Mesenchymal Stem Cell–Mediated Delivery of an Oncolytic Adenovirus Enhances Antitumor Efficacy in Hepatocellular Carcinoma

Liver cancer is the third most common tumor type and the second leading cause of death in Asia. The most common form of liver cancer, hepatocellular carcinoma (HCC), lacks effective treatment options (1, 2). Only a small fraction of patients with HCC can undergo surgical resection and tumor recurrence occurs in more than 50% of these patients (3). Chemotherapeutics, such as sorafenib, elicit limited therapeutic effect against advanced HCC (4, 5). Therefore, there is an urgent need for new HCC therapy.

Mesenchymal stem cells (MSC) have emerged as a promising systemic delivery tool for cancer therapeutics due to their tumor-homing abilities (6, 7). Furthermore, MSCs exhibit relatively low immunogenicity (8) and do not stimulate lymphocyte proliferation, thus avoiding immune rejection (9). Despite these beneficial attributes, MSCs cannot control tumor growth (10). To address this problem, MSCs can be genetically engineered to express therapeutic genes (11, 12), loaded with drugs (13), or be infected with oncolytic viruses (14, 15).

Oncolytic adenovirus (oAd), which exhibits cancer-specific replication, viral production, and cytopathic effects, is a promising therapeutic cargo to enhance the overall therapeutic efficacy of MSC (16, 17). When oAd is utilized as the therapeutic cargo, MSCs are eventually lysed by oAd replication, thus preventing any adverse side effects that are associated with prolonged survival of stem cells in vivo. Importantly, MSC-mediated systemic delivery of oAd can improve intratumoral accumulation of the virus, which is not feasible with naked virion. In specific, the viral capsid, which is easily recognized by immune system as pathogen, can trigger adverse inflammatory responses and swift clearance of oAd in bloodstream, thus resulting in poor intratumoral accumulation (18, 19).

HCC is known for its active Wnt signaling, which contributes to the rapid and uncontrolled proliferation of tumor cells (20). An activated Wnt signaling pathway promotes the epithelial-to-mesenchymal transition (EMT), which contributes to the metastatic and drug-resistant phenotypes of cancer (21). Thus, to enhance the therapeutic efficacy of oAd against HCC, we inserted a sequence encoding a Wnt-inhibiting decoy receptor (WNTi) into the previously reported oAd that targets alpha-fetoprotein (AFP)-positive HCCs (22), generating a WNTi-expressing and HCC-targeting oAd (HCC-oAd-WNTi).

In this report, HCC-oAd-WNTi was loaded into bone marrow–derived human MSCs (HCC-oAd-WNTi/MSC) and evaluated whether this delivery method would reduce the liver toxicity of systemically administered oAd while improving its accumulation in tumor tissues. We illustrate that systemically administered HCC-oAd-WNTi/MSC efficiently targets orthotopic HCC tumors, thus enhancing the tumor accumulation and antitumor efficacy of oAd. Furthermore, HCC-oAd-WNTi/MSC overcomes well-known limitations of systemically administered oAd, such as strong immunogenicity, short blood circulation time, nonspecific liver sequestration, and hepatotoxicity in vivo.

Cell lines HEK293, A549, Hep3B, and BJ were purchased from the ATCC. Cells were cultured in DMEM (Gibco BRL) supplemented with 10% FBS (Gibco BRL), penicillin (100 IU/mL), and gentamicin (20 μg/mL). All cell lines were maintained at 37°C in a humidified incubator with an atmosphere containing 5% CO₂. Cryopreserved human MSCs, which were isolated from bone marrow samples after aspiration from healthy adult male donors, were provided by Pharmicell Co., Ltd. MSCs were cultivated in low-glucose DMEM.

To construct a firefly luciferase–expressing HCC-targeting oAd vector, a firefly luciferase or WNTi expression cassettes were ligated into an Ad E3 shuttle vector (pSP72ΔE3; refs. 23, 24). These E3 shuttle vectors went through homologous recombination with linearized HCC-oAd (Ha2bm-d19-k35; ref. 22) to generate Ha2bm-d19-k35/Luc (HCC-oAd-Luc) or Ha2bm-d19-k35/WNTi (HCC-oAd-WNTi), respectively. Upon generation of viral particles (VP) in HEK293 cells by transfection, HCC-oAd-Luc and HCC-oAd-WNTi were propagated in A549 cells and purified by CsCl gradient centrifugation. The numbers of VPs were calculated from optical density measurements at 260 nm (OD₂₆₀), where an absorbance of 1 (OD₂₆₀ = 1) was equivalent to 1.1 × 10¹² VP/mL. While infectious titers were determined by the plaque formation assay, VP: infectious viral particle (PFU) ratio was estimated to be 100:1, as described previously (25), for the experiments in this study. The genomic DNA from newly generated viruses were sequenced using PCR, then analyzed with SnapGene Software (GSL Biotech LLC) to confirm genomic integrity of final construct. The purified viruses were stored at −80°C until use.

To optimize our transduction results, MSCs were transduced with serotype 5 Ad (Ad5) containing 5 different types of fibers: (i) a wild-type fiber (dE1/GFP), (ii) a fiber knob domain of Ad35 (dE1-k35/GFP), (iii) an RGD motif in the HI loop of the fiber (dE1-RGD/GFP), (iv) a full-length fiber shaft of Ad5 and the Ad3 fiber knob (dE1-k3L/GFP), and (v) a truncated fiber shaft of Ad5 and the Ad3 fiber knob (dE1-k3S/GFP). MSCs were then incubated for 48 hours at 37°C. GFP expression was observed with a fluorescence microscope (Olympus IX81; Olympus Optical). To assess luciferase expression by HCC-oAd-Luc, MSCs were infected with HCC-oAd-Luc at multiplicities of infection (MOI) of 50, 100, or 200. At 48 hours postinfection, luciferase signals were analyzed using the IVIS Imaging System (Xenogen Corp). The bioluminescence signal intensity was obtained as photons acquired per second (p/s) from the region of interest (26).

To assess the optimal dose of HCC-targeting oAd (HCC-oAd-Luc) for loading into MSCs, the viability of MSCs (5 × 10⁴ cells/well in 24-well plates) was evaluated after infection with HCC-oAd-Luc at MOIs ranging from 0.5 to 50. At 2 or 5 days postinfection, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT; Sigma-Aldrich) was added to each well, after which, the plates were incubated at 37°C for 4 hours. The supernatant was then removed, and the precipitate was dissolved in DMSO. The absorbance at 540 nm was measured on a microplate reader. The viability of untreated MSCs was considered to be 100%.

To assess the production of HCC-targeting oAd in MSCs, MSCs were seeded in 24-well plates and infected with HCC-oAd-Luc at MOIs ranging from 0.5 to 50. After 2 or 5 days of incubation, the cell pellets and supernatants were collected and freeze-thawed 3 times to harvest the virions. qRT-PCR (Q-PCR; TaqMan PCR detection; Applied Biosystems) was used to assess the number of viral genomes in each sample as described previously (27). Data were processed using the SDS 19.1 Software Package (Applied Biosystems). The results are representative of three independent experiments.

To quantify HCC-oAd-WNTi released from the MSC as a result of virus' lytic property and intracellular replication of the virus, 2 × 10⁴ MSC were infected with 5 MOI of HCC-oAd-WNTi to generate HCC-oAd-WNTi/MSC. The viral genome copy number in culture supernatant and cell lysates were analyzed by qPCR as described above at 6 hours to 6 days postinfection.

Hep3B cells were seeded on 100-mm plates containing DMEM supplemented with 5% FBS and infected with 1 MOI of H101, HCC-oAd, or HCC-oAd-WNTi. Untreated group served as a negative control. At 48 hours postinfection, cell lysates were obtained. Western blotting was performed with cell lysates in similar manner as described previously (28), and following primary antibodies were used: Wnt3a (Abcam), β-catenin (Cell Signaling Technology), epithelial cadherin (E-cadherin; Cell Signaling Technology), or phospho-MAPK/ERK (p-MEK; Cell Signaling Technology).

To evaluate the effectiveness of HCC-oAd-WNTi for HCC killing, the viability of Hep3B cells was evaluated after infection with H101, HCC-oAd, or HCC-oAd-WNTi at MOIs ranging from 1 to 50. To assess the responsiveness of HCC-oAd-WNTi to hypoxia, the viability of Hep3B cells was evaluated under normoxic or hypoxic conditions after infection with 2 MOIs of HCC-oAd-WNTi. The cell viability in both set of experiments was determined at 48 hours postinfection using MTT as described above. Untreated cancer cells were considered to be 100% viable. The IC₅₀ was determined by analysis of a dose–response curve in GraphPad Prism.

For the assessment of time-dependent change in MSC viability following infection with HCC-oAd-WNTi, MSCs (seeded on 12-well plate at 2 × 10⁴ cells/well) were infected with HCC-oAd-WNTi at an MOI of 5. At 6 hours to 6 days postinfection, cell viability was assessed by the MTT assay as described previously. The results are representative of three independent experiments performed in triplicates.

To evaluate the effectiveness of HCC-oAd-WNTi/MSC for killing cancer cells, Hep3B, A549, or BJ cells (seeded on 24-well plate at 2 × 10⁴ cells/well) were treated with MSC, HCC-oAd-WNTi, or HCC-oAd-WNTi/MSC for 1, 3, or 5 days either under normoxic or hypoxic conditions. For the preparation of MSC and HCC-oAd-WNTi/MSC treatments, MSC (seeded on 24-well plate at 2 × 10⁴ cells/well) were either treated with PBS or 5 MOI of HCC-oAd-WNTi for 2 days, respectively. Subsequently, MSC and HCC-oAd-WNTi/MSC went through several washing cycles with ice-cold PBS, then cells were detached from the plate, centrifuged, and treated to Hep3B, A549, or BJ cells. At 1, 3, or 5 days posttreatment, MTT assay was performed as described previously.

To generate orthotopic HCC tumors, 1 × 10⁶ Hep3B cells, which stably express the firefly luciferase gene, were injected into the left liver lobe of athymic nude mice. On day 7 after the cell injection, bioluminescence imaging was performed to confirm tumor establishment. Mice were anesthetized in a chamber filled with 2% isoflurane in oxygen and received d-luciferin (150 mg/kg; Caliper) by intraperitoneal injection. Orthotopic HCC tumor–harboring mice were randomized into four treatment groups; 200 μL of PBS, MSCs (1 × 10⁶ cells), HCC-oAd-WNTi (5 × 10⁸ VPs), or HCC-oAd-WNTi/MSC (1 × 10⁶ cells infected with 5 × 10⁸ VPs for 18 hours) was intravenously injected twice into each mouse, at 9 and 13 days postimplantation (n = 6/group). HCC-oAd-WNTi/MSC was prepared in similar manner as those described above to remove excess HCC-oAd-WNTi. Optical imaging was conducted every week following the first imaging with an IVIS SPECTRUM Instrument (Xenogen Corp.). Image signals were quantitatively analyzed with IGOR-PRO Living Image Software (Xenogen Corp). The in vivo bioluminescence signal intensity was obtained from a body region of interest (29). The livers from mice in each experimental group were harvested at 35 days posttumor implantation and incubated in PBS supplemented with d-luciferin (Caliper). Both optical and luminescent images were obtained using an IVIS Imaging System (Xenogen Corp). All procedures of animal research were provided in accordance with the Laboratory Animals Welfare Act, the Guide for the Care and Use of Laboratory Animals. This animal care and use protocol was reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) in College of Medicine, Hanyang University of Korea (Seoul, Korea).

Liver tissues were harvested from mice at 3 days after the last intravenous injection of PBS, MSCs, HCC-oAd-WNTi, or HCC-oAd-WNTi/MSC. The harvested liver tissues were fixed in 10% formalin, processed for paraffin embedding, and then cut into 5-μm sections. The sections were stained with hematoxylin and eosin (H&E), and then examined by light microscopy. The tumor area in the liver sections was also immunostained with proliferating cell nuclear antigen (PCNA)-specific antibody (DAKO) or CD90-specific antibody (Abcam).

To assess the rate of Ad clearance from the blood, qPCR was performed with whole-blood samples from oAd-treated mice as described previously (27). Blood was collected from retro-orbital plexus of mice at 5 minutes, 10 minutes, 30 minutes, 1 hours, 6 hours, 24 hours, and 48 hours postsystemic injection with 5 × 10⁸ VPs of naked HCC-oAd-WNTi or HCC-oAd-WNTi/MSC (n = 3).

Hep3B tumor–bearing mice were systemically treated twice in similar manner as above. The lung, heart, stomach, spleen, liver, tumor tissues, and blood were harvested from mice in each group at 24 hours after the second injection, then DNA extraction and qPCR were performed as described previously (27).

To assess the in vivo toxicity of each formulation, mice were intravenously injected twice with 200 μL of PBS alone (negative control) or MSCs (1 × 10⁶ cells), HCC-oAd-WNTi (5 × 10⁸ VPs), or HCC-oAd-WNTi/MSC (1 × 10⁶ cells infected with 5 × 10⁸ VPs for 18 hours) on days 9 and 13 (n = 3). The serum levels of alanine transaminase (ALT) and aspartate aminotransferase (AST) were measured at 3 days postinjection (30).

Data were expressed as the mean ± SD. Statistical significance was determined by the two-tailed Student t test or one-way ANOVA test (SPSS 13.0 Software; SPSS). P values less than 0.05 were considered as statistically significant.

To optimize the efficiency of Ad transduction into MSCs, we tested the transduction efficacy of GFP-expressing and replication-incompetent human serotype 5 Ads with 5 different types of fibers (dE1/GFP, dE1-k35/GFP, dE1-RGD/GFP, dE1-k3L/GFP, and dE1-k3S/GFP). As shown in Supplementary Fig. S1, the transduction efficacy of dE1/GFP and dE1-RGD/GFP was limited at all MOI in MSCs, whereas dE1-k35/GFP, dE1-k3L/GFP, and dE1-k3S/GFP led to higher GFP expression levels in a dose-dependent manner. Of note, dE1-k35/GFP showed the highest transduction efficacy in MSCs among the tested Ads with different fibers, suggesting that replacement of the fiber knob with that of Ad serotype 35 leads to optimal transduction of Ad into MSCs. On the basis of these results, we generated an oAd with an Ad type 35 knob; the oAd replicates under the control of the AFP-positive HCC-specific Ha2bm promoter and expresses luciferase (HCC-oAd-Luc; Fig. 1A). To evaluate the infection efficacy of HCC-oAd-Luc, MSCs were transduced with HCC-oAd-Luc at various MOIs. At 48 hours after infection, the luciferase signal from MSCs increased in a MOI-dependent manner (Fig. 1B). These results demonstrate that an HCC-targeting oAd can efficiently internalize into MSCs and express its transgene.

Optimization of the oAd dose for loading into cell carriers is critical for preventing premature cytolysis. Thus, the optimal oAd loading dose for MSCs was assessed. As shown in Fig. 2A, MSC viability was significantly lower at 5 versus 2 days postinfection for all MOIs (P < 0.001). These results suggest that infecting MSCs with HCC-oAd-Luc for 5 days is suboptimal because the majority of the MSCs were lysed by the virus.

For further optimization, we assessed the level of viral replication in MSCs (Fig. 2B). A 10-fold increase in the viral dose, from 0.5 to 5 MOI, resulted in a 5,000-fold higher level of viral production at day 2 postinfection, whereas a same fold increase in the viral dose, from 5 to 50 MOI, only led to a minimal increase (1.8-fold) in viral production, showing that increasing the viral dose above 5 MOI yields a diminishing return on viral production. Because the viral production level plateaued at MOIs higher than 5 and a significant dose-dependent reduction in MSC viability was observed at the same viral dose range, we chose day 2 of infection and an MOI of 5 as the optimal conditions, which balance viral replication and MSC viability, to generate oAd-loaded MSCs in our subsequent experiments.

HCC cells demonstrate rapid and uncontrolled growth driven by the highly activated Wnt pathway (20). Therefore, we previously developed an oAd that expresses WNTi and demonstrated that WNTi expression can effectively inhibit the overactive Wnt signaling pathway in cancer (31). To continue this line of work, in our current study we designed and generated a HCC-specific oAd that expresses WNTi by utilizing a enhancer region–modified AFP promoter (Ha2bm; ref. 22) to restrict oAd replication to AFP-positive HCCs, ultimately generating HCC-oAd-WNTi (Fig. 3A).

To assess whether the newly generated HCC-oAd-WNTi could suppress the Wnt signaling pathway, we investigated the expression levels of Wnt and various associated downstream factors in HCC-oAd- or HCC-oAd-WNTi–infected Hep3B cells via Western blotting. As shown in Fig. 3B, Hep3B cells treated with HCC-oAd-WNTi exhibited a significantly lower Wnt expression level than did cells infected with control oAd (HCC-oAd) or the negative control group (untreated). Furthermore, HCC-oAd-WNTi treatment significantly attenuated the expression levels of various downstream factors, such as β-catenin and p-MEK that are known to promote rapid cancer cell proliferation and EMT. In addition, the expression level of a mesenchymal marker, E-cadherin, was markedly higher in HCC-oAd-WNTi–infected Hep3B cells than in untreated or HCC-oAd–infected cells. Together, these data show that HCC-oAd-WNTi effectively inhibited the Wnt signaling pathway and decreased levels of an EMT-related protein in HCC cells.

To assess whether expression of WNTi by the HCC-targeting oAd would enhance the cell-killing effect of oAd in HCC cells, AFP-positive HCC cells were infected with a clinically approved oAd (H101), the cognate control HCC-oAd, or HCC-oAd-WNTi at various MOIs. As shown in Fig. 3C, HCC-oAd-WNTi elicited a dose-dependent and significantly greater cancer cell–killing effect than H101 or HCC-oAd at all MOIs (P < 0.001), suggesting that WNTi expression enhances the cancer cell–killing effect of HCC-oAd. The IC₅₀ values for the control HCC-oAd and HCC-oAd-WNTi were 10.3 and 1.6 MOI, respectively, showing that the expression of WNTi augments the anticancer effect of oAd. We had previously shown that the Ha2bm promoter–driven viral replication of HCC-oAd overcomes hypoxia-induced repression of Ad replication (22). Thus, we assessed whether HCC-oAd-WNTi elicits more potent cancer cell–killing effect under hypoxic conditions than normoxia. As shown in Fig. 3D, HCC-oAd-WNTi elicited greater cancer cell–killing efficacy than HCC-oAd under both hypoxia and normoxia. Furthermore, HCC-oAd-WNTi showed 1.2-fold greater cancer cell–killing efficacy under hypoxia than normoxia (P < 0.001). These findings suggest that HCC-oAd-WNTi elicits a potent cell-killing effect in AFP-positive HCCs and overcomes hypoxia-mediated downregulation of viral replication (22).

To evaluate the cancer-specific killing effect of HCC-oAd-WNTi/MSC (infected with 5 MOI of oAd), we cocultured a cancer or normal cell line [HCC (Hep3B), non-HCC cancer (A549), or normal (BJ)] with HCC-oAd-WNTi/MSC at a 1:1 MSC-to-cell ratio. As shown in Fig. 4, similar level of viability was observed for the untreated and MSC-treated groups under both normoxic and hypoxic conditions, suggesting that MSCs alone did not affect the cell growth rate. In contrast, Hep3B cells treated with HCC-oAd-WNTi or HCC-oAd-WNTi/MSC exhibited significantly reduced viability in comparison with untreated or MSC-treated cells, suggesting that both naked HCC-oAd-WNTi and the virus loaded in MSCs can kill AFP-positive HCC cells. At day 1 posttreatment, HCC-oAd-WNTi- and HCC-oAd-WNTi/MSC–treated Hep3B cells exhibited a similar level of viability to that of the untreated group, emphasizing that the reduction in cancer cell viability was due to replication-mediated cytopathic effects in the oAd-incorporated groups. In marked contrast, HCC-oAd-WNTi/MSC treatment elicited a significantly more potent HCC killing effect than HCC-oAd-WNTi on day 3 or 5 of treatment (P < 0.05, P < 0.01), showing that HCC-oAd-WNTi replication in MSCs enhances the cytopathic effect. These results are in agreement with results shown in Fig. 2B and S2. In AFP-negative A549 lung cancer cells, similar results were observed only under hypoxic conditions (P < 0.01), confirming the HRE-mediated enhancement of Ad replication (22, 26). These results suggest that HCC-oAd-WNTi/MSC can elicit cancer-specific cell-killing effect in response to hypoxia, which is a hallmark of solid tumors (32), regardless of cellular AFP level; the expanded cancer targeting by hypoxia-responsiveness of Ha2bm promoter would be beneficial in clinical setting where heterogeneity of tumors leads to large phenotypical variance in tumors, yet hypoxia is conserved in all types of solid tumors, including 30% of HCC patient cases with AFP-negative HCC tumors (22, 33). Furthermore, both HCC-oAd-WNTi and HCC-oAd-WNTi/MSC elicited minimal killing effects in the normal BJ cell line, showing that the oncolytic effects of HCC-oAd-WNTi/MSC occur with high specificity toward cancer cells.

Orthotopic tumor models are emerging as important cancer research models due to their clinical relevance (34). Solid tumors inaccessible by a needle, such as lung, liver, and pancreatic cancers, cannot be treated by intratumoral administration, and thus require systemic treatment. To evaluate the therapeutic efficacy of systemically administered oAd-loaded MSCs, luciferase-expressing orthotopic Hep3B tumors were treated with PBS, MSCs, HCC-oAd-WNTi (5 × 10⁸ VP), or HCC-oAd-WNTi/MSC (1 × 10⁶ MSC infected with 5 × 10⁸ VP of oAd; 5 MOI) via tail-vein injection. Infection of MSC with 5 MOI of HCC-oAd-WNTi for 18 hours (HCC-oAd-WNTi/MSC) had negligible impact on MSC viability, because more than 95% of MSCs were viable with these conditions (Supplementary Fig. S3). In addition, HCC-oAd-WNTi/MSC prepared by these conditions was determined to possess intracellular VP count that is equivalent to infection with 0.08 MOI (Supplementary Fig. S2), which would translate to 8.0 × 10⁶ VP being delivered by MSC at time of administration. These results indicate that 62.5-fold lower VP of HCC-oAd-WNTi/MSC were administered compared with HCC-oAd-WNTi. As shown in Fig. 5A and B, the systemic administration of HCC-oAd-WNTi/MSC resulted in markedly higher antitumor activity than either HCC-oAd-WNTi or MSCs alone at 35 days postimplantation (P < 0.01), showing 44.7-, 24.0-, or 8.1-fold greater therapeutic efficacy than PBS, MSCs, or HCC-oAd-WNTi, respectively.

Although HCC-oAd-WNTi did not induce tumor regression in this report due to extremely low viral dose deflating the overall antitumor effect to a suboptimal level, systemic administration of HCC-specific oAd replicating under the control of Ha2bm promoter at higher dose would likely achieve more potent tumor growth inhibition. Indeed, a systemic administration of HCC-oAd at 2.5 × 10¹⁰ VP has been reported to elicit potent antitumor effect against orthotopic HCC tumor (22). Ex vivo imaging of the livers further supports these data; PBS-, MSC-, or HCC-oAd-WNTi–treated livers showed a relatively larger tumor burden, whereas livers from HCC-oAd-WNTi/MSC–treated mice had a significantly lower tumor burden (Fig. 5C and D). Similar results were observed when the tumor burden was assessed by end-point tumor weight measurement (Fig. 5E). Of note, the HCC-oAd-WNTi/MSC was estimated to harbor 8.0 × 10⁶ VP of oAd based on the results from Supplementary Fig. S2, which meant that the injected dose was 2,500-fold lower than the oAd dose (∼2 × 10¹⁰ VPs) frequently utilized for systemic treatment of tumor-bearing mice (27). These results suggest that loading of oAd into MSCs augments the therapeutic efficacy of systemically administered virus while minimizing any potential dose-related side effects.

To further investigate the therapeutic effect of HCC-oAd-WNTi/MSC, liver tissues were harvested at 3 days after the last treatment (with PBS, MSCs, HCC-oAd-WNTi, or HCC-oAd-WNTi/MSC). As shown in Fig. 6A, H&E staining revealed large tumor areas in both PBS- and MSC-treated liver tissues. In contrast, HCC-oAd-WNTi/MSC–treated liver tissues showed no observable tumor cells. In support of the H&E staining results, HCC-oAd-WNTi/MSC treatment led to a markedly lower level of proliferating cells compared with PBS, MSC, or HCC-oAd-WNTi treatments (Fig. 6B), indicating that HCC-oAd-WNTi/MSC induces a potent antitumor effect by inhibiting HCC cell proliferation. Next, we assessed how the oAd affects MSC viability in the tumor tissues by staining for CD90, a marker of human MSCs. As shown in Fig. 6C, the tumor tissues from the mice treated with MSCs were CD90⁺, whereas no MSCs were observed in the liver tissues treated with HCC-oAd-WNTi/MSC.

For the efficient treatment of HCC tumors by systemic injection, the blood retention time of Ad must be prolonged (35). Therefore, we assessed whether MSC prolongs the blood retention time of HCC-oAd-WNTi. As shown in Fig. 7A, naked HCC-oAd-WNTi was rapidly cleared from the blood. In contrast, HCC-oAd-WNTi/MSC was retained at 17- and 3,200-fold higher levels in the blood compared with naked HCC-oAd-WNTi at 6 hours and 24 hours postinjection (P < 0.001 for HCC-oAd-WNTi vs. HCC-oAd-WNTi/MSC at 6 hours: 2.4 × 10³ vs. 4.0 × 10⁴ VP and 24 hours: 1.8 × 10¹ vs. 5.9 × 10⁴ VP), respectively, suggesting that MSCs efficiently extend the oAd blood circulation time.

Naked Ad is nonspecifically sequestered in the liver following intravenous injection due to interactions with Kupffer cells and coagulation factors (27), resulting in hepatotoxicity and a limited antitumor effect. To examine whether HCC-oAd-WNTi/MSC can circumvent hepatic sequestration while enhancing intratumoral oAd accumulation, biodistribution profiles of systemically administered MSCs, HCC-oAd-WNTi, or HCC-oAd-WNTi/MSC were analyzed in tumor-bearing mice. As shown in Fig. 7B, the liver uptake of HCC-oAd-WNTi/MSC was significantly less than that of naked HCC-oAd-WNTi (P < 0.001). Importantly, HCC-oAd-WNTi/MSC showed 4,824-fold higher intratumoral accumulation than naked HCC-oAd-WNTi (P < 0.001). These results show that MSCs can protect their oAd cargo effectively during systemic circulation and curtail native hepatic oAd tropism, leading to efficient intratumoral accumulation of the virus. In addition, these data suggest that MSC-mediated intratumoral oAd localization, even at a low viral dose, can lead to a potent antitumor effect as the HCC-oAd-WNTi can effectively replicate within tumor tissues. Consequently, the tumor-to-liver ratio of HCC-oAd-WNTi/MSC was 6,481-fold greater than that of naked HCC-oAd-WNTi, showing that MSC-based delivery of oncolytic Ad can overcome the hurdles of systemic administration to improve the therapeutic efficacy and safety profile of the virus.

Next, hepatotoxicity was analyzed by assessing the serum level of ALT and AST after the systemic administration of PBS, MSCs, HCC-oAd-WNTi, or HCC-oAd-WNTi/MSC in tumor-bearing mice, as well as in normal mice without tumor. As shown in Fig. 7C, mice treated with HCC-oAd-WNTi showed the highest AST levels (1.6-fold higher than in the PBS-treated group; P < 0.05). Conversely, we did not observe a significant increase in AST levels in mice treated with HCC-oAd-WNTi/MSC relative to levels in the normal mice group, showing that MSC-mediated tumor-specific oAd delivery prevents the hepatic damage that is traditionally associated with systemically administered Ad as well as with an HCC tumor burden (22, 36).

MSCs can infiltrate into tumor tissues and release therapeutic cargos, such as conventional chemotherapeutics or oncolytic viruses (37, 38). Although several types of drugs can be loaded into MSCs, efficient loading of Ad into MSCs is hampered by a low level of CAR expression on the MSC surface (39). This problem was addressed in this report by replacement of the endogenous Ad serotype 5 fiber knob with an Ad serotype 35 fiber knob to enhance Ad transduction into MSCs (Supplementary Fig. S1). Here, we employed other genetic modification strategies to further improve the therapeutic efficacy of oAd against HCC. We previously developed HCC-specific oAd (HCC-oAd), which utilizes modified AFP hypoxia–responsive and modified AFP enhancer regions to enhance transcription of the Ad E1A gene; this modification overcomes hypoxia-mediated downregulation of Ad replication in tumors. Building on this previous work, we have now inserted a sequence encoding WNTi, a soluble Wnt decoy receptor, into HCC-oAd (generating HCC-oAd-WNTi) to inhibit the hyperactive Wnt signaling pathway, a problem found in 50% of clinical HCCs that is associated with a poor patient survival rate (31, 40). The addition of WNTi as a therapeutic gene allowed for effective suppression of various Wnt signaling–related factors and improved the HCC killing potency of HCC-oAd-WNTi (Fig. 3). The disruption of the overactive Wnt signaling pathway in HCC is of critical importance as only few Wnt-inhibitory chemical drugs have reached phase I clinical trials to date, even though the Wnt/β-catenin pathway is frequently upregulated and implicated in drug resistance, tumor progression, and metastasis of HCC (41). Furthermore, oAd-mediated therapeutic gene expression has been shown to be highly localized to tumor tissues due to tumor-specific replication and gene expression by the virus (28), making oAd better suited for HCC therapy than chemical compounds with insufficient cancer specificity, given that Wnt/β-catenin signaling is critical for the maintenance of diverse liver functions in normal hepatic tissues (42).

One of the innovative aspects of using MSCs as a carrier for oncolytic viruses is that these cells could function as a biological factory to enable viral replication, suggesting that it may be possible for a small quantity of virus to be initially loaded into MSCs to deliver a sufficient viral dose to tumor tissues (15, 43). Even though replication of oncolytic virus in MSCs has been reported by our current study and others before (44, 45), the detailed mechanism is not clear. One plausible explanation is that Ha2bm promoter is predicted to possess several transcription factor–binding domains for C-JUN, FOXP3, and C/EBP-β, which are expressed in the MSC and upregulated during determination of MSC fate (46, 47), and these attributes may allow replication of HCC-oAd-WNTi in MSC through Ha2bm promoter activation.

Although viral replication within MSCs is a promising attribute, excessive viral replication may induce premature MSC lysis, which would decrease the overall efficacy of the system. Given this background, we carefully optimized the viral dose for MSC infection to balance the increase in total viral yield with the reduction in MSC viability (Fig. 2), showing that too high of an initial viral dose critically attenuates overall MSC viability with minimal improvement in total viral yield. Profiling of MCS survival and viral release also demonstrate that MSCs infected with HCC-oAd-WNTi (5 MOI; 18 hours or 2 days of infection; in vivo or in vitro preparation condition of HCC-oAd-WNTi/MSC, respectively) are healthy and viable enough to produce progeny viruses (Supplementary Figs. S3 and S2). In particular, plenty of viral progenies were released from the MSCs due to cytolytic property of oAd. These results are in line with a previously reported Ad production protocol in which a low initial dose of Ad (2 MOI) was utilized to optimize viral production in a virus producer cell line, suggesting that an excessive viral dose may lead to poor viral production and yield due to an asymmetrical compensation in cell viability (48).

Notably, the conditions for optimal viral replication and MSC viability (Fig. 2) enabled HCC-oAd-WNTi/MSC to elicit a more potent HCC killing effect than that elicited by naked HCC-oAd-WNTi (Figs. 3 and 4), showing that, over time, viral replication in MSCs may increase the final viral titer. The difference between the anticancer effects of naked HCC-oAd-WNTi and HCC-oAd-WNTi/MSC was even greater in vivo: systemic treatment with a substandard therapeutic dose of HCC-oAd-WNTi/MSC led to complete regression in a subpopulation of treated mice (Fig. 5), supporting in vitro findings (Figs. 2 and 4). In general, the systemic administration of various oAds for the treatment of orthotopic HCC tumors has required more than 1 × 10¹⁰ VPs to elicit a notable antitumor effect (22, 26), and to best of our knowledge there are no available reports of systemically administered oAd inducing complete HCC regression at a 40-fold lower viral dose than the conventionally reported dose of 2 × 10¹⁰ VPs. Such dose reduction is critical for the clinical development of new therapeutics, because it can greatly improve the safety profile.

The tumor-homing and cargo-protective properties of MSCs acting as a cell carrier are well known (15, 49). In present, our findings show that cargo-protective properties of MSC enabled the virions loaded in HCC-oAd-WNTi/MSC to efficiently accumulate in tumor tissues, in part, through enhanced virion retention in the bloodstream (Fig. 7A and B, respectively). In agreement with this finding, others have reported that several types of systemic carriers can improve the intratumoral accumulation of systemically administered oAd by prolonging the blood circulation of virions (27, 50). Carrier MSCs also attenuated hepatic sequestration of oAd and prevented hepatic damage (Fig. 7B and C), a promising result given that the hepatic tropism and inflammatory attributes of Ad are major concerns associated with the systemic delivery of Ad in clinical trials. These findings show that MSCs functioning as a systemic carrier of oAd can overcome conventional shortcomings associated with systemic oAd delivery, such as nonspecific hepatic sequestration, hepatotoxicity, and poor blood circulation (29, 30).

Finally, one of the major concerns about all stem cell–based therapy in clinical trials is that stem cells may uncontrollably differentiate into undesirable lineages, potentially leading to tumorigenesis and inflammation; thus, the FDA requires long periods of patient follow-up during and after clinical trials (51). Our findings provide strong evidence that this potential risk can be addressed by utilizing the cytolytic effects of oAd, because CD90⁺ MSCs were not detected in liver tissues from mice treated with HCC-oAd-WNTi/MSC at 3 days after administration, whereas MSCs remained in the liver tissues from mice treated with MSCs alone (Fig. 6C). Our in vitro findings shown in Fig. 2 also support this claim because active oAd replication in MSCs led to cell lysis, thus abolishing any potential risk that may arise due to engraftment of therapeutic MSCs into tumor tissues.

Collectively, we demonstrate that (i) HCC-targeted oAd can be efficiently loaded into MSCs through modification of the viral capsid, (ii) oAd loaded into MSCs can elicit potent cancer-specific killing effects through active viral replication in the MSC carrier, (iii) MSC tumor–homing tropism improves tumor-specific oAd accumulation, (iv) the cargo-protective attributes of MSCs prolong and enhance virion circulation in the blood, and (v) oAd-loaded MSC improves the safety profile of both oAd and MSCs by decreasing oAd hepatic sequestration and hepatotoxicity while promoting elimination of MSCs through viral replication. These findings show that MSCs loaded with HCC-targeting oAd can serve as a “stealth carrier” to preferentially deliver the virus to the tumor while attenuating any potential risk that may arise from systemic administration of naked virions in the clinical environment.

J. Hong is a researcher at GeneMedicine. C.-O. Yun is the CEO at GeneMedicine. H.C. Shin and H Lee are researchers at Phamicell. H.S. Kim is the CEO at Pharmicell Co. Ltd. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.-R. Yoon, H. Lee, C.-O. Yun

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.-R. Yoon, Y. Li, H.C. Shin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.-R. Yoon, J. Hong, Y. Li

Writing, review, and/or revision of the manuscript: A.-R. Yoon, J. Hong, C.-O. Yun

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.-R. Yoon, Y. Li, H. Lee, H.S. Kim, C.-O. Yun

Study supervision: C.-O. Yun

This work was supported by grants from the National Research Foundation of Korea (2016M3A9B5942352 to C.-O. Yun; 2016R1C1B2015558 to A.-R. Yoon) and Korea Drug Development Fund (KDDF) funded by MSIP, MOTIE, and MOHW (KDDF-201611-05, Republic of Korea, to A.-R. Yoon).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.