The transmission dynamics and diversity of human metapneumovirus in Peru

Background The transmission dynamics of human metapneumovirus (HMPV) in tropical countries remain unclear. Further understanding of the genetic diversity of the virus could aid in HMPV vaccine design and improve our understanding of respiratory virus transmission dynamics in low‐ and middle‐income countries. Materials & Methods We examined the evolution of HMPV in Peru through phylogenetic analysis of 61 full genome HMPV sequences collected in three ecologically diverse regions of Peru (Lima, Piura, and Iquitos) during 2008‐2012, comprising the largest data set of HMPV whole genomes sequenced from any tropical country to date. Results We revealed extensive genetic diversity generated by frequent viral introductions, with little evidence of local persistence. While considerable viral traffic between non‐Peruvian countries and Peru was observed, HMPV epidemics in Peruvian locales were more frequently epidemiologically linked with other sites within Peru. We showed that Iquitos experienced greater HMPV traffic than the similar sized city of Piura by both Bayesian and maximum likelihood methods. Conclusions There is extensive HMPV genetic diversity even within smaller and relatively less connected cities of Peru and this virus is spatially fluid. Greater diversity of HMPV in Iquitos compared to Piura may relate to higher volumes of human movement, including air traffic to this location.


| INTRODUCTION
Human metapneumovirus (HMPV) causes substantial morbidity particularly in the extremes of age and in the immunocompromised. 1 Despite the burden of disease due to this pathogen, there is a limited knowledge of the diversity and transmission dynamics of this virus in Peru and other tropical countries due to the low availability of molecular data. 2 Further understanding of the genetic diversity of the virus could aid in HMPV vaccine design and improve our understanding of respiratory virus transmission dynamics in low-and middle-income countries (LMIC). We therefore examined the evolution of HMPV in Peru through phylogenetic analysis of 61 full-genome HMPV sequences collected in three ecologically diverse regions of Peru (Lima, Piura, and Iquitos) during 2008-2012, comprising the largest dataset of HMPV whole genomes sequenced from any tropical country to date (Table 1). Using these data, we assessed (i) the extent of genetic diversity in each location, (ii) the extent of gene flow of HMPV between Peru and other regions globally, and (iii) whether viruses from three ecologically diverse Peruvian cities with distinctly different seasonal patterns of influenza-like illnesses were more similar to each other than from the global gene pool.

| Study population, fieldwork, and initial laboratory procedures
Study respiratory specimens were derived from influenza-like illness (ILI) sentinel and active surveillance studies in three Peruvian cities between 2008 and 2012 which yielded 87 HMPV PCR-positive specimens. Lima (n = 15 specimens) is the temperate urban capital and major air-hub of Peru, with a large population (~8.5 million). Lima receives >4 million international flights annually, and is a hub of transportation within Peru (Table 1) (3). Piura (n = 34 specimens) is a semi-arid regional capital of northern Peru with a fraction of Lima's population (377 000) and greatly reduced volume of air traffic (Table 1). 3 Iquitos (n = 38 specimens) is similar to Piura in terms of demography (population = 370 000), but is located deep in the Amazonian tropics and accessible only by air or river. 3 Previous studies have demonstrated that ILI activity peaks in winter in Lima, but is less seasonal in Piura and Iquitos, which tend to have year-round incidence (Y.Tinoco unpublished data). The field and laboratory methods of the surveillance studies yielding this biobank of HMPV-positive specimens from these three cities are described in full detail elsewhere. 2,4,5 Briefly, those presenting to healthcare settings in Lima (Hospital Nacional Daniel Alcides Carrion and Hospital Edgardo Rebagliati), Piura (Pachitea Health Center), and Iquitos (Hospital Regional, Hospital Apoyo, Quistococha Health Center, Tupac Amaru Health Center, San Antonio Health Center, Moronacocha Health Center and Bellavista-Nanay Health Center) were screened for study enrollment and included if they met criteria for influenza-like-illness, defined as fever 38°C plus either cough or sore throat. Naso-oropharyngeal swabs were selected from eligible participants, in addition to data on age, sex, and whether the patient was managed as an inpatient or outpatient at the time of collection. Importantly, there was heterogeneity in specimen collection by location with a highly engaged collaborator who provided a relatively large number of respiratory samples for Piura.
In addition to sentinel surveillance in each of the three Peruvian locations, specimens were also leveraged from active influenza-like illness surveillance in Iquitos as described by Forhsey et al. 5 Briefly, enrolled community-based participants were visited at their residence 3 times per week by healthcare workers who elicited whether participants had ILI as per the definition used for sentinel surveillance, with naso-oropharyngeal swabs and demographic data collected from eligible participants.
These specimens were sent in viral transport media at −70°C Original naso-oropharyngeal swab specimens which were positive for HMPV were then biobanked.  Table S3) from primers designed across a consensus of complete HMPV genome sequences using JCVI's automated primer design tool. 6 The four forward RT primers were diluted to 2 μmol/L and pooled in equal volumes. cDNA was generated from 4 μL un-  was subjected to manual inspection and quality control before submission to GenBank. All sequences generated as part of this study were submitted to GenBank as part of the Bioproject ID PRJNA237298.

| Evolutionary analyses
To understand how HMPV in Peru relates to viral populations sampled globally, a global background dataset of HMPV F-gene sequences was  Studying the spatial dynamics of HMPV in Peru is complicated by the low number of samples available from Lima, the largest and most interconnected city in the country (Table 1). Therefore, we examined only broad spatial patterns in our phylogeographic analysis, in which we reconstructed the spatial dynamics of HMPV using a discrete phy-

| Ethical considerations
The Naval Medical Research Unit Number-6 Institutional Review Board approved the field studies that yielded the HMPV-positive respiratory specimens.

| RESULTS
Of the 87 HMPV-positive specimens, 61 were able to be success- Of the four major HMPV lineages defined globally (A1, A2, B1, and B2 subclades), all but A1 viruses were identified in Peru ( Figure 1, Table 2, Figures S1, S2, and S3 present trees for these A2, B1, and B2 HMPV clades, respectively. Figure S4 displays the global phylogeny with all labeled taxa and bootstrap annotations). The A2, B1, and B2 subclades were observed even in the more isolated locales of Iquitos and Piura, and several clusters with high bootstrap support were identified that included viruses from all 3 Peruvian cities, an indication of intracountry movement of viruses between locations with markedly different climates.
We found that HMPV is frequently introduced into each of the 3 Peruvian localities, including undersampled Lima, based on the estimated number of location transitions ("Markov jumps") on the trees that were estimated using stochastic mapping techniques ( Table 2).
Estimates of viral gene flow into a location were not related to the number of sequences available from that location. In fact, the lowest viral gene flow was observed in Piura (Table 2), for which the largest number of sequences was available ( Table 1). The phylogeographic analysis also suggested that, overall, HMPV epidemics in Peru were more likely to be seeded from other locations in Peru rather than from a global gene pool (Table 2). These findings were consistent with the F I G U R E 1 Maximum likelihood inferred phylogeny of Peru (red) and reference global HMPV F-gene sequences (black), rooted by avian pneumovirus group C outgroup (GenBank accession AY579780, removed for clarity). Scale bar indicates number of nucleotide substitutions per site, and major clade identities are indicated in red (A1, A2, B1, B2). All taxa labels and bootstrap values are indicated in Figure S4 phylogeographic results on the F-gene dataset (Table 2), as well as the number of introductions estimated from the ML tree (Table 2). A higher degree of geographic clustering also was observed Piura compared to Iquitos, as assessed using the phylogeny-trait association test

| CONCLUSIONS
By sequencing 61 HMPV specimens in Peru, we have generated the largest dataset of whole-genome sequences from any tropical country for this important respiratory pathogen. Importantly, the case characteristics yielding these data (overall skew toward pediatric age-groups who are frequently managed on an ambulatory basis rather than requiring inpatient admission) are similar to those noted in HMPV study in other regions 1 and were derived from both populations presenting to health care as well as those actively identified in the community via active surveillance, thereby improving the generalizability of our results. Notably, the genetic diversity of HMPV in Peru spans almost the entire known global HMPV diversity, due to widespread gene flow within Peru and between Peru and other regions. The lack of data available from Lima has limited our ability to infer viral movements within Peru at a refined scale, and we suspect that increased sequencing in Lima will be central to understanding HMPV dynamics in Peru, given Lima's volume of domestic and international air traffic. This geospatially skewed data due to ascertainment bias (heterogeneity in specimen collection which doesn't correlate with disease incidence) are an unfortunately common limitation of many spatial phylodynamic studies of other respiratory RNA viruses. 13 We tried to explicitly account for this spatial skew and ascertainment bias in our analysis. Specifically, we did not make any specific conclusions of the  we estimate the absolute viral traffic in Lima or compare it to Piura and Iquitos.
Still, our conservative approach allowed us to identify extensive genetic diversity even within the smaller and relatively less connected cities. We further showed that there is limited persistence of HMPV in tropical locations and that this virus is spatially fluid. Furthermore, we demonstrate that Iquitos experiences greater HMPV traffic than the similarly sized city of Piura, which may relate to greater volumes of air traffic through Iquitos (Table 1), and indicates there is some correlation between human movement and HMPV dispersal, at least in these two locations that were well-sampled. The role of human movement in putatively driving greater than expected viral traffic through a locale has also been noted in recent phylogeographic studies of H3N2 influenza A virus in Peru. 14 The general lack of sustained persistence of HMPV in any of the three locations across years indicates that that viral diversity in any locale in Peru is likely maintained by continual migration and reintroduction (from either within or outside of Peru). This has also been demonstrated for both influenza and respiratory syncytial virus (RSV) epidemics in Peru. 14,15 While epidemics in discrete Peruvian locales were shown to be more likely to be seeded from elsewhere in Peru than a non-Peruvian country, a caveat is that countries immediately