key: cord-0012609-q8w3ii6i
authors: Bennett, Andrew J.; Goldberg, Tony L.
title: Pteropine Orthoreovirus in an Angolan Soft-Furred Fruit Bat (Lissonycteris angolensis) in Uganda Dramatically Expands the Global Distribution of an Emerging Bat-Borne Respiratory Virus
date: 2020-07-09
journal: Viruses
DOI: 10.3390/v12070740
sha: fe58914352af35cde4236cea9b87a87e5763abb1
doc_id: 12609
cord_uid: q8w3ii6i

Pteropine orthoreovirus (PRV; Reoviridae: Spinareovirinae) is an emerging bat-borne zoonotic virus that causes influenza-like illness (ILI). PRV has thus far been found only in Australia and Asia, where diverse old-world fruit bats (Pteropodidae) serve as hosts. In this study, we report the discovery of PRV in Africa, in an Angolan soft-furred fruit bat (Lissonycteris angolensis ruwenzorii) from Bundibugyo District, Uganda. Metagenomic characterization of a rectal swab yielded 10 dsRNA genome segments, revealing this virus to cluster within the known diversity of PRV variants detected in bats and humans in Southeast Asia. Phylogeographic analyses revealed a correlation between geographic distance and genetic divergence of PRVs globally, which suggests a geographic continuum of PRV diversity spanning Southeast Asia to sub-Saharan Africa. The discovery of PRV in an African bat dramatically expands the geographic range of this zoonotic virus and warrants further surveillance for PRVs outside of Southeast Asia.

Pteropine orthoreovirus (PRV; Reoviridae: Spinareovirinae) is a bat-borne zoonosis that causes upper respiratory tract infection (URTI), ranging from mild respiratory symptoms to influenza-like illness (ILI), in Southeast Asia [1] [2] [3] [4] . PRV is a fusogenic orthoreovirus with 10 dsRNA segments (S1-4, M1-3, L1-3) [5] . PRV is only distantly related to other mammalian orthoreoviruses (MRV), and is nested within a clade of fusogenic viruses that cause significant pathology in reptiles and birds [5] . First isolated in 1968 from a flying fox (Pteropus poliocephalus) in Australia [6] , variants of PRV have since been detected in diverse pteropodid bats across Asia [7] [8] [9] [10] and in residents of, and travelers to, Southeast Asia [1, [11] [12] [13] . Here, we describe the discovery of PRV in a bat from Equatorial Africa and discuss the implications of a dramatically expanded PRV geographic distribution.

The reoviruses (respiratory enteric orphan viruses; family Reoviridae) contain non-enveloped, segmented dsRNA viruses that cause distinct pathologies in diverse hosts [14] . The orthoreovirus genus is split into fusogenic and nonfusogenic viruses, determined by their capacity to form multinucleated syncytia mediated by uniquely simple fusion-associated small transmembrane (FAST) proteins for which there are no known homologs elsewhere in nature [15] . For nearly 30 years, the original variant of PRV, Nelson Bay virus (PRV1NB), remained a singular curiosity as the lone mammalian fusogenic orthoreovirus [16] . In June and July of 2017, we sampled Angolan soft-furred fruit bats (Lissonycteris angolensis ruwenzorii) in Kasitu Subcounty, Bundibugyo District, Western Region, Uganda, which was the site of a 2007-2008 outbreak of ebolavirus disease [22] . We sampled bats from a culvert beneath a road spanning a mountain stream that serves as a water source for nearby human communities. We then used metagenomic methods to survey for viruses that may have been shed into the stream. In one rectal swab, we identified the nearly complete genome of the first African bat PRV, which we provisionally named Kasama virus (PRV16K) (named for Kasitu Subcounty, Uganda, and the Konjo word for water, amaghetse).

The genome sequence of PRV16K is highly similar to that of Asian PRV variants and PRV16K clusters within the known diversity of Asian PRVs, despite its geographic distance from Asia (see below). In Southeast Asia, PRV variants appear to transmit readily between species of pteropid bats [23] . We hypothesize that this promiscuity could result in a multispecies geographic continuum of PRV diversity in pteropid bats from Southeast Asia to Equatorial Africa. In support of this hypothesis, we present evidence of isolation-by-distance (IBD) for PRV. Given the apparent frequency of PRV transmission from bats to humans in Southeast Asia, PRV may represent an unappreciated cause of ILI across Africa and Asia.

The bats were collected by mist-net (Avinet, Portland, ME, USA) as they exited their roost at dusk. The bats were roosting as a multispecies assemblage of~200 Angolan soft-furred fruit bats (2 nights of double-observer roost exit counts) and an unknown number of Sundevall's leaf-nosed bats (Hipposideros caffer). The roost was a stone and cement culvert beneath the Itojo-Sempaya Old Road in Bundibugyo District, Uganda (0 • 50'18.5"N and 30 • 10'22.1"E). The bats were held in individual cloth drawstring bags from the time of collection until sampling, they were provisioned with a sugar/water solution (30% m/v) following sample collection, and they were released on site.

Rectal swabs were collected from 11 Angolan soft-furred fruit bats using sterile rayon-polyester-tipped swabs, preserved in TRI Reagent (Zymo Research, Irvine, CA, USA), and frozen at −20 • C within 3 h of collection. All the samples were exported with permission of the Uganda Wildlife Authority and the Uganda National Council for Science and Technology, shipped in accordance with International Air Transport Association (IATA) regulations, and imported under permit number 2017-07-103 of the Centers for Disease Control and Prevention (Atlanta, GA, USA).

Total RNA was extracted from the bat rectal swab samples using the Direct-zol RNA MicroPrep kit (Zymo Research, Irvine, CA, USA). Libraries were prepared for high-throughput sequencing (HTS) as previously described [24] . Briefly, sequencing libraries were prepared from 1 ng of cDNA per sample using the Nextera XT DNA (Illumina, San Diego, CA, USA) sample kit and sequenced using the Illumina MiSeq platform (Illumina, San Diego, CA, USA). HTS reads were quality-trimmed, and contiguous sequences (contigs) were assembled de novo using CLC Genomics Workbench version 11.1 (CLC Bio, Aarhus, Denmark) as previously described [25] . The assembled contigs were compared at the nucleotide and deduced protein sequence level to virus-derived and bat mitochondrial gene sequence in GenBank using BLASTn and BLASTx homology searching algorithms, respectively [26] .

PRV has 10 dsRNA genome segments (S1-3, M1-3, L1-3), nine of which contain a single open reading frame (ORF) (S2-3, M1-3, L1-3) and one (S1) with 3 overlapping ORFs [27] . The ORFs of the 10 segments of PRV16K were aligned (GenBank Accession numbers MT505315-MT505324) with homologous ORFs of the 15 previously identified strains of PRV [28] (Table 1 ; Genbank Accession numbers in Supplemental Table S1) using the Prank algorithm implemented in TranslatorX. The phylogenies were inferred using PhyML with "Bayesian Information Criterion" (BIC) smart model selection, "general-time reversible model with incorporation of rate of variation across site and proportion of invariable sites" (GTR+Γ+I) [29] and these are visualized in FigTree v.1.4.3 with midpoint rooting. The pairwise p-distances and standard errors (estimated from 1000 bootstraps) were calculated for all available PRV segments using MEGA7 [30] .

Sixteen sequences (PRV1NB-PRV16K; Table 1 ) of the S1 ORF encoding the cell-attachment protein (σC) were aligned as previously described [31] . The ratio of nonsynonymous (Ka) to synonymous (Ks) nucleotide substitution rates was estimated and a Ka/Ks annotated tree was generated using the Ka/Ks Calculation Tool of the University of Bergen Computational Biology Unit [32] .

To examine the patterns of the geographic spread of PRVs, we measured the pairwise geographic distance (minimizing distance traveled over water) between the site of sample collection/reported sites of origin for 15 PRVs (excluding PRV12I, which was isolated from a bat imported to Italy) (Supplemental Table S1 ) using ArcGIS Online (ESRI, Redlands, CA, USA). We computed the patristic distances between PRV variants from the phylogenetic trees of segments S1-S4 using PATRISTIC [33] . Mantel tests of the matrix correlation [34] were then performed to compare matrices of pairwise geographic distance to matrices of pairwise patristic distance for S1-S4 using XLSTAT [35] . The confidence intervals and the significance of Mantel's r (rM) were computed using 10,000 Monte Carlo simulations (alpha = 0.05) and Mantel correlograms were generated using the Spatial Data Calculator [36] (20 intervals of 1000 km).

We collected rectal swabs from 11 bats, representing an estimated 5.5% of the approximately 200 Angolan soft-furred fruit bats in the roost. Metagenomic sequencing revealed the presence of PRV in one of these bats (prevalence 9.1%). In the infected bat, the average depth of coverage across the 10 PRV segments was 82.15 ± 32.7, with 1.2% of quality-trimmed reads mapping to a PRV segment. Gaps were filled via rt-PCR and Sanger sequencing. The coding-complete sequence of 9 out of 10 PRV segments was recovered. Based on the sequence data, PRV16K has the typical PRV genome organization including a tricistronic S1 and single ORFs on the remaining segments. For L3, 706 nt of the ORF encoding the major inner capsid protein (λ1) (nt 2544-3249 of L1) was lost to low coverage and could not be recovered by rt-PCR. This gap in the sequence alignment of L3 was excluded for all subsequent analyses. The mitochondrial gene sequences of cytochrome b (cytb) and cytochrome c oxidase subunit 1 (cox1) from the infected bat were also assembled as described above. These sequences (GenBank accession numbers MT708578 and MT707261) confirmed the bat to be an Angolan soft-furred bat of the Ruwenzorii subspecies (Lissonycteris angolensis ruwenzorii) (percent identity: cytb 99.56; cox1 100%).

Nucleotide-level analyses revealed different patterns for the 10 PRV segments (Figure 1 ). PRV S1 demonstrates the greatest diversity globally with an average pairwise distance of 35% ± 12.7%. The tricistronic S1 segment encodes the cell-attachment protein (σC), as well as two nonstructural proteins (p10: fusion-associated small transmembrane (FAST) protein, and p17 (unknown function in PRV)) [37] . Much of the global PRV S1 nucleotide diversity is found in the σC ORF (overall mean p-distance 40.6% ± 0.7%) as compared to P10 (24.1.x% ± 1.4%) and P17 (30.4% ±1.5%). PRV16K form a clade with bat variants PRV1NB, PRV12I, and PRV15G, but clusters with different human and bat variants across the remaining segments ( Figure 2 ). PRV16K S1 is most closely related to S1 of PRV12I, from a bat imported to Italy from Indonesia [9] (p-distance = 25.3% ± 1.1%). For the remaining segments, p-distance is remarkably low between PRV16K and known variants from humans (PRV3M, L3 (6.6% ± 0.5%); PRV4k, S2 (6.4% ± 0.4%), M2 (5.3% ± 0.5%), M3 (7.6% ± 0.5%); PRV8B, S3 (9.0% ± 0.6%); PRV10M, L2 (13.2% ± 0.6%)) and from bats (PRV2P, L1 (8.7% ± 0.5%); PRV11C, M1 (4.8% ± 0.4%); PRV12I, S4 (4.8% ± 0.4%)) of Southeast Asia ( Figure 1 ). Supplemental Table S1 . The missing segments of PRV isolates are indicated by black and white hashed cells. The outlined cells indicate the lowest p-distance to PRV16K for each segment. The heatmap color key shows the range of the p-distance values estimated.

The location of origin of 14 PRV variants were mapped with the color of points indicating p-distance relative to PRV16K of S1-4 ( Figure 3 ). Although p-distances generally increase with geographic distance for S1-S4, the low S1 pairwise distance between PRV16K and PRV1NB is an outlier for the segment. The global Mantel test of the correlation between geographic and patristic distance of PRV S1-4 indicated significant correlation for S1, S2, and S4 but not for S3, and correlation was strongest for S1 and S4 ( Table 2 ). The mantel correlograms ( Figure 4) show some distance decay of Mantel's r (rM), but genetic distance significantly declines with geographic distance out to the greatest pairwise geographic distance (19,420 km) for S2 and S4. The correlogram for S1 shows a a steep decline in rM from 2000 to 5000 km, which suggests that forces other than geographic distance, such as diversifying selection, may be driving geographic clustering of S1 genetic diversity. The global Mantel test for S3 was not significant, but the mantel correlogram for S3 shows significant correlation at short and intermediate distance intervals (0-4000 km, 5000-9000 km; Figure 4 ). Table 1 . The sequences for S2 and S4 were not available for (6) PRV6XR (indicated by a black hexagon). Supplemental Table S1 . The sequences for S2 and S4 were not available for (6) PRV6XR (indicated by a black hexagon). 

We detected strong positive selection (Ka/Ks = 1.231) at node 15, where the two major clades of PRV S1 diverged. The purifying selection predominates (Ka/Ks << 1) on all other branches aside from the more moderate positive selection (Ka/Ks = 1.007) on branch 2 of Node 7 (Supplementary Figure  1 ).

PRV was first discovered in 1968, with the isolation of PRV1NB from a grey-headed flying fox (Pteropus poliocephalus) [6] . PRV1NB was a novelty among mammalian orthoreoviruses, as it was capable of forming multinucleated syncytia via a FAST protein akin to the avian reoviruses (ARV) [5] . PRV2P, isolated from a small flying fox (Pteropus hypomelanus) in Malaysia in 1999, expanded the range of PRV in bats [10] , but the recognition of PRV as a cause of respiratory illness in humans did not come until the discovery of PRV3M and PRV4K associated with URTI in Malaysia [18, 19] . This realization led to further investigations showing that PRV may be transmitted regularly from bats to humans [1, 20] , placing it in a rare class of zoonotic bat viruses that transmit frequently from bats to humans [38] . The discovery of a closely related PRV strain in a Ugandan bat expands the geographic distribution of the virus to a new continent. The high similarity of PRV16K to Southeast Asian variants, and the strong and significant correlation between the geographic and genetic distances of 

We detected strong positive selection (Ka/Ks = 1.231) at node 15, where the two major clades of PRV S1 diverged. The purifying selection predominates (Ka/Ks << 1) on all other branches aside from the more moderate positive selection (Ka/Ks = 1.007) on branch 2 of Node 7 (Supplementary Figure S1 ).

PRV was first discovered in 1968, with the isolation of PRV1NB from a grey-headed flying fox (Pteropus poliocephalus) [6] . PRV1NB was a novelty among mammalian orthoreoviruses, as it was capable of forming multinucleated syncytia via a FAST protein akin to the avian reoviruses (ARV) [5] . PRV2P, isolated from a small flying fox (Pteropus hypomelanus) in Malaysia in 1999, expanded the range of PRV in bats [10] , but the recognition of PRV as a cause of respiratory illness in humans did not come until the discovery of PRV3M and PRV4K associated with URTI in Malaysia [18, 19] . This realization led to further investigations showing that PRV may be transmitted regularly from bats to humans [1, 20] , placing it in a rare class of zoonotic bat viruses that transmit frequently from bats to humans [38] . The discovery of a closely related PRV strain in a Ugandan bat expands the geographic distribution of the virus to a new continent. The high similarity of PRV16K to Southeast Asian variants, and the strong and significant correlation between the geographic and genetic distances of PRV segments, evince that PRV16K is not the extreme outlier its far-flung origin would suggest, but perhaps a distant point on a geographic continuum of PRV diversity. PRV16K, discovered in a rectal swab from an apparently healthy Angolan soft-furred fruit bat in Bundibugyo District, Uganda, is closely related to PRVs isolated from humans and bats in Asia Pacific (Figure 1 ). The ORFs of 8 out of 10 segments (S2-4, M1-3, L1, L2) all share ≥91% nucleotide identity with a known PRV variant. S1 identity is lower, but globally S1 appears to be less constrained than the remaining segments across PRVs. That no segments of PRV16K appear as an outlier to an Asian clade of PRVs ( Figure 2 ) suggests that the virus did not co-diverge with African fruit bats from their ancestral Asian lineages in the Miocene [39] , but rather spread between African and Asian bats more recently. The Angolan soft-furred bats are split into five subspecies. PRV16K was identified in the Ruwenzorii subspecies, which has been identified across sub-Saharan Africa [40] .

The genetic and geographic distances between PRVs correlated significantly for segments S1, S2, and S4 by the global Mantel test ( Table 2 ). Significant Mantel correlation supports an isolation-by-distance (IBD) hypothesis [41] , wherein genetic distance between PRV variants increases with geographic distance across a distribution of host species [42] . For S2 and S4, Mantel correlograms ( Figure 4) showed a relatively smooth decrease in rM, with significant correlation out to the greatest distance interval. This provides evidence that S2 and S4 diversity, across the geographic distribution of PRV, is not significantly structured by factors such as host species. In other words, PRV appears to have a broad host tropism [3] . We note that our analyses assume that human PRV infections were acquired locally, or, for imported cases, at the reported site of infection. Furthermore, PRV in bats may also be oversampled at some sites, which could cause other geographic biases. The strong association between PRV genetic and geographic distance that we documented may therefore change in magnitude as new viral variants are described from other locations.

Globally, PRV segments show different evolutionary histories (Figure 3 ), probably due to frequent inter-species transmission and genome reassortment during PRV co-infection [23] . A study of bat roosts in the Philippines found evidence of cross-species transmission of PRV isolates, neutralizing antibodies in the sera of infected and uninfected bats, and one bat apparently co-infected with two genetically divergent PRV isolates [7] . PRV has thus far been isolated from cave-and canopy-roosting frugivorous Pteropodid bats [43] . Cave-roosting bats show a higher propensity for interspecies virus transmission than do canopy roosting bats, due to the co-roosting of multiple species [44] . Although the mechanism by which PRV transmits between bat species across the range of the virus remains incompletely known, these studies suggested that co-roosting behavior is a major driver of such transmission. The transmission of PRV between related bat species should occur with relative ease under permissive ecological conditions, given the aforementioned zoonotic potential of PRV and further evidence of infection of wild crab-eating macaques (Macaca fascicularis) in Singapore [3] and Thailand [45] . We anticipate that future surveys of pteropids bats of south Asia, the Middle East, and the Horn of Africa may further expand the range of PRV and fill the gaps in the geographic continuum of genetic distance that we have documented.

The S1 Mantel correlogram ( Figure 4) shows some evidence of genetic clustering, due to the steep decline in rM in the intervals from 2000 to 5000 km. We hypothesized that diversifying selection may have significantly impacted the evolutionary history of S1. We found evidence of strong positive selection on σC, and in particular the ratio of nonsynonymous to synonymous mutations showed strong positive selection acting on the node separating the two S1 clades (Supplemental Figure S1 ). σC is necessary for cell attachment and may determine cell/host tropism, although the host ligand(s) for PRV σC is not certain [37] . The structural and functional homolog in the more thoroughly studied ARV (σ1) induces high levels of strain specific anti-ARV neutralizing antibodies in birds [46] , and so the strong positive selection on σC which orders the evolutionary history of S1 in PRV may be due to immune selection.

Although we do not yet have evidence that PRV16K is zoonotic, the high similarity of many PRV16K segments to PRVs isolated from humans with ILI suggests that further investigation is warranted. PRV infection dynamics are poorly understood, and the modes of transmission between bats, and to spillover hosts, are largely unknown, but the detection of virus RNA in bat excreta indicates that fecal-oral transmission could play a role. That the bat population in which PRV16K was discovered was associated with a human water source in Bundibugyo District, Uganda, is of concern. Reoviruses are notably stable in the environment due to their non-enveloped, "double-shelled" capsid [47] . Although waterborne transmission of PRV has not been reported, infectious mammalian reoviruses are frequently isolated from surface water [48, 49] . Additionally, the original zoonotic variants of PRV (PRV3M, PRV4K) were acquired via indirect contact with bats [18, 19] , implicating environmental transmission of the viruses. We also note that Bundibugyo District was the site of an ebolavirus disease outbreak in 2007-2008 [22] , suggesting a high risk for bat-associated zoonoses in this area.

Surveillance for influenza and ILI in many sub-Saharan African countries is poor, although the disease burden of ILI is significant [50] [51] [52] [53] . Even when active influenza surveillance is undertaken, differential diagnosis requires multiplex diagnostics to accurately identify the diverse etiologies of ILI [54] . It is therefore entirely conceivable that PRV is a cause of URTI in Africa that has gone undetected. Furthermore, given the pan-equatorial African distribution of the Angolan soft-furred fruit bat and the apparent ease of transmission of PRV between bat species reported in Southeast Asia, PRV may be broadly distributed across the African continent. The discovery of PRV in Africa dramatically expands the range of this bat-borne zoonosis and warrants further investigation to assess whether PRV is actively transmitted in sub-Saharan Africa, as it is in Southeast Asia.

The following are available online at http://www.mdpi.com/1999-4915/12/7/740/s1, Figure S1 : Ka/Ks annotated tree of cell-attachment protein ORF of Pteropine orthoreovirus Segment 1, Table S1 : Accession numbers and locations of PRVs used in this study. 

Pteropine orthoreovirus infection among out-patients with acute upper respiratory tract infection in Malaysia

Serological evidence of human infection by bat orthoreovirus in Singapore

Serological evidence and experimental infection of cynomolgus macaques with pteropine orthoreovirus reveal monkeys as potential hosts for transmission to humans

Pteropine orthoreovirus: An important emerging virus causing infectious disease in the tropics?

The diversity of the orthoreoviruses: Molecular taxonomy and phylogentic divides

Nelson Bay virus

First isolation and characterization of pteropine orthoreoviruses in fruit bats in the Philippines

Characterization of a novel orthoreovirus isolated from fruit bat, China

A new member of the Pteropine Orthoreovirus species isolated from fruit bats imported to Italy

Pulau virus; A new member of the Nelson Bay orthoreovirus species isolated from fruit bats in Malaysia

A novel reovirus isolated from a patient with acute respiratory disease

Rapid whole genome sequencing of Miyazaki-Bali/2007 Pteropine orthoreovirus by modified rolling circular amplification with adaptor ligation-Next generation sequencing

Viruses-DsRNA Viruses

A new class of fusion-associated small transmembrane (FAST) proteins encoded by the non-enveloped fusogenic reoviruses

Characterization of a Novel Syncytium-Inducing Baboon Reovirus

Xi River virus, a new bat reovirus isolated in southern China

A previously unknown reovirus of bat origin is associated with an acute respiratory disease in humans

Identification and Characterization of a New Orthoreovirus from Patients with Acute Respiratory Infections

Investigation of a Potential Zoonotic Transmission of Orthoreovirus Associated with Acute Influenza-Like Illness in an Adult Patient

Serological evidence of human infection with Pteropine orthoreovirus in Central Vietnam

Evolutionary relationship of the Land M-class genome segments of bat-borne fusogenic orthoreoviruses in Malaysia and Australia

Naturally Circulating Hepatitis A Virus in Olive Baboons

Metagenomic assessment of adventitious viruses in commercial bovine sera

Basic local alignment search tool

Orthoreoviruses and their replication

Viruses in bats and potential spillover to animals and humans

Smart Model Selection in PhyML

MEGA7: Molecular Evolutionary Genetics Analysis version 7.0 for bigger datasets

Phylogeny-aware alignment with PRANK

Computational Biology Unit-Ka/Ks Calculation Tool

A program for calculating patristic distances and graphically comparing the components of genetic change

Biometry: The Principles and Practice of Statistics in Biological Research

An online calculator for spatial data and its applications

Reverse Genetics for Fusogenic Bat-Borne Orthoreovirus Associated with Acute Respiratory Tract Infections in Humans: Role of Outer Capsid Protein σC in Viral Replication and Pathogenesis

Ecological dynamics of emerging bat virus spillover

The evolutionary history of the African fruit bats (Chiroptera: Pteropodidae)

Isolation by Distance

Surveillance of Bat Coronaviruses in Kenya Identifies Relatives of Human Coronaviruses NL63 and 229E and Their Recombination History

PREDICT Consortium A Comparative Analysis of Viral Richness and Viral Sharing in Cave-Roosting Bats

Mass spectrometry-based identification and whole-genome characterisation of the first pteropine orthoreovirus isolated from monkey faeces in Thailand

Avian Reovirus Proteins Associated with Neutralization of Virus Infectivity

In Viruses

Detection and Identification of Mammalian Reoviruses in Surface Water by Combined Cell Culture and Reverse Transcription-PCR

High Incidence of Mammalian Orthoreovirus Identified by Environmental Surveillance in Taiwan

The Unrecognized Burden of Influenza in Young Kenyan Children

Global burden of respiratory infections due to seasonal influenza in young children: A systematic review and meta-analysis

Influenza in Africa: Uncovering the Epidemiology of a Long-Overlooked Disease

Surveillance and management of influenza on the African continent

Population-based active surveillance cohort studies for influenza: Lessons from Peru

This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license

We would like to thank David Hyeroba (Makerere University) for his invaluable intellectual and logistic contributions to this project (in memoriam). We also thank Jessica Carag (University of Wisconsin-Madison) for assistance with sample collection and Erica Moulton (University of Wisconsin-Madison) for assistance with the preparation of figures for this manuscript.

The authors declare no conflict of interest.