Delivery of oncolytic vaccinia virus by matched allogeneic stem cells overcomes critical innate and adaptive immune barriers

Background Previous studies have identified IFNγ as an important early barrier to oncolytic viruses including vaccinia. The existing innate and adaptive immune barriers restricting oncolytic virotherapy, however, can be overcome using autologous or allogeneic mesenchymal stem cells as carrier cells with unique immunosuppressive properties. Methods To test the ability of mesenchymal stem cells to overcome innate and adaptive immune barriers and to successfully deliver oncolytic vaccinia virus to tumor cells, we performed flow cytometry and virus plaque assay analysis of ex vivo co-cultures of stem cells infected with vaccinia virus in the presence of peripheral blood mononuclear cells from healthy donors. Comparative analysis was performed to establish statistically significant correlations and to evaluate the effect of stem cells on the activity of key immune cell populations. Results Here, we demonstrate that adipose-derived stem cells (ADSCs) have the potential to eradicate resistant tumor cells through a combination of potent virus amplification and sensitization of the tumor cells to virus infection. Moreover, the ADSCs demonstrate ability to function as a virus-amplifying Trojan horse in the presence of both autologous and allogeneic human PBMCs, which can be linked to the intrinsic immunosuppressive properties of stem cells and their unique potential to overcome innate and adaptive immune barriers. The clinical application of ready-to-use ex vivo expanded allogeneic stem cell lines, however, appears significantly restricted by patient-specific allogeneic differences associated with the induction of potent anti-stem cell cytotoxic and IFNγ responses. These allogeneic responses originate from both innate (NK)- and adaptive (T)- immune cells and might compromise therapeutic efficacy through direct elimination of the stem cells or the induction of an anti-viral state, which can block the potential of the Trojan horse to amplify and deliver vaccinia virus to the tumor. Conclusions Overall, our findings and data indicate the feasibility to establish simple and informative assays that capture critically important patient-specific differences in the immune responses to the virus and stem cells, which allows for proper patient-stem cell matching and enables the effective use of off-the-shelf allogeneic cell-based delivery platforms, thus providing a more practical and commercially viable alternative to the autologous stem cell approach. Electronic supplementary material The online version of this article (10.1186/s12967-019-1829-z) contains supplementary material, which is available to authorized users.

The immunosuppressive properties of mesenchymal stem cells have been investigated extensively and supported by clinical data demonstrating their ability to avoid immune recognition and survive even in MHCmismatched allogeneic recipients [11,[46][47][48]. What remains still unclear, however, is how the immunosuppressive properties of MSCs will be affected by the progression of vaccinia virus infection especially in allogeneic MHC-mismatched settings. Importantly, MSCmediated immunosuppression is subject to differential modulation by both type I [36,49] and II interferons [22,[50][51][52][53][54][55], which are similarly involved in the control of vaccinia virus infection [2,3,56,57]. While type I interferons are secreted by any cell type in response to viral infection, including the MSC, Interferon gamma (IFNγ) is a type II interferon that is produced by a limited number of immune cells like NK, NKT and T cells in response to virus infection [58] or, as we demonstrate, the carrier stem cells, particularly in some unfavorable allogeneic MHC-mismatch settings. These conflicting roles of IFNγ to limit virus spread through the induction of an anti-viral state but at the same time to stimulate MSCmediated immunosuppression counteracting anti-viral immunity might prove to be a critically important determinant of therapeutic success.
We hypothesized that the complex interplay between vaccinia virus, ADSCs and immune cells could have significant impact on the ability of stem cells to function as a Trojan horse that can amplify and deliver oncolytic vaccinia virus. Here we show that interferons protect stem cells against vaccinia virus infection but compromise their function as a Trojan horse. The IFNγ responses to the virus and allogeneic stem cells alone or in combination, however, appear to be highly variable and patientspecific. Our studies indicate that these differences can be associated with subtle allogeneic NK-and T cellmediated cytotoxic and cytokine responses that can result in the inactivation or complete rejection of the Trojan horse. These findings have significant implications for the development of cell-based delivery platforms for oncolytic viruses and suggest the need for proper screening and patient-specific matching to enable the successful use of off-the-shelf allogeneic cell carriers, as opposed to the more expensive personalized autologous stem cell approach.

Adipose-derived stem cells isolation and culture
Non-cancer donor SVF and PBMC were obtained as part of an IRB-approved protocol after informed written consent (International Cell Surgical Society; IRB# ICSS-2016-024). Fresh SVF fractions were plated to attach overnight and next day were washed to remove unattached cells and debris. Media was changed every 3-4 days until the mesenchymal stem cells start to grow and reach 80% confluency. Cell were expanded to 80% confluency and passaged every 3-4 days using Try-pLE ™ Express (Life Technologies, (1×), no phenol red, Cat# 12604021, 3 min 37C incubator) for up to 10 passages. Note that P12 has normal mesenchymal look and morphology but manifested evidence of gradual loss of immunosuppressive ability. ADSCs were expanded and maintained in 5% Human Platelet Extract (Cook Regentec, Stemmulate, PL-SP-100) in DMEM supplemented with l-Glutamine and Pen/Strep.

Generation of adipose-derived MSC constitutively expressing eGFP
RM20 adipose-derived stem cells at passage 0 were engineered to express eGFP under the control of the CMV promoter. A Lentiviral vector (VectorBuilder) containing eGFP was used to introduce eGFP for constitutive expression. 10,000 eGFP-positive cells were sorted at passage 1 and subsequently at passage 2 using the Bio-RAD S3 Cell Sorter. eGFP expression was confirmed by flow cytometry and fluorescence microscopy using the Keyence All-in-one Fluorescence Microscope BZ-X700 Series.

PBMC assays
PBMC were isolated through standard Ficoll protocol (Ficoll-Paque Plus, GE Healthcare, cat# 95021-205) and co-cultured (100 μl) with ADSC (50 μl) plus minus vaccinia virus (50 μl) for 48 h on 96-well flat-bottom plates and in a total of 200 μl R10 medium (RPMI 1640 supplemented with 10% FBS, l-Glutamine and Pen/Strep). In some experiments the virus (50 μl) and stem cells (50 μl) were premixed and agitated on an orbital shaker at 37C (incubator) for 1 h, and then (100 μl of the mix) was added to the PBMCs without additional washing of any unbound virus. At the end of the 48 h incubation period the cells were recovered for staining and flow analysis directly or after an additional 4-5 h stimulation with K562 cells or PMA/Ionomycin (50 μl) with Monensin/ Brefeldin A (50 μl) as needed.

Intracellular stain
To evaluate IFNγ production in activated NK, NKT and T cells, in some experiments Brefeldin A was added at 1:1000 an hour after stimulation or 4 h prior to recovery and surface staining (BioLegend, cat# 420601-BL, 1000×). Monensin and Brefeldin A were added together when cells were to be evaluated for both IFNγ production and CD107a surface exposure. Following standard surface and viability staining cells were processed using the eBioscience Intracellular Staining Buffer Set (Ther-moFisher, cat# 00-5523). Briefly, following surface staining with or without anti-CD107a-AlexaFluor 488, cells were fixed for 30 min with 1 part Fixation/Permeabilization Concentrate (cat# 00-5123) and 3 parts of Fixation/Permeabilization Diluent (cat# 00-5223), washed twice with 200 μl/well Permeabilization Buffer 10× (cat# 00-8333), diluted 1:10 in double distilled water), and stained with anti-human IFNγ-APC antibody (Bio-Legend, cat# 502512, at 1:50) in Permeabilization Buffer for 1 h at RT. Cells were washed twice in Permeabilization Buffer, fixed in 2% PFA in 1xPBS for 15 min at RT, washed again with FACS Buffer to remove PFA and analyzed on BD FACSAria II.

NK cell immunosuppression
To evaluate the extent of ADSC-induced immunosuppression following the 48 h co-culture of PBMC with autologous or allogeneic ADSCs in the presence or

Microscopy
Time course microscopic observations of virus infection were done on a Keyence All-in-one Fluorescence Microscope BZ-X700 Series. ADSCs were engineered to express eGFP and were followed on the GFP channel (1 s exposure) while virus infection with the TurboFP635engineered L14 virus was monitored on the TRITC channel (3 s exposure). Images at 4× or 10× magnification were collected and overlaid with bright field (phase contrast, 1/50 s exposure).

HLA and KIR analysis
HLA and KIR/MIC typing analysis was done through NGS by ProImmune (Oxford, UK) and Scisco Genetics (Seattle, USA), respectively. The presence/absence of the known KIR ligands A3/A11 (HLA-A), Bw4 (HLA-B) and C1/C2 (HLA-C) epitopes in the HLA alleles of our PBMC and ADSC donors was taken from http://www. dorak .info/mhc/nkcel l.html. The − 21 M/T (Methionine/ Threonine) dimorphism at the anchor amino acid from the leader sequence that predicts strong/weak binding and presentation of HLA-B-derived leader peptides by HLA-E, which provides inhibitory signaling though the NKG2A/CD94 receptors on NK cells, was taken from the Immuno Polymorphism Database (IPD) at http://www. ebi.ac.uk/ipd/.

Data analysis and statistics
Data was plotted and analyzed for statistical significance using licensed Graph Prism software. Statistical significance was evaluated with the two-tailed Student's T-test, p < 0.05, and statistically significant correlation was shown with the Pearson coefficient and corresponding p values.

Adipose-derived stem cells provide potent amplification of vaccinia virus that can be restricted by the induction of IFN-mediated anti-viral state
Tumor cells frequently harbor defects in type I interferon signaling that render them sensitive to oncolytic virus infection [59]. In addition to type I interferons, the type II IFNγ is also known for its critical role in restricting vaccinia virus infection in vivo through the induction of the so-called anti-viral state in both healthy and some interferon-responsive tumor cell lines [4]. Adipose-derived stem cells are normal untransformed mesenchymal stem cells that can be rapidly expanded ex vivo and can be used as a vehicle for delivery of oncolytic vaccinia virus. We investigated the potential of these cells to amplify vaccinia virus as well as to respond to the protective anti-viral effects of interferons. Here, we demonstrate that adipose-derived stem cells are highly permissive for vaccinia virus infection, showing amplification potential equivalent to levels observed in the highly permissive A549 human lung carcinoma cells. Both type I and II interferons protected stem cells against vaccinia virus infection ( Fig. 1a, b, Additional file 1: Figure S1A), consistent with these cells being untransformed and having functional anti-viral interferon responses. Protection, however, was less efficient when interferon was given concurrently rather than 24 h prior to virus exposure. The combination of type I and II interferon didn't further enhance protection, indicating the absence of a significant synergistic effect (Additional file 1: Figure S1B). Of note, the anti-viral state induced by interferon treatment was stable, lasting for several days after a transient 24 h exposure to IFNγ (Fig. 1c, Additional file 1: Figure S1C). Thus, type I and II interferon responses, while protecting the stem cells against virus infection and potentially improving their immunosuppressive abilities [51], can have the unfortunate effect of also compromising their ability to deliver and amplify vaccinia virus in vivo.

ADSCs promote the oncolysis of resistant tumor cell lines through a combination of virus amplification, tumor cell recruitment and secretion of factors sensitizing the resistant tumor cells to virus infection
Human tumors demonstrate varying sensitivity to vaccinia virus infection, with some tumors being more resistant than others. We specifically wanted to evaluate the effect of using ADSC to enhance the delivery of vaccinia virus in the context of resistant tumors. Stem cell delivery of oncolytic viruses can be used as a strategy to assist the oncolysis of resistant tumors by the introduction of highly sensitive cells that provide and sustain higher initial local virus production, thereby facilitating virus spread throughout the tumor. The murine B16 melanoma cells are known to be resistant to vaccinia virus infection and we tested whether in the presence of human adipose-derived stem cells the B16 cells can be targeted more effectively (  we could observe that in confluent environments stem cells tended to cluster together while at the same time attracting the unlabeled melanoma cells. In the presence of the virus this attraction resulted in the formation of highly infected yellow stem cell clusters surrounded by intensively red-colored infected B16 cells. These effects were associated with a dramatically improved oncolysis of the monolayer of resistant murine B16 melanoma cells (Fig. 2a). Improved targeting of the resistant B16 cells was also observed with ADSC derived from another donor (Additional file 2: Figure S2B). Successful oncolysis, however, was compromised if the stem cells were pretreated with IFNγ (Additional file 2: Figure S2C) or were insufficient in numbers (Additional file 2: Figure S2D), suggesting that the observed antitumor potential was dependent on amplification of the virus by the stem cells.
Our findings indicate that ADSCs have the unique property of both amplifying the virus (approx. 10,000-fold or 5000 pfu/cell) and spreading it to the tumor cells, which can be attributed to higher local multiplicity of infection (MOI) as well as some form of chemoattraction. Importantly, the viral titers recovered from these cocultures demonstrated that the viral output of the mixed stem and melanoma cells was greater than the combined outputs of the individually infected cells (Fig. 2b), suggesting that the highly permissive cancer or stem cells can sensitize the resistant melanoma cells to infection with vaccinia virus. We could further demonstrate that this effect is at least in part due to the secretion of still unidentified soluble factors present in the supernatants of ADSCs (Fig. 2c). Supernatants from different ADSCs donors could provide similar sensitization of both the murine B16 and the extremely resistant human K562 cancer cells (Additional file 2: Figure S2E), but the observed potentiating effects on the frequency of infected cells and virus amplification were relatively small (approximately twofold) and importantly insufficient for the eradication of these resistant cancer cells (Additional file 2: Figure S2F-H), indicating that successful therapy of resistant tumors might require both the sensitization and the amplification properties of the stem cells, as well as their ability to recruit murine and human tumor cells (Fig. 2a, Additional file 2: Figure S2H).

ADSCs can be immunosuppressive towards NK cells in both autologous and allogeneic settings
We hypothesized that the immunosuppressive potential of MSCs might be a key factor for their use as a Trojan horse delivery method for oncolytic viruses. It is frequently assumed that the immunosuppressive properties of MSCs would allow them to evade immune recognition and rejection even in highly unfavorable allogeneic settings [11]. To evaluate the potential of ADSCs to overcome allogeneic immune barriers and function as an effective Trojan horse, we tested their ability to immunosuppress and amplify vaccinia virus in the presence of allogeneic human PBMCs. MHC mismatches can typically trigger both NK and T cell responses, with NK cells representing the most significant initial barrier to both vaccinia virus and the Trojan horse, which is linked to their innate nature, high frequencies and ability to immediately respond to "missing self " due to allogeneic differences or vaccinia virus-induced MHCdownregulation, and without the need for T cell clonal expansion [58,60]. Surface exposure and upregulation of CD69 and CD107a have been extensively used to study the immunosuppressive effects of MSCs on NK or T cell activation and cytotoxic functions, respectively [61][62][63]. Preliminary experiments with total PBMC, rather than purified immune subpopulations, indicated that a coculture of at least 48 h was necessary for consistent detection of largely PMA/Ionomycin-reversible stem cell-mediated immunosuppression (Fig. 3a, b). This was a critical improvement over previous reports that were largely focused on the suppressive effects of MSC on purified NK and T cell populations using longer incubation times and exogenous cytokine support, thus introducing activation bias and ignoring the important role of innate and adaptive immune cell crosstalk [20,39,64,65]. The co-culture with ADSCs demonstrated potent dosedependent immunosuppressive properties against NK cells stimulated by brief exposure to the physiologically relevant K562 target cells in both autologous and allogeneic settings (Fig. 3a, b, Additional file 3: Figure S3A). Of note, in the co-culture settings the allogeneic stem cells failed to trigger any direct NK or T cell responses, which were decreased below the background levels of activation as measured in the PBMC alone controls (Fig. 3a, b, black bars). We were encouraged by our initial data confirming the immunosuppressive potential of ADSC in allogeneic settings but also wanted to explore the possibility that the actual potency of immunosuppression may be patientspecific and subject to certain allogeneic restrictions. Our optimized co-culture assay was specifically designed to reveal patient/MSC recipient-specific differences in both the immunosuppressive and virus amplification abilities of the stem cells.

ADSCs can overcome allogeneic immune barriers and amplify VV even in the presence of some allogeneic PBMCs
We next evaluated the potential of ADSCs not only to immunosuppress but also to amplify vaccinia virus when co-cultured with PBMC from other individuals and against possible allogeneic immune barriers. We hypothesized that an unfavorable allogeneic PBMC environment might result in elimination or inactivation of the stem cells before they have a chance to amplify vaccinia virus. Surprisingly, the presence of allogeneic PBMCs didn't interfere with virus amplification, suggesting that at least in some "permissive" cases the allogeneic stem cells remain "under the radar" and avoid immune recognition even under conditions of highly inflammatory virus infection (Fig. 3c). Moreover, in the presence of the allogeneic PBMCs the overall virus output was increased relative to the combined output of stem cells and PBMCs when infected in separation, suggesting that the stem cells can sensitize some subpopulations of normal PBMC to vaccinia virus infection in a way equivalent to their effect on tumor cells (Fig. 2b, c). These observations raised the relevant question as to whether successful and unrestrained virus amplification by the Trojan horse could be attributed to potent stem cell-mediated inhibition of anti-viral responses irrespective of any allogeneic differences or was it patient/MSC recipient-specific and limited to cases of a particularly close MHC match, hence not necessarily applicable to a larger cohort of MHC-diverse recipients.

The potential of allogeneic stem cells to overcome immune barriers correlates with their ability to suppress virus-induced T, NK and NKT cell responses
The ability of the Trojan horse to amplify vaccinia virus in immunocompetent recipients might be critically dependent on the potential of stem cells to successfully inhibit anti-viral innate and adaptive immune responses. We therefore tested the ability of allogeneic ADSC from the RM35 donor to specifically suppress virus-induced responses mediated by NK and T cells in a new cohort of two PBMC donors/MSC recipients and with the goal to reveal possible patient-specific restrictions (Fig. 4a, Additional file 4: Figure S4A). Relative to T cells, NK cells were indeed the major responding population with the combination of virus and low doses of stem cells resulting in activation of more than 80% of the NK cells (Fig. 4b, Additional file 4: Figure S4B). Lower doses of the stem cells were insufficient for immunosuppression and unexpectedly increased virus-induced immune responses, possibly reflecting significantly augmented virus amplification. Higher doses of ADSCs, however, provided potent suppression of the weaker vaccinia virus-induced T cell responses in a dose-dependent fashion, indicating that stem cell-mediated immunosuppression overcomes the immune-stimulatory effect of augmented virus amplification. Suppression of anti-viral NK cell responses was evident only at the highest stem cell doses, and was not consistent across the two allogeneic blood donors tested. Interestingly, the NK cells from one of the blood donors responded directly to the allogeneic ADSCs (RM48, red empty bars). Using a standard 4 h K562 NK stimulation assay at the end of the 48 h PBMC/ADSCs/VV co-culture experiment, we demonstrate that the inconsistent suppression of the anti-viral NK responses might correlate with loss of the stem cells' immune-privileged status and immunosuppressive abilities (Fig. 4b, NK panel, solid red/black bars versus K562 CTRL). We also identified a NKp46 + CD3 + NKT-like population of cells that responded vigorously to virus infection with upregulation of activation markers. This was also the only population of cells that manifested ability for rapid and selective expansion in response to vaccinia virus (Additional file 4: Figure S4B), consistent with the already established role of NKT cells in the control of virus infections [66].

The potential of allogeneic stem cells to function as a Trojan horse is restricted by patient-specific differences suggesting that proper matching will be required
Our next goal was to evaluate the immunosuppressive and virus amplification abilities of the other allogeneic ADSC derived from the RM20 donor and to test it in a larger cohort of blood donors/MSC-recipients to get a better understanding of the scope and magnitude of patient-specific restrictions. The low immunogenicity of stem cells has been associated with the secretion of various immunosuppressive factors as well as their extremely low level of MHC class I expression combined with the total absence of MHC class II or co-stimulatory molecules like CD80/86 or CD40. However, the potential of allogeneic stem cells to immunosuppress and "remain under the radar" [11,67] can be significantly undermined by the progression of virus infection, which can not only wipe them out but also sensitize them to immune recognition by NK cells and T cells through Interferon-driven upregulation of MHC class I/II or surface expression of various infection-related stress molecules [24]. Thus, in the IFNγ-rich environments associated with vaccinia virus infection, the degree of matching between the allogeneic stem cells and the patient could control their ability to retain immunosuppressive properties and function as a Trojan horse. We validated our previous data (see Fig. 4b) showing that the same RM20 ADSC stem cells when tested against a panel of 4 new allogeneic PBMC donors can trigger direct NK, T, and NKT cell responses in a patient-specific manner (Fig. 5a, Additional file 5: Figure S5A). We observed a very high interpatient variability in the NK, NKT and T cell responses to both the allogeneic stem cells and to the virus. Moreover, the immunosuppressive properties of ADSCs appear to be failing in unfavorable allogeneic settings where the stem cells lose their immune privileged status and activate NK and T cells directly, even in the absence of the virus (Fig. 5b, Additional file 5: Figure S5B). We also demonstrate that these immunological differences have significant impact on the stem cells' virus amplification potential. Importantly, improperly matched stem cells and blood donors representing potential MSC-recipients can completely abrogate the virus amplification potential of the allogeneic ADSC, thus revealing critical patientspecific differences that could lead to "permissiveness" or "resistance" to the Trojan horse (Fig. 5c). Additional correlative analysis of the immune responses against the virus revealed that vaccinia virus induces highly coordinated NK, NKT and T cell responses (Fig. 5d, Additional file 5: Figure S5C). Similar correlation was evident for the NK and T cells responses against the allogeneic stem cells, but responsiveness to virus versus ADSCs was discordant and likely independent of each other (Additional file 5: Figure S5D). Of note, we identified a pair of blood donors who were respectively highly resistant (SIBD01) and permissive (SIBD02) to one of our established allogeneic ADSC lines (RM20) and these extreme cases were used to further analyze the underlying mechanisms of patient-specific resistance to the Trojan Horse.

Patient-specific immunological barriers regulate the survival of the Trojan horse and can limit the therapeutic potential of both genetically attenuated and wild type vaccinia virus strains
The total abrogation of the virus amplification potential of ADSC in certain allogeneic settings challenged their assumed immune privileged status and prompted us to investigate their survival in cocultures with allogeneic PBMC. We thought that immunological rejection could provide a possible explanation for the high resistance of some patients to an allogeneic Trojan horse. Monitoring stem cells' survival in complex cocultures is limited due to their strong surface adhesion and tendency to cluster together, making it challenging to disaggregate them without losing viability.
To visually evaluate the fate of allogeneic stem cells cocultured with PBMCs from resistant versus permissive donors and in the presence or absence of vaccinia virus infection, we utilized the eGFP-labelled RM20 ADSC line and the TurboFP635-engineered L14 (See figure on next page.) Fig. 5 The potential of allogeneic stem cells to function as a Trojan horse is restricted by patient-specific differences suggesting that proper matching will be required. a PBMC donors demonstrate highly variable responses to the allogeneic stem cells and the virus alone or in combination. Flow cytometry analysis of gated live NK, T, and NKT cells from 48 h cocultures of 250 k PBMCs from 4 different blood donors with 5 k-40 k allogeneic RM20 ADSC, and in the presence or absence of 5 k pfu of WT1 VV. Data show the percentage of CD69+ activated cells from each cell type. Bars represent duplicate measurements ± SD. Statistically significant differences (Student T-test, p < 0.05) versus corresponding PBMC without virus or stem cells controls (CTRL) or as indicated are marked with asterisks. b Comparison of the immunosuppressive potential of the allogenic RM20 ADSCs against the NK cells from two blood donors in which the stem cells demonstrate differential ability to "stay under the radar". Flow cytometry analysis of cocultures of PBMC from the SIBD01 and SIBD02 blood donors as in a followed by a 4 h stimulation with K562 cells to evaluate the extent of NK cell suppression. Bars represent duplicate measurements ± SD. Statistically significant differences (Student T-test, p < 0.05) versus the corresponding PBMC alone controls (CTRL) are marked with asterisks. In the K562 groups asterisks indicate statistically significant difference versus the corresponding K562 CTRL, which represents PBMC cocultured alone without ADSC, with or without virus (solid versus empty bars, respectively). c Plaque analysis of the 48 h cocultures as in b demonstrating that RM20 ADSC can amplify WT1 vaccinia virus only in the presence of allogeneic PBMC from the permissive SIBD02 but not the resistant SIBD01 blood donor. d Correlative analysis of NK, T, and NKT cell responses (% CD69 + normalized to untreated PBMC control) against the highest dose of the allogeneic ADSCs as in (a) and (Fig. 4b). Statistically significant correlations are indicated with corresponding p values and Pearson coefficients vaccinia virus. Parallel viral titer analysis of the cocultures also included the WT1(ACAM2000) virus, which is a clonal isolate of wild type vaccinia virus that lacks the genetically attenuating elimination of the TK locus. This is the FDA-licensed USA stockpile of smallpox vaccinia virus vaccine that has been extensively used in humans with a well-established safety profile. Comparative analysis of the survival and amplification potential of the allogeneic Trojan horse in the case of permissive versus resistant recipient indicated that resistance was associated with the rapid loss of the stem cells in the presence of the resistant allogeneic PBMCs and in the absence of virus infection. These same stem cells remained intact when cocultured with PBMCs from the permissive recipient (Fig. 6a). To make sure that the allogeneic immunological barriers cannot be easily overcome by simply giving a head start to the virus and to more closely mimic its potential clinical application in the future, we also infected the Trojan horse an hour before exposure to the PBMCs, which was sufficient for approximately half of the virus to get into the cells (Additional file 6: Figure S6A). Despite potentially enhancing stem cell infection and accelerating virus amplification, the 1 h head start had a very small overall effect on the amplification potential of the Trojan horse (Fig. 6b). Interestingly, the permissive SIBD02 PBMCs partially suppressed the amplification of the L14 but not the WT1 virus. Contrary to the Lister-based Turbo-FP635-engineered L14 virus, which has the TK locus inactivated/removed as in most genetically attenuated vaccinia virus strains, WT1/ACAM2000 is a wild type TK-positive Wyeth vaccinia virus that has demonstrated higher amplification potential and ability to overcome stronger allogeneic barriers, consistent with possible faster amplification cycle or augmented ability to evade anti-viral immunity. These advantages of the WT1 virus were nevertheless insufficient to overcome the immunological barriers in the resistant recipient and were only marginally improved by initiating virus infection of the stem cells an hour prior to co-culture with the allogeneic PBMCs. Consequently, patientspecific resistance to the Trojan horse represents an important and therapeutically significant barrier to the use of cell-based delivery vehicles for oncolytic viruses like vaccinia and might be even more relevant for viruses that are genetically attenuated to improve safety or tumor selectivity. The important patient-specific differences observed in our studies raised the question of whether resistance to an allogeneic Trojan horse reflects a simple random MHC mismatching phenomenon between the donor stem cells and certain recipients or that some patients manifest broader patterns of permissiveness and resistance to multiple allogeneic stem cell lines, which can be associated with other intrinsic and more complex immunological characteristics.

Patients' resistance to the Trojan horse is associated with possible HLA mismatches and the rapid induction of anti-stem cell cytotoxic and interferon responses
Comparative analysis of the resistant SIBD01 versus permissive SIBD02 blood donors tested against a panel of four available expanded allogeneic ADSCs indicated that they were broadly permissive and resistant, respectively (Fig. 7a). This broader permissiveness versus resistance correlated with partial matching mostly at the HLA-A and HLA-DP loci, with the broadly permissive donor having the most common HLA-A*02:01 allele, while the resistant one was HLA-A*01:01. HLA typing data alone are clearly insufficient to predict permissiveness versus resistance, as seen by the discordant case of an HLA-A*01:01 RM58 Trojan Horse (Fig. 7b). Rejection of closely HLA-matched cells can be alternatively explained by differences in the KIR Haplotype or in the balance of signaling through the NK cell inhibitory (KIR, NKG2A/ CD94) or stimulatory (KIR, NKG2D) receptors. Indeed, our analysis shows that the discordant HLA-A*01:01 case could be linked to an important C2 KIR ligand mismatch (Additional file 7: Figure S7A) that has already been associated with insufficient KIR inhibitory signaling and NK cell-mediated rejection of HLA-Haploidentical iPSC [68]. Further mechanistic studies revealed that, regardless of the degree of partial HLA or KIR/KIR Ligand match, the ability of the Trojan horse to efficiently amplify the virus was associated with the absence of significant anti-stem cell IFNγ (Additional file 7: Figure S7B) and cytotoxic (Additional file 7: Figure S7C) NK and T cell responses across all the four allogeneic ADSC lines tested (Fig. 7c, Additional file 7: Figure S7D, E). Analysis of the PBMCs from the highly resistant SIBD01 recipient reveals that the improperly matched allogeneic stem cells induce a detectable IFNγ response even in the absence of virus infection that appears to originate from both NK and T cells. While T cells represent the bulk of early IFNγ-producing cells, NK cells demonstrate the highest proportional cytotoxic activity suggesting the existence of important cross talk between the innate and adaptive immune cell populations. The PBMCs from the highly resistant SIBD01 donor also manifested a much higher frequency of CD3+ NKp46 + NKT-like cells relative to other blood donors tested, which could suggest an alternative explanation for the broader resistance to various allogeneic stem cell lines (Fig. 7d, Additional file 7: Figure S7F). The latter, however, were not found to be associated with significant IFNγ production or cytotoxic activity as measured by CD107a surface exposure at the 48 h coculture timepoint (Additional file 7: Figure S7B-D), but the potential involvement of NKT cells in the anti-viral response or the crosstalk between NK and T cells at earlier timepoints cannot be excluded. We specifically looked at anti-virus and anti-stem cell induced immune responses at earlier timepoints (Additional file 7: Figure S7G). Significant background immune cell activation was evident at 6 h that subsided by the 24 h timepoint. At 6 h only T and NKT cells from the resistant donor demonstrated stem cell-induced CD69 upregulation but not IFNγ responses. At 24 h all three subsets responded to the allogeneic stem cells by upregulation of the CD69 activation marker regardless of the virus. While the NK cell response was the most vigorous, only the NKT subset manifested statistically significant stem cell-induced IFNγ responses, suggesting that NKT cells might be the first immune cell subtype mounting effector cytokine responses against the allogeneic stem cells (Additional file 7: Figure S7G). The NK, NKT and T cells from the permissive donor didn't upregulate CD69 or produced IFNγ in response to the stem cells at any timepoint. Of note, at the 24 h timepoint the allogeneic stem cells were even able to suppress IFNγ production by the permissive donor NK cells below background (Additional file 7: Figure S7G).
To study the potential detrimental effects of IFNγ secretion on the interactions between the immune cells and the allogeneic stem cells that are likely to occur in vivo, we performed experiments where the stem cells were pretreated with IFNγ prior to coculture with the PBMCs in the presence or absence of vaccinia virus (Additional file 7: Figure S7H-J). Contrary to our expectations that IFNγ-pretreatment would suppress immune responses due to inhibited virus amplification and improved immunosuppressive potential, as previously reported in the literature [53], the immune responses against the infected stem cells were further increased, specifically in the permissive donors suggesting that virus-associated infection and IFNγ-production might have the unfortunate effect of sensitizing immune cells to otherwise small and "under the radar" allogeneic differences resulting in compromised rather than improved immunosuppression of T and NK cells. Consistent with previous experiments (Figs. 4b, 5b), we also observed that the NK-specific stimulator cell line K562 induced indirect activation of NKT and T cells (Additional file 7: Figure S7J), providing further support to the notion of intensive NK-NKT-T cell crosstalk and interdependence. Such a crosstalk might also be critically important for the magnitude and kinetics of anti-viral or anti-allogeneic responses, thus playing a key role in the patient-specific restrictions to the Trojan horse.
Thus, while HLA or KIR matching might be informative and indeed play a significant role in determining the permissiveness versus resistance, the ultimate therapeutic efficacy of an allogeneic Trojan horse can be influenced by multiple other patient-specific differences including innate and adaptive immune cell composition as well as activation status or sensitivity to vaccinia virus/ allogeneic mismatch. Such complex patient-specific differences would be much more properly evaluated by the development and application of robust companion diagnostic assays.

Discussion
The goal of the work presented in the current manuscript was to evaluate the potential of using off-the-shelf allogeneic adipose-derived stem cells as a delivery vehicle to potentiate oncolytic virotherapy, leveraging the unique abilities of mesenchymal stem cells to amplify vaccinia virus and to transiently suppress anti-viral immunity. The immune system is known to play a dual role in oncolytic virotherapy of cancer. While the induction or potentiation of tumor-specific immunity is believed to play an important role in the long-term therapeutic efficacy of oncolytic viruses [69], the presence of innate and adaptive anti-viral immune barriers can significantly restrict the extent of direct vaccinia virus-mediated oncolysis. Importantly, recent data indicate that tumor burden might directly correlate with inadequate responsiveness to immunotherapy [70]. Limited oncolysis and the lack of immediate reduction in tumor burden can therefore compromise the immuno-stimulatory effects of oncolytic viruses and greatly reduce their therapeutic efficacy.
Improving direct oncolysis is contingent on designing effective strategies to overcome the existent immunological barriers to oncolytic viruses that include complement/antibody neutralization, the intrinsic tumor cell anti-viral interferon responses, as well as the elimination of infected cells by innate and adaptive immune cell populations such as NK cells and T cells. The NK cell responses appear to be of particular interest due to their innate characteristics and immediate nature, in sharp contrast to the time needed for expansion of virusspecific T cells [71,72]. Of note, the latter branches of adaptive immunity appear to be critical for the ultimate clearance of the virus from the patient and warrant safety as well as therapeutic efficacy associated with the induction of tumor-specific immunity.
The therapeutic efficacy of oncolytic viruses is often restricted by the limited permissiveness of the tumor to virus infection and amplification [73], which reflects patient specific differences, including intrinsic tumor cell resistance [74], functional interferon anti-viral responses [4], or the presence of non-permissive tumor-associated stroma [75][76][77]. In tumors with relatively modest permissiveness, vaccinia virus might have insufficient time to efficiently colonize the tumor before it gets eliminated Augmenting the oncolytic potential of vaccinia virus in such circumstances might therefore require a strategy that combines efficient delivery with a boosted local virus amplification that together could provide a favorable environment where virus spread and colonization occurs faster than the induction of protective anti-viral state. Ex-vivo expanded autologous or allogeneic mesenchymal stem cells represent a unique delivery platform that offers the combined power of boosted local virus amplification and the ability to transiently suppress anti-viral innate and adaptive immunity, which is key to the success of oncolytic virotherapy in vivo. Given that the application of personalized autologous MSCs is an approach that is rather expensive and impractical for clinical development, the feasibility and limitations of using established and easily available allogeneic stem cell lines needs to be evaluated and investigated further. In the current study, we demonstrate that the impressive ability of mesenchymal stem cells to amplify vaccinia virus as well as to recruit and sensitize tumor cells to virus infection might significantly improve the therapeutic potential and broaden efficacy against resistant and low-permissive tumors. In addition, MSCs demonstrate the ability to effectively suppress anti-viral NK cell responses that represent the earliest and most significant innate immune barrier to tumor colonization and oncolysis [71,72]. Of note, while MSCs appear to be immunosuppressive against NK cells in both autologous and allogeneic settings, as to date it has been shown in the literature, we demonstrate that their immunosuppressive ability and virus amplification potential remain subject to significant allogeneic barriers associated with the production of IFNγ and direct cytotoxicity by both NK and T cells. Such rapid and exaggerated allogeneic responses or the outright rejection of the stem cells, therefore, represent a significant obstacle to the use of off-the-shelf MSC-or alternative cell-based delivery platforms in the clinic.
The hypoimmunogenic and immunosuppressive characteristics of mesenchymal stem cells have justified their extensive use in the treatment of various autoimmune and inflammatory conditions, demonstrating similar persistence and equivalent therapeutic efficacy in both autologous and allogeneic settings. Conflicting reports demonstrate the superiority of using autologous stem cells, which probably reflects different disease background or requirements for brief immunosuppression versus long-term persistence and tissue repair [78]. In fact, numerous studies have already challenged the paradigm that allogeneic stem cells can be used in a one-size-fits-all universal donor setting, due to their hypoimmunogenic features. Instead, it becomes increasingly clear that MSC are not immune privileged and can induce allogeneic responses resulting in their ultimate rejection [79,80].
Here, we argue that, to successfully utilize the potential of MSCs for the purposes of oncolytic virotherapy, it is necessary to build a much deeper understanding of the underlying mechanisms that control the balance between their immune evasive and immunogenic characteristics, and specifically how these are affected by the inflammatory environment associated with vaccinia virus infection. Importantly, long-term persistence of the allogeneic stem cells is irrelevant for the purposes of oncolytic virotherapy, while their short-term immunosuppressive and virus amplification potentials are essential and indispensable. Our demonstration that the combination of IFNγ production and direct cytotoxicity are associated with inability of the allogeneic Trojan horse to deliver and amplify vaccinia virus is critically important for the advancement of this therapeutic modality, as IFNγ plays a key role in both the control of vaccinia virus infection and the regulation of the immunosuppressive properties of MSCs. Multiple studies have demonstrated that exposure to IFNγ can significantly improve the immunosuppressive functions of MSCs by upregulation of IDO, iNOS/NO, COX2/PGE, and PD-L1 [50]. This would give stem cells the unique ability to support virus colonization by subverting IFNγ-driven immune activation and anti-viral immunity, but previous studies as well as our own findings indicate that IFNγ can also exacerbate small allogeneic differences and increase the immunogenicity of stem cells, potentially through up-regulated expression of MHC Class I/II or other costimulatory molecules. It should be emphasized that, while IFNγ-pre-treatment might be beneficial to enhance MSC-mediated immunosuppression for the treatment of autoimmune and inflammatory conditions, where long-term engraftment is unnecessary or unaffected by minor allogeneic differences, this is certainly not the case with oncolytic virus approaches that are critically dependent on the ability of the Trojan horse to amplify the virus. Accordingly, inappropriate and rapid IFNγ responses against the allogeneic stem cells might be sufficient to inactivate the Trojan horse, even in the absence of outright rejection and compromised persistence in vivo.
Our findings reveal that the responses to vaccinia virus and allogeneic stem cells alone or in combination are highly patient-specific, thus demonstrating the need for further mechanistic studies aimed to validate the relative contribution of IFNγ and direct cytotoxicity for the inactivation of the allogeneic Trojan horse. Understanding the basis for this patient-specific permissiveness versus resistance to the Trojan horse is challenging as it can reflect multiple sources of variability, including MHC I/II mismatches, differences in the MHC-binding Killer Cell Immunoglobulin-like receptors (KIR Haplotype or ligands) or differential frequency/activation status of innate immune cell populations such as NK or NKT cells. The use of allogeneic stem cells for oncolytic virus delivery is greatly facilitated by the lack of requirement for long-term survival and engraftment of the cells, making this approach likely to work across insignificant MHC mismatch barriers and restricting patient-specific resistance to relatively small groups of patients, who can be excluded from clinical trials with the use of simple diagnostic assays, as the ones presented in this study. Alternatively, a more frequent or broader patient-specific resistance to allogeneic Trojan horses would require the establishment of allogeneic MSC banks and matching the patients to the available stem cell lines, based on MHC typing data or in vitro diagnostic assays, which take into account all of the complex immunological characteristics of patients and particularly those conferring rapid and exaggerated anti-viral/-stem cell responses able to significantly limit therapeutic efficacy in highly resistant individuals.
Given that amplification of the virus in the stem cells proceeds within hours to several days after infection, while it takes more than a week for the adaptive T cell responses to contain virus spread and oncolysis, it becomes evident that the kinetics of the anti-viral/ stem cell responses within the first few days of treatment would be critical for the ability of the Trojan horse to have a significant therapeutic effect in vivo. Consequently, the crosstalk and interplay between different innate and adaptive branches of the immune system need to be further investigated as these might be directly associated with the speed and magnitude of the anti-viral/ stem cell immunity. The combination of exaggerated innate immune mechanisms sensitizing adaptive immunity against the virus or the allogeneic stem cells can be particularly detrimental. We were very interested by the finding that a patient can be broadly resistant to several allogeneic stem cell lines and that this broad resistance was also associated with unusually high frequency of NKp46 + CD 3+ NKT-like cells, which were also the only population that expanded in numbers in response to both the virus and the allogeneic stem cells. Unfortunately, our attempts to link this NKT-like population of cells with effector functions able to directly interfere with the Trojan horse were unsuccessful, as these cells didn't manifest any significant contribution to IFNγ production or cytotoxic activity. NKT cells are known to play a role in the control of viral infections, but their involvement in vaccinia virus responses and immunity hasn't been investigated fully yet. The fact that this population of cells is unlikely to be responsible for the direct rejection or inactivation of the Trojan horse does not eliminate the possibility that NKT or NKT-like cells are critically important for directing and accelerating coordinated NK and T cell responses against the stem cells or the virus. Detailed kinetic and mechanistic studies would be necessary to evaluate the possibility that innate immune cells like NKT cells provide early cytokine help sensitizing NK and T cell to the presence of the virus and/or potential allogeneic differences, as suggested by some of our preliminary data (Additional file 7: Figure S7D). The importance of the crosstalk between the innate (NK/NKT cells) and adaptive (B/T cells) arms of the immune system, as revealed by our data, suggests the existence of correlation between patient-specific NK and T cell responses. Despite the rather limited cohort of PBMC donors tested, it becomes evident that patients mount coordinated NK, NKT and T cell responses against the virus and allogeneic stem cells (Fig. 5d). On the other hand, the comparative responsiveness to the virus and stem cells appears discordant and highly patient-specific (Additional file 5: Figure S5C), which cannot be explained by HLA typing data alone and suggests the involvement of innate immunological differences that might not be the same with respect to the virus or possible MHC mismatches. These innate immunological differences can be associated with differences in the NK or NKT cells and more precisely with the balance of signaling through their activating versus inhibitory receptors such as NKG2A/NKG2C/ NKG2D or KIR receptors, which like certain HLA alleles are highly patient-specific and have already been linked to resistance/susceptibility to infectious agents, autoimmune diseases, and cancer [81][82][83][84][85][86].
Overall, our data indicate that while autologous mesenchymal stem cells are potentially the optimal vehicles for the delivery and amplification of oncolytic viruses, the use of properly matched allogeneic stem cell lines in combination with robust companion diagnostic assays could provide a more practical and commercially viable alternative that guarantees consistent stem cell quality and validated amplification potential. This approach also provides the unique opportunity to utilize readily available off-the-shelf cell-based delivery platforms in a highly efficient personalized fashion.

Conclusions
We demonstrate that mesenchymal stem cells, in addition to protecting naked viruses, can amplify vaccinia virus, enhance its delivery, and suppress anti-viral innate and adaptive immunity to potentiate oncolytic virotherapy. Our data also reveal that the ability of stems cells to amplify and deliver vaccinia virus is subject to allogeneic barriers that require proper patient-to-stem cell matching. We have also demonstrated that the resistance of patients to allogeneic carrier cells is associated with the induction of anti-stem cell IFNγ and cytotoxic responses, likely reflecting patient-specific HLA or KIR Haplotype mismatches. This study provides fundamental understanding of the molecular principles behind patient-specific resistance and will guide the future clinical development of cell-based delivery platforms for oncolytic virus therapy of cancer.

Additional files
Additional file 1: Figure S1. were infected with L14 VV at MOI of 0.1 as in (E) but instead of supernatants K562 cells were cocultured with 5000 or 20,000 RM20-eGFP ADSCs in triplicates. Fluorescence imaging and flow cytometry analysis were used to show that the green fluorescent stem cells attract the unlabeled/grey K562 cells and dramatically increase the percentage of infected eGFPnegative TurboFP635 + K562 cells. Despite the potentiated infectivity of the highly resistant K562 cells, the stem cells ultimately fail to eradicate or significantly impact their overall survival, consistent with the minimal ability of these cells to amplify vaccinia virus, as shown in the NCI-60 human cell line screen previously. Statistically significant differences (Student T-test, p < 0.05) based on duplicates or triplicates versus control or as indicated are marked with asterisks.
Additional file 3: Figure S3. ADSC are suppressive against NK cells and can overcome allogeneic immune barriers. (A) ADSC-mediated immunosuppression does not affect the frequency of NK and T cells. Note that the PMA-Ionomycin treatment causes downregulation of the NKp46 marker used to identify and gate on NK cells, resulting in "disappearance" of the most activated NK cells.
Additional file 4: Figure S4. The potential of allogeneic stem cells to overcome immune barriers correlates with their ability to suppress virusinduced T, NK and NKT cell responses. (A) Gating strategy used to evaluate the effect of ADSC and vaccinia virus on activation of T, NK, and NKT-like cells as measured by upregulation of surface CD69 expression. (B) Summary of the modulation of the three immune cell populations as percentage of gated live lymphocytes in patients RM047 and RM048 (See Fig. 4b).
Additional file 5: Figure S5. The potential of allogeneic stem cells to function as a Trojan horse is restricted by patient-specific differences suggesting that proper matching would be required.