In our previous survey of human epithelial cancers, we graded the expression of Trask in about 90 carcinomas of the breast, colon, and lung as well as more than 30 surgical or biopsy specimens of normal breast, colon, and lung tissues (Wong et al 2009
). This analysis, which included pre-invasive tumors as well as primary invasive tumors and tumor metastases did not reveal an increase in Trask expression in cancers, although phosphorylation of Trask was often seen in tumors. In fact some tumors appeared to have a Trask expression score that was lower than normal tissues. The normal tissue samples and the tumor tissue samples were from different patients in these studies. To more accurately and directly determine whether Trask expression is reduced or lost in some tumors we undertook to compare the expression of Trask in a series of breast and colon cancers compared with their adjacent normal epithelial tissue counterparts. We specifically looked for specimens in which both normal epithelium and cancer can be seen on the same slide so as to minimize variations in immunostaining intensity. In a survey of breast cancer tissues there is much variation in the expression of Trask. Some cancers have preserved Trask expression compared to the normal ductal epithelium, some cancers have reduced Trask expression, and some cancer have lost Trask expression (). In a survey of colon cancer tissues there is also variation in the expression of Trask with preservation of expression in some cases, and a reduction in expression in many cases ( and supplementary figure S1
). The expression of Trask in tumors is often heterogeneous with a patchy distribution. We do not see an overexpression of Trask in these cancers.
Comparative expression of Trask in normal and malignant human epithelium
We have also determined the expression of Trask in a panel of epithelial cancer cell lines by western blotting. There is wide variation in the expression of Trask, including some cancer cells with very low or no expression (). None of the cancer cell lines show Trask expression that is significantly greater than non-cancer cells such as MCF10A or HaCaT cells. We also considered that some cancers may have impaired Trask function rather than reduced expression. Although we don’t yet fully understand all the functions of Trask, we know that Trask is phosphorylated by Src kinases during anchorage deprivation and p-Trask functions in reciprocity with focal adhesion signaling and functions to inhibit integrin clustering and cell adhesion (Spassov et al 2009
, Spassov et al 2011b
). This appears to be a general attribute of epithelial cells and can be experimentally induced in cultured cells or in the mouse epidermis, or seen in detached epithelial cells in vivo (Spassov et al 2009
, Spassov et al 2011b
). Therefore we sought to determine whether the detachment-induced phosphorylation of Trask is preserved in all cancer cells. In our analysis of a panel of epithelial cancer cell lines we identified cancer cell lines that do express Trask, but fail to phosphorylate it when detached (, see MDA-361, MDA-453, T47D, BT474, ZR75-1, SkOv3). RT-PCR amplification and sequencing of the Trask mRNA showed no mutations associated with Trask in these cells. These data from our comparative immunohistochemical survey of tissues and from the cell line panel suggest that Trask may contribute to tumor progression through a reduction or loss of function.
Expression of Trask in human epithelial cancer cell lines
In order to experimentally interrogate the role of Trask in tumor growth and progression we re-expressed Trask in a tumor cell line that lacks its expression. MCF-7 cells have no expression of Trask protein () and no expression of Trask mRNA (). This is likely due to methylation silencing of the Trask promoter region as determined by southern blot analysis of the Trask promoter region using methylation-sensitive restriction enzymes (). The promoter region of Trask is dense with many CpG repeats and its methylation silencing in cancer cells has been previously shown (Ikeda et al 2006
). Treatment of MCF-7 cells with 5-azacytidine induces the re-expression of Trask, confirming that its silencing in these cells is mediated through genome methylation (). MCF-7 cells were engineered to express the luciferase gene to aid with in vivo
imaging, and also engineered to express Trask in a tet-inducible fashion. When Trask is induced to express in MCF-7/Luc/TR/Trask cells, it is constitutively phosphorylated, similar to its overexpression in other cancer cell lines () (Spassov et al 2011b
). This may be due to the high activity of Src kinases in these tumor cells and/or the saturation of dephosphorylation mechanisms. When grown as orthotopically implanted tumors in mice, Trask expression can be induced in the MCF-7/Luc/TR/Trask tumors in vivo
by administration of doxycycline to the mice (). The expression of Trask in MCF-7/Luc/TR/Trask tumors in vivo
has no significant effect on tumor growth (). To determine whether tumor metastasis is affected by Trask expression mice were sacrificed at 7 weeks of tumor growth post-implantation, and the development of tumor metastases was assessed by necropsy analysis assisted by ex-vivo bioluminescence imaging. Metastases were identified at necropsy in a variety of organs, including lungs, liver, bone, muscle, lymph nodes and spleen (supplementary figure S3
). Quantifying the number of mice with or without mets shows a reduction due to Trask expression such that 35.7% of the control mice had detectable metastases whereas only 12.5% of doxycycline-treated mice had detectable metastases (). The difference was evident across all organs. Quantifying the number of metastases per mouse also shows a significant decrease in metastatic burden due to Trask expression with the doxycycline-fed mice having 4 times less metastatic disease (p=0.03; supplementary figure S4
The induction of Trask expression in MCF-7 breast cancer cells
To further interrogate the role of Trask in tumor growth and progression we conducted a loss-of-function experiment in vivo
. L3.6pl pancreatic cancer cells have expression of Trask and proper phosphorylation of Trask when anchorage deprived or when grown as orthotopically implanted tumors in the mouse pancreas (). L3.6pl cells were engineered to express the firefly luciferase gene as well as a control non-silencing shRNA (L3.6pl/Luc/shControl) or either of two Trask shRNA sequences (L3.6pl/Luc/shTrask-1 and -2) leading to near-complete knockdown of Trask protein expression (). The engineered L3.6pl/Luc/shTrask cells and controls were grown as orthotopically implanted tumors in the mouse pancreas and tumor growth rate and the development of tumor metastases was monitored by bioluminescence imaging. The L3.6pl/Luc/shTrask tumors showed slightly slower tumor growth rate compared with the L3.6pl/Luc/shControl tumors (). Mice were sacrificed at six weeks and the development of tumor metastases was assessed on necropsy analysis including ex-vivo bioluminescence imaging. At necropsy, metastases were frequently seen through peritoneal dissemination, as well as in the lungs and liver ( and supplementary figure S5
). Trask knockdown tumors had an increased chance of metastasizing to the lung or liver, although not statistically significant (). But peritoneal dissemination was significantly increased in the Trask knockdown tumors. Peritoneal metastases was quantitatively assessed by bioluminescence imaging of the body cavity after removal of the primary tumor by pancreatectomy. Mice bearing tumors with Trask knockdown had significantly increased peritoneal metastatic disease compared with control mice (). To further interrogate the metastatic potential of this tumor model, L3.6pl/Luc/shControl and L3.6pl/Luc/shTrask-1 tumor cells were introduced into the systemic circulation of mice by tail-vein injection and the development of systemic metastasis was assessed by weekly in vivo
bioluminescence imaging. The development of tumor metastases was significantly accelerated in the Trask knockdown tumors compared with controls (, p=0.02 by chi square test).
The knockdown of Trask expression in L3.6 pancreatic cancer cells
Metastases in L3.6pl-Luc-shRNA tumors
Since the data suggests that the functions of Trask may be tumor suppressing, we sought additional evidence in a third experimental model of in vivo tumor metastasis. v-src transformed fibroblasts are highly metastatic in a tail-vein injection model and have low expression of Trask. We engineered 3T3v-src cells to express firefly luciferase as well as myc-tagged Trask in a doxycycline-inducible fashion (). The overexpression of Trask in these cells leads to abundant tyrosine phosphorylation of Trask similar to other cancer cells with active Src kinases (). 3T3v-src/Luc/TR/Trask cells were introduced into the systemic circulation of mice by tail-vein injection in two arms consisting of mice treated with doxycycline or control. Doxycycline treatment was initiated the day prior to injection. Mice were monitored for the development of metastases weekly by in vivo bioluminescence imaging. Control mice develop metastases with high frequency and short latency, and the induction of tumor Trask expression is associated with a significant reduction in tumor metastastic burden (). Both the rate of development and the burden of tumor metastasis is significantly reduced by the induction of Trask expression ().
The induction of Trask overexpression in v-src transformed cells
These experiments, in three in vivo
models, show that the functions of Trask are negatively associated with tumor progression, a function consistent with a tumor suppressing function. We recently described that Trask, when phosphorylated, functions to inhibit integrin clustering and outside-in integrin signaling (Spassov et al 2011b
). Since integrin signaling is important in tumor progression and metastasis, we looked to determine whether the induction of Trask overxpression in our experimental models is also associated with the suppression of integrin outside-in signaling. In both the MCF-7/Luc/TR/Trask cells and in the 3T3v-src
/Luc/TR/Trask cells, the induction of Trask expression and phosphorylation by doxycycline leads to the dephosphorylation of FAK, direct evidence of the inhibition of integrin outside-in signaling ().
The inhibition of integrin signaling by Trask overexpression