High Expression of PRL-3 Promotes Cancer Cell Motility and Liver Metastasis in Human Colorectal Cancer: A Predictive Molecular Marker of Metachronous Liver and Lung Metastases

Colorectal cancer is the third most common malignant neoplasm worldwide (1) and the second leading cause of death due to cancer in the United States (2). Despite recent advances in diagnostic and therapeutic measures, the prognosis of colorectal cancer patients with distant metastasis still remains poor. In addition, not a few colorectal cancer patients suffer from the unexpected development of occult metastases, especially in the liver and lung, after the curative resection of their primary tumors. Therefore, it is necessary to clarify the molecular mechanism(s) involved in metastasis and to identify the specific biomarkers of colorectal cancer metastasis. To identify the consistent genetic alterations associated with the transition from primary colorectal cancers to liver metastases, Saha et al. (3) performed global gene expression profiles using a serial analysis of gene expression approach and found that PRL-3 (phosphatase of regenerating liver-3/PTP4A3) was frequently overexpressed in the liver metastases studied, but expressed at lower levels in primary tumors and normal colorectal epithelium.

Protein tyrosine phosphatases play a fundamental role in regulating diverse proteins that essentially participate in every aspect of cellular physiologic and pathogenic processes (4). PRL-1, -2, and -3 represent a novel class of protein tyrosine phosphatase superfamily members in that they possess a unique COOH-terminal prenylation motif with a protein tyrosine phosphatase-active site signature sequence CX₅R (5, 6). PRLs were found to be associated with the early endosome and plasma membrane in their prenylated state, whereas nuclear localization of these phosphatases may occur in the absence of prenylation (7). Although the PRLs share 75% amino acid sequence similarity, the ScanProsite analysis revealed that the potential sites of phosphorylation by several kinases are quite different (6, 8). Moreover, Northern blot analysis has demonstrated that the preferential mRNA expression pattern of these PRLs also differed among organs, indicating that PRLs are quite divergent in their functions (6, 9). PRL-1, the founding member of PRL phosphatases, was originally identified as an immediate early gene, the expression of which was induced in mitogen-stimulated cells and in the regenerating liver (10, 11). Overexpression of PRL-1 and PRL-2 has been found to transform mouse fibroblasts and hamster pancreatic epithelial cells in culture and to promote tumor growth in nude mice, suggesting that both of these PRLs may participate in tumorigenesis (5, 8). Similarly, PRL-3 has been found to enhance the growth of human embryonic kidney fibroblasts (9). Although the expression of PRL-1 and PRL-2 has been detected widely in various organs, human PRL-3 is expressed predominantly in the heart, striated muscle cells, and smooth muscle cells, with lower level of expression in the pancreas (9). Zeng et al. (12) demonstrated recently that overexpression of PRL-3 in Chinese hamster ovary cells enhanced the motility and invasive ability of these cells, suggesting that high expression of PRL-3 phosphatase may be one of the key alterations contributing to the metastasis of the transformed cells.

In the current study, we evaluated the role of PRL-3 in human colon cancer DLD-1 cells, especially targeting their proliferation, motility, and hepatic colonization by down-regulating PRL-3 expression with small interfering RNA. We also examined the levels of PRL-3 expression in both primary and metastatic human colorectal cancers and investigated the relationships with clinicopathological features, including patient outcome.

Four human colon cancer cell lines (DLD-1, HCT-15, LoVo, and SW480) were routinely maintained in RPMI 1640 (Invitrogen Co., Carlsbad, CA) supplemented with heat-inactivated 10% (v/v) fetal bovine serum (Invitrogen Co.) at 37°C in a humidified atmosphere of 95% air and 5% CO₂.

A total of 197 patients who underwent surgical resection of primary and/or metastatic colorectal cancer between January 1998 and December 2002 at Kobe University Hospital were investigated. Formalin-fixed and paraffin-embedded specimens from 177 colorectal cancer patients who underwent surgical resection for primary tumors and from 30 colorectal cancer patients who underwent surgery for the resection of metastatic tumors were collected (Tables 1 and 2). Both primary and metastatic tumor specimens were collected from 10 of these patients. In brief, the lesions consisted of 177 primary and 92 metastatic colorectal cancers (lymph node metastases, 59 cases; liver metastases, 23 cases; lung metastases, 6 cases; and peritoneal dissemination, 4 cases). Informed consent was obtained from all of the patients, and no patient received any type of therapy pre- or postsurgery. Histologic classification and clinicopathological staging were performed according to the General Rules for Clinical and Pathological Studies on Cancer of the Colon, Rectum, and Anus (13) along with the classification of the International-Union Against Cancer (14).

For the RNA interference analyses, human PRL-3–specific small interfering RNA (5′-GUGACCUAUGACAAAACGCTT-3′ and 5′-GCGUUUUGUCAUAGGUCACTT-3′) and human PRL-1–specific small interfering RNA (5′-GAUGCAGUUCAGUUUAUAATT-3′ and 5′-UUAUAAACUGAACUGCAUCTT-3′) were designed and synthesized based on the coding sequence of human PRL-3 and PRL-1. Control small interfering RNA targeted Luciferase (Luc-small interfering RNA: 5′-CGUACGCGGAAUACUUCGATT-3′ and 5′-UCGAAGUAUUCCGCGUACGTT-3′), control scrambled small interfering RNA for PRL-3 (scramble small interfering RNA 1: 5′-ACGCUAUAGCUAGAGCAACTT-3′ and 5′-GUUGCUCUAGCUAUAGCGUTT-3′), and control scrambled small interfering RNA for PRL-1 (scramble small interfering RNA 2: 5′-GUCAUUAAGUGUACUAGAUTT-3′ and 5′-AUCUAGUACACUUAAUGACTT-3′) were also synthesized. All of the small interfering RNA sequences were subjected to basic local alignment search tool search to confirm the absence of homology to any additional known coding sequences in the human genome. DLD-1 cells were seeded in 24-well plates (1 × 10⁵ cells/well) in RPMI 1640 plus 10% fetal bovine serum. The following day, each small interfering RNA was added to the culture at a final concentration (0, 0.5, and 5 nmol/L) in 500 μL of RPMI 1640 without antibiotics and serum in the presence of 0.8% Oligofectamine (Invitrogen Co.). After small interfering RNA transfection, cells were incubated in RPMI 1640 plus 1% fetal bovine serum. The medium was renewed 72 and 144 hours afterward. DLD-1 cells treated only with Oligofectamine were also used as a control (mock transfection). We confirmed that the transfection of PRL-small interfering RNAs and these control small interfering RNAs did not affect the levels of β-actin using reverse transcription-PCR (RT-PCR) analyses.

Using the RNeasy Mini kit (Qiagen, Hilden, Germany), each total RNA was isolated from nontreated human colon cancer cell lines (DLD-1, HCT-15, LoVo, and SW480; 1 × 10⁵) and small interfering RNA-transfected DLD-1 cells (1 × 10⁵). Then, quantitative real-time RT-PCR analyses were performed using the ABI PRISM 7700 Sequence Detection System and the QuantiTect SYBR Green RT-PCR kit (Qiagen). Primer sets used for RT-PCR amplification of PRL-3 and PRL-1 were as follows: PRL-3/forward, 5′-GGGACTTCTCAGGTCGTGTC-3′; PRL-3/reverse, 5′-AGCCCCGTACTTCTTCAGGT-3′; PRL-1/forward, 5′-ATGGCTCGAATGAACCGCCCAG-3′; and PRL-1/reverse, 5′-TTATTGAATGCAACAGTTGTTT-3′. As a control, the levels of β-actin expression were also analyzed (β-actin/forward, 5′-CCACGAAACTACCTTCAACTCC-3′; β-actin/reverse, 5′-TCATACTCCTGCTGCTTGCTGATCC-3′). According to the manufacturer’s instructions, a master-mix (50 μL) of the following reaction components was prepared to the indicated end concentration: 25 μL of 2 × QuantiTect SYBR Green RT-PCR Master Mix, 0.5 μmol/L of each forward and reverse primer, 0.5 μL of QuantiTect RT Mix, 10 μL (10 ng) of total RNA, and the proper amount of RNase-free water. After an initial incubation at 50°C for 30 minutes and denaturation at 95°C for 15 minutes, the following cycling conditions (40 cycles) were used: denaturation at 94°C for 15 seconds, annealing at 60°C for 30 seconds, and elongation at 72°C for 1 minute. All of the experiments were performed in triplicate.

To exclude the possibility that the protein kinase R-dependent interferon pathway activated by small interfering RNA transfection induced broad and complicating effects (15), we investigated the expression of the phosphorylated forms of protein kinase R and the protein kinase R substrate eukaryotic inhibition factor 2α in DLD-1 cells treated with or without small interfering RNA. For Western blotting, the cells (1 × 10⁵) were lysed in a buffer containing 50 mmol/L Tris-HCL (pH 7.4), 125 mmol/L NaCl, 0.1% Triton X-100, and 5 mmol/L EDTA containing both 1% protease inhibitor (Sigma, St. Lois, MO) and 1% phosphatase inhibitor mixture II (Sigma). Protein was separated by SDS-PAGE followed by electrotransfer. Anti-protein kinase R, phospho-protein kinase R^Thr446, phospho-protein kinase RR^Thr451, eukaryotic inhibition factor 2α, and phospho-eukaryotic inhibition factor 2α^Ser51 polyclonal antibodies (1:1000 dilution; Cell Signaling, Beverly, MA) were used in the primary reaction. Horseradish peroxidase-conjugated goat antirabbit IgG (Jackson ImmunoResearch Laboratories Inc., West Grove, PA) was used as a secondary antibody.

Cell growth and survival in the presence or absence of each PRL-small interfering RNA transfection were determined using the Premix WST-1 Cell Proliferation Assay System (Takara Biochemicals, Tokyo, Japan) as described elsewhere (16). Forty-eight hours after small interfering RNA transfection, an aliquot of 1 × 10⁵ cells (100 μL volume/well) were inoculated to triplicate wells and maintained in phenol red-free medium for 48 hours. After incubation, 10 μL of Premix WST-1 was added to each microculture well, and the plates were incubated for 30 minutes at 37°C, after which absorbance at 450 nm was measured using a microplate reader. The absorbance in the cells without small interfering RNA transfection (1.272 ΔOD) was considered to be 100%.

Cell motility and invasive activity were estimated using Transwells (6.5 mm in diameter; polycarbonate membrane, 8 μm pore size) coated with extracellular matrix gel obtained from Chemicon (Temecula, CA). Forty-eight hours after small interfering RNA transfection, an aliquot of 1 × 10⁵ cells was placed in the upper chamber with 0.5 mL serum-free medium, whereas the lower chamber (24-well plate) was loaded with 1 mL of medium containing 10% fetal bovine serum. After 48 hours of incubation at 37°C with 5% CO₂, the cells were fixed with 4% paraformaldehyde and then counterstained with hematoxylin. The cells that had migrated into the lower chamber were observed and counted under a light microscope.

The protocol was approved by the Kobe University Health Sciences Animal Care Committee and Japanese Governmental Law 105. Eight-week–old BALB/cA Jcl-nu female mice (housed 5 per cage) were used in this study. Mice were locally injected to the spleen with 3 × 10⁵ viable DLD-1 cells treated under each condition (nontreated group: n = 5; PRL-3–small interfering RNA group: n = 5; and PRL-1–small interfering RNA group: n = 5), and the mice were sacrificed under anesthesia on day 30. Liver and spleen tissues were fixed in 10% buffered formalin (pH 7.4) and processed for routine histology. The number and diameter of metastatic foci in 5 sections per liver was determined, and the volume (V) of these foci was calculated using the equation V = 1/2 × A × B², where A and B indicate long and short diameters, respectively.

The specific antisense oligonucleotide DNA probe for PRL-3 (5′-GTTGATGGCTCC GCGGCG-3′) was designed complementary to the mRNA transcripts of the PRL-3 gene according to the GenBank database. The specificity of the oligonucleotide sequences was initially determined by a Gen-EMBL database search using the FastA algorithm (17), which showed minimal homology with the PRL-1 and PRL-2 genes and other nonspecific mammalian gene sequences. All of the probes were synthesized with six biotin molecules (hyperbiotinylated) at the 3′ end via direct coupling using standard phosphormidine chemistry (18, 19). We then used multibiotinylated poly(dT)₂₀ oligonucleotides (Invitrogen Co.) to verify the integrity of mRNA in each sample. The lyophilized probes were reconstituted to a 1 μg/μl stock solution in 10 mmol/L Tris-HCl (pH 7.6) and 1 mmol/L EDTA. The stock solution was diluted with Brigati Probe Diluent (Invitrogen Co.) immediately before use.

In situ hybridization was performed using manual capillary action technology (20) on the Microprobe Staining System (Fisher Scientific, Pittsburgh, PA). ProbeOn Plus slides (Fisher Scientific) were placed in the MicroProbe slide holder so as to make a 150 μm gap, and all of the subsequent reagents were placed onto and drained from the slides by capillary action. The tissue sections were dewaxed with xylene and rehydrated with 1 × Tris-buffered saline-Tween 20. The tissue sections were digested with a stable pepsin solution (DAKO, Carpinteria, CA), which was used at full strength for 3 minutes at 100°C. Hybridization of the probes was carried out for 80 minutes at 60°C, and the samples were then washed three times with 2 × SSC for 2 minutes at 45°C. The samples were incubated with alkaline phosphatase-labeled avidin (Biomeda, Foster City, CA) for 30 minutes at 45°C, briefly rinsed in 20 × Tris-buffered saline-Tween 20, rinsed with alkaline phosphatase enhancer (Invitrogen Co.) for 1 minute, and finally incubated with the chromogen substrate Fast Red (Biomeda) for 20 minutes at 37°C. Hybridization of the samples with biotinylated poly(dT)₂₀ probes was always performed to verify the integrity of mRNA. To analyze the specificity of the hybridization signal, we performed RNase pretreatment of the tissue sections and competition assays with an unlabeled antisense probe as control procedures. Then, as controls for endogenous alkaline phosphatase activity, we performed treatment of the samples in the absence of the biotinylated probes and the use of chromogen in the absence of any probes. We confirmed that no signal was detected under any of these conditions.

For the evaluation of in situ hybridization reactivity, normal colorectal epithelium tissue, smooth muscle cells of the vessel, and lymphocytes were used as internal controls. The reactivity of in situ hybridization was graded as follows: high PRL-3 expression, > 10% cancer cells showed PRL-3 expression exceeding that of the internal controls; and low PRL-3 expression, > 90% cancer cells showed no increase in the expression of PRL-3 compared with the internal controls. The levels of in situ hybridization reactivity were evaluated independently by three pathologists (H. K., S. S., and U. A. M.).

Among our cohort of 150 colorectal cancer patients clinically diagnosed as free of distant metastases at the time of curative resection of their primary tumors, 104 had complete follow-up information. For these 104 patients, we evaluated the associations between clinicopathological features, including the levels of PRL-3 expression in the primary tumors and the incidence of metachronous liver and/or lung metastasis after curative surgery for primary tumors. All of the patients underwent imaging examinations (computed tomography and ultrasonography) at regular intervals. The median follow-up period was 2.8 years (range, 5 months to 6.1 years).

The results of the in vitro assays, in vivo metastasis assay, and the in situ hybridization studies were investigated by χ² test. The time to the appearance of metachronous metastasis in cases with high and low expression of PRL-3 after surgical resection for primary tumor was compared by performing a Kaplan-Meier analysis and testing the results with the log-rank statistic. We also evaluated the association between the incidence of metachronous metastasis and the conventional indicators, such as primary tumor size, angiolymphatic invasion, and the presence of lymph node metastasis by Kaplan-Meier analyses. A P < 0.05 was regarded as statistically significant.

We first examined the expression of PRL-3 and PRL-1 in 4 human colon cancer cell lines and normal colonic epithelium by quantitative real-time RT-PCR analyses. Although PRL-3 and PRL-1 were expressed in all of the cell lines, the expression levels varied significantly. The DLD-1 cells demonstrated the highest level of PRL-3 expression, which was equivalent to about three times that of the normal colonic epithelium, whereas the LoVo and SW480 cells showed low levels of PRL-3 expression that were almost equal to that of the normal colonic epithelium (Fig. 1,A). As shown in Fig. 1 B, DLD-1 cells also demonstrated high PRL-1 expression equivalent to about twice that of the normal colonic epithelium, and therefore we performed the following RNA interference studies using the DLD-1 cells.

Transfection of PRL-3 and PRL-1–small interfering RNAs into DLD-1 cells down-regulated the level of the expression of each type of PRL in a concentration-dependent manner (Fig. 2, A and B). After the transfection of PRL-3– and PRL-1–small interfering RNAs (5 nmol/L), the expression of each PRL reached a minimal level at 48 to 96 hours, which gradually recovered to almost the baseline level at 168 hours (Fig. 2, C and D). According to these results, we performed Western blotting analysis, WST-1 cell proliferation assay, cell motility/invasion assay, and hepatic metastasis assay experiments using DLD-1 cells 48 hours after small interfering RNA (5 nmol/L) transfection. Western blotting analysis showed that the transfection of each small interfering RNA (5 nmol/L) had no influence on the levels of expression of the phosphorylated forms of protein kinase R and eukaryotic inhibition factor 2α (Fig. 2,E), thereby demonstrating that the exogenously transfected small interfering RNAs (5 nmol/L) did not activate the protein kinase R-dependent interferon pathway in the DLD-1 cells. In contrast, the DLD-1 cells treated with 0.1 μmol/L Calyculin A (serine/threonine phosphatase inhibitor; Cell Signaling) demonstrated quite high levels of expression of the phosphorylated forms of protein kinase R and eukaryotic inhibition factor 2α (Fig. 2 E).

As a starting point of our attempt to determine the significance of PRL-3 in the progression and development of human colorectal cancer, we evaluated the roles of PRL-3 in cell proliferation and motility/invasion using an RNA interference technique. First, we attempted to analyze the effect of PRL-3 and PRL-1 on cell growth in vitro. Transfection of these PRL-small interfering RNAs did not alter the cell growth of DLD-1 cells, as compared with the nontreated control cells (PRL-3–small interfering RNA: 100% and PRL-1–small interfering RNA: 98%; Fig. 3,A). We next focused on determining the roles of PRL-3 and PRL-1 on cell migration and invasive activity in vitro. As shown in Fig. 3,B, the number of DLD-1 cells that had migrated into the lower chamber was statistically smaller in the group treated with these PRL-small interfering RNAs than in the nontreated group (PRL-3–small interfering RNA: 32% and PRL-1–small interfering RNA: 53%). Interestingly, remarkable structural alterations were observed in DLD-1 cells treated with PRL-3–small interfering RNA: the cells had a compact shape and lacked ruffles, protrusions, or other membrane processes on their surfaces (Fig. 3,C). However, DLD-1 cells treated with PRL-1–small interfering RNA did not demonstrate such morphologic changes (Fig. 3 C).

To investigate whether decreased levels of PRL-3 expression suppress the establishment of colorectal cancer liver metastasis, we used hepatic metastasis model mice. DLD-1 cells (3 × 10⁵) were maintained for 48 hours after small interfering RNA transfection and were then locally injected to the spleen. After the injection of tumor cells with or without small interfering RNA treatment, all of the mice survived until sacrifice on day 30. All of the mice used in this analysis developed localized splenic tumors that were almost equal in size, whereas the incidence of liver metastasis varied in each group; although all of the mice in the nontreated group (5 of 5, 100%) demonstrated the formation of metastatic foci in the liver (Fig. 4,A) with an average number (volume) of 11.0 foci per liver (201.3 μm³; Fig. 4, B and C), the incidence of liver metastasis was significantly decreased in both groups treated with PRL-3– (1 of 5, 20%) and PRL-1– (2 of 5, 40%) small interfering RNAs (Fig. 4,A). Moreover, in these groups the average number (volume) of metastatic foci was significantly smaller than that in the nontreated group: the number volume in the PRL-3–small interfering RNA group was 0.2 foci per liver (7.9 μm³) and the number volume in the PRL-1–small interfering RNA group was 4.0 foci per liver (1.6 μm³; Fig. 4, B and C). By in situ hybridization methods, we examined the expression of PRL-3 in all of the splenic (data not shown) and hepatic (Fig. 4 A) tumors and confirmed that the DLD-1 cells forming the tumors demonstrated high expression of PRL-3, irrespective of treatment with PRL-small interfering RNAs. In this experiment, formation of metastasis was not observed in any organs except the liver.

We next investigated the levels of PRL-3 expression in human colorectal cancers using in situ hybridization methods. Weak PRL-3 expression was detected in every sample of normal colorectal epithelium analyzed, and the levels were almost the same in each sample (Fig. 5). However, in the primary colorectal cancer cells, PRL-3 expression varied significantly among cases and was distributed heterogeneously. The results are summarized in Table 3. In general, a high level of PRL-3 expression was observed in 79 (44.6%) of 177 primary colorectal cancers. The frequency of high expression of PRL-3 in cases with distant metastasis (liver: 84.4% and lung: 88.9%) was statistically higher than that in the cases without distant metastasis [liver: 35.9% (P < 0.001) and lung: 42.3% (P = 0.006)]. Simultaneously, high expression of PRL-3 was detected more frequently in the cases with venous invasion (52.2%) than in the cases without venous invasion (31.2%, P = 0.007). Cancer cells forming intravenous tumor emboli showed strikingly increased PRL-3 expression (Fig. 6 A). However, the levels of PRL-3 expression showed no correlation with any other clinicopathological features such as depth of invasion, tumor size, lymphatic invasion, or the presence of lymph node metastasis.

In metastatic colorectal cancer lesions, high expression of PRL-3 was found in 21 of 23 (91.3%) cases of liver metastasis and in 6 of 6 (100%) cases of lung metastasis (Table 4). However, in lymph node metastasis and peritoneal dissemination, high expression of PRL-3 was observed in only 47.5% and 50.0% of the total number of cases, respectively. In liver and lung metastatic lesions, almost all of the colorectal cancer cells homogeneously demonstrated high expression of PRL-3. Regarding the 10 cases for which serial analyses of both primary and metastatic tumors could be performed, all of the specimens showed high expression of PRL-3 (Fig. 6, C–F).

Finally, we performed a prognosis study to determine whether PRL-3 could be used as a biomolecular marker to monitor the risk of metachronous metastasis after curative surgery for primary colorectal cancer. Overall, postoperative development of occult liver and/or lung metastasis appeared in 14 of 104 cases (13.5%). As shown in Fig. 7, Kaplan-Meier analysis showed that cases with high PRL-3 expression had a greater risk for the development of metachronous metastasis than those with low PRL-3 expression (P < 0.0001). However, the incidence of metachronous metastasis was not statistically related to the conventional indicators such as primary tumor size, angiolymphatic invasion, and the presence of lymph node metastasis (data not shown).

Using serial analysis of gene expression technology, which is a powerful strategy for the detection of altered gene expression, PRL-3 was recognized as the most important molecule that was consistently and specifically activated in liver metastases of human colorectal cancers (3). As a member of the PRL family of phosphatases, PRL-3 has a catalytic active signature motif (C104S; refs. 9, 21). Stable expression of wild-type active PRL-3 dramatically enhanced cell migration, whereas the catalytically inactive PRL-3 (C104S) mutant greatly reduced the promotion of cell migration (12). These results indicated that the ability of PRL-3 to promote cell migration depended on its phosphatase activity. Protein prenylation is important in targeting proteins to intracellular membranes and in protein-protein interactions (22, 23). PRL-3 has been reported to be a member of the prenylated protein phosphatase family, and the metastatic properties of PRL-3, like those of PRL-1, are dependent on its prenylation activity (7, 12). In the current study, we investigated the roles of PRL-3 and PRL-1 in human colon cancer DLD-1 cells, targeting in particular their metastasis-related activities. Decreased levels of the expression of these PRLs suppressed cell motility/invasiveness, especially when the cells were treated with PRL-3–small interfering RNA. However, they had no influence on cell proliferation. To our knowledge, this is the first report to study the roles of PRL-3 and PRL-1 in a human colon cancer cell line by inhibiting endogenous PRL-3 and PRL-1 expression using an RNA interference technique.

Metastasis consists of a series of sequential steps, all of which must be successfully completed. These include the shedding of cells from a primary tumor into the circulation, survival of the cells in the circulation, arrest in a new organ, extravasation into the surrounding tissue, initiation and maintenance of growth, and vascularization of the metastatic tumor (24). Zeng et al. (12) exhibited that Chinese hamster ovary cells exogenously expressing PRL-3 and PRL-1 induced metastatic tumor formation in vivo. In our study, down-regulation of the expression of these PRLs in DLD-1 cells suppressed metastatic tumor formation in vivo. These results indicate that both PRL-3 and PRL-1 can regulate not only cell motility but also the formation of metastatic lesions. However, the suppression of endogenously expressed PRLs by this RNA interference technique is a transient suppression. As expected, irrespective of treatment with PRL-3–small interfering RNA, PRL-3 expression in the DLD-1 cells forming splenic and hepatic tumors was re-up–regulated 32 days after the treatment. This phenomenon suggests that PRL-3 may contribute to the establishment of colorectal cancer liver metastasis, especially at the step in which cancer cells leave the circulation to extravasate into the liver tissue.

PRL-3 was located at the cytoplasmic membrane and in the early endosome when prenylated and was shifted into the nucleus when unprenylated or lacking the COOH-terminal prenylation signal (7). Chinese hamster ovary cells exogenously expressing PRL-3 were enriched in several membrane processes, including protrusions, ruffles, and some vacuolar-like membrane extensions, which have been reported to play a role in cell motility and invasion (12). Interestingly, DLD-1 cells treated with PRL-3–small interfering RNA demonstrated morphologic alterations, showing compact cytoplasm but not wide processes and ruffles. However, such morphologic alterations were not observed in cells treated with PRL-1–small interfering RNA. It has been considered that PRL-3 may promote cell motility and metastatic tumor formation more effectively than PRL-1 (12). Although we did not confirm the localization of the PRL-3 protein in these cultured cells, several studies have reported that PRL-3 was localized at the plasma membrane of the foot processes and assisted with the cellular motility machinery (7, 12), findings that may be related to the functional differences between PRL-3 and PRL-1. PRL-3 may play a key role in the cytoskeletal remodeling that is required for cancer cell motility. However, the signal transduction pathways and the cytoskeletal alterations associated with PRL-3 are largely unknown. Additional investigations are required to clarify the mechanism in which PRL-3 controls cell motility/invasiveness.

In human tissue samples, increased levels of PRL-3 expression were significantly correlated with liver and lung metastases. Although cancer cells showing high expression of PRL-3 were detected heterogeneously in primary tumors, most of the metastatic tumor cells demonstrated high expression of PRL-3 homogeneously. Moreover, in the primary tumors, high expression of PRL-3 was significantly correlated with venous invasion. These results indicate that PRL-3 may also contribute to the establishment of colorectal cancer metastasis at the step in which cancer cells intravasate into venules at the primary site. Using in situ hybridization methods, we found that PRL-3 was expressed not only in the colorectal epithelium but also in the smooth muscle cells of vessels and lymphocytes. Due to such contamination, it would be inappropriate to evaluate the levels of PRL-3 expression in the total RNA isolated from resected colorectal cancer tissue using any of the standard methods. In situ hybridization analysis has the advantage of avoiding such issues. Using in situ hybridization methods, Bardelli et al. (25) indicated recently that PRL-3 was expressed in colorectal cancer metastatic lesions but not in the normal colorectal epithelium, nonmetastatic colorectal cancer, or gastric cancer. In the present study, on the other hand, we observed PRL-3 expression in the normal colorectal epithelium and in nonmetastatic colorectal cancer tissue. This inconsistency may have been due to differences in the number of cases studied, the criteria for the evaluation of in situ hybridization signals, or the specificity of the antisense DNA probe for PRL-3. However, our in situ hybridization analyses enabled the detection of slight differences in PRL-3 expression among primary colorectal cancer cases, and we also found PRL-3 expression in cases of gastric cancers using RT-PCR and in situ hybridization analyses (data not shown).

Unexpectedly, colorectal cancer patients develop metachronous liver (8.9%) and/or lung (6.0%) metastasis after curative surgery for the primary tumors (26). Thus, it would be of great value to identify a promising biomarker for the metachronous metastasis of colorectal cancer. In addition to our in situ hybridization analyses of colorectal cancer tumor samples, our prognosis study suggested that the level of PRL-3 expression in the primary colorectal cancer lesion is a more promising predictor of the postoperative development of metachronous liver and/or lung metastasis than such conventional predictors as tumor size, angiolymphatic invasion, or the presence of lymph node metastasis. High expression of PRL-3 in surgical and biopsied colorectal cancer specimens may provide clinicians useful information not only for identifying occult metastases but also for initiating adjuvant chemotherapy after the surgical treatment of the primary tumor. Moreover, PRL-3 may provide a novel therapeutic target for intractable colorectal cancer metastasis. Although the specific protein substrate for PRL-3 has not yet been identified, the function of PRL-3 in metastasis could be blocked or reduced by inhibiting prenylation and/or inactivating the catalytic function of the PRL-3 phosphatase active site (12). Additional investigations will be necessary to clarify the role(s) of PRL-3 in the process of colorectal cancer metastasis and to develop inhibitors against PRL-3 itself.

The authors would like to thank Dr. Hiroki Kuniyasu (Department of Oncological Pathology, Nara Medical University) for his valuable input, and Ms. Akiko Obata for her technical assistance.