A divergent hepatitis A virus lineage in wild western lowland gorillas and central chimpanzees supports localized circulation and expands HAV diversity

Abstract

Hepatitis A virus (HAV) is a ubiquitous human pathogen that may cause acute liver disease. Despite the availability of effective vaccines, high prevalence rates are still reported in Southeast Asia and Africa, where sanitation conditions are poor. Although seven genotypes are known, including three in humans, three in non-human primates (NHP), and one recently identified in alpacas, the diversity, ecology, and evolutionary history of HAV remain poorly understood. In a pilot study of virome molecular characterization of wild African apes by using oligonucleotide probes enrichment, HAV sequences were detected in two animals (2/105; 1.9%): one from a western lowland gorilla (Gorilla gorilla gorilla) from southern Cameroon, sampled in 2006, and another from a central chimpanzee (Pan troglodytes troglodytes) in northeastern Gabon, sampled in 2011, approximately 450 km apart. Two near-full-length viral genomes were successfully reconstructed using specific primer sets that were developed. Phylogenetic analyses revealed that these sequences cluster within the simian HAV clade, forming a distinct cluster closely related to environmental strains from the Sanaga River in Cameroon and South Africa, yet clearly divergent from genotype V sequences (≥15.8% nucleotide divergence in the VP1 protein). Amino acid comparisons identified residues conserved in simian HAVs and, subsequently, only in African HAV strains. These results support the classification of a supplementary African simian genotype, provisionally designated HAV-VIII. HAV-positive samples (13/227; 5.7%) were detected in the same field sites, suggesting localized viral circulation in these populations. Bayesian time-scaled phylogenetic analysis estimated the most recent common ancestor of the human and simian HAV lineages at ~ 2 291 years ago and that of simian lineages at ~ 1900 years ago, with a more recent diversification (~100 years ago) within the identified lineage. This study highlights the presence of HAV in wild ape populations for the first time and also the first HAV detection among NHP species in Central Africa. These findings contribute to our current understanding of the diversity and evolution of HAV and highlight the need to extend surveillance across NHP species to better assess potential zoonotic risks.

Introduction

Hepatitis A virus (HAV) is a 7.5-kb non-enveloped RNA virus. Within the family Picornaviridae and the genus Hepatovirus, HAV is classified as species Hepatovirus ahepa, according to the International Committee on Taxonomy of Viruses (ICTV Report Consortium 2025). Clinical manifestations range from asymptomatic to acute hepatitis, with symptomatic forms occurring more frequently in adults than in children (Lemon et al. 2018). Although HAV infection can lead to fulminant hepatitis, it is generally self-limiting and does not lead to chronic disease compared to other viral hepatitis (Lemon et al. 2018). The faecal-oral route is the major transmission mode, either directly through contaminated water or indirectly through consumption of seafood or poorly washed foods such as vegetables or fruits (Gaillot et al. 1988). Southeast Asia and Africa have higher prevalence rates than other regions, principally because of low sanitation standards (Van Damme et al. 2023). Approximately 150 million human infections and 39 000 deaths were reported worldwide in 2019, despite the availability of effective vaccines.

Until quite recently, six HAV genotypes were described, genotypes I–III in humans, each subdivided into two subgenotypes (A and B) and genotypes IV–VI in non-human primates (NHP) (Cristina and Costa-Mattioli 2007). But an additional genotype, VII, was recently reported in alpacas and to date, it is the only HAV genotype reported outside the order Primates (Veith et al. 2023). All simian HAV strains possess a signature sequence at the VP3/VP1 junction distinguishing them from human strains (Brown et al. 1989, Nainan et al. 1991, Tsarev et al. 1991). The three simian HAV genotypes have been isolated only from Old World monkeys. Whereas genotypes IV and VI were identified in Asian NHP species (Macaca fascicularis), genotype V was detected in African NHP species on several occasions (Table 1). More recently, HAV infection was reported in captive crab-eating macaques (M. fascicularis) from Mauritius and considered a divergent HAV genotype (Mecklenburg et al. 2024).

Table 1.

Summary of currently reported simian HAV strains (host species, host status at the detection and geographic origin). For HAV-V, host species, host status and geographic origin are presented in corresponding order.

HAV genotype	Host species	Host status	Geographic origin	References
HAV-IV	Macaca fascicularis	Captive	Imported into the USA from the Philippines	Nainan et al. 1991
HAV-V	Chlorocebus spp., Macaca spp.; Papio anubis	Captive (wild-caught, captive-born); free-living	Kenya, Tanzania, Russian centre, Indian NHP facility; Uganda,	Andzhaparidze et al. 1987; Tsarev et al. 1991; Bennett et al. 2016; Dogadov et al. 2018
HAV-VI	M. fascicularis	Captive	Imported into Russia from Indonesia	Andzhaparidze et al. 1985; Nainan et al. 1991
Mecklenburg HAV strain	M. fascicularis	Captive	Imported into Germany from a breeding facility in Mauritius	Mecklenburg et al. 2024

Antibodies against HAV have been detected in different NHP species, i.e. in wild-caught chimpanzees (Kessler et al. 1982) as well as in other wild NHP species living close to human populations in Asia (Burke and Heisey 1984), in South Africa (Drewe et al. 2012), and America (Svoboda et al. 2016). Chimpanzees have been observed to be susceptible to HAV, including human genotype I and simian genotype V strains from African green monkeys (Abe and Shikata 1982, Emerson et al. 1996, Lanford et al. 2011, Theamboonlers et al. 2012).

The paucity of current knowledge about HAV in NHPs limits our understanding of the diversity and the evolutionary trajectory across human and NHP species. Because most of our knowledge regarding simian HAV is based on only a few instances of HAV detection in captive animals, it is unclear whether simian HAV circulates among wild animals. In this study, we identified a divergent HAV genotype in wild ape populations in their natural habitat in Central Africa that supports a viral circulation among these populations, and we further document the genetic diversity and evolution of simian HAV strains.

Methods

Study population

As part of a descriptive and exploratory pilot study to better characterize the virome in wild ape populations in their natural environment in Africa, 117 faecal samples were selected from a panel of over 10 000 samples collected during previous studies on the origins of human immunodeficiency virus (HIV) (Keele et al. 2006, Van Heuverswyn et al. 2006, Van Heuverswyn et al. 2007, D’arc et al. 2015). Samples were selected to maximize the geographic representation of wild apes across the sampling sites in Central Africa [Cameroon, Democratic Republic of the Congo (DRC), Gabon] (Fig. 1, Table 2, Supplementary Table S1). Faecal samples were collected from wild-living (non-habituated) apes and preserved at −80°C in an RNA-stabilizing solution (RNA-Later, Ambion, Austin, TX, USA) as previously described (Keele et al. 2006, Van Heuverswyn et al. 2006). Only samples for which species and subspecies assignment was previously confirmed by mitochondrial DNA (mtDNA) methods or confidently extrapolated based on field information were included. In addition, for some samples of this study, individual characterization by microsatellite analysis was also available as previously reported (Keele et al. 2006, Van Heuverswyn et al. 2006, Van Heuverswyn et al. 2007, D’arc et al. 2015). For each sample, visual species identification, collection date, GPS coordinates, and ecological environment were recorded in the field. The faecal samples were collected and studied with approval from different ministries and ethics committees from the different countries as stated in the acknowledgements.

Figure 1.

Map of sampling sites in Central Africa. The letter codes of the 35 sampling sites correspond to those previously used (Keele et al. 2006; Van Heuverswyn et al. 2007; Neel et al. 2010; Prugnolle et al. 2010; Liu et al. 2014; Boué et al. 2015; D’arc et al. 2015) and are coloured distinctly by ape species studied at each site. Species and subspecies assignments were confirmed by mtDNA molecular analysis (Keele et al. 2006, Van Heuverswyn et al. 2006). The geographical distribution of African ape species and subspecies is shown, with dotted or solid lines indicating gorilla species and solid areas for chimpanzee species, except for the western chimpanzee (Pan troglodytes verus), which is located in the northwest of the continent. Samples from KIN are from an NHP sanctuary. The base map was created using the software QGIS (https://github.com/qgis/QGIS/blob/master/INSTALL.md). The satellite imagery is from Esri World Imagery and is used under license. Copyright © 2026 Esri and its licensors. All rights reserved.

Table 2.

Characteristics of the faecal samples analysed in this pilot study. The letter codes of the 35 sampling sites correspond to those previously used and shown in Fig. 1. The values in brackets indicate the corresponding number of individuals when different from the number of samples.

Collection location	Sampling sites	G. g. gorilla	G. g. diehli	Pan paniscus	P. t. ellioti	P. t. schweinfurthii	P. t. troglodytes	Total
Cameroon	AL	2	0	0	0	0	0	2
Cameroon	BQ	3	0	0	0	0	4	7
Cameroon	BP	2	0	0	0	0	0	2
Cameroon	BM	1	0	0	0	0	2	3
Cameroon	BY	2	0	0	0	0	2	4
Cameroon	CP	4	0	0	0	0	1	5
Cameroon	DD	2	0	0	0	0	0	2
Cameroon	DJ	4	0	0	0	0	2	6
Cameroon	DP	0	0	0	0	0	2	2
Cameroon	EK	0	0	0	0	0	3	3
Cameroon	GB	2	0	0	0	0	0	2
Cameroon	KG	0	0	0	0	0	1	1
Cameroon	KK	2(1)	0	0	0	0	0	2(1)
Cameroon	LB	3	0	0	0	0	2	5
Cameroon	LD	2	0	0	0	0	0	2
Cameroon	LM	2	0	0	0	0	0	2
Cameroon	MB	2	0	0	0	0	8(7)	10(9)
Cameroon	MF	0	0	0	2	0	0	2
Cameroon	MG	0	0	0	0	0	2	2
Cameroon	MLD	2	0	0	0	0	0	2
Cameroon	MM	2	0	0	0	0	0	2
Cameroon	MS	2	0	0	0	0	0	2
Cameroon	MP	0	0	0	1	0	0	1
Cameroon	MT	1	0	0	0	0	0	1
Cameroon	NK	1	0	0	0	0	1	2
Cameroon	SL	0	0	0	0	0	1	1
Cameroon	TK	0	1	0	0	0	0	1
DRC	BG	0	0	0	0	2(1)	0	2(1)
DRC	KIN	0	0	3	0	0	0	3
DRC	LK	0	0	6(3)	0	0	0	6(3)
DRC	LP	0	0	1	0	0	0	1
DRC	LUKI	0	0	0	0	0	1	1
DRC	ML	0	0	7(3)	0	0	0	7(3)
DRC	MZ	0	0	11	0	0	0	11
Gabon	MA	0	0	0	0	0	10(8)	10(8)
Total		41(40)	1	28(21)	3	2(1)	42(39)	117(105)

Sample preparation, viral detection, and NGS sequencing

To extract nucleic acids from faecal samples, the NucliSens miniMAG kit (bioMérieux, Durham, NC) was used, a semi-manual system based on magnetic silica particle technology (Boom et al. 1990). cDNA was produced with the LunaScript RT SuperMix Kit (NEB, Ipswich, MA, USA) and the dsDNA synthesis with the NEBNext® Ultra™ II Non-Directional RNA Second Strand Synthesis Module (NEB, Ipswich, MA, USA). The dsDNA was then purified using magnetic beads (DNA Purification Beads, South San Francisco, CA, USA) and stored at +4°C. Libraries were prepared using the Twist Comprehensive Viral Research panel workflow (Twist Bioscience, South San Francisco, CA, USA) according to the manufacturer’s instructions, and pools of eight samples were enriched with the probe panel. They were loaded onto an Illumina MiSeq™ sequencer using a MiSeq Reagent Kit v2 (2 × 150 bp paired-end reads) (Illumina, San Diego, CA, USA), and ~1.5 million paired-end reads per library were generated. To improve the viral genome coverage, additional enrichment was also performed for a subset of samples with the Viral Surveillance Panel (VSP2) (Illumina, San Diego, CA, USA) according to the manufacturer’s protocol. FASTQ files were analysed using an online pipeline, Genome Detective (https://www.genomedetective.com/db/ui/login;  Vilsker et al. 2019), which assembles and identifies a wide diversity of known viruses available in GenBank. Subsequently, consensus sequences with at least 3× sequencing depth were generated for samples with the presence of HAV using the GeVarLi pipeline (https://forge.ird.fr/transvihmi/nfernandez/GeVarLi), after read alignment with the BWA aligner (Li and Durbin 2009) to several HAV reference genomes (accessions: NC_001489.1; M14707; D00924.1). The pipeline also performs a quality control step to eliminate low-quality reads, followed by a trimming step to remove unnecessary nucleotide sequences added during library preparation.

Characterization and sequencing of full-length Hepatitis A virus genomes

To obtain high-quality full-length HAV sequences, 15 primer sets, mostly overlapping, based on the next-generation sequencing (NGS) consensus sequences at 3× and 1×, were designed (Supplementary Table S2, Supplementary Fig. S1). Partial NGS consensus sequences used for primer design are provided as supplementary data. cDNAs were produced with the GoScript™ Reverse Transcription System A5001 kit (Promega, Madison, WI, USA) according to the manufacturer’s recommendations using random primers and a final concentration of MgCl₂ at 4.0 mM. HAV amplifications were done after an activation of the HotStarTaq® DNA polymerase (Qiagen, Hilden, Germany) at 95°C for 15 min followed by 45 cycles with denaturation at 94°C for 20 s, a hybridization step at the optimal temperature for each primer pair (Supplementary Table S2), then elongation at 72°C for 45 s and 5 supplementary minutes for a final extension during the last cycle. Amplification mixtures had a total volume of 50 μL and included 5 μL of cDNA, 25 μL of HotStarTaq® Master Mix (which contains 200 μM of each dNTP, 1.5 mM of MgCl₂, 2.5 units of HotStarTaq® Polymerase), 5 μg of BSA (bovine serum albumin) (Ambion, Austin, TX, USA), and 0.4 μM of each primer. The amplicons were visualized on a 1% agarose gel with 0.005% BET (Ethidium bromide). PCR products at the expected size were purified using the GeneClean® Turbo Kit (MP Biomedicals, Irvine, CA, USA) and Sanger sequencing was done on the ABI 3500 equipment (Applied Biosystems, Foster City, CA, USA) using the BigDyeTerminator v3.1 Sequencing Kit (Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s protocol. ABI-format chromatograms obtained were inspected, edited, and assembled into consensus sequences using SeqMan Ultra™, version 18.1.0 (DNASTAR, Inc., Madison).

Reconstruction of Hepatitis A virus full-length genome sequences

To reconstruct full-length HAV sequences, the fragments obtained by Sanger sequencing were aligned using MAFFT v7.256 (Multiple Alignment using Fast Fourier Transform; Rozewicki et al. 2019) with the NGS consensus (3×) and a HAV reference genome (D00924.1) to guide the alignment and ensure the correct positioning. For each sample, a final consensus sequence was generated using Geneious Prime® 2026.0.2. Sanger sequences were used to resolve positions not covered (N) in the NGS consensus or below 3× depth. When a nucleotide position was covered by a maximum of three times due to the overlap of certain fragments, a threshold of 65% (2/3) was considered to call the base in the final consensus. For the gorilla sample, several non-consecutive gaps (totaling 138 bp) located between positions 3105 and 4038 of the final consensus could not be resolved by Sanger sequencing due to amplification failure and were filled in by nucleotide reads at 1× depth in the NGS consensus.

Maximum likelihood and nucleotide identity analyses

To analyse the phylogenetic relationships of the HAV sequences produced in the study, the final consensus sequences were aligned with reference sequences representing the known diversity of Hepatovirus ahepa species (Supplementary Table S3) using MAFFT. Several HAV reference sequences isolated from human samples collected in countries where ape samples were collected were included, with collection dates close to those available in GenBank. Phylogenetic analysis was performed using the Maximum Likelihood (ML) method on the VP1 protein used to classify the HAV strains (Costa-Mattioli et al. 2002), on the VP1-2A junction to increase the number of HAV strains, and on the 5′ untranslated region (UTR) for positive supplementary samples. We selected the best evolutionary model using ModelFinder (Kalyaanamoorthy et al. 2017) within IQ-TREE 3 (Wong et al. 2025). Branch supports were estimated by Ultrafast Bootstrap (UFBoot), and Shimodaira-Hasegawa approximate likelihood-ratio test (SH-aLRT) with 1000 replicates (iqtree3 -s alignment.fasta -m MFP -bb 1 000 -alrt 1 000 -T AUTO), and only branches with UFBoot ≥95% and SH-aLRT ≥80% were considered well-supported. The trees were annotated using iTOL v6 (Letunic et al. 2024). Genetic distances were calculated using p-distance method using MEGA 12.1 (Stecher et al. 2026).

Bayesian phylogenetic analyses

To estimate the divergence age of the ape HAV sequences, an alignment of full-length HAV genomes available on GenBank (Supplementary Table S3) was generated using MAFFT. The human sequences were selected to ensure sufficient temporal diversity in the sampling year and to represent diverse geographical areas for each of the human genotypes, based on the dataset by Wang et al. 2015, and all complete simian sequences available in GenBank were selected. As recombination can influence molecular clock analyses, a Simplot v.3.51 (Lole et al. 1999) and RDP5 analyses (Martin et al. 2020) were performed to detect recombination sites within the alignment. The final alignment was explored to identify the presence of temporal signal within the known lineages of human and simian HAV using TempEst (Supplementary Fig. S2) (Rambaut et al. 2016). Bayesian phylogenetic analyses were performed using BEAST software v10.5.0 (Baele et al. 2025) with a gradual calibration strategy. First, a dataset containing only genotype I HAV sequences from humans for which the highest numbers are available in GenBank and which also cover an extended sampling period (61 years) was selected to build a reference-time tree dataset with strong temporal signal (R² = 0.70), as reported in Supplementary Figs S2 and S3. Additionally, a date randomization test (DRT) was performed on this dataset to directly assess the temporal signal underlying the calibration strategy, as reported in Supplementary Fig. S4.

The HAV I time to the most recent common ancestor (tMRCA) mean (1657) and substitution rate (2.382 × 10⁻⁴) obtained in this dataset were used as priors to calibrate nodes in a dataset with the other human genotypes. Then, the estimated tMRCA for genotypes I, II, III, and all human strains, were used as priors in the dataset including all human and simian genotypes. This approach allowed for improved convergence for all parameters (ESS > 200). Two independent Markov chain Monte Carlo (MCMC) of 50 000 000 generations were run for the human genotype I dataset, with sampling every 5000 steps. For the human genotypes I–III dataset and the complete dataset (human and simian genotypes), a single MCMC of 50 000 000 generations was sufficient to achieve convergence. GTR + I + Γ4 substitution model was applied with a relaxed molecular clock (uncorrelated lognormal) and an exponentially growing population (Wang et al. 2015). The burn-in period was set to 10% of the generations when combining runs with LogCombiner. Convergence and chain mixing were evaluated using Tracer v1.7.2 (Rambaut et al. 2018), which ensured that the independent runs converged to the same distribution a posteriori. The maximum credibility tree of clades (MCC) was generated using TreeAnnotator v10.5.0 (Drummond and Rambaut 2007), visualized and annotated with iTOL v6 (Letunic et al. 2024). BEAST files are provided as supplementary data.

PCR detection of HAV in faecal samples from apes

To further evaluate the presence of HAV in additional ape faecal samples, we designed HAV primers to amplify all HAV genotypes in the highly conserved 5′ UTR as shown in Supplementary Fig. S1 by a nested approach (HAV_DIAG_238F1:5′-AGCTGTAGGAGTCTAAATTG-3′ and HAV_DIAG_664R1:5′-TGAGGAAAAACCTAAATGC-3′ for the first round; HAV_DIAG_242F2:5′-TGTAGGAGTCTAAATTGG-3′; HAV_DIAG_539R2: 5′-TTTGCCCTAAGCACAGAGAG-3′ for the second round). The two obtained HAV consensus sequences were also included in the primer design. The first PCR round (426 bp) was done after activation of the HotStarTaq® DNA Polymerase (Qiagen, Hilden, Germany) at 95°C for 15 min followed by 10 cycles with denaturation at 94°C for 20 s, a hybridization step set to a touchdown at 55°C allowing a temperature decrease of 0.5°C per cycle of 30 s followed by elongation at 72°C for 30 s.

An additional 35 cycles were added under the same conditions, with the hybridization temperature maintained at 50°C for 30 s and a final elongation of 5 min at the last cycle. The second PCR round (297 bp) differed only in the hybridization temperature, which was lowered by a touchdown from 58 to 52°C for 12 cycles, then maintained at 53°C for an additional 23 cycles. Amplification mixtures had a total volume of 50 μL and included 5 μL of cDNA or 2 μL of first-round DNA, 25 μL of HotStarTaq® Master Mix, 5 μg of BSA for the first-round, and 0.4 μM of each primer. The amplicons were visualized on a 2% agarose gel with 0.005% BET, and PCR products at the expected size were purified when needed using the GeneClean® Turbo Kit (MP Biomedicals, Irvine, CA, USA) before Sanger sequencing on the ABI 3500 equipment (Applied Biosystems, Foster City, CA, USA) using the BigDyeTerminator v3.1 Sequencing Kit (Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s protocol. ABI-format chromatograms obtained were inspected, edited, and assembled as described above. Confidence intervals for HAV detection rates were calculated using the Wilson method.

Results

Detection and characterization of near-full-length HAV sequences in wild ape populations in Central Africa

The 117 faecal samples from apes were selected based on dates, localities, and microsatellite analysis data, so that as many as possible were derived from different individuals and covered a large geographic range. The samples were collected between 2003 and 2017 at 27 sites in Cameroon, 7 in DRC, and 1 in Gabon. The details are shown in Fig. 1, Table 2, and Supplementary Table S1. The 117 samples correspond to 105 different individuals, based on collection date and site, or by microsatellite analysis when collected at the same site: 47 chimpanzee samples from 43 animals, 42 gorilla samples from 41 animals, and 28 bonobo samples from 21 animals. Gorilla samples are from western lowland gorillas (Gorilla gorilla gorilla), except one from a Cross River gorilla (Gorilla gorilla diehli) located at the Cameroon–Nigeria border. Among the 47 chimpanzee samples, 42 are from central chimpanzees (Pan troglodytes troglodytes) in Cameroon and Gabon, corresponding to 39 animals; 2 are from eastern chimpanzees (Pan troglodytes schweinfurthii) in DRC, corresponding to 1 animal; and 3 are from Nigeria–Cameroon chimpanzees (Pan troglodytes ellioti), corresponding to 3 distinct animals.

NGS sequencing using probe viral enrichment (Twist Comprehensive Viral Research Panel) detected HAV sequences in 2/117 samples (1.7%; 95% CI: 0.47–6.02) corresponding to 2/105 individuals (1.9%; 95% CI: 0.52–6.68): a western lowland gorilla from southwestern Cameroon, in MM sampling site and a central chimpanzee from northeastern Gabon, in MA sampling site (Fig. 1). These HAV-positive samples were collected at ~450 km distance and a 5-year interval (Fig. 1, Supplementary Table S1). Because no full-length genome sequences were obtained with this approach, both samples were resequenced with another viral probe enrichment panel (VSP2) to reduce off-target viral capture (200 viruses targeted) and to try to increase genome coverage and depth. The results of both approaches showed partial genome coverage with a heterogeneous distribution of reads across the viral genomes (24%–72% coverage at a depth ≥ 3×, depending on the probe panel). Details of coverage and depth NGS profiles for both samples are shown in Supplementary Fig. S5. HAV-specific primers allowed the amplification of overlapping PCR fragments, especially those not covered by NGS results, and to obtain near-full-length HAV sequences. By combining different sequencing approaches, two near-full-length HAV sequences of 7403 and 7405 bp, respectively, from a chimpanzee and a gorilla were obtained.

Phylogenetic analyses of the HAV sequences including ape HAV sequences and amino-acid signatures: identification of a divergent genotype

To characterize the HAV sequences produced in this study within the known HAV diversity, phylogenetic analysis was performed using the ML method on the full VP1 coding sequence (900 bp) (Fig. 2), which is used for genotype classification (Costa-Mattioli et al. 2002). The two ape HAV sequences from a chimpanzee and a gorilla fall into the simian clade. They form a well-supported independent cluster with a strain detected in irrigation water in South Africa (Rachida et al. 2016), which is distantly related to HAV-V sequences previously reported in free-living and wild-caught NHPs from East Africa (Andzhaparidze et al. 1987, Tsarev et al. 1991, Bennett et al. 2016), with strong support for the corresponding node. HAV-V is the only genotype that contains sequences from African NHPs. The HAV sequences from the gorilla and chimpanzee show nucleotide distances of 8.3% and 7.8% with the South African strain, respectively, and exhibit nucleotide distances of 15.9% and 15.8% with HAV-V strain, respectively (Table 3). The observed divergence from genotype V exceeds the divergences within subgenotypes (8.4%–11.1%, Table 3) and falls within the range of intergenotype divergences (14.1%–23%), suggesting that the ape HAV strains form a distinct genotype based on VP1 classification criteria (Costa-Mattioli et al. 2002) named tentatively HAV-VIII. The Mecklenburg HAV strain previously detected in M. fascicularis clusters with the strain Cy145 of HAV-IV and does not form an independent lineage. The nucleotide divergence of 13.2% observed from this sequence does not fall within the range of intergenotype divergences and, therefore, suggests that it belongs to HAV-IV (Fig. 2, Table 3).

Figure 2.

Genome organization of Hepatovirus ahepa (A) and maximum-likelihood phylogenetic reconstruction (B) of the complete VP1 region (900 bp). The genome region included in the phylogenetic analysis is shown in panel A. The VP1 extraction was performed by alignment against an annotated reference genome. The best-fit substitution model, as determined by ModelFinder in IQ-TREE, is TIM2 + F + I + R2. The tree was rooted using a Hepatovirus devoli sequence. Branch supports were estimated by SH-aLRT and UFBoot with 1000 replicates. Only branches with SH-aLRT ≥80% and UFBoot ≥95% were considered well-supported and are shown. Each sequence is labelled with host species, the sampling year (when available), and the country of origin, except for sequences sampled from captive animals for which the country was not indicated. Human genotypes (I–III), simian genotypes (IV, V), and alpaca genotype (VII) are indicated. HAV-VI could not be integrated into the analysis because the GenBank sequence is too short to cover the region analysed. The two ape HAV sequences are depicted by an ape symbol. An HAV sequence identified in wastewater collected in South Africa is also included, depicted in bold. The tree was visualized and edited using iTOL v6 (Letunic et al. 2024).

Table 3.

Pairwise nucleotide distance (p-distance) of the complete VP1 coding sequence (900 bp) between the ape HAV sequences and HAV strains representing the diversity of established (sub)genotypes. One HAV sequence per (sub)genotype was used. HAV-VI could not be integrated into the analysis because the available GenBank sequence is too short.Distance values between the two ape HAV sequences and genotype V, South African strain, are shown in bold, and also between the Mecklenburg strain and HAV-IV sequence. Evolutionary analyses were conducted in MEGA 12.1 (Stecher et al. 2026). Accession numbers used: IA, AB020564.1; IB, MH685714.1; IIA, MH933768.1; IIB, AY644670.1; IIIA, AB279732.1; IIIB, AB279735.1; IV, M59286.1; V, D00924.1; VII, OR452344.1; Mecklenburg strain, OQ559662.1; South African strain, KP875236.1.

	G. g. gorilla	P. t. troglodytes	South African strain	HAV-V	HAV-IV	Mecklenburg strain	HAV-VII	HAV-IIIB	HAV-IIIA	HAV-IA	HAV-IB	HAV-IIB
G. g. gorilla
P. t. troglodytes	0.032
South African strain	0.083	0.078
HAV-V	0.159	0.158	0.159
HAV-IV	0.198	0.199	0.202	0.218
Mecklenburg strain	0.201	0.207	0.207	0.203	0.132
HAV-VII	0.193	0.195	0.204	0.212	0.212	0.209
HAV-IIIB	0.194	0.186	0.193	0.213	0.193	0.190	0.187
HAV-IIIA	0.207	0.202	0.213	0.23	0.203	0.199	0.197	0.111
HAV-IA	0.180	0.175	0.184	0.183	0.201	0.209	0.194	0.182	0.186
HAV-IB	0.186	0.189	0.204	0.188	0.203	0.212	0.194	0.188	0.190	0.084
HAV-IIB	0.206	0.210	0.214	0.20	0.213	0.210	0.190	0.192	0.193	0.153	0.160
HAV-IIA	0.204	0.198	0.201	0.201	0.203	0.208	0.193	0.187	0.199	0.141	0.159	0.092

Detailed amino acid analysis of the VP1 protein (Fig. 3) shows that the two ape HAV sequences repeatedly share several residues that are only observed in simian HAV and not found in human HAV. These include amino acids at positions 1 (Valine to Threonine), 102 (Serine to Alanine), 105 (Lysine to Arginine), 178 (Serine to Threonine), and 188 (Valine to Isoleucine). Two other positions, 271 (Serine to Valine) and 294 (Glutamic acid to Alanine), are shared only with simian HAV strains from Africa (genotype V and the South African strain). To investigate whether the amino acid substitutions identified in the ape HAV VP1 protein could affect VP1 structure, a 3D prediction was performed and is shown in Supplementary Figs S6 and S7. Most specific substitutions were predicted to occur in solvent-exposed regions, several of them mapping to surface-exposed loops.

Figure 3.

Amino acid sequence alignment of the complete VP1 protein (300 amino acids). The translation was performed with Geneious Prime® 2026.0.2. Residue positions and the consensus sequence are shown above the alignment identity. Conserved residues are depicted by a dot. The 24 HAV sequences representing human (H-HAV), simian genotypes (HAV-IV, HAV-V), and alpaca genotype (HAV-VII) are shown on the left side with accession numbers. HAV-VI could not be integrated into the analysis because the available GenBank sequence is too short. The ape viral sequences are listed first, depicted in bold and with an ape symbol. The South African strain is depicted by a farm symbol. Above the positions, several symbols are marked: a triangle for residues shared by all simian strains and a diamond for residues shared by African simian strains. All these positions are outlined. Position 221 was not highlighted because the substitution is not shared by all simian strains. The figure was created using Geneious Prime® 2026.0.2.

Phylogenetic analysis was also performed on partial HAV VP1-2A sequences (517 bp) to include a recently reported HAV sequence in GenBank of environmental origin, sampled from a clam (Egeria radiata) in the Sanaga River, Cameroon (Bonny et al. 2020) (Fig. 4). The ape HAV sequences, as well as the wastewater strain from South Africa, form a well-supported cluster with the recently reported strain from a clam, with nucleotide distance ranging from 3.2% to 3.9%, respectively, with the gorilla and chimpanzee HAV sequences (Table 4).

Figure 4.

Genome organization of Hepatovirus ahepa (A) and maximum-likelihood phylogenetic reconstruction (B) of the VP1-2A junction region (517 bp). The genome region included in the phylogenetic analysis is shown in panel A. The best-fit substitution model, as determined by ModelFinder in IQ-TREE 3, is TIM2 + F + G4. The tree was rooted using a Hepatovirus devoli sequence. Branch supports were estimated by SH-aLRT and UFBoot with 1000 replicates. Only branches with SH-aLRT ≥80% and UFBoot ≥95% were considered well-supported and are shown. The HAV-VI sequence, Sanaga clam strain, Chlorocebus pygerythrus strain, and the strain Cy145 cover 167 bp, 440 bp, 321 bp, and 340 bp of the region, respectively. Each sequence is labelled with host species, the sampling year (when available), and the country of origin, except for sequences from captive animals for which the country was not indicated. Human genotypes (I–III), simian genotypes (IV–VI), and alpaca genotype (VII) are indicated. The two ape HAV sequences are depicted by an ape symbol. HAV strains from irrigation water in South Africa and from a clam in the Sanaga River in Cameroon are highlighted in bold and depicted by symbols. The tree was visualized and edited using iTOL v6 (Letunic et al. 2024).

Table 4.

Pairwise nucleotide distance (p-distance) of the VP1-2A junction region (517 bp) between the ape HAV sequences and HAV strains representing the diversity of established (sub)genotypes. The pairwise deletion option was applied, and all sites containing gaps were excluded only from the corresponding pairwise comparison. One HAV sequence per (sub)genotype was used. Distance values between the two ape HAV sequences, the Sanaga clam and South African strains, are shown in bold. Evolutionary analyses were conducted in MEGA 12.1 (Stecher et al. 2026). Accession numbers used: IA, AB020564.1; IB, MH685714.1; IIA, MH933768.1; IIB, AY644670.1; IIIA, AB279732.1; IIIB, AB279735.1; IV, M59286.1; V, D00924.1; VI, L07731.1; VII, OR452344.1; Sanaga clam strain, MT185676.1; South African strain, KP875247.1.

	G. g. gorilla	P. t. troglodytes
G. g. gorilla
P. t. troglodytes	0.036
Sanaga clam strain	0.039	0.032
South African strain	0.088	0.074
HAV-V	0.170	0.170
HAV-IV	0.256	0.259
HAV-VI	0.185	0.179
HAV-VII	0.244	0.244
HAV-IIIA	0.258	0.258
HAV-IIIB	0.238	0.235
HAV-IB	0.244	0.24
HAV-IA	0.249	0.242
HAV-IIA	0.227	0.218
HAV-IIB	0.22	0.233

Assessment of the extent of Hepatitis A virus infection in ape communities

To assess the extent of HAV circulation in areas where the two above-described positive animals were identified, 227 faecal samples, including the 2 initial positive samples, were tested at the same collection sites using a newly designed specific PCR system. Overall, 140 gorilla samples collected in the same year (June–October 2006) in the sampling site MM (Cameroon) were tested. In the sampling site MA (Gabon), 87 samples were also tested: 62 chimpanzee samples (46 individuals) and 25 gorilla samples, collected in 2011–13 (Supplementary Table S4). Based on microsatellite analyses previously conducted on a subset of samples from the same biobank (Neel et al. 2010) estimated that each individual western lowland gorilla was represented by an average of 1.8 faecal samples. Applying this sampling ratio to the present dataset suggests that the analysed samples correspond to ~77 and 13 individual gorillas from MM and MA, respectively. HAV was detected and confirmed by Sanger sequencing in 13 samples: 5 chimpanzee samples in Gabon corresponding to 4/46 molecularly confirmed individuals (8.69%; 95% CI: 3.43%–20.32%) as reported in Supplementary Table S4 and 8 gorilla samples in Cameroon, corresponding to at least 4/77 animals (5.2%; 95% CI: 2.04%–12.61%) (Supplementary Table S4). The positive chimpanzees in MA in Gabon were collected at 3 different GPS and/or time points. They are within a radius of ~12 to 15 km from the village, except one, and are detected at 2 different periods (April–October 2011) (Supplementary Fig. S8). A phylogenetic analysis was performed on the 5′ UTR, including the supplementary HAV sequences produced, and the 11 additional ape HAV Sanger sequences clustered with the 2 near-full-length ape HAV sequences and are well-supported, at a distance from genotype V sequences (Fig. 5).

Figure 5.

Genome organization of Hepatovirus ahepa (A) and maximum-likelihood phylogenetic reconstruction (B) of complete 5′ UTR (609 bp). The genome region included in the phylogenetic analysis is shown in panel A. The best-fit substitution model, as determined by ModelFinder in IQ-TREE 3, is TPM2u + F + I + R2. The tree was rooted using a Hepatovirus devoli sequence. Branch supports were estimated by SH-aLRT and UFBoot with 1000 replicates. Only branches with SH-aLRT ≥80% and UFBoot ≥95% were considered well-supported and are shown. The 11 additional HAV ape sequences cover a length of between 196 and 302 bp of the region. Each sequence is labelled with host species, the sampling year (when available), and the country of origin, except for sequences from captive animals for which the country was not indicated. Human genotypes (I–III), simian genotypes (IV–V), and alpaca genotype (VII) are indicated. The strain Cy145 and HAV-VI sequence could not be used because the sequences do not cover this region. The ape HAV sequences are depicted by an ape symbol. The tree was visualized and edited using iTOL v6 (Letunic et al. 2024).

Time-scaled phylogenetic analysis of the full-length Hepatitis A virus genomes

To estimate the divergence date of the characterized HAV genomes, a Bayesian analysis was performed using BEAST. Because recombination can bias molecular clock analyses, a Simplot analysis was performed, and no robust evidence of recombination was detected between the two ape HAV genomes, as reported in Supplementary Fig. S9. These results support the use of the complete genomes for subsequent analyses. A Bayesian phylogeny based on near-full-length HAV sequences was obtained (Fig. 6), including 67 complete HAV sequences representing human and simian HAV genotypes (5 simian sequences and 62 human sequences) and spanning 64 years (1957–2021). The mean nucleotide substitution rate estimated is 1.566 × 10⁻⁴ substitutions/site/year [95% HPD (Highest Posterior Density): 1.2663 × 10⁻⁴–1.8976 × 10⁻⁴]. The phylogeny shows the distinction between human and simian genotypes in two monophyletic clades, as well as the position of the two ape HAV sequences, which are distantly related to genotype V sequences. The MRCA between human and simian HAV sequences is estimated at ~2291 years (95% HPD: 1777–2899), corresponding to 270 bce (95% HPD: −878–243). The MRCA of the simian HAV sequences is estimated at 1900 years (95% HPD: 1375–2569). The MRCA mean of genotype V and ape HAV strains from this study is estimated at 1098 years (95% HPD: 757–1 503), corresponding to the year 923 (95% HPD: 517–1 263). The MRCA between the chimpanzee and gorilla HAV is estimated at 103 years ago (95% HPD: 68, 147), corresponding to the year 1918 (95% HPD: 1873–1952), while for the 2 genotype V strains, it is estimated at 203 years ago (95% HPD: 134, 286).

Figure 6.

Bayesian MCMC phylogenetic reconstruction of near-full-length HAV nucleotide sequences based on 7191 bp. The tree was constructed with BEAST software 10.5.0 (Baele et al. 2025) using a relaxed clock (uncorrelated lognormal) with exponential population growth, the best combination, and one MCMC of 50 000 000 generations. The best substitution model was GTR + I + Γ4. The nucleotide substitution rate estimated is 1.565 × 10⁻⁴ substitutions/site/year (95% HPD: 1.2674 × 10⁻⁴–1.8872 × 10⁻⁴). The support for the branches is indicated by posterior probabilities using a colour gradient. Node labels indicate the estimated calendar year of the MRCA, with 95% HPD intervals shown as bars. The horizontal axis represents time in calendar years. The tree is based on 67 sequences (5 simian sequences and 62 human sequences). Each genotype is indicated on the side. The genotype I clade was collapsed to improve tree visualization. The strain Cy145 and genotype VI sequence could not be used because the sequences available in GenBank are not complete. Genotype VII was not included because the analysis focused on divergence times between human and simian HAV lineages. HAV sequences from NHPs are depicted by an ape symbol, and the box indicates the clade formed by the ape HAV sequences. The tree was visualized and edited using iTOL v6 (Letunic et al. 2024).

Discussion

Detection and phylogenetic characterization of a distinct HAV in ape populations

This study demonstrates for the first time the presence of HAV in wild apes in their natural habitat in Africa. Near-full-length HAV genomes were recovered in faecal samples from one western lowland gorilla (G. g. gorilla) and one central chimpanzee (P. t. troglodytes) ~450 km apart and sampled 5 years apart (Fig. 1). This also marks the first evidence of HAV detection among NHP species in Central Africa; the two other known African HAV strains were isolated from wild-captured vervet monkeys (Chlorocebus spp.) and free-living olive baboons (Papio anubis) in East Africa (Andzhaparidze et al. 1987, Tsarev et al. 1991, Bennett et al. 2016, Dogadov et al. 2018).

Phylogenetic analysis based on complete VP1 protein (Fig. 2), which is commonly used to distinguish HAV genotypes (Costa-Mattioli et al. 2002), shows that the two identified ape HAV sequences form a separate cluster from other HAV genotypes, including the genotype V strains that circulate among African NHP species in East Africa (Andzhaparidze et al. 1987, Tsarev et al. 1991, Bennett et al. 2016, Dogadov et al. 2018). Moreover, the observed divergence in the VP1 (>15.8%) and amino acid signatures (Table 3, Fig. 3) supports the proposal of a supplementary genotype, potentially named HAV-VIII, for the strains described in wild ape species, consistent with the VP1 classification criteria (Costa-Mattioli et al. 2002). However, with regard to the Mecklenburg simian strain, the findings of our analysis based on the intergenotype divergence range appear to contradict the previous conclusion (Mecklenburg et al. 2024) that this strain represents an additionally HAV genotype, and so we suggest it should be reclassified into HAV-IV. It is possible that the Mecklenburg strain and strain Cy145 share a common ancestor that circulated in M. fascicularis populations in Southeast Asia before the introduction of cynomolgus macaques to Mauritius in the 17th century (Sussman and Tattersall 2008).

The ape HAV strains form a well-supported cluster with two other recently described HAV strains obtained from environmental sampling (Fig. 4), i.e. an HAV strain sampled from irrigation water in South Africa in 2013 (Rachida et al. 2016) and another HAV strain identified in clam tissues populating the Sanaga River in Cameroon (Bonny et al. 2020). The HAV strain in clam tissues was detected in samples collected between 2018 and 2019, about 10 kilometres upstream from the Atlantic Ocean, in the western part of the country (Fig. 7). The lower Sanaga watershed is characterized by a mosaic of tropical forest, protected areas, agricultural land, but also cities such as Edéa upstream. Local populations eat these clams, and although they are smoked or boiled before consumption, food contamination cannot be ruled out (Ajonina et al. 2005). As a result, this HAV strain could originate from wildlife, including NHP which live in the watershed, but anthropogenic contamination through human wastewater cannot be excluded (Bonny et al. 2020). The Sanaga River crosses the country from east to west, suggesting a potential circulation of this distinct HAV beyond southern Cameroon.

Figure 7.

Geographic distribution of reported HAV detections in Africa among NHP and environmental samples and those identified in the study. Countries where the divergent HAV lineage has been reported are shown in blue; those with previous reports of simian HAV-V are shown in grey. Triangles indicate the locations where the divergent ape HAV lineage was detected in the present study. The map was created using the free access software QGIS (https://github.com/qgis/QGIS/blob/master/INSTALL.md).

Moreover, the circulation of a closely related strain in irrigation water in southern Africa suggests a large geographic spread of this HAV genotype across the African continent because the geographic range of apes is restricted to Central Africa (east to west) and some isolated populations in West Africa (Fig. 7). The limited size of chimpanzee and gorilla populations, which are critically endangered species (Strindberg et al. 2018), may alone be insufficient to explain the apparent geographic distribution of this HAV genotype across Africa. The recent identification of HAV-VII in alpacas (Veith et al. 2023) further demonstrates that members of the species Hepatovirus ahepa are not exclusively restricted to primates, supporting the possibility that additional, as yet unidentified, animal hosts may exist. Another major question concerns the health implications of such circulation. If this virus is circulating among wild NHP populations, are apes and/or other NHP population densities sufficiently high to contaminate environmental water bodies at levels allowing viral persistence after wastewater treatment and contamination of water destined to be re-used for domestic or agricultural purposes, like in South Africa (Pintó et al. 2012).

The detection of HAV RNA in additional chimpanzee and gorilla stool samples (Fig. 5), which are geographically close, further adds weight to the hypothesis that HAV may infect and persist in ape populations, as supported by the different time periods reported (July and August 2006; April and October 2011). Further studies to detect HAV RNA or antibodies in the serum of apes would be needed to demonstrate infection. Some of these samples are collected near villages, and potential spillover events to humans cannot be ruled out. The detection of this divergent genotype in the environment could also suggest that this HAV genotype is already present in human populations in Africa and has not yet been reported because sequencing is still limited among samples that are positive in diagnostic tests. Although this divergent HAV was detected in only two regions of Central Africa, the virus could also circulate in other areas as suggested by previous environmental detections, including those investigated in the pilot study, but at viral loads too low to be detected by NGS. This can partly explain the lack of HAV detection in bonobos, particularly because they were less well represented in the pilot study than chimpanzees and gorillas.

Time-scaled analysis of HAV lineages

The Bayesian time-scaled analysis (Fig. 6) shows a clear separation between human and simian HAV genotypes, with the MRCA estimated at ~2291 years, consistent with previous studies based on a smaller dataset (Kulkarni et al. 2009). Within the limitations of the dataset, the estimated divergence among simian HAV lineages (1900 years) appears to predate the diversification of human HAVs and may be consistent with a long-term HAV circulation in NHP. This idea is consistent with broader phylogenetic data assuming that HAV-like viruses may have emerged even earlier in small mammals, particularly rodents, where viral diversity is significantly greater (Drexler et al. 2015). However, the direction and frequency of host transitions cannot be inferred from the present data alone. Reconstruction of ancestral host states would be required to better resolve the evolutionary relationships between human and simian lineages. In addition, the relatively recent MRCA estimated for the ape HAV strains (110 years) suggests recent diversification within this lineage, in contrast to the older divergence observed for genotype V strains. The lack of congruence between host and viral phylogenies suggests that HAV evolution cannot be explained only by host co-divergence and is more consistent with a history involving cross-species transmission events. Nevertheless, the divergence times inferred in the present study are probably underestimated because the substitution rate (1.565 × 10⁻⁴ substitutions/site/year) was estimated from sequences sampled over only 62 years. Because of the time-dependent rate phenomenon, substitution rates estimated over short time scales tend to be higher than long-term evolutionary rates, which may lead to an underestimation of deeper divergence times (Ho et al. 2011, Aiewsakun and Katzourakis 2016). The severe undersampling of historical samples makes the analysis highly sensitive to which historical sequences happen to be available and so including older genomes may therefore refine estimates of substitution rates and divergence times.

This study revealed, for the first time, the presence of HAV in two ape species in Central Africa in their natural habitat. Moreover, the ape HAV strains represent a supplementary HAV genotype, HAV-VIII, which is also detected in irrigation water in South Africa and clam tissues in Cameroon. Further studies on HAV prevalence and genetic diversity across different NHP species over a broader geographic range are needed to better understand HAV diversity, which may be more extensive than that found in humans and remains largely under-sampled. This is essential not only for refining evolutionary scenarios for this virus but also for better understanding current zoonotic risk to local human populations. Such an assessment is currently complicated by the lack of genotype confirmation of diagnostic samples from humans; we therefore cannot be certain that the currently known human genotypes reflect the entire genetic diversity of HAV infecting human populations. It is also essential to study HAV prevalence and genetic diversity in human populations living close to infected wildlife in more detail to evaluate past and/or recent cross-species transmissions of HAV between humans and NHP.

Conflicts of interest

None declared.

Funding

This work was supported in part by grants from the National Institute of Health (RO1 AI 50529), the Agence Nationale de Recherches sur le SIDA (ANRS 12125/12182/12325), by the Centre International de Recherches Médicales de Franceville (CIRMF) and the Institut de Recherche pour le Développement (IRD).

Data availability

Final consensus sequences and partial sequences have been deposited in the NCBI/GenBank database under accession numbers PZ362759-PZ362771. Partial genomes generated by NGS are provided as supplementary data. Raw sequencing reads are available from the lead contact upon request.