The human renal cancer cell line Caki-1, human lung cancer cell line A549, human colorectal adenocarcinoma cell line HCT116, cervical cancer cell line CaSki, human breast cancer cell line BT549 and human embryonic kidney cell line HEK293T were acquired from National Collection of Authenticated Cell Cultures (Beijing, China). Caki-1, A549 and HCT116 cells were maintained in McCoy’s 5A medium modified (KeyGEN BioTECH, Jiangsu, China). CaSki and BT549 cells were cultured in RPMI-1640 medium (SIGMA, Vienna, Austria). HEK293T were maintained in DMEM medium (Gibco, California, USA). All mediums contained 10% fetal bovine serum (FBS, Ausbian, Australia) and 1% penicillin–streptomycin mixture (Solarbio, Beijing, China). All cell lines were incubated at 37°C with 5% CO₂ in a humidified incubator.

MDK antibody (1:1000 dilution, 11009-1-AP) was purchased from Proteintech Group Inc (Rosemont, IL, USA). E-cadherin antibody (1:1000 dilution, 3195T), ZO1 antibody (1:1000 dilution, 8193T), Vimentin antibody (1:1500 dilution, 5741T) and Snail antibody (1:500 dilution, 3879T) were from Cell Signaling Technology (Danvers, Massachusetts, USA). STAT1 antibody (1:5000 dilution, ab109320) and phospho-STAT1 (phosphor Y701) antibody (1:1000 dilution, ab109457) were purchased from Abcam (Cambridge, UK). β-actin antibody (1:1000 dilution, AF5001) was from Beyotime Biotechnology (Shanghai, China). Recombinant human IFN-γ was acquired from PeproTech (Rocky Hill, NJ, USA) and used at a concentration of 50 ng/ml for 48h. The MDK inhibitor iMDK was bought from Merck Millipore (Darmstadt, Germany) and diluted to a final concentration of 100nM in growth mediums. The STAT1 inhibitor Fludarabine was purchased from Med Chem Express (New Jersey, USA) and used at a concentration of 5 μM for 48h.

Total RNA was extracted using TRIeasy reagent (Yeasen Biotechnology, Shanghai, China) according to the manufacturer’s instruction. The concentration and purity of RNA were determined by Nanodrop 1000. The ratio of OD 260/OD 280 falling between 1.8 and 2.0 indicates acceptable values. Reverse transcription was achieved with a total of 1 μg RNA using the cDNA Synthesis SuperMix (Yeasen Biotechnology, Shanghai, China). The primers (5′-3′) synthesized by Rui Biotech (Beijing, China) were listed in Table S1 . QRT-PCR was performed using qPCR SYBR Green Master Mix (Yeasen Biotechnology, Shanghai, China). Each sample was run in 3 duplicate wells, and the relative RNA expression levels between different samples were analyzed by 2 ^–ΔΔCt method, with β-actin as an internal control.

The human MDK overexpression lentiviral vector (YOE-LV004-hMDK) with neomycin resistance and the corresponding control vector (YOE-LV004-Ctrl) were purchased form UBIGENE (Guangzhou, China). Lentivirus was produced in HEK293T cells by co-transfecting the lentiviral vector with the psPAX2 and pMD2G packaging vectors using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA). Caki-1, A549, HCT116, CaSki and BT549 cells were infected with the packaged lentiviruses with polybrene, and then selected with G418 (Solarbio, Beijing, China) at a concentration of 400 μg/ml. The empty plasmid (YOE-LV004-Ctrl) was used as a control.

Transwell assays were used to assess the invasion and migration abilities of cancer cells. Briefly, 1×10⁵ cells in serum-free medium were seeded into the upper chamber of transwell plates (Corning, NY, USA), while the lower chamber was added with 10% FBS medium. For invasion assay, the transwell chambers were pre-coated with Matrigel (BD Biosciences, Palo Alto, CA). Approximately 16-24 h after seeding, the cells remaining on the upper chamber were carefully wiped off with cotton swabs, while the migrated cells were fixed in 4% paraformaldehyde for 20 minutes and then stained in 0.2% crystal violet solution for 20-25 minutes. After cleaning the chamber with wash buffer for three times, the migrated cells were counted under an inverted bright-field microscope.

Cell lysates were prepared in RIPA lysis buffer (Beyotime Biotechnology, Shanghai, China) supplemented with 1x protease inhibitor cocktail (Beyotime Biotechnology) and 1x phenylmethylsulfonyl fluoride (PMSF, APPLYGEN, Beijing, China). Protein concentrations were measured with a BCA assay kit (KeyGEN BioTECH, Jiangsu, China). The protein samples (50 μg) were boiled at 100°C for 10 minutes, separated by 10-12% SDS-PAGE, and then transferred onto PVDF membranes (Merck Millipore, MA, USA) via semi-dry transfer unit. After blocked in 5% skim milk at room temperature for 1h, the membranes were incubated overnight at 4°C with primary antibodies, and then incubated with anti-rabbit or mouse IgG-HRP secondary antibody for 2h at room temperature. Membranes were visualized using Baygene Chemilmaging system (Baygene Biotech, Beijing, China) with Super ECL Detection Reagent (Yeasen Biotechnology, Shanghai, China).

All statistical analysis was performed in GraphPad Prism 8.0 (GraphPad Software, La Jolla, CA, USA). Unpaired and multiple T-test were used to assess the differences between control and experimental groups. P value < 0.05 was considered statistically significant. The results were presented as mean ± standarddeviation (SD). *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

Interferon-gamma (IFN-γ) is a pleiotropic cytokine with antiproliferative, pro-apoptotic and immunomodulatory functions. IFN-γ has been used to treat a variety of malignancies in pre-clinical and clinical trials; however, limited benefits have been achieved, presumably due to its pro-tumor adverse effects, including eliciting metastasis (23). Prior to deciphering the mechanism underlying IFN-γ-induced cancer metastasis, we first verified the pro-metastatic effect of IFN-γ in five cell lines of different human cancers ( Figure 1A ): the kidney cancer cell line Caki-1, the lung cancer cell line A549, the cervical carcinoma cell line CaSki, the breast cancer cell line BT549, and the colon cancer cell line HCT116. Transwell assays demonstrated that IFN-γ exposure obviously increased the migration and invasion abilities of all the five cancer cell lines ( Figure 1A ). This is in line with the previous findings in colon cancer (16), prostate cancer (17), non-small cell lung cancer (20), and melanoma (19), suggesting that IFN-γ treatment could globally enhance metastasis of cancers.

EMT is a common mechanism driving metastasis of cancers (24). We then examined the role of IFN-γ in the EMT programme. Western blotting results showed that IFN-γ treatment led to decreased expression of epithelial markers, including ZO-1 and E-cadherin, and concomitant upregulation of mesenchymal markers, such as Vimentin, Snail and Slug in the above cell lines at the protein level ( Figure 1B ). The alterations in expression of EMT markers were further verified at the mRNA level ( Figure 1C ). These data indicate that IFN-γ may promote metastasis by activating the EMT programme in cancers.

MDK is a heparin-binding growth factor well-documented to promote EMT and cancer metastasis (25). Though there is no evidence suggesting the regulatory relationship between IFN-γ and MDK in the cancerous context, we found a previous report indicating that the MDK expression in lymphocytes is upregulated in response to IFN-γ treatment (26). This gave rise to the speculation that whether IFN-γ treatment in cancers would result in MDK activation, thereby triggering EMT and metastasis. To verify this hypothesis, we treated different cancer cell lines with IFN-γ (50 ng/ml) and then examined the alterations of MDK expression in the mRNA and protein levels. Real-time qPCR assays demonstrated a dramatic upregulation of MDK mRNA by IFN-γ in all examined cancer cell lines ( Figure 2A ), including the kidney cancer Caki-1, the lung cancer A549, the cervical carcinoma CaSki, the breast cancer BT549, and the colon cancer HCT116, which was further validated by Western blotting at the protein level ( Figure 2B ), suggesting that this is a shared regulation across different types of cancers.

To further validate the IFN-γ-MDK regulatory axis, we treated the HCT116 cancer cell line with different concentrations of IFN-γ and verified that activation of MDK by IFN-γ was dose-dependent at both the mRNA ( Figure 3A ) and protein ( Figure 3B ) levels. Furthermore, IFN-γ treatment also caused an obviously time-dependent upregulation of MDK at both the mRNA ( Figure 3C ) and protein ( Figure 3D ) levels, as well as time-dependent activation of EMT evidenced by time-dependent downregulation of E-cad and ZO-1 but upregulation of Slug and Snail ( Figures 3D, E ).

STAT1 is a key downstream effector in the IFN-γ signalling (2), which leads us to speculate that the IFN-γ-MDK regulation may rely on IFN-γ-induced STAT1 activation. This speculation was preliminarily supported by the significant correlations between STAT1 and MDK in various cancers in the TCGA database ( Figure 4A ). Moreover, as reported, real-time qPCR and Western blotting assays showed that IFN-γ exposure caused a significant upregulation in levels of STAT1 abundance and phosphorylation ( Figure 4B ). Then we blocked STAT1 activation using a specific STAT1 inhibitor Fludarabine and assessed its influence on IFN-γ activation of MDK. Notably, STAT1 inhibitor remarkably diminished IFN-γ-induced STAT1 activation, and substantially abrogated IFN-γ-induced MDK activation at the mRNA level in all examined cell lines ( Figure 4C ), which was further validated at the protein level by Western blotting analyses ( Figure 4D ). All these demonstrate that IFN-γ activates MDK via STAT1 in cancer cells.

To further explore the possibility that MDK is a shared mechanism underlying IFN-γ-triggered metastasis in various cancers, we first examined the effect of MDK on cancer metastasis in the above cancer cell lines. Transwell assays clearly showed that MDK overexpression ( Figure 5A ) enhanced the migration and invasion capacities of all five cancer cell lines ( Figure 5B ). In accordance, MDK overexpression resulted in activation of the EMT programme, as evidenced by loss of epithelial markers, and concomitant gain of mesenchymal markers at the both protein ( Figure 5C ) and mRNA ( Figure 5D ) levels. To further validate that MDK is necessary for cancer metastasis, we silenced MDK expression using a specific MDK inhibitor iMDK, which demonstrated high efficiency in decreasing endogenous MDK expression ( Figure 5E ) and subsequent EMT activation in all five cell lines ( Figure 5F ). In accordance, iMDK suppressed the migration and invasion in all five cell lines in transwell assays ( Figure 5G ). In line with its oncogenic role, MDK is elevated in the majority of cancer types in the TCGA dataset ( Figure S1 ).

To test whether MDK inhibition would attenuate IFN-γ-induced metastasis, we added iMDK (100 nM) to the IFN-γ-treated cancer cells, which resulted in no conspicuous cell death under microscope. As expected, iMDK efficiently eliminated IFN-γ-induced MDK expression at both the mRNA and protein levels ( Figures 6A, B ), reversed IFN-γ-driven EMT activation determined by Western blotting and RT-qPCR ( Figure 6B ), and subsequently abrogated IFN-γ-triggered migration and invasion as shown in transwell assays ( Figure 6C ) in all examined cancer cell lines. These data suggest that MDK confers the IFN-γ-elicited metastasis in various cancers, and that pharmacologically inhibiting MDK can broadly and efficiently abrogate IFN-γ treatment-induced cancer metastasis.