChip Fluidics Station 450 and loaded on the GeneChip Scanner 3000 7G (Affymetrix) where fluorescence intensity was scanned to generate array images (DAT files).

Next, array fluorescence intensity data (CEL) files were generated with an Affymetrix GeneChip Command Console (AGCC, Affymetrix) and used to produce OSCHP files and QC metrics through OncoScan Console Software (Biodiscovery). The standard Affymetrix reference control file for OncoScan data was used for processing the arrays. Samples passing QC criteria (MAPD ≤0.3, ndSNPQC ≥26) were further analyzed using tumor Scan (TuScan) and BioDiscovery’s SNPFASST2 algorithm using the Nexus Express for OncoScan software version 3.0 and 7.5 (Biodiscovery). The TuScan algorithm creates segmentation to differentiate between adjacent clusters of probes and determines the copy number changes. SNP-FASST2 uses a Hidden Markov Model (HMM)-based approach to identify larger copy number segments based on a log-ratio threshold derived from all probes in a given region. Ratios are the log2 ratios of the normalized intensity of the sample over the normalized intensity of a reference with further correction for a sample-specific variation. The Median Log2 Ratio was computed for each segment detected in the analysis. The significance threshold for segmentation was set at 1.0e − 5 also requiring a minimum of three probes per segment and a maximum probe spacing of 1000 kbp between adjacent probes before breaking a segment. Segments were classified as having gains when the Log2Ratio (L2R) exceeded 0.2, losses when L2R was < − 0.2, and with high copy gains and homozygous losses being called when L2R was > 0.6 and < − 1.0, respectively. Differences in copy number changes between samples from a tumor for individual genes were counted when one sample had a change as defined above and another either did not have that change or had a different change. B-allele frequency (BAF) information was also generated for each tumor. BAF is a normalized measure of the allelic intensity ratio of two alleles (A and B), such that a BAF of 1 or 0 indicates the complete absence of one of the two alleles (e.g. AA or BB), and a BAF of 0.5 indicates the equal presence of both alleles (e.g. AB). Median BAF is reported for each segment and is the median BAF of the markers identified as heterozygous. If the number of heterozygous markers in the segment is below 10 or the percent of homozygous markers is above 85% no value is reported. The B-allele frequency values are used to determine whether a segment is in a loss of heterozygosity (LOH) or an allelic imbalance state. By default, probe sets were automatically centered to the median for all samples by the Nexus software. For individual samples where the median probe set value was not diploid, specified regions of balanced heterozygosity were manually identified by visual inspection of L2R and BAF plots and defined as diploid regions, permitting the Nexus software to reset the entire probe set to the newly defined areas [40]. Finally, we carefully analyzed the CNVs involving the X chromosome. In detail, we used a different threshold for the CNVs of the X chromosome in males and females (Copy number = 0, for a deletion identified in a male; Copy number = 2 or greater than 2, for duplications identified in a male subject. Copy number = 1 or 0 for a deletion in a female; Copy number = 3 or greater than 3 for a duplication in a female subject).

Clinical samples and histopathology

Sixty-six subjects affected by CRC (Additional file 1: Table S2) were retrospectively selected from the files of the Unit of Surgical Pathology of the Azienda Ospedaliero-Universitaria Pisana. For each patient, specimens (normal mucosa, early adenomas and carcinomas) were surgically obtained and fixed in 10% neutral-buffered formaldehyde and embedded in paraffin. Routine hematoxylin and eosin staining were performed on microtomic sections for histopathological examination. Histological diagnoses, reviewed independently by experienced pathologists (A.S., G.F.), were formulated according to the 2010 World Health Organization (WHO) Classification (4th Edition). Approval was granted by Azienda Ospedaliero-Universitaria Pisana Ethics Committee (protocol number 3867).

Mutation analysis for SMC1A

DNA was extracted from embedded paraffin samples by the NucleoSpin Tissue kit (Macherey-Nagel) according to the manufacturer’s protocol. Primer pairs (Additional file 2: Table S3) were designed to amplify exons, exon/intron boundaries and short flanking intronic sequences. Amplified PCR products were purified and sequenced.

Immunohistochemistry

SMC1A immunohistochemical analysis was performed on 3-μm tissue sections. After deparaffinization via serial xylene baths, samples were rehydrated via a graded ethanol series. Then the sections were heated to 98 °C for 40 min to unmask target antigens, cooled in solution at room temperature and washed with phosphate-buffered saline (PBS) for 3 min. After treatment with peroxide block, sections were washed again in PBS and incubated with power block reagent (Biogenex Laboratories) for 30 min. Primary mouse monoclonal antibody against human SMC1A (diluted 1:400, Bethyl Laboratories) was applied overnight at 4 °C. After washing with PBS, immunoreactivity was obtained by using the “Super Sensitive Polymer-HRP Detection System” (Biogenex), following the manufacturer’s instructions. Thereafter, samples were incubated for 20 min at room temperature with Super Enhancer Reagent, followed by incubation with Poly-HRP reagent. The reaction was developed using 0.05% 3,3′-diaminobenzidine tetrahydrochloride until adequate color development was seen. A negative control was obtained by substituting primary antibody with PBS. The staining percentage was graded as 0 (0–5%), 1 (6–20%), 2 (21–60%) and 3 (61–100%), and the staining intensity was graded as 0 (negative), 1 (+, weak), 2 (++, moderate) and 3 (+++, strong), as previously described [41, 42].

SMC1A cDNA mutagenesis and cell transfection

Site-directed mutagenesis of the SMC1A cDNA (OriGene) was performed with QuikChange Site-Directed Mutagenesis Kit (Stratagene) according to the manufacturer’s instructions. Coding sequences for SMC1A wild type or mutated were inserted into the Not I site of the pcDNA3.1 plasmid (Invitrogen, ThermoFisher Scientific). The presence and orientation of the insert was confirmed by enzyme restriction digestion.

Human colorectal carcinoma HCT116 cells were maintained in culture according to American Type Culture Collection specifications. Briefly, cells were grown in McCoy’s Medium with 10% fetal calf serum and antibiotics in a humidified 5% CO₂ atmosphere. HCT116 cells were transfected with the pcDNA3.1 vectors expressing SMC1A wild-type or SMC1A mutant. Stable expressing clones were obtained by selection in G418 (800 μg/ml) for 2 weeks.

Animal care and experiments

Twenty-one immunocompromised NOD-SCID female mice 5–6 weeks of age were purchased from the Animal Care Facility of the IRCCS - Ospedale Policlinico San Martino of Genova, where animals were then housed and maintained in pathogen-free conditions. All experiments were reviewed and approved by the internal Review Board (OPBA) and authorized by the Italian Ministry of Health, accordingly with the current national and European regulations and guidelines for the care and use of laboratory animals (D.L. 26/2014; 86/609/EEC Directive).

Mice were bilaterally injected subcutaneously with 2 × 10⁶ HCT116 cells, HCT116 overexpressing SMC1A wild-type or HCT116 harboring SMC1A c.A2027G mutation, suspended in 0.1 ml of PBS. The mice were then regularly palpated to assess tumor latency. Tumor growth was recorded measuring nodule size with calipers three times per week. When the first nodule reached the volume of 300 mm³, all animals were sacrificed and tumors were excised, measured, photographed, weighed and formalin-fixed and snap-frozen in liquid nitrogen for further analyses.

For histology analysis, half of the samples were fixed in 4% neutral-buffered formalin, embedded in paraffin and cut to obtain 3- to 4-mm-thick sections. Slides were then stained with hematoxylin and eosin for histopathological examination; the other halves were snap-frozen in liquid nitrogen and kept at − 80 °C for molecular investigation.

Library preparation and RNA-sequencing (RNA-seq)

Three tumors deriving from the inoculation of HCT116 (907_1, 907_2 and 907_3), four deriving from HCT116 overexpressing SMC1A wild-type (907_4, 907_5, 907_6 and 907_7) and four deriving from HCT116 harboring SMC1A c.A2027G mutation (907_8, 907_9, 907_10 and 907_11) were separately processed for RNA-seq analyses.

Library preparation was obtained using the TruSeq Stranded mRNA Sample Prep kit (Illumina). The poly-A mRNA was fragmented for 3 min at 94 °C and every purification step was performed using 1X Agencourt AMPure XP beads. Both RNA samples and final libraries were quantified using the Qubit 2.0 Fluorometer (Invitrogen) and the quality was tested using the Agilent 2100 Bioanalyzer RNA Nano assay (Agilent). Libraries were then processed with Illumina cBot for cluster generation on the flowcell, following the manufacturer’s instructions and sequenced on single-end mode on HiSeq 2500 (Illumina). The CASAVA 1.8.2 version of the Illumina pipeline was used to process raw data for format conversion and de-multiplexing.

RNA-Seq analysis

To avoid low-quality data, adapters were removed by Cutadapt 1 and lower quality bases were trimmed by ERNE2. For the analysis of differentially expressed genes, the quality-checked reads were processed using the TopHat version 2.0.0 package (Bowtie 2 version 2.2.0) as FASTQ files [43–45]. Reads were mapped to the human reference genome GRCh37/hg19. Read abundance was evaluated and normalized by using Cufflinks 3 for each gene and Cuffdiff from the Cufflinks 2.2.0 package was used to calculate the differential expression levels and evaluate the statistical significance of detected alterations. Numbers of reads per sample are shown in Additional file 3: Table S4. Only protein-coding genes were considered and gene level expression values were determined by fragments per kilobase million (FPKM) mapped. All genes with FPKM > 1 were designated as expressed and analyzed with an established p-value < 0.05.

cDNA synthesis and quantitative real-time PCR (qRT-PCR)

Total RNA was extracted by RNAeasy Mini-kit (Qiagen) and cDNA was synthesized with SuperScript™ II reverse transcriptase using oligo-dT (Invitrogen). PCR analyses were performed using Rotor Gene 3000 (Corbett). qPCR reactions were run in duplicate and normalized with respect to HPRT. Primers used for mRNA expression analysis are listed in Additional file 4: Table S5.

Pathway analysis and function

The differentially expressed genes were functionally analyzed for biological processes using Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.8 (https://david.ncifcrf.gov). For each term, the p-value was calculated and a term with p < 0.05 was considered to be enriched.

Statistical analysis

LOH, CNVs and genome changed data were analyzed by Kruskal-Wallis test, a non-parametric method for comparing two or more independent samples of equal or different sample sizes. Immunohistochemistry data were analyzed by the Wilcoxon signed-rank test, a non-parametric method used for comparing related samples or matched samples. All other data were analyzed by Student’s t-test or the chi-squared test. P-values of < 0.05 were considered statistically significant.

Accession number

The data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO Series accession number GSE113678.

OncoScan analysis during CRC development highlights the presence of extra-copies of cohesin genes

In order to identify copy number variations (CNVs) and loss of heterozygosity (LOH), we performed molecular cytogenetic analysis on normal mucosa, early adenoma, and carcinoma from 16 subjects affected by CRC. Genomic DNA was interrogated using OncoScan FFPE Array. We found an increase in LOH at the carcinoma stage (p = 0.0015, Kruskal-Wallis test, Table 1, Additional file 5: Figure S1a). In addition, most of the subjects (13/16) showed a marked increase in chromosome gains or losses proceeding from normal mucosa until the carcinoma stage (p = 0.0006, Fig. 1a-c, Additional file 5: Figures S1b–S4,). Common events in carcinoma were the gain of whole chromosomes 7, 13, and X; gain of the long arm of chromosomes 8 and 20; loss of chromosome 18 and loss of the short arm of chromosomes 8 and 17 (Fig. 1c, Additional file 5: Figure S4). Of note, few events were identified in both normal mucosa and early adenomas (Fig. 1a-b, Additional files 5: Figures S2 and S3). The distribution of the events on chromosomes during cancer development is shown in Additional file 5: Figure S5a, b and c. Globally, the fraction of the genome that was altered ranged from 8.99% in normal mucosa to 68.9% in carcinoma (p < 0.0001, Table 1, Additional file 5: Figure S1c).

Sample	CNV events	LOH (%)	Genome changed (%)
Mucosa	1–47	0–29.8	0.0058–8.99
Adenoma	3–71	0.1–24.1	0.0054–15.6
Carcinoma	1–156	0.3–34.9	0.012–68.9

Next, we manually searched for genes belonging to the cohesin pathway. This allowed us to identify genomic regions containing HDAC8 (chromosome X, gain in 62.5% of subjects), RAD21 (chromosome 8, gain in 75% of subjects), STAG2 (chromosome X, gain in 62.5% of subjects) and SMC1A (chromosome X, gain in 50% of subjects). This finding indicates that cohesin genes go through extra-copies during CRC development.

SMC1A sequencing reveals high frequency of mutations in carcinoma during CRC development

SMC1A expression increases from normal mucosa to carcinoma during CRC development

SMC1A overexpression reduces the latency period of cancer formation in vivo

To determine the effects of altered cohesin expression in vivo, we introduced both SMC1A wild-type gene and SMC1A c.A2027G (leading to p.E676G change, located in the coiled-coil domain near the hinge domain) mutation, which induced chromosomal instability when transiently transfected in vitro [30], in HCT116 cells, a near-diploid human colorectal cancer cell line with stable karyotype. Western blots were then performed to document that the vectors led to overexpression of both wild-type and mutated SMC1A proteins (Fig. 2a). The effect of SMC1A overexpression was analyzed in vivo using immunocompromised mouse models. The time-dependent analysis showed that the development of tumors peaked after 11 and 13 days (expressed as 50% of survival) with SMC1A c.A2027G mutation and SMC1A wild-type respectively, whereas the latency was longer (18 days) with HCT116 cells (p < 0.05, Fig. 2b). Furthermore, both the weight and volume of the tumors significantly increased in mice inoculated with SMC1A wild-type (p = 0.011, p = 0.0079) and SMC1A c.A2027G (p = 0.013, p = 0.0022) when compared with HCT116 cells (Fig. 2c-g, Additional file 11: Figure S11). No difference was found between SMC1A wild-type and SMC1A c.A2027G, though both the weight and volume of the latter were higher (Fig. 2f-g). Tumors derived from the three cell lines were morphologically indistinguishable from one another, and presented large necrotic areas and several mitotic figures (Fig. 2h).

Next, we measured global gene expression profiles in eleven tumors induced by parental and transduced HCT116 cells. In particular, three tumors induced by parental cells (907_1, 907_2 and 907_3), four induced by SMC1A wild-type transduced cells (907_4, 907_5, 907_6 and 907_7) and four induced by SMC1A c.A2027G transduced cells (907_8, 907_9, 907_10 and 907_11). Unsupervised sample clustering by principal component analysis (PCA) showed that tumors induced by mutant-SMC1A transduced cells differed to a greater extent than those induced by parental and wild-type SMC1A transduced cells (Additional file 12: Figure S12).

SMC1A wild-type and SMC1A c.A2027G displayed 744 (401 down and 343 up) and 742 (486 down and 256 up) dysregulated genes respectively when compared with parental HCT116-induced tumors (Fig. 3a). The transcriptional effects were small, with fold changes ranging from + 0.9 to − 0.17 and from + 0.79 to − 1.00 for SMC1A wild-type and SMC1A c.A2027G respectively (Additional file 12: Tables S8, S9). Dysregulated genes belong to many biological processes (Additional file 13: Figures S13, S14, Tables S10, S11). Furthermore, 68 dysregulated genes were shared in common between tumors induced by SMC1A wild-type and SMC1A c.A2027G (Fig. 3b). All of them displayed only minor fold changes that ranged from + 0.8 to − 0.8 when compared with the baseline of parental HCT116 (Fig. 3c). RNA-seq data was validated in eleven genes (ATG12, CEP55, CHAF1A, CLDN4, H19, HIF1A, KLF5, SAT1, STEAP4, TACSTD2 and TOMM40) by quantitative RT-PCR experiments (Additional file 14, Figures S15, S16). These genes were chosen because they showed differential expression between SMC1A wild-type and SMC1A c.A2027G tumors (Fig. 3c). In addition, since most of them are involved in cancer development their differential expression could explain the aggressiveness of SMC1A c.A2027G tumors.