Non-structural genes of novel lemur adenoviruses reveal codivergence of virus and host

Abstract

Adenoviruses (AdVs) are important human and animal pathogens and are frequently used as vectors for gene therapy and vaccine delivery. Surprisingly, there are only scant data regarding primate AdV origin and evolution, especially in the most basal primate hosts. We detect and sequence AdVs from faeces of two Madagascan lemur species. Complete genome sequence analyses define a new AdV species with a particularly large gene encoding a protein of unknown function in the early gene region 3. Unexpectedly, the new AdV species is not most similar to human or other simian AdVs but to bat adenovirus C. Genome characterisation shows signals of virus–host codivergence in non-structural genes, which show lower diversity than structural genes. Outside a lemur species mixing zone, recombination less frequently separates structural genes, as in human adenovirus C. The evolutionary history of lemur AdVs likely involves both a host switch and codivergence with the lemur hosts.

Introduction

Adenoviridae is a family of non-enveloped viruses with linear, double-stranded DNA genomes, which cause mostly mild, respiratory and gastrointestinal disease in humans but severe disease in immunosuppressed individuals (Khanal, Ghimire, and Dhamoon 2018). Human and chimpanzee adenoviruses (AdVs) have proven useful as gene therapy and vaccine vectors (Bulcha et al. 2021; Knoll and Wonodi 2021). The family comprises six genera—Atadenovirus, Aviadenovirus, Ichtadenovirus, Mastadenovirus, Siadenovirus, and Testadenovirus—and currently over eighty species, as defined by the International Committee on Taxonomy of Viruses (ICTV) (Benkő et al. 2022). AdVs have only been identified in vertebrates and are generally species-specific (Benkő et al. 2022). All currently described primate AdVs are closely related and belong to the genus Mastadenovirus (Benkő et al. 2022). Primates can be divided into simians—including humans, apes, and old- and new-world monkeys—and prosimians—including lemurs. Among primate AdVs, seven species (human AdVs (HAdVs) A–G) infect both humans and simians, and nine (simian AdVs A–I) have been found only in old-world monkeys and apes (Benkő et al. 2022). To date, only two complete genomes of new-world monkey AdVs have been published (Chen et al. 2011; Rogers et al. 2020), and no AdV of any prosimian host has been sequenced completely.

AdVs in mouse lemurs in Madagascar

Lemurs, small prosimian primates endemic to Madagascar, belong to the Strepsirrhini, the most ancient extant primate suborder. Strepsirrhini are estimated to have diverged from other primates around 74 million years ago (mya) (Pozzi et al. 2014). Lemurs colonised Madagascar 60–50 mya (Poux et al. 2005; Masters et al. 2021) where they underwent adaptive radiation and allopatric speciation and offer a useful model for evolutionary studies, including the study of virus evolution. Here, we focus on the lemur species Microcebus (M.) griseorufus and M. murinus in southern Madagascar. Following the phylogenetic analyses performed by Poelstra et al. (2022), we include the recently defined species M. ganzhorni in M. murinus. M. griseorufus and M. murinus are estimated to have diverged around 600,000 years ago, assuming no migration (Poelstra et al. 2022).

AdVs have recently been identified in various lemur species by polymerase chain reaction (PCR) and amplicon sequencing (Zohdy et al. 2015; Podgorski et al. 2018; Wasimuddin et al. 2019). In a PCR screening of faeces of lemurs including M. rufus, AdVs were detected in 19 of 77 (25%) samples (Zohdy et al. 2015). In a PCR screening of faeces from a range of prosimian hosts, AdVs were detected in 10 per cent of samples (Podgorski et al. 2018), and partial sequences were generated. Although there are no data regarding AdV pathogenicity in lemurs, AdVs have been linked to an altered lemur gut microbiome (Wasimuddin et al. 2019).

How do AdVs evolve?

Evolution of novel AdV species

Based on phylogenetic trees of AdV genera and mastadenoviruses, AdVs are hypothesised to have coevolved with their animal hosts. Most AdVs are found in reptiles, birds, and mammals. Only one fish and one frog AdV have been found despite intensive screening of fish and amphibians (Harrach, Tarján, and Benkő 2019). Therefore, it has been suggested that AdVs developed in or invaded the vertebrates during the separation of amniotes from amphibians (Harrach, Tarján, and Benkő 2019). Primate AdV phylogeny also supports a hypothesis of coevolution, as the known primate AdVs are monophyletic and their phylogeny roughly mirrors that of the primate phylogenetic tree (Harrach, Tarján, and Benkő 2019). However, most of these analyses are based on simian AdVs only. The only analysis including prosimian AdVs, published in 2018, presents 252–base pair (bp) prosimian AdV sequences that fall basal to simian AdVs in phylogenetic analyses (Podgorski et al. 2018). No full phylogenetic analysis of prosimian AdVs has been conducted, due to the lack of a complete genome.

Despite a background pattern of codivergence and the observation that AdVs are generally species-specific (Benkő et al. 2022), host switching has occurred many times, and our interpretation of extant patterns must take this into account. It is hypothesised that the single fish and amphibian AdVs resulted from host switches (Harrach, Tarján, and Benkő 2019). The skunk AdV A is closely related to, and likely originates from, a bat AdV (BtAdV) and was also found in African pygmy hedgehogs (Kozak et al. 2015; Madarame et al. 2019; Needle et al. 2019; Ochiai et al. 2020). Canine AdVs may have similarly originated from BtAdVs (Kohl et al. 2012). Within the mastadenoviruses, a review found sixteen studies showing AdV transmission between humans and other animals, most often between non-human primates and humans, and thirteen studies of cross-species transmission between animals other than humans (Borkenhagen et al. 2019). One notable instance was an outbreak of the titi-monkey AdV (TmAdV) in a primate research centre, where TmAdV infected a researcher who then infected a family member (Chen et al. 2011). Apart from the fact that there is limited data, the combination of codivergence, cross-species transmission, and even that the taxonomic unit of the viral host is not categorically limited to one species, make it difficult to draw conclusions regarding adenovirus evolution.

Molecular mechanisms of AdV evolution

The molecular mechanisms of AdV evolution have mostly been studied in HAdVs. AdVs evolve through single nucleotide mutations, deletions, insertions, duplications, and recombination (Ma and Mathews 1996; Szpara and Van Doorslaer 2021). The HAdV-5 mutation rate has been estimated at 1.31E−07 per base per cell infection cycle (Risso-Ballester, Cuevas, and Sanjuán 2016). Substitution rates have been estimated to be 7.20E−05 substitutions per site per year (s/s/y) for HAdV-B and 3.46E−05 s/s/y for HAdV-C (Firth et al. 2010; Peck and Lauring 2018) in an analysis of modern AdVs, and 2.52E−10 s/s/y in an analysis of two ancient HAdV-C genomes (Nielsen et al. 2021), which is similar to the human rate of 4.18E−10 s/s/y (Bergeron et al. 2023). In HAdV-C, non-structural proteins (polymerase and early gene 1A (E1A)) are more conserved than structural proteins (hexon and fiber) (Lukashev et al. 2008).

Inter- and intraspecies recombination occurs frequently in AdVs (Rainbow and Castillo 1992; Ebner, Pinsker, and Lion 2005; Lukashev et al. 2008). Recombination is well researched in HAdVs, but we are not aware of any study investigating AdV recombination in a non-human host. AdV recombination can confer new phenotypic properties: a recombinant HAdV-D was identified in an outbreak of keratoconjunctivitis, where the recombined partial hexon resulted in a new neutralisation profile and likely corneal tropism (Walsh et al. 2009). Within the diverse HAdV-D, which contains 78 out of 113 currently recognised HAdV types (HAdV Working Group, 2022), recombination frequently separates three genes encoding structural proteins: penton base, hexon, and fiber (Robinson et al. 2009, 2013; Walsh et al. 2009; Singh et al. 2013; Gonzalez et al. 2014). Because novel AdV types are defined based on the combination of these three genes, this results in frequent naming of novel types. HAdV-D recombination also frequently occurs in the E3 and E4 regions (Robinson et al. 2009, 2013; Walsh et al. 2009; Singh et al. 2013; Gonzalez et al. 2014). A contrasting pattern occurs in HAdV-C, where recombination breakpoints rarely separate the hexon and fiber genes (Lukashev et al. 2008; Dhingra et al. 2019; Mao et al. 2019; Ji et al. 2021); therefore, new types are infrequently defined, with currently only eight HAdV-C types designated. Only one study reports high recombinational activity in HAdV-C within the hexon and fiber genes (Rivailler et al. 2019). Instead, there is frequent recombination in non-structural gene regions such as E1A, E1B, DNA polymerase, E3, and E4 (Lukashev et al. 2008; Dhingra et al. 2019; Mao et al. 2019; Ji et al. 2021). Recombination separating hexon and fiber genes may be selected against, as both proteins need to interact with the penton base to produce a viable capsid (Lukashev et al. 2008). It is not clear why this evolutionary pressure could be stronger for HAdV-C than HAdV-D. Fewer recombination breakpoints in structural than non-structural genes have also been observed in single-stranded DNA viruses (Lefeuvre et al. 2009).

Here, we describe five complete lemur AdV (LAdV) genomes. We show that these define a novel mastadenovirus species and analyse LAdVs in the context of mastadenovirus phylogeny, suggesting a host switch. We analyse the phylogeny and recombination of LAdVs in detail based on partial gene sequences of sixty samples, from which we draw conclusions about LAdV evolution and suggest codivergence of LAdVs with M. griseorufus and M. murinus hosts.

Results

New AdV species in lemurs, according to ICTV criteria

Faecal samples from wild M. griseorufus and M. murinus living at three sites in southern Madagascar were collected between 2013 and 2017. Of 297 samples, 60 (20.2%) were positive for LAdVs (Fig. 1). Using high-throughput sequencing (HTS) and Sanger sequencing, we produced complete viral genomes for five LAdVs (IDs 204, 354, 400, 814, and 2019).

Figure 1.

The topographical map of sampling sites. Dark green refers to heights of 350–500 m, yellow to 550–650 m, red to 750–1,100 m, and grey to >1,100 m, with rivers marked in blue. Microcebus griseorufus is found at sites A and B, shown in dark red, and M. murinus at sites B and C, shown in light red. The number of AdV-positive samples at each site is given for each lemur species.

These are the first complete AdV genomes from a prosimian host. We therefore analysed whether LAdVs represent a novel mastadenovirus species according to the nine ICTV species demarcation characteristics (Benkő et al. 2022): phylogenetic distance (particularly in the DNA polymerase amino acid (aa) sequence), nucleotide composition, ability to recombine, number of viral-associated (VA) genes, genome organisation (especially in the E3 region), host range, oncogenicity in rodents, cross-neutralisation, and haemagglutination. For LAdVs, we only have information regarding the first five criteria.

In order for a newly-discovered mastadenovirus to qualify as a new species, the ICTV specifies a minimum aa difference of 10–15 per cent between the DNA polymerase of the novel sequence and the closest defined AdV species. The LAdV DNA polymerase aa sequences are 35.5–36.2 per cent distant from the most similar DNA polymerase sequence, bat AdV C (BtAdV-C) Rc-kw20 (Katayama et al. 2022), in pairwise comparison, greatly exceeding the ICTV cut-off (Table 1). At 51.5–52.5 per cent, the LAdV GC content is within the range of mastadenovirus GC contents (Table 1). Analyses of the five complete LAdV genomes using RDP5 (distance analysis) (Martin et al. 2021) and DualBrothers (recombination analysis) (Suchard et al. 2002) both indicate the ability to recombine, identifying similar recombination breakpoints on both sides of the E3 region and the fiber gene, and before or within the penton base (Figs 2A, S1). DualBrothers analysis also identified a recombination breakpoint within the E1 region (Fig. S1B). Overall genome organisation of LAdVs is similar to that of other mastadenoviruses, including HAdV-1 (Fig. 2A), with differences only in the number of VA genes and the E3 and E4 regions. Previously described primate AdVs have one or two VA genes (Benkő et al. 2022), which produce non-coding RNAs that interfere with host cell processes, promoting viral replication (Vachon and Conn 2016). LAdVs however lack VA genes (Table 1). In the E4 region, LAdVs have ten or eleven predicted open reading frames (ORFs), more than other mastadenoviruses (Fig. 2A, Table 2). The LAdV E3 region is particularly long (4,809–5,109 bp) and holds only two ORFs (Fig. 2B): the conserved 12.5k and a long (4,431–4,731 bp) unknown ORF making up most of the E3 region. Other primate AdVs have shorter E3 regions (1,700–4,500 bp) with more ORFs (five to nine) (Davison, Benkő, and Harrach 2003). The LAdV E3 organisation however resembles that of BtAdV-C (Tan et al. 2016; Ai et al. 2022; Katayama et al. 2022). Both have relatively large E3 regions with few ORFs, one of which is particularly long (Fig. 2B, Table 2). According to the ICTV characteristics analysed, LAdVs differ even from other primate adenoviruses and therefore likely represent a new adenovirus species.

Table 1.

Summary table of the five complete LAdV genomes, IDs indicated. % GC: GC content of the entire AdV sequence, calculated in Geneious; Pol. distance: DNA polymerase amino acid (aa) distance to the most closely related AdV DNA polymerase aa sequence (Rc-kw20, species bat mastadenovirus C, which had the highest per cent identity in aBLASTp search against the non-redundant protein database), calculated in Geneious; VA RNA genes: number of VA RNA genes, determined in Geneious. ICTV mastadenovirus species demarcation criteria are noted where applicable.

Sample ID		204	354	400	814	2019
Host species		M. murinus	M. murinus	M. griseorufus	M. murinus	M. griseorufus
Sampling location		Mandena	Mandena	Tsimanampetsotsa	Mandena	Mangatsiaka
Pol. distance, % (ICTV: >10-15)		36.4	36.7	36.4	36.6	36.3
% GC		51.7	52.5	51.5	52.5	51.8
VA RNA genes (ICTV: 0-2)		0	0	0	0	0
E3l protein Ig-like domains		aa 340–739 Interleukin-1 receptor family	aa 1,209–1,296 Zig-8 family	aa 1,173–1,284 Basigin-like family	aa 1,173–1,284 Basigin-like family	Related to immunoglobulin Fc receptor

Figure 2.

Genome organisation of LAdV ID 204 compared to (A) HAdV-1 along the complete genome and (B) compared to HAdV-1, TsAdV, and BtAdV-C in the organisation of the E3 region. The reference position is relative to HAdV-1. ‘x’ in LAdV denotes detected recombination breakpoints (Fig. S1). Genes in bold were used for PCR design for amplicon analysis (Figs 5–8).

Table 2.

Summary E3 and E4 region gene content.

	LAdV	BtAdV-C	HAdV-1	TsAdV
E3 region	2 ORFs (12.5k and e3l) 4,809–5,109 bp	3 ORFs (12.5k, e3l, and e3s) 4,446–5,361 bp	7 ORFs ∼3,000 bp	3 ORFs ∼1,300 bp
E4 region	10–11 ORFs	7 ORFs	6 ORFs	8 ORFs

What could be the function of LAdV E3l?

Analogous to the naming of the large E3 ORF in BtAdV-C (Tan et al. 2016), we call the LAdV large E3 ORF e3l. No protein similar to LAdV E3l was found in a BLASTp search of the National Center for Biotechnology Information (NCBI) non-redundant protein database. Although similar in E3 region organisation, LAdV E3l and BtAdV-C E3l only have 10–11 per cent aa identity. LAdV e3l are diverse withan aa identity of 43–100 per cent among themselves. AdV E3 region proteins have been described as having immune antagonism functions (Fessler, Delgado-Lopez, and Horwitz 2004), so we searched for known sequence motifs in LAdV E3l. In all five LAdV E3l proteins, we detected five to twelve transmembrane (TM) and five to seven immunoglobulin-like (Ig-like) domains (Table 1). Ig-like motifs mediate adhesive interactions, and combinations of Ig-like and TM motifs are found in virus immune modulator proteins (Tan et al. 2020). We predicted the protein structure of LAdV ID 400 E3l using AlphaFold (Jumper et al. 2021; Varadi et al. 2022) and compared it to structures in the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB). Nectin-1, an Ig-like cell adhesion molecule (Takahashi et al. 1999), was the structurally most similar protein to E3l (Z-score 11.4), with the caveat that only the Ig-fold-1 of nectin-1 overlaps with a small part of E3l. Nectin-1 is a receptor of herpes simplex virus 1, binding the virus glycoprotein with its Ig-fold-1 (Di Giovine et al. 2011; Zhang et al. 2011). Nectins also have a well-established role in regulating natural killer cells (NKC) (Martinet and Smyth 2015). Nectin-1 binds the NK regulatory receptor CD96 at its Ig-fold-1, likely activating NKC, and dimerises with higher affinity (Harrison et al. 2012; Holmes et al. 2019). A variety of mechanisms can be proposed for E3l to antagonise NKC function, including competitive binding and decoy effects. These analyses suggest a potential role for LAdV E3l in immune antagonism.

LAdVs are not immediately basal to human and simian AdVs

We performed phylogenetic analyses of the five complete LAdV genomes in the context of other mastadenovirus species. LAdVs are monophyletic, with bootstrap support (BS) of 100 per cent, even in otherwise weaklysupported phylogenetic trees of the hexon, DNA polymerase, and IVa2 aa sequences (Figs 3A, S2). Contrary to expectation based on the primate host phylogenetic tree (Fig. 3B), LAdVs are not monophyletic with other primate AdVs as they do not fall immediately basal to other primate AdVs (Fig. 3A). LAdVs share a most recent common ancestor with primate AdVs, BtAdV-C, porcine AdV A (PAdV-A), and murine AdV B (MuAdV-B). We additionally included short LAdV gene fragments of IVa2 (Podgorski et al. 2018) into phylogenetic analysis (Fig. 3C). The samples were from lemur species other than Microcebus spp.: Eulemur spp., Hapalemur griseus, Lemur catta, and Varecia variegata. Low BS values prevent conclusions regarding relatedness (Fig. 3C). However, a strongly-supported phylogenetic clade including non-lemur primate AdVs, PAdV-A, and MuAdV-B emphasises that LAdVs do not fall immediately basal to other primate AdVs.

Figure 3.

Maximum likelihood (ML) phylogenetic trees created using raxml-ng (Kozlov et al. 2019) with the substitution model GTR + I + G. (A) Complete genome nucleotide sequences. (B) A sketch of a simplified primate phylogenetic tree adapted from Yoder and Yang (2004), wherein the category of old-world monkeys excludes apes and humans. (C) Eleven published 252-bp IVa2 gene fragments from lemur AdVs are included in the phylogenetic tree. Branch BS between 70 and 90 is shown as a white circle and ≥90 as a black circle. All trees were rooted with aviadenoviruses and atadenoviruses (not shown).

Our initial expectation was that LAdVs would be most similar to other primate AdVs, but they are in fact more similar to BtAdV-C. A BLASTn search against the NCBI non-redundant nucleotide database finds the LAdV complete genomes to be most similar to BtAdV-C WIV10, with 51–52 per cent pairwise nucleotide identity calculated in Geneious. A distance plot (Fig. 4) comparing LAdV ID 400 to BtAdV-C WIV10, tree shrew AdV (TsAdV), TmAdV, and HAdV-1 confirms that overall, the LAdV ID 400 sequence is most similar to BtAdV-C WIV10. The patristic distance in maximum likelihood (ML) analyses is smallest between LAdVs and BtAdV-C (Fig. 3A). Previously published short LAdV fragments (Podgorski et al. 2018) have the lowest patristic distance to BtAdV-C and LAdVs (Fig. 3C). LAdV E3 region organisation also resembles that of BtAdV-C. In phylogenetic analyses, however, LAdVs are equally closely related to BtAdV-C and primate AdVs.

Figure 4.

Nucleotide distances along a complete genome alignment, comparing LAdV ID 400 as representative against BtAdV-C WIV10, TsAdV, TmAdV, and HAdV-1. The analysis was carried out with a sliding window size of 500 nucleotides and a step size of 150 nucleotides. The background colour represents the AdV most similar to LAdV ID 400 in each window.

Genome characterisation shows signals of virus–host codivergence

To analyse evolution and diversity within LAdVs, and because we have evidence of recombination, we designed PCR assays for six regions between the recombination breakpoints identified (Fig. S1), for three structural (hexon, penton base, and fiber) and three non-structural proteins (E1A, DNA polymerase, and E4 34K). We performed PCR and Sanger-sequenced the partial gene sequences. Diversity of LAdV structural genes, similar to HAdV-C type structural gene diversity (Table 3A), suggests that LAdV consists of several types (Table 3B). We went on to investigate LAdV evolution within the three lemur populations sampled in this study (Fig. 1): the western sites, Tsimanampetsotsa and Miarintsoa in close vicinity, are inhabited by M. griseorufus. The easternmost site, Mandena, is inhabited by M. murinus. Mangatsiaka, the geographical mixing zone located ∼50 kilometres (km) west of Mandena, is home to a mixed population of both M. murinus and M. griseorufus. We analysed LAdV phylogeny of the six genes of sixty LAdV-positive samples. We defined seven LAdV clades (I–VII) based on the hexon gene phylogenetic tree (Figs S3B, S4B) and named clades in the other five trees according to LAdV ID clade majority (Figs S3, S4).

Table 3.

(A) HAdV-C-type nucleotide distance matrix of concatenated and aligned penton–hexon–fiber sequence fragments from the regions where sequence data are available for LAdVs. (B) LAdV concatenated and aligned penton–hexon–fiber sequence distance matrix.

(A)	HAdV-C108	HAdV-C2	HAdV-C89	HAdV-C104	HAdV-C1	HAdV-C57	HAdV-C6
HAdV-C2	99.0
HAdV-C89	97.6	97.1
HAdV-C104	91.0	90.0	89.2
HAdV-C1	86.9	86.0	85.1	95.7
HAdV-C57	86.2	85.4	84.3	86.2	87.5
HAdV-C6	85.7	85.1	84.3	85.9	87.1	92.2
HAdV-C5	83.7	84.2	83.1	83.5	83.8	84.7	84.3
(B)	LAdV 204	LAdV 400	LAdV 620	LAdV 610	LAdV 2019	LAdV 63	LAdV 605	LAdV 616	LAdV 2008	LAdV 2016	LAdV 354	LAdV 858
LAdV 400	92.1
LAdV 620	92.2	99.9
LAdV 610	85.2	90.3	90.2
LAdV 2019	71.4	71.7	71.7	81
LAdV 63	69.3	70.1	70.1	79.5	90.8
LAdV 605	69.2	70	70	79.5	90.8	99.8
LAdV 616	68.7	69.5	69.4	78.9	90.1	99.2	99.3
LAdV 2008	70.9	71.3	71.2	77.4	87.9	86	86	85.5
LAdV 2016	70.6	71.1	71.1	77.2	89.2	87.1	87.1	87.6	97.3
LAdV 354	70.2	70.3	70.3	75.6	79.2	77.5	77.5	76.9	85.5	85.4
LAdV 858	69.8	70	70.1	75.4	79.1	77.3	77.3	76.8	85.6	85.4	99.6
LAdV 76	68.3	69.3	69.3	75.7	79.2	87.3	87.3	86.7	76.3	76.2	86.2	86

Phylogenetic analysis reveals codivergence signals. All clades in all trees were specific for samples from either the eastern sites or the western sites (Figs S3, S4). This agrees with reports of the rarity of long-distance dispersal of mouse lemurs (Schliehe-Diecks, Eberle, and Kappeler 2012). We find markedly different evolutionary histories for non-structural and structural genes (Fig. 5A vs 5B, Fig. S4A vs S4B). Phylogenetic trees of non-structural genes resemble each other (Figs 5A, S4A), and show a clear binary split between LAdVs of M. griseorufus and of M. murinus, supported by at least 99 per cent BS in all three analyses. This split is by host species and not by sampling site, as shown by LAdVs of M. griseorufus in the East clustering with LAdVs of M. griseorufus in the West, not the M. murinus LAdVs in their vicinity. This suggests that non-structural genes diverged together with the lemur host species. Phylogenetic trees of structural genes also resemble each other but differ from those of non-structural genes (Figs 5, S4). There is no binary lemur species split in phylogenetic trees of structural proteins, rather clades from the East or West are often sister clades (Figs 5B, S4B). Clades VI and VII, from the East and West, respectively, are sister clades in all three structural protein trees (BS ≥ 97%). In the hexon tree, Clades IV and V are sister clades (BS = 98%) and the remaining Clades I, II, and III cluster together. This would accord with codivergence from a starting population of three ancestral mouse lemur hexon sequences, with additional subsequent evolutionary mechanisms. We thus see a clear signal for codivergence of LAdVs with the two mouse lemur species in the phylogenetic trees of non-structural genes, but not in the trees of structural genes (Fig. 5).

Figure 5.

Tanglegram sketches based on ML phylogenetic trees. (A) Non-structural genes, DNA polymerase as representative (B) Structural genes, hexon as representative. The M. griseorufus and M. murinus tree sketch is based on Poelstra et al. (2022). The complete trees and all tree sketches can be found in Figs S1 and S2. Dark red corresponds to M. griseorufus and light red to M. murinus. Only BS values > 70 are shown.

Non-structural genes are less diverse than structural genes

As a rough estimate of gene diversity in non-structural and structural genes, we measured root-to-tip distances for the phylogenetic trees of the six genes (Fig. S3). We compared the average root-to-tip distances for each gene (Fig. 6A) and the average distances for non-structural or structural genes (Fig. 6B). Overall, the DNA polymerase showed the lowest, and the hexon the highest root-to-tip distances of the partial gene sequences analysed. Non-structural gene root-to-tip distances (average 0.05) are smaller (two-sided t-test, P = 4 x 10⁻⁸³) than structural gene root-to-tip distances (average 0.41) by a factor of 8.2.

Figure 6.

Root-to-tip patristic distances of ML trees for (A) individual gene fragments and (B) non-structural and structural genes. Error bars represent 1 standard deviation. A two-sided t-test with Bonferroni correction was performed, and **** denotes p-values < 0.0001.

Signals of recombination are common, but breakpoints separating structural genes are rare

Another factor that might contribute to the differential evolution of non-structural and structural genes is recombination, and we thus aimed to analyse recent recombination in LAdVs. The prerequisite of recombination is a double infection, but double infections can be mistaken for recombination. Therefore, we checked our sequencing data for corresponding signals, such as double peaks in Sanger sequencing. We rarely found signals for double-infected individuals (4% of samples) in our dataset. A potential double infection in a small percentage of samples might still influence recombination analysis, as non-overlapping PCRs may favour different viruses occurring in a double infection. This is a clear limitation for our recombination analysis. We define any incongruities in the clade classification of the six gene fragment sequences of one LAdV sample as an indication of recent recombination. For fifty-five out of the sixty LAdVs (91.7%), we recovered gene sequence information for at least two genes and were able to include these into our analysis of recombination. Twenty-four of the fifty-five LAdVs (43.6%) show a pattern of recent recombination (Fig. 7, Table S4). All detected recent recombination patterns are of LAdVs changing clades within either eastern or western clades (Fig. 8A, B). This suggests that we are looking at recent recombination, assuming allopatric host speciation without migration. We detect patterns of recombination between LAdVs of different lemur species in Mangatsiaka and therefore suggest that LAdV recombination is not species-specific. In our analysis, the detected recombinational breakpoints less frequently separate structural genes (Fig. 8D). Only 8 per cent (western clades) and 0 per cent (eastern clades, excluding Clade III) of recombination breakpoints were detected in regions between structural genes. Recombination breakpoints were more frequently found in regions between non-structural genes or regions between non-structural and structural genes.

Figure 7.

The heatmap of clade assignment of six PCR sequence fragments for each LAdV, grouped by (A) non-recombinant (n = 31) and (B) recombinant (n = 24) LAdVs. The corresponding data can be found in Table S4, and these data were taken from the phylogenetic trees shown in Fig. S3. Five LAdVs were excluded from recombination analysis because we obtained only the sequence of the hexon gene. Wherever we were not able to generate PCR product and sequence data, the field is left empty, as the sequence is not available (n.a.).

Figure 8.

(A) The map of sampling sites showing clade location. Western sites are in dark blue, eastern sites are in bright blue, and Clade III is in turquoise. (B) The boxes symbolise the seven clades. A line between them stands for the detection of patterns, suggesting recombination involving the two clades. (C) Average patristic nucleotide distances of each clade to the phylogenetically closest clade in geographic vicinity, separated by structural (s) and non-structural (non-s) genes and grouped into clades non-III and III. For Clade III, the most closely related clades are II (penton), II (hexon), V (fiber), VI (E1A), V (DNA pol), and VI (E4 34k). A two-sided t-test with Bonferroni correction was performed, ns denotes p-values > 0.05, and ** denotes p-values < 0.01. (D) The schematic of LAdV genome. Each connecting line indicates the location of one detected recombination breakpoint (Fig. 7 and Table S4). The potential recombination breakpoint lies anywhere between the two connected genes.

The cross-lemur-species recombination pattern in the geographical mixing zone has a distinct pattern

Despite limited sequence data from the geographical mixing zone Mangatsiaka, there are two points of note. First, the identified recombination breakpoints in the geographical mixing zone, between Clades I and III, occur between structural genes in 75 per cent of observed cases (Fig. 8D). This cross-lemur-species LAdV recombination pattern differs from the within-lemur-species LAdV recombination seen in western and other eastern clades. Because recombination is based on homology, we analysed the non-structural and structural gene genetic distance of eastern clades, western clades, and Clade III to the closest LAdV from the same area (Fig. 8C). Eastern and western LAdVs encounter LAdVs with very different structural and very similar non-structural genes, consistent with recombination rarely separating structural genes. Clade III LAdVs however encounter LAdVs with less similar non-structural genes (Fig. 8C). Second, recombination between Clades I and III in the geographical mixing zone exchanged exclusively structural genes. All LAdVs found in M. griseorufus in Mangatsiaka have Clade III non-structural as well as penton and fiber genes, but some have Clade I hexon genes (Fig. 7, Table S4). LAdVs of M. murinus in Mangatsiaka have Clade I non-structural or fiber genes, but penton and hexon genes from Clade III were found in M. murinus in Mangatsiaka (Fig. 7, Table S4).

Discussion

LAdVs were detected in 20.2 per cent of mouse lemurr faecal samples, comparable to the 25 per cent positivity rate of a study screening AdVs in lemur faecal samples (Zohdy et al. 2015). Illumina and Oxford Nanopore HTS enabled us to identify LAdVs, which we propose as a novel mastadenovirus species including at least three types. LAdVs have a DNA polymerase over 36 per cent different from other known DNA polymerase aa sequences, expected GC content, the ability to recombine, and no VA genes. LAdVs also have an exceptionally long E3 region, comparable in its organisation to the BtAdV-C E3 region, with a large unknown putative protein with potential immune antagonism function. This qualifies LAdVs as a novel mastadenovirus species according to ICTV criteria (Benkő et al. 2022). As the primate AdV most distant from HAdVs, LAdVs may be an interesting addition to available adenoviral vectors. Pre-existing neutralising antibodies or cellular immunity against AdV vectors are a problem, as they can decrease gene therapy or vaccine efficiency (Manno et al. 2006; Fausther-Bovendo and Kobinger 2014). This would likely be a smaller problem for LAdVs, only present in Madagascar. It is however unclear whether LAdVs would be safe and of any use for gene therapy and related techniques.

LAdVs are not immediately basal to human and simian AdVs

According to the hypothesis of AdV coevolution with primates, one would expect LAdVs to be most similar to simian AdVs and to fall immediately basal to simian AdVs in phylogenetic analysis. Importantly, animal AdVs are severely undersampled, which is a major limitation for evolutionary inference based on genetic information. Taking into account currently available AdV sequences, we found LAdVs to be similar to BtAdV-C in patristic distance and E3 genome organisation. BtAdV-Cs have so far been detected in Asia, and, to our knowledge, no BtAdV-Cs from Madagascar or prosimian AdVs from Asia have been described. As more animal AdVs are described, AdV sequences more similar to LAdVs may be discovered. We do not observe LAdVs to fall immediately basal to other primate AdVs in complete genome analysis. Instead, BtAdV-C, PAdV-A, and MuAdV-B all share a more recent common ancestor with primate AdVs other than LAdVs than they do with LAdVs. An explanation for the observed phylogeny could include one or more host switches. The directionality, timing, AdV, and host species of any such switch(es) are unclear.

Molecular evolution of LAdVs

We observe clear codivergence signals in LAdV non-structural genes, where phylogenies are clearly split by lemur species. Possible scenarios that might explain this are that the mouse lemur ancestor had AdVs with only one conserved set of non-structural genes, or that a bottleneck led to the loss of ancestral diversity in non-structural genes, and LAdVs subsequently codiverged along with allopatric speciation of their hosts. These scenarios assume that the two evolving lemur species did not encounter one another during speciation. Today, M. griseorufus and M. murinus do encounter one another, at least in the mixing zone Mangatsiaka investigated here, making recombination between LAdVs of the two species possible. Non-structural genes are estimated to be less diverse than structural genes in root-to-tip analysis, and conservedness can facilitate homologous recombination. While we observe patterns suggesting non-structural gene switches through recombination in LAdVs within each lemur species, we do not observe patterns suggestive for non-structural gene switches through recombination between AdVs of different lemur species in the mixing zone, despite this observation for structural genes. This may be due to undersampling. Non-structural genes may however also confer host-species-specific evolutionary advantages, as non-structural proteins need to interact with highly conserved host proteins and participate in crucial and complex functions, such as replicating the AdV genome. Non-structural genes may be subject to purifying selection.

In contrast, structural genes do not show a clear pattern of codivergence. It is possible that the ancestral Microcebus species already harboured LAdVs with different variants of structural genes. Eastern and western clades are the closest relatives in phylogenetic analyses, and the well-supported sister clade relationship of western Clade VI and eastern Clade VII support codivergence from an ancestral clade. Thus, structural gene variants present in the ancestral lemur species may have undergone allopatric evolution with their lemur hosts, but other evolutionary mechanisms, including recombination, likely occurred to produce the complex phylogenetic trees we observe. Structural gene diversity may be evolutionarily advantageous. Even though structural proteins need to be able to bind to each other and cellular receptors, diversity may allow exposed regions of proteins to evade prevailing adaptive immunity. Long-term selection pressure favouring diversity may have resulted in an overwriting of codivergence signals.

Recombination events and breakpoints observed in LAdVs mostly resembled the pattern seen in HAdV-C. However, a recombination pattern similar to HAdV-D was also observed in LAdVs from the geographical mixing zone of both lemur species, involving virus Clades I and III. To understand intrinsic and extrinsic factors restricting and facilitating recombination in LAdVs, more analyses using complete genomes and more importantly including LAdVs from more species and geographic sites are needed, especially as sampling in our study was limited to three sites and two mouse lemur species. A broader sample of AdV sequences from other lemurs and other Madagascan and Indian mammalian species, such as bats, would also aid the understanding of AdV origin and evolution in early primate species. Virus sequences more similar to LAdVs may be found, clarifying the timing and directionality of the proposed host switch.

Conclusion

LAdV genomes, the first completely sequenced prosimian AdVs, reveal the complexities of AdV evolution. The primate AdV evolutionary history cannot be explained by codivergence alone, and the evolution of LAdV possibly involves a host switch. LAdVs have however likely codiverged with the mouse lemur hosts, as is primarily evident in the non-structural genes. A combination of evolutionary mechanisms is necessary to resolve the evolution of lemur and primate AdVs.

Materials and methods

Sample collection

160 faeces samples of M. griseorufus were collected at three sites (Table S3) in southern Madagascar in 2013, 2014, and 2015 as analysed and described (Wasimuddin et al. 2019). In 2013 and 2017, additional 137 faecal samples of M. murinus were collected from two sites in the same manner as described by Wasimuddin et al. (2019) (Table S3). Microcebus spp. were live-trapped in six trapping grids in spiny forest around the Andranovao Research Camp at the western edge of the Tsimanampetsotsa National Park, referred to as Tsimanampetsotsa in this study (Ratovonamana et al. 2011; Scheel et al. 2015). Further sample sites were dry deciduous shrub close to Miarintsoa at the western site of Tsimanampetsotsa (Feldt and Schlecht 2016), dry deciduous and spiny forest of Mangatsiaka in Parcel II of the Andohahela National Park (Rakotondranary, Hapke, and Ganzhorn et al. 2011), and the humid littoral forest of Mandena (Andriamandimbiarisoa et al. 2015). A topographical map of the sampling sites was created using the open source tool QGIS, based on OpenStreetMap data (© OpenStreetMap contributors; openstreetmap.org), licensed as CC BY-SA, and the open source topographical data from Viewfinder Panoramas by Jonathan de Ferranti (http://www.viewfinderpanoramas.org/Coverage%20map%20viewfinderpanoramas_org3.htm). Tsimanampetsotsa and Miarintsoa are not separated by geographical barriers and are therefore referred to as one site in this paper. Elevations of >1,000 m are found between Mangatsiaka and Mandena, and they are separated by five rivers. The distance between western (Tsimanampetsotsa and Miarintsoa) and eastern (Mangatsiaka and Mandena) sites is >300 km, and hills as well as at least four rivers separate them.

Screening

From 50 µl of RNAlater (Invitrogen, Life Technologies, Karlsruhe, Germany) faeces suspension, total nucleic acids (NA) were extracted using the Roche MagNAPure 96 and the Viral NA Small Volume Kit (Roche, Mannheim, Germany). Initial HTS and subsequent hexon gene screening were as described previously (Wasimuddin et al. 2019).

Complete genome sequencing

Five complete LAdV genomes (of LAdV IDs 204, 354, 400, 814, and 2019) were created using HTS and Sanger sequencing. Libraries were prepared from native NA extracts using KAPA HyperPrep (Roche, Basel, Switzerland) followed by v3 2 × 300 bp (MiSeq) and v2 2 × 75 bp (NextSeq) sequencing chemistry on Illumina sequencers (Illumina, San Diego, USA). AdV NA–positive samples were PCR-amplified using a set of full-genome amplification primers designed based on the LAdV ID 204 complete genome sequence (Table S1) and sequenced as mentioned earlier on Illumina MiSeq. Remaining sequence gaps were filled with PCR and Sanger sequencing. The PCR used 10x Platinum Taq-Buffer (Invitrogen, Life Technologies, Karlsruhe, Germany), 0.5 U Invitrogen Platinum™ Taq DNA (Invitrogen, Life Technologies, Karlsruhe, Germany), 400 nM of forward and reverse primers each (Table S1), 200 nM of dNTPs each, 2.5 mM MgCl₂, 40 ng/ml BSA, and 2.5 µl sample extract. We used a PCR temperature protocol of 95 °C for 3 min, forty-five cycles of 95 °C for 15 sec, 58 °C for 15 sec, elongation 72 °C for 30 sec to 2 min (depending on expected amplicon size), and a final extension step at 72 °C for 2 min. The second round of PCR was carried out if needed, using the same protocol with 1 µl of the first round product as a template. The PCR products were visualised on agarose gels stained with Midori Green (Nippon Genetics, Düren, Germany) and were Sanger-sequenced (SEQLAB Sequence Laboratories, Göttingen, Germany). For a particularly large and non-conserved sequence gap in LAdV ID 2019, Expand™ High Fidelity (Sigma-Aldrich, Burlington, USA) PCR was carried out, using 10x buffer (Sigma-Aldrich), 2.45 U Expand™ High Fidelity enzyme (Sigma-Aldrich), 200 nM dNTPs each, 500 nM of each of the forward primer AdV-TV-102 F and reverse primer AdV-TV-99 R, as well as 3 µl sample extract. The temperature protocol used was 94 °C for 2 min for Mix 1 (no enzyme), 80 °C for 1 min and hold for manual hot start, forty-five cycles of 94 °C for 20 sec, 56 °C for 20 sec, 68 °C for 4 min 20 sec, and a final elongation step of 72 °C for 3 min. The resulting PCR amplicon was MinION-sequenced using the kit SQK-PSK004 (Oxford Nanopore Technologies, Oxford, UK) according to the manufacturer’s instructions.

Complete genome analysis

Generated sequencing reads were mapped, and complete genome sequences were generated, aligned, and annotated using Geneious Prime (Geneious Prime 2020 2.5 https://www.geneious.com), Porechop (https://github.com/rrwick/Porechop), minimap2 (Li 2018), and ClustalW (Sievers and Higgins 2018). BLASTn (Altschul et al. 1997) was used to search the NCBI nucleotide database with standard parameters and the ‘discontiguous megablast’ option. Recombination analysis of complete genome sequences was carried out using ClustalW (Sievers and Higgins 2018), the Geneious Prime DualBrothers Plugin (Suchard et al. 2002), and RDP5 (Martin et al. 2021). The distance plot in Fig. 4 was created using RDP4 (Martin et al. 2017). ML trees were constructed using Geneious Prime Tree Builder (Geneious Prime 2020 2.5 https://www.geneious.com) and raxml-ng (Kozlov et al. 2019), using models GTR + I + Gamma (GTRIG) as indicated by jModelTest (Guindon and Gascuel 2003; Darriba et al. 2012). We calculated up to 1,000 bootstrap replicates. The accession numbers of viral sequences used for phylogenetic trees are given in Table S2. The phylogenetic tree of primates is adapted from Yoder and Yang (2004).

Recombination PCR and analysis

PCR using Invitrogen Platinum™ Taq DNA (Invitrogen, Life Technologies, Karlsruhe, Germany) was carried out and Sanger-sequenced as described earlier using primers for E1A, DNA polymerase, penton base, fiber, and E4 34k genes (Table S1). Amplicon recombination analysis was done using MAFFT (Katoh et al. 2002; Nakamura et al. 2018), Geneious Prime (Geneious Prime 2020 2.5 https://www.geneious.com), and raxml-ng (Kozlov et al. 2019). Models GTR + Gamma (GTRG), GTRIG, TIM2 + I + G, and Blosum62 as indicated by jModelTest (Guindon and Gascuel 2003; Darriba et al. 2012) were used, and up to 1,000 bootstrap replicates were calculated. Root-to-tip distances were computed using the Python ETE-3 package (Huerta-Cepas, Dopazo, and Gabaldón 2010).

Data analysis

BLASTp (Altschul et al. 1997) was used to search the NCBI non-redundant protein database with standard parameters for proteins similar to E3l. Translations of LAdV E3l and BtAdV-C E3l (MZ683934, MW597741, and NC_029899) were aligned using MAFFT (Katoh et al. 2002; Nakamura et al. 2018), and aa identity was calculated using Geneious Prime (Geneious Prime 2020 2.5 https://www.geneious.com). The online tool PredictProtein (Bernhofer et al. 2021), the TMSEG algorithm (Bernhofer et al. 2016), and InterPro (Blum et al. 2021) were used to analyse the LAdV E3 large ORF aa sequence. The protein structure of LAdV ID 400 E3l was predicted using AlphaFold (Jumper et al. 2021; Varadi et al. 2022) and compared to structures in the RCSB PDB (Berman 2000) using DALI (Holm 2020). Plots were created using Inkscape and the Python libraries matplotlib v3.3.4 (Hunter 2007), seaborn v0.11.1 (Waskom 2021), and pandas v1.2.2 (McKinney 2010). Statistical analyses were performed in Python, using SciPy 1.6.0 (Virtanen et al. 2020) and statannot 0.2.3 (Webermarcolivier 2022).

Data availability

All generated LAdV sequence data genomes can be found on the NCBI GenBank under the accession numbers OQ081771–OQ081775 for complete LAdV genomes, OQ081776–OQ081817 for amplicon sequences of the E1A gene, OQ081818–OQ081872 for amplicon sequences of the hexon gene, OQ081873–OQ081917 for amplicon sequences of the penton base gene, OQ081918–OQ081957 for amplicon sequences of the E4 34k gene, OQ081958–OQ081992 for amplicon sequences of the fiber gene, and OQ081993–OQ082035 for amplicon sequences of the DNA polymerase gene.

Supplementary data

Supplementary data are available at Virus Evolution online.

Funding

The fieldwork was supported by the Deutsche Forschungsgemeinschaft (DFG Ga 342/19; Priority Program SPP 1596/2 Ecology and Species Barriers in Emerging Infectious Diseases, DR 772/8-1, SO 428/9-1) and the Landesforschungsförderung Hamburg. The study was carried out in a collaboration between Madagascar national parks (MNPs), the Department of Animal Biology and Department of Plant Biology and Ecology (Antananarivo University, Madagascar), and the Department of Animal Ecology and Conservation (Hamburg University, Germany) and authorised by Madagascar’s Ministère de L’Environnement, de l’Écologie, de la Mer et des Forêts. T.C.J. is funded in part through the NIAID-NIH CEIRS contract HHSN272201400008C. V.M.C. is supported by the Berlin Institute of Health (BIH) Charité Clinician Scientist program.

Conflict of interest:

None declared.

Ethics approval

This research was approved by the Ethics Committee of the Institute of Zoology of Hamburg University, the University of Antananarivo, and Madagascar national parks. Autorisation de Recherche No. 54/13/MEF/ SG/DGF/DCB.SAP/SCB of 22 February 2013 issued by the Direction Générale des Forêts and the Direction de la Conservation de la Biodiversité et du Système des Aires Protégées of the Ministère de l’Environnement, et des Forêts, and exported to Germany under the CITES permit 576C-EA09/MG14. We confirm that all methods were performed in accordance with the relevant guidelines and regulations.