Matrix Metalloproteinase 2 Is A Direct Target of The RAN-GTP Pathway And Mediated The Invasion, Migration And Metastasis of Human Breast Cancer Cells

Background: Ras-related nuclear protein (RAN) causses increases in invasion in vitro and is associated with early breast cancer patients deaths in vivo. However, the underlying mechanism is unknown. Methods: Effect of RAN expression on potential targets MMP2, ATF3, CXCR3 was measured by Real-Time PCR / Western blots. Effects of MMP2 and RAN expression on cancer cell lines (e.g. MDA-MB231) was mearured by soft-agar, cell adhesion, and invasion assays. Correlation between MMP2 and RAN and patient survival times was examined in breast cancer patients. Results: Knockdown of RAN lead toreduction of MMP2 and its potential regulators ATF3 and CXCR3 in breast cancer cell lines. Knockdown of ATF3 or CXCR3 downregulates MMP2 without affecting RAN, indicating that RAN regulates MMP2 through ATF3 and CXCR3. Both knockdown of RAN and MMP2 reduced cell adhesion, migration and growth in agar invasion whilst overexpression of MMP2 reversed the knockdown of RAN. Moreover, the level of immunoreactive RAN and MMP2 are positively associated with each other and with patient survival times, respectively in breast cancer specimens, suggesting that a high level of RAN may be a pre-requisite for MMP2 overexpression. Conclusions: Our results suggest that MMP2 expression can stratify progression of breast cancers with a high and low incidence of RAN and both RAN and MMP2 in combination can be used for accurate patient stratication of breast cancer metastasis. a matrix metalloproteinase (MMP), MMP2, and its upstream regulators, activating transcription factor 3 (ATF3) (4–10) and chemokine receptor 3 (CXCR3) (11, 12) that are targets for the RAN-induced increasese in metastatic-related properties in vitro and show a signication association between RAN, MMP2 and survival time of breast cancer patients in vivo. a signicant decrease in cell adhesion by nearly 5-fold in invasion by 4-fold compared scrambled shRNA transfected overexpressed MDA-MB231-shRAN there signicant cell adhesion cancer cell migration and invasion by upregulating MMP2 expression (21). Our results are consistent with these reports. et Classical Hodgkin lymphoma is characterized by high constitutive expression of activating transcription factor 3 (ATF3), which promotes viability of Hodgkin/Reed-Sternberg cells.


Western blot
Total protein was extracted using N-Per kit from Pierce to visualise the antibodies (SupplementaryTable 4). Data are the mean of three independent experiments ± SD. All results were normalized using a housekeeping gene, β-actin and quanti ed using densitometry readings.

Cell adhesion assay
For the cell the adhesion assay, 40000 cells/well in normal medium were seeded in a 96-well plate coated with bronectin and allowed to settle for 30 min. Suspended cells were removed by washing 4 times with PBS. Adhered cells were xed and stained with crystal violet. The excess dye was washed out and the retained dye was extracted. The absorbance at 595nm was measured in a microplate reader. Data are the mean of three independent experiments ± SD.

Boyden chamber migration and invasion assays
Migration and invasion assays were performed as previously described (18). Brie y, 5000 and 50000 cells in serum-free conditions were seeded into the upper Boyden chamber (Millipore) on top of the membrane with or without a Matrigel coating, respectively, for migration and invasion assays. The cells were allowed to migrate/invade towards the underside for 24 hours with 10 ng/ml HGF as a chemoattractant. Cells on the underside of the membrane were xed and stained with crystal violet solution. Data are the mean of three independent experiments ± SD.

Soft agar assay
Soft agar assay was performed as previously described (19). 5000 suspension cells in normal medium containing 0.35% (w/v) lowmelting-point agarose were overlaid onto a solidi ed normal medium containing 0.7% (w/v) low-melting-point agarose. Cells were incubated at 37 o C with 5% (v/v) carbon dioxide for 2 to 3 weeks. Colonies were visualized by staining with crystal violet and counted.
Data are the mean of three independent experiments ± SD.
Patients and specimens. A retrospective study was undertaken using samples of 181 primary tumours from unselected breast cancer patients as described previously ( . Patients who died from causes other than cancer were censored. Unadjusted relative risk (RR) for survival with 95% con dence interval (95% CI) was calculated using Cox's univariate analysis (Rudland et al, 2010). Association of IHC staining for RAN or MMP2 with other tumour variables was assessed by cross-tabulations using Fishers Exact test (2-sided) using either 1% or 5% cut-offs. For multiple comparisons the resultant P values were corrected by the Holm-Bonferroni formulae of I-(I-P) n , where n is the number of tumour variables. Binary Logistic Regression was used for calculation of the relative independent association (RA) of staining for one protein with the remaining proteins in the group. To determine if the association of patient survival with RAN, MMP2 etc. was signi cant within a group of proteins, Cox's multivariate analyses were performed on 181 patients, incomplete data arose mainly from lack of sampling (de Silva Rudland et al, 2011). Data analysis was performed using Excel (Microsoft, Redmond, WA), and SPSS version 22 (SPSS, Chicago, IL). The sensitivity and speci city, positive predictive response (PPR) and negative predictive response (NPP) were calculated and compared between 1% and 5% of RAN and MMP-2 cut-offs, respectively.

Statistical Analysis
Statistical analysis was performed using SPSS 19.0 software (IBM, Armonk, NY). Differences between groups in in vitro experiments were tested by Student's t test (two groups) and analysis of variance (ANOVA) with post hoc Games-Howell. Differences in expression levels between groups/ samples in the human specimens were analyzed by Chi 2 , Fisher exact test, or Mann-Whitney U tests, where applicable. The association between the expression level and patient survival was recorded by Kaplan-Meier plots and compared by Wilcoxon-Gehan tests. A P value of less than 0.05 was considered statistically signi cant. All statistical tests were two-sided.

Results
Knockdown of RAN results in the downregulation of MMP2 in cancer cell lines Previously, we have shown that knockdown of RAN using potent shRNA results in apoptosis and changes of cell properties including cell adhesion, migration and invasion (18,19). Using these shRNAs, we now have found that silencing RAN by shRNA speci c for RAN

Reduction of MMP2 expression by RAN knockdown reduces cell adhesion and invasion
In a previous study MDA-MB231-shRAN cells resulted in a signi cant decrease in cell adhesion and colony formation compared to MDA-MB231-shScr control cells (17)(18)(19). In this study MDA-MB231-shMMP2 cells produced a signi cant decrease in MMP2 by 2.7 fold We then further investigated the importance of MMP2 in two more breast cancer cell lines, MCF-7 and T47D was investigated.
Knockdown of MMP2 in these two breast cancer cell lines ( Fig. 3A and C, P < 0.0001) resulted in a statistically signi cant decrease in their ability to produce MMP2 by 3.9-fold and by 2-fold, respectively, and to form colonies in soft agar by 6.6 and 7.1 fold, respectively ( Fig. 3B and D, P < 0.05). These results suggest that expression of MMP2 is important for these breast cancer cell lines to maintain their tumorigenic properties in vitro. RAN knockdown reduced the number of colonies growing in soft agar for MCF-7-shRAN cells compared to vector alone shScr controls by nearly 40% (Fig. 3E, P < 0.0001 ). When the MCF-7-shRan cells were overexpressing MMP2 (Fig. 3E), there was no signi cant change in cell adhesion and in the number of colonies in soft agar compared to the vector alone controls ( Fig. 3F, p > 0.5). This result showed that knockdown of RAN led to reduced colony formation, but this reduction could be overcome when the levels of MMP2 were raised. Collectively, our results suggest that RAN precedes MMP2 in a signalling pathway which controls, in part, the invasive properties of breast cancer cells.
Association of RAN-related molecules with patient survival times in human breast cancer Next, we investigated the relationship in human breast cancer between RAN and other potential prognostic molecules/markers with patient demise probably as a result of metastasis. To simplify subsequent statistical analyses, the individual carcinoma cell staining groups for each molecule were separated into two categorical groups using previously-determined cut-offs of 1% or 5% to separate  f Relative Association (RA) and 95% con dence interval (95%CI) from binary Logistic Regression Analysis.

Association of RAN and target molecules in primary breast tumours
Results of IHC staining in primary tumours for RAN and its relationship with that of other molecular tumour markers showed that RAN was very signi cantly associated with c-Met (P = 6.6 x 10 − 5 ), cMyc (P = 4.4 x 10 − 5 ), MMP2 (P = 5.7 x 10 − 6 ), and CK5/6 (P = 5.5 x 10 − 5 ) but not at all with Ki67, ERα, c-erbB-2, tumour size and histological grade (P ≥ 0.94) and only of possible borderline signi cance with TRNBC (P uncorrected = 0.06) and involved lymph nodes alone (P uncorrected = 0.037). The most signi cant association of staining for RAN was with that for MMP2 (P = 5.7x10 − 6 ) ( Table 2). When staining for MMP2 was tested for its relationship with staining for the other tumour variables, it was strongly signi cantly associated with the same variables as RAN: cMet (P = 6.4 x 10 − 9 ), cMyc (P = 1.  When staining for RAN was tested for its relative probability of association (RA) with that of its potential target molecules cMet, cMyc, MMP2 and Ki67 using binary logistic regression, the RAs with cMet and MMP2 were the strongest (RA = 3.0 to 3.4), but that with Ki67 was not signi cant (RA = 1.12, P = 0.81) ( Table 1). Moreover, when staining for c-Met was analysed, it also showed the strongest associations with that for RAN (RA = 3.41, P = 0.019) and for MMP2 (7.9, P < 0.001), but the strongest association for cMyc was with MMP2 and that for MMP2 was with cMet (RA = 7.7) ( Table 1), suggesting a closer association between these 3 molecules than with RAN itself.

Association Of Ran, Mmp2 And Patient Survival
When staining for RAN, cMet, cMyc and MMP2 were tested together for independent association with patient survival times using Cox's multivariate analysis, they all showed some signi cant degree of independence (P ≤ 0.036) with similar relative risks (RR) for patient demise of 3.1 to 3.7 fold (Table 2). These RRs were considerably less than the 7 to 15-fold decreases obtained in univariate analyses (SupplementaryTable 1). When analysed in binary combinations of staining for RAN with that for cMet, for cMyc or for MMP2, the RR for patient demise was suppressed from 14.9 to 7.6-7.8 fold for RAN with cMet or with MMP, but only to 9.8 fold with cMyc (Fig. 5, SupplementaryTable 3). The effects of staining for RAN and for MMP2 on RR were synergistic, increasing from 17.1 and 23.1, respectively, to 82.1 or, in terms of patients surviving, from 64% and 60%, respectively, to only 6% surviving after nearly 20 years (Fig. 5).
The sensitivity, speci city, positive predictive response (PPR) and negative predictive response (NPR) of IHC scoring for RAN, MMP2 and RAN with MMP2 are presented in Table 3. The NPR is shown to be the best performing for the two biomarkers RAN and MMP2 taken together than when one of them is used alone for the prediction of patients alive (Table 3). Patients monitored up to 20 years from diagnosis and free from radiotherapy, chemotherapy and hormone therapy, 2 cases of death by other causes have been excluded, 3 PPR =positive percentage response, 4 NPR= negative percentage response.

Discussion
Previously, we have shown that knockdown of RAN by shRNA results in reduction of in vitro cell biological properties including cell adhesion, colony formation and cell invasion (17)(18)(19) as well as in vivo metastasis (3). Here RAN knockdown in breast cancer cells reduces MMP2 mRNA and protein levels, probably via ATF3 and CXCR3 which in turn results in a signi cant reduction in cell adhesion and colony formation in breast cancer cell lines. However, overexpression of MMP2 in RAN knocked-down breast cancer cells results in overcoming RAN silencing and this led to increases in cell adhesion and cell invasion. The fact that transfection of pBabe MMP2 overcomes the knockdown effect of RAN on the levels of MMP2 and consequent biological effects is probably due to the natural promoter being different from that of pBabe.
In this study, knockdown of CXCR3 results in reduction of mRNA / protein levels of MMP2 with no changes in ATF3 expression. However, silencing of AFT3 caused a reduction in both MMP2 and CXCR3 mRNA expression. Thus, it is probable that the RAN/MMP2 pathway is connected in the order of RAN → ATF3 → CXCR3 → MMP2. It has been established that CXCR3 is suppressed in cardiomyocytes and macrophages from ATF3-knockout mice and is positively regulated by ATF3 through an ATF3 transcriptional response element found in its proximal promoter (22). In another study, knockdown of ATF3 using siRNA reduced the expression of MMP2 and inhibited the growth of U373MG cells grown in vivo xenografts in nude mice (8). It has also been shown that CXCR3 promotes gastric cancer cell migration and invasion by upregulating MMP2 expression (21). Our results are consistent with these reports.
Although previous publications (17)(18)(19) and work in this paper have established a causal relationship between RAN, cMet, cMyc, MMP2 and properties related to metastasis, these studies have been undertaken in cell line models of breast cancer. By using IHC staining of primary breast cancers we have also previously shown that RAN (18), cMet (17), and cMyc (19) are on their own signi cantly associated with patient demise from metastatic breast cancer. Now we have shown that increased staining for RAN is very signi cantly associated with staining for proteins in the cell-signalling pathway containing cMet, cMyc, and MMP2 that is linked to increases in the metastatic properties of cultured cells. Since increased IHC staining is related to increased levels of protein in the carcinoma cells (18), protein levels have also therefore increased by similar levels of at least 5-10 folds between tumours whose patients are at low and those who are at high risk of dying from metastatic disease. These fold increases in cellular levels are su cient to cause the increases in metastatic properties observed in stably transfected cells in culture as outlined above. There is no signi cant association of staining for RAN with that for Ki67, consistent with little increase in cell proliferation being observed in RAN transfected cells. The fact that staining for RAN, cMet, cMyc and MMP2 are all very signi cantly associated with that for CK5/6, but not with that for ERα or c-erbB-2 (SupplementaryTable 2) suggests that these proteins occur mainly in the Basal Cell Type of breast cancers. This subgroup of breast cancers overlaps considerably with the triple receptor negative breast cancer (TRNBC) group (de Silva et al, 2011) and hence may explain the observed borderline association of staining for RAN with the TRNBC subgroup alone (SupplementaryTable 2). The fact that there is a stronger relative association (RA) between staining for MMP2, cMet and cMyc than with that for RAN (

Conclusion
In this article, we have established in cell line models of breast cancer a direct relationship between Ran and MMP-2 and properties related to metastasis. This inclusion facilitates a more accurate prognosis and further identi es a subgroup of patients that could bene t more from chemotherapy and from therapy directed against both proteins than against either one alone.

Consent for publication
The authors consent for publication Availability of data and material

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.