Introduction

Oncolytic viruses are viruses that are capable of selective or preferential replication in cells with a malignant phenotype [13]. These may be either pathogenic or vaccine strain viruses, usually with attenuating genetic alterations that limit their ability to replicate in normal tissues (including strains of herpes simplex virus [4], poliovirus [5], and measles [6]), or viruses that cannot normally replicate in human cells, but that are capable of replicating in cancer cells due to the reduced ability of the tumor to induce an anti-viral response (including animal pathogens such as vesicular stomatitis virus [7] or myxoma virus [8] and orphan viruses such as reovirus [9] or Seneca Valley Virus [10]). Because these viruses are capable of replication only in the tumor cells, the therapeutic itself becomes amplified where it is needed most, and viral replication ultimately leads to direct destruction of infected cells (viral-mediated lysis) with subsequent spread of the therapy within the tumor. Furthermore, as the mechanisms of targeting and tumor cell killing are typically distinct from traditional chemotherapies and radiotherapies, these therapeutics often synergize with more traditional therapies [1116] while any toxicities are typically not additive. Indeed, the therapeutic benefits of some oncolytic viruses are only evident when used in combination with other therapies.

The first clinical use of a logically designed and genetically engineered viral vector used as an oncolytic agent incorporated an adenovirus serotype 5 backbone (dl1520, ONYX-015) [17, 18]. Although this virus has since gone on to become the first approved oncolytic viral therapy (H101 or Oncorine for head and neck cancers in the Chinese market [19]), it also demonstrated many of the limitations of this approach, including a limited ability to deliver the virus systemically especially in pre-immunized hosts, while the slow spread of this virus often led to clearance of the therapy prior to tumor eradication. This resulted in an initial focus in the field on “overcoming” the immune response against the virus, so as to delay viral clearance and increase the therapeutic potential of these agents through enhancing their oncolytic potential. There was also a general move away from adenovirus-based vectors for oncolytic use, with the ideal virus being one that was well studied (with known pathogenesis and ideally with anti-virals available), having a rapid spread, an innate potential to traffic within the blood stream and the capability to express foreign, therapeutic transgenes. As such, vaccinia virus has become one of the most promising backbones; in addition to its widespread historical use in humans [20] and the availability of approved anti-virals [21], it displays a rapid spread within susceptible host tissues leading to a high degree of tissue destruction [22] and has a large cloning capacity [23] while it has evolved to spread through the blood as a part of its natural life cycle [24].

It has also become apparent (especially when viruses capable of replicating in mouse cells were studied, so facilitating the pre-clinical study of immunocompetent animal tumor models) that the immune response induced by the virus can provide a beneficial anti-tumor effect [2, 25]. In particular, viruses expressing a variety of selected cytokines have been found to frequently display enhanced anti-tumor effects in pre-clinical and clinical studies [2630], while animal models in which an initial implanted tumor has been completely cleared are often resistant to re-challenge with the same tumor cell line [26, 31], indicating that a long-term memory immune response targeting tumor-associated antigens must have been induced by the viral infection within the tumor. This is perhaps unsurprising, as viral replication will typically lead to direct destruction of tumor cells, releasing tumor antigens and other danger signals into the extracellular environment, while the ultimate clearance of these viruses from the tumors indicates that the viral infection is capable of at least transiently overcoming localized immunosuppression within the tumor. Furthermore, the recent clinical success with viruses such as ONCOVEX and JX-594 (both of which have recently entered Phase III clinical testing) [32] is at least partly due to the fact that these viruses express the cytokine GM-CSF as a transgene, so enhancing their immunotherapeutic potential.

A more focused attempt to harness the immune response as an additional and therapeutic mechanism of action for oncolytic viruses has therefore begun, while additional focus has been applied on strategies for combining these viruses with other immunotherapies. However, the immune response remains a double-edged sword, and some approaches that may enhance the immunotherapeutic potential of these therapies will likely also reduce their direct oncolytic effects as a result of premature clearance of the viral therapy. Striking the correct balance between these two mechanisms of action will therefore be necessary for this therapeutic platform to reach its full potential.

We have focused on developing the immunotherapeutic potential of oncolytic vaccinia virus strains and have described several techniques that appear to directly enhance this potential in pre-clinical models, while having minimal or no detrimental effect on the oncolytic potential of these therapies. These approaches may be applied to other oncolytic vectors and will be covered in this review.

Enhancing viral immunotherapy

Transgene expression

A variety of transgenes have been proposed as candidates for the expression from oncolytic viruses, including vaccinia virus. Generally these are used to enhance the tumor-killing potential of the agent by providing a bystander effect (leading to the additional destruction of non-infected cells within the tumor). For example, these might be directly cytolytic or act as prodrug-converting enzymes (capable of converting a systemically delivered, non-toxic prodrug to a toxic product within the tumor) [3335]. Other commonly used transgenes have functions that include anti-angiogenesis [36] and proteolytic activity (usually designed to target and degrade the extracellular matrix and so enhance viral spread within the tumor) [37]. Alternatively, reporter genes might be used for imaging purposes, so as to follow patterns and kinetics of viral gene expression profiles either within a patient or in a pre-clinical mouse model [3841].

However, the most successful transgenes have been those with products with immunomodulatory function, especially cytokines. Ever since Dranoff reported on the ability of GM-CSF expression to prevent the formation of B16 tumors in mouse models [42], this cytokine has been incorporated into different oncolytic vectors to boost their immunotherapuetic potential. We have demonstrated that GM-CSF expression from vaccinia can indeed boost its therapeutic potential in pre-clinical models [27], and the vaccinia strain JX-594 that expresses GM-CSF has shown exciting potential in Phase I trials targeting hepatocellular cancers [30] and is currently entering Phase III clinical testing. However, more recent data demonstrating the role of GM-CSF in inducing proliferation of monocyte-derived suppressor cells (MDSCs) [43] may indicate that the exact role of this cytokine needs to be re-examined in the context of oncolytic viral expression. It is possible that GM-CSF in combination with the highly immunogenic viral infection within the tumor has an immune-activating function, while the same cytokine used without viral infection may have more suppressive properties. Further investigation is therefore needed to determine the exact role of this cytokine.

Another commonly used cytokine family is that of the type I interferons. The therapeutic potential of these cytokines is well known [44, 45]; however, the toxicities involved with delivering recombinant IFN systemically to cancer patients have limited its use. One advantage of expressing this cytokine from an oncolytic virus is that transgene expression will occur exclusively within the tumor, so limiting systemic toxicity and increasing the concentration of the therapeutic within the tumor target itself [26]. Indeed, we have also shown advantages of expressing type I interferons from vaccinia in pre-clinical models, including enhanced anti-tumor effects and a reduction in viral infection of non-tumor tissues [26].

It is likely that as our understanding improves of how cytokine networks can be critical for overcoming immune suppression within even large solid tumors [46], future oncolytic vectors will express combinations of cytokines to further enhance specific immune pathways, notably targeting a Th1 cytolytic effector cell-mediated immune response, while reducing the induction of a potentially less beneficial Th2, antibody-based immune response. One advantage of the use of vaccinia as an oncolytic agent in this area is that its large cloning capacity (greater than 25 KB) [23] means that multiple immunomodulatory transgenes might be expressed from the same vector (in addition to other transgenes, such as reporter or suicide genes).

Although selected cytokine expression from oncolytic vaccinia and other viral therapies will frequently enhance their overall therapeutic potential, a careful examination of viral gene expression and replication demonstrates that this is typically at a cost to the replication and oncolytic potential of the viruses [47]. We have therefore looked to develop approaches that would allow the full potential of both oncolytic and immunotherapeutic modes of action to be achieved. In particular, we have incorporated technologies to exogenously control protein function at the level of protein stability into the cytokine transgenes expressed from vaccinia strains [47]. This approach utilizes a degradation domain that can be fused with any protein of interest and that misfolds upon translation, targeting the protein for rapid proteasomal degradation and so ablating protein function [48]. This process is reversible through the addition of a small molecule regulator that binds and shields the degradation domain from the proteasomal pathway. In this way, we demonstrated that it was possible to down-regulate the function of different cytokine transgenes for the initial period after viral delivery, so allowing a more complete viral infection and spread within the tumor over the first few days of action of the therapeutic. After this time, a controlled up-regulation of cytokine function allows the induction of a more localized immune response of greater magnitude exclusively from within the tumor. Together, this resulted in a significantly greater overall therapeutic benefit in pre-clinical models. We predict that many next-generation vectors might take advantage of this or similar mechanisms to regulate viral mechanisms of action, so creating viruses capable of switching from a primarily oncolytic to a primarily immunotherapeutic mechanism of action in a controlled fashion.

Deletion of viral virulence genes

Vaccinia is a large virus with multiple virulence genes [49]. The majority of these act (1) inside the cell, primarily to block apoptotic pathways; (2) at the cell surface to activate or suppress cell signaling pathways; or (3) extracellularly, to block the immune response to the virus, primarily through the secretion of decoy receptors that sequester cytokines and chemokines from the extracellular environment. Because many of these virulence factors also target the hallmarks of cancer [50], their functions are frequently redundant in cancer cells, and deletion of many of these virulence genes will attenuate the virus exclusively in non-tumor cells and so creates an oncolytic agent with enhanced specificity and greater therapeutic index.

The most frequently used mutation engineered into oncolytic vaccinia strains is in the viral thymidine kinase gene [51], as loss of this gene function creates a virus dependent on cellular thymidine kinase, a protein commonly overexpressed in cancer cells. Replication of this virus can be further restricted to cancer cells through combination with deletion of the viral growth factor gene [27, 52]. The viral growth factor (VGF) binds to the EGF receptor, activating this signaling cascade and inducing cellular proliferation (and also cellular thymidine kinase production) [53], so its deletion further restricts viral replication to cells with mutations in the EGF-R signaling pathways (as found in the vast majority of cancers). Although the replication of this virus is highly restricted to malignant cells and it has a potent oncolytic effect, the continued expression of viral-encoded immunosuppressive virulence genes may reduce its immunotherapeutic potential.

As such, it has become evident that the mechanism of tumor targeting employed not only will affect the phenotype of the cancer that is most effectively targeted but also will influence the immunotherapeutic potential of the virus. For example, we have also reported on the construction of a virus with a deletion in the viral B18R gene [26]. This gene product is secreted and binds type I interferons [54, 55]. It was supposed that a B18R deletion would be most effectively used in combination with IFNb expression (as otherwise the B18R protein would reduce the effectiveness of the transgene). However, it was also noted that the B18R deletion in itself was tumor targeting and created a more cancer-specific oncolytic vector. More recently, we have demonstrated that viruses carrying the B18R deletion have significantly greater immunotherapeutic potential than the traditional thymidine kinase/VGF double-deleted vaccinia backbone, and so we believe the viral backbone must also be a critical consideration when designing next-generation oncolytic vectors specifically to enhance the immunotherapeutic potential of these therapies.

The large array of immunosuppressive virulence genes encoded by vaccinia means that it might be possible to selectively delete viral virulence genes that together target cytokines critical for the induction of specific arms of the immune response and so more specifically induce particular types of immune response within the tumor. However, as before, the effects of extensive viral gene deletions on the oncolytic potential of the therapeutic will need to be continuously assessed.

Combining oncolytic vaccinia with immune cell therapies

Cell-based carrier vehicles

One issue that has traditionally hindered the use of oncolytic viruses has been the limited ability to deliver these viruses systemically, especially in the face of a pre-existing anti-viral immune response. Because a viral particle is an inert object outside of a target cell, the virus has no inherent means to actively target the tumor. In some viruses, attempts have been made to partially address this through the incorporation of ligands on the virus surface that target specific receptors, typically in conjunction with removal of the natural receptor-targeting ligands on the viral surface. In this way, it is possible to limit the population of cells that are susceptible to viral infection to tumor cells [56, 57]. However, this approach has met limited success as (1) the receptor needs to be expressed exclusively on tumor cells, to not be down-regulated as a result of negative selection pressure, and be exposed within the vasculature for systemic delivery while (2) the reticuloendothelial system will continue to remove virus from circulation, meaning that the short (typically less than 5 min) half life of the viruses in circulation is not significantly altered [58]. Vaccinia has no cognate receptor, having the advantage of being able to infect any cell type. However, this does mean that only a small percentage of any systemically delivered inoculum will actually infect tumor cells (with the majority of the virus delivered into the blood stream forming non-productive infections in normal cell types). This percentage is dramatically further reduced when circulating neutralizing antibody is present, further limiting the systemic delivery potential of the virus.

Although some local (intratumoral injection to accessible tumors) or local regional (such as intraperitoneal delivery to ovarian cancers or intrahepatic arterial delivery to hepatocellular cancers) delivery methods can help delivery in some situations, this will still not allow effective delivery to disseminated micrometastases. An improved ability to actively traffic viruses to their tumor targets is therefore needed. We have developed an approach whereby a tumor-targeting immune cell population can be used as a delivery vehicle to actively carry the viral particles to their tumor targets [5961]. In this approach, we have utilized cytokine-induced killer (CIK) cells as carrier vehicles [62, 63]. These are a therapeutic immune cell population (currently undergoing clinical evaluation in the US) [64] that possess NK-T cell phenotypic markers and can be expanded from peripheral blood (or mouse splenocytes) through a simple cytokine enrichment protocol. They can effectively traffic to tumor targets [65], including micrometastases, and possess cytolytic activity through the recognition of NKG2D ligands on the tumor cell surface [66]. However, they do not typically possess sufficient cytolytic capacity to destroy large solid tumors. We therefore proposed that if these cells could be pre-infected with virus, they might act as a delivery vehicle, actively targeting the viral therapy to the tumor, potentially even in the face of an anti-viral immune response. This was indeed found to be the case, with a further synergy identified between the action of the immune cells and the viral therapy [59]. This synergy was partly mediated by the viral-induced increase in NKG2D ligand expression on the tumor cell surface, leading to greater CIK-cell activity, as well as the ability of the CIK cells to deliver virus throughout the tumor, rather than merely at or around the vasculature. Finally, we also demonstrated that viral delivery could be achieved within pre-infected CIK cells even in the face of pre-existing anti-viral immunity [67].

Furthermore, we found that as long as we could deliver the virus to the tumor in a pre-immunized animal model, we could produce significant therapeutic benefit. This anti-tumor effect was in some cases even greater than that for naïve animals [67] and was despite the fact that imaging studies demonstrated that virus replication and gene expression in the tumors of pre-immunized mice was less than 5% of the level of that in naïve animals. We therefore hypothesized that when virus was delivered successfully to the tumor in pre-immunized mice, the presence of anti-tumor effects (despite the loss of viral replication) meant that the mechanism of action of the viral therapy was switched from an oncolytic to an immunotherapeutic mode of tumor destruction. In support of this, we saw a large infiltration of CD4+ and CD8+ cells into the tumor and a reduction in the tumor of cell types typically associated with an immunosuppressive phenotype, including regulatory T cells, MDSC, and Th2-weighted macrophages. Further understanding of this reduction of immune suppression and implementation of approaches to successfully harness it might significantly benefit both viral therapies and other immunotherapeutic approaches.

In a further study, we found that CIK-mediated delivery of oncolytic vaccinia virus could clear residual lymphoma disease, whereas either therapy alone could not [31]. Because the residual lymphoma cells reliably relapsed without intervention, despite there being less than 1,000 cells present in the mouse model, this indicated that they represented a population of cells with cancer-initiating properties. This tumor-clearing capability of the combined therapy again appeared to be mediated by the immune response as the same CIK–vaccinia dual biotherapy could not clear the residual disease in immunodeficient mice [31]. Treatment with the dual biotherapy not only prevented relapse in these mice but could both protect mice against re-challenge with the same lymphoma cell line and induced a long-term memory T-cell response that recognized antigens on the lymphoma cells. In addition, we observed that pre-infection of CIK cells with oncolytic vaccinia resulted in increased production of Th1 cytokines (including IL-2 and IFN-gamma) and chemokines (IP-10 and ITAC). It therefore appears that a key factor in the therapeutic benefits of combined CIK/vaccinia dual biotherapy is mediated through the induction of a beneficial anti-tumor immune response.

Immune cell and vaccine combinations

The observation that the initial success of the use of CIK cells in combination with oncolytic vaccinia went beyond their use as carrier vehicles demonstrated the potential for synergy between oncolytic vaccinia and immune cell therapies. We have since looked to further develop this interaction through both enhancing the cross talk between virus and an immune cell therapy and incorporating different immunotherapeutic approaches in combination with oncolytic vaccinia virus, including vaccine therapies.

In initial studies, we looked to express specific chemokines from the oncolytic virus that are known to have cognate receptors expressed on the surface of CIK cells, including CCL5 (RANTES) [68]. The expression of CCL5 from oncolytic vaccinia was itself sufficient to enhance the therapeutic effects of the virus, presumably through enhanced immune infiltration into the tumor (as such, we demonstrated enhanced CD4+ T-cell and DC infiltration into vvCCL5-infected tumors relative to virus without CCL5 expression). However, of particular interest, the delivery of CIK cells systemically to mice bearing bilateral tumors, one of which was infected with the basic oncolytic vaccinia strains, the other with the same virus expressing CCL5, resulted in preferential CIK-cell homing to the tumor with the CCL5 expression and an enhanced therapeutic benefit when the two therapies were used in combination [68]. This demonstrated the potential to further modify oncolytic vaccinia viruses specifically so they might more favorably interact with adoptively transferred immune cell therapies.

We further looked at the potential for these viruses to work in combination with vaccine therapies. In particular, DC1 vaccines [69, 70] (developed at the University of Pittsburgh, see Pawel Kalinski’s article in this issue) are known to produce cytolytic T lymphocytes that target tumor-associated antigens, as well as expressing the receptor for CCL5 and CCR5 on their surface. The induction of high levels of circulating tumor antigen-specific T cells after DC1 vaccination has been demonstrated in the clinic [71, 72]; however, therapeutic responses were not as robust as predicted. One reason for this is believed to be due to the low level of tumor infiltration seen with the CTLs produced, an issue common to therapeutic cancer vaccines. It was therefore hypothesized that pre-treatment of the tumor with vaccinia strains expressing CCL5 might enhance the subsequent trafficking of CTLs raised by DC1 vaccination into the tumor and so enhance the therapeutic response. This was indeed found to be the case in pre-clinical mouse models [68] and raises the possibility of further enhancing the interaction between oncolytic vaccinia strains and vaccine therapies.

Discussion

The potential for oncolytic viruses to act as dual mechanistic cancer killers, combining direct oncolytic effects with a potent immunotherapeutic potential, has only recently been appreciated. A panel of studies by ourselves and others have begun to elucidate pathways and approaches that might most beneficially enhance the immunotherapeutic potential of these viruses without severely and adversely effecting their ability to directly destroy tumor cells through an oncolytic mechanism. There is still much to do, but the beginning of an understanding of how the host–pathogen relationship might be used to overcome localized immunosuppression within a tumor and be used to raise the most beneficial adaptive immune response targeting the tumor itself has already led to greatly improved next-generation oncolytic viral vectors in pre-clinical studies and in the design of viruses specifically intended to enhance the therapeutic effects of other immunotherapies. It is believed that the future clinical translation of these approaches might allow us to develop the full potential of this promising therapeutic platform.