A genome‐wide linkage map for the house sparrow (Passer domesticus) provides insights into the evolutionary history of the avian genome

The house sparrow is an important model species for studying physiological, ecological and evolutionary processes in wild populations. Here, we present a medium density, genome wide linkage map for house sparrow (Passer domesticus) that has aided the assembly of the house sparrow reference genome, and that will provide an important resource for ongoing mapping of genes controlling important traits in the ecology and evolution of this species. Using a custom house sparrow 10 K iSelect Illumina SNP chip we have assigned 6,498 SNPs to 29 autosomal linkage groups, based on a mean of 430 informative meioses per SNP. The map was constructed by combining the information from linkage with that of the physical position of SNPs within scaffold sequences in an iterative process. Averaged between the sexes; the linkage map had a total length of 2,004 cM, with a longer map for females (2,240 cM) than males (1,801 cM). Additionally, recombination rates also varied along the chromosomes. Comparison of the linkage map to the reference genomes of zebra finch, collared flycatcher and chicken, showed a chromosome fusion of the two avian chromosomes 8 and 4A in house sparrow. Lastly, information from the linkage map was utilized to conduct analysis of linkage disequilibrium (LD) in eight populations with different effective population sizes (Ne) in order to quantify the background level LD. Together, these results aid the design of future association studies, facilitate the development of new genomic tools and support the body of research that describes the evolution of the avian genome.


| INTRODUC TI ON
Recent technological advances have decreased the cost of next generation sequencing and made the technology broadly available for non-model species. Whole genome sequencing and the construction of draft genome assemblies have become routine and are currently applied to non-commercial species for the purpose of studies in ecology and evolution (Ellegren, 2014). While a draft genome assembly accurately describes the physical position of genes along scaffolds, it does not provide information about the order and orientation of scaffolds and high-level chromosome organization.
As such, evolutionary information cannot be found in highly fragmented assemblies. In contrast, linkage maps are inaccurate at a fine scale due to difficulties of resolving recombination between adjacent markers, but highly valuable to describe genome structure at the chromosome level. Combining the two resources can be used to order and orientate scaffolds along linkage groups to generate chromosome-scale assemblies (Fierst, 2015).
The comprehensive information in a reference genome cannot be fully utilized without further knowledge about the genetic distance between loci and the pattern of inheritance. Linkage maps represent an invaluable tool in this respect, as they describe the genetic distance along the chromosomes; that is the likelihood that two loci are inherited together and the frequency at which their association is broken by crossing over during meiosis. Recombination is a fundamentally important force in evolution, as it creates new haplotypes by breaking up associations between linked loci, thus increasing haplotype variation and improving the efficiency of natural selection.
Whilst advantageous gene combinations may be disrupted, recombination may also break associations between disadvantageous alleles and favourable mutations at linked loci (Kawakami et al., 2017). The rate of recombination is therefore expected to correlate with rate of adaptive evolution.
The rate of recombination is not uniform, but elevated in genomic hotspots, also called 'fragile sites' characterized by specific repeats (see Aguilera & Gomez-Gonzalez, 2008 for review). Genome-wide and site-specific recombination rates are species specific and information in linkage maps cannot be extrapolated between taxa, although some general trends exist. Due to at least one obligate crossover for each chromosome pair prior to segregation, the rate of recombination per physical unit of distance generally increases with decreasing chromosome size (Ellegren, 2010;Kaback, Guacci, Barber, & Mahon, 1992). In the zebra finch (Taeniopygia guttata), hotspots were detected on average every 215 kb, with a total of nearly 4,000 hotspots (Singhal et al., 2015). These were in the proximity of GC rich regions, gene promotor regions and transcription start/stop sites. In zebra finch and collared flycatcher (Ficedula albicollis), these hotspots are often associated with chromosome ends, whilst areas near centromeric regions are characterized as recombination deserts (Backström et al., 2010;Bolívar, Mugal, Nater, & Ellegren, 2016;Kawakami et al., 2017). Different species of finches show greater conservation in the location of recombination hotspots than do mammals (Singhal et al., 2015), perhaps in concordance with the higher degree of synteny in bird genomes compared to mammals (Backström et al., 2010;Ellegren et al., 2012). Moreover, several studies spanning a wide variety of taxa have reported sex specific recombination rates, also termed heterochiasmy (Hansson, Åkesson, Slate, & Pemberton, 2005;Johnston, Huisman, Ellis, & Pemberton, 2017;Lien et al., 2011), although no clear pattern exists with respect to the sex that displays the elevated recombination rate.
Linkage maps were previously limited to humans (Homo sapiens) (Weissenbach et al., 1992) and model or life stock species such as mice (Mus musculus) (Copeland et al., 1993), sheep (Ovis aries) (Crawford et al., 1995), zebra finch (Backström et al., 2010) or Atlantic salmon (Salmo salar) (Lien et al., 2011). With the advancement of genotyping technology, linkage maps from many wild populations have become available, and include great reed warbler (Acrocephalus arundinaceus) (Åkesson, Hansson, Hasselquist, & Bensch, 2007), red deer (Cervus elaphus) (Johnston et al., 2017), collard flycatcher (Ficedula albicollis) (Kawakami et al., 2014), great tit (Parus major) (van Oers et al., 2014) and Soay sheep (Slate et al., 2010). These linkage maps greatly facilitate association studies in wild populations, often with the aim of mapping genes for ecologically important phenotypes. In this context, the extent of linkage disequilibrium (LD) is important: LD describes the nonrandom association between alleles at different loci. A long LD range suggests that fewer markers are needed to detect an association, but it also implies that the underlying causative variants will be more difficult to identify (Lindblad-Toh et al., 2005).
Higher LD is the result of genetic drift, natural selection, and population structure/admixture, whilst the breakdown of LD is mainly caused by recombination (De La Vega et al., 2005). Because the effective size of a population reflects the rate of genetic drift, LD is expected to increase with decreasing population size. Information about LD in different focal populations is therefore important when carrying out association studies. Importantly, linkage maps provide insight into the evolution of the avian genome, which is characterized by a high chromosome number and a karyotype consisting of a mixture of microchromosomes and a few larger macrochromosomes that may differ by as much as an order of magnitude in size (Rodionov, 1996). Despite long evolutionary divergence times, the avian karyotype has proved to be remarkably well conserved compared to other amniotic taxa such as mammals (Backström et al., 2010(Backström et al., , 2008Ellegren et al., 2012;Kawakami et al., 2014). Passerine birds diverged from Galliformes approximately 100 million years ago (MYA) (Tuinen, Sibley, & Hedges, 2000), and while Passeriformes is the largest and most diverse bird clade, the synteny is similar in all species investigated (Åkesson et al., 2007;Backström et al., 2010;Kawakami et al., 2014;van Oers et al., 2014).
The house sparrow (Passer domesticus) is a small passerine bird that has served as a model species of a number of evolutionary, physiological and ecological studies (Anderson, 2006). The house sparrow is native to Europe and Asia, and due to introductions aided by humans, the species is now established on all major habitable land masses across the globe (Anderson, 2006). House sparrows are nonmigratory with a low dispersal rate from their natal area (Pärn, Ringsby, Jensen, & Saether, 2012) and populations and individuals can be monitored over multiple years. The house sparrow is commensal with humans, individuals are easily captured using mist nets, and the species will readily make use of nest boxes, which again facilitates the access to eggs and chicks.
Together, these characteristics make populations of house sparrows suitable for pedigree studies. Until recently, genomic resources for the house sparrow were lacking and the genetic tools were limited to microsatellites (Dawson et al., 2012). A custom made 10 K Illumina SNP array represented the first genomic resource (Hagen et al., 2013) and has greatly facilitated further development of genetic tools. A major recent achievement has been the house sparrow reference genome (Elgvin et al., 2017), which was developed in concert with the house sparrow linkage map

| Blood sampling and pedigree
The samples used in the linkage analysis were collected from 1910 individuals from four different island populations: Aldra (N = 348), Hestmannøy (N = 448), Leka (N = 511) and Vega (N = 603) (see Table 1 Table 1. Extraction of the blood samples and subsequent genotyping on the 10 K Illumina iSelect HD BeadChip is described in detail in Hagen et al. (2013). The results were scored using GenomeStudio (Illumina) following the manufacturer's guidelines (Illumina, Inc. 2010). We used plink version 1.07 (Purcell et al., 2007) for further quality control. Markers with a minor allele frequency (MAF) < 0.01 were removed. First, we removed markers that deviated significantly from Hardy-Weinberg equilibrium frequencies within the four populations used in linkage mapping (n = 163). Next, markers with a minor allele frequency (MAF) < 0.005 (n = 1,462), and individuals with >10% missing genotypes were removed. Finally, we removed SNPs with >20 Mendelian errors based on parent-offspring links in the microsatellite pedigrees (n = 73). After quality control our data set consisted of 6,764 (autosomal) SNPs typed in 1,898 individuals.

| Construction of the linkage map
The linkage map was constructed in concert with the house sparrow reference genome (Elgvin et al., 2017). A modified version of the crimap 2.4 software (Green, Falls, & Crooks, 1990), including added utilities that prepare text files to facilitate downstream analysis of large marker sets provided by Xuelu Liu and Michael Grosz Additionally, 456 SNPs were assigned to an unmapped linkage group corresponding to the Z chromosome. After the initial grouping of SNPs, markers on the 29 autosomal LGs were ordered using the BUILD and FLIPSN options in CRIMAP. Following this, 120 bp flanking each SNP was positioned in the house sparrow genome assembly using BLAST and used to assign scaffolds to LGs, order and orientate scaffolds within LGs, and build sequences for 29 autosomal chromosomes in P. domesticus (Elgvin et al., 2017). If a scaffold contained only one marker in the linkage map, the scaffold was assigned the same orientation as in the zebra finch genome. Following the construction of chromosome sequences, the SNP-order within each LG was fine-tuned using physical positions of the SNPs in the chromosome. The CHROMPIC option in CRIMAP was then used to phase genotypes within linkage groups. CHROMPIC output were utilized to identify double recombination events caused by single SNPs. Such events are considered highly dubious, most probably caused by genotyping errors, and the genotypes were zeroed out before constructing the final version of linkage maps. Finally, multipoint linkage maps for the 29 autosomal linkage groups in the linkage map were constructed using the FIXED option of CRIMAP and the Kosambi map function (Kosambi, 1943). Chromosome nomenclature in house sparrow was determined by alignments against the zebra finch, flycatcher and chicken genomes and named Pdom 1-15, Pdom 17-28, Pdom 1A and LGE22, where Pdom is derived from the scientific name.

| Analysis of recombination rate and genome structure
We investigated the relationship between physical length and genetic length of all mapped chromosomes using linear regression in R version 3.5.3 (R Core Team, 2018). The mean recombination rate of macro-and microchromosomes (larger and smaller than 35 Mbp), according to the karyotype described in Bulatova, Radjabli, and Panov (1972) was compared using t-statistics in R. To characterize how recombination rate varies with distance from chromosome ends, recombination rates were characterized in 1 Mbp bins along the chromosomes and the physical length of each chromosome adjusted to a relative size of 0 (start) to 1 (end). Due to size dependent recombination patterns, we visualised the recombination landscape of macrochromosomes (Pdom 1A and 1-8) and microchromosomes < (Pdom 9-28) separately.

| Chromosomal rearrangements
For visualization of the alignments of the house sparrow genome with the genomes of zebra finch, chicken and collared flycatcher, we used circos version 0.67-7 (Krzywinski et al., 2009). To obtain input files for circos, we aligned the house sparrow genome (GenBank accession MBAE00000000) with each of the three target genomes (GenBank accession numbers AGTO02000000, ABQF00000000 and AADN00000000) using Satsuma synteny version 2.0 (Grabherr et al., 2010). To get a measure of the percentage of interchromosomal rearrangements on the 29 mapped house sparrow chromosomes compared to zebra finch, chicken and collared flycatcher, we estimated the pairwise length of rearranged sequence for each chromosome. Chicken, flycatcher or zebra finch sequences that were in opposite direction than in the house sparrow genome were defined as "negative length", and the total length of "negative sequence" was summarised for each species.
Due to chromosomal rearrangements between species, house sparrow chromosomes 1A and 1 were compared with chicken chromosome 1, and house sparrow chromosome 8 was compared with chicken chromosomes 4 and 8. House sparrow chromosome 8 was compared with zebra finch and collared flycatcher chromosomes 4A and 8.

| Validation of interchromosomal fusion
We utilized re apr (v 0.1.9) (Hunt et al., 2013) and the 8 and 10 kb mate-pair sequencing libraries from the house sparrow genome assembly to further validate the fusion of chromosomes 4A and 8 identified in the linkage map. re apr uses short reads to score bases and long insert libraries to identify assembly errors. To identify potential assembly errors, both mate-pair libraries were separately mapped to the house sparrow assembly using the default settings of bwa-mem (v 0.7.17) (Li, 2013) with the -M parameter added for Picard compatibility. The resulting bam files were sorted with Picard (v 1.9.6) SortSam and duplicates were marked with Picard MarkDuplicates with validation stringency set to "lenient". The full re apr pipeline of assembly evaluation was then run using the re apr pipeline

| Analysis of linkage disequilibrium
For the analysis of LD, we used 6,490 SNPs for 28 autosomal LGs (excluding the eight SNPs on LGE22), and genotypes from individuals that originate from the following eight populations: Aldra, Gjaerøy, Hestmannøy, Indre Kvarøy, Leka, Myken, Nesøy and Vega (see Table 1).
The genotypes were based on the data set described in Hagen et al. (2013) with additional 271 samples genotyped on the house sparrow 10K Illumina SNP array. The additional samples were stored, extracted and genotyped as previously described in Hagen et al. (2013). Analysis of LD were carried out in plink version 1.07 (Purcell et al., 2007) and R ver-

| Linkage analysis
The linkage analysis mapped 6,498 SNPs to 29 autosomal chromosomes (1-15, 17-28, 1A and LGE22) (Figure 1, Table 2). For accurate physical and genetic marker positions and information on number of informative meioses for each marker, see Table S1. The sex averaged linkage map had a length of 2,004.5 cM, with a longer map for females (2,239.7 cM) than for males (1,801.3 cM), a difference corresponding to a 1.24 times higher recombination rate in females than males (Table 2).

| Recombination rate and genome structure
It is apparent that house sparrow chromosomes lack the strong sshaped or sigmoid recombination rate pattern that is seen in zebra finch ( Figure 2). Instead, recombination rates along the chromosomes more closely resembles the linear pattern that is seen in chicken.
However, the recombination rate varied both within and between chromosomes in the house sparrow genome: recombination rates were greater for microchromosomes (6.41 cM/Mbp) than for macrochromosomes (1.78 cM/Mbp) (one-sided t test, p = .001) (Figure 3).
Additionally, the recombination pattern along the chromosomes differs between micro-and macrochromosomes. Macrochromosomes TA B L E 2 Number of SNP markers, physical size (bp), average distance between markers (bp), and male, female and average recombination rate (cM) for each autosomal chromosome in the house sparrow linkage map

| Chromosomal rearrangements
In construction of the linkage map, we detected a novel interchromosomal rearrangement in the house sparrow genome: the fusion of the Passerine chromosome 8 and the Passerine chromosome

F I G U R E 2
The relationship between the physical and genetic position for all markers on each mapped autosomal chromosome for three avian species. Black dots are house sparrow (this study), red dots are chicken (Groenen et al., 2009) and blue dots are zebra finch (Backström et al., 2010). The x-axis represents the physical size of the chromosomes, based on the house sparrow genome assembly (Elgvin et al., 2017). Due to the chicken linkage map being significantly longer than the two passerine species, the chicken genome is always presented on a secondary y-axis (right), and for chromosome 1 and 4, a secondary x-axis (top). House sparrow lacks chromosome 4A, which for house sparrow is fused with chromosome 8. Only chromosomes with 15 or more markers are included. For zebra finch, data was not available for chromosomes 21,22,24,26,27  and collared flycatcher ( Figure 6). As in zebra finch and collared flycatcher, chromosome 1 in house sparrow is split into 1 and 1A, whilst chicken has retained the ancestral chromosome 1.
Beyond this, the house sparrow chromosome karyotype largely resembles that of zebra finch and collared flycatcher ( Figure 6).
The separate linkage group LGE22, which is found in zebra finch, great tit and flycatcher was also detected in house sparrow and is homologous to regions in the chicken linkage groups LG2 and LGE22C19W28_E50C23.
Despite the high degree of conserved synteny seen between house sparrow, zebra finch, collared flycatcher and chicken, several inversions were detected when intrachromosomal marker orders were compared (Table S3). When comparing the percentage of inverted sequence between house sparrow and chicken, collared flycatcher and zebra finch respectively, we found that only a few chromosomes had near complete conserved synteny of marker order between the different species. Chromosomes 6, 9 and 17 showed a high degree of conserved synteny between all three species. The total percentage of "negative sequence" between house sparrow and chicken was 47%, for collared flycatcher 51% and for zebra finch 50% (see Table S3).

| Linkage disequilibrium
Our results show that LD (r 2 ) decreased from values of 0.15-0.20 for markers up to 5 Kbp apart to 0.11-0.14 in the 10-15 kbp bin

| D ISCUSS I ON
We have mapped 29 autosomal chromosomes for the house sparrow ( Figure 1, Table 2). Additionally, 456 markers have been assigned to an unmapped reference sequence that corresponds to the zebra finch Z chromosome (Elgvin et al., 2017). As the house sparrow has an expected haploid chromosome number of 38 (Bulatova et al., 1972), it is apparent that eight chromosomes (linkage groups) have not been mapped. It is likely that these represent a corresponding number of microchromosomes, which by nature are particularly challenging to map and assemble: they are often rich in tandem repeats, have a high GC content and often contain gene families that are difficult to sequence such as the MHC gene complex on Gga16, which has proved difficult to assemble (Warren et al., 2017). Despite a high-density linkage map and excellent reference genome for collared flycatcher, for which the karyotype is unknown, Kawakami et al. (2014) estimated that approximately five or six microchromosomes remain unmapped.
The mean recombination rate reported in this study is comparable to that found in chicken (6.02 cM/Mbp) (Groenen et al., 2000), zebra finch (3.18 cM/Mbp) (Stapley, Birkhead, Burke, & Slate, 2008) and collared flycatcher (3.1 cM/Mbp) (Kawakami et al., 2014). Recombination rates in the zebra finch have been shown to be strikingly heterogeneous along chromosomes, with most recombination events being concentrated towards the chromosome ends (Backström et al., 2010), i.e., plots of recombination rate against physical distance have a sigmoidal shape. Even though there is a tendency for increased recombination rate towards telomeric ends on the large chromosomes (Figure 4a), the house sparrow does not seem to have the pattern of very low recombination rate in middle parts of the chromosomes that is seen in zebra finch ( Figure 2).
Instead, the house sparrow displays a recombination pattern that resembles the uniform distribution seen in collared flycatcher and chicken (Kawakami et al., 2014) (Figure 2). As expected (Groenen et al., 2000;Stapley et al., 2008), there was an inverse relationship between chromosome size and recombination rate as well as a near linear relationship between genetic distance and physical size of chromosomes ( Figure 3) similar to what has been reported for other Passerine species (Backström et al., 2010;Kawakami et al., 2014).
Of the 10 largest house sparrow chromosomes, eight are either submetacentric or subtelocentric, whilst two are telocentric (Bulatova et al., 1972). Apart from chromosome 2, which is the largest, the identity of the different chromosomes cannot be discerned based on karyotype. Also, the position of the centromere for the 28 smaller microchromosomes is unknown (Bulatova et al., 1972).
As recombination rate increases with distance from centromere (Akhunov et al., 2003), we expect the different arms on acrocentric chromosomes to have different recombination rates. It is therefore possible that recombination rate as a function of distance from both the centromere and chromosome ends would provide the F I G U R E 6 Comparative circular visualization of the house sparrow genome with the genomes of (a) chicken, (b) collared flycatcher and (c) and zebra finch. The plots show the full house sparrow reference genome with the 29 mapped autosomal chromosomes as well as the unmapped Z chromosome, for which scaffold assembly was based on synteny with the zebra finch genome [Colour figure can be viewed at wileyonlinelibrary.com] most accurate model for describing the heterogeneous recombination rates along house sparrow macrochromosomes. Therein lies the challenge of achieving an accurate estimate for the centromere position for the different chromosomes; data which cannot be discerned from the karyotype described in Bulatova et al. (1972).
Although the avian karyotype is highly conserved across species (Derjusheva, Kurganova, Habermann, & Gaginskaya, 2004), several rearrangements have been reported, especially involving chromosome 4. Chicken chromosome 4 is likely to be a result of a fusion between the ancestral chromosome 4 and another chromosome (Griffin, Robertson, Tempest, & Skinner, 2007). In zebra finch, great tit and collared flycatcher, chromosome 4 is split into 4 and 4A, respectively (Kawakami et al., 2014;Stapley et al., 2008;van Oers et al., 2014). Microsatellite-based linkage maps suggest that chromosome 4 is split into two or more linkage groups in blue tit as well (Hansson et al., 2009;Nietlisbach et al., 2015). in these species (Pala et al., 2012). In the current study, we have shown that house sparrow lacks the corresponding chromosome 4A, due to a fusion of 4A with chromosome 8 (Figure 6). These results enlighten the interesting evolutionary history of the avian chromosome 4.
As expected, when the house sparrow linkage map and reference genome are compared to that of other species, several intrachromosomal inversions are found (see Table S3). Only three chromosomes (6, 9 and 17) have a similar and conserved marker order along the chromosomes across species. This lends support to the findings in Kawakami et al. (2014) where the same chromosomes in collared flycatcher were involved with few inversions relative to the rest of the genome when compared to chicken and zebra finch. Divergence times between house sparrow and chicken, collared flycatcher and zebra finch are approximately 100, 40 and 35 MYA, respectively (Griffin et al., 2007;Tuinen et al., 2000). When the house sparrow genome is compared to the three species, we find that approximately 50% of the genome is involved with intrachromosomal inversions in all three cases, despite differing divergence times.

F I G U R E 6 (Continued)
Within Passeriformes, most research suggests higher recombination rates in females. The most extreme heterochiasmy has been detected in the great reed warbler, with female recombination rate being 2.15 times higher than in males (Hansson et al., 2005).
Female recombination rate also exceeds that of males in blue tits with a ratio of 1.18 (Hansson et al., 2009), and in zebra finches with a ratio of 1.02 (Backström et al., 2010). In the current study, recombination rate was found to be 1.24 times higher in female house sparrows compared to males. Interestingly, Kawakami et al. (2014) found the opposite in collared flycatchers, where males recombine more frequently than females, with a ratio of 1.  (Haldane, 1922) and haploid selection (Lenormand, 2003) should lead to decreased recombination rates in the heterogameic sex. Our results, where female are the heterogameic sex, do not support these hypotheses. Moreover, Trivers (1988) suggested that in species where one sex experiences more variation in reproductive success than the other, the sex with higher variation should recombine less to maintain successful haplotypes. In house sparrows, the difference in sex specific variance in reproductive success is small, but slightly male biased (Jensen et al., 2004). Thus, based on this scenario, males are expected to experience a relatively small evolutionary advantage by recombining less. An alternative hypothesis predicts that there should be higher recombination rate in the sex with the most variation in reproductive success, because an elevated recombination rate tends to coevolve with any trait under selection (Korol & Iliadi, 1994). The overall lower recombination rate seen in house sparrow males lends little support to this hypothesis, unless specific chromosomes are considered. The most pronounced heterochiasmy was seen on the macrochromosomes, and on chromosomes Pdom 2-8 in particular, where females had a recombination rate 1.4 times higher than males. Interestingly, the reverse pattern was seen for the smallest chromosomes Pdom 25 and Pdom 28, respectively (Table 2). Directional selection on a trait can increase recombination (Otto & Lenormand, 2002)  More significant hits were found when the denser panel of SNPs and an average intermarker distance ca. 6 Kbp was used (Lundregan et al., 2018), thus illustrating our conclusion that an average intermarker distance of 18 Kbp is not dense enough to detect associations in GWA studies given the level of LD in the house sparrow populations used in these studies. We have presented a genome wide, medium-density linkage map for an ecological and genetic model species, the house sparrow. These data will greatly facilitate ongoing association studies in wild populations and aid our understanding of evolutionary adaptations in a Passerine species. Moreover, we have characterized a novel interchromosomal translocation, and thus added to the body of research that elucidates the evolution of the avian genome.

ACK N OWLED G EM ENTS
We thank fieldworkers and students for their contributions to fieldwork, and Randi Røsbak for her assistance with laboratory analyses.
Thanks also to Arild Husby, Takeshi Kawakami, Anna Santure, and Helle T. Baalsrud for fruitful discussions. Finally, we extend our gratitude to the residents of the study system, whose cooperation and friendly attitude has been invaluable. This study was supported by  H.J. designed the study and is the senior scientist.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data supporting the findings of this study are available in the Dryad Digital Repository with the identifier https ://doi.org/10.5061/ dryad.00000 000n.