Pharmacological Targeting of Midkine (MDK) Reveals Stiffness-Dependent Control of Hepatocellular Carcinoma Invasiveness

Abstract

Metastasis accounts for most cancer-related deaths and hepatocellular carcinoma (HCC) is no exception. Midkine (MDK) is a multifunctional secreted protein elevated in HCC with a vague role in HCC. In this study, we used bioinformatics to verify MDK expression in HCC tumors, and next, we inhibited the MDK protein in invasive Hep3B cells using an MDK inhibitor (iMDK) both in vitro and in vivo. Our results showed that iMDK promoted cell migration and enhanced lamellipodia formation while at the same time downregulating the expression of cell–matrix adhesion genes. In order to also consider forces exerted by the surrounding matrix, we performed cell adhesion, transwell invasion, and 3D tumor spheroid invasion assays in two different stiffness conditions. Adhesion and invasion always exhibited opposite patterns, with adhesion being inhibited in soft matrix environments, accompanied by increased invasion, and a reverse effect in stiff environments. In vivo experiments where cells pre-treated with iMDK were implanted to zebrafish embryos showed overall reduced metastasis, verifying that MDK is a central mechanotransduction regulator that enables HCC cells to adapt their metastatic strategies to ECM stiffness. Thus, MDK inhibition effectively disrupts mechanosensitive coordination during metastasis, highlighting its potential as a therapeutic target.

1. Introduction

Cancer, defined by the uncontrolled proliferation and division of cells [1], remains one of the deadliest diseases globally and poses a significant challenge to public health. Hepatocellular carcinoma (HCC) represents a major form of liver cancer, ranked sixth among the most diagnosed cancer types worldwide and the third leading cause of cancer-related deaths [2,3].

Recent statistics from the United States estimated that in 2024, 41.630 new cases of liver and intrahepatic bile duct cancer were reported, while approximately 29.840 deaths were attributed to these diseases [4]. Notably, the incidence of new HCC cases is projected to increase by 55% over the next 20 years [3]. Given that extrahepatic metastasis is associated with a poorer prognosis [5], and that the therapeutic benefit for metastatic patients remains limited, it is vital to better understand the determinants of metastatic HCC, as it usually reflects the aggressiveness of the primary tumor [6]. Moreover, a deeper understanding of metastasis is essential for developing more effective treatment strategies and improving patients’ lives.

Midkine (MDK), a heparin-binding growth factor, has gained attention due to its overexpression in HCC and its reported association with tumor aggressiveness and poor prognosis [7], suggesting its potential utility as a diagnostic biomarker. While in normal adult tissues, MDK expression is typically undetectable [8], it is found to be highly elevated in infants [9], and most remarkably in various cancers, including HCC, particularly as the tumor progresses to advanced stages [7]. MDK interacts with proteoglycans, syndecans, integrins α4β1 and α6β1, as well as cell–extracellular matrix (ECM) adhesion proteins such as paxillin (PXN) [7]. These receptor interactions trigger the activation of key pro-survival signaling cascades, eventually affecting fundamental cellular processes such as cell proliferation, survival, adhesion, migration, angiogenesis, and epithelial to mesenchymal transition (EMT), a key process during cancer cell metastasis [10,11]. However, the exact role of MDK in metastasis remains unclear.

Therefore, in the present study, we investigated the role of MDK in the metastatic properties of highly invasive HCC cells by blocking it using a known and effective pharmacological inhibitor (iMDK). Metastatic properties of cells were then evaluated using multiple in vitro approaches, both in 2D and 3D culture systems, as well as an in vivo zebrafish model of metastasis.

2. Results

2.1. MDK Expression Is Significantly Upregulated in HCC Tumors Compared to Normal Tissue

Since MDK seems to play a critical role in cancer, we first tested MDK expression using available bioinformatics databases. First, we used the Uniprot [12] and Human Protein Atlas [13] databases to obtain more information about the localization of MDK. As shown in Figure 1A,B, MDK is mostly localized in the extracellular space as well as in intracellular vesicles, which is consistent with a secreted type of protein. We then searched publicly available transcriptomic datasets and cancer gene expression platforms. Specifically, the TNM plot database [14] and OncoDB [15] were used to compare MDK gene expression in HCC and normal liver tissues. Analysis using the OncoDB database showed that MDK was significantly upregulated in HCC samples (n = 371) compared to that in normal liver tissues (n = 50) (Figure 1B). Consistently, RNA-seq data from the TNM plot database also demonstrated higher MDK expression in liver tumor samples (n = 371) than in normal liver tissues (n = 225) (Figure 1C) [14], indicating that MDK is implicated in liver cancer pathogenesis. Moreover, in an effort to better define MDK expression in liver pathology, we used the GepLiver database, which integrates publicly available liver RNA-sequencing datasets [16], and compared MDK mRNA expression in normal liver samples (n = 362) and samples from fibrotic (n = 20) and cirrhotic (n = 73) livers as well as HCC (n = 724). As shown in Figure 1E, MDK mRNA expression was low in normal liver tissue, increased in fibrotic and cirrhotic samples, and markedly elevated in HCC, indicating a progressive increase in MDK expression from normal liver tissue to HCC. Notably, this is also compatible with studies on liver biomechanics reporting that tissue stiffness increases from the healthy liver, which is generally below 6 KPa, to the fibrotic liver, which is around 8–12 KPa, with HCC tissue being much stiffer, often above 20 KPa [17]. In fact, when MDK expression was compared between cirrhotic liver samples (Group A) and HCC samples (Group B) using data from the GepLiver database [16], MDK expression was found to be higher in HCC than in cirrhosis, with median expression values of approximately 6 and 2.5, respectively (Figure 1F). This difference is statistically significant (t-test), indicating increased MDK expression in HCC compared to cirrhotic liver tissue.

2.2. iMDK Promotes Cell Migration in Hep3B Cells

To functionally assess the role of MDK in HCC metastasis, we selected iMDK, a small-molecule cell-permeable pharmacological inhibitor specifically designed to directly bind to the MDK protein and inhibit it [18]. iMDK has been shown to inhibit MDK expression in cancer models such as non-small-cell lung cancer [18,19,20], prostate cancer stem cells [21], and oral squamous cell carcinoma [22], leading to reduced proliferation, migration, and angiogenesis via a yet unidentified mechanism (EMD Biosciences, Inc. (Calbiochem), San Diego, CA, USA). However, little is known about HCC.

In this study, we selected Hep3B HCC cells, which are considered highly invasive, and treated them with 1 μΜ iMDK for 48 h following a series of viability assay experiments using a range of concentrations (0.5–10 μΜ) of iMDK (Supplementary Figure S1). Since iMDK is known to bind directly to the MDK protein and inhibit it, we first performed an immunoblotting experiment to assess the MDK protein expression level. As shown in Figure 2A, MDK protein expression was significantly decreased, indicating that iMDK effectively inhibited the MDK protein.

Next, we tested whether iMDK treatment affected the migratory capacity of Hep3B cells using a standard transwell migration assay. Surprisingly, iMDK treatment significantly enhanced Hep3B cell migration compared to the untreated control (Figure 2B,C), with a quantitative analysis of nine (9) independent experiments confirming a statistically significant increase in cell migration following treatment (Figure 2C).

2.3. iMDK Induces Cytoskeletal Remodeling and Formation of Lamellipodia While Downregulating Cell–Matrix Adhesion-Related Genes

As cell migration is the result of intense cytoskeletal remodeling and includes the formation of actin-rich membrane protrusions such as lamellipodia, we also examined cytoskeletal organization following iMDK treatment. Hep3B cells were stained with fluorescein-labeled phalloidin, known to bind and label filamentous actin (F-actin). As shown in Figure 3, iMDK-treated cells exhibited notable changes with multiple well-developed lamellipodia, indicative of an active migratory phenotype that supports the migration data (Figure 3, see white arrows in the iMDK-treated cells, compare A and B, and higher magnification C and D).

Intrigued by this finding and considering that MDK is associated with integrins and cell–matrix adhesion proteins [23,24], we further investigated the molecular basis of this phenotype by assessing the mRNA expression of key cell–matrix adhesion-associated genes using qRT-PCR. Specifically, we evaluated the mRNA expression of five key focal adhesion genes, namely Integrin-Linked Kinase (ILK), LIM Zinc Finger Domain Containing-1 (LIMS1), Parvin alpha (PARVA), Ras Suppressor-1 (RSU1), and paxillin (PXN). All five genes were significantly downregulated following iMDK treatment (Figure 3E), supporting the migration assay results (Figure 2), as for the cell to be able to migrate, connection to the substrate must be weakened [25].

2.4. Inhibition of MDK Modulates Invasion and Adhesion of Hep3B Cells in a Stiffness-Dependent Manner

Next, we proceeded to test the invasion capacity of cells using the standard transwell invasion assay in which the transwells used were the same as in the migration assay, with the difference being that the porous membrane of the transwell was coated with a layer of diluted Matrigel (1:25 dilution with cold DMEM) [26]. As shown in Figure 4, consistent with our migration findings, iMDK-treated cells displayed a significant increase in invasion compared to control cells (Figure 4A–C).

Although our migration and invasion findings were in agreement with each other, we were still puzzled by the fact that all our experimental results suggested that the inhibition of MDK increases the metastatic properties of HCC cells, contrary to what is known from the literature about other cancer cell types. Thus, before concluding that this is a cell-type-specific response, as occurs in many other cases [27,28], we wondered whether the changes observed in cell invasion are also affected by other factors critical for metastasis, such as matrix mechanical properties and especially stiffness of the surrounding ECM. Since stiffness can be increased by increasing the ECM concentration [29], we repeated the transwell invasion assay with a higher concentration of Matrigel coating conditions (1:10 dilution with DMEM). Indeed, as shown in Figure 4D–F, iMDK-treated cells exhibited significantly reduced invasion compared to control in stiffer ECM conditions.

Interestingly, when Hep3B cells were treated with iMDK and subjected to a cell adhesion assay on substrates of different Matrigel concentrations and consequently different stiffness [29], their adhesive properties also changed. Specifically, iMDK treatment significantly increased cell adhesion on stiffer ECM (Matrigel diluted 1:10), while it decreased cell adhesion in soft environments (Matrigel diluted 1:25) (Figure 4G). This supports the idea that for the cell to invade through the surrounding matrix, cell adhesion needs to be reduced and vice versa [30,31,32].

2.5. Inhibition of MDK Differentially Modulates 3D Tumor Spheroid Invasion Depending on Matrix Stiffness

To further validate our findings using another approach, we employed a tumor spheroid invasion assay [26] to evaluate the effect of MDK inhibition in a 3D setting. To this end, cells were pre-treated with iMDK for 24 h, and then tumor spheroids were generated using the hanging drop method [29]. Spheroids were then embedded into Matrigel that was diluted 1:10 or 1:25 with serum-free DMEM [26]. Pictures were taken at time 0 and 24 h later. As shown in Figure 5, tumor spheroids formed from iMDK-treated cells exhibited increased invasion capacity when embedded in a less stiff matrix (Figure 5A–E), whereas they exhibited significantly decreased invasion capacity when embedded in a stiffer matrix (Figure 5F–J). This is in complete accordance with both cell adhesion and transwell invasion experimental findings.

2.6. iMDK Pre-Treatment Reduces Metastatic Dissemination of Hep3B Cells in Zebrafish Xenografts

To investigate whether MDK inhibition affects the metastatic potential of Hep3B cells in vivo, iMDK pre-treated cells were implanted in Tg(flk1:GFP) zebrafish embryos and compared with untreated control cells at 3 days post-injection (dpi) (Figure 6). Overall, animals injected with iMDK-pre-treated Hep3B cells exhibited significantly fewer metastatic foci than controls (median 2 vs. 5, p = 0.046; Figure 6A–E). Notably, although a subset of iMDK xenografts exhibited higher metastatic foci counts, elevating the group mean, the overall distribution in the group shifted downward, as reflected by the significantly lower median. This indicates that although iMDK does not abolish dissemination in all cases, it consistently reduces the metastatic burden in the majority of xenografts. The reduction in metastatic foci was accompanied by a decrease in the average pixel fluorescence intensity (median 25.0 vs. 27.5, p = 0.039; Figure 6F) and lower average integrated density values (median 2455 vs. 1058, p = 0.0478; Figure 6G), reflecting a diminished overall metastatic burden. The total area of metastatic foci showed important downward trends in the iMDK group (median 465 vs. 104) but did not reach statistical significance (p = 0.09).

Taken together, these data demonstrate that iMDK pre-treatment attenuates both the frequency and cellular density of metastatic deposits, resulting in a measurable reduction in the overall metastatic burden in the zebrafish xenograft model. The primary tumor area at the yolk sac implantation site was not different at 1 dpi in a randomly sampled subset of embryos, but was reduced in the iMDK-pre-treated group at 3 dpi (Supplementary Figure S2). Dissemination analyses were restricted to caudal metastatic foci, with the yolk sac implantation site excluded from the ROI. Representative dual-channel images (Figure 6A,C) further illustrate these differences. In control xenografts, red fluorescent Hep3B foci were frequently located along the caudal vessels, consistent with vascular-associated dissemination. In contrast, iMDK-pre-treated xenografts displayed fewer and smaller foci, which often appeared more isolated from vascular structures. Tumor-only images highlighted these differences in the metastatic burden with greater clarity, complementing the quantitative analyses (Figure 6B,D).

3. Discussion

MDK is increasingly recognized as a multifunctional protein with growth factor/cytokine characteristics that is implicated in tumor progression. However, the contribution of MDK to HCC biology has not been adequately explored. In the current study, we investigated the role of MDK in the metastatic properties of Hep3B cells following pharmacological targeting using iMDK.

Our initial finding that iMDK treatment promoted HCC cell migration and the formation of migration-related actin-rich lamellipodia seems to contradict reports in other cancer types where iMDK inhibited migration, including prostate cancer stem-like cells [21], ovarian (SKOV3) [27], renal (Caki-1), lung (A549), colorectal (HCT116), cervical (CaSki), and breast (BT549) cancer cells [28].

Interestingly, however, our findings regarding cell adhesion as well as 2D and 3D invasion assays indicated a response that is greatly dependent on matrix stiffness. Specifically, iMDK differentially modulated HCC cell adhesion depending on matrix stiffness while also affecting known cell–ECM adhesion proteins ILK, LIMS1, RSU1, PARVA, and PXN mRNA levels.

Moreover, both the 2D transwell invasion assay and the 3D tumor spheroid assay showed that iMDK treatment significantly enhanced invasion in soft ECM environments, while it had an inhibitory effect in stiffer conditions, consistent with previous reports in ovarian (SKOV3) [27], renal (Caki-1), lung (A549), colorectal (HCT116), and breast (BT549) cancer cells [28].

Finally, our experiment in the zebrafish xenograft model of metastasis revealed that pre-treatment of Hep3B HCC cells with iMDK significantly reduced metastatic dissemination, as shown by fewer micrometastatic foci, lower average fluorescence intensity, and decreased integrated density per focus in the majority of fish despite inter-individual heterogeneity. The zebrafish embryo xenograft model enables functional assessment of human HCC cell dissemination and metastatic burden in a living vertebrate system and was therefore selected to address early metastatic events relevant to human cancer biology, particularly in soft tissue environments. Moreover, our data highlight the utility of zebrafish embryos for the functional assessment of micrometastasis, since disseminated foci detected in the caudal region are consistent with processes including escape from the implantation site, survival during transport, arrest/extravascular transition, and early niche establishment at distant sites, serving as a proxy for early metastatic events rather than established secondary tumors. However, because our analysis is based on foci quantification at a single terminal time point and does not distinguish intra- versus extravascular localization, the present data do not allow attribution of the observed reduction to a specific step of the metastatic cascade. More importantly, zebrafish embryonic tissues exhibit sub-kilopascal stiffness values, typically between 50 and 500 Pa, as determined by atomic force microscopy [33,34], which places the zebrafish xenograft model within the range of soft extracellular environments, comparable to the low-stiffness (1:25 Matrigel) condition in vitro. However, it should be noted here that it is often difficult and sometimes impossible to directly relate values measured in vivo to those obtained ex vivo, or even more so to those from in vitro matrix models [35,36,37].

Hence, our findings indicate that MDK plays a complex role in regulating cellular dynamics critical for metastasis, including migration, adhesion, and invasion, and the final outcome with regard to metastasis is determined by the balance between adhesion and invasion, always in relation to the stiffness of the surrounding matrix. This explains why the inhibition of MDK in the stiff matrix promotes cell adhesion and inhibits cell invasion, while the effect in soft matrix environments is the opposite (Figure 7A). Notably, however, real metastasis in vivo involves many steps, several of which include cell adhesion to the surrounding matrix or tissue and local invasion (Figure 7B), and of course, additional microenvironmental factors, such as hemodynamic forces and immune interactions, which limit cancer cell dissemination in vivo. Additionally, cancer cells in vivo need to invade multiple tissue types to effectively metastasize, and each of those may have different stiffness.

Overall, although performed in one cell line, our study has utilized many different experimental approaches, both in vitro and in vivo, and has revealed the MDK protein as a central mechanotransduction regulator that enables HCC cancer cells to adapt their metastatic strategies to ECM stiffness. Of course, more studies need to be performed in other cell lines, other cancer types, and human samples to make a more generalized conclusion, but the current study has demonstrated that although the inhibition of MDK may transiently enhance migration and invasion under certain conditions, its overall effect impairs the coordinated processes required for successful metastatic dissemination in vivo. This suggests that disrupting mechanosensitive signaling is a promising therapeutic approach to eliminate cancer dissemination across diverse tissue environments.

4. Materials and Methods

4.1. Cell Culture

Hep3B cells were obtained from the American Type Culture Collection (HB8064) and cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% Fetal Bovine Serum (FBS) and 1% antibiotic/antimycotic and incubated at 37 °C in a humidified incubator with 5% CO₂.

4.2. MDK Inhibitor (iMDK) Treatment

iMDK was purchased from EMD Biosciences, Inc. (Calbiochem), San Diego, CA, USA (cat. #5080520001) and dissolved in dimethyl sulfoxide (DMSO). Hep3B cells were seeded in 6-well plates and treated with 1 μΜ iMDK for 48 h.

4.3. Western Blotting

For the analysis of protein expression, a standard immunoblotting protocol was used, as described previously [29]. Total cell lysates were prepared using RIPA buffer containing 1% sodium dodecyl sulfate (SDS). Protein concentration was determined using the Bradford assay, and 40 μg of protein per sample was loaded on a 15% SDS-polyacrylamide gel to be transferred to PVDF membranes (Merck Millipore, Burlington, MA, USA) using a Bio-Rad wet transfer system. Membranes were blocked with 5% non-fat milk in Tris-Buffered Saline containing 0.1% Tween-20 (TBST) for 1 h at room temperature and incubated overnight at 4 °C with primary antibodies diluted in 5% milk/TBST (1:1000 for anti-MDK and 1:500 for anti-β-actin). Anti-MDK antibody (cat. #PA5-115560, Thermo Fisher Scientific Inc., Waltham, MA, USA) and anti-β-actin antibody (cat. #sc-47778, Santa Cruz Biotechnology Inc., Dallas, TX, USA) were used. After washing, membranes were incubated with the appropriate HRP-conjugated secondary antibodies. Protein bands were detected using the SuperSignal West Femto Maximum Sensitivity Substrate (cat. #ZJ399323, Thermo Fisher Scientific, Inc., Waltham, MA, USA) and visualized using a Bio-Rad Gel Documentation System (Image lab 5.0).

4.4. RNA Isolation and Real-Time PCR

Total RNA was extracted from Hep3B cells using the QIAzol Lysis Reagent (QIAGEN, N.V, Venlo, The Netherlands). RNA concentration and purity were assessed using a Nanodrop spectrophotometer (IMPLEN GmbH, Munich, Germany). cDNA synthesis was performed using Superscript Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s guidelines. The resulting cDNA was diluted to 1:10 and used as a template for real-time PCR using KAPA SYBR Green FAST Master Mix (Roche, F. Hoffmann–La Roche AG, Basel, Switzerland). Reactions were performed on a CFX96 thermal cycler (Bio-Rad Laboratories, Inc., Hercules, CA, USA), with β-actin as the reference gene. All reactions were performed in triplicate, and the data were derived from at least three independent experiments. Primer sequences are listed in

Table 1. Primers sequences used in real-time PCR.

Gene Name	Primer Sequence
ILK	Forward: 5′ GAC ATG ACT GCC CGA ATT AG 3′ Reverse: 5′ CTG AGC GTC TGT TTG TGT CT 3′
LIMS1	Forward: 5′ CCG CTG AGA AGA TCG TGA AC 3′ Reverse: 5′ GGG CAA AGA GCA TCT GAA AG 3′
PARVA	Forward: 5′ CAA TTC GAC TCC CAG ACC AT 3′ Reverse: 5′ TGG TCG AAC AAG GTG TCA AA 3′
RSU1	Forward: 5′ AGG CCA CAG AGC AAG GTC TA 3′ Reverse: 5′ CGT GCA ATC TCA AAA GCT CA 3′
PXN	Forward: 5′-ACGTCTACAGCTTCCCCAACAA-3′ Reverse: 5′-AGCAGGCGGTCGAGTTCA-3′
β-actin	Forward: 5′ CGA GCA CAG AGC CTC GCC TTT GCC 3′ Reverse: 5′ TGT CGA CGA CGA GCG CGG CGA TAT 3′

. Relative gene expression was calculated using the ΔΔCt method [38], while cells treated only with DMSO served as a calibrator.

4.5. Cell Migration Assay

A cell migration assay was performed using transwell chambers with 8 μm pore-size membranes, as described previously [26]. A total of 1 × 10⁵ cells suspended in serum-free DMEM were seeded in the upper chamber, and 600 μL of complete DMEM culture medium containing 10% FBS was added to the lower chamber. Hep3B cells were pre-treated with iMDK 24 h prior to seeding in the transwell and were left to migrate for another 24 h in serum-free DMEM with iMDK. Non-migrating cells were then removed from the upper surface of the transwell membrane using a cotton swab and cells on the bottom side were fixed in 4% paraformaldehyde (PFA) for 20 min and stained with 0.1% crystal violet for 20 min. Transwells were then washed multiple times with ddH₂O, and images were taken from 5 randomly selected optical fields under a Motic AE2000 microscope (Motic, Xiamen, China) equipped with the Moticam 5+ 0.5 MP camera (Motic, Xiamen, China). The average number of migrating cells per transwell was calculated, and at least three independent experiments were performed per condition.

4.6. Cell Invasion Assay

A cell invasion assay was performed similarly to the migration assay [26], with the difference that transwell membranes were coated with Matrigel (CORNING) diluted 1:10 and 1:25 in cold plain DMEM one hour prior to cell seeding, and cells were treated with iMDK for 24 h and then left to invade through the Matrigel layer for another 30 h. After incubation, non-invaded cells remaining on the upper surface of the membrane were removed with a cotton swab moistened with Phosphate-Buffered Saline (PBS) while the invaded cells were fixed and stained as described above. Quantification was performed in the same manner as in the migration assay.

4.7. Cell Adhesion Assay

Hep3B cells were cultured in a 6-well plate until they reached 70% confluence. The following day, cells were treated with iMDK for 48 h. Cells were then detached using trypsin, counted, and seeded at a density of 5 × 10⁴ cells per well into a 96-well plate pre-coated with Matrigel (CORNING Inc, New York, NY, USA), diluted 1:10 and 1:25 in cold plain DMEM to be subjected to a cell adhesion assay [39]. The plate was incubated for 3 h at 37 °C. Following the incubation period, the medium was carefully removed, and the wells were washed twice with PBS to remove non-adherent cells. Adherent cells were fixed and stained with a solution containing 10% ethanol and 0.2% crystal violet for 20 min at room temperature. Absorbance was measured at 570 nm using a Varioskan LUX Multimode Plate Reader (Thermo Fisher Scientific, Waltham, MA, USA).

4.8. Cell Staining

Cells were seeded in a 6-well plate at a 70% confluency, onto glass coverslips pre-coated with 0.1% gelatin. The following day, cells were treated with iMDK and incubated for 48 h at 37 °C. After treatment, cells were washed with PBS and fixed in 4% PFA for 20 min at room temperature. Fixed cells were permeabilized with a buffer containing 0.1% Triton X-100 and 2 mg/mL of Bovine Serum Albumin (BSA) in PBS for 35 min. F-actin was stained using Phalloidin Fluorescein conjugate (BIOTIUM, Inc. Fremont, CA, USA) for 45 min at room temperature in the dark. Coverslips were then mounted on microscope slides using a DAPI-containing mounting medium (ibidi, GmbH Gräfelfing, Gräfelfing, Germany). Fluorescence imaging was performed using a Zeiss LSM900 confocal microscope (Zeiss AG, Oberkochen, Germany) with 20X and 63X objective lenses.

4.9. Tumor Spheroid Invasion Assay

Tumor spheroids were generated using the hanging drop method [29]. Cell suspension was prepared at a concentration of 2.5 × 10⁴ cells/mL, and 20 μL drops (containing 500 cells each) were placed on the inner surface of a culture dish lid and incubated at 37 °C for 24 h. After spheroid formation, the spheroids were carefully transferred into the wells of a 96-well plate using P20 plastic tips. Each spheroid was transferred to a single well of a 96-well plate, and diluted Matrigel (CORNING) (1:10 or 1:25, as indicated) in serum-free cold DMEM was added on top. Spheroids were then allowed to invade the Matrigel layer for 24 h. Images were captured at time zero, and the number of wells corresponding to each image was recorded. After 24 h, images were captured from the same well to assess cell invasion. For experiments involving iMDK treatment, the inhibitor was added at the time of the spheroid formation. All photos were taken using a Motic AE2000 microscope (Motic, Xiamen, China) equipped with a Moticam 5+ 0.5MP camera. Cell invasion through the surrounding Matrigel was measured using Fuji/ImageJ software (Wayne Rasband & National Institute of Health (NIH), Bethesda, MD, USA; version 1.54p), and the final spheroid size (average of the major and minor axis length) was compared to the initial size at time zero, and the % difference was taken, as described previously [26]. At least 13 spheroids were analyzed per condition, and at least two independent experiments were conducted.

4.10. Cell Pre-Treatment and Staining Prior to In Vivo Experiments in the Zebrafish Model

Following treatment of Hep3B cells with 1 μΜ iMDK for 48 h, cells were detached with 0.05% Trypsin–EDTA, washed twice with PBS, and labeled with the red fluorescent dye CellTracker CM-DiI (Thermo Fisher Scientific, C7000), according to the manufacturer’s instructions. Labeled cells were resuspended in PBS containing 0.025% EDTA to minimize aggregation. The final suspension was adjusted by dilution to yield ~300 cells per 4.2 nL of injection volume.

4.11. Zebrafish Housing and Handling

All animal housing, maintenance, and breeding procedures followed established protocols as described in the RSPCA guidelines [40] and were reported in accordance with ARRIVE guidelines. Transgenic Tg(flk1:GFP) zebrafish (Danio rerio; Wik strain), which enable visualization of blood and lymphatic vessels via eGreen Fluorescent Protein (eGFP) expression under the fli1 promoter [41], were obtained from bleached eggs provided by the European Zebrafish Resource Centre (Institute of Biological and Chemical Systems, Eggenstein-Leopoldshafen, Germany). Embryos and adults were maintained at the University of Cyprus zebrafish facility, licensed by the Veterinary Services of the Ministry of Agriculture, Rural Development, and Environment of the Republic of Cyprus (CY/EXP/111). Fish were housed in an automated ZebTEC StandAlone system with ‘Active Blue’ recirculating water technology (Techniplast, Buguggiate, Italy) under optimal conditions (28.5 °C, pH 7.5, conductivity 550 μS) and a 14:10 h light/dark photoperiod. Feeding was performed twice daily with formulated diets rich in soluble hydrolyzed marine proteins, HUFAs, phospholipids, and algae (GEMMA 300 micro for adults; GEMMA 75 micro for larvae, Skretting Ltd., Stavanger, Norway), supplemented with freshly hatched Artemia nauplii (Ocean Nutrition Ltd., Somerset West, South Africa) two days prior to breeding events. Breeding was carried out in on-bench breeding tanks (Techniplast, Italy) using system water at 28 °C. Eggs were collected 1–2 h after the onset of the light cycle, and fertilized eggs were selected and incubated in system water containing 0.00005% v/v methylene blue for ~72 h until hatching. All procedures complied with the European Directive 2010/63/EU for animal experimentation and the relevant national legislation.

4.12. Zebrafish Xenografting

On day 2 of post-fertilization (2 dpf), Tg(flk1:GFP), zebrafish embryos were manually dechorionated using fine forceps and anesthetized with 0.02% tricaine (MS-222) prior to implantation. The embryos were randomly allocated to a control group (injected with untreated Hep3B cells) or an experimental group (injected with iMDK-pre-treated Hep3B cells). Approximately 300 CM-DiI-labeled Hep3B cells were injected into the yolk sac of each embryo in a 4.2 nL volume using custom-pulled borosilicate glass capillary needles (Sutter Instruments, Novato, CA, USA) mounted on a manual micromanipulator and a microinjector system. Successful loading of cell suspensions into the needle and injection placement were verified under a Nikon SMZ800N (Nikon, Tokyo, Japan) fluorescence stereomicroscope equipped with FITC (GFP) and TRITC (CM-DiI, Thermo Fisher Scientific, Waltham, MA, USA) filter sets. Following injection, embryos were allowed to recover overnight at 28 °C before being transferred to 34 °C for the remainder of the experiment. At day 1 post-injection (1 dpi), embryos were screened under fluorescence microscopy to confirm the correct localization of implanted cells. Embryos showing leakage or disseminated fluorescence signals outside the yolk sac (e.g., circulation-based dispersal) were excluded. The primary tumor area at 1 dpi did not differ significantly between groups in a randomly sampled subset of embryos (control median 40,968 px; iMDK mean 44,197 px; Mann–Whitney test with exact p = 0.097). After screening, 27 embryos remained in the control group and 51 in the experimental group. At the end of the experiment (3 dpi), surviving embryos were imaged, then euthanized by tricaine overdose (0.05% MS-222) and fixed in 100% methanol at −20 °C for 5 days prior to disposal. No specific humane endpoints were defined beyond predefined experimental termination; embryos were monitored daily for general morphology, viability, and signs of distress, and experiments were terminated at 3 dpi. No expected or unexpected adverse events were observed during the course of the zebrafish experiments.

4.13. Zebrafish Imaging and Analysis

For imaging at 3 dpi, zebrafish xenografts were anesthetized and immobilized with 6% methylcellulose. Whole-body images were captured for every zebrafish at the green (for vasculature) and red (for tumor) fluorescence channels, as well as in the brightfield (for topology confirmation). Quantitative analysis of the images was performed using the Fuji/ImageJ software (Wayne Rasband & NIH, USA; version 1.54p). The regions of interest (ROIs) were drawn in the caudal region (posterior to the cloaca) while explicitly excluding the primary yolk sac tumor. Automated macros were used to standardize the data extraction across all images and experimental groups. After thresholding and applying a minimum area filter, the following metrics were extracted per zebrafish: (a) Metastatic Foci (count): number of discrete fluorescent foci within the caudal ROI, reflecting the frequency of disseminated metastatic foci per zebrafish, as a proxy of metastatic incidence. (b) Total Area (pixels): the cumulative area of all metastatic foci, reflecting the cumulative size of metastatic lesions, providing a measure of metastatic burden in terms of area. (c) Integrated Density: the sum of all pixel intensities inside ROIs summed per zebrafish (equivalent to Total Area × Mean Gray Value) after subtraction of the background signal, reflecting the overall metastatic fluorescent burden and used as a composite measure of total metastatic load and an indirect proxy of relative cell density within metastatic foci. (d) Average Pixel Fluorescence Intensity: the average of mean pixel intensity values (fluorescent signal) across all metastatic foci, proportional to the number of fluorescent cells in the foci. (e) Average Integrated Density: calculated by the ratio of the integrated density over the number of metastatic foci, reflecting the typical metastatic burden per focus, integrating size and intensity to approximate the average lesion load.

4.14. Statistical Analysis

All in vitro data presented in this work followed a normal distribution and therefore were statistically analyzed using excel while the t-test was used for comparison of means between two groups (control vs. treated cells). In all cases, a p-value < 0.05 was considered statistically significant and denoted by an asterisk in the graphs. For the in vivo experiments, statistical analyses were performed using GraphPad Prism version 8.0.1 (GraphPad Software, San Diego, CA, USA). Data were first assessed for normality using the D’Agostino–Pearson and Shapiro–Wilk tests. As most datasets did not follow a normal distribution, comparisons between groups were performed using the non-parametric Mann–Whitney U test. For normally distributed datasets, unpaired two-tailed Student’s t-tests were applied. Bars represent mean values ± values ± standard error of the mean (SEM) with individual data points overlaid. Statistical comparisons were performed using the Mann–Whitney test on distributions, as the data were not normally distributed. Statistical significance was defined as p < 0.05. In the figures, significance is indicated by (*). Sample sizes (n) refer to the number of individual fish analyzed and are indicated in the figure legends.

4.15. Bioinformatics Analysis

Bioinformatics analyses were performed to investigate the subcellular localization and gene expression profile of MDK. Specifically, MDK subcellular localization was assessed using the UniProt database [12] to obtain curated protein annotation data and the Human Protein Atlas to evaluate experimentally supported protein localization patterns in human tissues and cells. Gene expression analyses were conducted using multiple RNA sequencing-based resources. Differential MDK mRNA expression between HCC and normal liver tissues was first analyzed using the OncoDB database [15], including 371 HCC samples and 50 normal liver samples. Independent validation was performed using the TNMplot database [14], comparing RNA-seq data from 371 primary HCC tumors and 225 normal liver tissues. In addition, MDK mRNA expression across disease progression stages was examined using RNA-seq datasets available through the GepLiver database [16], encompassing normal liver tissues (n = 362), fibrotic liver samples (n = 20), cirrhotic liver samples (n = 73), and HCC samples (n = 724). Statistical comparisons between cirrhotic liver samples (Group A) and HCC samples (Group B) were conducted within the GepLiver dataset, and significance was defined as p < 0.05. All datasets were accessed and analyzed using the default processing and normalization pipelines provided by each database, with data last accessed in February 2026.