The discovery that binding of the FX GLA domain to the hexon protein on Ad5 played a role in liver transduction has opened the door to developing new ways to reduce viral uptake in the liver. In this study, we have taken a different approach and have tested an alternative use for FX with Ad5. We hypothesized that fusing the hexon-binding GLA domain of FX to ligands might be able to retarget the virus to new receptors. We also hoped that this would enable the use of potent ScFv ligands with the virus by bypassing the compatibility problem between secreted ligands and intracellularly assembled Ad.
Our data show that both the GLA and GLA-EGF1 domains of human coagulation FX can bind to Ad5 and redirect the virus to new receptors when they are fused to separate ScFvs. It is interesting to note that the GLA protein retargets virus more efficiently when coupled with the EGF1 domain from FX. These data suggest that binding of the GLA domain to Ad5 may be dependent on binding of calcium to the adjacent EGF1 domain or may simply be due to increasing the flexibility of the fusion protein. We show that three separate ScFvs targeting three different receptors worked with GLA-EGF1, suggesting that it should be a generally useful platform for many ligands.
First proof of principle was demonstrated with cell supernatants from transfected cells. This demonstrated that the GLA fusion could be combined with three different ScFvs to mediate targeting. Testing with purified GLA-αEGFR demonstrated that increasing amounts of the fusion protein mediated increased transduction. Importantly, this failed when GLA-αEGFR was combined with Ad5 whose hexons were modified with a large BAP, demonstrating that the GLA–hexon interaction is pivotal to the retargeting effect.
We hypothesized that the GLA affinity for hexon might be sufficient to block binding of native FX to the virus after intravenous injection and reduce hepatocyte infection. Unfortunately, preincubation of purified GLA-αEGFR failed to mediate this type of hepatocyte detargeting after intravenous injection. This failure could simply be due to the exceptionally high levels of FX in the blood (2
) that may still rapidly replace GLA protein on the virus after injection. This is supported by the observation that inactivation of endogenous FX with warfarin before injection ablated liver infection. Regardless of the explanation, GLA fusions do not appear to detarget liver, so they are likely to be most useful for retargeting to new receptors.
To sidestep any need for protein purification and to allow for multiple generations of targeting by viruses in vivo, GLA fusion proteins were inserted into replication-defective and replication-competent oncolytic Ad to allow these vectors to retarget vectors in situ. Expression from a replication-defective Ad allows this virus to provide targeting protein in trans to an oncolytic Ad or another Ad gene therapy vector. Expression directly from an oncolytic Ad allows the virus to produce its own cell-targeting proteins. Under these conditions, we show in vivo that both replication-defective and replication-competent Ads expressing the EGFR-targeting ScFv fused to GLA reduce tumor size in an ovarian cancer model and that the replication-competent virus has markedly stronger effects on improving survival of mice bearing these tumors. These data suggest that GLA targeting with ScFvs may be an effective Ad-retargeting scheme for oncolytic therapy, cancer gene therapy, or other applications.
Unlike other methods that have been used to target adenovirus via fiber conjugates, the GLA fusion is conjugated to hexon. Evidence suggests that release of the overproduced fiber protein after lysis of adenovirus-infected cells inhibits viral spread. The proposed mechanism is that free-floating fiber binds to and blocks fiber receptors, like CAR, on uninfected cells (Rebetz et al.
). This implies that most fiber-targeted adenoviruses (including wild type) will spread poorly. In fact, this has been a major problem for adenoviral therapy, as exemplified by oncolytics that fail to spread beyond the needle track. Interestingly, free GLA fusion protein should not block adenovirus binding after binding to its ScFv receptor. Rather, receptor-bound GLA fusion protein could act as an attachment site for any adenovirus not saturated with GLA proteins. Because each adenovirus has 720 hexon proteins, it is not likely to become saturated with GLA fusion protein in an extracellular environment. Thus, GLA fusion-targeted adenovirus may be superior to other targeted adenovirus in its ability to spread to uninfected cells.
ScFvs have arguably been the most difficult ligand to use with Ad5 because of their need for disulfide formation and glycosylation. Therefore, successful retargeting with ScFvs suggests that the use of GLA or other hexon interacting domains with other ligands will likely mediate retargeting. High-affinity peptides, proteins, full-length antibodies, drug compounds, or magnetic particles could also be linked to the GLA domain. Because GLA can bind to any Ad5 or serotype that binds FX, this targeting approach can be applied to an array of adenoviruses to modify tropism. For those that do not bind FX, the development of other hexon display approaches may also enable retargeting of other Ads.
Previous attempts to target with high affinity antibodies bound to hexon via the high affinity avidin-biotin interaction (10−15M
) failed (Campos and Barry 2006
). In contrast, antibody targeting using the lower affinity GLA-hexon interaction (10−9M
) succeeded here. This suggests that ligand-receptor and ligand-virus interactions must be sufficiently stable to target in vivo
, but that if any of these interactions are too strong, the virus may become trapped on the receptor or fail in downstream interactions. This paradigm appears to apply for targeting via complex formation and will likely apply when inserting ligands genetically into Ad and other viral capsids (i.e.
, AAV). In these cases, insertion of a low affinity peptide ligand into the capsid of hexon on Ad or VP on AAV will likely fail to target in vivo
. In contrast, insertion of too strong of a ligand or complex forming domain will likely fail due to the virus being trapped on receptors.