Advances in mesenchymal stem cell-mediated gene therapy for cancer

Mesenchymal stem cells have a natural tropism for tumours and their metastases, and are also considered immunoprivileged. This remarkable combination of properties has formed the basis for many studies investigating their potential as tumour-specific delivery vehicles for suicide genes, oncolytic viruses and secreted therapeutic proteins. The aim of the present review is to discuss the range of approaches that have been used to exploit the tumour-homing capacity of mesenchymal stem cells for gene delivery, and to highlight advances required to realize the full potential of this promising approach.


Introduction
Despite signifi cant advances in the fi eld of gene therapy for cancer, two major obstacles remain that continue to limit the clinical potential of this approach: lack of tumour tropism of vectors, and stimulation of an immune response. Th ese barriers preclude systemic administration of current vectors to effi ciently target metastatic disease. Th e combination of cellular therapy and gene delivery is an attractive option as it will potentially protect the vector from immune surveillance, and will support targeted delivery of a gene or therapeutic protein to the tumour site.

Mesenchymal stem cells
Mesenchymal stem cells (MSCs) are nonhaematopoietic stem cells that have generated a signifi cant amount of interest as a result of their apparent ability to home to the tumour site following systemic delivery. MSCs have an inherent ability both to self-renew and to diff erentiate into multiple lineages including osteoblasts, chondrocytes and adipo cytes [1]. Th e cells are readily isolated from the stromal compartment of bone marrow, along with a number of other sources including adipose tissue, trabecular bone and skeletal muscle [2]. Although a single marker for MSCs has not been isolated, a panel of specifi c antigens has been identifi ed, including expression of CD105, CD73 and CD90 in >95% of the culture, and an absence of CD14, CD34, CD19, HLA-DR and CD45 [3]. When introduced systemically to healthy animals, MSCs have been shown to home preferentially to the lung, liver and bone, and were found to a lesser extent in other tissues. Upon injury, however, the migratory pathway changes to preferentially target sites of injury [4].
Although MSCs have potential uses in regenerative medicine and a number of diff erent disease models, the present review will specifi cally focus on their potential for targeted gene delivery in the context of cancer. Th is is an exciting area of research that has gained considerable momentum in recent years, with studies reporting engineered MSCs specifi cally targeting multiple tumour types followed by local secretion of therapeutic proteins (IFNβ [5-7], IL-2 [8,9], IL-12 [10-12], pigment epitheliumderived factor [13], NK4 [14], TNF-related apoptosis inducing ligand (TRAIL) [15][16][17][18]), expression of prodrug activating suicide genes (herpes simplex virus-thymidine kinase [19][20][21], cytosine deaminase [22]), and delivery of replicating oncolytic viruses [16,19,[23][24][25]. A major advan tage of MSCs in this setting is that they are considered immunoprivileged, possibly due to low expression of Ag(HLA) MHC class 1, and no expression of CD40, CD80 and CD86 [4]. Th e cells are also known to secrete prostaglandin, transforming growth factor beta and hepatocyte growth factor, which regulate the T-cell immune response, thereby decreasing the probability of a cytotoxic T-cell response to transduced cells [17].
Resident MSCs suppress both transient and continuous immune surveillance, which aims at facilitating the healing process [26]. Th is immune privilege in the context of cancer, however, has the potential to support tumour progression. Djouad and colleagues reported growth of B16 melanoma cells in allogenic animals only in the presence of MSCs, suggesting that protection from the host immune response supported tumour establishment [27]. Further studies by the same group revealed that MSCs administered in low numbers with Renca adenocarcinoma cells actually induced tumour rejection [28]. MSCs were also shown to inhibit outgrowth of colon carcinoma in rats, with complete inhibition seen when the number of MSCs were at least equal to the number of tumour cells. Tumour establishment using the mixed cell population was found to induce more infi ltration of monocytes and granulocytes than the individual populations alone [29].
Th is observation may be explained by the fact that high numbers of MSCs have been shown to suppress alloreactive T cells, with very low numbers found to stimu late lymphocyte proliferation [30]. Indeed, additional evidence suggests that the context with which MSCs are introduced in vivo may infl uence their immune pheno type [26].

Tumour tropism
Tumour-specifi c migration of MSCs is not completely understood, but appears to be dependent upon the biological properties of the tumour microenvironment, as well as the native tropism of selected cells. Integration of MSCs into the tumour stroma is thought to be mediated by high local concentrations of infl ammatory chemokines and growth factors. Th e tumour microenvironment is considered a site of chronic infl ammation [31]. Th is environment may mediate MSC migration through secretion of soluble factors such as epidermal growth factor, vascular endothelial growth factor-A, fi broblast growth factor, platelet-derived growth factor, stromal-derived growth factor-1α (SDF-1α/CXCL12), IL-8, IL-6, granulocyte-macrophage colony-stimulating factor, granulocyte colony-stimulating factor, Ang1, monocyte chemoattractant protein-1 (CCL2), haematopoietic growth factor, transforming growth factor beta-1 and urokinase-type plasminogen activator [32][33][34][35][36][37].
Th e process of MSC mobilization to the tumour is thought to be regulated similarly to leukocyte migration through integrins and adhesion molecules [38]. Molecules involved in leukocyte traffi cking -such as tethering, rolling, adhesion and transmigration from the bloodstream to the tissue -are expressed on MSCs. Th ese include integrins, selectins and chemokine recep tors. Both P-selectin and vascular cell adhesion molecule-1 have been found to infl uence the adhesion of MSCs in endothelium [39].
Th e most documented chemokine receptor implicated in targeted homing of MSCs is CXCR4, which has potential in cell mobilization and homing [45]. A study by Wynn and colleagues reported that CXCR4 is highly expressed on MSCs, but mainly intracellularly (83 to 98%) rather than on the surface [46]. Another study reported no detectable CXCR4 expression on MSCs [42]. Variable expression of CXCR4 detected in diff erent studies has been suggested to be related to sensitivity of the trypsin digestion procedure used [44], diff erences in culture conditions, and heterogeneity of MSC populations. In vitro three-dimensional culture of MSCs as spheroids was shown to increase SDF-1α signalling, which restored functional expression of its receptor CXCR4 and homing potential that is crucial for therapeutic applications [47].
Although the tumour tropism of MSCs is generally accepted, it is certainly dependent on the tumour model. Variation in levels of MSC engraftment reported in diff erent studies may be explained by diff erences in MSC isolation, culture conditions and experimental protocols used. Within individual studies, however, variable levels of MSC engraftment have been reported in diff erent tumour types, most probably due to diff erences in the microenvironment created by the tumour in question [48]. Th e proportion of MSCs engrafted was not found to be related to tumour size [48].
A recent study further highlighted the role that the degree of infl ammation in a tumour microenvironment plays in the level of MSC recruitment [7]. In a study of MSC-IFNβ-mediated therapy of pancreatic cancer, treatment with an anti-infl ammatory agent resulted in reduction of MSC engraftment in the tumour, and reversed the tumour inhibitory eff ects observed [7].

Modifi cation of the tumour microenvironment
Th e apparent role of infl ammation in MSC tumour tropism has also been harnessed to increase engraftment through tumour irradiation, which is associated with release of several cytokines from exposed tissue [48,49]. Klopp and colleagues found that low-dose irradiation of the tumour microenvironment enhanced MSC tropism and engraftment at the tumour site [49]. Irradiation resulted in apoptosis and increased release of infl ammatory signals at the site of radiation, including TNFα, platelet-derived growth factor, as well as chemokines CCL2 and CCR8 [49]. Th e eff ect of tumour radiotherapy on localization of lentivirus-transduced MSCs in a variety of tumour types has also been reported [48]. Irradiation increased MSC localization in LoVo, HT-29 (colon) and MDA-231 (breast), but not UMSCC1 (head and neck) xenografts. Th is study also reported a modest elevation in CCL2 expression in irradiated tumours, although it was not found to correlate with MSC infi ltration [48]. Infl ammation plays a critical role in tumour progression [50], and therefore stimulation to support MSC homing to tumours would not be a viable option. Radiotherapy is frequently a component of cancer therapy, however, and therefore could work in combination with MSC-based gene delivery to support improved targeting of MSCs to tumours.

Modifi cation of the mesenchymal stem cell surface
While variations in MSC engraftment have been observed in diff erent tumour models, attempts are being made to improve tumour tropism and infi ltration through modifi cation of the MSC surface. Cell rolling is a critical step of the adhesion cascade supporting rapid deceleration of cells from the bloodstream, and is mediated by selectins expressed on the endothelium of the target organ. Immobilized sialyl Lewis X on MSCs was shown to induce cell rolling on the P-selectin surface under dynamic shear fl ow conditions in vitro, and may have potential applications in improving MSC engraftment in vivo [51]. In one study where native MSC tropism for the tumour of interest was not detected, MSCs were engineered to overexpress the epidermal growth factor receptor -which binds transforming growth factor alpha and epidermal growth factor. Transduced MSCs had enhanced migratory properties towards GL261 gliomas or B16 melanoma in vivo [52]. Following establishment of improved engraftment, the cells were further engineered to secrete IFNγ, resulting in increased animal survival [52].
Th is approach has evolved to include the use of conditionally replicating adenoviruses, which support delivery of an increased viral load specifi cally to the tumour site [23][24][25]. Clearly the timing is important here to avoid toxicity to MSCs prior to engraftment at the target site. Th e cycle of MSC adenovirus replication has been reported to have relatively slow kinetics, which may allow time for MSCs to reach the target site before replication causes cell death [56]. Th e delivery of oncolytic viruses does not rely on long-term survival and proliferation of cellular vehicles, as they are destroyed by viral replication. Capsid-modifi ed oncolytic adenoviruses have been coupled with the use of transcription-specifi c promoters to limit ectopic viral amplifi cation in nontarget cells [55]. MSCs have also been engineered to express the herpes simplex virus-thymidine kinase followed by adminis tration of the prodrug ganciclovir for targeted cancer suicide gene therapy [19][20][21]. Based on similar principles, retrovirus transduction of adiposederived MSCs to express cytosine deaminase, followed by systemic adminis tration of the prodrug 5-fl uorocysteine, mediated a strong anti-tumour eff ect in vivo [22].

Localized delivery of therapeutic proteins
Along with their tumour tropism, MSCs have been shown to integrate into and persist in the tumour stroma [5]. Th is integration has supported their use as delivery vehicles for various biological agents, whose systemic administration is precluded due to their short half-life and toxicity at the doses required for therapy. MSCs can effi ciently produce biological products at tumour sites and so have the potential to improve pharmacokinetics of secreted agents [5].
In a number of tumour models, MSCs expressing IFNβ have been shown to result in decreased tumour burden and increased animal survival [5][6][7]. Increased systemic levels of IFNβ or secretion at sites distant from the tumour were not eff ective, indicating that regional secretion was required [5][6][7]. MSCs engineered to secrete IL-12 and embedded in a matrix adjacent to tumours were also reported to have a signifi cant therapeutic eff ect [10]. Similar to fi ndings in the case of IFNβ, regional secretion was required, with no reduction in growth observed when the implant was placed in the opposite fl ank to the tumour [10].
MSCs expressing the hepatocyte growth factor antagonist NK4 in vivo were also found to prolong animal survival by inhibiting tumour-associated angiogenesis, lympho angiogenesis and induction of cancer cell apoptosis [14]. Local secretion of pigment epithelium-derived factor in a model of hepatocellular carcinoma through lentivirus transduction of MSCs similarly resulted in lower tumour volume, reduced lung metastases and improved survival through inhibition of tumour angiogenesis [13].
Further, MSCs secreting IL-2 [8,9] or IL-12 [10,11] were shown to elicit an immunological reaction, and to stimulate infl ammatory cell infi ltration of the tumour tissue. Th e observed anticancer eff ect was shown to be immune mediated and absent in immunodefi cient animals [10]. Delivery of MSC-IL-12 did not cause systemic toxicity, and resulted in increased serum and tumour levels of IL-12. In contrast, administration of Ad-IL-12 only increased serum IL-12 levels and induced systemic toxicity [11]. Th erefore it appears that MSCmediated local delivery of a therapeutic agent may be better tolerated by the host without inducing an unacceptable immune response [11].
TRAIL induces caspase-mediated apoptosis in tumour cells that overexpress the receptor. Like most healthy tissues, MSCs are resistant to TRAIL-induced apoptosis due to their very low levels of active receptors [17]. As a result of this, MSCs secreting TRAIL have been used in models of lung cancer, breast cancer, cervical cancer and brain cancer in vivo, resulting in signifi cant anti-tumour eff ects [15][16][17][18]53]. In one study using a lentiviral vector, TRAIL expression was placed under the control of a tet promoter, supporting conditional activation using doxycycline [16]. In an animal model of lung metastases of breast cancer, this controlled, local delivery of TRAIL completely cleared metastatic disease in a selection of animals [16]. Interestingly, when MSC-TRAIL cells were co-injected with tumour cells for subcutaneous tumour formation, only doxycycline-mediated activation on the day of tumour cell inoculation (day 0) caused a signifi cant decrease in tumour weight. Activation following tumour establishment (day 25) did not result in a change in tumour burden [16].

Potential role in tumourigenesis
Although beyond the scope of the current review, the potential role of MSCs in tumour initiation or promotion is a signifi cant concern that must be addressed fully to allow MSC-mediated therapy for cancer to realize its full potential. Th is role remains a topic of continued debate. Expansion of MSCs in vitro will be required for therapeutic application and so their stability in culture is paramount. Spontaneous transformation of human MSCs has been reported following long-term passage in vitro [57,58], while Bernardo and colleagues found no evidence of human MSC transformation [59]. Indeed the majority of studies have shown that human MSCs are stable, while murine MSCs are more prone to genetic transformation during in vitro culture, and may be capable of forming sarcomas in vivo [59][60][61][62][63]. Although transformation of human MSCs appears unlikely, and very rare, these studies certainly emphasize the importance of stringent monitoring of MSCs, including karyotyping, before application in the clinical setting.
MSCs have also been implicated as tumour supportive when co-injected in the presence of a variety of tumour cell types, including breast [64][65][66][67], ovarian [68], melanoma [27], glioma [69,70] and colon [71,72] tumour cells. Th e majority of these studies, however, used an equal or even excess number of MSCs over tumour cells. Th e data generated provide important information on interactions between MSCs and tumour cells, although the models are unlikely to refl ect the in vivo situation. MSCs were shown to integrate into the tumour stroma and were demonstrated to exert their eff ects at least partly through secretion of paracrine factors including CCL5, IL-6 and SDF-1α [64,65,68]. Th ere is also evidence that MSCs may serve as precursors for carcinoma-associated fi broblasts and/or pericytes, playing a potentially important role in tumour angiogenesis through diff erentiation and the release of proangiogenic factors [67][68][69][71][72][73][74][75][76]. Additionally, as previously mentioned, the immunosuppressive qualities of MSCs may support tumour development and progression through protection of cancer cells from immune surveillance [27].
Conversely, co-injection of MSCs has also been shown to result in tumour suppression in a model of colon cancer [29], hepatoma [77] and melanoma [78].
In terms of MSC-mediated gene delivery, under standing the role of MSCs following engraftment at the site of a pre-established tumour is required. Th e majority of studies outlined here, using MSCs engineered to deliver therapeutic agents, have resulted in signifi cant antitumour eff ects in vivo. Unmodifi ed MSCs were also shown to result in tumour suppression in some cases [7,8,79], with the majority showing no eff ect on tumour progression following engraftment at the site of an established tumour [13,18,23,53,55,69,75]. Repeat intravenous administration of MSCs over 3 weeks, however, was shown to stimulate increased tumour growth in a model of pancreatic cancer [21]. Similar to the level of MSC engraftment in tumours, it seems that the eff ect of MSCs following engraftment will be tumour specifi cprobably dependent on a range of factors including the method of MSC isolation and culture, the experimental model, the number of cells engrafted in the tumour, and the milieu of growth factors and infl ammatory cytokines present within the tumour microenvironment.

Conclusion
Th e studies outlined highlight very promising potential for MSC-mediated delivery of therapeutic agents directly to tumour tissue, with remarkable progress made in the past decade. Clearly MSCs have a number of advantages as cellular vehicles -they are relatively easy to isolate and expand, specifi cally target tumours and their metastases following systemic delivery, can be transduced effi ciently with a range of vectors, have immuno suppressive properties, have the ability to express therapeutic proteins in secretory form and can support amplifi cation of oncolytic viruses. Th e potential for MSC-mediated tumour promotion, however, must be addressed. Further understanding the biology of MSCs, and the specifi c combination of factors controlling their tumour-specifi c migration and persistence, will support translation to the clinical setting.

Competing interests
The authors declare that they have no competing interests.