Background
Colorectal carcinoma (CRC) is the third most lethal malignancy in the United States for both women and men, with an overall 5-year survival rate of around 60% [
1]. 106,680 cases of colon and 41,930 cases of rectal cancer are expected to occur in 2006. It is estimated that 55,170 deaths from CRC will occur in 2006, accounting for 10% of all cancer deaths [
2]. At present, the only curative treatment is surgical resection: however, it is often impossible to remove all cancer cells, especially those that have invaded the surrounding tissues. The penetration of tumor cells into lymphoid vessels and blood vessels leads to tumor metastasis and ultimately the tumor becomes fatal [
3]. The current major method for assessing the risk of metastatic recurrence and need for adjuvant chemotherapy is to examine tumor resection specimens for evidence of metastasis to local lymph nodes. However, this approach may be of limited prognostic value as a sizeable fraction of colorectal carcinomas have innate resistance to chemotherapy and 25% to 30% of the patients presenting with lymph-node negative tumors also develop fatal disease [
4]. Therefore, there is an urgent need for more accurate and informative methods of individual risk assessment for patients with CRC, some of which might be based on the molecular properties of the primary tumor itself [
5].
Tumor invasion and metastasis are the result of highly coordinated processes that involve multiple intracellular and extracellular factors [
6‐
8]. In part, carcinoma cell migration is enabled by the altered differentiation status of the epithelial cells that includes changes in cell-cell and cell-matrix adhesion properties and in the organization of the actin cytoskeleton [
9‐
12]. With regard to the composition of the cytoskeleton of carcinoma cells, the actin-bundling protein, fascin, has become of great interest due to its functional involvement in cell adhesion and motility [
13,
14]. Fascin is expressed in mature dendritic cells, mesenchymal cells, endothelial cells and neurons during development and in the adult [
15,
16]. It is absent from most normal epithelia, but is expressed in multiple epithelial neoplasms, including carcinomas of the pancreas, lung, esophagus, stomach and breast [
17‐
24]. Most strikingly, fascin expression has been associated with a poorer prognosis in carcinomas of the lung, esophagus, stomach and breast [
19‐
21,
23]. In node-negative, invasive hereditary breast carcinomas, fascin is frequently expressed by BRCA1-associated tumors [
24]. Fascin has also been identified as a component of a gene signature that correlates clinically with breast cancer metastasis to the lung [
25].
In cell culture, expression of recombinant fascin in fascin-negative colonic adenocarcinoma cells correlated with increased proliferation, altered beta1 integrin distribution, increased invasive capacity and altered differentiation status [
26]. Similar findings have been obtained in other epithelial cells, suggesting that fascin may contribute to a more aggressive tumor phenotype by facilitating carcinoma cell migration and invasion [
20,
27]. However, with regard to colorectal cancer, an initial study of tumor specimens examined only 10 cases without regard to tumor stage or clinical annotation [
26]. Thus, the clinical relevance of fascin expression in CRC remains unclear and it is also unknown whether fascin plays any role in the early development of colorectal carcinoma.
Studies of fascin in multiple cell types have established that its actin-binding properties are regulated by extracellular cues acting both through adhesion receptors and receptor tyrosine kinases [
14,
16,
28‐
30]. Furthermore, several studies have indicated that fascin expression may be related to the proliferative status of carcinomas. In mouse xenografts, cells from fascin-positive human ovarian carcinomas were more tumorigenic than fascin-negative lines [
31]. Colonic epithelial cells engineered to over-express fascin proliferated faster in culture than control cells [
26]. Despite these findings, the relationship between fascin expression and cell proliferation in human cancers is currently unclear. In non-small cell lung carcinoma, highly fascin-positive tumors tended to be highly proliferative, as established by Ki67 antibody staining. However, it was also noted that individual Ki67-positive cells stained less strongly for fascin than surrounding tumor cells [
19]. In gastric carcinoma, the reverse trend was observed, with a higher Ki67 index in fascin-positive areas compared to fascin-negative areas [
21].
Here, we have examined the clinicopathological significance of fascin and Ki67, singly and in combination, in a series of colorectal tumor specimens. By examining whole paraffin-embedded sections, we closely reviewed the localization and topographic relationship of fascin and Ki67 in colonic adenomas and colorectal adenocarcinomas. The potential prognostic significance of fascin expression was assessed by using clinically-annotated samples in a CRC tissue microarray of 158 colorectal adenocarcinomas and 15 adenomas. We report that fascin and Ki67 are most frequently inversely correlated at the cellular level. Fascin is upregulated at the adenoma stage and is of potential prognostic significance as a marker of aggressive colorectal adenocarcinomas.
Methods
Patients and surgical specimens: conventional sections
We studied 142 tumor samples that included 107 adenomas (89 sporadic and 18 from patients with familial adenomatous polyposis (FAP)) and 35 adenocarcinomas from an unselected series of cases seen at the Cleveland Clinic between 2004 and 2006. The adenoma samples were derived from 76 patients, comprising 65 patients with sporadic adenomas and 11 FAP patients. Adenomas with high grade dysplasia were not included in the set and only moderately- or poorly-differentiated adenocarcinomas were collected. 76 of the sporadic adenomas were < 2 cm in diameter and 13 were > 2 cm. 16 of the FAP were <2 cm and 2 were > 2 cm in diameter. One representative section of each specimen was selected for study. In adenocarcinomas, this represented the portion of deepest extent. In adenomas, the sections were selected to demonstrate the entire lesion including the head and stalk of the adenoma, where available.
Tissue microarrays
A custom built instrument equipped with stainless steel thin-wall needles (Beecham Instruments, Hackensack, NJ), was used to take core tissue biopsies from carefully selected, morphologically representative areas of the original paraffin blocks ("donor" blocks) and arrayed into a new "recipient" paraffin block. The precision digital device that equips this instrument allows for precise placement and spacing of the tissue cores into the recipient block. The resulting tissue microarray (TMA) contained a total of 374 cores, from 14 normal colonic epithelia, 15 adenomas and 158 colorectal adenocarcinomas that were diagnosed at The Cleveland Clinic between 1993 and 1999. The majority of the adenocarcinomas were moderately-differentiated and six were poorly-differentiated. Tumors were classified according to standard TNM staging guidelines [
32], and the location of tumors was divided into two groups, proximal (cecum, ascending colon, hepatic flexure and transverse colon) and distal (splenic flexure, descending colon, sigmoid colon and rectum). Of the 158 adenocarcinomas, 131 samples were available for scoring after fascin staining; this study group derived from 79 males and 52 females. The median age was 64.5 years (range 32–95 years). Of the 62 patients with stage III or stage IV tumors in the dataset, 19 received no adjuvant treatment, 27 received adjuvant chemotherapy and 16 received combined adjuvant chemotherapy and radiotherapy. To minimize sampling errors, two separate large diameter (1.5 mm) tissue cores of each adenocarcinoma were included in the array, totaling a surface area of 3.5 mm
2 per case. Each separate tissue core was assigned a unique TMA location number which was subsequently linked to a CCF Institutional Review Board-approved (IRB-5085) database containing a mean 38 months of clinical follow-up.
Immunohistochemical staining
Both the conventional paraffin sections and the TMAs were treated similarly. Immunohistochemistry was carried out using a fully automated Ventana Benchmark system (Ventana, Tucson, AZ). Briefly, 4 μm thick unstained sections were placed onto a electrostatically charged glass slide and baked to allow for tissue adherence. The glass slides were pretreated with the recommended pretreatment solution provided by Ventana for tissue deparaffinization and antigen retrieval. After primary antibody incubation and a secondary biotinylated antibody/streptavidin amplification step, antigen detection was carried out by peroxidase/3,3'-diaminobenzidine development. Primary antibodies used in this study were directed against fascin (DAKO, clone 55k-2, dilution 1:50) and Ki67 (Novacastra, clone MM1, dilution 1:5).
Evaluation of immunohistochemical stainings and scoring
A section without the primary antibody was used as a negative control in each case. A normal tonsil tissue was used as a positive control to confirm fascin immunoreactivity in tonsillar dendritic cells in each series of experiments. Fascin immunoreactivity in each specimen was verified by the staining of endothelial cells in microvessels. Fascin-positive adenoma or adenocarcinoma samples were defined as those showing a cytoplasmic pattern of expression. For the conventional sections, each section was recorded on the basis of the area of fascin staining and was scored on a four point scale: 1), negative; 2), 1% to 10% positive; 3), 10% to 50% positive, and 4), more than 50% positive. For the TMA, this standard scoring scheme was used with minor modifications. Each separate section or tissue core was recorded on basis of the area and intensity of the staining and scored on a 0 to 3+ scale, (0, no staining; 1+, less than 10% of cells moderately to strongly stained, or a weak reactivity in any % of cells; 2+, 10% to 50% of cells moderately or strongly stained; 3+, more than 50% of cells with moderate or strong staining), and the results entered into the research database. In all samples, Ki67 index was scored according to the percentage of cells with positive nuclear staining and was divided into three groups: 1+ (less than 10% of cells with positive nuclei); 2+ (10% to 50% of cells with positive nuclei), and 3+ (more than 50% of cells with positive nuclei).
All slides were evaluated independently by two investigators (YH and MS) without any prior knowledge of the clinical information. When the opinions of the two evaluators differed, a consensus agreement was reached by re-review of the slides and thorough discussion. Where scores differed between two cores on the TMA, the average score was taken.
Statistical analysis
The immunohistochemical staining scores for fascin and Ki67 in relation to clinicopathological factors were analyzed using chi-squared. The overall survival was defined as that from the date of the operation to the date of death due to cancer. The Kaplan-Meier method was used to determine the probability of survival and the data were analyzed by the log-rank test. Multivariate analysis was performed using the Cox regression model to study the effect of different variables on overall survival. The software StatView for Windows version 5 (SAS Institute, Cary, NC) was used for the analysis. A p value of < 0.05 was considered significant.
Discussion
Our study provides several novel insights into the clinical relevance of fascin in colorectal adenocarcinoma. Most significantly, fascin was strongly expressed in similar proportions of adenomas and adenocarcinomas, (16% versus 17% to 26%, respectively), and fascin expression in invasive stage III and IV adenocarcinomas correlated significantly with decreased survival. Given the known roles of fascin in cytoskeletal organization and cell migration, these findings indicate a potential clinical significance of fascin as a marker or prospective therapeutic target for the most aggressive forms of colorectal adenocarcinoma. Our study also clarifies that fascin protein is upregulated in precancerous lesions, of both inherited and sporadic origin. Adenomas are known to be of polyclonal origin [
33,
34], and additional experimental research will be needed to establish whether and how fascin-staining adenomas give rise to strongly fascin-staining carcinomas. If the two groups are indeed related, fascin could have value as a novel prognostic marker, for example to identify those individuals who should receive additional monitoring or treatment after the detection and surgical removal of an adenoma or invasive adenocarcinoma.
A second novel finding of our study was the highly significant correlation of strong fascin expression with tumor location in the proximal colon (Table
3). The proximal and distal colon have different embryonic origins, distinct blood supplies and innervation [
35,
36]. Much evidence supports the concept that tumors of the proximal colon have specific clinical and pathological characteristics and arise through distinct genetic and molecular processes [
35‐
37]. Eighty-seven percent of sporadic tumors with microsatellite instability (MIN or MSI tumors) occur in the proximal colon, whereas tumors with chromosomal instability (CIN tumors) predominate in the distal colon [
36,
37]. It has been suggested that methylating carcinogens are responsible for the origin of MIN tumors in the proximal colon [
38]. In general, MIN tumors have a better prognosis [
39]. Differing responses of CIN or MIN tumors to chemotherapy have been reported [
39,
40]. However, in our dataset, those patients with strongly fascin-positive advanced tumors had poorer survival (Figure
3). From Kaplan-Meier analyses we found that fascin-high, proximal tumors did not correlate with reduced survival (unpublished observation). Fascin-high distal tumors were not represented in large enough numbers in the current dataset to obtain a statistically meaningful comparison for this group alone. Studies of a larger dataset that include assessment of the MIN status of tumors, for example by staining for the
hMLH1 gene product, [
41], will be needed to pursue this interesting correlation in depth.
Our study also revealed that stromal fascin staining was elevated in 47% of colorectal adenocarcinomas. This staining was independent of whether the adenocarcinoma itself was fascin-positive or -negative and tended to be strongest in the stroma adjacent to the invading front of the tumor. We postulate that increased stromal fascin represents an aspect of the host/tumor interaction. Many properties of the stroma, including extracellular matrix composition, density of immune and fibroblastic cells, angiogenesis and the production of angiogenic and chemoattractant factors, are known to be regulated as a result of crosstalk between the tumor and its surrounding host tissue [
6]. The increased stromal fascin staining could reflect either a higher density of fascin-positive cells, most likely fibroblasts, or an increased fascin content per cell. From the Ki67 staining studies, it was clear that elevated stromal cell fascin did not correlate with zones of increased stromal cell proliferation. Further studies are needed to examine whether elevated stromal fascin staining is specific to colorectal adenocarcinomas or is also associated with carcinomas in other tissues.
Previous studies have suggested a possible relationship between fascin expression and increased proliferation of epithelial cells. Over-expression of fascin in colonic epithelial or its depletion in esophageal carcinoma cells correlated, respectively, with increased or decreased cell proliferation in culture [
26,
42]. The most strongly staining fascin-positive non-small cell lung carcinomas tended to be the most highly proliferative tumors, but cells high for Ki67 tended to be low for fascin [
19]. However, the reverse trend was documented in gastric tumors, which had a higher Ki67 index in fascin-positive areas compared to fascin-negative areas [
21]. Our analysis of colorectal adenomas and adenocarcinomas established that, in the adenomas, fascin-positive cells and crypts were clearly less proliferative than fascin-negative crypts (Fig.
1m, 1n). In the adenocarcinomas, distinctions between the fascin and Ki67 staining patterns were less clear-cut, but the overall trend was for fascin-positive cells to be low or negative for Ki67 and
vice versa. Overall, our results do not support the hypothesis that up-regulation of fascin correlates positively with cell proliferation. There is accumulating evidence that migration and proliferation are exclusive behaviors for carcinoma cells and that migrating cells tend to be concentrated at the invading edge of tumors [
43,
44]. Thus, tumors containing a large fraction of fascin-positive cells might have a high potential for invasive behavior. Indeed, in breast carcinomas, fascin has been identified as a component of a gene signature that correlates clinically with tumor metastasis to the lung [
25]. This concept would be in agreement with the correlation of fascin expression with an aggressive subset of advanced colon adenocarcinomas, as we have established from our TMA study.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
YH carried out scoring of sections, conducted the statistical analyses, prepared figures and tables and helped draft the manuscript. MS built the TMA, collected tumor specimens, organized the stainings, scored sections and contributed to the design of the study and the drafting of the manuscript. ICL contributed tumor specimens and in the setupof the CCF IRB-approved database. ALM participated in the preparation of the TMA and setup of the CCF IRB-approved database. GC participated in the preparation of the TMA and setup of the CCF IRB-approved database and contributed to the coordination of the study. JCA designed the study, participated in data analysis and drafted the manuscript. All authors read and approved the final manuscript.