Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Jan 16:7:e31257.
doi: 10.7554/eLife.31257.

MERS-CoV spillover at the camel-human interface

Gytis Dudas  1 , Luiz Max Carvalho  2 , Andrew Rambaut  2   3 , Trevor Bedford  1
Affiliations

MERS-CoV spillover at the camel-human interface

Gytis Dudas et al. Elife. .

Erratum in

Abstract

Middle East respiratory syndrome coronavirus (MERS-CoV) is a zoonotic virus from camels causing significant mortality and morbidity in humans in the Arabian Peninsula. The epidemiology of the virus remains poorly understood, and while case-based and seroepidemiological studies have been employed extensively throughout the epidemic, viral sequence data have not been utilised to their full potential. Here, we use existing MERS-CoV sequence data to explore its phylodynamics in two of its known major hosts, humans and camels. We employ structured coalescent models to show that long-term MERS-CoV evolution occurs exclusively in camels, whereas humans act as a transient, and ultimately terminal host. By analysing the distribution of human outbreak cluster sizes and zoonotic introduction times, we show that human outbreaks in the Arabian peninsula are driven by seasonally varying zoonotic transfer of viruses from camels. Without heretofore unseen evolution of host tropism, MERS-CoV is unlikely to become endemic in humans.

Keywords: MERS; coronavirus; epidemiology; evolutionary biology; genomics; global health; phylodynamics; phylogenetics; structured coalescent; virus; zoonosis.

PubMed Disclaimer

Conflict of interest statement

GD, LC, AR, TB No competing interests declared

Figures

Figure 1.
Figure 1.. Typed maximum clade credibility tree of MERS-CoV genomes from 174 human viruses and 100 camel viruses.
Maximum clade credibility (MCC) tree showing inferred ancestral hosts for MERS-CoV recovered with the structured coalescent. The vast majority of MERS-CoV evolution is inferred to occur in camels (orange) with human outbreaks (blue) representing evolutionary dead-ends for the virus. Confidence in host assignment is depicted as a colour gradient, with increased uncertainty in host assignment (posterior probabilities close to 0.5) shown as grey. While large clusters of human cases are apparent in the tree, significant contributions to human outbreaks are made by singleton sequences, likely representing recent cross-species transmissions that were caught early.
Figure 1—figure supplement 1.
Figure 1—figure supplement 1.. Evolutionary history of MERS-CoV partitioned between camels and humans.
This is the same tree as shown in Figure 1, but with contiguous stretches of MERS-CoV evolutionary history split by inferred host: camels (top in orange) and humans (bottom in blue). This visualisation highlights the ephemeral nature of MERS-CoV outbreaks in humans, compared to continuous circulation of the virus in camels.
Figure 1—figure supplement 2.
Figure 1—figure supplement 2.. Posterior backwards migration rate estimates for two choices of prior.
Negligible flow of MERS-CoV lineages from humans into camels is recovered regardless of prior choice (note that rates are backwards in time). Plots show the 95% highest posterior density for the estimated migration rate from the human deme into the camel deme looking backwards in time (orange) and vice versa (blue). Dotted lines indicate exponential priors specified for migration rates, with mean 1.0 (bottom) or 10.0 (top).
Figure 1—figure supplement 3.
Figure 1—figure supplement 3.. Maximum clade credibility (MCC) tree with ancestral state reconstruction according to a discrete trait model.
MCC tree is presented the same as Figure 1 and Figure 1—figure supplement 4, with colours indicating the most probable state reconstruction at internal nodes. Unlike the structured coalescent summary shown in Figure 1 where camels are reconstructed as the main host where MERS-CoV persists, the discrete trait approach identifies both camels and humans as major hosts with humans being the source of MERS-CoV infection in camels.
Figure 1—figure supplement 4.
Figure 1—figure supplement 4.. Maximum clade credibility (MCC) tree of structured coalescent model with enforced equal coalescence rates.
MCC tree is presented the same as Figure 1 and Figure 1—figure supplement 3, with colours indicating the most probable state reconstruction at internal nodes. Similar to Figure 1—figure supplement 3 enforcing equal coalescence rates between demes in a structured coalescent model identifies humans as a major MERS-CoV host and the source of viruses in camels.
Figure 1—figure supplement 5.
Figure 1—figure supplement 5.. Maximum likelihood (ML) tree of MERS-CoV genomes coloured by origin of sequence.
Maximum likelihood tree shows genetic divergence between MERS-CoV genomes collected from camels (orange tips) and humans (blue tips).
Figure 2.
Figure 2.. Seasonality of MERS-CoV introduction events.
(A) Posterior density estimates partitioned by month showing the 95% highest posterior density interval for relative odds ratios of MERS-CoV introductions into humans. Posterior means are indicated with circles. Evidence for increased or decreased risk (95% HPD excludes 1.0) for introductions are indicated by black or white circles, respectively. Hatched area spanning October to February indicates the camel calving season. (B) Sequence cluster sizes and inferred dates of introduction events. Each introduction event is shown as a vertical line positioned based on the median introduction time, as recovered by structured coalescent analyses and coloured by time of year with height indicating number of descendant sequences recovered from human cases. 95% highest posterior density intervals for introductions of MERS-CoV into humans are indicated with coloured lines, coloured by median estimated introduction time. The black dotted line indicates the joint probability density for introductions. We find little correlation between date and size of introduction (Spearman ρ=0.06, p=0.68).
Figure 3.
Figure 3.. Monte Carlo simulations of human transmission clusters.
Leftmost scatter plot shows the distribution of individual Monte Carlo simulation sequence cluster size statistics (mean and skewness) coloured by the R0 value used for the simulation. The dotted rectangle identifies the 95% highest posterior density bounds for sequence cluster size mean and skewness observed for empirical MERS-CoV data. The distribution of R0 values that fall within 95% HPDs for sequence cluster size mean, standard deviation, skewness and number of introductions, is shown in the middle, on the same y-axis. Bins falling inside the 95% percentiles are coloured by R0, as in the leftmost scatter plot. The distribution of total number of introductions associated with simulations matching MERS-CoV sequence clusters is shown on the right. Darker shade of grey indicates bins falling within the 95% percentiles. Monte Carlo simulations indicate R0 for MERS-CoV in humans is likely to be below 1.0, with numbers of zoonotic transmissions numbering in the hundreds.
Figure 3—figure supplement 1.
Figure 3—figure supplement 1.. Monte Carlo simulations of human transmission clusters.
From top to bottom each row corresponds to departures from completely random sequencing efforts with respect to case cluster size (bias parameter = 1.0) to sequencing increasingly biased towards capturing large case clusters (bias = 2.0, bias = 3.0). Leftmost scatter plots show the distribution of individual Monte Carlo simulation sequence cluster size statistics (mean and skewness) coloured by the R0 value used for the simulation. The dotted rectangle identifies the 95% highest posterior density bounds for sequence cluster size mean and skewness observed for empirical MERS-CoV data. The distribution of R0 values matching empirical data are shown in the middle, on the same y-axis across all levels of the bias parameter. Under unbiased sequencing (bias = 1.0) only 0.45% of simulations fit our phylogenetic observations, while 1.79% and 1.67% of simulations fit for bias levels of 2.0 and 3.0, respectively. Correspondingly, we estimate 11.6% support for a model with bias level 1.0, 45.7% support for a model with bias level 2.0, and 42.7% support for a model with bias level 3.0. Bins falling inside the 95% percentiles are coloured by R0, as in the leftmost scatter plot. While the 95% percentiles for R0 values are close to 1.0 (0.71–0.98) for the unbiased sequencing simulation (i.e. uniform sequencing efforts, in which every case is equally likely to be sequenced), we also note that increasing levels of bias are considerably more to likely to generate MERS-CoV-like sequence clusters. The distribution of total number of introductions associated with simulations matching MERS-CoV sequence clusters is shown in the plots on the right, on the same y-axis across all levels of bias. Darker shade of grey indicates bins falling within the 95% percentiles. The median number of cross-species introductions observed in simulations matching empirical data without bias are 346 (95% percentiles 262–439). These numbers jump up to 568 (95% percentiles 430–727) for bias = 2.0 and 656 (95% percentiles 488–853) for bias = 3.0 simulations. Model averaging would suggest plausible numbers of introductions between 311 and 811.
Figure 3—figure supplement 2.
Figure 3—figure supplement 2.. Monte Carlo simulation schematic.
Case clusters are simulated according to Equation 1 until an outbreak size of 2000 cases is reached. We sample 174 cases from each simulation to represent sequencing of human MERS cases. ‘Sequencing’ is carried out by using multivariate hypergeometric sampling, representing sampling cases without replacement to be sequenced. Sequencing simulations take place at three levels of bias: 1.0, where every case is equally likely to be sequenced, and 2.0 and 3.0, where cases from larger clusters are increasingly more likely to be sequenced. The distribution of simulated sequence clusters is summarised by its mean, median and standard deviation. A simulation is considered to match if the mean, median and standard deviation of its sequence cluster sizes falls within the 95% highest posterior density interval of observed MERS-CoV sequence clusters. R0 values that ultimately generate data matching empirical observations, as well as associated numbers of ‘introductions’ are retained as estimates. These estimates are summarised in Figure 3.
Figure 3—figure supplement 3.
Figure 3—figure supplement 3.. Results of Monte Carlo simulations with vast underestimation of cases.
The plot is identical to Figure 3—figure supplement 1, but instead of 2000 cases, simulations were run with 4000 cases. With more unobserved cases the R0 values matching observed MERS-CoV sequence clusters can only be smaller, with a corresponding increase in numbers of zoonotic transmissions. However, the numbers of simulations that match MERS-CoV data go down as well.
Figure 3—figure supplement 4.
Figure 3—figure supplement 4.. Boxplots of matching simulated case and sequence cluster distributions.
Boxplots indicate frequency of case (blue, top) and sequence (red, bottom) cluster sizes across simulations at different bias levels, marginalised across R0 values. Outliers are shown with transparency, medians are indicated with thick black lines. Case clusters exhibit a strong skew with large numbers of singleton introductions and a substantial tail at higher levels of bias.
Figure 3—figure supplement 5.
Figure 3—figure supplement 5.. Quantile-quantile (Q-Q) plot of empirical and simulated sequence cluster sizes.
Density of sequence cluster size percentiles (1 st–99th, calculated across a grid of 50 values) calculated for random states from the posterior distribution (x-axis) and matching simulations (y-axis). Most values fall on the one-to-one line, with a heavier tail in mid-sized sequence clusters in empirical data, manifesting as a greater density of points below the one-to-one line in the middle.
Figure 3—figure supplement 6.
Figure 3—figure supplement 6.. Numbers of epidemiological simulations conforming to empirical observations.
Numbers indicate the total number of epidemiological simulations under each combination of bias and dispersion parameter ω that result in MERS-CoV-like sequence cluster sizes. More simulations match observations with bias >1 and ω≈0.1.
Figure 4.
Figure 4.. Recombinant features of MERS-CoV phylogenies.
(A) Marginal posterior probabilities of taxa collected from humans belonging to the same clade in phylogenies derived from different parts of the genome. Taxa are ordered according to phylogeny of fragment 2 (genome positions 21001 to 29364) reduced to just the human tips and is displayed on the left. Human clusters are largely well-supported as monophyletic and consistent across trees of both genomic fragments. (B) Tanglegram connecting the same taxa between a phylogeny derived from fragment 1 (left, genome positions 1 to 21000) and fragment 2 (right, genome positions 21001 to 29364), reduced to just the human tips and branches with posterior probability <0.1 collapsed. Human clusters exhibit limited diversity and corresponding low levels of incongruence within an introduction cluster.
Figure 4—figure supplement 1.
Figure 4—figure supplement 1.. Tests of recombination across MERS-CoV clades.
Maximum clade credibility tree of MERS-CoV genomes annotated with results of two recombination detection tests (PHI and 3Seq) applied to descendent sequences of each clade. Both tests identify large portions of existing sequence data as containing signals of recombination. Note that markings do not indicate where recombinations have occurred on the tree, merely the minimum distance in sequence/time space between recombining lineages.
Figure 4—figure supplement 2.
Figure 4—figure supplement 2.. MERS-CoV genomes exhibit high numbers of non-clonal loci.
Ancestral state reconstruction (right) identifies a large number of sites in which mutations have occurred more than once in the tree (homoplasies, orange) or are reversions (red) from a state arising in an ancestor. Mutations that apparently only occur once in the tree (synapomorphies) are shown in grey. The maximum likelihood phylogeny on the left is coloured by whether sequences were sampled in humans (blue) or camels (orange).
Figure 4—figure supplement 3.
Figure 4—figure supplement 3.. Human clade sharing between genomic fragments 1 and 2.
Central scatter plot shows the posterior probability of human clades shared between genomic fragments 1 and 2, in their respective trees. Left and bottom scatter plots track the posterior probability of human clades only observed in fragment 2 (left) or fragment 1 (bottom). The cumulative probability of human clades present in either tree are tracked by plots on the right (fragment 2) and top (fragment 1). Most of the probability mass is concentrated within human clades that are present in trees of both genomic fragment 1 and 2 (0.9701 and 0.9474 of all human clades across posteriors, respectively).
Figure 5.
Figure 5.. Demographic history of MERS-CoV in Arabian peninsula camels.
Demographic history of MERS-CoV in camels, as inferred via a skygrid coalescent tree prior (Gill et al., 2013). Three skygrid reconstructions are shown, red and orange for each of the stationary distributions reached by MCMC with the whole genome and a black one where the genome was split into ten partitions. Shaded interval indicates the 95% highest posterior density interval for the product of generation time and effective population size, Ne⁢τ. Midline tracks the inferred median of Ne⁢τ.
Figure 5—figure supplement 1.
Figure 5—figure supplement 1.. Skygrid comparison between whole and fragmented genomes.
Inferred median Ne⁢τ recovered using a skygrid tree prior on whole genome (bottom) and ten genomic fragments with independent trees (left), coloured by time. Dotted line indicates the one-to-one line.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Abdallah H, Faye B. Typology of camel farming system in Saudi Arabia. Emirates Journal of Food and Agriculture. 2013;25:250. doi: 10.9755/ejfa.v25i4.15491. - DOI
    1. Adney DR, van Doremalen N, Brown VR, Bushmaker T, Scott D, de Wit E, Bowen RA, Munster VJ. Replication and shedding of MERS-CoV in upper respiratory tract of inoculated dromedary camels. Emerging Infectious Diseases. 2014;20:1999–2005. doi: 10.3201/eid2012.141280. - DOI - PMC - PubMed
    1. Ali MA, Shehata MM, Gomaa MR, Kandeil A, El-Shesheny R, Kayed AS, El-Taweel AN, Atea M, Hassan N, Bagato O, Moatasim Y, Mahmoud SH, Kutkat O, Maatouq AM, Osman A, McKenzie PP, Webby RJ, Kayali G, Systematic KG. Systematic, active surveillance for Middle East respiratory syndrome coronavirus in camels in Egypt. Emerging Microbes & Infections. 2017;6:e1. doi: 10.1038/emi.2016.130. - DOI - PMC - PubMed
    1. Almutairi SE, Boujenane I, Musaad A, Awad-Acharari F. Non-genetic factors influencing reproductive traits and calving weight in Saudi camels. Tropical Animal Health and Production. 2010;42:1087–1092. doi: 10.1007/s11250-010-9529-y. - DOI - PubMed
    1. Antia R, Regoes RR, Koella JC, Bergstrom CT. The role of evolution in the emergence of infectious diseases. Nature. 2003;426:658–661. doi: 10.1038/nature02104. - DOI - PMC - PubMed

Publication types

MeSH terms