Advanced Generation Adenoviral Virotherapy Agents Embody Enhanced Potency Based upon CAR-Independent Tropism

Gene therapy has become an area of focus in the development of new cancer treatment regimens based on the recognition that innovative therapeutic approaches are required to improve current survival rates. Virotherapy marks the newest approach, in which a replicating virus itself is the anticancer agent. In this regard, virotherapy exploits the lytic property of virus replication itself to kill tumor cells. Because this approach relies on viral replication, the virus can self-amplify and spread in the tumor from an initial infection of only a few cells. Although attempted in the past, and abandoned because of toxicity and inefficacy (2), the virotherapy approach has reemerged with great promise, in large part due to better understanding of virus biology and the ability to genetically modify viruses.

The adenovirus-based vector has emerged as a leading candidate for in vivo oncolytic virotherapy. Adenoviruses are attractive vectors because they can be produced in high titers, do not integrate into the host chromosome, and have a broad tropism. In addition, adenoviral vectors infect both dividing and nondividing cells, have high stability in vivo and have a high capacity for gene transfer. Another beneficial attribute contributing to their employment in antitumor therapy is that adenoviruses possess a lytic life cycle that can be exploited for oncolysis. Although adenoviruses do not have a natural tendency to replicate in tumor cells, they can be rendered to do so.

Adenoviral vectors based on the native adenoviral serotype 5 (Ad5) are the most commonly used for construction of virotherapy agents. An early generation adenoviral virotherapy agent (dl1520 or ONYX-015) has a deletion in the E1B region, which serves to accomplish tumor-specific replication. In clinical trials, this agent has shown clear benefits for recurrent head and neck cancer in combination with chemotherapy (3). Another early Ad5 virotherapy agent is the AdΔ24 vector, containing a 24-bp deletion in the adenoviral E1A gene that targets replication to Rb gene–defective tumor cells (4). We have developed a number of adenoviral virotherapy agents, which harbor the E1A gene under the control of tumor-specific promoters/elements. These include the cyclooxygenase-2 promoter (5), midkine promoter (6), secretory leukocyte protease inhibitor promoter (7), survivin promoter (8), human telomerase reverse transcriptase promoter (9), tyrosinase enhancer/promoter (10), and the vascular endothelial growth factor promoter (11).

One of the attractive features of adenoviral agents for virotherapy is their unparalleled efficiency in accomplishing in vivo infection. Indeed, of all of the currently available vector approaches, adenoviral vectors possess the highest capacity to achieve in vivo infection of tumors. Despite this capacity, overall efficacy of adenoviral virotherapy-based cancer clinical treatment regimens remains limited by suboptimal infectivity. A primary discrepancy between poor response in clinical trials and very encouraging results in preclinical studies using either nonreplicative or replicative Ad5-based systems has been identified. This difference has now been understood to result from relatively low expression of the primary Ad5 receptor, the coxsackie and adenovirus receptor (CAR), in primary tumors relative to their cell line counterparts (12). Cancer cells have been specifically shown to be profoundly resistant to Ad5 infection, based upon a relatively low expression of CAR (13). This CAR deficiency may be explained by the finding that CAR has been shown to exhibit tumor suppression activity (14). Based on this block in viral cell attachment, it has been proposed that gene delivery via CAR-independent pathways may be required to circumvent tumor cell deficiency of CAR (15, 16). Thus, it is clear that augmenting infectivity of adenoviral vectors for cancer targets is of fundamental importance to derive their full benefit as adenoviral virotherapy agents.

Native Ad5 tropism is mediated by two capsid proteins: the fiber and the penton base. These proteins bind to the primary high-affinity cellular receptor CAR and to the integrins α_vβ₃ and α_vβ₅, respectively. Based on an understanding of these molecular interactions, a concerted effort has been made to modify Ad5 tropism, resulting in enhanced tumor cell transduction by retargeting cellular entry through heterologous pathways. Ad5 retargeting studies have established the concept that altered cellular tropism can be accomplished with genetic modifications of the viral domains that dictate cellular attachment. A trimeric fiber protein protrudes from each of the 12 vertices of the icosahedral adenoviral particle, where it is attached noncovalently to the penton base (17). It has been shown that the globular COOH-terminal knob domain of the adenoviral fiber protein is the ligand for attachment to the Ad5 cellular receptor CAR, the first step in infection (18–21). Based on the demonstration that infection can be blocked (22), it is recognized that the knob is necessary for Ad5 binding to host cells. The NH₂-terminal tail domain of the fiber is separated from the knob domain by a long rod-like shaft comprising a 15-amino-acid residue β-spiral motif repeated 22 times in Ad5 (23). Following attachment, the next step in Ad5 infection is internalization of the virion by receptor-mediated endocytosis. This process involves the interaction of Arg-Gly-Asp (RGD) sequences in the penton base with the host internalization cell receptors, the integrins α_vβ₃ and α_vβ₅ (24). The understanding of adenoviral structure-function and the cellular entry pathway has facilitated attempts to modify the tropism of Ad5 vectors and permit the enhanced transduction of tumor cells.

The first efforts to modify Ad5 infection to expand their tropism to CAR-independent entry pathways focused on the incorporation of peptide ligands into the COOH terminus or HI-loop of the fiber protein. However, the addition of various targeting ligands to the Ad5 fiber defined a clear size limit for incorporated heterologous sequences (25–27), effectively limiting the repertoire of motifs to small peptides. In view of these structural limitations, more dramatic efforts to genetically modify adenoviral vectors for CAR-independent tropism were done by complete replacement of the Ad5 knob domain.

As the primary determinant of viral tropism, the capsid fiber protein has represented the most logical site for genetic engineering for targeted transduction. Trimerization of the fiber is essential for capsid incorporation and thus cannot be perturbed. One effective strategy is to capitalize on the structural homology present between Ad5 and the other adenoviral serotypes. Because a high degree of structural similarity exists in the adenoviral fiber proteins, chimeric vectors employing whole fibers or even the distal knob domains from other adenoviral serotypes represent an efficient and natural means of achieving CAR-independent tropism and alternate receptor recognition. Of course, this serotype fiber/knob “switching” approach is relevant to those adenoviral serotypes whose native tropism is associated with receptors other than CAR. Fifty distinct serotypes of human adenoviruses have been identified; these serotypes have been classified into six subgroups (A-F) based on sequence comparisons, each subgroup with different tropisms. Adenoviral serotype 5 is classified as subgroup C, which also includes serotypes 1, 2, and 6. The subgroup B adenoviruses can be divided into subgroup B1 (serotypes 3, 7, 16, 21, and 50) and B2 adenoviruses (serotypes 11, 14, 34, and 35). These subgroups display different organ tropisms in vivo, suggesting a difference in receptor usage. In this regard, the knob domains of adenoviruses of serotype 5 (subgroup C) and serotype 3 (subgroup B1) have been shown to bind to different cellular receptors (28). It is now apparent that the subgroup B adenoviruses can use multiple cellular attachment receptors, including CD46, CD80, and CD86 (29, 30).

Based on differences in receptor use (31), it was shown that the native tropism of adenoviral subgroup C could be altered by genetic incorporation of an alternate fiber knob domain from subgroup B. For example, replacement of the Ad5 fiber knob domain with that of adenoviral serotype 3 fiber resulted in enhanced cytopathicity of an adenoviral virotherapy agent to primary melanoma cells, which was at least four orders of magnitude higher than wild-type Ad5 (32, 33). This transductional enhancement was similar for melanoma cells, ovarian cancer cells (7, 34), renal cancer cells (16), and squamous cell carcinoma (35). Interestingly, the genetic shortening of the Ad5 fiber shaft in an Ad5/Ad3 chimera also significantly reduced liver tropism (36). Importantly, the chimeric Ad5/Ad3 virus has been shown to be progressively more efficient at each step of the replication cycle compared with its Ad5 counterpart (35).

In the current issue of Clinical Cancer Research, Reddy et al. (1) replaced the fiber knob of an Ad5 virotherapy agent with the knob domain of either Ad3 or Ad35 from subgroup B. The infectivity of Ad5/3 fiber chimera used in this study was further enhanced by insertion of an RGD motif at the COOH terminus of the Ad3 knob. The authors showed that these fiber modifications resulted in dramatically improved transduction, replication, and cytotoxicity in vitro and enhanced antitumor efficacy and survival advantage in vivo using models of melanoma and head and neck cancer. These results obtained from this strategy of adenoviral tropism modification to use alternate cellular attachment receptors would predict an improved therapeutic index in clinical trials.

CD46 has been identified as one of the cellular attachment receptors for the subgroup B adenoviruses Ad3 (29), Ad11 (37), and Ad35 (37, 38). In addition, the subgroup D adenovirus Ad37 has been shown to use CD46 as an attachment receptor (39). CD46 is ubiquitously expressed and belongs to the regulators of complement activation gene family, whose biological role is to protect cells from complement-mediated attack. Recent evidence has suggested an additional role for CD46 in linking innate and acquired immunity by signaling events induced in macrophages and lymphocytes (40). It is now apparent that malignant tumor cells also express these proteins to escape complement attack. CD46 consists of (a) four NH₂-terminal copies of an ∼60-amino-acid structural motif termed the short consensus repeat (SCR; also called complement control protein repeat), (b) one to three serine-threonine-proline-rich domains, (c) a short region of unknown function, (d) a transmembrane spanning domain, and (e) a COOH-terminal cytoplasmic tail. A common feature of proteins with SCR domains is that they recognize virus particles through two or more domains.

Cryoelectron microscopy single-particle reconstruction has provided moderate resolution structures for a variety of adenoviral vectors, adenovirus/antibody, and adenovirus/integrin complexes (41). Cryoelectron microscopy studies combined with functional assays have shown the importance of fiber shaft flexibility for interaction with CAR (42). In particular, the Ad5 fiber is long (>300 Å) and flexible at the third β-spiral repeat of the fiber shaft, which corresponds to a point ∼60 Å above the penton base (43). An X-ray crystal structure of a complex of the Ad12 fiber knob with one domain of CAR showed conclusively that it is the side of the fiber knob that interacts with CAR (44). Thus, the bend in the fiber shaft of many CAR-using adenoviral serotypes may be essential for formation of the proper geometric relationship between the fiber knob and CAR (43).

Thus far, no structure has been determined for a complex of an adenoviral fiber with CD46; however, structural information is available for several CD46-binding fibers and for two domains of CD46. The CD46-binding fibers of Ad37 and Ad3 have been shown to be both short (150 and ∼120 Å, respectively) and rigid by cryoelectron microscopy reconstruction (43, 45). The Ad35 fiber is also short (∼85 Å) and somewhat flexible (46). Before CD46 was identified as an adenoviral attachment receptor, it was hypothesized that fiber flexibility might not be required for interaction with the alternate attachment receptor if the critical binding residues were on the top, rather than the side, of the fiber knob (43, 47). In addition, it was proposed that the length of the Ad37 fiber might be optimal for concurrent binding of fiber with its attachment receptor and penton base with α_v integrin (43). Since then, the identity of the alternate receptor has been established and an X-ray crystal structure determined for two domains of CD46, SCR I and II (48). Analysis of point mutations and domain mutations in CD46 has indicated that it is the SCR I and II domains that interact with the Ad35 fiber (49). The same study also served to map the putative Ad35-binding surfaces of CD46 SCR I and II. One residue near the top of the Ad37 fiber knob, Lys²⁴⁰, was shown to be important for binding to conjunctival cells (47) and presumably is within the CD46-binding interface. Although the precise protein-protein interactions among the Ad3, Ad35, and Ad37 fibers with CD46 have yet to be determined by a structural study, a “schematic” model can be built based on the available structural and mutagenetic information (Fig. 1).

A cryoelectron microscopy reconstruction of Ad37F, an adenoviral vector composed of the Ad5 capsid and the Ad37 fiber, is shown in Fig. 1A. The full length of the fiber was reconstructed, including the distal knob, indicating that it is rigidly straight. Figure 1B shows a structure-based diagram of the interaction of the Ad37 fiber with CD46. The four homologous SCR domains are represented as forming a somewhat compact rather than fully extended structure on the cell surface. Some compaction is reasonable to assume given the flexibility at the domain interface indicated by the crystal structure. Consideration of the structural geometry involved makes it plausible that α_vβ₅ integrin could simultaneously bind to the penton base while the Ad37 fiber is attached to CD46 (Fig. 1C). It is likely that an optimized dual receptor interaction leads to efficient adenoviral cell entry. In support of this hypothesis, the length of chimeric CD46-CD4 receptor molecules was found to affect the efficiency of Ad35 uptake (49).

In the work of Reddy et al. (1), human tumor xenograft models in immunodeficient mice were used to quantitate efficacy of the chimeric adenoviral virotherapy agents in vivo. Although important preclinical studies using adenoviral virotherapy agents have been derived from such models, their major limitation is that the mouse host is generally regarded as a nonpermissive species for adenoviral replication. Thus, important for the future development of novel adenoviral virotherapy agents is the identification of tumor model systems using hosts that are permissive for adenoviral replication to more closely resemble the human clinical disease. In this regard, the use of an ex vivo model system that involves the evaluation of CRAd toxicity and therapeutic efficacy in thin, precision-cut slices of human primary tumor and liver may more closely predict therapeutic index than traditional cell culture models (50, 51). The cotton rat (a rodent species that is semi permissive for human adenoviral replication) has shown use as an in vivo model to test the selectivity, immunogenicity, and efficacy of oncolytic adenoviral vectors (52). More recently, the Syrian hamster has been used as an immunocompetent and replication-permissive model to evaluate an adenoviral virotherapy agent. Cross-species replication of Ad5 has also been shown in canine cells (53). Because the biological behavior and clinical presentation of certain dog tumors closely resemble those of their human counterparts, these results raise the possibility of exploiting canine models for preclinical analysis of candidate adenoviral virotherapy agents designed for human use. In conclusion, the development of in vivo and ex vivo models for evaluation of efficacy and toxicity may a powerful tool to aid in the clinical translation of advanced generation adenoviral virotherapy agents.

In the work of Reddy et al. (1), adenoviral virotherapy efficacy is determined in vivo by measuring changes in tumor volume, and viral replication and spread is determined by immunohistochemistry in xenograft tumors. The completed gene therapy clinical trials, thus far, have also had to rely on conventional histology of biopsy specimens and analysis of body fluids for the detection of virus. Such static assessments fall short of accurately depicting the dynamic mechanism of replicative agents. Critical for future development of novel adenoviral virotherapy agents is the ability to ascertain in vivo quantitative assessments of replication and spread of the adenoviral virotherapy agents. Clinical trials to date have relied on conventional histology of biopsy specimens and analysis of body fluids for the detection of virus; such static assessments fall short of accurately depicting the dynamic mechanism of replicative agents. Several studies have attempted to address this problem, using the detection of reporter transgene expression. Despite their use for assessing gene delivery and expression, these reporters by themselves are not suitable for monitoring activity of adenoviral virotherapy agents. The essence of oncolytic virus function is to infect and kill target cells, a concept that is at odds with reporter gene expression. In addition, reporter gene expression may not truly represent the underlying level of viral replication or the physical distribution of viral progeny.

The combination of a noninvasive imaging modality with a genetic adenoviral labeling system for detection of replication and progeny localization would provide a powerful means to monitor adenoviral virotherapy agent in vivo. This strategy for genetic labeling of adenovirus through the fusion of a reporter with a structural protein has led to investigation of the capsid protein IX as a potential locale for labeling the adenovirus through the fusion of reporter proteins. In an adenoviral vector construct that incorporated a fusion between the COOH terminus of pIX and the red fluorescent protein (RFP), the pIX-RFP fusion protein was efficiently incorporated into an adenoviral virotherapy agent and had little effect on viral DNA replication, thermostability, CAR binding, or cytopathic effect binding (54). In addition, the RFP signal correlated with viral DNA synthesis and infectious progeny production both in vitro and in vivo. A recent report also described the incorporation of the herpes simplex virus type 1 thymidine kinase (TK) protein as a fusion to pIX (55). In this study, the pIX-TK fusion protein incorporated within the adenoviral virions functioned as a reporter gene for micro-positron emission tomography imaging. Although noninvasive imaging of adenoviral vector biodistribution poses formidable challenges, the genetic adenovirus pIX-labeling system may have an important effect in the field of adenoviral virotherapy. Other conventional imaging systems designed for the detection of transgene expression of such reporters as green fluorescent protein (56), luciferase (57), sodium iodide symporter (58), somatostatin receptor type 2 (59), and thymidine kinase (60) highlight the need for surrogate end points in monitoring replication and spread of adenoviral virotherapy agents.

The strategy of infectivity enhancement modification by serotype knob switching has allowed dramatic augmentations in gene delivery to tumor targets in preclinical studies, with a specificity that would predict an improved therapeutic index. In particular, the work of Reddy et al. (1) showed that intratumoral administration of oncolytic adenoviruses were effective for the treatment of locally confined tumors. As advanced generation adenoviral virotherapy agents are just entering clinical trials, the gains in therapeutic efficacy through infectivity enhancement strategies should soon become apparent. However, whereas these modifications have shown use to enhance virotherapy potency based upon CAR-independent tropism, they still are not able to target metastatic disease effectively. In this regard, it may be possible to combine infectivity enhancement with tumor-specific targeting by incorporating high-affinity targeting ligands. This concept of a complex mosaic approach was recently tested by incorporating the RGD motif into HI loop, at the COOH terminus, or both locales of the Ad3 knob, in the context of Ad5/3 chimera fiber, to simultaneously retarget the adenoviral vector to integrins and to the Ad3 attachment receptor CD46 (61). This study showed that complex mosaic modification can function via dual-receptor targeting. Reddy et al. (1) also showed enhanced infectivity of an Ad5/Ad3 chimera by insertion of an RGD motif at the COOH terminus of the Ad3 knob. Such strategies to combine the targeting specificities of short peptide ligands together with fiber knob switching of various adenoviral serotypes have the potential for further reduce deleterious side effects and increase the therapeutic index of virotherapeutic agents in vivo. As their development proceeds, these novel adenoviral virotherapy agents will likely be of universal relevance to a broad spectrum of potential anticancer targets and sites for in vivo gene delivery to patients.