The specificity of adenoviral infections in vitro is dictated by the presence of the CAR receptor and integrins. The role of these receptors in vivo is less certain as the cell types to which a vector is exposed becomes a critical issue. Because a majority of vector administered intravenously in rodents is found in the liver, gene therapy for other tissues requires both detargeting from the normal trafficking route to the liver and a delivery system to the target (116,117). For lung epithelium, intratracheal administration is feasible. Direct injection is possible in some other applications, for example, into the myocardium or directly into a tumor. But vector that is inadvertently injected into capillaries or drains into the circulation through the lymphatic system will find its way to the liver. Alternatively, there are cells types such as lymphocytes (38) that are very difficult to infect with type 5 Ad vectors. These considerations raise questions about whether vectors can be retargeted to cells or tissues of interest, or at least whether the expression of the transgene could be limited to that tissue.
Many vectors use the CMV immediate/early promoter/enhancer, which was chosen on the basis that it directs a high level of transgene expression and is expressed in most tissues studied. However, often expression in a specific tissue or cell type is more desirable and expression in other tissues might be toxic. Therefore, the promoters of genes specific to a cell type have sometimes been used for specific applications (118-120). For example, carcinoembryonic antigen and alpha-fetoprotein (AFP) are tumor-specific antigens that are not expressed by normal cells. When a therapeutic gene is expressed from an Ad vector with an AFP promoter, expression should be confined to specific tumor cells expressing AFP and not normal cells. Kaneko et al. (118) showed this theory to be correct and further demonstrated that the expected selectivity is maintained in vivo. In this context, the vector Av1AFPTK1 [expressing thymidine kinase from herpes simplex virus (HSV-TK) from the AFP promoter] can prevent tumor growth in gancyclovir-treated nude mice implanted with a AFP-expressing tumor cell line but not in identical mice implanted with a control (non-AFP expressing) tumor cell line. In contrast, the vector Av1TK1, which expresses HSV-TK from the Rous sarcoma virus promoter, protects gan-cyclovir-treated nude mice regardless of which cell line is used to transduce the tumor.
Alternatively, modifications of the fiber/high-affinity receptor interaction or the penton-integrin interaction might be used to modify tissue tropism. In this context, the seroswitch vectors described above (114), as well as capsid chimeras with part of the Ad3 fiber (121,122) or the fiber gene from Ad17 (123), might be more effective in certain tissues because, a priori, different serotypes of wild-type adenoviruses with different known pathologies should target different tissues. In an extensive survey of the tropism of Ad5-derived vectors, but with fibers derived from different serotypes, the fiber gene of Ad16 was found to be better at targeting fibroblasts and chondrocytes, that of Ad35 better at targeting dendritic cells and melanocytes, and that of Ad50 better at targeting myoblasts and hematopoietic stem cells (124).
Some groups have taken the approach of directly screening for serotypes that replicate preferentially in brain or lung epithelium. In both cases, wild-type strains were screened for efficient replication and certain subgroup D viruses, including serotype 17, were identified to replicate more efficiently. On this basis, a serotype 2 virus with the fiber from Ad17 was constructed and used to study infection of various cell types in vitro. This hybrid is much more efficient at gene transfer to human umbilical vein endothelial cells, neurons, glioma cell lines, and lung epithelial cells than the pure Ad2 gene transfer vector (123). However, it is sufficiently proficient in infecting 293 cells that it can still be propagated and titered for production purposes. It is not known if this translates into better gene transfer efficiency in vivo.
Numerous other approaches to modifying tropism have been identified. For example, the fiber protein can tolerate some manipulation without impairing virus production. The determination of the 3-dimensional structure of fiber in conjunction with mutagenesis studies (125) assists in the identification of amino acids essential for the interaction with CAR and of domains where insertions might be tolerated without grossly affecting structure. An early modification to fiber was the addition of an oligolysine motif to the C-terminal of the fiber protein, giving the virus an affinity for polyanions such as heparin sulfate (126). This profoundly affects the cell types that can be infected in vitro, allowing cells lacking CAR, such as vascular smooth muscle cells and B cells, to be infected. It has also been shown that this oligolysine addition allows for more efficient gene transfer to smooth muscle cells in vivo. Additional manipulations to either the C-terminus or in the HI loop of fiber (summarized in Table 2) have been described to modify the tropism of Ad vectors. The most widely used is the addition of an additional RGD integrin-binding motif to the fiber knob domain, thereby increasing the efficiency of infection of some important cell types, including ovarian cancer cells, fibroblasts, and dendritic cells (127-129). In a study that may point to future developments, phage display technology was used to identify peptide motifs that preferentially target human umbilical vein endothelial cells. When this peptide sequence was incorporated into the HI loop of fiber, aden-oviruses were generated with a high preference for endothe-lium (127,128).
Bispecific antibodies have been used as a reagent to direct Ad toward particular cell types (38,130,131). For example, using a bispecific antibody conjugate with one arm binding the Ad fiber and the other binding the epidermal growth factor (EGF) receptor (130), is was possible to increase the specificity of Ad vectors toward glioma cell lines with low levels of CAR but high levels of the EGF receptor.
The route by which Ad reach the tissue and cell type is complex, especially after intravenous injection. The half-life of Ad in blood is about 2 min, and the interactions with cellular and protein components of blood are unknown (132). In addi tion the degree to which Ad penetrate endothelium in various tissues is unknown, meaning that in vivo tissue tropism is not a simple outcome of the abundance of CAR primary receptor and integrin secondary receptor. Moreover, in humans there will generally be antibodies against Ad so the formation of complexes with antibody will further complicate tropism. If retargeting is to be successful after an intravenous injection, both detargeting from the normal pathways (117), as well as retargeting to the novel pathways, must be achieved.
Therefore, a number of groups have eliminated the amino acids in the knob of fiber that are essential for interaction with CAR. This makes the resultant vectors difficult to propagate in 293 cells and therefore substitute pseudoreceptor-ligand systems have to be devised. For example, the addition of an HA epitope from the influenza hemagglutinin gene permits a 293 derivative with an anti-HA single-chain antibody to be used for propagation of vectors unable to interact with CAR or integrins (117). The interaction of integrin and penton base has been modified both by elimination of the RGD motif in penton base and by its replacement by the LDV motif, which should promote interaction with a4p1 integrins, characteristic of lymphocytes and monocytes (17).
Was this article helpful?