The goal of this study was to determine if latent AAVs exist in humans, and if so, to characterize their structural, serologic, and functional properties.
RESULTS AND DISCUSSION
Human tissues were obtained from a variety of sources, and DNAs were evaluated for endogenous AAV sequences by PCRs with oligonucleotides specific to homologous regions of the cap gene. Figure 1
summarizes a portion of the results of this screen of 259 human samples of 18 different tissue types derived from 250 individuals. The data were compared to similar studies using an expanded pool of tissues from rhesus monkeys, cynomolgus and pigtailed macaques, baboons, and chimpanzees. The prevalence of AAV sequences in human and nonhuman primate tissues was similar (19 and 18%, respectively). Livers and spleens were the predominant sites of AAV infection in both human and nonhuman primates, although endogenous sequences were also frequently found in colons and bone marrow from humans and lymph nodes from nonhuman primates. Quantitative PCR studies indicated that endogenous AAV in human tissues is present in low quantities and unlikely to be present as a result of germ line transmission (data not shown).
To better understand the origins and consequences of endogenous AAVs in humans, we attempted to recover and fully sequence full-length cap structures from human tissues; isolates from nonhuman primates, in addition to what was previously described, were also included. A total of 108 new and unique isolates (from 55 human and 53 nonhuman primates) were identified (clones from the same individual with fewer than four amino acid differences were deemed redundant and eliminated from the analysis).
This pool of primate AAV cap sequences was analyzed for phylogenetic relationships by using a variety of computational approaches. Sequences were aligned with the ClustalX1.81 program and phylogenies were assessed by the neighbor-joining, maximum parsimony, and maximum likelihood algorithms (18
). Each method yielded a similar clustering of sequences. The phylogeny of AAV was further evaluated for evidence of recombination through a sequential analysis of split decomposition (2
) and bootscanning (20
). Split decomposition analysis depicts parallel events in a set of sequences with a tree-like network rather than a bifurcating tree. The bootscanning algorithm then further verifies these putative recombination events by visualizing the mosaic structure of a given sequence. A number of different cap sequences amplified from eight different human subjects showed phylogenetic relationships to AAV2 (5′) and AAV3 (3′) around a common breakpoint at position 1400 of the cap DNA sequence, consistent with recombination and the formation of a hybrid virus (Fig. 2
). This is the general region of the cap gene in which recombination was detected in isolates from a mesenteric lymph node of a rhesus macaque (9
The phylogenetic analyses were repeated, excluding the clones that were positively identified as hybrids. In this analysis, goose and avian AAVs were included as outgroups (6
). Figure 3
summarizes a neighbor-joining tree; similar relationships were obtained by maximum parsimony and maximum likelihood analyses. These analyses demonstrated 11 phylogenetic groups, which are summarized in Table 1
. When a group contained nonredundant but phylogenetically similar members from three or more sources, it was called a clade; otherwise, it was called a clone or set of clones. The clades were defined as clades A to F, as shown in Table 1
. The five categories of clones are as follows: the previously described AAV3 and AAV5 clones that are clearly distinct from one another but were not detected in our screen; a group of three similar clones from a rhesus macaque that are closely related to AAV4; two similar clones, represented by rh.8, from different rhesus macaques that are not related to previously described AAV serotypes; and a single unique clone from chimpanzees, called ch.5. The previously described AAV1 and AAV6 clones are members of a single clade for which four isolates were recovered from three humans. The clades representing AAV2 and the AAV2-AAV3 hybrid are the most abundant of those found in humans (22 isolates from 12 individuals for AAV2 and 17 isolates from 8 individuals for the AAV2-AAV3 hybrid). A clade containing AAV7 is unique to rhesus and cynomolgus macaques, with 15 members being isolated from 10 different animals. The clade containing AAV8 is interesting because it is found in both human and nonhuman primates: 9 isolates were recovered from 7 humans and 21 isolates were obtained from 9 different nonhuman primates, including rhesus macaques, a baboon, and a pigtail monkey.
The last clade was derived from isolates from three humans and did not contain a previously described serotype. Polyclonal antisera were generated against a representative member of this clade, and a comprehensive study of serologic cross-reactivity between the previously described serotypes was performed (Table 2
). The isolate from this clade was serologically distinct from the other known serotypes and therefore the clade was called the AAV9 clade.
Additional experiments were performed to evaluate the relationship of phylogenetic relatedness to function, as measured by serologic activity and tropism. Polyclonal antisera generated against the nine known serotypes were used to evaluate cross-neutralization (Table 2
). For the purposes of discussion, we defined a new serotype as one for which the neutralization titer by heterologous sera was at least 16-fold less than the neutralization titer against the homologous vector in reciprocal titrations. These data confirmed the phylogenetic groupings of the different clones and clades except for an unanticipated serological reactivity of the structurally distinct AAV5 and AAV1 serotypes (the ratios of heterologous to homologous titers were 1/4 and 1/8, respectively, in reciprocal titrations). It should be noted that the previously described AAV1 and AAV6 serotypes do not segregate by either their phylogeny (Fig. 3
) or their serology (the ratios of heterologous to homologous titers were 1/2 and 1/4, respectively, in reciprocal titrations).
The biological tropisms of AAVs were studied by generating transencapsidated vectors in which recombinant AAV2 genomes expressing either GFP or the secreted reporter gene A1AT were packaged with capsids derived from the various clones or clades. The vectors were evaluated for their transduction efficiency in vitro, based on GFP transduction, and their transduction efficiency in vivo in the liver, muscle or lung. Vectors expressing EGFP were used to examine their in vitro transduction efficiencies in 84-31 cells and to study their serological properties. For in vivo studies, human A1AT was selected as a sensitive and quantitative reporter gene for the vectors and was expressed under the control of the cytomegalovirus-enhanced chicken β-actin promoter. Four to 6-week-old NCR nude mice were treated with novel AAV vectors at a dose of 1011
genome copies per animal through intraportal, intratracheal, and intramuscular injections for liver-, lung-, and muscle-directed gene transfers, respectively. Serum samples were collected at different time points after the gene transfer, and A1AT concentrations were determined by an enzyme-linked immunosorbent assay. A representative set of assay results is shown in Fig. 4
In order to compare unique profiles of transduction, we developed a grading system to characterize the relative transduction efficiency of each in vitro and in vivo model (from 0 [lowest] to 3 [highest]). The cumulative functional difference between two vectors with capsids A and B is the sum of the absolute values of differences between the individual assays as follows: cumulative functional difference = (in vitro A − in vitro B) + (liver A − liver B) + (lung A − lung B) + (muscle A − muscle B). Smaller cumulative functional differences indicate similar profiles with regard to transduction efficiency. Table 3
summarizes the cumulative functional difference scores as well as the % differences in VP1 amino acid sequences in pairwise comparisons.
Unique profiles of biological activity, in terms of the efficiency of gene transfer, were demonstrated for the different clones and clades of AAVs, with substantial concordance between members of a set of clones or a clade (data not shown). This suggests that biological pressures drive the evolution of AAVs.
Our studies point out a number of issues that are relevant to the study of parvoviruses in humans. The prevalence of endogenous AAV sequences in a wide array of human tissues suggests that natural infections with this group of viruses are quite common. The wide tissue distribution of viral sequences and their frequent detection in the liver, spleen, and gut suggest that transmission may occur via the gastrointestinal tract and that viremia may be a feature of the infection. Some earlier reports also documented the detection of AAV sequences in the human female genital tract and suggested that sexual contact could be another route of transmission (8
). However, the clinical consequences of infection with AAV have yet to be delineated.
An inspection of the topology of the phylogenetic analysis revealed insight into the relationship between the evolution of the virus and its host restriction. The entire genus Dependovirus
appears to be derived from avian AAVs, consistent with the work of Lukashov and Goudsmit (15
). After the emergence of AAV4 and AAV5, the family diverged into two monophylic groups (Fig. 3
), with one containing clades that are specific to humans (clades A, B, and C) and the other comprised of a mixture of clades that were isolated exclusively from humans (clade F), exclusively from nonhuman primates (clade D), or from both human and nonhuman primates (clade E).
The presence of latent AAVs that are widely disseminated throughout human and nonhuman primates and their apparent predisposition to recombine and to cross species barriers raise important issues. This combination of events has the potential to lead to the emergence of new infectious agents with modified virulence. Assessments of this potential are confounded by the fact that the clinical sequelae of AAV infections in primates have yet to be defined. In addition, the high prevalence of AAV sequences in the liver may contribute to dissemination of the virus in the human population in the setting of allogeneic and xenogeneic liver transplantation. Finally, the finding of endogenous AAVs in humans has implications for the use of AAV for human gene therapy. The fact that wild-type AAV is so prevalent in primates without ever being associated with a malignancy suggests that it is not particularly oncogenic. In fact, the expression of AAV rep
genes has been shown to suppress transformation (12
). A potential complication of AAV gene therapy, however, could be recombination between the vector and endogenous genomes. This could lead to swapping of the inverted terminal repeats, rearrangement in the transgene cassettes, a loss of regulatory elements for regulated gene expression, and other effects.