Recently, the use of viruses as potential tools to complement the conventional treatment options for cancer therapy has gained considerable interest (28
). In this study, we demonstrate that the mouse coronavirus MHV may be an attractive candidate for this purpose. MHV has a narrow species tropism and a strong capacity to kill cells. These features might be used to advantage by controlled targeting of the virus to human tumor cells. Using the specificity of the mCEACAM1a MHV receptor, we constructed bispecific adapter molecules composed of the S protein-binding fragment of this receptor coupled to an antibody fragment directed to the human EGFR, which is known to be overexpressed on cells of many human cancers. The adapter proteins were found to effectively mediate infection of EGFR-expressing human A431 cancer cells. Cell entry appeared to be effected through the normal spike protein-driven membrane fusion mechanism, and the infection resulted in efficient formation of syncytia, which we observed earlier to ultimately lead to efficient eradication of the culture (39
We have shown recently that targeted entry of the feline coronavirus FIPV and chimeric fMHV could also be accomplished by making use of an adapter composed of two single-chain antibodies, one directed against the feline viral spike and the other to the EGFR (39
). Like the soR-425-mediated MHV infections, the bispecific antibody-mediated fMHV infections were blocked by an HR-mimicking peptide derived in this case from the feline S protein (39
). These observations suggest that binding of the adapters to particular regions of the S protein—i.e., the receptor-binding domain or a certain epitope—results in conformational changes that, upon binding of the virus-adapter complex to a cellular target receptor, facilitate insertion of the fusion peptide into the cell membrane. Binding of these adapters apparently does not inactivate the virus, as is also indicated by studies in which MHV, purified by sedimentation through a sucrose cushion after preincubation of the virus with the soR-425 adapter protein, maintained its infectivity for EGFR-positive cells (our unpublished data). These results suggest that the adapter proteins, once bound to the virus, may freeze the S protein in a stable prefusion conformation ready to proceed to its structural rearrangements upon contacting the target cell. What the actual trigger is that activates the fusion process is unknown, but it is quite likely that the insertion of the fusion peptide into the cellular membrane is a critical event.
An interesting paradox that we noticed in our studies was the opposite efficiency of hinge-containing and hingeless adapter molecules in neutralization and targeting. Thus, the virus-neutralizing capacities of the adapters soR-h and soR-h-425 were significantly higher than those of the soR and soR-425 molecules, respectively, while the reverse was true of the relative targeting efficiencies of soR-h-425 and soR-425 (Fig. to ). The reasons for these differences appear to be unrelated. This can be concluded from the differential effects of reducing agent treatment of the adapters on the two activities. Whereas this treatment did not affect the targeting efficiencies of the adapters, it reduced the virus-neutralizing capacities of the hinge-containing adapters to about the level of their hingeless counterparts, the activity of which was not significantly changed by the treatment. Dimeric forms of the adapters apparently neutralize the infectivity of the virus more efficiently than monomers, possibly by their stronger interaction with the spikes or their ability to cross-link spikes on the viral particle. Remarkable as it may seem, these effects apparently have no bearing on the efficiency of the targeting of such viral particles or on the subsequent steps leading to membrane fusion.
Previous studies have shown that the N domain of the mCEACAM1a receptor is sufficient by itself to induce conformational changes in the MHV spike protein (22
). Remarkably, however, this same N domain, when linked directly to the transmembrane domain, thus lacking the intermediate CEACAM1a A2 domain (Fig. ), failed to function as a receptor for MHV (9
). A possible explanation for this observation might be that the functional structure of the N domain is affected by linking it to the transmembrane region. This appeared, however, not to be the case when the N domain was replacing the corresponding domain in the mouse poliovirus receptor homolog (25
), which also belongs to the family of Ig-like viral receptors (23
). The resulting chimeric receptor molecule functioned as a receptor for MHV (6
). It is conceivable that the CEACAM1a N domain needs a certain distance from the cellular membrane to allow interaction with the S protein to lead to effective membrane fusion. Interestingly, the efficiency of the targeted infections described here was significantly higher with the soR-425 adapter than with the soR-h-425 adapter, and this difference was not accounted for by the dimeric state of a fraction of the latter, as the targeting efficiency of the hinge-containing adapter was unaffected by dissociation through DTT treatment. These observations indeed suggest that also the dimensions of our adapters may affect the efficiency of their targeting. Whether this is really the case, and how, remains to be investigated.
The ability to target viruses by design to preselected cells is a tremendous challenge with far-reaching implications for the development of virus-based therapies. We previously established the principle of retargeting of coronaviruses by exchanging spike ectodomains (16
). Subsequent attempts to modify the tropism of these viruses further by incorporating tumor-binding ligands into different parts of the spike protein were unsuccessful, as no viable viruses could be rescued (M. H. Verheije, T. Würdinger, and P. J. M. Rottier, unpublished data). As an alternative and potentially also versatile targeting approach, we therefore aimed at the development of bispecific adapters. The use of soluble receptor-based adapters to expand virus tropism has been described before for retroviruses (32
) and adenoviruses (20
). Recently, also the herpes simplex virus was redirected to a new receptor via the variable domain of its cellular receptor, nectin 1, fused to a single-chain antibody (27
). However, in contrast to these viruses, the coronavirus MHV is normally incapable of infecting human cells, due to its restricted tropism. Hence, its native tropism does not need to be eliminated to obtain a truly targeted virus, which avoids possible reduction of its cytotoxicity to the specific target cells. In order to fully exploit the oncolytic properties of MHV, multiround genetic targeting will be essential. As we described earlier, incorporation of the gene encoding the bispecific single-chain antibody s11-425 into the adenovirus genome leads to persistent self-targeting of the virus to a specific receptor (35
). Also others showed virus retargeting by genomically expressed bispecific adapters, though this resulted in a reduction of the oncolytic potency of the adenovirus in the case of the sCAR-EGF adapter (19
). The same approach seems also feasible for the development of tumor-targeted coronaviruses. As we and others showed recently, MHV can tolerate and express foreign genes from various insertion sites of its genome (4
). Therefore, our next step in the development of oncolytic coronaviruses will be to introduce the gene encoding a bispecific adapter molecule into the virus genome.