Insufficiency of hepatocyte growth factor activator inhibitor‐1 confers lymphatic invasion of tongue carcinoma cells

Abstract Hepatocyte growth factor (HGF) activator inhibitor type‐1 (HAI‐1), encoded by the SPINT1 gene, is a transmembrane protease inhibitor that regulates membrane‐anchored serine proteases, particularly matriptase. Here, we explored the role of HAI‐1 in tongue squamous cell carcinoma (TSCC) cells. An immunohistochemical study of HAI‐1 in surgically resected TSCC revealed the cell surface immunoreactivity of HAI‐1 in the main portion of the tumor. The immunoreactivity decreased in the infiltrative front, and this decrease correlated with enhanced lymphatic invasion as judged by podoplanin immunostaining. In vitro homozygous deletion of SPINT1 (HAI‐1KO) in TSCC cell lines (HSC3 and SAS) suppressed the cell growth rate but significantly enhanced invasion in vitro. The loss of HAI‐1 resulted in enhanced pericellular activities of proteases, such as matriptase and urokinase‐type plasminogen activator, which induced activation of HGF/MET signaling in the co‐culture with pro‐HGF‐expressing fibroblasts and plasminogen‐dependent plasmin generation, respectively. The enhanced plasminogen‐dependent plasmin generation was abrogated partly by matriptase silencing. Culture supernatants of HAI‐1KO cells had enhanced potency for converting the proform of vascular endothelial growth factor‐C (VEGF‐C), a lymphangiogenesis factor, into the mature form in a plasminogen‐dependent manner. Furthermore, HGF significantly stimulated VEGF‐C expression in TSCC cells. Orthotopic xenotransplantation into nude mouse tongue revealed enhanced lymphatic invasion of HAI‐1KO TSCC cells compared to control cells. Our results suggest that HAI‐1 insufficiency leads to dysregulated pericellular protease activity, which eventually orchestrates robust activation of protease‐dependent growth factors, such as HGF and VEGF‐C, in a tumor microenvironment to contribute to TSCC progression.


| INTRODUC TI ON
Squamous cell carcinoma is the major histological type of oral cancer.
It is highly frequent in South Central Asia and ranks 18th in incidence of cancers worldwide. 1 Tongue squamous cell carcinoma (TSCC) accounts for approximately 40% of oral cancers. 2 Although early stage TSCC is usually amenable to surgery, most patients present at an advanced disease stage. Surgical techniques, chemoradiotherapy and several other targeted therapies have advanced over the past two decades, but the 5-year survival rate for TSCC patients is still below 50%. 3 Rapid local invasion and lymph node metastasis are the main challenges in managing TSCC, and the underlying mechanisms of the aggressive features of this disease remain unclear.
Hepatocyte growth factor (HGF) activator inhibitor type-1 (HAI-1) is a type I transmembrane Kunitz-type serine protease inhibitor. HAI-1 is expressed in most epithelial cells and placenta trophoblasts. 4 Initially identified as an endogenous inhibitor of serum HGF activator, 5 type II transmembrane serine proteases (TTSPs) and glycosylphosphatidylinositol-anchored serine protease prostasin are physiological targets of HAI-1 in epithelial cells. 4,6 Among TTSPs, matriptase has been identified as the cognate target protease of HAI-1. 4,6,7 Abnormal matriptase activity contributes to the development and progression of squamous cell carcinoma of the skin. [8][9][10] Matriptase and HGF activator stimulate extracellular activation of the proform of HGF (pro-HGF). 4,11 Mature HGF activates the MET tyrosine kinase receptor pathway, which contributes to multiple cellular processes in cancers, including invasion, metastasis, and angiogenesis as well as protection from apoptosis. [12][13][14] Other growth factors, such as platelet-derived growth factor-C and -D and macrophage stimulating protein (a ligand of RON tyrosine kinase receptor), are activated by matriptase, [15][16][17] as are proteaseactivated receptor-2 (PAR-2) and other protease zymogens, such as the proform of urokinase-type plasminogen activator (pro-uPA). [18][19][20][21] Together, these targets of HAI-1 contribute to neoplastic progression, and consequently, accumulating experimental evidence has indicated suppressive roles for HAI-1 in carcinogenesis and metastasis. 4,8,[22][23][24][25] We have reported that HAI-1 immunoreactivity in oral cancer tended to be reduced at the invasion front of infiltrating-type cancers and was accompanied by increased numbers of cancerassociated fibroblasts through enhanced activation of matriptase/PAR-2 axis 21 and increased risk of lymph node metastasis. 26 However, the molecular mechanism underpinning the increased risk of lymph node metastasis associated with decreased HAI-1 remains to be clarified. In this study, we examined the effects of HAI-1 loss on the invasive growth of TSCC cells in vitro as well as on tumorigenesis, lymphatic invasion and lymph node metastasis in vivo using an orthotopic xenotransplantation method. Our results suggested that HAI-1 deficiency induced dysregulated pericellular protease activities that eventually promoted robust activation of growth factors possibly involved in lymphangiogenesis and invasion in the context of the tumor microenvironment.

| Clinicopathological study cohort
The study protocol was in accordance with the revised Helsinki

| Immunohistochemical study
Immunohistochemistry of paraffin embedded tissues was performed as described previously. 26,28 The HAI-1 immunoreactivity score was calculated as the sum of the following point values: cytoplasmic staining in >50% of tumor cells, 0.5 point; distinct membranous staining in 20%-50% of tumor cells, 1 point; and distinct membranous staining in >50% of tumor cells, 2 points. The immunoreactivity score of the whole tumor tissue was compared with that of the tumor invasion front. Then, the degree of reduction of HAI-1 levels in the invasion front (HAI-1 reduction level) was calculated by subtracting the invasion front HAI-1 score from the whole tumor area HAI-1 score. 26 Lymphatic permeation of cancer cells and the number of lymph vessels were evaluated by podoplanin immunohistochemistry.

| Cell culture
Human TSCC cell lines, HSC3 and SAS, were obtained from the Riken Cell Bank and the Cell Resources Center for Research (Tohoku University, Sendai, Japan), respectively. The human embryonic kidney cell line, HEK293T, and human fibroblast cell line, MRC5, were from the Riken Cell Bank and the Japanese Cancer Research Resources Bank, respectively, and cultured in DMEM containing 10% FBS (Sigma) at 7°C in a fully humidified atmosphere of 5% CO 2 in air.

| Knockout of the SPINT1 gene and knockdown of matriptase in cultured cells
Genome editing using clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR-associated proteins 9 (Cas9) was applied for SPINT1 (HAI-1) gene knockout (KO) in the HSC3 cell line and the SAS cell line as described previously. 29 KO of the SPINT1 gene was confirmed by genome sequencing and immunoblot analysis. For transient silencing of matriptase, two kinds of siRNA (siMat#1 and siMat#2; Table S1) were used as described. 21 Transfection of siRNA was performed using the RNAiMax reagent (Invitrogen).

| Evaluation of cellular proliferation, migration, and invasion
Cell proliferation was evaluated using CCK-8 (DOJINDO Laboratories) according to the manufacturer's instructions. For the colony forming assay, cells were seeded into six-well plates at 100 cells per well and cultured for 10 days. Plates containing colonies were washed with PBS and fixed, followed by staining with crystal violet. The number of visible colonies was counted, and the size of each colony was measured by microscopy. A wound healing assay was used to determine cellular migration activity, as described prevously. 29 To evaluate the invasive activity, trans-well chambers (ThinCert, Greiner Bio-One) were coated with 25 μg/ filter of Matrigel (BD Biosciences). A total of 1 × 10 5 cells suspended in DMEM with 0.1% BSA were seeded in the upper chamber. The lower chamber was filled with medium supplemented with 5% FBS. Cells were incubated at 37° for 24 h and invading cells on the inferior surface of the filters were fixed and stained with hematoxylin.

| Protease activity assays
To measure matriptase activity, serum-free culture conditioned medium was mixed with the synthetic substrate Boc-Glu-Ala-Arg-AMC (Peptide Institute; 10 μM). Fluorescence release was measured using a FlexStation 3 (Molecular Devices) for 30 min at 37℃, and the velocity max was calculated. Pericellular activities of urokinase-type plasminogen activator (uPA) and plasmin were measured with the synthetic substrates S-2444 (pyro-Glu-Gly-Arg-p-nitroanilide) and S-2251 (H-D-Val-Leu-Lys-p-nitroanilide dihydrochloride), respectively (KabiVitrum). Briefly, after the cells reached 80% confluency in the 96-well plate, each well was washed with PBS three times. For uPA activity assay, each well was filled with 100 µL PBS containing 10 µg/mL S-2444 and incubated for 90 min at 37°C. For plasmin activity assay, cells in 96-well plates were pretreated with 1 µg/mL plasminogen (R&D systems) for 3 h and then incubated with 10 µg/mL S-2251 for 90 min at 37°C. After incubation, optical density at 405 nm of each well was measured.

| Immunoblotting, RT-PCR and quantitative RT-PCR
Cell lysates were collected in 1% Triton X-100 in PBS with protease inhibitor cocktail (Sigma). Equal amounts of proteins were subjected to standard SDS-PAGE and processed for a standard immunoblot analysis as described previously. 29 The labeled proteins were visualized with a chemiluminescence reagent (PerkinElmer Life Sciences).
RT-PCR and quantitative real-time RT-PCR were performed as described previously. 21 The primer sequences were indicated in Table S1.

| Animal experiments
All the animal work was performed using protocols approved by the University of Miyazaki Animal Research Committee, in accordance with international guidelines for biochemical research involving animals. For the present study, 3 × 10 5 of HSC3 or SAS cells were laterally implanted into the muscle layer of tongues of 6 week-old male nude mice (BALB/cAJc1-nu; Kyudo) with or without 3 × 10 5 MRC5 fibroblasts. During the observation period, the body weight was measured every 2 days until the animals were killed based on weight loss (>20% of preinjected weight) or study timeline criteria. After study termination at 21 days, all mice were killed using somnopentyl, and the tongues, cervical lymph nodes, lungs, and livers were collected. The tissues were fixed with 10% formalin in PBS, and paraffin-embedded sections were prepared. The tumor size was evaluated as the ratio of the tumor area to the whole tongue area in an H&E section prepared from a cut surface that showed the maximum tumor diameter.

| Cell surface HAI-1 immunoreactivity is decreased in cancer cells at the invasion front of tongue squamous cell carcinoma
Initially, we performed an immunohistochemical study of HAI-1 using formalin-fixed paraffin-embedded tissue sections from 42 surgically resected TSCC cases. The cancer cells in the main tumor portion generally showed obvious cell surface HAI-1 immunoreactivity, particularly in well-differentiated carcinomas.
In contrast, infiltrating cancer cells at the invasion front frequently showed decreased membranous HAI-1 immunoreactivity ( Figure 1A,B). We semi-quantitatively assessed the decrease in HAI-1 immunoreactivity at the invasion front (i.e., HAI-1 reduction level) and performed correlation analysis with each clinicopathological parameter. The HAI-1 reduction level was correlated with the presence of lymphatic invasion, as judged by immunohistochemistry of podoplanin (P = 0.008; Figure 1C and Table 1).
Although tumors with a high degree of reduction in HAI-1 levels at the invasion front tended to show less differentiated morphology as well as lymph node metastasis at the time of surgery, the correlations were not statistically significant (P = 0.0721 and 0.0927, respectively; Table 1).
The establishment of a SPINT1-deleted SAS subline (SAS/HAI-1KO) and a mock-transfected control subline (Figure 2A) has been reported previously. 29 We then analyzed the role of HAI-1 in TSCC cell proliferation in vitro. The growth rate of both HSC3 and SAS cells was significantly reduced by the complete loss of HAI-1 ( Figure 2B).
In a colony forming assay, the colony number was not altered by loss of HAI-1, but the colony sizes of HAI-1KO cells were significantly smaller than those of control cells ( Figure 2C).
We next analyzed the effects of HAI-1 loss on the migration and invasion of TSCC cells. Wound healing assays showed that HAI-1 loss significantly enhanced migration of both HSC3 and SAS cells ( Figure 3A). Similarly, in vitro Matrigel invasion was significantly enhanced by the loss of HAI-1 ( Figure 3B).

| Enhanced pericellular proteinase activities in the absence of HAI-1
Hepatocyte growth factor activator inhibitor type-1 insufficiency is reported to result in enhanced activity of matriptase in keratinocytes and cancer cells, including SAS. 26,30 In fact, both HSC3 and SAS cell lines and their sublines expressed matriptase ( Figure 4A), and hydrolysis of the synthetic matriptase substrate Boc-Glu-Ala-Arg-AMC was increased in the absence of HAI-1 ( Figure 4A). As matriptase is known to activate other zymogens, such as pro-uPA, 18-20 and considering the well-established pro-invasive roles of uPA in many cancers, 31 we then analyzed the effects of HAI-1 loss on pericellular uPA activity using the synthetic substrate S-2444. Both HSC3 and SAS cell lines expressed uPA and uPA receptor mRNAs ( Figure 4A), and pericellular S-2444 hydrolysis was significantly enhanced by the loss of HAI-1 ( Figure 4B). Consequently, higher plasmin activity was induced by the addition of exogenous plasminogen in HAI-1KO TSCC sublines compared to control sublines ( Figure S2 and Figure 4C). This induction was inhibited, at least in part, by siRNA-mediated silencing of matriptase ( Figure 4C).

| HAI-1 insufficiency confers activation of growth factors related to lymphangiogenesis
Considering the observed correlation between HAI-1 reduction  33 VEGF-C is secreted as a non-active proform (pro-VEGF-C), and proteolytic activation of pro-VEGF-C is required to transduce its activity through VEGF receptor-3 (VEGFR-3). 34 Plasmin is a wellknown activator of pro-VEGF-C, which generates N-terminal active fragment around 20 kDa. 34,35 The culture supernatants of HAI-1KO sublines cleaved pro-VEGF-C in the presence of plasminogen, generating a presumed active VEGF-C fragment that was similar to that seen for plasmin treatment ( Figure 6C). 34 This activity was barely detectable in the supernatant of control sublines or in the absence of plasminogen ( Figure 6C). In addition, recombinant matriptase did not generate the N-terminal activated fragment ( Figure S3). These observations suggested that HAI-1 loss induced pericellular protease activities that activated HGF and VEGF-C pathways in the tumor microenvironment in vivo, where stromal pro-HGF and serum-derived plasminogen are expected to be present. Moreover, HGF/MET signaling may amplify VEGF-C-mediated lymphangiogenesis via upregulation of VEGF-C expression.

| Orthotopic xenotransplantation of tongue squamous cell carcinoma cells indicated enhanced tumor growth and lymphatic invasion in the absence of HAI-1
Finally, we examined the effect of HAI-1 deficiency on TSCC cells in vivo. Because subcutaneous transplantation of HSC3 in nude mice was unsuccessful in our previous work, 26 here we applied an orthotopic intra-tongue transplantation method. Considering the potential harm and benefit associated with intra-tongue xenotransplantation experiments on mice, mice were killed based on weight loss (>20% of preinjected weight) or study timeline criteria (i.e., 21 days after transplantation; Table 2). The tumor size at the time of sacrifice was significantly larger in both HAI-1KO SAS and HSC3 sublines compared to corresponding control sublines TA B L E 1 Correlation between hepatocyte growth factor activator inhibitor type-1 (HAI-1) reduction level and clinicopathological parameters ( Figure 7A). Unlike SAS, HSC3 cells showed stromal pro-HGFdependent MET phosphorylation, as evidenced by the co-culture study with MRC5 human fibroblasts ( Figure 5B). As murine HGF does not efficiently transduce signals through human MET, 36 we also performed a co-xenotransplantation experiment with HSC3 and MRC5 cells. As expected, co-transplantation of MRC5 promoted MET phosphorylation of HSC3 cells ( Figure 7B). The number of lymphatic invasion foci was significantly increased in HAI-1KO groups compared to the control group, particularly in the model of HSC3+MRC5 co-culture ( Figure 7C). The number of lymph vessels tended to be increased by the loss of HAI-1, which was statistically significant in SAS cells ( Figure S4). With SAS cell transplantation, the incidence of lymph node metastasis at the time of termination (21 days after transplantation) was higher for the HAI-1KO group (9/10) compared to the control group (5/10), although this difference did not reach statistical significance (P = 0.0571). Although loss of HAI-1 may enhance lymph node metastasis of HSC3 cells ( Table 2), 33% and 50%-70% in the mice in HSC3/HAI-1KO group and the HSC3/HAI-1KO+MRC5 co-transplantation group, respectively, were euthanized due to weight loss. Thus, we could not evaluate and compare the precise incidence of lymph node metastasis of the HSC3 sublines. In addition, we observed HAI-1 reduction at the invasion front of control HSC3 cells co-transplanted with MRC5, which likely mimicked the in vivo tumor microenvironment of human TSCC. The HAI-1 reduction level tended to correlate with the frequency of lymphatic invasion ( Figure 7D). Kaplan-Meier survival analysis indicated that the loss of HAI-1 resulted in worse survival when HSC3 cells were co-transplanted with MRC5 cells (Figure 8).

| DISCUSS ION
Hepatocyte growth factor activator inhibitor type-1/SPINT1 is a membrane-anchored Kunitz-type serine protease inhibitor expressed on the surface of normal and neoplastic epithelial cells.  sublines.
Our current study has the following limitations. First, our immunohistochemical study of TSCC tissues included a limited number of patients (42 cases), and there was no formal sample size determination. Therefore, we could not perform rigorous statistical analysis regarding the impact of HAI-1 reduction levels on patients' prognosis.
Second, the precise molecular mechanisms by which loss of HAI-1 induced invasiveness but decreased the growth rate of TSCC cells in vitro remain to be explored. The enhanced invasion observed in HAI-1KO TSCC cells may be caused by matriptase-dependent pro-uPA activation, as silencing of matriptase reportedly inhibits tumor cell invasion through suppression of uPAR-bound pro-uPA activation. 49 The growth disadvantage following HAI-1 loss in vitro was likely compensated in vivo by tumor microenvironmental factors, as HAI-1KO cells showed a tendency of enhanced tumor growth after orthotopic xenotransplantation. Finally, an in vitro lymphangiogenesis assay to confirm VEGF-C activity was not carried out for this study. Nonetheless, the present study provides, for the first time, evidence that HAI-1 insufficiency may contribute to tumor lymphangiogenesis and lymphatic invasion.
In summary, the results of this study indicate that the HAI-1 deficiency in TSCC confers excess pericellular proteolysis that could orchestrate robust activation of the HGF/MET pathway and VEGF-C in the tumor microenvironment where stromal pro-HGF and plasma-derived plasminogen are likely present. Activation of cascades of proteases and growth factors following HAI-1 loss shown in this study likely contributes to the lymphatic spreading of cancer cells and may provide potential targets for innovative therapies for TSCC.

ACK N OWLED G EM ENTS
We thank Ms Junko Kurogi for her excellent technical assistance.
This work was supported by Japan Society for the Promotion of Science KAKENHI 16H05175 (H.K.) and 20K18483 (K.Y.).

D I SCLOS U R E
The authors have no conflict of interest.