Identification and validation of genes involved in cervical tumourigenesis

Cervical cancer is the second most common cancer among women worldwide and the most common cancer in Indian women [1]. In most developing countries there are no organized screening programmes, as a result most patients report to tertiary centres in locally advanced stages.

Human papilloma viruses (HPV) have been shown to play a major role in the pathogenesis of cervical cancer, but it alone is not sufficient [2]. Additional events, activation of proto-oncogenes and inactivation of tumour suppressor genes, are required in the induction of cervical cancer.

Cervical cancer goes through a series of pre-malignant stages - Cervical Intraepithelial Neoplasia (CIN) 1, 2 and 3. In general it takes upto about 10 - 15 years for the normal cervical epithelial cell to become a malignant one. However, some CIN2 lesions may develop soon after HPV infection, suggesting that there could be alternate pathways involved. CIN1 and 2 have a higher rate of spontaneous reversion compared to CIN3 [3]. The CIN3 then progresses to invasive carcinoma, which can then metastasize to regional lymph nodes and distant organs (e.g. lung).

The advent of microarray based technology has helped study the expression patterns of more than 40,000 genes at a time [4]. Several groups have used microarray based technology to look for differentially expressed genes in the different stages of cervical tumorigenesis [5, 6]. Few studies have followed up and validated the microarray data in a large number of genes [7, 8]. The objective of our study was to identify genes differentially expressed between normal cervix, CIN1/CIN2, CIN3/CIS and invasive cervical cancer, using oligo-microarray technique, validate the genes so identified using Relative quantitation Real Time Polymerase Chain Reaction (RQ-RT-PCR) and detect potential biomarkers for early diagnosis and therapeutic targets.

Archival total RNA extracted from punch biopsy samples from patients with cervical cancer, collected in RNA later (Ambion, Austin, USA; Cat no: AM7021) and stored in the tumour bank after an informed consent were used, after obtaining the Institutional Ethical committee's approval for the study. The RNA had been extracted from the biopsy samples using the RNeasy RNA extraction kit (Qiagen, Gmbh, Hilden; Cat no: 74106) as per the manufacturer's instructions.

Twenty eight cervical cancer patients' samples were included in the study. The criteria for inclusion in the study were as follows: 1. good quality RNA as assessed by Bio-analyser (RIN 6 or above); 2. paired paraffin block having at least 70% tumour cells; 3. sufficient quantity of RNA be available; 4. patient should have completed prescribed radiotherapy and follow-up information till death/last disease free status be available.

In addition, 5 normal cervix tissues from women who underwent hysterectomy for non-malignant conditions or for non-cervical cancer were included. Four CIN1/CIN2 and 4 CIN3/CIS (one CIN3/CIS was included for RQ-RT-PCR analysis directly) were also included after informed consent. The Normal and CIN samples underwent frozen section to confirm their histopathologic status and the samples were immediately snap frozen in liquid nitrogen. RNA was extracted from the samples using the RNeasy RNA extraction kit, as described above.

The stage distribution of the invasive cancer cases was as follows: IB - 2, IIA - 4, IIB - 18 and IIIB - 4. Twenty seven of the tumours were Squamous cell carcinomas (18 Large cell non-keratinizing, 5 large cell keratinizing and 4 unspecified) and one was a poorly differentiated carcinoma. Eighteen were HPV16 positive, 6 were HPV18 positive and 4 were HPV16 and 18 subtype negative (but HPV positive). All the Normals were HPV negative while one CIN1/2 and all the CIN3/CIS were HPV16 positive.

Using different methods, as described above, genes that were found to be differentially expressed between the four classes (Normal, CIN1/2, CIN3/CIS and Cancer) were identified. We did not use a Training set and a Test set for the Class Prediction model but used LOOCV for cross-validation and obtain the mis-classification error. The list of genes significant by different methods of microarray analysis is given in the Additional File 1 (AF1).

Sixty nine genes were selected for further validation by RQ-PCR using the Taqman Low Density Array card (TLDA) format (Additional File 2). These 69 genes formed part of the 95 genes selected for analysis using the TLDA format. The additional genes were those which had been found to be differentially expressed between the responders and non-responders to radiotherapy only treatment. Apart from the mandatory endogenous 18S rRNA included in the TLDA cards, based on the microarray data, UBC and β2 microglobulin, were included as additional endogenous reference genes.

Two of the samples CXL19-hov160 and CXM024-hov210 which had worked in microarray did not amplify satisfactorily in the RQ-TLDA assay and had to be removed from further analysis. In addition, RPS3A gene did not amplify in any of the samples.

The RQ values after calibrating with the Normal samples (Mean) for all the 94 genes showed 8 additional genes to be overexpressed; 4 (ASB16, CCL18, FST, THOC6) in Cancers, 1 (KLK9) in CIN3/CIS and 3 (RASSF6, TMEM123 and GLB1L3) in CIN1/2 samples. These 8 genes had initially been chosen for validation of the differentially expressed genes between responders and non-responders to radiotherapy. After excluding the genes which did not amplify, we now had 76 genes for further analysis.

Of the 31 genes which had been selected based on a greater than 2 fold difference between cancer versus CIN1/2 & Normal, 28 were concordant between the microarray data and the RQ-RT-PCR (Concordant rate of 90%). Three of four genes selected based on higher level of expression in Normals compared with all other classes showed concordance between the different methods of analysis. In the case of CIN1/2, concordance was seen in 6/7 genes (86%). However, with CIN3, this dropped to 41% (11/27). In four additional genes, there was a two fold greater difference between CIN3/CIS and Normal but not with CIN1/2. The overall concordance rate between the microarray data and the RQ-RT-PCR was 70% (48/69).

The list of genes validated and found to have a greater than 2 fold difference compared to the Normal, in the 3 different classes (Cancer, CIN3/CIS and CIN1/2) is given in Table 1. Figure 1 provides the fold change relative to Normal for these genes.

The genes were grouped on the basis of whether or not they were known to be involved in cervical tumorigenesis (Tables 2 and 3). Gene Ontology mapping was done using Babelomics software [18], which showed an over-representation of genes involved in cell cycle, cell division, catabolic process and multi-cellular organismal metabolic process. The genes identified to be differentially expressed were then analysed for specific pathways of relevance by manual curetting of data from published literature and online databases. The genes were grouped under the following categories: 1. Cell cycle regulatory genes (n = 13); 2. Interferon induced genes (n = 5); 3. Ubiquitin pathway (n = 5); 4. Myc Pathway [19] (n = 12); 5. HPV-E6/E7 related genes [20] (n = 14); 6. RNA targeting genes (n = 3) (details are given in Additional File 3). In addition, 40 genes in our list were found to be potentially regulated by p53 family of genes [21] (Additional File 4). Using GeneGo's Metacore software (Trial version) (url: http://​www.​genego.​com), the relationship of our validated genes with known Transcription factors was analyzed. Based on this and from the manually curetted information, we then attempted to construct relationship chart (Figure 2) providing information on the gene interactions.

Using IHC, we studied the protein expression for MMP3 in 5 normal cervical tissues, 30 dysplasias of varying grades and 27 invasive cancers. Using a semi-quantitative scoring system and a cut-off threshold set based on the normal cervical tissue staining, 6/30 dysplasias and 11/27 invasive cancers were found to overexpress MMP3 protein (Figure 3A). Among the patients whose tumours had been treated only with radical radiotherapy and had been followed up for a minimum period of 3 years, over-expression was seen in a greater number of tumours that failed treatment (6/9) compared to those free of disease at 3 years (2/12) (p = 0.03). p16 was found to be overexpressed in 19 of 31 dysplasias of varying grade and in 27/29 cancers (p = 0.005) (Figure 3B).

Using IHC, we found UBE2C to be overexpressed in 28/32 cancers, 2/11 CIN3/CIS and none of the CIN1 or 2 (Fisher's exact test p = 2.2 e^-11) (Figure 3C). Using RQ RT-PCR, UBE2C was found to be overexpressed by more than 2 fold in SiHa, HeLa, C33A and ME180 relative to the HEK293 cells (Figure 4A). The growth of SiHa cells transfected with dominant negative UBE2C was significantly reduced at 48 and 72 hours compared to SiHa WT and SiHa transfected with pcDNA vector alone (p < 0.001) (Figure 4B).

There was good overall concordance between the microarray and the RQ-RT-PCR data. The lower concordance rate seen with the CIN3/CIS may be due to the additional CIN3 sample processed directly using RQ-RT-PCR. The relative quantitation values with and without the additional sample is given as Additional File 5. The concordance rate between microarray and semi-quantitative RT-PCR in the study by Gius et al [8] was less than 50%, using the standard microarray data analysis package.

There were several instances, wherein, a small difference in Microarray (above the 2 fold mandatory criteria) sometimes translated to large differences with RQ-RT-PCR (e.g. p16, MMP1, MMP3) and vice versa (e.g. CD36). This reinforces the point about the limitation of the microarray technique and it does emphasize the need for further validation, using assays like RQ-RT-PCR.

HPV16 was the predominant subtype seen in the invasive cancers and CIN3/CIS. However, we did not look for all the high risk subtypes and hence cannot exclude multiple subtype infection. Four of the cancers were HPV positive but HPV16 and 18 negative, suggesting that other high risk subtypes could be involved. None of the normal cervical tissues were HPV positive.

The genes that were for the first time, found to be over-expressed in cervical cancers compared to Normal cervix, is given along with information in which other cancers they have been reported to be overexpressed (Table 2A). Our study, for the first time, has identified 20 genes to be up-regulated in cervical cancers and 5 in CIN3; 14 genes were found to be down-regulated. In addition, 26 genes identified by other studies, as to playing a role in cervical cancer, were also confirmed in our study. UBE2C, CCNB1, CCNB2, PLOD2, NUP210, MELK, CDC20 were overexpressed in tumours and in CIN3/CIS relative to both Normal and CIN1/CIN2, suggesting that they could have an important role to play in the early phase of tumorigenesis. Among the genes which were up-regulated in cancers compared to that of Normal, CIN1/2 or CIN3/CIS, IL8, INDO, ISG15, ISG20, AGRN, DTXL, MMP1, MMP3, CCL18, TOP2A AND STAT1 are likely to play an important role in the progression of the disease.

STAT1 gene has a bi-phasic level, a rise in CIN1/2, drop in CIN3/CIS and a significant rise in invasive cancers. STAT1 has been considered generally to be a tumour suppressor, while STAT3 and STAT5 are known to be proto-oncogenes. However, recent studies have shown STAT3 to have both oncogenic and tumour suppressor function [22]. It could be that in cervical cancer, STAT1 may be protective in the early phase of HPV infection but could function as a proto-oncogene in the invasive stages of the disease. Highly invasive melanoma cell lines had high levels of STAT1 and c-myc [23].

The study by Lessnick et al., [24] showed that introduction of the potentially oncogenic EWS-FLI transcript into the fibroblasts, resulted in growth arrest rather than transformation. Knocking out p53 using HPV E6 helped overcome the growth arrest but was not sufficient to induce malignant transformation. The study used microarray to identify genes differentially expressed between the EWS-FLI transfected and the mock transfected cell line and found several genes related to growth promotion down-regulated. Our study had several genes [19] overlapping with theirs. Thirteen genes from our study were found to be HPV E6/E7 related genes[20] and 40 of the genes in our list were found to be potential p53 Family Target genes[21] (Additional File 3). In addition, there were 12 myc regulated genes, (MYC Cancer database at http://​www.​myc-cancer-gene.​org/​) of which CSTB which has been reported to be down-regulated by myc, was down-regulated in CIN3/CIS and in Cancer [19].

p16 gene, a tumour suppressor has been reported to be over-expressed in dysplasias and invasive cancer of the cervix. Several studies have tried to use this as a marker in the PAP smears for more reliable interpretation of the smear. von Knebel's group from Germany [25], had developed an ELISA to detect p16 in the cervical cell lysates, and reported a 96% sensitivity to pick up high grade dysplasias. Subsequently, the p16 ELISA assay was compared with Hybrid Capture 2 and was found to have comparable sensitivity and a slightly better specificity (46.9% versus 35.4%) [26]. Our RQ-RT-PCR data shows a gross over-expression of p16 in the CIN3 and invasive cancers (>250 fold). In our series of dysplasias and cancers, p16 protein was found to be overexpressed in invasive cancers compared to the dysplasias.

Figure 2 shows the inter-relationship of our genes with E6 and E7 protein and other known Transcription factors including p53, E2F, c-myc, B-MYB and c-Jun. The important genes in our list MELK, ISG15, STAT1, IL8, MMP1 and MMP3, could be playing critical roles in the tumorigenic pathway and could be potential targets for newer therapies.

UBE2C is an E2 enzyme involved in the process of ubiquitination. Townsley et al. [17] had developed a dominant negative UBE2C which lacks the catalytic activity. When the dominant negative UBE2C was expressed in SiHa cells, which have nearly 4 fold greater levels of UBE2C compared to HEK293 cells, it produced a significant growth inhibition (Figure 4B), indicating that the dominant negative UBE2C is competing with the wild type UBE2C, and can interfere with cell proliferation. Additional studies will be required to understand the mechanism by which this effect occurs.

Our study has helped identify newer genes which could play a role in the cervical tumorigenesis and could offer the potential of developing newer diagnostic markers and therapeutic targets. We have confirmed over-expression of MMP3, UBE2C and p16 in tumours, by IHC. This will need to be validated further in a larger series of tumours and dysplasias. UBE2C will need to be studied further to assess its potential as a target for the treatment of cervical cancer.

The authors declare that they have no competing interests.