Predicting Heterosis and Inbreeding Depression from Population Size and Density to Inform Management Efforts

Received: 10 January 2020 | Accepted: 27 March 2020 DOI: 10.1111/1365-2664.13643

RESEARCH ARTICLE

Predicting heterosis and inbreeding depression from population size and density to inform management efforts

Linus Söderquist | Anna Broberg | Viktor Rosenberg | Nina Sletvold

Plant Ecology and Evolution, Department of Ecology and Genetics, Evolutionary Abstract Biology Centre, Uppsala University, Uppsala, 1. Effective population size should be positively related to census size and den- Sweden sity, and it is expected to influence the strength of genetic drift, inbreeding Correspondence and response to selection, and thus the distribution of the genetic load across Linus Söderquist Email: [email protected] populations. 2. We examined whether census population size and density predict the strength of Funding information Svenska Forskningsrådet Formas, Grant/ inbreeding depression, heterosis and population mean fitness at the seed stage Award Number: 2014-601 in the terrestrial orchid Gymnadenia conopsea by conducting controlled crosses

Handling Editor: Jin-Tian Li (self, outcross within and between populations) in 20 populations of varying size (7–30,000 individuals) and density (1–12.8 individuals/m2). In the largest population, we also examined how local density affects the occurrence of self-pollination with a pollen staining experiment. 3. The majority of populations expressed strong inbreeding depression at the seed stage (mean δID: min–max = 0.26: −0.53 to 0.51), consistent with a mainly outcrossing mating system and substantial genetic load. The effect of between- population crosses varied from strong outbreeding depression to heterosis (mean δOD: min–max = 0.05: −0.22 to 0.92), indicating varying influence of drift and selection among populations. 4. Census population size did not significantly predict the strength of inbreeding depression, heterosis or population mean fitness. However, inbreeding depression was positively and heterosis negatively correlated with population density. The proportion of self-massulae deposition was three times higher in sparse patches compared to dense ones (41% vs. 14%). 5. Combined effects of density-dependent pollinator behaviour and limited seed dispersal may cause stronger genetic sub-structuring in sparse populations and reduce the strength of the correlation between census and effective population size. The results point to the importance of considering population density in addition to size when evaluating the distribution of recessive deleterious alleles across populations. 6. Synthesis and applications. Management plans for threatened species often involve crosses between populations to restore genetic variation, a process termed genetic

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Journal of Applied Ecology published by John Wiley & Sons Ltd on behalf of British Ecological Society

wileyonlinelibrary.com/journal/jpe | 1459 J Appl Ecol. 2020;57:1459–1468. 1460 | Journal of Applied Ecology SÖDERQUIST et al.

rescue. This study indicates that such conservation efforts should be more successful if designed on the basis of population density in addition to population size, because we found population density predicted both the strength of heterosis and inbreeding depression across populations of Gymnadenia conopsea.

KEYWORDS density-dependent mating, genetic rescue, Gymnadenia conopsea, heterosis, inbreeding depression, population density, population size, selfing rate

1 | INTRODUCTION selection decreases (Lynch & Gabriel, 1990). Beneficial alleles may be lost via random drift, and inbreeding can lead to exposure of During the last century, many species have experienced significant recessive deleterious alleles that cause inbreeding depression reductions in abundance and local population sizes due to habitat (i.e. inbreeding load; Husband & Schemske, 1996). The strength of degradation and fragmentation (Ceballos, Ehrlich, & Dirzo, 2017; inbreeding depression should be weaker in small relative to large Sánchez-Bayo & Wyckhuys, 2019). A reduction in population size is populations because the inbreeding load should be reduced by ge- expected to lead to increased extinction risk and lowered adaptation netic drift (Hedrick & García-Dorado, 2016). Drift is expected to potential (Leimu & Fischer, 2008). Hence, the census size of a pop- randomly fix mildly deleterious alleles (and lose strongly deleterious ulation is regularly used as a measure of its viability, and it is an im- ones), reducing the difference in expression of recessive deleterious portant determinant of a species' conservation value in management alleles between progeny resulting from selfing versus outcrossing. decisions (IUCN, 2012). How suitable census size is as a measure of Closer kinship due to biparental inbreeding in small populations will a species' conservation value will depend on its correspondence to also contribute to diminished differences between outcrossing and effective population size, determined by the number of individuals selfing. Additionally, inbreeding load can be reduced by selection that actually contribute offspring to the next generation, and the against alleles when expressed as homozygotes (Keller & Waller, equality of their contributions (Wright, 1938). 2002). However, the removal of deleterious alleles through selection In most natural populations, census population size will be (i.e. purging) requires a relatively large population size to be effec- larger than effective population size, as not all adult individuals tive, and should mainly act on strongly deleterious alleles (García- breed. More interesting from a conservation viewpoint is that pop- Dorado, 2015; Winn et al., 2011). Fixation of deleterious alleles via ulations of equal census size can differ in several ways influencing drift and inbreeding can ultimately reduce population fitness and effective population size. One factor that may strongly affect the lead to a decrease in adaptation potential (Ellstrand & Elam, 1993; mating structure and thereby the effective size is population den- Young, Boyle, & Brown, 1996), increasing the extinction probability sity (Levin, 1988). In flowering plants, several studies have docu- of small populations (Whitlock, 2000). mented density-dependent pollinator behaviour, where pollinators Besides reducing population size, habitat fragmentation typ- visit fewer plants (Kunin, 1997) and more flowers per plant at low ically also leads to increased isolation. The combination of re- density relative to high (Grindeland, Sletvold, & Ims, 2005; Karron, duced gene flow and increased inbreeding and genetic drift will Thumser, Tucker, & Hessenauer, 1995). Density-dependent be- lead to stronger genetic differentiation between populations haviour should lead to shorter pollen dispersal distances, more (Templeton, Shaw, Routman, & Davis, 1990). As deleterious al- geitonogamy and lower outcrossing rates (Karron et al., 1995; Peter leles are fixed at random by genetic drift, different populations & Johnson, 2009), and stronger fine-scale genetic structure in can, by chance, fix different alleles (Keller & Waller, 2002). sparse populations (or patches) compared to dense ones (reviewed in In such cases, between-population crosses can restore het- Loveless & Hamrick, 1984; Vekemans & Hardy, 2004). For example, erozygosity and lead to increased fitness, that is, heterosis spatial aggregation was found to increase the magnitude of genetic (Crow, 1948; Whitlock, Ingvarsson, & Hatfield, 2000). Heterosis structuring within natural populations of Silene ciliata (Lara-Romero is expected to be strongest in small populations, where random et al., 2016). Knowledge of density-dependent mating patterns is fixation of mildly deleterious alleles should be common (Oakley important for understanding the correspondence between census & Winn, 2012; Spigler, Theodorou, & Chang, 2017). Crosses and effective population size, and ultimately for guiding the design between small populations can thus be used as a conservation of conservation plans. effort, that is, genetic rescue (Ingvarsson, 2001, reviewed in Effective population size influences many ecological and genetic Frankham, 2015). However, populations can also be adapted to processes. Small populations typically have low levels of genetic differing local conditions, where between-population crosses variation (Leimu, Mutikainen, Koricheva, & Fischer, 2006). The in- would lead to outbreeding depression, that is, lowered fitness fluence of genetic drift and inbreeding also increases in small pop- as compared to within-population outcrossing (Oakley, Ågren, & ulations (Ellstrand & Elam, 1993), whereas the efficiency of natural Schemske, 2015; Templeton et al., 1986). As small populations SÖDERQUIST et al. Journal of Applied Ecolog y | 1461 are expected to show low potential for local adaptation (Leimu & Fischer, 2008), the strength of outbreeding depression is predicted to increase with population size. To separate the inbreeding and drift load across populations, it is important to combine knowledge on the strength of inbreeding depression versus heterosis (Keller & Waller, 2002). Low inbreeding depression and no heterosis indicate a history of purging of deleterious alleles, whereas low inbreeding depression and high heterosis imply that drift is the dominating process determining population genetic structure (Paland & Schmid, 2003). This information is of vital importance for guiding management of threatened species and for the success of management efforts such as population rein- forcements or genetic rescue. In this study, we examined the effect of census population size and population density on the strength of inbreeding depression and heterosis (or outbreeding depression) via controlled crosses between and within 20 populations of the orchid Gymnadenia con opsea (L.) Br. R. Additionally, we conducted a pollen staining experiment in dense and sparse patches in a large population to examine if low density is associated with higher selfing. We predict to see lower mean fitness, lower inbreeding depression and higher heterosis in small populations and/or in populations with low population density, reflecting more inbreeding, stronger genetic drift and weaker response to natural selection, relative to large and dense populations.

2 | MATERIALS AND METHODS FIGURE 1 Map of Öland showing the location of the 20 Gymnadenia conopsea populations included in this study. Focal populations (n = 20) are marked with a circle, donor populations 2.1 | Study system and sites (n = 3) with a star

Gymnadenia conopsea (L.) Br. R. is a terrestrial orchid distributed across Eurasia, highly associated with hay meadows and pastures. 2.2 | Population size and density Gymnadenia conopsea has historically been common, but is currently declining in both number of populations and population sizes In June 2017, 2018 and 2019, we estimated census population sizes. (Meekers, Hutchings, Honnay, & Jacquemyn, 2012). It is a long-lived In small populations (<1,000 individuals), we counted all flowering perennial, non-clonal orchid that produces one inflorescence with individuals. In larger populations, we counted the number of flower- around 10–100 flowers. Flowering lasts for up to a month and flow- ing individuals in five to nine 1-m2 subplots and extrapolated to the ers open successively from the bottom of the inflorescence. Flowers total area, assuming equal density across the total area. We deter- contain two pollinia that attach to pollinators in pair or one by one, mined the total area by encircling the population using a GPS and and may be deposited on multiple flowers (Sletvold, Grindeland, QGIS (QGIS Development Team, 2009). We defined a distance of Zu, & Ågren, 2012). The main pollinators are diurnal and nocturnal 50 m without any flowering individuals as a population boundary. lepidopterans, and pollen limitation of seed production is typically The three yearly census population size estimates were highly cor- 10%–25% (Chapurlat, Ågren, & Sletvold, 2015). Fruits are mature related (r = 0.85–0.87, all p < 0.001), and we use the 2017 estimate 3–6 weeks after pollination. in all analyses. This study was conducted in 20 populations of G. conopsea, To quantify density of flowering individuals, we laid out three with varying population size and population density, located on the 1 × 2 m plots in 16 of the 20 populations (Table S1), and counted Swedish island Öland (Table S1; Figure 1). We selected the popula- the number of inflorescences in each plot. The mean of the three tions to cover a wide range of census sizes and an even geograph- plots was used as density. The plots did not overlap with the focal ical distribution on Öland. The populations were separated by plants in the crossing experiment (see below), but were haphazardly 2.4–232 km. Populations are mainly found in pastures, meadows, selected within 30 m of the focal plants, in three patches of plants mires and on alvar ground. If grazers were present, we fenced focal representative of the population as a whole. In three populations plants to avoid herbivory. (Resmo, Mensalvaret and Lilla Horn), limited population size or 1462 | Journal of Applied Ecology SÖDERQUIST et al. density constrained sampling and density plots therefore included that is, outbreeding depression. The mean across maternal families the focal plants. was used as the estimate of population inbreeding depression and outbreeding depression.

2.3 | Crossing experiment 2.4 | Pollen staining experiment To estimate inbreeding depression and heterosis, we conducted controlled crossings in the field during 18–27 June 2017. We hap- To quantify the proportion of self-massulae deposition in dense hazardly chose 16 focal individuals in each population, except in the and sparse patches of G. conopsea, we conducted a pollen staining smallest, Resmo, where we only could include four individuals (total experiment in the largest population, Störlinge, during 19–25 June n = 308). Before flowers opened, we caged plants with a mosquito 2018. We collected inflorescences in the field, and in the labora- net to prevent pollinator access and marked six consecutive, un- tory we removed wilted flowers and flowers with massulae depos- opened flowers in the lower half of the inflorescence. When flowers ited or pollinia removed. We stained both pollinia on every second had opened, we randomly allocated three different crossing treat- of the remaining flowers using histochemical stains that do not ments to the six flowers, each treatment replicated on two flowers: affect pollen transfer (Kropf & Renner, 2008; Peakall, 1989). We selfing (S), within-population outcrossing (W) and between-popula- used three different colours (fast green 1%, gentian violet 1% and tion outcrossing (B). We ensured that focal flowers had two intact rhodamine B pink 0.2%) and injected 0.3–0.5 µl stain into each pollinia and no massulae deposition, and emasculated the flowers to anther sac with a microsyringe. Staining efficiency was estimated avoid unintended selfing. We saturated the stigmas with pollen from to 0.63 ± 0.2 (M ± SD), see Appendix S3 for details on the staining each respective pollen donor. For S, we used pollen from the focal method. flower or from another flower on the same individual. For W, we In the field, we estimated patch density (individuals/m2) by used pollen from several other plants within the population. Donors counting inflorescences within a 2 × 2 m quadrat. Based on the ob- for the W treatment were haphazardly selected among plants not served range in flowering densities among populations (Table S1), part of the experiment and at least 5 m away from the recipient plant we selected patches corresponding to the extremes, with densities (if possible). For B, we used a mix of pollen from multiple individu- of either 1–3 individuals/m2 (sparse) or more than 8 individuals/m2 als from each of the three large populations (>2,000 flowering indi- (dense), separated from other included patches by at least 50 m. In viduals) as donor for all focal populations (Figure 1). Specific donor each patch, we placed three inflorescences stained with different individuals varied among recipient populations. We collected inflo- colours in plastic vials filled with nutrient solution (Figure S1). After rescences from each of the three donor populations the day before 48-hr exposure in the field, we scored pollinia removal and massu- use. We put the inflorescences on floristic foam, transported them lae deposition for each flower under a microscope. For each patch in cooled, Styrofoam boxes and stored them overnight in the fridge. density, we calculated the proportion of plants with self-massulae We did all crossings in one population on the same day, and caged deposition, the proportion of flowers with self-massulae deposition plants again until we collected the fruits around 3 weeks later (5–13 per plant and the proportion of self-massulae deposition per flower July). Some individuals dried out and wilted without fruits, and we in plants with self-deposition (Appendix S2). In total, we stained included 254 individuals in the final analyses. Differences in donor 327 flowers on 98 inflorescences, placed in 14 dense and 18 sparse inflorescence age (B treatment stored overnight) did not affect pol- patches over 7 days. len viability and seed production (one-way ANOVA; F3,56 = 0.707, p = 0.552), see Appendix S2 for details on this experiment. We estimated fitness (W) by counting the number of seeds con- 2.5 | Statistical analyses taining well-formed embryos for one randomly chosen fruit per treatment and individual (n = 254). For each fruit, we spread out the All statistical analyses were done in r, version 3.3.2 (R Core seeds evenly on a petri dish with a grid pattern. We counted the Team, 2016). To determine whether seed production differed be - number of seeds containing well-formed embryos in one-quarter tween crossing treatments, we used a linear mixed model and the of the grids and extrapolated to the whole disc. We calculated pop- package lme4 (Bates, Mächler, Bolker, & Walker, 2015). Number of ulation mean fitness as the mean number of seeds in the within- seeds with well-formed embryos was used as response variable population outcrossing (WW). Inbreeding depression (δID) and het- and population, crossing treatment (S, W and B) and their interac- erosis/outbreeding depression (δOD) were estimated as relative per- tion as fixed effects. Individual nested in population was included formance for each maternal family: δID = (WW − WS)/max(WW, WS) as a random effect. Statistical significance was determined using and δOD = (WB − WW)/max(WB, WW). Relative performance gives a a type three ANOVA from the lmerTest package (Kuznetsova, value between −1 and 1 (Ågren & Schemske, 1993). For δID, a posi- Brockhoff, & Christensen, 2017). We used planned contrasts treat- tive value means that W outperformed S, that is, inbreeding depres- ing the within-population cross (W) as a control (lsmeans package, sion. For δOD, a positive value means that B outperformed W, that Lenth, 2016) to determine in which populations inbreeding depres - is, heterosis, and a negative value means that W outperformed B, sion (S vs. W) and heterosis or outbreeding depression (B vs. W) SÖDERQUIST et al. Journal of Applied Ecolog y | 1463 were statistically significant. Number of seeds was square root 3 | RESULTS transformed to improve homogeneity of variance across populations (cf. Ives, 2015). Census population size ranged from 7 to c. 30,000 flowering in- We examined whether census population size (ln transformed) dividuals, and mean population density ranged from 1.0 to 12.8 and/or population density predicted the strength of inbreeding de- individuals/m2 (Table S1). Population size and density were signifi- pression, heterosis and mean fitness (square root transformed) using cantly positively correlated when the smallest population was in- multiple linear regression analyses. We initially tested for an inter- cluded (r = 0.49, p = 0.032, n = 16), but not when excluded (r = 0.43, action between population size and density, but this was never sig- p = 0.111, n = 15). The number of seeds with well-formed embryos nificant, and was removed from the final models. We conducted all per fruit varied among individuals, populations and crossing treat- multiple regressions first including and then excluding the smallest ments (Figure S2). The lowest number was zero, the highest 3,472 population, because its means were based on only four individual es- and the mean (±SE) across all treatments and populations was timates and were found to have high leverage in the analyses. To ex- 721.2 ± 24.7. amine whether high mean fitness is associated with weak heterosis The effect of crossing treatment on number of seeds varied and strong inbreeding depression, we calculated their correlations among populations (Table 1). Mean δID (±SE) across all popula- across populations. tions was 0.26 ± 0.03, and ranged from −0.53 ± 0.23 to 0.51 ± 0.15 To determine whether self-massulae deposition was more com- (Table S2, Figure 2). Statistically significant inbreeding depression mon in sparse patches relative to dense, we used a generalized lin- (significant contrast between WB and WS; Table S3) was found in ear mixed model including population density (dense vs. sparse) as a eight populations, and was bordering significance in Gårdbyvägen fixed factor, the number of open flowers per plant as a covariate and (p = 0.0506). Mean δOD (±SE) across all populations was 0.05 ± 0.03, date of placement in the field as a random factor. The probability and ranged from −0.22 ± 0.11 to 0.92 ± 0.05 (Table S2, Figure 2). of self-massulae deposition per plant was analysed with a binomial Statistically significant heterosis (WB > WW; significant contrast be- distribution, and the proportion of flowers with self-deposition per tween WW and WB; Table S3) was documented in two populations, plant and the proportion of self-massulae deposited per flower were and approached significance in the smallest population with only analysed with a normal distribution. four individuals (Resmo, p = 0.082). The estimate of outbreeding

TABLE 1 The effect of population, NumDF DenDF MS F-value p crossing treatment (S, W and B) and their interaction on the number of seeds with Population 19 234 618.02 9.955 <0.001 well-formed embryos per fruit (square Crossing 2 468 2,860.99 46.084 <0.001 root transformed) analysed with a linear Population:Crossing 38 468 89.67 1.444 0.0456 mixed model (n = 762). Individual nested in population was included as a random Abbreviations: DenDF, denominator degrees of freedom; MS, mean squares; NumDF, numerator factor degrees of freedom.

FIGURE 2 Relative performance (δID and δOD) of each focal Gymnadenia conopsea population ordered by increasing (a) census population size (RM < SL) and (b) population density. For δID, a positive value indicates inbreeding depression. For δOD, a positive value indicates heterosis, and a negative value indicates outbreeding depression. Significant differences identified with planned contrasts are marked with *, error bars indicate SE. RM, Resmo; SB, Skogsby; GS, Gillsättra; GV, Gårdbyvägen; MA, Mensalvaret; LH, Lilla horn; LR, Lindreservatet; LL, Långlöt; KL, Kristinelund; NB, Näsby; TB, Triberga borg; SA, Sandbyborg KV; Kvinneby; SG, Segerstad; AS, Alvara strand; AB, Alböke; ML, Melösa; VS, Vässby strandängar; HM, Hörninge mosse; SL, Störlinge 1464 | Journal of Applied Ecology SÖDERQUIST et al.

depression (WB < WW) was highest in the largest population, but was Figure S3) and excluding (r = −0.55, p = 0.02) the smallest population. not statistically significant in any population (Table S3). Mean pop- High population mean fitness was associated with high δID (r = 0.74, ulation fitness (±SE) varied 100-fold, from 16 ± 13 to 1,731 ± 195 p < 0.001, Figure S4a; r = 0.56, p = 0.01 excluding Resmo) and low δOD (Table S2). Mean population fitness across all population means was (i.e. outbreeding depression; r = −0.50, p = 0.03, Figure S4b), but the 758 ± 101. latter relationship was driven by the smallest population (r = −0.10, Population density, but not population census size, significantly p = 0.68 excluding Resmo). predicted the strength of inbreeding depression and heterosis Pollinia were removed and/or massulae received in 77 of 89 (Table 2; Figure 3). Mean δID increased with population density inflorescences with stained flowers, giving a visitation rate of (Figure 3a), whereas mean δOD decreased with density (Figure 3c), 87%. Visitation was observed on all dates and in all replicates. In and these results were consistent and stronger when excluding the sparse patches, self-massulae contributed to 41% of all massulae smallest population (Table 2). Neither population size nor population transferred, whereas the corresponding number for dense patches density was significantly related to mean population fitness (Table 2). was 14% (Figure S5). This increase was a result of a 72% higher Estimates of inbreeding depression and heterosis were negatively probability of self-massulae deposition per plant (Figure S6a) and correlated across populations, both including (r = −0.82, p < 0.0001, a 75% higher proportion of self-massulae deposition per flower

TABLE 2 Multiple regressions (n = 16) of δID, δOD or W (mean fitness, square root transformed) on census population size (ln transformed) and population density including and excluding the smallest population Resmo

Population size Population density

R2 β t-value p β t-value p VIF

δID 0.29 0.005 0.179 0.86 0.046 2.296 0.04 1.41 δID (excl. Resmo) 0.34 −0.026 −1.622 0.13 0.035 3.014 0.01 1.22 δOD 0.25 −0.007 −0.255 0.80 −0.044 −2.066 0.06 1.41 δOD (excl. Resmo) 0.28 0.026 1.550 0.15 −0.033 −2.672 0.02 1.22 W −0.01 0.612 0.507 0.62 0.758 0.810 0.43 1.41 W (excl. Resmo) −0.14 −0.887 −0.347 0.73 0.427 0.526 0.61 1.22

Abbreviations: R2, adjusted r square; VIF, variance inflation factor; β, regression coefficient.

FIGURE 3 Partial regression plots of (a, b) inbreeding depression (δID) and (c, d) heterosis (δOD) against (a, c) population density and (b, d) population size (ln) in Gymnadenia conopsea. Trend lines indicate a significant association. Corresponding statistical tests can be found in Table 2. The smallest population, Resmo, was excluded. SB, Skogsby; GS, Gillsättra; GV, Gårdbyvägen; MA, Mensalvaret; LH, Lilla horn; LR, Lindreservatet; LL, Långlöt; KL, Kristinelund; NB, Näsby; TB, Triberga borg; SA, Sandbyborg; KV, Kvinneby; SG, Segerstad; AS, Alvara strand; AB, Alböke; ML, Melösa; VS, Vässby strandängar; HM, Hörninge mosse; SL, Störlinge SÖDERQUIST et al. Journal of Applied Ecolog y | 1465

(Figure S6b) in sparse patches compared to dense ones (Table S4). In natural populations, population size and density are often pos- The proportion of flowers receiving self-pollen in inflorescences itively correlated, which could lead to apparent relationships be- with self-deposition was 33% higher in sparse compared to tween size and inbreeding depression that in reality are driven by dense patches, but this difference was not statistically significant correlations with density. (Table S4). Interestingly, we did not detect any association between mean fitness and population density or population size. Perhaps our estimate of fitness, number of seeds with well-formed embryos, is a 4 | DISCUSSION poor proxy for population performance. Plant fitness will not be proportional to seed production if germination and seedling survival Population census size is relatively straightforward to estimate for are strongly density-dependent (Campbell, Brody, Price, Waser, & many organisms, and it is regarded a major determinant of population Aldridge, 2017), or if the availability of suitable germination sites status and viability. Interestingly, the present study shows that popula- limits population growth rate more than seed number (Eriksson tion density is superior to size when predicting both heterosis and in- & Ehrlén, 1992). We are currently collecting demographic data to breeding depression in the perennial plant Gymnadenia conopsea. If this link population size and density as well as selfing rate to population is common, conservation efforts designed on the basis of population viability. size alone will be less successful than those that also consider density. The majority of the 20 studied G. conopsea populations ex- With increasing population density, the strength of inbreeding pressed strong inbreeding depression at the studied life stage, seed and outbreeding depression increased in parallel in G. conopsea. This production. This is consistent with a high outcrossing rate and a is consistent with an increasing inbreeding load expressed during substantial genetic load expressed during selfing. Mean inbreed- selfing, and with stronger selection leading to adaptive genetic dif- ing depression across all populations was 26%, which is lower than ferentiation with increasing population density. Several processes previously reported from two Norwegian G. conopsea populations that influence the magnitude of the genetic load may be density- (δID = 0.41–0.67; Sletvold et al., 2012), but similar to other orchid dependent. First, pollinators often probe more flowers per inflores- species (Smithson, 2006). Outcrossing species are expected to ex- cence in sparse patches or populations, potentially leading to more press strong inbreeding depression at early fitness components geitonogamous self-pollination and higher inbreeding at low density (Husband & Schemske, 1996), but inbreeding depression can also (Bernhardt, Mitchell, & Michaels, 2008; Grindeland et al., 2005; be strong at later life stages (Fischer & Matthies, 1998; Sletvold, Mustajärvi, Siikamäki, Rytkönen, & Lammi, 2001). This is supported Mousset, Hagenblad, Hansson, & Ågren, 2013; Smithson, 2006). by the 193% higher proportion of self-massulae among pollen de- In the Norwegian G. conopsea populations, estimates at the germi- posited in sparse G. conopsea patches compared to dense ones, nation stage were equally strong as at the seed production stage documented in the staining experiment in the largest population (Sletvold et al., 2012), indicating that the present study probably (Figure S5). Second, although the small seeds of orchids are capa- underestimates total inbreeding depression across the life cycle. An ble of long distance dispersal, the majority of seeds end up within a ongoing cultivation experiment may bring more clarity into effects few meters from the maternal plant (Brzosko et al., 2017; Jacquemyn on later life stages. et al., 2007, 2009). Short-range seed dispersal will likely lead to a The effect of between-population crosses was much less decrease in kinship with increasing distance between plants, and consistent than within-population crosses, suggesting variable substantial genetic substructuring has been reported from many or- influence of drift versus selection. As expected, the strongest chid populations (Chung, Nason, & Chung, 2004; Jacquemyn, Brys, heterosis effect was found in the smallest population, Resmo Vandepitte, Honnay, & Roldán-Ruiz, 2006; Jacquemyn et al., 2007, (δOD = 0.93). In total, substantial heterosis combined with weak 2009; Machon et al., 2003). We expect this substructuring to be inbreeding depression suggests fixation of deleterious alleles weaker in dense populations of G. conopsea, as the seeds near a by drift in four populations (Resmo, Sandbyborg, Vässby and given plant is expected to come from a higher number of maternal Hörninge) that vary widely in census size (7–20,000) but are all plants and represent a higher paternal diversity. This could be tested sparse (1–5 individuals/m2, no density for Sandbyborg). Two other with molecular markers. populations (Gillsättra and Lilla Horn) showed weak inbreeding In contrast, we found no statistically significant association depression and no heterosis, which could indicate a history of between census population size and inbreeding depression or het- purging of deleterious alleles. The largest (and dense) population, erosis. This further supports that mating is non-random in the pop- Störlinge, tended to show outbreeding depression in combination ulation, that is, that census size is invalid as a proxy for effective with strong inbreeding depression, which suggests efficient selec- population size. Previous studies have reported inconsistent relation and a low impact of drift. Outbreeding depression is often tionships between census population size and inbreeding depres- not observed until the F2 or later generation (Edmands, 2002), but sion, including positive (Angeloni, Ouborg, & Leimu, 2011), negative has been documented in F1 progeny in a few plant species (Oakley (Spigler et al., 2017) and no relationship at all (Angeloni, Vergeer, et al., 2015), including G. conopsea (Sletvold et al., 2012). The Wagemaker, & Ouborg, 2014; Hens, Pakanen, Jäkäläniemi, Tuomi, & strongest estimate of outbreeding depression found in this study Kvist, 2017; Michaels, Shi, & Mitchell, 2008; Oakley & Winn, 2012). (δOD = −0.22) is comparable to that reported for two G. conopsea 1466 | Journal of Applied Ecology SÖDERQUIST et al. populations separated by only 1.6 km in Norway (δOD = −0.23 to We thank J. Braun, A. Costa, E. Hellkvist, M. Uscka-Perzanowska, −0.27; Sletvold et al., 2012). On Öland, we used a common pol- L. Vikström, M.-H. Westlund and F. White for fieldwork assistance, len donor pool comprising three large populations from the en- and two anonymous reviewers for helpful comments on the manu- tire study range to maximize genetic diversity in the donor group. script. This study was funded by a grant from The Swedish Research The varied effects of between-population crossings are therefore Council Formas to N.S. (2014-601) and from Regnells botaniska re- likely to reflect variation in the genetic load among the focal popu- sestipendium, R. Sernanders, Tullbergs f. biol. forskn., B. Lundmans lations. We are currently analysing SNP data that can reveal if dif- Fund for Botanical Studies and Stiftelsen Th. Kroks donation to L.S. ferent deleterious alleles are fixed in the focal populations and if the variation in crossing effects can be related to genetic distance AUTHORS' CONTRIBUTIONS between populations. L.S. and N.S. conceived the study questions and design, L.S., A.B. Studying the association between contemporary population size and V.R. collected the data, conducted the analyses and wrote the and the fitness effects of controlled crosses involves some caveats. first draft of the manuscript, and all authors contributed substan- Genetic drift will usually require generations to pass before fixa- tially to revisions. tion of deleterious alleles, and if a reduction in size is too recent, we might not see pronounced effects yet. The considerable life span of DATA AVAILABILITY STATEMENT G. conopsea (up to 50 years, N. Sletvold, pers. comm.) means that Data available from the Dryad Digital Repository: https://doi. the flowering plants seen today could have established at a differ- org/10.5061/dryad.ncjsxksrf (Söderquist, Broberg, Rosenberg, & ent population size. Genetic data can inform on historical popula- Sletvold, 2020). tion sizes and inbreeding rates. Another possibility could be that the small populations included in this study are too few or too large to ORCID detect a strong signal of census size. Finally, populations that ex- Linus Söderquist https://orcid.org/0000-0002-9894-4119 perience fluctuations in size are expected to show weaker inbreed- Nina Sletvold https://orcid.org/0000-0002-9868-3449 ing depression and higher heterosis than stable populations (Spigler et al., 2017). Many orchids flower intermittently, and even though REFERENCES our three estimates of population size were highly correlated, longer Ågren, J., & Schemske, D. W. (1993). Outcrossing rate and inbreed- time series are necessary to determine the accuracy of the relative ing depression in two annual monoecious herbs, Begonia hir suta and B. semiovata. Evolution, 47, 125–135. https://doi. ranking of population sizes. org/10.1111/j.1558-5646.1993.tb01204.x Because small populations are expected to experience fit- Angeloni, F., Ouborg, N. J., & Leimu, R. (2011). Meta-analysis on the ness reductions caused by random fixation of deleterious alleles, association of population size and life history with inbreeding de- crosses between them are used for genetic rescue (reviewed in pression in plants. Biological Conservation, 144, 35–43. https://doi. org/10.1016/j.biocon.2010.08.016 Frankham, 2015). In G. conopsea, the 33%–92% improvement in Angeloni, F., Vergeer, P., Wagemaker, C. A. M., & Ouborg, N. J. (2014). seed production observed in the three populations with strongest Within and between population variation in inbreeding depression heterosis suggests that such crosses may indeed be efficient. In in the locally threatened perennial Scabiosa columbaria. Conservation addition, signs of outbreeding depression were restricted to three Genetics, 15, 331–342. https://doi.org/10.1007/s10592-013-0541-4 larger populations, suggesting minor concerns about upsetting local Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67, genetic adaptation at this life stage. Moreover, the results points to 1–48. https://doi.org/10.18637/jss.v067.i01 the importance of considering population density when designing Bernhardt, C. E., Mitchell, R. J., & Michaels, H. J. (2008). Effects of pop- such conservation efforts. If the significant association between ulation size and density on pollinator visitation, pollinator behavior, density and strength of inbreeding depression and heterosis reflects and pollen tube abundance in Lupinus perennis. International Journal of Plant Sciences, 169, 944–953. https://doi.org/10.1086/589698 stronger fine-scale genetic structuring in sparse populations caused Brzosko, E., Ostrowiecka, B., Kotowicz, J., Bolesta, M., Gromotowicz, A., by density-dependent mating patterns, similar effects can be ex- Gromotowicz, M., … Wojdalska, M. (2017). Seed dispersal in six spe- pected for many plants, with considerable implications for conser- cies of terrestrial orchids in Biebrza National Park (NE Poland). Acta vation management. We suggest that between-population crosses Societatis Botanicorum Poloniae, 86, 3557. https://doi.org/10.5586/ asbp.3557 are underused as a conservation measure for small populations Campbell, D. R., Brody, A. K., Price, M. V., Waser, N. M., & Aldridge, G. (cf. Frankham et al., 2011; Ralls et al., 2018), and that density (2017). Is plant fitness proportional to seed set? An experiment and should be considered in combination with census population size a spatial model. The American Naturalist, 190, 818–882. https://doi. in plant conservation plans involving crosses and reintroductions org/10.1086/694116 Ceballos, G., Ehrlich, P., & Dirzo, R. (2017). Biological annihilation via (cf. Espeland et al., 2017; Kirchner, Robert, & Colas, 2006) as well as the ongoing sixth mass extinction signaled by vertebrate population in red list criteria. losses and declines. Proceedings of the National Academy of Sciences of the United States of America, 114, E6089–E6096. https://doi. ACKNOWLEDGEMENTS org/10.1073/pnas.1704949114 Chapurlat, E., Ågren, J., & Sletvold, N. (2015). Spatial variation in pollina- We thank landowners for allowing us to work on their land, and tor-mediated selection on phenology, floral display and spur length in A. Wilks at Station Linné for all her help in organizing field work. SÖDERQUIST et al. Journal of Applied Ecolog y | 1467

the orchid Gymnadenia conopsea. New Phytologist, 208, 1264–1275. recruitment in the terrestrial orchid Orchis purpurea. New Phytologist, https://doi.org/10.1111/nph.13555 176, 448–459. https://doi.org/10.1111/j.1469-8137.2007.02179.x Chung, M. Y., Nason, J. D., & Chung, M. G. (2004). Spatial genetic struc- Jacquemyn, H., Wiegand, T., Vandepitte, K., Brys, R., Roldán-Ruiz, I., & ture in populations of the terrestrial orchid Cephalanthera longibrac Honnay, O. (2009). Multigenerational analysis of spatial structure in the teata (Orchidaceae). American Journal of Botany, 91, 52–57. https:// terrestrial, food-deceptive orchid Orchis mascula. Journal of Ecology, doi.org/10.3732/ajb.91.1.52 97, 206–216. https://doi.org/10.1111/j.1365-2745.2008.01464.x Crow, J. F. (1948). Alternative hypotheses of hybrid vigor. Genetics, 33, Karron, J. D., Thumser, N. N., Tucker, R., & Hessenauer, A. J. (1995). The 477–487. influence of population density on outcrossing rates in Mimulus rin Edmands, S. (2002). Does parental divergence predict reproductive gens. Heredity, 75, 175–180. https://doi.org/10.1038/hdy.1995.121 compatibility? Trends in Ecology & Evolution, 17, 520–527. https://doi. Keller, L. F., & Waller, D. M. (2002). Inbreeding effects in wild populations. org/10.1016/s0169 -5347(02)02585 - 5 Trends in Ecology & Evolution, 17, 230–241. https://doi.org/10.1016/ Ellstrand, N. C., & Elam, D. R. (1993). Population genetic consequences s 0 1 6 9 -5347(02)02489 - 8 of small population size: Implications for plant conservation. Kirchner, F., Robert, A., & Colas, B. (2006). Modelling the dynamics of in- Annual Review of Ecology and Systematics, 24, 217–242. https://doi. troduced populations in the narrow-endemic Centaurea corymbosa: A org/10.1146/annurev.es.24.110193.001245 demo-genetic integration. Journal of Applied Ecology, 43, 1011–1021. Eriksson, O., & Ehrlén, J. (1992). Seed and microsite limitation of re- https://doi.org/10.1111/j.1365-2664.2006.01179.x cruitment in plant populations. Oecologia, 91, 360–364. https://doi. Kropf, M., & Renner, S. (2008). Pollinator-mediated selfing in two de- org/10.1007/bf00317624 ceptive orchids and a review of pollinium tracking studies address- Espeland, E. K., Emery, N. C., Mercer, K. L., Woolbright, S. A., Kettenring, ing geitonogamy. Oecologia, 155, 497–508. https://doi.org/10.1007/ K. M., Gepts, P., & Etterson, J. R. (2017). Evolution of plant mate- s00442-007-0919-4 rials for ecological restoration: Insights from the applied and basic Kunin, W. E. (1997). Population size and density effects in pollination: literature. Journal of Applied Ecology, 54, 102–115. https://doi. Pollinator foraging and plant reproductive success in experimental org/10.1111/1365-2664.12739 arrays of Brassica kaber. Journal of Ecology, 85, 225–234. https://doi. Fischer, M., & Matthies, D. (1998). Effects of population size on perfor- org/10.2307/2960653 mance in the rare plant Gentianella germanica. Journal of Ecology, 86, Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest 195–204. https://doi.org/10.1046/j.1365-2745.1998.00246.x package: Tests in linear mixed effects models. Journal of Statistical Frankham, R. (2015). Genetic rescue of small inbred populations: Meta- Software, 82, 1–26. https://doi.org/10.18637/jss.v082.i13 analysis reveals large and consistent benefits of gene flow. Molecular Lara-Romero, C., García-Fernández, A., Robledo-Arnuncio, J. J., Roumet, Ecology, 24, 2610–2618. https://doi.org/10.1111/mec.13139 M., Morente-López, J., López-Gil, A., & Iriondo, J. M. (2016). Individual Frankham, R., Ballou, J. D., Eldridge, M. D., Lacy, R. C., Ralls, K., Dudash, spatial aggregation correlates with between-population variation M. R., & Fenster, C. B. (2011). Predicting the probability of outbreed- in fine-scale genetic structure of Silene ciliata (Caryophyllaceae). ing depression. Conservation Biology, 25(3), 465–475. https://doi. Heredity, 116, 417–423. https://doi.org/10.1038/hdy.2015.102 org/10.1111/j.1523-1739.2011.01662.x Leimu, R., & Fischer, M. (2008). A meta-analysis of local adaptation in García-Dorado, A. (2015). On the consequences of ignoring purging plants. PLoS ONE, 3, e4010. https://doi.org/10.1371/journal.pone. on genetic recommendations for minimum viable population rules. 0004010 Heredity, 115, 185–187. https://doi.org/10.1038/hdy.2015.28 Leimu, R., Mutikainen, P., Koricheva, J., & Fischer, M. (2006). How gen- Grindeland, J. M., Sletvold, N., & Ims, R. A. (2005). Effects of floral display eral are positive relationships between plant population size, fitness size and plant density on pollinator visitation rate in a natural popu- and genetic variation? Journal of Ecology, 94, 942–952. https://doi. lation of Digitalis purpurea. Functional Ecology, 19, 383–390. https:// org/10.1111/j.1365-2745.2006.01150.x doi.org/10.1111/j.1365-2435.2005.00988.x Lenth, R. V. (2016). Least-squares means: The R package lsmeans. Journal Hedrick, P. W., & García-Dorado, A. (2016). Understanding inbreeding of Statistical Software, 69, 1–33. https://doi.org/10.18637/jss.v069. depression, purging, and genetic rescue. Trends in Ecology & Evolution, i01 31, 940–952. https://doi.org/10.1016/j.tree.2016.09.005 Levin, D. A. (1988). The paternity pools of plants. The American Naturalist, Hens, H., Pakanen, V. M., Jäkäläniemi, A., Tuomi, J., & Kvist, L. (2017). 132, 309–317. https://doi.org/10.1086/284854 Low population viability in small endangered orchid popula- Loveless, M. D., & Hamrick, J. L. (1984). Ecological determinants of ge- tions: Genetic variation, seedling recruitment and stochasticity. netic structure in plant populations. Annual Review of Ecology and Biological Conservation, 210, 174–183. https://doi.org/10.1016/j. Systematics, 15, 65–95. https://doi.org/10.1146/annurev.es.15.110 biocon.2017.04.019 184.000433 Husband, B. C., & Schemske, D. W. (1996). Evolution of the magnitude Lynch, M., & Gabriel, W. (1990). Mutation load and the survival of small and timing of inbreeding depression in plants. Evolution, 50, 54–70. populations. Evolution, 44, 1725–1737. https://doi.org/10.1111/ https://doi.org/10.2307/2410780 j.1558-5646.1990.tb05244.x Ingvarsson, P. K. (2001). Restoration of genetic variation lost – The ge- Machon, N., Bardin, P., Mazer, S. J., Moret, J., Godelle, B., & Austerlitz, F. netic rescue hypothesis. Trends in Ecology & Evolution, 16, 62–63. (2003). Relationship between genetic structure and seed and pollen https://doi.org/10.1016/s0169 -5347(00)02065 - 6 dispersal in the endangered orchid Spiranthes spiralis. New Phytologist, IUCN. (2012). IUCN red list categories and criteria: Version 3.1 (2nd ed.). 157, 677–687. https://doi.org/10.1046/j.1469-8137.2003.00694.x Gland, Switzerland and Cambridge, UK: IUCN. Meekers, T., Hutchings, M. J., Honnay, O., & Jacquemyn, H. (2012). Ives, A. R. (2015). For testing the significance of regression coefficients, Biological flora of the British Isles: Gymnadenia conopsea s.l. Journal of go ahead and log-transform count data. Methods in Ecology and Ecology, 100, 1269–1288. https://doi.org/10.1111/j.1365-2745.2012. Evolution, 6, 828–835. https://doi.org/10.1111/2041-210x.12386 02006.x Jacquemyn, H., Brys, R., Vandepitte, K., Honnay, O., & Roldán-Ruiz, I. Michaels, H. J., Shi, X. J., & Mitchell, R. J. (2008). Effects of popula- (2006). Fine-scale genetic structure of life history stages in the tion size on performance and inbreeding depression in Lupinus pe food-deceptive orchid Orchis purpurea. Molecular Ecology, 15, 2801– rennis. Oecologia, 154, 651–661. https://doi.org/10.1007/s0044 2808. https://doi.org/10.1111/j.1365-294x.2006.02978.x 2-007-0861-5 Jacquemyn, H., Brys, R., Vandepitte, K., Honnay, O., Roldán-Ruiz, Mustajärvi, K., Siikamäki, P., Rytkönen, S., & Lammi, A. (2001). I., & Wiegand, T. (2007). A spatially explicit analysis of seedling Consequences of plant population size and density for plant–pollinator 1468 | Journal of Applied Ecology SÖDERQUIST et al.

interactions and plant performance. Journal of Ecology, 89, 80–87. Spigler, R. B., Theodorou, K., & Chang, S.-M. (2017). Inbreeding depres- https://doi.org/10.1046/j.1365-2745.2001.00521.x sion and drift load in small populations at demographic disequilib- Oakley, C. G., Ågren, J., & Schemske, D. W. (2015). Heterosis and out- rium: Inbreeding in small populations. Evolution, 71, 81–94. https:// breeding depression in crosses between natural populations of doi.org/10.1111/evo.13103 Arabidopsis thaliana. Heredity, 115, 73–82. https://doi.org/10.1038/ Templeton, A. R., Hemmer, H., Mace, G., Seal, U. S., Shields, W. M., & hdy.2015.18 Woodruff, D. S. (1986). Local adaptation, coadaptation, and popula- Oakley, C. G., & Winn, A. A. (2012). Effects of population size and tion boundaries. Zoo Biology, 5, 115–125. isolation on heterosis, mean fitness, and inbreeding depression Templeton, A. R., Shaw, K., Routman, E., & Davis, S. K. (1990). The ge- in a perennial plant. New Phytologist, 196, 261–270. https://doi. netic consequences of habitat fragmentation. Annals of the Missouri org/10.1111/j.1469-8137.2012.04240.x Botanical Garden, 77, 13–27. Paland, S., & Schmid, B. (2003). Population size and the nature of genetic Vekemans, X., & Hardy, O. J. (2004). New insights from fine-scale load in Gentianella germanica. Evolution, 57, 2242–2251. https://doi. spatial genetic structure analyses in plant populations. Molecular org/10.1554/02-663 Ecology, 13, 921–935. https://doi.org/10.1046/j.1365-294x.2004. Peakall, R. (1989). A new technique for monitoring pollen flow in orchids. 02076.x Oecologia, 79, 361–365. https://doi.org/10.1007/bf00384315 Whitlock, M. C. (2000). Fixation of new alleles and the extinction of Peter, C. I., & Johnson, S. D. (2009). Reproductive biology of Acrolophia small populations: Drift load, beneficial alleles, and sexual selection. cochlearis (Orchidaceae): Estimating rates of cross-pollination in Evolution, 64, 1855–1861. https://doi.org/10.1554/0014-3820(2000) epidendroid orchids. Annals of Botany, 104, 573–581. https://doi. 054[1855:fonaat]2.0.co;2 org/10.1093/aob/mcn218 Whitlock, M. C., Ingvarsson, P. K., & Hatfield, T. (2000). Local drift load QGIS Development Team. (2009). QGIS geographic information system. and the heterosis of interconnected populations. Heredity, 84, 452– Open Source Geospatial Foundation. Retrieved from www.qgis.org/ 457. https://doi.org/10.1046/j.1365-2540.2000.00693.x R Core Team. (2016). R: A language and environment for statistical comput Winn, A. A., Elle, E., Kalisz, S., Cheptou, P.-O., Eckert, C. G., Goodwillie, ing. Vienna, Austria: R Foundation for Statistical Computing. C., … Vallejo-Marín, M. (2011). Analysis of inbreeding depression in Ralls, K., Ballou, J. D., Dudash, M. R., Eldridge, M. D. B., Fenster, C. B., mixed-mating plants provides evidence for selective interference Lacy, R. C., … Frankham, R. (2018). Call for a paradigm shift in the ge- and stable mixed mating. Evolution, 65, 3339–3359. https://doi. netic management of fragmented populations. Conservation Letters, org/10.1111/j.1558-5646.2011.01462.x 11, 1–6. Wright, S. (1938). Size of population and breeding structure in relation to Sánchez-Bayo, F., & Wyckhuys, K. A. G. (2019). Worldwide decline of evolution. Science, 87(2263), 430–431. the entomofauna: A review of its drivers. Biological Conservation, 232, Young, A., Boyle, T., & Brown, T. (1996). The population genetic 8–27. https://doi.org/10.1016/j.biocon.2019.01.020 consequences of habitat fragmentation for plants. Trends in Sletvold, N., Grindeland, J. M., Zu, P., & Ågren, J. (2012). Strong inbreed- Ecology & Evolution, 11, 413–418. https://doi.org/10.1016/0169- ing depression and local outbreeding depression in the rewarding 5347(96)10045-8 orchid Gymnadenia conopsea. Conservation Genetics, 13, 1305–1315. https://doi.org/10.1007/s10592-012-0373-7 Sletvold, N., Mousset, M., Hagenblad, J., Hansson, B., & Ågren, J. (2013). SUPPORTING INFORMATION Strong cumulative inbreeding depression over two growing seasons Additional supporting information may be found online in the in two Scandinavian populations of the self-incompatible peren- Supporting Information section. nial herb Arabidopsis lyrata. Evolution, 67, 2876–2888. https://doi. org/10.1111/evo.12174 Smithson, A. (2006). Pollinator limitation and inbreeding depression in orchid species with and without nectar rewards. New Phyto How to cite this article: Söderquist L, Broberg A, Rosenberg V, logist, 169, 419–430. https://doi.org/10.1111/j.1469-8137.2005. Sletvold N. Predicting heterosis and inbreeding depression 01592.x from population size and density to inform management Söderquist, L., Broberg, A., Rosenberg, V., & Sletvold, N. (2020). Data efforts. J Appl Ecol. 2020;57:1459–1468. https://doi. from: Predicting heterosis and inbreeding depression from population size and density to inform management efforts. Dryad Digital org/10.1111/ Repository, https://doi.org/10.5061/dryad.ncjsxksrf 1365-2664.13643