Despite great efforts, the treatment options for patients with malignant brain tumors are still very limited. Due to their often widely ramified nature and their anatomical location, complete surgical resection of these tumors is often not feasible. Hence, creative, minimally invasive new strategies are required among which virotherapy seems, at least conceptually, the most straightforward approach. We continued here our explorations of the potential of a new candidate oncolytic virus in vivo. We demonstrate that the nonhuman coronavirus MHV can be redirected to the EGFR by using adapter proteins consisting of a soluble mCEACAM1a receptor arm fused to the EGF protein. Genetic incorporation of the adapter sequence in the viral genome was tolerated, and the virus retained its ability to induce cell-cell fusion and produce progeny virus. Moreover, this recombinant virus mediated efficient cell killing of the human glioblastoma cell line U87ΔEGFR. Most importantly, the redirected virus MHVsoR-EGF was effective in eradicating the highly aggressive U87ΔEGFR tumor in an orthotopic mouse model. To our knowledge, this is the first evidence demonstrating in an animal model that coronaviruses have potential as antitumor agents.
Recently, we showed that the species barrier of coronaviruses can be alleviated by using adapter proteins (43
). These adapter proteins were constructed in such a way that either both arms encoded single chain antibody fragments (scFv; e.g., the bispecific scFvs directed against the feline spike protein and the EGFR [48
]), both arms encoded “regular” genes (e.g., the soluble mCEACAM1a and the artificial His tag [43
]), or a combination of the two (e.g., soluble mCEACAM1a fused to the scFv directed against EGFR [47
]). It appeared that the incorporation of adapter proteins consisting of either one or two scFvs in the viral genome was not tolerated by MHV; viruses expressing the adapter gene could not be stably maintained (unpublished data). Fortunately, such viruses could be prepared and maintained when the adapter gene was composed of sequences encoding the soluble mCEACAM1a receptor linked to an artificial His tag (43
) or to EGF (the manuscript). The recombinant MHVs grew with characteristics comparable to those of MHVd2aHE, a similar virus lacking the adapter genes. Most likely, it is the sequence composition of the scFvs rather than insert size constraints that resist stable maintenance in the viral genome, since the introduction of significantly larger sequences has been successfully achieved earlier (8
The in vivo oncolytic activity of our redirected MHV against the highly aggressive brain tumor U87ΔEGFR (31
) appeared to be fast and lasting. Soon after virus inoculation, most of the tumor was found to be destroyed, and no recurrent tumor was observed in this orthotopic model in six of seven MHVsoR-EGF-treated animals. The single animal that had neoplastic tissue lacked any immunohistochemical staining for MHV, suggesting that the redirected virus had not been applied correctly. One other striking outlier was the animal in the control virus-treated group that survived the longest and in which no tumor tissue was observed, implying that this single U87ΔEGFR implantation had not been successful, since all other animals in this group developed large neoplasms.
One obvious drawback to studying the oncolytic capacity of MHV in a mouse model is the virus’ natural tropism. It has been shown that MHV strain A59, the genetic background of our control and retargeted viruses, is weakly neurovirulent and causes demyelination in immunocompetent mice (38
). However, no demyelination was observed in the brains of mice treated with the retargeted virus, since immunohistochemical staining with an antibody to the myelin basic protein appeared similar in infected and PBS control mouse brains (data not shown). Our recombinant MHV-A59 viruses lack the virulence gene cluster 2a/HE, a deletion that appeared to attenuate the virus as judged after intracranial inoculation in C57BI/6 mice (7
). However, mouse tissue is likely to be still permissive to the recombinant viruses since they can be propagated in murine LR7 cells in vitro. Furthermore, our histopathological studies revealed cyst formation and secondary inflammatory changes. Moreover, we did observe substantial loss of tissue within the injected area resulting in the formation of cysts with secondary local granulomatous and neutrophilic inflammation in both acute and later stages. The cysts most likely resulted from liquefactive necrosis of neoplastic tissue and not of necrosis of preexistent neural tissue. This is suggested by the absence of morphological indications of degeneration, necrosis, or inflammation in preexisting areas in the brain where MHV antigen was detected immunohistochemically. Liquefactive necrosis within the central nervous system typically results in cystic structures since the formation of granulation tissue, as is observed in other tissues during damage repair, does not occur in the central nervous system. In conclusion, the neurological symptoms recorded in the animals are probably related to growth of the neoplasms and/or formation of cystic structures in the brain tissue. Most likely, replication of the targeted MHV in nonneoplastic brain tissue caused the neurological symptoms observed in the mice treated with MHVsoR-EGF.
Several other viruses have been explored for their potential as oncolytic agents against glioblastomas. The first one genetically engineered and proven to be efficient in an experimental glioblastoma model was herpes simplex virus (27
). Among the ones that followed were lentiviruses (40
), vesicular stomatitis virus (24
), myxoma virus (25
), measles virus (1
), Newcastle disease virus (5
), and adenoviruses (reviewed in references 17
). One major advantage of nonhuman coronaviruses is that they have a very short replication cycle, resulting in fast clearance of the tumor cells (already observed at day 5 postinoculation, Fig. ). Their ability to cause cell-cell fusion of infected with noninfected neighboring cells is likely to contribute to this. Furthermore, MHV does not infect humans, so its natural tropism does not need to be ablated, nor will preexisting antibodies dampen its reproduction. The EGFR is an intriguing target in high-grade glioblastomas because it is often abundantly overexpressed (9
). Moreover, the most frequently identified mutant EGFR in malignant glioblastomas is EGFRvIII, as is the case in U87ΔEGFR cells (11
). EGFR is, however, also present to some extent on nonneoplastic tissue in the brain, indicating that improving the selectivity of targeting might be required.
Intriguingly, overcoming the entry barrier was not the only requirement for establishing a productive MHV infection. Although postentry blocks of viral replication have been described for several viruses (reviewed in reference 2
), this has never been observed for MHV. Rather, the presence of the viral receptor mCEACAM1a thus far appeared to be the sole requirement for MHV multiplication. As we found out when inoculating several EGFR-expressing human glioblastoma cells with our self-targeted MHVsoR-EGF, successful infection including the induction of syncytium formation was achieved in all cases. Surprisingly, however, the infection did not spread and eradicate the culture except in one case, the U87ΔEGFR cells, which were efficiently killed by the virus. Although not studied here, it is unlikely that binding to the mutant EGFR (ΔEGFR or EGFRvIII) additionally expressed in the U87ΔEGFR cell line accounts for this difference since EGFRvIII is unable to bind EGF as a ligand (31
), and EGFR levels are comparable for U87MG and U87ΔEGFR (31
). Thus, although alleviation of the tropism barrier seems to be a prerequisite for MHV infection, its oncolytic properties might depend on other cellular signaling events, such as interferon signaling (as observed for poxvirus, vesicular stomatitis virus, and Newcastle disease virus), p53 (30
), or Ras (4
). For instance, Ras activation mediates reovirus oncolysis by enhancing virus uncoating, particle infectivity, and apoptosis-dependent release (26
). Moreover, it has been shown that EGFR overexpression or its aberrant functioning can lead to Ras activation (23
). Hence, since U87ΔEGFR is constitutively active in EGFR signaling due to its deletion of exon 2-7, it will be of interest to investigate whether coronavirus replication is also affected by the Ras pathway.
The significant oncolytic activity of the new, redirected animal coronavirus observed in the present study seems to warrant further studies. In view of the complications associated with using a murine virus in a murine model, shifting to a relevant nonmouse model, such as immune deficient rats, seems the most logical option, since MHV does not replicate in other mammalian species. Such studies will hopefully allow us to establish oncolytic efficacy of the recombinant MHV in immunocompetent animals, as well as in other tumors.