Intense Browsing by Sika Deer (Cervus Nippon) Drives the Genectic Differentiation of Hairy Nettle (Urtica Thunbergiana) Populations

Many studies have inferred the way in which natural selection, genetic drift and gene �ow shape the population genetic structures, but very few have quanti�ed the population differentiation under spatially and temporally varying levels of selection pressure, population �uctuation and gene �ow. In Nara Park (6.6 km 2 ; NP), central Japan, where several hundred sika deer (Cervus nippon) have been protected for more than 1,200 years, heavily-or moderately-haired nettle (Uritica thunbergiana) populations have evolved probably in response to intense deer browsing. Here, we analysed the genetic structure of two populations from NP and �ve from surrounding areas using ampli�ed fragment length polymorphism markers. A total of 546 marker loci were genotyped from 210 individuals. A Bayesian method estimated 5.5% of these loci to be outliers, which are putatively under natural selection. Neighbour-joining, Bayesian clustering and principal coordinates analyses using all-loci, non-outlier loci and outlier loci datasets showed that the two populations from NP formed a cluster distinct from the surroundings. These results indicate the genome-wide differentiation of the populations from NP and the surroundings. Moreover, these imply that: (1) gene �ow is limited between these populations and thus genetic drift is a major factor causing the differentiation; and (2) natural selection imposed by intense deer browsing has contributed to some extent to the differentiation. In conclusion, sika deer seems to have counteracted genetic drift to drive the genetic differentiation of hairy nettles in NP. This study suggests that a single herbivore species could lead genetic differentiation among plant populations.


Introduction
Natural selection, genetic drift and gene ow interplay to in uence genetic differentiation of populations (Templeton 2006; Futuyma and Kirkpatrick 2018; Freeland 2020).Natural selection can lead to local adaptation and then cause differentiation of traits or genes involved in the adaptation, while genetic drift can promote differentiation of neutral genes.On the other hand, gene ow generally impedes population differentiation.In recent decades, numerous empirical studies have analysed how natural selection, genetic drift and gene ow have shaped the genetic structure of plant populations occurring under different abiotic conditions such as altitude, temperature, precipitation and soil type using newly developed or improved molecular genetic techniques assisted with advanced mathematical and statistical methods (Műller et al. 2008 Many studies have also examined how biotic factors such as herbivory affect local adaptation (Pollard and Briggs 1982, 1984a, 1984b Leinonen et al. 2009).In this herb, populations under heavy herbivory by the diamond-back moth Plutella xylostella are monomorphic with trichome-producing form, while those under its low herbivory are polymorphic in this trait.However, there is little genetic differentiation at the genome-wide level between monomorphic and polymorphic populations.These results suggest that gene ow may prevent the differentiation of most genes despite the fact that differential herbivory by the moth species leads to the differentiation of genes responsible for the production of trichomes.
The Japanese stinging nettle Urtica thunbergiana (Rosales, Urticaceae) is a wind-pollinated, perennial herb with stinging hairs on leaves and stems.Nettle populations in Nara Park (hereafter NP; 6.6 km 2 ), which is adjacent to the urban area of Nara City, central Japan, are heavily haired (Fig. 1A, B), whereas those in the surrounding areas are lightly haired (Fig. 1C, D; Kato et al. 2008;Shikata et al. 2013).In NP, several hundred sika deer (Cervus nippon (Artiodactyla, Cervidae) have been protected for more than 1,200 years as sacred animal.Our previous work has revealed that heavily-haired nettles in NP are resistant to sika deer but not herbivorous insects such as the red admiral Vanessa indica (Kato et al. 2008;Iwamoto et al. 2014) and that variation in hairiness among nettle populations are genetically based (Kato et al. 2008;Hirata et al. 2019).These ndings suggest that heavy hairiness has evolved through natural selection under intense browsing by sika deer.In addition, nettles from NP exhibit lower plant height, higher degree of branching, earlier and longer owering, higher resource allocation to leaves and in orescences, and higher proportion of monoecious individuals than those from a surrounding area where deer are rarely found (Hirata et al. 2019).Most of these traits are assumed to have evolved to cope with intense browsing by deer as well as hairiness, although some traits may also have been in uenced by other factors (Hirata et al. 2019).Nettles are patchily distributed in NP and surrounding areas, and thus their evolution and differentiation should be in uenced by genetic drift and gene ow.Therefore, nettle populations from NP and surrounding areas provide an excellent system for studying how natural selection, genetic drift and gene ow interplay to shape the evolution of phenotypic traits.
This study aims to assess the relative importance of natural selection, genetic drift and gene ow to the genetic differentiation of heavily-, moderately-and lightly-haired populations from NP and surrounding areas.For this purpose, we performed a comprehensive genetic analysis using ampli ed fragment length polymorphism (AFLP) markers.More speci cally, we (1) compared genetic diversity among nettle populations, (2) examined the degree of genetic differentiation among populations of NP and the surrounding areas, (3) identi ed F ST outlier loci, or signals of natural selection, on AFLP markers, (4) examined phylogenetic relationships among populations, and (5) quanti ed the population genetic structure.

Sampling Sites
Nara Park is adjacent to the town of Nara City, central Japan, having an area of 6.6 km 2 (Fig. 2).World Heritage Sites such as Todaiji Temple, Kofukuji Temple and Kasugataisha Shrine are located on the west side of NP, and grassy and woody areas spread in the remaining areas.In NP, hundreds of sika deer (Cervus nippon) have been protected for 1200 years for religious reasons (Torii and Tatsuzawa 2009), and its number reached 1095 in 2011 (Foundation for the Protection of Deer in Nara Park 2020).In the grassy and woody areas, patchy colonies of U. thunbergiana are found at a distance of one to a few hundred metres from one another (Kato 2001;Shikata et al. 2013).Here we refer to each colony as a population for expediency.
We sampled leaves of U. thubergiana for DNA extraction from seven sites (Fig. 2): Todaiji (110 m a.s.l.) located in the central area of NP; Notogawa (150 m) in a marginal area of NP; Kohbohji (100 m) in a vacant lot of a residential area close to NP; Nara Women's University campus (80 m) in the urban area near NP; Sakurai (140 m), Takatori (550 m) and Kouchidani (280 m) at the edges of cedar or cypress forests located 17 to 28 km from NP (see Shikata et al. 2013 for topographical details).Of them, the Todaiji population (hereafter TO; patch size 4 × 10 3 m 2 ) is subjected to the most intensive browsing by sika deer (Shikata et al. 2013).The Notogawa population (NO; 1.4 × 10 3 m 2 ), which grows on steep banks of Notogawa Stream, avoids intense deer browsing.The Kohbohji population (KO; 0.38 × 10 3 m 2 ) is almost free from deer browsing because a wire fence prevents the access by deer.The population in Nara Women's University campus (NWU; 2.4 × 10 3 m 2 ) also scarcely suffers deer browsing because the campus is enclosed by woody fences.On the other hand, no or few deer were found around the Sakurai (SA; 0.2 × 10 3 m 2 ), Takatori (TA; 22.5 × 10 3 m 2 ) and Kouchidani (KC; 2.5 × 10 3 m 2 ) populations before 2011 (Shikata et al. 2013), but they have been frequently observed around TA and KC since 2016 (HS, personal observation).

Leaf traits
We determined the among-population difference in leaf area and the number of stinging hairs on a leaf by common garden experiments and re-analysis of Shikata et al's (2013) data on wild populations.Shikata et al. (2013) randomly chose 32 U. thunbergiana individuals for each of the above-mentioned seven populations and collected two intact, fully grown leaves on the second or third node from the tip of main shoots of each individual in late May 2008.We carried out common garden experiments according to Hirata et al. (2019).Brie y, we collected 20 to 40 seedlings from all locations except Kouchidani in mid-March 2015 and were grown in a greenhouse at Nara Women's University (nettles from Kouchidani were not raised because of the limitation of greenhouse space).The structure and light conditions of the greenhouse were described in Hirata et al. (2009).We planted seedlings individually in owerpots (12.5 cm in diameter and 13.3 cm in depth) with potting soil (Takii-no-baiyodo, Takii & Co., Ltd., Kyoto, Japan).
We paired two seedlings from different sites and placed in a plastic tray (30 cm in width, 42 cm in length and 5 cm in depth) lled with water.We then arrayed trays on tables in the greenhouse and rotated once every week to avoid position effects.On the same day, we supplied water to trays.On 18 June 2015, we collected two leaves on the second or third node from the tip of the main shoot from each surviving individual.
We scanned the upper surface of collected leaves at 600 dpi using an image scanner (EPSON GT-X770, Seiko Co., Suwa, Japan).We visually counted the number of stinging hairs on the leaf surface from the digital images, and measured leaf area using free software LIA for Windows95 ver.0.373 (Yamamoto 1997).We calculated the number of stinging hairs per unit area (i.e., stinging hair density) by dividing the number of hairs by leaf area.We tested among-population differences in leaf area, stinging hair number and stinging hair density using sequential Šidák multiple comparison post hoc test (α = 0.05) implemented in generalized linear mixed models (GLMMs).All GLMMs included population as xed factor and individual as random factor.In the analysis for stinging hair density, we used its number as response variable and the logarithm of leaf area as offset term.Error distributions and link functions were Gaussian and identity-link for leaf area, and Poisson and log-link for stinging hair number and density.

DNA extraction and AFLP analysis
We randomly chosen 30 U. thubergiana plants in each population on 1 or 2 June 2010, and sampled one young leaf from each plant for DNA analyses.We extracted total genomic DNA using the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) with the manufacturer's protocol.

Analyses of genetic Data
We quanti ed genetic diversity within populations by the proportion of polymorphic loci and expected average heterozygosity.In addition, we examined genetic differentiation among populations by pairwise F ST value and Nei's genetic distance (Nei 1972).We computed these genetic diversity and differentiation indices using the AFLP-SURV ver.1.0 (Vekemans 2002).
To test the signi cance of the isolation-by-distance (Wright 1943) among nettle populations, we examined the correlation between pairwise F ST and pairwise geographic distance using a Mantel test with 10,000 permutations as implemented in the vegan package ver.2.5-6 (Oksanen et al. 2019) in R ver.3.6.2(R Core Team 2019).
To detect F ST outlier loci, which are putatively under natural selection, we employed a Bayesian approach implemented in BayeScan ver.2.1 (Foll and Gaggiotti, 2008) using the all-loci dataset.We run this analysis with a burn-in of 100,000 iterations, a thinning interval of 20, and a sample size of 300,000.We considered loci to be outliers when the logarithm of the posterior odds [log 10 (PO)] (i.e., the ratio of the posterior probability of the selection model to the neutral model) was greater than 0.5.
We inferred phylogenetic relationships among populations by the neighbour-joining method based on Nei's genetic distance for the all-loci, non-outlier loci and outlier loci datasets using PHYLIP ver.3.6 (Felsenstein 2005).
To detect population genetic structure, we applied the Bayesian model-based cluster analysis implemented in STRUCTURE ver.2.3.4 (Pritchard et al. 2000) for the three datasets.We did not determine the optimal number of actual genetic clusters using Ln P(D) (Pritchard et al. 2000) or ΔK (Evanno et al. 2005) and, instead, examined grouping patterns across k values (cluster numbers).This is because there is a large degree of uncertainty in the estimation of k and there are few biological reasons to assume an only single value for k (Meirmans 2015).

Results
The pattern of among-population variation in leaf area differed between wild and common-garden nettles (Fig. 3A, B).Wild nettles from TO and SA had signi cantly smaller leaf area than those of the other populations NO, KO, NWU, TA and KC, but common-garden nettles of the ve populations TO, SA, NO, KO and TA did not signi cantly differ from each other (Fig. 3A).In contrast, wild and common-garden nettles showed similar variation patterns in stinging hair number and density (Fig. 3C-F); whether wild or raised in the common garden, TO was most heavily haired followed by NO and KO, while the other were only lightly haired.
We genotyped a total of 546 AFLP loci in 210 individuals from the seven populations.The proportion of polymorphic loci was highest in TO (75.5%), followed by NO (67%), and lowest in NWU (31.7%) (Table 1).Likewise, expected heterozygosity was highest in TO (0.200), followed by NO (0.195), and lowest in NWU (0.128).
Table 1 Genetic diversity based on 546 AFLP loci N, sample size; PLP, percentage of polymorphic loci; H j , expected average heterozygosity or Nei's gene diversity.See Fig. 2 for abbreviations of populations All pairwise F ST values among the seven populations were signi cantly different from zero (P < 0.001; Table 2), indicating that genetic differentiation occurred among populations overall.However, the value between TO and NO in NP was very low (0.031; Table 2).Despite that, there was no signi cant correlation between pairwise F ST values and pairwise geographic distances (Mantel's r = 0.066, P = 0.44; Suppl 2); that is, the isolation-by-distance model (Wright 1943) was not supported in this study.The BayeScan programme identi ed 30 AFLP loci (5.5% of total loci) as F ST outliers with explicit differentiation among populations (log 10 (PO) > 0.5; Suppl 1).These outlier loci may be linked to genes that are under natural selection.On the other hand, the non-outlier loci could be putatively regarded as neutral with respect to selection.
We constructed neighbour-joining trees with three datasets (i.e., those which are composed of all loci, non-outlier loci or outlier loci).TO (heavily haired population) and NO (one of the two moderately haired populations) located in NP were grouped together into a strongly supported cluster in the trees constructed with the all-loci and non-outlier loci datasets (bootstrap values were 100%; Fig. 4A, B), whereas their cluster was moderately supported in the tree constructed with the outlier-loci dataset (77%; Fig. 4C).The other ve populations did not form distinct clusters (bootstrap values < 73%).Note that KO (another moderately-haired population) and NWU (a lightly haired population) close to or near NP were consistently distinct from TO and NO in all trees.
We analysed the genetic structure of the seven populations using STRUCTURE for the three datasets.In the results, TO and NO in NP consistently remained in the same group irrespective of datasets and k values, and they formed a group together at k ≥ 3 in the all-loci and non-outlier loci datasets (Figs.4A, B and Suppl 3a, b) and at k ≥ 4 in the outlier loci dataset (Figs.4C and Supple 3c).
We also examined the genetic similarity among the populations using PCoA for the three datasets.In biplots for the all-loci and non-outlier loci, TO and NO in NP were clearly separated from the other populations along the rst principal component, forming a group, while NWU near NP was distinguishable from the others along the second component (Figs.6A, B and Suppl 4a, b).For the outlier loci dataset, TO and NO were identi able along not the rst but second component, while NWU was separated along the rst component (Figs.6C and Suppl 4c).

Discussion
Our previous study have reported that nettles in NP are heavily haired, while those in the surrounding areas are lightly haired, and also suggested that this variation is genetically based (Kato et al. 2008;Shikata et al. 2013;Hirata et al. 2019).This notion is supported by the present study; common-garden individuals showed the pattern of among-population differences similar to wild individuals (Fig. 3C-F).On the other hand, leaf area varied a little among common-garden individuals from different populations compared with wild individuals (Fig. 3A, B).This suggests that among-population variation in leaf area is in uenced largely by environmental factors.Despite that, the among-population variation in the number of stinging hairs on a leaf was robust to environmental differences.The present study populations, therefore, can be assigned TO as heavily haired, NO and KO moderately haired, and NWU, SA, TA and KC lightly haired.
The TO in NP is conspicuously, heavily haired compared to the other populations (Figs. 1 and 3C-F).Evidently, this characteristic has evolved as a result of adaptation to intense deer browsing from our present and previous ndings that (1) stinging hairs serve as defence against sika deer but not herbivorous insects such as the red admiral butter y (Kato et  Another population from NP, or NO, is moderately haired.This is probably ascribed to the fact that NO is less accessible to deer because it grows on steep banks (Shikata et al. 2013).
All pairwise F ST values among the seven nettle populations based on the all-loci dataset were signi cant (Table 2), indicating that these populations are well differentiated from one another.TO and NO in NP are, however, relatively close according to neighbour-joining analysis, STRUCTURE and PCoA (Figs. 4A, 5A and 6A).The differentiation among populations cannot be explained by the isolation-by-distance theory (Wright 1943), because the genetic and geographical distances were not signi cantly correlated (Suppl 2).Instead, the differentiation appears to be related to the difference in stinging hair number and density (Fig. 3C-F).Genetically similar TO and NO are heavily or moderately haired, while the other populations but KO are lightly haired.This would lead to the inference that the differentiation of TO and KO in NP from the others is in uenced by natural selection.
In populations that are adapted to different habitats but experience gene ow, genes involved in the expression of adaptive traits may show larger differentiation than those which are neutral to natural selection.Such genes under strong selection could be detected as outlier loci (Holderegger et al. 2010;Nosil 2012).Neutral genes and DNA regions, meanwhile, would be assigned as non-outlier loci, and they are expected to show non-adaptive differentiation because of the interplay of gene ow and genetic drift (Nosil 2012;Freeland 2020).In the following sections, we discuss how natural selection, gene ow and genetic drift interplay to in uence the genome-wide differentiation of NP populations on the basis of results obtained by neighbour-joining analysis, STRUCTURE and PCoA for the non-outlier and outlier loci datasets.

Natural selection
The neighbour-joining analysis, STRUCTURE and PCoA for the non-outlier loci dataset indicate that TO and NO in NP are genetically separated from the other populations including KO and NWU located close to NP (Figs. 4B, 5B and 6B), implying that gene ow occurs more frequently between the two populations from NP than between populations in NP and the surrounding areas.Similarly, those analyses for the outlier dataset show similar patterns, but somewhat unclearly; in the neighbour-joining tree, the two populations from NP form a moderately-de ned clade with a bootstrap value of 77% (but 100% for the non-outlier loci datasets; Fig. 4); in STRUCTURE, they belong to the same genetic cluster at k ≥ 4 (but k ≥ 3; Figs. 5 and One Resource 3B); in PCoA, they are separated from the other populations along the second principal component (but the rst one; Fig. 6 and Suppl 4b).Thus, the genetic differentiation of the two populations from NP in outlier loci would not be so large as that in non-outlier loci.This could be explained by a substantial difference in the selection pressure imposed by deer browsing between the two populations from NP.As mentioned above, TO is subjected to intense deer browsing.This should exert a strong selective pressure on this population to maintain not only heavy hairiness but also small stature, high degree of branching, early and long owering period, high resource allocation to leaves and owers, and dominance of monoecy (Hirata et al. 2019).These characteristics would help to enhance resistance and tolerance to sika deer and to reduce damages on growth and reproduction (Hirata et al. 2019).In contrast, NO suffers only moderate browsing.Therefore, if alleles casing less effective defence are brought into the populations by gene ow, they would be rapidly eliminated from TO but may remain longer in NO, resulting in the differentiation of genes involved in the defensive traits (i.e., candidates of outlier loci) between the two populations.Another possible explanation may be that either or both of TO and NO have been affected by genetic drift.However, it is unlikely that genetic drift drives the differentiation of all outlier loci underlying defensive traits.Effects of genetic drift will be considered again later.
The presence of outlier loci may be a result of natural selection (Nosil 2012).The proportion of loci identi ed as outliers in this study was 5.5%.This gure is roughly in concordance with a survey of Strasburg et al. (2012), who found that the proportion of outliers ranged from 0.4 to 35.5% with a median of 4.2% through review of 22 plant genome scan studies.It is not known what functions our outlier loci have, but some may be related to the production of stinging hairs, while others to the expression of the plant defensive traits mentioned above.Loci underlying these traits remain to be identi ed.Nevertheless, our sampling design was insu cient to lter out outlier loci undergoing selection, because nettles with defensive traits against deer browsing are limited to occur in the narrow area of NP.Hence, some of these outlier loci probably also contain region-speci c loci that are unrelated to adaptation to deer browsing.

Gene ow
The neighbour-joining analysis, STRUCTURE and PCoA for the non-outlier dataset (Figs.4B, 5B and 6B) show that the genetic cluster of TO and NO in NP are separated distinctly from the surrounding populations.This implies that gene ow between populations in NP and the surrounding areas is limited.
The KO is interesting, because despite being moderately haired and occurring close to NP, it does not belong to the genetic cluster of NP populations regardless of datasets (Figs. 4, 5 and 6).This population is protected from deer browsing by a wire fence.Hence its moderate hairiness cannot be regarded as local adaptation to deer browsing.Hairiness of KO could be accounted for by founder effect.KO colonized at a vacant lot in 1980's when the house was removed and the ground was cleaned up.Its founders may have originated from a hairy population of NP which has a different set of neutral genes from those which TO and NO have.Although their descendants were released from selection imposed by deer browsing, the time since colonization may be too short for the reduction of hairs to proceed.This hypothesis can be con rmed by searching populations in NP with the same set of neutral genes as KO.
It is also worth noting that although TO and NWU are only 1.1 km apart from each other, they are morphologically and genetically distinct (Figs.4-6).This implies that gene ow between the two populations is hindered despite the spatial closeness.Given this, it is natural that gene ow is almost absent between populations from NP and more distantly surroundings such as SA, TA or KC.Such a situation would act to maintain the genetic distinctness of populations in NP.
Gene ow in nettles can occur through pollen and seed dispersal; pollen is dispersed by wind, while seeds are probably transported by mammals (Pollard and Briggs 1984b).Pollen of European stinging nettle, U. dioica, may be blown several hundred metres in open lands (Pollard and Briggs 1984b).Because nettle fruits are very small (ca. 1 mm in length) and their surface (perianth) is hairy, they would adhere to the fur of mammals.In addition, mature seeds falling on the ground would adhere to mammal's toes with soil.In the present study system, however, pollen dispersal should be hindered considerably in urban areas and forests (Holderegger et al. 2010;Miles et al. 2019), and seed vectors may hardly migrate across these areas.
The level of gene ow among populations in NP should be high, because TO located in the centre of NP and NO located in a marginal area of NP (1.9 km apart; Table 1, Fig. 2) have similar compositions of neutral loci (Figs.4B, 5B and 6B).Probably nettle populations in NP form a metapopulation in which constituent populations are connected with one another by gene ow via intervening populations.
Nevertheless, gene ow level may not be so high as to hinder their genetic differentiation from one another; otherwise, the pairwise F ST between TO and NO should have been statistically insigni cant (Table 2).

Genetic drift and genetic diversity
In general, the effect of genetic drift on the differentiation of neutral loci becomes apparent under reduced gene ow.As mentioned earlier, non-outlier loci identi ed in this study account for 94.5% and are substantially differentiated between the populations from NP and the surrounding areas.Considering that gene ow is almost absent or very low between the two groups of populations and that non-outlier loci are probably selectively neutral (Nosil 2012), the differentiation of these non-outlier loci may have been generated mostly by genetic drift.
The TO and NO in NP show higher proportions of polymorphic loci and higher degrees of expected heterozygosity than the surroundings, having higher genetic diversity (Table 1).This is apparently inconsistent with the hypothesis that natural selection operates strongly in nettle populations of NP, because selection for ttest genotypes should lead to the elimination of all the other genotypes (Freeland 2020).However, given that the proportion of outlier loci, which are putatively subject to natural selection, is 5.5%, this inconsistency may be negligible.There are also three other possible explanations for higher genetic diversity in the populations from NP. First, as mentioned before, patchily distributed populations in NP are connected to one another by weak gene ow, forming a large meta-population.This can reduce the effect of genetic drift and help to maintain relatively high genetic diversity.Second, populations in NP as a whole seem to have experienced neither bottlenecks nor resulting loss of genetic variation in the past, because the environment and landscape of NP are believed to have persisted without considerable changes for 1200 years.Third, although deer browsing pressure is intense in NP as a whole, the extent actually varies according to ne-scale topography.For instance, nettles growing on steep slopes, such as NO on the banks of a stream, can avoid heavy browsing, and accordingly they are moderately haired (Shikata et al. 2013).In addition, the surrounding populations (NO, NWU, SA, TA and KC) are isolated and their colony sizes are far small compared with that of the metapopulation of NP (see "Sampling Sites").Thus, high genetic diversity of TO and NO could be explained by meta-population structure, population stability and environmental heterogeneity.In plant populations, polyploidy has also been known as a factor causing high genetic diversity (Sampson and Byrne, 2012).However, polyploidy seems to be unlikely in the two populations from NP, because nettles from TO and a surrounding population, TA, produce fertile seeds in reciprocal crosses (unpublished data).

Conclusion
This study con rmed the Hirata e al's (2019) notion that among-population variation in the number of stinging hairs on a leaf in nettles is genetically based.Furthermore, we demonstrated that U. thunbergiana populations from NP, which clearly exhibit local adaptation to deer browsing, have differentiated genome-widely from the surroundings which are almost free from browsing.In addition, our data show relatively ne-scale genetic differentiation; it occurs between populations located 1.1 km apart from each other.Based on the proportion of putatively neutral loci (94.5%), divergence of most loci may be due to genetic drift under limited gene ow.Differentiation of outlier loci (5.5%), which are putatively subjected to selection, should have been promoted partly by one herbivorous animal, sika deer.Several studies demonstrated that a single herbivore species causes plant species to evolve resistant traits (e.g., the relationship between diamond-back moth Plutella xylostella and perennial herb Arabidopsis lyata; see Introduction) but no studies revealed that the herbivore even leads genetic differentiation among plant populations.Nettles under intense browsing by sika deer exhibit not only heavy hairiness but also low plant height, high degree of branching, early and long owering, high resource allocation to leaves and in orescences, and dominance of monoecy compared with those almost free from deer browsing (Hirata et al. 2019).These characteristics would help to enhance resistance and tolerance to sika deer and to reduce damages on growth and reproduction (Hirata et al. 2019).The expression of these traits should be under the control of many genes, most of which may be located on the outlier loci assigned in this study.
Identi cation of these genes remains to be performed but is very challenging.

Declarations
Authors' declaration authors declare that they have no con ict of interest.
; Voure'h et al. 2001; Arany et al. 2009; Garrido et al. 2012; Didano et al. 2014; Lehndal and Ågren 2015; Martin et al. 2015; Skaien and Arcese 2020), yet very few have evaluated the relative importance of natural selection, genetic drift and gene ow in shaping population genetic structure at different levels of their effects.A rare example is a series of studies of northern European populations of the perennial herb Arabidopsis lyata which consists of glabrous and trichome-bearing forms (Kärkkäinen and Ågren 2002; Kärkkäinen et al. 2004; Kivimäki et al. 2007; Løe et al. 2007; Gaudeul et al. 2007;

Table 2
Geographic distances (km; upper triangle) and pairwise F ST values (lower triangle) between seven nettle populations in Nara Park (TO and NO) and its surrounding populations (KO, NWU, SA, TA and KC) F ST values are based on 546 AFLP loci, and all of them were signi cantly different from zero with P < 0.001.See Fig. 2 for abbreviations of populations al. 2008; Iwamoto et al. 2014), (2) amongpopulation variation in stinging hair number and density is genetically based (Fig. 3C-F; Kato et al. 2008; Hirata et al. 2019), and (3) populations subjected to intense deer browsing are heavily haired, whereas populations almost free from browsing are lightly haired (Fig. 3C-F; Kato et al. 2008; Shikata et al. 2013).

Figures
Figures