Introduction

Familial adenomatous polyposis (FAP) is a rare form of colorectal cancer caused by germline mutations in the adenomatous polyposis coli (APC) gene. Approximately 70–90% of FAP patients have identifiable germline mutations in APC [1, 2]. FAP is clinically characterized by the formation of hundreds to thousands of adenomas that carpet the entire colon and rectum [3]. Although initially benign the risk of malignant transformation increases with age such that, if left untreated, colorectal carcinoma usually develops before the age of 40 years [4].

Loss of APC results in dysregulation of the Wnt signalling pathway that leads to the constitutional activation of the transcription factor Tcf-4, which has been associated with adenoma formation [5]. Alterations in Wnt signalling cause stem cells to retain their ability to divide in the upper intestinal crypt, thereby forming monocryptal adenomas [6]. Eventually the adenomas may acquire metastatic potential, resulting in carcinoma development [7]. Not all adenomas will progress to malignant tumours; however, due to the abundance of adenomas carcinoma development is virtually assured [8].

Apart from the apparent loss of APC function, little is known about the molecular processes involved in adenoma initiation [6]. Similarly, the molecular events occurring during the transformation of adenomas into carcinomas are poorly understood, as are the mechanisms that underlie the development of extra-colonic disease in FAP.

It is well established that FAP patients are susceptible to benign extra-colonic tumours, including desmoid tumours [3]. Although rare in the general population, desmoids occur in approximately 10% of FAP patients and they are the second most common cause of death [9]. Desmoid tumours are poorly encapsulated and consist of spindle-shaped fibroblast cells with varying quantities of collagen [10]. Despite their apparent inability to metastasize, desmoid tumours can be extremely aggressive [11].

It has been speculated that desmoid formation is a result of an abnormal wound healing response [12]. Desmoids can affect surrounding viscera, causing potentially fatal complications [13]. FAP-associated desmoid tumours are usually associated with germline APC mutations [14], but somatic APC mutations have been detected in sporadic desmoid tumours [15].

Microarray technology has an enormous potential for applications in the endeavour to better understand tumours and their development [16]. The ability to detect expression levels of thousands of genes can identify particular genes that are either up- or down-regulated in different tumour types [17]. Tumours that are currently categorized by similar morphology, such as desmoid tumours, may be more usefully divided into subtypes according to their expression profiles [18]. Particular expression profiles in tumours may also be capable of predicting the clinical outcome in specific patients in the early stages of tumour development [18]. In colorectal cancer, gene expression profiles of adenomas and adenocarcinomas have been compared and subsets of genes expressed at common levels in both lesions have been identified as well as expression patterns that are unique to each [19]. Gene expression profiling has the potential to identify factors involved in the malignant transformation of adenomas, and may aid in the diagnosis of benign versus malignant disease.

Although genome-wide expression studies have been reported on FAP adenomas and desmoid tumours, the present one of the first to compare the two tissue types. The first aim of this study was to identify distinct gene expression profiles for colorectal and stomach FAP adenomas and desmoid tumours. The second aim was to determine the similarity between the gene expression profiles in FAP adenomas and desmoid tumours to identify mechanisms important in regulating formation of these lesions. To achieve this, mRNA from normal colon, FAP stomach and colon adenomas and desmoid tumours was measured using whole human genome expression BeadChips (Illumina). The findings of this study further our understanding of the mechanisms underlying FAP and desmoid tumour formation.

Materials and methods

FAP adenoma and tumour tissue and controls

Frozen adenoma tissue from 4 FAP patients was available for this study. Colorectal FAP adenoma A was from an individual aged 40 at the time of surgery. Genetic testing revealed a heterozygous A5465T change in the APC gene, causing a missense change from aspartic acid to valine at position 1822 in the amino acid sequence. The specimen obtained for this study was obtained as a result of a proctocolectomy. The pathology report indicated that over 100 tubulovillous adenomas were present in the original specimen, with no evidence of invasive tumour. Patients B, C and D harboured the same frameshift mutation, a 4 base pair deletion at position 3462–3465 of the APC gene. Patient B was diagnosed with FAP at the age of 11 years, patient C at 13 years of age, and patient D at the age of 37 years. One gastric adenoma was obtained from patient D, in addition to a colonic adenoma. Normal colon tissue from 7 healthy individuals with no history of FAP or desmoid disease was used as a mixed reference sample for this study.

Desmoid Disease Tissue

Desmoid tumour tissue from two individuals was available for this study. Patient A had FAP-associated desmoid disease. There was a family history of FAP, but no known history of desmoid disease. The individual harboured a 1 bp deletion in exon 15 of the APC gene resulting in a frameshift that introduced a premature stop codon at amino acid position 964. Patient B had a family history of FAP and desmoid disease. This patient harboured a 17 bp duplication in exon 15 of the APC gene, which introduced a premature stop codon at amino acid position 1969. A previously established fibroblast cell line from a healthy individual with no history of FAP or desmoid disease was used as a control for this study. The fibroblast cell line was cultured in 1× Complete DMEM media at 37°C (5% CO2).

RNA Extraction

2–3 mm2 pieces of fresh frozen FAP adenoma and desmoid tumour tissue were cut from the original sample and transferred immediately to 1 ml Trizol reagent (Invitrogen, USA). Similarly, approximately 1–10 × 106 control fibroblast cells were lysed in 1 ml Trizol reagent (Invitrogen, USA). RNA was extracted per manufacturer's instructions. The RNA pellet was washed with 75% ethanol, before being dissolved in 20 μl water.

The total RNA was purified using a Qiagen RNeasy MiniElute Cleanup Kit as per manufacturer's instructions. The concentration of the purified total RNA samples was measured using a Quant-It RiboGreen RNA Assay Kit (Invitrogen, USA) and a fluorometer (Fluostar OPTIMA) as per manufacturer's instructions.

RNA amplification

To synthesise first and second strand cDNA and amplify biotinylated cRNA from the total RNA, an Illumina Totalprep RNA Amplification Kit was used as per manufacturer's instructions.

The purified cRNA samples were quantified to determine the volume required for the BeadChip hybridisation step via the Quant-iT RiboGreen RNA Assay Kit as described previously.

Illumina BeadChip Procedure

Hybridisation to the Illumina Sentrix 8 BeadChip was performed according to the manufacturer's instructions without modification. The Sentrix 8 BeadChips were read using an Illumina Beadarray reader (San Diego, CA, USA).

Data Analysis

Analysis and normalisation of expression data from the 24,000 transcripts was carried out using BeadStudio 2.0 (Illumina, San Diego, CA, USA). The t-test error model and cubic spline normalisation was used for all samples. A differential analysis was applied to all adenoma and tumour samples using the Illumina custom test of significance, utilising the mixed normal colon control as the reference group. GeneSpring 5.0 (Agilant, Santa Clara, CA, USA) used standard correlation and distance to create dendrograms (Experiment trees) to show relationships between gene expression profiles. A second dendrogram (Gene tree) was created for each gene list using standard correlation and distance to show relationships between the expression levels of genes across the groups.

Results

Gene expression data from over 23,000 genes on Illumina HumRef-8 BeadChips was analysed and normalised using Illumina BeadStudio 2.0 software. Cubic spline normalisation and the t-test error model were employed for all the FAP adenoma, normal colon and desmoid tumour samples. Correlation analyses identified the average R2 value of the duplicates for each sample as 0.950 ± 0.04. An average of each duplicate pair was then taken before additional analysis was carried out.

Differential gene expression analysis in FAP adenomas and healthy colon tissue

Differential analysis using the mixed normal colon control as the reference group was applied to all adenoma and tumour samples. Genes in each analysis were excluded if their fluorescence detection score was less than 0.99, and if their differential score was less than 13 (p > 0.05). From the genes that met the exclusion criteria, according to detection and differential scores, lists were generated for genes both up- and down-regulated more than 2-fold in the FAP adenoma samples compared to the mixed normal colon control. The genes commonly up- and down-regulated across all the FAP adenomas are shown in Tables 1 and 2 and genes that were commonly up- or down-regulated across the 4 colorectal FAP adenomas only are shown in Tables 3 and 4 respectively.

Table 1 Genes commonly up-regulated more than 2-fold in all FAP polyps compared to normal colon
Full size table
Table 2 Genes commonly down-regulated more than 2-fold in all FAP polyps compared to normal colon
Full size table
Table 3 Genes commonly up-regulated 2-fold or more in colorectal FAP polyps compared to normal colon
Full size table
Table 4 Genes commonly down-regulated 2-fold or more in colorectal FAP polyps compared to normal colon
Full size table

Cluster analysis was performed using GeneSpring 5.0 software in order to further characterise the similarity across the FAP samples and to determine if there was differential gene expression compared to healthy colon tissue. The stomach FAP duplicates display profiles slightly distinct from the other FAP adenomas. The normal colon duplicate profiles are unique to all other profiles (Figure 1).

Figure 1
figure 1

Cluster analysis of FAP polyps and mixed normal colon. The columns represent the gene expression profiles of each sample. Green – low expression level, yellow – medium expression level, red – high expression level. The relationships between each sample are shown by the upper dendrogram. The colouring in the upper dendrogram represents the sample type: green (left) – normal colon; blue – colorectal FAP polyps; yellow – stomach FAP. 1 – Normal Colon Duplicate; 2 – Normal Colon Duplicate; 3 – Colorectal FAP Polyp A Duplicate; 4 – Colorectal FAP Polyp A Duplicate; 5 – Colorectal FAP Polyp D Duplicate; 6 – Colorectal FAP Polyp D Duplicate; 7 – Colorectal FAP Polyp B Duplicate; 8 – Colorectal FAP Polyp B Duplicate; 9 – Colorectal FAP Polyp C Duplicate; 10 – Colorectal FAP Polyp C Duplicate; 11 – Stomach FAP Polyp D Duplicate; 12 – Stomach FAP Polyp D Duplicate.

Full size image

Differential gene expression analysis in desmoid tumours and control fibroblasts

The average expression in the desmoid tumours was compared to the control fibroblast cell line and significantly altered expression identified by differential gene expression analysis. Genes in each analysis were excluded if their fluorescence detection score was less than 0.99, and if their differential score was less than 13 (p > 0.05). Genes with differential expression and up- or down-regulated more than 2-fold in the desmoid tumour samples compared to the normal fibroblast cell line were compiled into lists (Tables 5 and 6).

Table 5 Genes commonly up-regulated 2-fold or more in desmoid tumours compared to normal fibroblast cells
Full size table
Table 6 Genes commonly down-regulated 2-fold or more in desmoid tumours compared to normal fibroblast cells
Full size table

To reveal any correlation between the expression profiles of desmoid tumours and FAP adenomas, the data from each group were compared. In the upper dendrogram (Figure 2) it can be seen that all the FAP adenomas cluster in the same group. The desmoid tumours and the normal fibroblast cell line clustered in an entirely different group to the FAP samples. The FAP adenomas and the normal colon have distinct gene profiles compared to the desmoid tumours and the normal fibroblasts. Within the FAP adenomas, the stomach adenoma and the normal colon have slightly different gene profiles compared to the colorectal adenomas.

Figure 2
figure 2

Cluster analysis of FAP polyps, normal colon, desmoid tumours and normal fibroblasts. The columns represent the gene expression profiles of each sample. Green – low expression level, yellow – medium expression level, red -high expression level. The relationships between each sample are shown by the upper dendrogram. The colouring in the upper dendrogram represents the sample type: green (left) – normal colon; blue – colorectal FAP polyps; orange – stomach FAP polyp; green (right) – desmoid tumours; purple – fibroblast cell line. 1 – Normal Colon Duplicate; 2 – Normal Colon Duplicate; 3 – Colorectal FAP Polyp A Duplicate; 4 – Colorectal FAP Polyp A Duplicate; 5 – Colorectal FAP Polyp D Duplicate; 6 – Colorectal FAP Polyp D Duplicate; 7 – Colorectal FAP Polyp B Duplicate; 8 – Colorectal FAP Polyp B Duplicate; 9 – Colorectal FAP Polyp C Duplicate; 10 – Colorectal FAP Polyp C Duplicate; 11 – Stomach FAP Polyp D Duplicate; 12 – Stomach FAP Polyp D Duplicate; 13 – Desmoid Tumour A Duplicate; 14 – Desmoid Tumour A Duplicate; 15 – Desmoid Tumour C Duplicate; 16 – Desmoid Tumour C Duplicate; 17 – Fibroblast Cell Line Duplicate; 18 – Fibroblast Cell Line Duplicate.

Full size image

Discussion

In this study, 24 K Illumina HumRef-8 BeadArrays were used to compare gene expression of FAP adenomas, desmoid tumours and normal fibroblasts. To date there have been a number of small scale gene expression studies on FAP adenoma tissue, the vast majority of which have employed immunohistochemistry (IHC). Most of these studies have been performed on individual genes that include E-cadhein, α-, β- and -catenin, COX-1, COX-2, and c-myc [20–25]. In addition, one study used semi-quantitative RT-PCR to study GKLF [26]. The only report examining global gene expression in human FAP adenoma tissue identified 84 differentially expressed genes in adenomas compared to normal colon tissue [27].

In this study, the gene expression profiles obtained from the FAP adenomas indicate that colorectal adenomas are similar but distinctly different to the stomach adenomas. There were a large number of commonly expressed genes identified across the colorectal FAP adenomas, but when the differentially expressed genes from the stomach adenoma were included in the analysis the number of commonly expressed genes decreased dramatically. The genes that were differentially expressed in the four colonic adenomas and one stomach adenoma were investigated more closely in an attempt to identify common genetic features in FAP. From this analysis genes involved in the cell cycle, transcription and metabolism were the most frequently up-regulated. The most frequently down-regulated genes included those involved in metabolism, cell adhesion, signal transduction, transcription and transport. Since adenomas develop due to a breakdown in the fidelity of the Wnt signalling pathway it was not surprising to observe the over-expression of genes involved in cell cycle progression.

Altered Expression of Wnt/β-catenin Target Genes in Colorectal FAP Adenomas

It has been long established that deregulation of the Wnt signalling pathway due to APC mutations plays a major role in the progression of FAP [5]. The Wnt/β-catenin signalling pathway is involved in the control of expression of Sox9, PTTG1 and EphB2, all of which were found to be up-regulated by more than 2-fold in all the colorectal FAP adenomas compared to the normal colon.

PTTG1 is regulated by a TCF binding sequence in its promoter region [28]. The normal function of PTTG1 is to regulate chromosome segregation during cell division [29]. Over-expression of PTTG1 has been reported frequently in various types of cancer, including colorectal, and has been associated with angiogenesis [30–32]. The role of PTTG1 in angiogenesis is thought to be a result of its part in mediating the secretion of the basic fibroblast growth factor into the extracellular matrix, which promotes proliferation and migration of colorectal cancer cells [30, 31].

The Sox9 gene encodes a transcription factor that is required for chondrogenesis and male gonad development [32], which is under the control of the Wnt signalling pathway [33]. The expression of the Sox9 gene in the intestine is dependent on the activity of the β-catenin/TCF-4 complex, although it is unknown whether this complex interacts directly with the Sox9 promoter or through another of its targets [33].

The EphB2 gene encodes the Eph receptor B, which has been shown to be a target of the Wnt signalling pathway [34]. There is evidence to suggest that normal patterning in the epithelium of the intestinal crypts is coordinated by EphB2 and its ligand, ephrin B [34]. Over-expression of EphB2 is often found in colorectal cancers, but there is confusion about its role in tumourigenesis. Many studies on other tumours have reported EphB2 over-expression as a marker of poor prognosis, but recent studies in colorectal cancer have suggested otherwise [35, 36].

Altered Expression of Cell Cycle-Related Genes in Colorectal FAP Adenomas

A number of genes found to be commonly up-regulated in the adenomas used in this study have previously been reported as being over-expressed in various types of cancers. These genes include the cell cycle-related genes Chromosome condensation protein G (HCAP-G), Protein regulator of cytokinesis 1 (PRC1), SMC4 structural maintenance of chromosome 4-like 1 (SMC4L1) and Cyclin B2 (CCNB2) [37–39]. Although these genes are associated with tumour development none have been thoroughly characterized in FAP to date.

Altered Gene Expression in Desmoid Tumours

A limited number of gene expression studies have been performed on desmoid tumours, primarily due to the difficulties in obtaining tissue. Two reports have studied gene expression in desmoid disease using 6.8 K, 19 K and 33 K Affymetrix microarrays [40, 41]. Skubitz and Skubitz (2004) [40] reported that ADAM12, WISP-1, Sox-11 and fibroblast activation protein-a are uniquely expressed in desmoids. Denys et al. (2004) identified 69 differentially expressed genes in desmoid tumour tissue compared to normal fibroblasts, before focusing on the down-regulation of IGFBP-6 [41].

A number of genes that were identified as being differentially expressed in desmoid tumours in this study have been reported previously. The over-expressed genes include transforming growth factor β3 (TGFβ3), a distintegrin and metalloproteinase domain 19 (ADAM19), chimerin 1 (CHN1), and ephrin-B3 (EFNB3) [40, 41]. The under-expressed genes include quiescin Q6 (QSCN6), prostaglandin I2 synthase (PTGIS), proenkephalin (PENK), keratin 18 (KRT18), cytokine receptor-like factor 1 (CRLF1), pentaxin-related gene (PTX3) and endoglin (ENG) [41].

Ephrin-B3, a Wnt Target Overexpressed in Desmoid Tumours

The known Wnt/β-catenin target gene ephrin-B3 [42] has been found in this study to be up-regulated more than 2-fold in desmoid tumours compared to normal fibroblasts. The ephrins are ligands for the EPH receptor family, whose normal function is to organize cell patterning in the intestinal crypts [34]. In addition, more recent observations suggest that ephrins are tumour suppressors, although the mechanism by which this is affected remains to be clarified [3, 43, 44]. Further investigation into the precise role of ephrin-B3 is required before any conclusions can be made regarding its role in desmoid disease.

Wound Healing-Associated Genes Differentially Expressed in Desmoid Tumours

Two genes, transforming growth factor β-3 (TGFβ3) and pleiotrophin (PTN), were found to be differentially expressed in desmoid tumours. Both genes are associated with wound healing and could potentially explain the growth advantage of desmoid tumours [45].

TGFβ3 is a multifunctional protein, having roles in cell proliferation and differentiation during embryogenesis and wound healing [46]. Pleiotrophin has been reported to be strongly expressed in many human cancers, and is thought to promote malignant transformation and angiogenesis [47]. It is also frequently found to be up-regulated during the wound healing process [48].

In this study, three genes associated with negative regulation of the wound response have been identified as being under-expressed in desmoid tumours. The three genes are: signal transducer and activator of transcription 1 (STAT1), mothers against decapentaplegic homolog 3 (MADH3 or Smad3) and mothers against decapentaplegic homolog 6 (MADH6 or Smad6). STAT1 enhances transcription in response to interferon-, an action which has been shown to inhibit the wound healing response by preventing phosphorylation of Smad2 and Smad3 [49]. This in turn inhibits the action of TGFβ on the wound response [50]. The role of Smad3 in the wound response is not entirely understood; however, the absence of Smad3 causes an accelerated healing response, even though its over-expression has also been shown to promote healing [51, 52]. Smad6 is a known inhibitor of TGFβ, and has shown to be down-regulated in keloids [53].

The abundance of wound response-related genes found to be deregulated in the desmoid tumours in this study adds to the notion that desmoid formation is an abnormal wound response. The finding of over-expressed genes involved in fibroblast proliferation and migration could explain the abnormal proliferation and local invasiveness of desmoid tumours. The down-regulation of angiogenesis-associated genes could account for the poor vascularisation of desmoids.

The limiting factor in this study of desmoid tumours is the small number of desmoids available. In order to reach more conclusions regarding the exact molecular nature of desmoids and their growth mechanisms, a much larger sample size would be required.

Comparison of FAP Adenoma and Desmoid Tumour Molecular Profiles

It has long been recognized that desmoid tumours occur with a much higher frequency in FAP patients than in the general population. The apparent role of aberrant Wnt signalling in both diseases could indicate a molecular similarity between the two. Although Wnt target genes were identified as being up-regulated in both tumour types in this study, the specific genes were different in the two groups. The finding of different Wnt targets could be attributed to the use of different control groups for the FAP adenomas and desmoid tumours. Nevertheless, the molecular profiles obtained using cluster analysis clearly demonstrated that FAP adenomas and desmoid tumours display distinctly different gene expression profiles.