Homologous Recombination in E3 Genes of Human Adenovirus Species D

Abstract

Genes within the E3 transcription unit of human adenoviruses modulate host immune responses to infection. A comprehensive genomics and bioinformatics analysis of the E3 transcription unit for 38 viruses within human adenovirus species D (HAdV-D) revealed distinct and surprising patterns of homologous recombination. Homologous recombination was identified in open reading frames for E3 CR1α, CR1β, and CR1γ, similar to that previously observed with genes encoding the three major structural capsid proteins, the penton base, hexon, and fiber.

TEXT

Human adenoviruses (HAdVs) infect a wide array of mucosal surfaces, including the respiratory and gastrointestinal tracts and the eye, and cause infections that vary in severity from mild to life-threatening (1–3). Sixty-nine HAdV genotypes are now formally recognized in GenBank; they have been classified into seven species (A to G), with species HAdV-D containing the most members. HAdV-D includes a substantial number of viruses identified during the first 2 decades of the AIDS epidemic (4). HAdVs are preferred candidate vectors for gene therapy trials because of their broad tissue tropism and ease of manipulation (5). These viruses are nonenveloped with an icosahedral capsid and contain a linear, double-stranded DNA genome of 34 to 36 kb. The recent completion of whole-genome sequencing and analysis for all prototype HAdV genomes (6) has contributed to renewed understanding of the role of homologous recombination in the evolution of HAdVs, particularly species D, but also other HAdV species (7–16). Genes for the three major capsid proteins, the hexon, penton base, and fiber, all readily recombine among viruses within HAdV-D (13, 17–25). However, the evolution of another region of HAdV-D genomes with relative hypervariability, the E3 transcription unit, has not been as well studied. Although E3 is labeled an early transcription region, its transcripts are expressed both early and late during viral infection (26–28). E3 gene products are not required for viral replication in cultured cells (29), but they act to inhibit cellular and cytokine-mediated host immune responses to infected cells (30–32). Specific deletions of E3 have been shown to affect the activity of bioengineered oncolytic HAdV in tumor models (33), and exogenous expression of E3 proteins can prolong xenograft transplant survival (34). These data suggest that the E3 transcription unit is important to viral pathogenesis.

The E3 transcription units in different HAdV species vary in both length and number of open reading frames (ORFs) and are areas of major sequence divergence (35, 36). The DNA sequence of the E3 transcription unit of HAdV-C2 was the first to be determined (37, 38), followed by that of HAdV-C5 (39), and consists of seven ORFs. From the 5′ end of the E3 coding region, these are 12.5K (or gp12.5 kDa), CR1α (also known as 6.7K), 19K (or gp19K), CR1β (11.6K), RIDα (10.4K), RIDβ (14.9K), and 14.7K. The RID proteins were named for their activity in receptor internalization and degradation (40–42), and CR1 proteins were named for conserved region 1 within E3 (43). The E3 transcription unit from HAdV-D contains eight ORFs, including an additional ORF not found in HAdV-C, CR1γ (30K). HAdV-D gene orthologs tend to vary in size relative to HAdV-C (43, 44).

With notable exceptions (25, 43–45), most of what is known about the functions of specific E3 proteins is derived from studies of HAdV-C. For example, HAdV-C2 E3 glycoprotein CR1α directs E3 19K to the endoplasmic reticulum (46). There, it inhibits cell lysis caused by cytotoxic T cells by binding to and retaining major histocompatibility complex (MHC) class I proteins in the endoplasmic reticulum (47), thereby blocking presentation of viral peptides within MHC class I at the cell surface (48–52). E3 19K also sequesters natural killer (NK) cell ligands (53), reducing NK cell recognition of infected cells. The HAdV-C E3 14.7K protein inhibits cell lysis by tumor necrosis factor alpha (TNF-α) and lymphotoxin (54–56). CR1α, RIDα, and RIDβ cooperate to evade TNF-α-related apoptosis through TRAIL (57–60), and the RID complex independently mediates loss of cell surface Fas (58, 61), blocks TNF-α-induced NF-κB activation (62–64), and downregulates the epidermal growth factor receptor (40, 65). HAdV-C CR1β (66), also called the adenovirus death protein, is expressed at greatest abundance late in infection (67) and is required for efficient cell lysis (68) and viral spread (69) in the final phase of viral infection. Notably, the immune evasion functions of E3 gene products are not universal to all cell types (70), and substantial differences may exist between orthologs from E3 regions of different species (45).

The recent availability of high-quality whole-genome sequences with validated annotations for all previously typed viruses within HAdV-D (6) now permits the comparative examination of the HAdV-D E3 genomic region. To form an overview of possible evolutionary relationships between E3 regions within HAdV-D, the full E3 regions of 38 viruses (Table 1) were subjected to bootstrap analysis through the neighbor-joining method, using MEGA (Molecular Evolutionary Genetics Analysis 4.0.2) (71) (Fig. 1). In numerous instances, paired viruses formed subclades on the neighbor-joining tree, for example, HAdV-D types 15 and 53, types 19a (now type 64 [23]) and 37, types 24 and 46, types 9 and 56, and types 33 and 43. These data add to information from prior analyses suggesting homologous recombination in the E3 region between types 19a (type 64) and 37 (23) and between types 9 and 56 (20). The E3 region of HAdV-D53 E3 was previously reported to be most closely related to that of type 8 (22), but at the time of that analysis, many fewer E3 regions had been sequenced, and the full E3 sequence for type 15 had not yet been released. Phylogenetic data also suggest that E3 recombination may in some cases occur for the entire E3 transcription unit, as previously suggested for HAdV-D63, which evolved as the result of a single recombination event between types 29 and 30, encompassing about 25% of their genomes and including the entire E3 and fiber ORFs (24).

Table 1.

GenBank accession numbers for viruses analyzed in this study

Type	GenBank accession no.
HAdV-D8	AB448767
HAdV-D9	AJ854486
HAdV-D10	JN226746
HAdV-D13	JN226747
HAdV-D15	JN226748
HAdV-D19p	JQ326209
HAdV-D19a	EF121005
HAdV-D20	JN226749
HAdV-D22	FJ619037
HAdV-D23	JN226750
HAdV-D24	JN226751
HAdV-D25	JN226752
HAdV-D26	EF153474
HAdV-D27	JN226753
HAdV-D28	FJ824826
HAdV-D29	JN226754
HAdV-D30	JN226755
HAdV-D32	JN226756
HAdV-D33	JN226758
HAdV-D36	GQ384080
HAdV-D37	AB448775
HAdV-D38	JN226759
HAdV-D39	JN226760
HAdV-D42	JN226761
HAdV-D43	JN226762
HAdV-D44	JN226763
HAdV-D45	JN226764
HAdV-D46	AY875648
HAdV-D47	JN226757
HAdV-D48	EF153473
HAdV-D49	DQ393829
HAdV-D51	JN226765
HAdV-D53	FJ169625
HAdV-D54	AB333801
HAdV-D56	HM770721
HAdV-D58	HQ883276
HAdV-D59	JF799911
HAdV-D60	HQ007053

Fig 1.

Bootstrap-confirmed (500 replicates) neighbor-joining phylogenetic tree of whole E3 transcription units from 38 viruses within HAdV-D. Bootstrap values below 80 are indicative of low confidence.

To further parse possible recombination between individual HAdV-D E3 ORFs, proteotype analysis was applied (72). In this method, amino acid alignments for all viruses under consideration are performed using MEGA with the ClustalW option to generate maximum likelihood trees for each putative protein. For visualization in the alignment, the amino acid residues at each position in the proteins of all 38 viruses are color-coded: each amino acid is arbitrarily assigned the same unique color for all occurrences across all the viruses, and then the consensus amino acid at each position in the alignment is changed to white. Gaps in the alignment are colored black. A 10% sequence divergence threshold from consensus is applied to confirm unique proteotypes (6). Those genes/proteins within a specific proteotype are assumed to have recombined with one another for that gene or region, or are all offspring of recombination with another as-yet-uncharacterized (parent) virus. Proteotype analysis was performed for all eight ORFs in the E3 region within HAdV-D. Only one proteotype was identified for each of five proteins: 12.5K, 19K, RIDα, RIDβ (data not shown), and 14.7K (Fig. 2A), consistent with very high sequence conservation for their respective genes. In contrast, CR1α, CR1β (6), and CR1γ each express multiple unique proteotypes (6, 14, and 15 proteotypes, respectively) (Fig. 2B and C), suggesting high levels of recombination in those genes. To better visualize the results of proteotyping, viruses were sorted using the CR1α proteotype as the reference and then numbered (and colored) according to proteotype (Fig. 3). Remarkably, 37 of 38 CR1α protein sequences were shared between at least two viruses, with only one unshared proteotype, that from HAdV-D13. Therefore, all but 1 of the 38 HAdV-D viruses analyzed appeared to be recombinant for the CR1α gene. Seventeen of 38 viruses appeared to be the product (or source) of recombination for the E3 region as a single unit (∼5,000 bp), and 10 viruses (5 pairs) showed evidence for a single recombination spanning the entire E3 region and adjacent fiber gene (∼6,200 bp) (Fig. 3). Based on prior analysis (6), the genomic region with the next highest degree of homologous recombination after CR1α is the penton base RGD loop, with 10 proteotypes among 38 viruses and only 3 unshared. In contrast, the hexon protein, which contains the serum neutralization determinant, was shown to have 28 proteotypes among 38 viruses.

Fig 2.

Proteotyping analysis for select open reading frames of the E3 transcription unit of HAdV-D: (A) E3 14.7K; (B) CR1α; (C) CR1γ. Proteotyping analysis of CR1β was previously published (6). Maximum likelihood phylogenetic trees are shown to the left for each putative protein, and amino acid signatures are on the right. The scale bar at the bottom left of each panel denotes the phylogenetic distance reflected in the horizontal dimension of the corresponding tree. To construct the amino acid signatures shown, each amino acid was assigned a unique color (see the legend in the upper right corner), consensus amino acids at each position across all 38 viruses were assigned white, and gaps in the alignment are shown in black.

Fig 3.

Assortment analysis of putative proteotypes for hexon, E3 CR1, and fiber open reading frames, organized according to the putative CR1α protein. Each unique proteotype was assigned a number and color for accounting and visual representation, respectively. The seven vertical red bars on the left side of the figure denote those virus groups that demonstrate homologous recombination across the entire E3 transcription unit. The five blue bars on the right side of the figure identify those virus groups with evidence for recombination inclusive of all E3 and adjacent fiber genes. Inclusion of a protein in a given proteotype required an amino acid content that was <10% different from other members of the proteotype. The number recombined reflects the total number of viruses within any proteotype with at least two members.

When considered together, phylogenetic and proteotype analyses provide strong evidence for E3 recombination among viruses within HAdV-D. To further explore this, genome sequence recombination analysis using SimPlot (http://sray.med.som.jhmi.edu/SCRoftware/simplot/) (73) was performed for the entire E3 transcription unit along with 800 additional nucleotides on either end of E3, which included the pVIII and (partial) fiber genes on the 5′ and 3′ ends of the E3 region, respectively (representative examples are shown in Fig. 4). This software compares multiple reference genomic sequences to a query sequence to allow visualization of possible recombined regions on the basis of high nucleotide similarity between two or more sequences. The SimPlot for HAdV-D23 (Fig. 4A) demonstrated relatively high similarity with multiple other viruses for CR1α but a unique sequence for CR1β and -γ. These data are consistent with proteotype analysis that showed the CR1α amino acid sequence of HAdV-D23 was shared among 12 viruses (Fig. 3). SimPlot analysis for HAdV-D24 demonstrated high similarity with type 46 across all three CR1 ORFs and the adjacent fiber gene (Fig. 4B), consistent with phylogenetic and proteotype analyses (Fig. 1 and 3). HAdV-D27 (Fig. 4C) appeared to be recombinant in CR1α (multiple) and possibly CR1β (with type 28), but not in CR1γ. HAdV-D36 (Fig. 4D), which has been associated with obesity (74), showed evidence for recombination in CR1α (with multiple viruses) and CR1γ (with type 60 and possibly type 19p), but not CR1β. HAdV-D42 (Fig. 4E) demonstrated recombination with type 15 across the entire E3 transcription unit and the adjacent fiber gene and also recombination with type 53 for the entire E3 region but not the fiber gene. Notably, types 42, 15, and 53 all appeared closely related by phylogenetic analysis (Fig. 1). SimPlot analysis results for HAdV-D44 (Fig. 4F) also were consistent with the phylogenetic and proteotype analyses, demonstrating recombination with type 48 for CR1β, CR1γ, and the fiber gene.

Fig 4.

SimPlot analysis of the E3 transcription unit, including 800 nucleotides on either side, for 38 viruses within HAdV-D. (A) HAdV-D23; (B) HAdV-D24; (C) HAdV-D27; (D) HAdV-D36; (E) HAdV-D42; (F) HAdV-D44. A window size of 200 bp and step size of 20 bp were used for the analysis. The color code for each virus in the analysis is shown on the right margin. Arrows point to examples of homologous recombination.

These analyses of the E3 transcription unit from 38 fully sequenced and annotated viruses within HAdV-D represent powerful evidence that evolution of the CR1 ORFs is driven by homologous recombination, and the results provide new insights into HAdV-D evolution. The HAdV-D E3 CR1 genes are among the most highly recombinant in the entire HAdV-D genome. Given that the E3 transcription unit of HAdV-C is not essential to viral replication in vitro (29), yet individual components can facilitate immune evasion by infected cells, these data suggest that evolution of E3 CR1 genes in HAdV-D may be driven by host factors that encourage the evolution of unique proteotypes for these genes. Considering that E3 genes do not encode proteins known to elicit neutralizing antibodies, the selection pressure imposed on this region likely occurs at the cellular level. Studies of E3 CR1 functions have focused almost entirely on HAdV-C, in which CR1α contributes to inhibition of TNF-α-induced apoptosis (57) and CR1β appears to assist in cell lysis at the final stages of infection (68). Not as much is known about the function of CR1 gene products from HAdV-D. It is noteworthy that the CR1β gene of HAdV-D codes for a much larger protein (49K) than in HAdV-C (11.6K) (25, 44). Also, CR1γ, which is not present in HAdV-C, has no known function (75). A stereotypical flux in GC nucleotide content was recently proposed as a potential marker for homologous recombination sites in HAdV-D (6), including the E3 region, but the exact mechanism for homologous recombination in adenoviruses has yet to be determined.

Finally, recognition of promiscuous recombination among CR1 genes in HAdV-D E3 should be of interest to those investigators using HAdV-D for development of viral gene vectors (76–79), particularly when the transgene is placed within the E3 cassette (77, 80–84). HAdV-D-based gene vectors with a transgene within the E3 region might be susceptible to recombination into wild-type HAdV-D during a coincidental coinfection, resulting in an infectious, replication-competent virus with potential expression of the transgene.