APOBEC3 deaminase editing in mpox virus as evidence for sustained human transmission since at least 2016

Abstract

Historically, mpox has been characterized as an endemic zoonotic disease that transmits through contact with the reservoir rodent host in West and Central Africa. However, in May 2022, human cases of mpox were detected spreading internationally beyond countries with known endemic reservoirs. When the first cases from 2022 were sequenced, they shared 42 nucleotide differences from the closest mpox virus (MPXV) previously sampled. Nearly all these mutations are characteristic of the action of APOBEC3 deaminases, host enzymes with antiviral function. Assuming APOBEC3 editing is characteristic of human MPXV infection, we developed a dual-process phylogenetic molecular clock that-inferring a rate of ~6 APOBEC3 mutations per year-estimates that MPXV has been circulating in humans since 2016. These observations of sustained MPXV transmission present a fundamental shift to the perceived paradigm of MPXV epidemiology as a zoonosis and highlight the need for revising public health messaging around MPXV as well as outbreak management and control.

Fig. 1.. Specific enrichment of APOBEC3-type mutations in MPXV samples collected since 2017.

(A) MPXV genetic diversity is categorised into Clade I (predominantly sequences from the DRC), Clade IIa (predominantly West African sequences) and Clade IIb. Within Clade IIb is a sub-clade of genomes sampled from 2017–2022 that show distinct mutational patterns to the other two clades. (B) We catalogue single nucleotide mutations across the phylogenies of Clade IIb, Clade IIa and Clade I (top to bottom). For Clade IIb, we include samples from 2017–2022 and only a single representative of the global lineage B.1. Of 120 reconstructed mutations that occurred on internal branches of the Clade IIb phylogeny (so are observed transmitted mutations), 109 are consistent with APOBEC3 editing (90.8% of mutations). Individual proportions of G→A and C→T mutations shown above the respective bars. Ancestral state reconstruction performed across Clade IIa and Clade I does not produce the same enrichment of mutations consistent with APOBEC3 editing, with only 27 of 207 observed mutations (13%) and 38 of 463 Clade I mutations (8%) fitting the dinucleotide pattern. (C) Observed heptamers of C→T or G→A mutated sites of Clade IIb, IIa and I phylogenies (top to bottom). Heptamers associated with G→A mutations have been reverse-complemented to reflect deamination on the negative strand. For Clade IIb, most C→T mutations are present in a TC dimer context, consistent with APOBEC3 editing (107 of 115 mutations, or 93%). However, the same is not seen for Clades IIa and I, in which 29 of 149 (19%) mutations and 42 of 256 (16%) have the dinucleotide context of APOBEC3 respectively, which is what we would predict under standard models of nucleotide evolution. *Only mutations occurring on internal branches of the Clade IIb phylogeny included.

Fig. 2.. Observed APOBEC3-type mutations are not merely a product of available target sites.

(A). Consequence of hypothetical APOBEC3 mutations at target dimer site (either the C in the TC target site or G in the GA target site) in the coding regions of the NCBI reference MPXV genome for Clade II (Accession NC_063383) and those observed APOBEC3 mutations across the coding regions of the Clade IIb phylogeny (not including the outgroup branch leading to the 1971 genome sequence). These are categorised into non-synonymous (altered amino acid), synonymous (amino acid remaining unchanged), nonsense (editing producing a stop codon) and intergenic (not present in a coding sequence). (B) The proportion of target sites edited for each target site percentile window across the MPXV genome. The teal shaded regions represent the binomial confidence interval around observations. Masked regions indicated by vertical grey bars. Observed edits include data from the Clade IIb phylogeny, with a single representative of lineage B.1 and not including the branch leading to the outgroup 1971 genome sequence. (C) Hypothetical amino acid changes for codons overlapping with TC and GA target sites in a reference MPXV genome (Genbank accession number: NC_063383) if APOBEC3 edited those dimers to TT and AA. Amino acid changes are coloured by Grantham Score (0–50 conservative: dark blue; 51–100 moderately conservative: light blue; 101–150 moderately radical: light red; >150 radical: dark red; synonymous: grey).

Fig. 3.. Estimating the time of MPXV emergence into the human population from the accumulation of APOBEC3-type mutations.

(A) MPXV genomes sampled from human infections from 2017–2022, with an outgroup sequence from an outbreak in Nigeria in 1971 (n=44 including outgroup). Lineages indicated as per nomenclature proposed by Happi et al., (7). Mutations along each branch are indicated with circles coloured by whether it is putatively APOBEC3 edited (TC→TT and GA→AA; red) or whether it is another mutation type (yellow). The break in the basal branch illustrates the assumption made in the regression model in panel B, that all APOBEC3 mutations occurred after emerging into the human population, however we do not know the precise distribution of red or yellow mutations. (B) APOBEC3 mutations from MPXV genomes sampled since 2017. The reconstructed most recent common ancestor (MRCA) of the panel A phylogeny is used as the root in the root-to-tip plot and the y-intercept is used as a proxy for time of emergence which is inferred by fitting a Bayesian regression to the sequence dates from panel A. Intersects with y=1, 2, & 3 are also shown as it is likely that a small number of the APOBEC3-type mutations are actually earlier replication errors and not induced by APOBEC3. (C) Maximum Clade Credibility (MCC) phylogeny of MPXV Clade IIb with absolute time shown on the X-axis. We separated the alignment into an APOBEC3 and a non-APOBEC3 partition and modelled the substitution process in each independently. We used an epoch model with two outgroup sequences (not shown in panel) representing the first epoch and hMPXV-1 ingroup sequences representing MPXV post-emergence into the human population with an exponential growth model. The probability density distributions show the estimated time of the most recent common ancestor (tMRCA) of the ingroup as well as the estimated transition time that represents the time of emergence into the human population. (D) Estimated effective population size of the outbreak using a non-parametric coalescent Skygrid model with 11 change points over a period of 8.5 years. This reconstruction falls within the bounds of the exponential growth model estimated from the second epoch in panel C, suggesting that the MPXV population has been exponentially growing since at least 2016.