In this study we report that cells which do not express CAR can be efficiently transduced by Ad5. Adenovirus entry is not dependent on fiber binding to cells but instead is blocked by an RGD peptide that interferes with the RGD domain on the adenovirus penton base binding cellular integrins. Further, we find that binding to low CAR cells is inhibited specifically by a blocking antibody to integrin αvβ5, demonstrating that integrin αvβ5 is required for Ad5 attachment to these cells. The binding mediated by integrin αvβ5 is extremely high affinity, in the picomolar range. Our data further challenges the prevailing model of adenovirus infection, in which binding to a primary receptor, CAR, is required in order for subsequent interactions between adenovirus and integrins to initiate viral entry.
Ad5 binding to low CAR cells is independent of fiber and instead depends on an interaction with integrin αvβ5. Therefore, Ad5 does not require an independent binding receptor to dock it to the cell before it can interact with internalization receptors. Other viruses are also reported to use both primary binding and internalization receptors. HIV-1 first binds to CD4 followed by binding to the chemokine receptors CCR5 or CXCR4, which trigger membrane fusion [49
]. Binding to CD4 induces conformational changes in the HIV protein gp120, revealing the previously hidden binding site for its coreceptors [6
]. Variants with mutations in gp120 allowing for direct interaction with coreceptors have been isolated in vitro
; however, these variants are sensitive to neutralizing antibodies and therefore selected against in vivo
]. Therefore, the role of CD4 binding in HIV-1 infection may be particularly critical in evading the immune system of the host.
Unlike HIV-1 binding to CD4, Ad5 binding to CAR does not induce conformational changes in viral proteins, thus facilitating subsequent entry steps [51
]. Rather, CAR is thought to facilitate a high affinity interaction to fiber, thus docking the virus to the cell surface and allowing the subsequent interaction between the penton base and integrins to initiate internalization. Indeed, only the extracellular domain of CAR is required for CAR-mediated adenovirus entry [52
]. However, previous studies have observed CAR-independent infection and the work in this paper demonstrates that Ad5 can bind directly to integrin αvβ5, previously identified as an internalization receptor [13
]. Therefore, our results and the results of others suggest this initial binding step is not required for entry. Nevertheless, CAR binding is conserved in a number of adenovirus serotypes and the fiber-CAR interaction is one that is well characterized and is itself of high affinity [10
]. Therefore, CAR binding clearly plays an important role in the adenovirus infection cycle. One possibility is that, even when CAR is not needed to enter cells, CAR functions as an exit receptor [54
]. Walters et al report that when Ad5 lyses a cell, excess fiber is released and through binding to CAR, disrupts neighboring cell-cell junctions, allowing for release of the virus back to the apical surface where it may continue infecting cells [54
]. Additionally, recent work has demonstrated that adenovirus binding to CAR may induce downstream signaling events that increase integrin activation, thus promoting infection[55
]. Therefore, binding to CAR may facilitate entry in ways beyond simply docking the virus to the cell surface. Indeed, at least two serotypes of adenovirus, Ad9 and Ad37, have fibers which bind CAR but do not use CAR as an attachment receptor, supporting the idea that CAR binding is important for steps other than attachment [10
]. Further, it is likely advantageous to the virus to be able to use multiple entry routes, enabled by its ability to engage multiple different receptors.
Both HIV, as evidenced by CD4-independent variants isolated in vitro
, and Ad5, as evidenced by our results and the results of others can infect cells without binding to their so-called primary receptors. Binding to these receptors, instead of being strictly required for infection, may contribute to other necessary parts of the virus infection cycle, such as evading the host immune system or facilitating virus escape. Many other viruses with less characterized receptors seem to also use multiple receptors, some classified as binding receptors [57
]. For example, rotaviruses are thought to first bind to a sialic acid (SA)-containing molecule, which anchors the virus to the cell, and then bind to coreceptors to initiate viral entry [57
]. Mutant variants of rotaviruses that are SA-independent and interact directly with coreceptors have been isolated in vitro
, suggesting that similarly to HIV and Ad5, binding to the primary receptor is not strictly required for infection [58
]. Therefore, the interaction between rotaviruses and SA-containing molecules may facilitate an as yet unidentified aspect of rotavirus infection. As more functional roles of virus receptors in infection are elucidated, the use of binding receptors in other aspects of the viral life cycle may emerge as a general principle of viral pathogenesis.
In addition to being used as a model system for viral entry, much effort has been put into developing adenoviruses, especially species C adenoviruses including Ad5, as vectors for gene therapy. In fact, adenoviral vectors have been used in more than one quarter of gene therapy trials worldwide [60
]. Cancer is one of the most common targets of adenovirus-mediated gene therapy. As mentioned previously, CAR expression is often lost as cancers progress and this loss has been viewed as a major hurdle to using adenovirus-based therapies in cancer [31
]. However, integrin αvβ5 has been reported to be overexpressed in cancers [61
]. Therefore, our conclusion that Ad5 can use integrin αvβ5 to bind to and infect cells lacking CAR suggests that cancer cells having lost CAR expression may still be good targets for adenovirus-based therapies. Recent work has shown that erythrocytes sequester adenovirus by binding CAR, thus limiting systemic infection; therefore, using CAR-ablated vectors, a strategy many groups are attempting, may improve delivery for gene therapy for reasons beyond changing receptor interactions [63
]. We also observed what may be an as yet unidentified obstacle to these therapies, however. T47D cells, which express CAR (Figure ) and integrin αvβ5 (data not shown) are still resistant to Ad5 infection (Figure ). Wang et al showed the cellular protein CEACAM6 blocks adenovirus trafficking to the nucleus in human pancreatic cancer cell lines[65
]. Future studies to determine if this protein blocks infection in T47D cells, or if resistance is due to a novel mechanism are needed.