Annexin A5 regulates hepatocarcinoma malignancy via CRKI/II-DOCK180-RAC1 integrin and MEK-ERK pathways

As a calcium-dependent phospholipid binding annexin protein, annexin A5 (Anxa5) links to the progression, metastasis, survival, and prognosis of a variety of cancers. Current work showed ANXA5 overexpression was positively correlated with the upregulations of CRKI/II and RAC1 in hepatocarcinoma (HCC) patients’ tissues, which potentially enhanced the clinical progression and lymphatic metastasis of HCC. The role and action mechanism of ANXA5 in hepatocarcinoma was then investigated using a hepatocarcinoma Hca-P cell line, an ideal and well-established murine cell model with 100% inducible tumorigenicity of implanted mice with low (~25%) lymph node metastatic (LNM) rate. In vitro evidences indicated ANXA5 stable knockdown resulted in decreased proliferation, migration, invasion and adhesion to lymph node (LN), and increased intercellular cohesion behaviors of hepatocarcinoma Hca-P cells. Consistently, stable ANXA5 knockdown led to reduced in vivo tumorigenicity and malignancy, LNM rate and level potentials of Hca-P- transplanted mice via inhibiting CD34 and VEGF3. The levels of CRKI/II and RAC1 were reduced in tumor tissues from mice transplanted with Hca-P cells with stable ANXA5 knockdown. Molecular action investigation further showed ANXA5 downregulation apparently suppressed the expressions of molecules CRKI/II, DOCK180, RAC1 in integrin pathway, p-MEK, p-ERK, c-Myc, and MMP-9 in MEK- ERK pathway together with VIMINTIN in Hca-P cells in appropriate to knockdown extent. Collectively, Anxa5 was able to mediate HCC carcinogenesis via integrin and MEK-ERK pathways. It is of potential use in the research and treatment of HCC.


Introduction
Listed as 5th and ranked as 3rd highest mortality among the most common cancers, hepatocarcinoma is an increasing health problem worldwide with high occurrence, metastasis and poor prognosis 1,2 . Generally, the early malignancy of lymphatic metastasis is clinically considered as a major prognostic indicator of cancers 3,4 . The lymph node metastasis (LNM) leads to a decrease of 50% for the prognosis of cancer patients. The study on the metastasis mechanism, specifically on the lymphatic metastasis, benefits better understanding of the diagnosis, prognosis and treatment of hepatocarcinoma.
Anxa5 is a member of the group A of annexin family that are Ca 2+ -regulated phospholipid-binding proteins with 12 members, annexin A1-A11 and A13. Commonly activated by higher cellular concentration of Ca 2+ or higher content of phosphatidylserine (PS), monomeric ANXA5 spontaneously forms trimer through binding to the cell membrane for exhibiting its functions 5,6 in physiological and pathological processes such as cell differentiation, apoptosis, signal transduction, inflammation and coagulation [7][8][9][10] . The dysexpression of Anxa5 is associated with the development, progression, metastasis, treatment and prognosis of a variety of cancers [11][12][13][14] . As we summarized the researches on Anxa5 in carcinogenesis 15 , Anxa5 expression level were commonly correlated with increased progression, metastasis, poorer survival and prognosis of most types of cancers. However, the underlying molecular regulation mechanism and clinical significance of Anxa5 in cancer progression and metastasis, especially in cancer lymphatic progression and metastasis, are poorly understood.
Research work from our group established the positive correlation of Anxa5 expression level with hepatocarcinoma progression and lymphatic metastasis [16][17][18][19] . Higher expression level enhanced the in vitro and in vivo tumor malignant progression and LNM potential of Hca-P, a murine hepatocarcinoma cell line with 100% tumorigenicity and~25% LNM rate being proved as an ideal model for mimicking the initial lymphatic metastasis of hepatocarcinoma cell and hepatocarcinoma progression in clinical [20][21][22][23][24] . In current work, we found ANXA5 was overexpressed in tumorous tissues from HCC patients and correlated with CRKI/II and RAC1. Cellular experimental results showed that stable knockdown of ANXA5 led to decreased in vitro migration and invasion capacities as well as clearly reduced in vivo tumorigenicity velocity and malignancy, LNM rate and level potentials of Hca-P-transplanted mice via inhibiting the expressions of CD34, VEGF3, CRKI/II, and RAC1. Current work also revealed the critical molecules CRKI, CRKII, DOCK180, and RAC1 in integrin pathway, as well as p-MEK, p-ERK, c-Myc, and MMP-9 in MEK-ERK pathway were liked to ANXA5 dysexpression in Hca-P cells. Anxa5 expression mediates the progression and metastasis of HCC via integrin and MEK-ERK pathways. It is of potential use in HCC diagnosis and treatment.

ANXA5 overexpression correlates with HCC progression and metastasis
A tissue microarray composed of 46 paired tumorous and adjacent normal tissues from HCC patients was employed to address the expression alteration of ANXA5. As shown in Fig. 1a and Table 1, ANXA5 was upregulated in patients' tumorous tissues. The change was analyzed for its relevance with the clinicopathologic parameters of HCC patients. As shown in Table 2, its dysexpression was related to TNM stage (P = 0.015). Among 46 HCC patients, 3 patients diagnosed with LNM all showed immunoreactivities against ANXA5 with the highest degree +++ (P= 0.014). The results implicated ANXA5 level was positively correlated with the progression and metastasis of HCC.

ANXA5 correlates with Hca-P cell growth
The suppression of ANXA5 on Hca-P growth was related with its downregulation extent. For colony forming capacity, a colony was defined as over 30 cells originally derived from a single cell. During a 10-day-colony-growth incubation, the colonies for Hca-P-ANXA5-shRNA1 and Hca-P-ANXA5-shRNA2 were less than the control, Fig. 2c. Compared with Hca-P-ANXA5-control, Hca-P-ANXA5-shRNA1 showed a decreased colony forming ability by 65.5% (P = 0.0003, Fig. 2c). Interestingly, the partial downregulation of ANXA5 expression (~53%) didn't affect the colony forming ability of Hca-P cells, in fact no apparent difference was measured between Hca-P-ANXA5-shControl and Hca-P-ANXA5-shRNA2 (P > 0.05, Fig. 2c). Moreover, the colony size of Hca-P-ANXA5-shRNA1 was smaller than Hca-P, Hca-P-ANXA5-shControl and Hca-P-ANXA5-shRNA2 cells (Fig. 3c). Trypan blue staining assay on the influence of ANXA5 downregulation on in vitro Hca-P proliferation was consistent with its effect on Hca-P colony forming capacity. As shown in Fig. 2d, no clear differences were measured for the proliferations among Hca-P-ANXA5-shRNA2, Hca-P-ANXA5-shControl and Hca-P. While, compared with Hca-P-ANXA5-shControl, the proliferation of Hca-P-ANXA5-shRNA1 was obvious decreased by ANXA5 knockdown. The suppression of ANXA5 knockdown on Hca-P growth relying on its downregulation extent.
ANXA5 knockdown attenuates Hca-P adhesion to LN and promotes cell-cell adhesion ANXA5 deregulation affects the in situ adhesion ability of Hca-P to peripheral lymphatic endothelium. The adhesion abilities (numbers) of Hca-P-ANXA5-shRNA1 and Hca-P-ANXA5-shRNA2 to LN decreased by~61% ( P= 0.0014) and 35% (P = 0.0109) than Hca-P-ANXA5-shControl cells (Fig. 3a). A higher degree knockdown of ANXA5 led to a more reduced in situ adhesion of Hca-P to LN. Fig. 2 The influence of stable knockdown of ANXA5 on the cell growth of Hca-P cells. a Western blotting assay indicated that ANXA5 was down-regulated by~99% (P < 0.001) and~53% (P < 0.001) in Hca-P-Anxa5-shRNA1 and -shRNA2 than the control group cells. Levels are normalized to α-Tubulin. Values mean ± S.D. of three independent experiments. b Compared with Hca-P-Anxa5-shControl, qRT-PCR assay showed Anxa5 levels decreased by~91% (P = 0.0008) and~17% (P = 0.0455) in Hca-P-Anxa5-shRNA1 and -shRNA2 cell lines. c The inhibition of ANXA5 knockdown on the colony forming capacity of Hca-P depending on its knockdown extent. Compared with Hca-P-Anxa5-shControl cells, the colony forming activity of Hca-P-Anxa5-shRNA1 decreased by 65.5% (P = 0.0003), while, no stastical difference was between Hca-P-Anxa5-shControl and -shRNA2. d MTT assay for the effect of ANXA5 knockdown on the proliferation of Hca-P cells. The proliferation decreases of Hca-P-Anxa5-shRNA1 compared with Hca-P-Anxa5-shControl were of stastical significations (P < 0.01). No proliferation differences were observed among Hca-P, Hca-P-Anxa5-shControl and -shRNA2. Values mean ± S.D. of three independent experiments We then evaluated the effect of ANXA5 knockdown on the intercellular adhesion property of Hca-P cells by hanging drop aggregation assay. The transfection of Hca-P with mock vector did not alter the cell cluster distributions. Hca-P-ANXA5-shControl group cells was dominated with 1 and 2 cells per aggregate ( P < 0.001, Fig. 3 ANXA5 suppression on the LN, cell-cell adhesion, migration, and invasion abilities of Hca-P. a The in situ adhesion ability of Hca-P to peripheral LN was decreased following ANXA5 inhibitions (P < 0.01). b ANXA5 suppression alters growth status of Hca-P. Hca-P-ANXA5-shRNA1 and shRNA2 grows close together easily, whereas Hca-P-ANXA5-shControl cells growing dispersively. Hca-P-ANXA5-shControl cells grow dominantly and evenly with one and two cells per aggregate. Following ANXA5 suppression, Hca-P-ANXA5-shRNA1 and -shRNA2 cells grow dominantly with 5 and > 5 cells per aggregate. c Phalloidin cytoskeleton staining assay of the morphology and structural changes of Hca-P cells following ANXA5 suppression. Its knockdown enhances intercellular adhesion ability of Hca-P. Following ANXA5 suppression, d the in vitro migration abilities and e invasion capacities of Hca-P-ANXA5-shRNA1 and Hca-P-ANXA5-shRNA2 cells decreased by~69% (P = 0.0003) and 39% (P = 0.0026), and decreased by~71% (P = 0.0002) and~56% (P = 0.0018) than Hca-P-ANXA5-shControl cells, respectively. Data are expressed as mean ± S.D. of three experiments  Fig. 3b). Controversially, following ANXA5 knockdown in Hca-P cells, the aggregates of Hca-P-ANXA5-shRNA1 and -shRNA2 were dominated with 5 and >5 cells (P < 0.001, Fig. 3b). The percentage of Hca-P-ANXA5-shRNA1 and -shRNA2 with >5 cells per aggregate increased by~4 and 2.5 times than Hca-P-ANXA5-shControl, respectively ( P< 0.001, Fig. 3b). The number of aggregate cell clusters were more with more ANXA5 knockdown extent.
Phalloidin cytoskeleton staining assay was performed to detect the morphology changes and structural evidences for increased intercellular adhesion ability of Hca-P cells following ANXA5 knockdown. More spreading lamellipodia (sheet-like protrusions from a cell) and more junction sites were found formed at the surfaces of Hca-P-ANXA5-shRNA1 and -shRNA2 cells than Hca-P-ANXA5-shControl cells (Fig. 3c). DAPI staining assay clearly showed more cells were included in each aggregate of Hca-P-ANXA5-shRNA1 and -shRNA2 than Hca-P-ANXA5-shControl cells (Fig. 3c). ANXA5 knockdown enhances intercellular adhesion ability of Hca-P cells.
ANXA5 knockdown reduces the in vitro migration and invasion of Hca-P ANXA5 knockdown significantly inhibits the in vitro migration and invasion abilities of Hca-P. ANXA5 knockdown suppressed the in vitro migration ability of Hca-P (Fig. 3d). The numbers of migrated Hca-P-ANXA5-shRNA1 and Hca-P-ANXA5-shRNA2 cells were~31% (P = 0.0003) and 61% (P = 0.0026) of that of Hca-P-ANXA5-shControl cells. ANXA5 downregulation also suppressed the in vitro invasion capacity of Hca-P (Fig. 3e). More extent knockdown of ANXA5 resulted in more decreased migration of Hca-P cells.

ANXA5 downregulation reduces the in vivo LNM rate of Hca-P transplanted mice
To study the role of ANXA5 in tumorigenesis and lymphatic metastasis, the lymph nodes (LNs) from inguinal, popliteal, axillary, iliac arterial and pararenal of the tumor-bearing mice were collected in 10 (soon after tumor formation) and 21days. As shown in Fig. 4c, d, the average masses of popliteal, inguinal, axillary LNs of Hca-P-ANXA5-shRNA1 group were lighter than Hca-P-ANXA5-shControl group. Histological examination (H. E. Staining) shows (Fig. 4e) the number of LNs invaded with Hca-P-ANXA5-shRNA1 cells was less than that of LNs invaded with Hca-P-ANXA5-shControl cells. The structure of the metastatic LNs from Hca-P-ANXA5-shControl-tansplanted mice became vague on 21 st day, nevertheless, the structure of LNs from Hca-P-ANXA5-shRNA1-transplanted mice were still distinct with clear and deep nucleus staining (Fig. 4e). ANXA5 knockdown reduces LNM potential of Hca-P cells in the development and progression of tumorigenicity.
CD34 was locally expressed in cell membrane and cytosol. Although without statistical significance (Fig. 5a, Table 4, P= 0.72), CD34 expression level in primary (see figure on previous page) Fig. 4 ANXA5 suppression reduces in vivo tumorigenicity and LNM rate of Hca-P. The influence of ANXA5 suppression on in vivo tumor formation volume and mass for mice transplanted with Hca-P-ANXA5-shControl and Hca-P-ANXA5-shRNA1 cells in 10 (a) and 21 (b) days. Tumor volume (V) was calculated as length × width × depth × π/6. Tumorigenicity of Hca-P-ANXA5-shRNA1 was decreased than Hca-P-ANXA5-shControl with ** P < 0.01. ANXA5 suppression decreases LNM malignancy and rate of tumor cell-bearing mice. The mass reductions of popliteal and inguinal LNs from Hca-P-ANXA5-shRNA1-tansplanted mice in c 10 and d 21 d following transplantation are of stastical significance than control group mice (P < 0.05). Data are expressed as mean ± S.D. e Popliteal and inguinal LNs with H.E staining. There were few morphological changes of the Hca-P-ANXA5-shRNA1-transplanted mice, including the chromatin condensed in border and the pathological karyokinesis reduced obviously than Hca-P-ANXA5-shControl-transplanted mice. The metastatic LNs from Hca-P-ANXA5-shControl-transplanted mice became more vague and abnormal on 21 st d than LNs from Hca-P-ANXA5-shRNA1-transplanted mice tumor tissues from Hca-P-ANXA5-shRNA1 group mice on the 10 th day following cell implantation showed a reduction tendency than Hca-P-ANXA5-shControl group mice. On 21 st day, CD34 expression was depressed in the tissues from the Hca-P-ANXA5-shRNA1 group mice significantly (Fig. 5a, Table 4, P = 0.002).
VEGF-3 expression was located in cytosol (Fig. 5b, Table 4). Similar result was obtained on the influence of ANXA5 knockdown on VEGF-3 expression in primary tumor induced by Hca-P cell transplantation in mouse.
These findings indicate ANXA5 knockdown suppresses in vivo HCC progression and metastasis via downregulating the expressions levels of CD34 and VEGF-3 in tumorigenesis.
ANXA5 knockdown potentially downregulated CRKI/II and RAC1 in suppressing tumorigenicity of Hca-P-transplanted mice ANXA5 knockdown in Hca-P led to reduced expressions of CRKI/II and RAC1 in its induced primary tumors of mice. ANXA5 was distributed in cell membrane and cytosol. ANXA5 expression levels in tumorous tissues on both 10 th and 21 st d were decreased (Fig. 6a, Table 4, P = 0.001).
Although without statistical significance, CRKI/II and RAC1 showed decreased expression tendencies in tumor tissues from Hca-P-ANXA5-shRNA1 than Hca-P-ANXA5-Control group on 10 th d (Fig. 6b, c, Table 4, P = 0.309). On 21 st d following tumor cell implantation, the decreases of CRKI/II and RAC1 expressions in primary tumors of Hca-P-ANXA5-shRNA1 group mice were of statistical significances.

ANXA5 mediates Hca-P behaviors viaCRK/DOCK180/RAC1 and pMEK-ERK pathways
Tumor migration and invasion are strictly regulated by actin cytoskeleton modulated by Rho family of small GTPases, including RAC1, Cdc42, and RhoA in space timely. DOCK180 is a guanine-nucleotide exchange factor for mediating RAC1 and Rap1. CRKI and CRKII, two alternative splicing products of signaling adaptor CRK family, delivery signals to GTPases through its downstream effector DOCK180. Based on previous results on the positive correlations of ANXA5 with RAC1 and CRKI/II in HCC progression and metastasis, we proposed and investigated the molecular action mechanism and association of ANXA5 with CRKI/II, RAC1 and DOCK80 using Hca-P cells.
Taken together, ANXA5 inhibition potentially suppresses the critical molecules CRKI, CRKII, DOCK180 and RAC1 in integrin pathway, p-MEK, p-ERK1, and p-ERK2 in ERK pathway, as well as VIMENTIN in EMT for inhibiting signal transduction into nuclear membrane to invoke MMP9 and C-MYC, as summarized and schemated in Fig. 8, which results in decreased in vitro proliferation, migration and invasion, and in vivo tumorigenicity velocity, malignancy and metastasis of hepatocarcinoma cells.

Discussion
As a member of annexin A family, Anxa5 plays important roles in vesicle trafficking and cell behaviors [25][26][27] , in the development, progression and diagnosis of some cancers [6][7][8][9][10] . At present, the research of Anxa5 in cancer lymphatic metastasis is limited. Previous work from our group demonstrated ANXA5 up-regulation increased the in vitro and in vivo malignancy behavior 16 . In current work, we investigated the associations of ANXA5 overexpression with the clinical progression and lymphatic metastasis in hepatocarcinoma patients' tissue microarray, and the action mechanism in vivo and in vitro.
We found ANXA5 overexpression contributed to the progression and metastasis of HCC patients. It was correlated with TNM stage advance. Our work showed the positive correlations of ANXA5 level with CRKI/II and RAC1, critical molecules in integrin pathway, in tumorous tissues from HCC patients. As two alternative splicing products of v-CRK 28 , CRKI and CRKII have been linked to a variety of human cancers [28][29][30] . Through DOCK180, CRKI/II was suggested to deliver signals to small GTPases, RAC1, Cdc42 and RhoA, for modulating actin cytoskeleton organization in tumor malignancy 31 . The overexpression of CRKI/II and RAC1 commonly lead to relative higher tumor cell migration and invasions [32][33][34] . Compared with the adjacent normal tissues, IHC assay showed CRKI/II and RAC1 were apparently overexpressed in HCC tumorous tissues and also positively correlated with patients' TNM stage. Mutual interpositive correlations were apparently established for the levels of ANXA5, CRKI/II and RAC1 in tumorous tissues from patients. Interestingly, all of the three protein showed highly expression level in three patients diagnosed with LNM. The above results implicate ANXA5 upregulation that is positively correlated with CRKI/II and RAC1 potentially promotes the clinical progression and metastasis of HCC.
We then investigated its potential action mechanism in HCC carcinogenicity and LNM using Hca-P cell line model. We constructed Hca-P cell lines with ANXA5 stable knockdowns to investigate consequent effects on the in vitro and in vivo malignant behaviors of Hca-P with underlying action mechanism. Compared with Hca-P-ANXA5-shControl, the monoclonal Hca-P-ANXA5-shRNA1 showed more decreased degree of ANXA5 expression than Hca-P-ANXA5-shRNA2 cell lines. ANXA5 suppression inhibited the in vitro growth, migration and invasion, in situ cell adhesion to LN as well as promoted the in vitro cell cohesion and aggregate capacities of Hca-P related to its knockdown extent. The cohesive and aggregative growth of Hca-P promoted by ANXA5 downregulation was further supported by phalloidin staining assay. The above results suggest ANXA5 might mediate Hca-P malignant properties via affecting its adhesion to LNs, and intercellular cohesion and aggregate capacity. Consistently, a greater knockdown extent of ANXA5 led to more clearly significant suppressions on in vitro malignant property and metastasis ability of Hca-P cells.
Then, consistent with the in vitro evidences, we found ANXA5 suppression led to apparently decreased the in vivo tumorigenicity and LNM potential of Hca-Ptransplanted mice. The attenuation of angiogenesis and lymphangiogenesis attenuate tumor development and metastasis 35,36 . Current work showed ANXA5 mediated Hca-P-transplantation induced murine tumorigenesis and metastasis via CD34 and VEGF-3, two well-known angiogenesis and lymphangiogenesis indicators 37,38 . As ANXA5, CRKI/II and RAC1 expressions were upregulated and positively correlated in tumorous tissues from HCC patients, we then measured ANXA5, CRKI/II and RAC1 expression levels in primary tumors from transplanted mice. The results showed ANXA5, CRKI/II, and RAC1 were reduced in Hca-P-shANXA5-transplanted mice.
CRKI/II usually binds to DOCK180 to activate RAC1 39 , resulting in activated MEK-ERK signaling transduction through PAK to mediates cell migration [40][41][42] . Hence, we proposed ANXA5 suppression might inhibit Hca-P malignant properties via CRK/DOCK180/RAC1 and MEK-ERK pathways. We found CRKI, CCRKII, DOCK180, RAC1, p-MEK (Ser217/221) and p-ERK1 levels reduced in Hca-P -ANXA5-shRNA1 and Hca-P-ANXA5-shRNA2 cells than Hca-P-ANXA5-shControl cells. Interestingly, p-ERK2 (Thr202/Tyr204) level was significantly reduced in Hca-P-ANXA5-shRNA1 cells, Fig. 6 Deregulating ANXA5 reduced the In vivo metastasis of Hca-P transplanted mice through CRKI/II and RAC1. IHC results shown ANXA5 was downregulation in tumor body both in 10 and 21 d (P= 0.001, A, Table 4). CRKI/II expression was suppressed in Hca-P-ANXA5-shRNA1 group in 21 d with significantly statistics (P = 0.041, B, Table 4). RAC1 expression had a decreased trend in 10d, and clear decrease in 21 d following transplantation (P = 0.013, C, Table 4) while, indistinctly altered in Hca-P-ANXA5-shRNA2 than Hca-P-ANXA5-Control cells. These results helped to understand why Hca-P-ANXA5-shRNA1 showed much more decreased malignant properties than Hca-P-ANXA5-shRNA2 cells. MMP9 and C-MYC are two critical molecules in MEK/ERK cascade-mediated cell migration and proliferation 43,44 . p-MEK phosphorylates both the Ser/Thr and Tyr residues of ERK1/2, transduces signals from cell surface into cytosol and nucleus, results in reduced transcriptional activities of MMP9 and C-MYC [45][46][47][48] . Our work showed ANXA5 knockdowns resulted in decreased expressions of C-MYC and MMP9 in Hca-P-ANXA5-shRNA1 and Hca-P-ANXA5-shRNA2 cells. Greater extent suppressions of these molecules were detected in Hca-P-ANXA5-shRNA1 than Hca-P-ANXA5-shRNA2, which consequently contributed in the greater extent suppressions of malignant phenotypes and behaviors of hepatocarcinoma cells, which also implicated the linkage of ANXA5 overexpression positively correlated with RAC1 and CRKI/II in promoting the clinical progression and metastasis of HCC. More scientific and clinical significances for investigating the role and action mechanism of ANXA5 in the progression and metastasis of human cancers and cancer cell lines should be addressed in the further. The new interesting oncotargets such as ANXA5 partners to be identified by highthroughput techniques including shotgun proteomics, comparative genomics, RNA-omics should provide complementary oncotargets to ANXA5 and supply new clues to tumorigenesis and the underlying mechanism.

Materials and methods
Cell culture, experimental animal and HCC tissue array Hca-P, a murine hepatocarcinoma cell line with~25% lymph node metastasis (LNM) rate was established and maintained in our research group at Dalian Medical University. 2 × 10 6 Hca-P cells were implanted into the abdominal cavity of 615-mouse (aged 6 w, 20 ± 2 g; Certificate number: SCXK (Liao) 2008-0002). The Hca-P cells taken from mouse abdominal cavity were purified via 2 passages and cultured in RPMI 1640 (Gibco, USA) with 15% FBS at 37°C with 5% CO 2 . Experimental mice supplied by the SPF Animal Laboratory Center of Dalian Medical University were fed, treated and sacrificed according to approved protocols by Experimental Animal Ethical Committee of Dalian Medical University (Permit Number: L2012012).
HCC tissue microarray was from Super Biotek (Shanghai, China). It was composed of 46 cases of HCC tumorous tissues including 3 HCC tissues with lymphatic metastasis.
Quantitative real-time PCR (qRT-PCR) assay qRT-PCR assay was performed for analyzing the level changes of Anxa5, MMP9, and Vimentin among different group cells using GAPDH as internal reference. Total RNA was extracted from each group cells using Trizol TM reagent (Life Technologies, USA). cDNA was prepared with Primescript RT Reagent (TaKaRa, Japan). Quantitative

SDA-PAGE and Western blotting assay
The total protein extraction and western blot were performed, as previously described 17

Cell growth assay
For colony forming, the base layer was prepared by mixing 1 mL 1.2% melted soft agar (AMRESCO, USA) with 1 mL 2 × medium (15%FBS in RPMI-1640). 1 mL 0.7% soft agar was mixed with 1 mL 2 × medium (15% FBS in RPMI-1640), mixed well with 1000 cells from each group, and seeded evenly onto the top of the basal layer as the upper layer in a 6-well plate. The plates were then incubated with 5% CO 2 at 37°C for 10 d. Cell colonies were fixed in methyl hydrate for 10 min and stained in 0.1% crystal violet, counted using a microscope and compared. For trypan blue staining assay, 1 mL (5 × 10 4 cells/mL) different group cells were added into 24-well plate and incubated at 37°C with 5% CO 2 . Surviving cells were counted using a cytometer with trypan blue staining every day for five consecutive days.
In situ cell adhesion potential to lymph node (LN) assay Frozen fresh LNs from 615 mice were sectioned into 10 µm slices. The surfaces of frozen slices of inguinal and popliteal LNs were covered in 200 μL serum-free RPMI 1640 containing 2 × 10 5 cells of Hca-P, Hca-P-ANXA5-shRNA1, -shRNA2, and shControl at 37°C with 5% CO 2 for 24 h, fixed in 95% ethanol, washed with flowing distilled water for 3 s and stained with hematoxylin eosin (HE). The adherent cells was counted and averaged from five random fields using a microscope (Olympus, Japan).

Hanging drop aggregation assay for intercellular adhesion
The effect of ANXA5 knockdown on the cell-cell adhesion ability of Hca-P cells were determined by hanging drop aggregation assay 48 . In total 30 μL of medium containing 5000 cells was loaded onto the inner surface of the cover of a 100 mm diameter Petri dish and incubated the hanging drop cells overnight at 37°C with 5% CO 2 . The number of cells in cell aggregates were carefully counted by two investigators, separately, using an inverted microscope (Olympus, Japan) by randomly selecting 100 attached sites. Triplicate experiments were performed for each assay.

Phalloidin cytoskeleton staining assay for cellular morphology
In total 2 × 10 5 cells in 200 μL RPMI-1640 with 15% FBS was loaded onto the center of a glass slide, put into a wet box, incubated in a humidified environment at 37°C with 5% CO 2 for 24 h. The cells were then anchored in 4% paraformaldehyde solution for 10 min, and dehydrated with pre-cool acetone (−20°C) for 5 min. Next incubated with 400 nM FITC-phalloidin (Sigma, USA) for 30 min in darkness at RT. The slides were counterstained in 100 μL 10 μg/ml Hoechst 33258 (Sigma, USA) for cell nucleus. Images were taken by a fluorescence microscope (Olympus, Japan) with ×1000.

In vitro cell migration and invasion assays
The 24-well transwell units with 8 mm I.D. polyester membrane (Corning, USA) were employed. For migration assay, 600 μL RPMI-1640 with 15% FBS and 6 × 10 4 cells from each group in 200 μL RPMI-1640 were loaded into the lower and upper chambers respectively, and incubated at 37°C with 5% CO 2 for 24 h. The cells migrated to lower surface of the filter were fixed in 4% paraformaldehyde for 20 min, stained in 0.1% crystal violet for 20 min, counted by randomly selecting 5 fields per well with a light microscope at a magnification of ×200.
For invasion assay, the fitter surface of an insert transwell unit was first coated with 50 μL ice-cold ECM gel (1:5 dilution with RPMI 1640, Sigma-Aldrich, USA) by incubating at 37°C for 8 h. The rest steps were the same as in migration assay.

In vivo tumorigenicity and LNM investigations
Thirty-six 615 mice (4-6 w, 21 ± 2 g), half male and half female, were used for in vivo tumorigenicity subjects of Hca-P cells. These mice were randomly divided into two groups with 18 mice per group, half male and half female. 2 × 10 6 of Hca-P-ANXA5-shControl and Hca-P-ANXA5-shRNA1 cells in 25 μL medium were transplanted into the footpads of mice. The length, width, depth of primary tumor formed on the footpad was measured using a vernier calliper. Tumor volume (V) was calculated according to the formula: V = length × width × depth × π/6. The primary neoplastic tissues and LNs from the inguinal, popliteal, axillary, iliac arterial, pararenal of the tumorbaring mice were isolated, weighted and photographed on 10 th and 21 st d following tumor cell implantation. The tissues were fixed in neutral formalin for 48 h and embedded in paraffin for later use.

Data processing and statistical analysis
The data were presented as mean ± SD of at least three independent experiments. Comparisons between two groups were performed by Student's t-test. For IHC assay, a Mann-Whitney U or Fisher exact test and Spearman rank correlation was employed. SPSS17.0 was used for statistical analysis. Results with P < 0.05 were of significance.