Evaluation of Human Mesenchymal Stromal Cells as Carriers for the Delivery of Oncolytic HAdV-5 to Head and Neck Squamous Cell Carcinomas

Human multipotent mesenchymal stromal cells (hMSCs) are of significant therapeutic interest due to their ability to deliver oncolytic adenoviruses to tumors. This approach is also investigated for targeting head and neck squamous cell carcinomas (HNSCCs). HAdV-5-HexPos3, a recently reported capsid-modified vector based on human adenovirus type 5 (HAdV-5), showed strongly improved infection of both hMSCs and the HNSCC cell line UM-SCC-11B. Given that, we generated life cycle-unmodified and -modified replication-competent HAdV-5-HexPos3 vector variants and analyzed their replication within bone marrow- and adipose tissue-derived hMSCs. Efficient replication was detected for both life cycle-unmodified and -modified vectors. Moreover, we analyzed the migration of vector-carrying hMSCs toward different HNSCCs. Although migration of hMSCs to HNSCC cell lines was confirmed in vitro, no homing of hMSCs to HNSCC xenografts was observed in vivo in mice and in ovo in a chorioallantoic membrane model. Taken together, our data suggest that HAdV-5-HexPos3 is a potent candidate for hMSC-based oncolytic therapy of HNSCCs. However, it also emphasizes the importance of generating optimized in vivo models for the evaluation of hMSC as carrier cells.


Introduction
In 2020, more than 250,000 people died from head and neck squamous cell carcinomas (HNSCCs), a cancer type derived from the mucosal epithelium of the larynx, the pharynx, or the oral cavity [1,2]. Despite surgical advances and aggressive multimodality therapy, the prognosis for advanced HNSCCs is poor (<25% survival rate after five years) [2,3]. One promising approach to improve the treatment of HNSCCs is oncolytic virotherapy, which aims to destroy tumors by direct virus-mediated cell lysis and by the induction of antitumor immunity [4], as it is actively pursued with oncolytic adenoviral vectors (oAVs) [5][6][7].
Although oAVs may harbor significant potential for the lysis of tumor cells in vivo, clinical efficacy is limited for several reasons. One reason is an unfavorable in vivo biodistribution pattern of oAVs: upon systemic administration, the viral particles encounter several barriers that lead to extensive off-target sequestration, including binding to human erythrocytes [8,9], neutralization by natural IgMs [10,11], complement proteins [12], and pre-existing IgGs [13], transduction of hepatocytes [14,15], and scavenger receptormediated sequestration by macrophages [16][17][18]. Although off-target sequestration can be reduced by intratumoral injection of oAVs, this route of application limits the therapeutic efficacy due to the restricted spread of viral particles within the tumor tissue and the inability to reach metastases [19]. Hence, a strategy enabling the systemic administration of oAVs without off-target sequestration might improve the efficiency of oncolytic virotherapy.
One strategy is based on the use of human mesenchymal stromal cells (hMSCs) as carriers to transport viral particles to the tumor site while shielding them from blood components. In this approach, hMSCs are infected with oAVs ex vivo and subsequently administered systemically to a patient. Attracted by chemokines and growth factors released from the tumor, hMSCs are believed to infiltrate the tumor site and are expected to be lysed by the replicating oAVs in direct vicinity of the targeted tumor cells. Consequently, the released viral progeny can infect and lyse tumor cells, ideally leading to tumor necrosis and the exposition of cancer-associated antigens, which can initiate the reactivation of the immune system [20]. This approach showed encouraging results in many preclinical studies and is already being evaluated in the clinic (e.g., NCT03896568, NCT02068794, NCT01844661) [21,22].
The overall aim of our study was to evaluate the therapeutic potential of hMSCs loaded with replication-competent HAdV-5-HexPos3 vectors for treating HNSCCs in a preclinical model. We generated several replication-competent HAdV-5-HexPos3 vectors harboring different life cycle modifications to assess their replication behavior within bone marrow-(BM) and adipose tissue-derived (A-) hMSCs. Moreover, we analyzed the migration of noninfected and infected hMSCs toward different HNSCC cell lines. Although we observed robust migration toward these tumor cells in vitro, no migration of vector-loaded hMSCs to HNSCC xenograft tumors was detected in two mouse models and a chorioallantoic membrane (CAM) model. Our results provide evidence that replication-competent HAdV-5-HexPos3 vectors might be a valuable tool for hMSC-mediated oncolytic virotherapy and highlight the need to improve site-directed homing of hMSCs toward tumors.
∆CAR vectors had a point mutation in the fiber knob (Y477A), which significantly reduces binding of the HAdV-5 fiber knob to CAR [37].
Total physical/particle titers were determined by optical density measurement at 260 nm wavelength [38]. SDS-PAGE with subsequent silver staining [39] was performed to confirm the purity of viral vectors. Moreover, vector genome integrity was verified by restriction enzyme digestion and sequencing of isolated vector DNA.

Transduction/Infection of hMSCs with HAdV-5 Vectors
Transduction or infection of hMSCs was performed in 24-(3 × 10 4 cells/well, 1 mL medium, ThermoFisher Scientific #142485) or 6-(1 × 10 5 cells/well, 3 mL medium, Ther-moFisher Scientific #40685) well plates. One day after seeding, cells were washed with PBS, and transduced/infected with the indicated HAdV-5 vector and physical/particle multiplicity of infection (pMOI) in supplement-free medium. In the case of transduction using human blood coagulation factor X (hFX) as a transduction enhancer, viral particles were preincubated with 4 fg of hFX/viral particle (30 min, 37 • C.) Three hours post-transduction, the vector-containing medium was removed, cells were washed with PBS, and fresh medium was added. For in vivo, in ovo, and in vitro migration analysis, hMSCs were washed with PBS 3 h post-infection. Subsequently, cells were detached from the plates using TrypLE select (ThermoFisher Scientific, #12563029), counted with a Neubauer counting chamber, and diluted to the respective cell concentration in culture medium (in vitro experiments) or PBS (in vivo and in ovo experiments).

HNSCC Xenograft Mouse Model
All animal experiments were conducted in an Association for Assessment and Accreditation of Laboratory Animal Care International-accredited facility in accordance with the European Union (EU) Directive 2010/63/EU for animal experiments. Authorization was granted by the Animal Care Commission of the government of Baden Württemberg, Germany (Regierungspräsidium Tübingen, TVA #1459).
2 × 10 6 UD-SCC-2 or UD-SCC-6 cells were diluted in 50 µL PBS, mixed with 50 µL growth factor-reduced matrigel (Merck, Burlington, MA, USA, #CLS356252), and injected subcutaneously into the left flank of NSG mice (bred in the animal research center Ulm). Subsequently, mice were monitored daily (general health status, body weight, tumor volume). Using a digital caliper, the tumor volume was determined by measuring the greatest longitudinal diameter (length) and the greatest transverse diameter (width). Subsequently, the tumor volume was calculated by the modified ellipsoidal formula: tumor volume = 0.5 × length × width 2 .

Tissue Preparation for Immunohistochemistry
At the end of the in vivo NSG mice experiments, mice were sacrificed, and organs and tumors were harvested and frozen in liquid nitrogen. Moreover, tumor tissue samples were fixed (2% PFA in PBS, 4 • C, 24 h) and incubated in cryoprotectant (20% sucrose, 4 • C, 24 h). Subsequently, samples were embedded in TissueTek (Sakura Finetek, Staufen im Breisgau, Germany, #4583) and frozen at −80 • C before the preparation of cryosections. Cryosections (6 µm) of tumor tissue were prepared in a cryostat cryotome (Leica, Wetzlar, Germany), transferred to microscope slides (ThermoFisher Scientific, #10149870), and used for immunohistochemical staining.

DNA Extraction from Tissue Samples and Analysis of Adenoviral E4 Copy Number Using Quantitative Real-Time Polymerase Chain Reaction (qPCR)
In vivo biodistribution of hMSC was analyzed by the detection of the adenoviral genome copy number in tumors and organs. First, DNA was extracted from tumors and organs using the GenElute Mammalian Genomic DNA Miniprep Kit (Sigma Aldrich, St. Louis, MO, USA, #G1N350-1KT) following the manufacturer's instructions. After eluting the DNA in 10 mM Tris, pH 8.0, DNA concentration was determined photometrically at 260 nm. To determine the number of adenoviral genomes in the samples, qPCR analysis targeting the adenoviral E4 gene was performed using KAPA SYBR ® FAST (Kapa Biosystems, Wilmington, MA, USA#KK4502) according to the manufacturer's instructions and a CFX Connect qPCR Cycler (BioRad, Hercules, CA, USA). As a template, 20 ng of total DNA (dissolved in 2 µL 10 mM Tris, pH 8.0) was used. Detected adenoviral genome copy numbers were normalized to detected actin copy numbers. Primers specific for the adenoviral E4 region: forw.: 5 TAGACGATCCCTACTGTACG 3 ; rev.: 5 GGAAATAT-GACTACGTCCGG 3 . Primers specific for β-actin: forw.: 5 GCTCCTCCTGAGCGCAAG 3 ; rev.: 5 CATCTGCTGGAAGGTGGACA 3 .

Virus Replication Assay
3 × 10 4 hMSCs were seeded in 24-well plates (Thermo Fisher Scientific, #142485). The next day, hMSCs were infected with indicated pMOIs. Six days post-infection, hMSCs were washed and fixed with 4% PFA in PBS (15 min, RT). After washing with PBS, hMSCs were stained with 0.1% crystal violet dissolved in H 2 O (2 min, RT). After three washing steps with PBS, the stained plate was air-dried for some minutes. Subsequently, pictures of the plates were taken.

Quantification of Virus Replication in hMSCs by qPCR
Human MSCs were infected with the indicated replication-competent HAdV-5-HexPos3 vectors. At the indicated time points, supernatants and cells (detached with a cell scraper) were harvested, adjusted to a volume of 2 mL, and snap-frozen. Viral particles were released from the cells by three freeze-thaw cycles. To digest all non-encapsidated adenoviral genomes, 25 µL of the samples were transferred to new tubes and treated with 5 units of DNAse I (37 • C for 30 min). DNAse I and the viral particles were denatured at 95 • C for 10 min. Subsequently, 2 µL of the DNAse I-treated samples were used for qPCR analysis to quantify E4 copy numbers (described in 2.10). Subsequently, based on the E4 copy numbers detected in 2 µL of sample, the total amount of E4 copies in the hMSC lysate (total volume 2000 µL) was calculated and divided by the total number of cells per well (3 × 10 4 hMSCs) to determine the total amount of genome-containing adenoviral particles produced per hMSC.

Analysis of Lactate Dehydrogenase (LDH) Release to Cell Culture Supernatant
The amount of LDH released to the cell culture supernatant of hMSC cultures upon HAdV-5 infection was determined using the LDH-Glo™ Cytotoxicity Assay (Promega, #J2380) following the manufacturer's instructions. Briefly, hMSCs were infected with pMOI 900 of the indicated replication-competent HAdV-5-HexPos3 vectors. Starting 24 h postinfection, 50 µL of supernatant were collected daily and centrifuged (400× g, 10 min). A total of 20 µL of the supernatant were diluted in 480 µL of LDH Assay storage buffer and used for analysis. Luminescence signals were quantified using the GloMax ® luminometer (Promega).

Analysis of Virus Spread by Plaque Assays
Two hundred hMSCs were infected with the indicated HAdV-5 vectors (pMOI 900, in supplement-free medium) and seeded on top of UD-SCC-2 cells (seeded the day before in 6-well plates, 1 × 10 6 cells/well). The next day, the medium was removed, and cells were overlayed with 3 mL of 0.75% low-melting agarose (Biozym, Hessisch Oldendorf, Germany, #840101) solved in MEM medium (ThermoFisher Scientific, #31095-029) supplemented with 10% FBS. Subsequently, cells were incubated at 37 • C and analyzed for the spread of eGFP signal by fluorescence microscopy daily until 216 h post-infection. The percentage of non-lysed BM-hMSCs was determined by setting the number of single cells (=non-lysed cells) and the number of eGFP-positive plaques in proportion. Moreover, the diameter of the occurring plaques was determined.

Migration Assays In Vitro
For Boyden chamber migration assays, 2 × 10 4 hMSCs (non-infected or 3 h postinfection with pMOI 900) were seeded into the upper chamber of tissue culture (TC)-inserts with a pore size of 8 µm. The TC-inserts were previously placed into 24-well plates harboring either (i) tumor cell-conditioned medium (harvested from tumor cell lines UD-SCC-2 or UD-SCC-6 cultured on 15 cm dishes for three-four days in DMEM without any additives) or (ii) the respective culture medium (DMEM without any additives) as a negative control.
After 18 h, non-migrated hMSCs were removed by thoroughly cleaning the upper chamber with cotton swabs. Subsequently, migrated cells at the bottom side of the TC insert were fixed with 4% PFA in PBS (15 min, RT). Cell nuclei were stained with 4 ,6-diamidino-2-phenylindole (DAPI, Sigma Aldrich, #D9542-1MG; 1 µg/mL) to allow for the counting of migrated hMSCs. Subsequently, three images of different areas of the TC-inserts were taken using a fluorescence microscope (10-fold magnification). The total number of migrated hMSCs per TC insert was calculated based on the average number of cell nuclei counted per microscope image.

Statistical Analysis
Experiments performed in this study were repeated at least three times independently unless described otherwise. Statistical tests used for analysis are stated in the figure legends and were performed using GraphPad Prism software version 8.4.2 (GraphPad Software LLC, San Diego, CA, USA).

Replication-Competent HAdV-5-HexPos3 Particles Efficiently Replicate in BM-and A-hMSCs
Modification of the HAdV-5 hexon protein by deleting a stretch of 13 amino acids and replacing them with four consecutive lysine residues was previously shown to result in very efficient infection of hMSCs and the HNSCC cell line UM-SCC-11B (for both ≥80% eGFP-positive cells with pMOI 1000) [29]. Thus, we hypothesized that combining oncolytic HAdV-5-HexPos3 vectors with hMSCs as carrier cells might be a promising approach for the treatment of HNSCCs. We generated three different types of replication-competent vectors: (i) vectors harboring the complete wildtype HAdV-5 genome (referred to as 'no life cycle modification'), (ii) HAdV-5-E1A-∆24 vectors harboring a 24 base pair (bp) deletion in the E1A 12S gene, which is supposed to restrict virus replication to proliferating cells, and (iii) HAdV-5-E1A-∆24-E1B-∆19K vectors, which carry, additionally to the E1A 12S modification, a deletion of the E1B 19K gene sequence, supposed to restrict virus replication to cells with mutated apoptotic pathways.
To analyze the ability of the three vector types to replicate in hMSCs, bone marrow-(BM-) and adipose tissue-derived (A-) hMSCs were infected. In addition to the infection with replication-competent HAdV-5-HexPos3 particles, hMSCs were also infected with their wildtype-capsid counterparts (both pMOI 1000). To overcome the inefficient infection of hMSCs with wildtype-capsid HAdV-5 particles, those particles were pre-incubated with the transduction enhancer human blood coagulation factor X (hFX) [23]. For both, transduction with HAdV-5 (combined with hFX) or HAdV-5-HexPos3 (without enhancer), efficiencies of ≥80% were achieved for BM-hMSCs and ≥50% for A-hMSCs in previous studies [23,29]. Six days post-infection, the supernatants were discarded, and remaining cells were fixed and stained with crystal violet (Figure 1). In this assay, high levels of crystal violet staining correspond to a high number of remaining cells indicating a low level of virus replication. In the samples in which wildtype-capsid replication-competent HAdV-5 vectors were used for infection, no lysis of hMSCs was detected independent of pMOI or life cycle modification. In contrast, strong virus-mediated cytotoxicity was observed when HAdV-5 was combined with the transduction enhancer hFX or when HAdV-5-HexPos3 particles were used, independent of life cycle modification. Interestingly, while no differences were observed between vectors without life cycle modification and with E1A-Δ24 modification, infection of cells with E1A-Δ24-E1B-Δ19K vectors resulted in a different phenotype independent of the pMOI used (between 1 and 1000). Upon infection with replicationcompetent vectors without life cycle modification or with the E1A-Δ24 modification, hMSCs show a roundish phenotype and slowly detach from the culture plate. In contrast, upon infection with vectors carrying the E1A-Δ24-E1B-Δ19K life cycle modification, hMSCs seem to be destroyed by the virus infection without previously detaching from the culture plate (Figure 2). Notably, the cytotoxic effects of HAdV-5-HexPos3 particles were significantly stronger compared to HAdV-5 particles combined with hFX, which is why only HAdV-5-HexPos3 particles were used for all further experiments. In the samples in which wildtype-capsid replication-competent HAdV-5 vectors were used for infection, no lysis of hMSCs was detected independent of pMOI or life cycle modification. In contrast, strong virus-mediated cytotoxicity was observed when HAdV-5 was combined with the transduction enhancer hFX or when HAdV-5-HexPos3 particles were used, independent of life cycle modification. Interestingly, while no differences were observed between vectors without life cycle modification and with E1A-∆24 modification, infection of cells with E1A-∆24-E1B-∆19K vectors resulted in a different phenotype independent of the pMOI used (between 1 and 1000). Upon infection with replicationcompetent vectors without life cycle modification or with the E1A-∆24 modification, hMSCs show a roundish phenotype and slowly detach from the culture plate. In contrast, upon infection with vectors carrying the E1A-∆24-E1B-∆19K life cycle modification, hMSCs seem to be destroyed by the virus infection without previously detaching from the culture plate ( Figure 2). Notably, the cytotoxic effects of HAdV-5-HexPos3 particles were significantly stronger compared to HAdV-5 particles combined with hFX, which is why only HAdV-5-HexPos3 particles were used for all further experiments. Next, we determined the amount of viral progeny produced at different time points by hMSCs infected with HAdV-5-HexPos3 with or without life cycle modification. Every 24 h, cells and supernatant were harvested, and the number of viral particles produced per hMSC was quantified by qPCR ( Figure 3).
Generation of viral progeny from hMSCs was demonstrated in all samples, with no significant differences between the different replication-competent vectors or the pMOI used for infection. In BM-hMSCs, a maximum of about 1 × 10 4 viral particles produced per cell was measured, whereas production efficiency in A-hMSCs was slightly lower. Independent of the hMSC origin, the maximum amount of viral progeny was already reached 48-72 h post-infection. Thus, infection of hMSCs with different replicationcompetent HAdV-5-HexPos3 vectors results in equally efficient viral replication.

Replication-Competent HAdV-5-HexPos3 Vectors Carrying the E1B-Δ19K Mutation Show Accelerated Replication and Release of Mature Viral Progeny in hMSCs
To confirm that viral progeny released from hMSCs are mature and infectious, we infected BM-and A-hMSCs with the different replication-competent HAdV-5-HexPos3 vectors (pMOI 900) and seeded them onto a dense layer of UD-SCC-2 cells. As all vectors used for infection carried an eGFP transgene, infected hMSCs were detected as single eGFP-expressing cells in the wells. This enabled the analysis of virus release kinetics: if only single cells are visible in the well, the hMSCs have not yet been lysed. The appearance of eGFP-positive plaques, indicates hMSC lysis and spread of viral progeny to tumor cells.
To assess viral spread to the UD-SCC-2 cells, the number of single cells (= non-lysed hMSCs) and the number of eGFP-positive plaques were counted using a fluorescence microscope and set into proportion every 24 h. In addition, the plaque diameter was determined ( Figure 4A-C). Next, we determined the amount of viral progeny produced at different time points by hMSCs infected with HAdV-5-HexPos3 with or without life cycle modification. Every 24 h, cells and supernatant were harvested, and the number of viral particles produced per hMSC was quantified by qPCR ( Figure 3).
Generation of viral progeny from hMSCs was demonstrated in all samples, with no significant differences between the different replication-competent vectors or the pMOI used for infection. In BM-hMSCs, a maximum of about 1 × 10 4 viral particles produced per cell was measured, whereas production efficiency in A-hMSCs was slightly lower. Independent of the hMSC origin, the maximum amount of viral progeny was already reached 48-72 h post-infection. Thus, infection of hMSCs with different replication-competent HAdV-5-HexPos3 vectors results in equally efficient viral replication.

Replication-Competent HAdV-5-HexPos3 Vectors Carrying the E1B-∆19K Mutation Show Accelerated Replication and Release of Mature Viral Progeny in hMSCs
To confirm that viral progeny released from hMSCs are mature and infectious, we infected BM-and A-hMSCs with the different replication-competent HAdV-5-HexPos3 vectors (pMOI 900) and seeded them onto a dense layer of UD-SCC-2 cells. As all vectors used for infection carried an eGFP transgene, infected hMSCs were detected as single eGFP-expressing cells in the wells. This enabled the analysis of virus release kinetics: if only single cells are visible in the well, the hMSCs have not yet been lysed. The appearance of eGFP-positive plaques, indicates hMSC lysis and spread of viral progeny to tumor cells.
To assess viral spread to the UD-SCC-2 cells, the number of single cells (=non-lysed hMSCs) and the number of eGFP-positive plaques were counted using a fluorescence microscope and set into proportion every 24 h. In addition, the plaque diameter was determined ( Figure 4A-C).  vector at pMOI 100, 300, or 900. Every 24 h, cells and supernatant were harvested together to determine the amount of encapsidated adenoviral genomes by qPCR. Therefore, the harvested cells were disrupted in three freeze-thaw cycles and the lysate was treated with DNAse I to digest nonencapsidated viral genomes. Subsequently, samples were analyzed by qPCR to determine the number of adenoviral genomes using primers binding to the adenoviral E4 region. Results are shown as mean ± standard deviation of biological duplicates, each determined in technical triplicates.   The plaque assays revealed a highly accelerated release of viral progeny for hMSCs infected with HAdV-5-HexPos3-E1A-∆24-E1B-∆19K. Already 48 h post-infection, first plaques were visible, and most of the single cells were lysed 96 h post-infection. In contrast, in samples with hMSCs infected with HAdV-5-HexPos3 (no life cycle modification) or HAdV-5-HexPos3-E1A-∆24, only single eGFP-expressing cells were detected until 72 h post-infection, and even after more than 200 h, single eGFP-expressing hMSCs were still visible. This difference in replication kinetics was also reflected in the determined plaque diameters. While the plaque diameter was already >1 mm at 216 h after infection of hMSCs with the E1B-∆19K mutant, plaques induced by hMSCs infected with the two other viruses were still <0.5 mm in diameter. Results were comparable independent of hMSC origin (data not shown).

Migration of hMSCs Is Not Inhibited by Virus Infection In Vitro
Equally essential as efficient virus replication in hMSCs is their migration toward tumor cells. As virus infection results in substantial interference with cellular homeostasis, we aimed to investigate whether the migration behavior of hMSCs was altered upon virus infection. First, we analyzed the migration of non-infected BM-and A-hMSCs toward the conditioned medium of the HNSCC cell lines UD-SCC-2 and UD-SCC-6 using a Boyden chamber assay ( Figure 5A,B). Human MSCs were attracted by the conditioned media, as significantly more cells had migrated toward the tumor cell-conditioned medium compared to the unconditioned control medium. No difference was observed between the conditioned medium of the two HNSCC cell lines. Next, we analyzed the migration of hMSCs infected with the HexPos3 variant of the different replication-competent HAdV-5 vectors. For both types of hMSCs, no statistically significant difference in migration toward UD-SCC-2 conditioned medium was detected when infected hMSCs were compared to their non-infected counterparts ( Figure 5C,D).

No Migration of hMSCs toward a Xenograft HNSCC Tumor In Vivo
In order to evaluate hMSC migration toward HNSCC tumors in vivo, two different xenograft mouse models were established. To this end, 10 HNSCC cell lines were screened to determine how well they are transduced with HAdV-5-HexPos3 particles (Supplementary Figure S1A). Out of this portfolio, the HNSCC cell lines UD-SCC-2 and UD-SCC-6 were selected to establish a tumor model in NSG mice as these cell lines were efficiently transduced with both HAdV-5 and HAdV-5-HexPos3 vectors, showed suitable but not too fast cell growth in vitro, and represent the two major groups of HNSCCs: human papilloma virus (HPV)-positive (UD-SCC-2, tested positive for HPV-16) or -negative (UD-SCC-6) tumors (Supplementary Figure S1B) [1]. Upon subcutaneous injection of 2 × 10 6 cells into the left flank of the NSG mice, we observed a 100% take rate, homogeneous tumor growth (Supplementary Figure S1C,D), and a high level of vascularization (Supplementary Figure S1E).

No Migration of hMSCs toward a Xenograft HNSCC Tumor in Vivo
In order to evaluate hMSC migration toward HNSCC tumors in vivo, two different xenograft mouse models were established. To this end, 10 HNSCC cell lines were screened to determine how well they are transduced with HAdV-5-HexPos3 particles (Supplementary Figure S1A). Out of this portfolio, the HNSCC cell lines UD-SCC-2 and UD-SCC-6 were selected to establish a tumor model in NSG mice as these cell lines were efficiently transduced with both HAdV-5 and HAdV-5-HexPos3 vectors, showed suitable but not too fast cell growth in vitro, and represent the two major groups of HNSCCs: human papilloma virus (HPV)-positive (UD-SCC-2, tested positive for HPV-16) ornegative (UD-SCC-6) tumors (Supplementary Figure S1B) [1]. Upon subcutaneous injection of 2 × 10 6 cells into the left flank of the NSG mice, we observed a 100% take rate, In order to analyze the migration of BM-hMSCs toward these xenograft tumors, cells were transduced with a replication-deficient HAdV-5 vector carrying a firefly luciferase gene (pMOI 1000). To enable efficient transduction of hMSCs (≥80% transduced cells), hFX was used as a transduction enhancer. Three hours post-transduction, BM-hMSCs were injected into the tail vein of tumor-bearing mice (14 or 21 days of tumor growth). Subsequently, firefly luciferase-expressing hMSCs (and a negative control having received non-transduced MSCs) were tracked using the IVIS 200 system. In animals injected with firefly luciferase-expressing BM-hMSCs, a decreasing luminescence signal was detected at the injection site and in the lungs, being strongest at 24 h post-hMSC injection ( Figure 6A). Unexpectedly, no luminescence signal was detected at the tumor site at any time, although suitable vascularization of the tumors was confirmed by immuno-histochemical staining on the endothelial cell marker CD31 (Figure 7A,B).  As an alternative approach, we analyzed whether i.p.-injected BM-hMSCs migrate toward the subcutaneous xenograft tumors (injection after 21 days of tumor growth). Strong luminescence signals were detected in the abdomen and scrotum. (Figure 6B). Tumors of some i.v.-or i.p.-injected mice were removed 72 h after BM-hMSCs injection to analyze the luminescence of the isolated tumors directly. No luminescence signal was detected in the isolated tumors (Supplementary Figure S2). Quantitative PCR analysis of tumor tissue was performed to determine intratumoral vector genome copy numbers. However, viral DNA was not detected in any of the samples ( Figure 7C). In contrast, adenoviral DNA was detected in the lungs of mice injected with transduced hMSCs i.v. No differences were observed between the UD-SCC-2 and UD-SCC-6 tumor models. Due to the lack of hMSC migration to the xenograft tumors, we did not perform any long-term efficacy studies with hMSCs infected with replication-competent oncolytic adenovirus in this model.

No Migration of hMSCs toward a Xenograft HNSCC Tumor in a Chorioallantoic Membrane Model
In addition to the murine tumor models, we investigated the migration of BM-hMSCs in a recently published HNSCC CAM model [40]. For the engraftment of a tumor on the CAM, UM-SCC-11B cells were used. Four days after tumor engraftment, hMSCs (nontransduced or transduced with an eGFP-expressing HAdV-5-HexPos3 vector) and pure HAdV-5 vector were injected i.v. Additionally, transduced eGFP-expressing hMSCs were injected intratumorally. Delivery of pure HAdV-5 vector to the tumor upon i.v. injection was detected by qPCR and fluorescence microscopy 48 h post-injection ( Figure 8). In addition, intratumorally injected hMSCs were detected by fluorescence microscopy in the xenograft tumor ( Figure 8B). However, neither qPCR analysis nor fluorescence microscopy revealed any tumor infiltration by i.v.-injected hMSCs.

Discussion
Using hMSCs as carrier cells for oncolytic adenoviral vectors (oAVs) holds promise to enable systemic delivery and expands therapeutic options for the treatment of a range of cancer entities, including head and neck squamous cell carcinomas (HNSCCs). In this study, we investigated whether the recently published Hexon-modified vector HAdV-5-HexPos3 can serve as a tool for hMSC-based oncolytic therapy [29]. Moreover, we analyzed hMSC migration toward HNSCC xenografts in vivo and in ovo.
Investigation of different replication-competent HAdV-5-HexPos3 vectors revealed that increased infection efficiency of Hexon-modified particles led to efficient virus replication in hMSCs. Moreover, mature viral progeny are released and can re-infect HNSCC cell lines in vitro. This holds not only true for life cycle-unmodified HAdV-5-HexPos3 but also for the life cycle-modified conditionally replicating vectors HAdV-5-HexPos3-E1A-∆24 and HAdV-5-HexPos3-E1A-∆24-E1B-∆19K. The E1B-∆19K modification hereby highly accelerated the release of viral progeny. This is in line with previous reports demonstrating that E1B-∆19K vectors are very potent oncolytic adenoviruses, showing increased tumor cell killing as well as accelerated virus release in tumor cells and hMSCs compared to life cycle-unmodified HAdV-5 [26,36,[41][42][43][44]. However, the mechanisms leading to this accelerated virus release and subsequent spread in tumor tissue remain unclear. As the E1B 19K protein is a Bcl-2 analog functioning as an apoptosis inhibitor, increased apoptosis in HAdV-5-E1B-∆19K-infected cells might explain accelerated virus release [45]. Moreover, E1B 19K is assumed to inhibit the proteolytic degradation of the nuclear lamins, a component of the nuclear membrane important for the stability of the nucleus [46,47]. The absence of E1B 19K could lead to proteolysis of the nuclear lamins, resulting in the disruption of the nuclear envelope and accelerated release of viral progeny. In addition, there is evidence that E1B 19K counteracts the adenovirus death protein (ADP), which is critical for the lysis of infected cells by adenoviruses [48]. Independently of the exact understanding of the mechanism, our results clearly demonstrate that the total number of viral particles produced per cell is not impeded. Although this might be an advantage concerning possible oncolytic effects, it is essential to consider that accelerated release of viral progeny also implies reduced time for carrier cells to migrate toward their target.
In our studies, we observed that non-infected and infected hMSCs efficiently migrated toward HNSCC cell lines in vitro. However, this tumor-homing capability was neither observed in two human HNSCC xenograft tumor mouse models nor in a human HNSCC xenograft tumor chorioallantoic membrane (CAM) model. In the mouse models, we detected most of the hMSCs in the lungs after i.v. injection, an observation reported by other researchers, too [49,50]. Nevertheless, successful delivery of oncolytic adenovirus using hMSCs as carrier cells was previously reported [21,25,[50][51][52]. For example, Zielske et al. showed that hMSCs administered i.v. into the tail vein infiltrated the tumor tissue of subcutaneous colon cancer, breast cancer, and HNSCC xenografts in mice [52]. Interestingly, the lowest number of hMSCs was detected in the HNSCC xenografts, suggesting that hMSCs might be comparatively weakly attracted by this type of tumor cells [52]. During the establishment of our Boyden chamber migration assay, we also observed that BM-hMSCs migrated stronger to Huh7 or A549 cells compared to the HNSCC cell lines UD-SCC-2, UD-SCC-6, and UM-SCC-11B, supporting this hypothesis (data not shown). Reports of MSCs migrating to the tumor site have not only been published after i.v. but also after i.p. administration of the carrier cells [25,51,53]. For example, Moreno et al. showed high antitumor efficacy of menstrual blood-derived hMSCs loaded with oncolytic adenovirus upon i.p. injection in a lung adenocarcinoma xenograft tumor model in mice [51]. Thus, in contrast to our observations, several publications show that migration of MSCs toward (xenograft) tumors have been observed in murine models. Kaczorowski et al. reported migration of i.v.-injected hMSCs toward MiaPaCa-2 cells in the CAM model; however, we did not observe any migration of hMSCs to our xenograft tumors in ovo [54].
Apparently, the HNSCC cell lines used in our study did not sufficiently attract hMSCs in vivo and in ovo. Indeed, inefficient homing of hMSCs toward the target cells has often been described in the literature, and most studies show that only a limited proportion of hMSCs infiltrate the tumor tissue [49,50,52]. Human MSCs were shown to be capable of infiltrating a variety of tissues, including tumors, kidneys, lungs, liver, thymus, and skin. Nonetheless, achieving a more targeted migration of MSCs toward the tissue of interest is one of the major challenges in the use of hMSCs as carriers [55,56]. To reach this goal, researchers evaluate different strategies, such as increasing the chemokine secretion of tumor cells by radiation or improving the inherent migration ability of hMSCs by increasing the expression of proteins relevant for their homing such as the C-X-C chemokine receptor type 4 (CXCR-4) [52,[57][58][59][60][61]. However, to guide hMSCs specifically and efficiently to a target tissue, a better understanding of the mechanisms underlying hMSC migration is mandatory and must be further investigated. Apart from the tumor cells, also the cells of the tumor microenvironment (TME), including endothelial cells, fat cells, and immune cells, are known to secrete chemoattracting molecules [62]. A combination of these molecules derived from the different cell types of the TME might be essential in the context of hMSC-based oncolytic virotherapy to efficiently attract hMSCs to the tumor. Since we used an immuneincompetent model, the lack of immune cell-associated chemoattracting molecules might be one possible explanation for the lack of hMSC homing, as these cells are known to secrete high amounts of cytokines and growth factors [63][64][65]. While immunocompetent (HNSCC) tumor models are available for several species (mice, hamsters, nonhuman primates . . . ), and the isolation of MSCs from different species could be easily implemented, the use of such a model introduces a different problem: most organisms are not (completely) permissive for HAdV-5 infection. Thus, infection of MSCs from another species with HAdV-5 would not result in the release of viral progeny, disabling the evaluation of oncolytic effects. An intensive effort has been made to find a permissive animal model; however, to date, no really suitable animal model could be identified [66]. To sum up, this illustrates the complexity of preclinical hMSC-based oncolytic adenovirus therapy and highlights that the establishment of a more suitable in vivo model is indispensable but not yet realizable.

Conclusions
In summary, we show that the Hexon-modified HAdV-5-HexPos3 mutant vector is a powerful tool for the development of hMSC-based oncolytic virotherapy. Virus replication, the efficient spread of viral progeny to HNSCCs, and the maintained migration ability of infected hMSCs to HNSCCs were demonstrated in vitro. Moreover, our data highlight the need to gain a better understanding of mechanisms underlying the homing of hMSCs in vivo and shows that the suitability of immune-incompetent xenograft tumor models for the analysis of tumor homing of i.v.-or i.p.-administered hMSCs is questionable.

Patents
A patent application has been filed relating to this work by Ulm University (European patent application #19204420.4).

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/v15010218/s1, Figure S1: Establishment of a xenograft HNSCC model in NSG mice; Figure S2: HNSCC tumors from mice injected with firefly luciferase-expressing hMSCs did not show intratumoral luciferase activity.