• Nina Sletvold,

    1. Plant Ecology and Evolution, Department of Ecology and Genetics, Evolutionary Biology Centre, Uppsala University, Norbyvägen, 18 D, SE-752 36 Uppsala, Sweden
    Search for more papers by this author
  • Mathilde Mousset,

    1. Plant Ecology and Evolution, Department of Ecology and Genetics, Evolutionary Biology Centre, Uppsala University, Norbyvägen, 18 D, SE-752 36 Uppsala, Sweden
    2. Current address: Institut des Sciences de l’Évolution de Montpellier, Université Montpellier II, Place Eugène Bataillon, Montpellier Cedex 5, France
    Search for more papers by this author
  • Jenny Hagenblad,

    1. Plant Ecology and Evolution, Department of Ecology and Genetics, Evolutionary Biology Centre, Uppsala University, Norbyvägen, 18 D, SE-752 36 Uppsala, Sweden
    2. IFM Biology, Linköping University, SE-581 83, Linköping, Sweden
    Search for more papers by this author
  • Bengt Hansson,

    1. Molecular Ecology and Evolution Lab, Department of Biology, Lund University, Ecology Building, Sölvegatan 37, Lund, Sweden
    Search for more papers by this author
  • Jon Ågren

    1. Plant Ecology and Evolution, Department of Ecology and Genetics, Evolutionary Biology Centre, Uppsala University, Norbyvägen, 18 D, SE-752 36 Uppsala, Sweden
    Search for more papers by this author


Inbreeding depression is a key factor influencing mating system evolution in plants, but current understanding of its relationship with selfing rate is limited by a sampling bias with few estimates for self-incompatible species. We quantified inbreeding depression (δ) over two growing seasons in two populations of the self-incompatible perennial herb Arabidopsis lyrata ssp. petraea in Scandinavia. Inbreeding depression was strong and of similar magnitude in both populations. Inbreeding depression for overall fitness across two seasons (the product of number of seeds, offspring viability, and offspring biomass) was 81% and 78% in the two populations. Chlorophyll deficiency accounted for 81% of seedling mortality in the selfing treatment, and was not observed among offspring resulting from outcrossing. The strong reduction in both early viability and late quantitative traits suggests that inbreeding depression is due to deleterious alleles of both large and small effect, and that both populations experience strong selection against the loss of self-incompatibility. A review of available estimates suggested that inbreeding depression tends to be stronger in self-incompatible than in self-compatible highly outcrossing species, implying that undersampling of self-incompatible taxa may bias estimates of the relationship between mating system and inbreeding depression.

Plants exhibit extraordinary mating system variation, and studies of realized mating patterns and associated traits have been central to our understanding of the evolution of plant reproduction (Darwin 1876; Stebbins 1957; Lloyd and Schoen 1992; Barrett 2003, 2010). Selfing carries a twofold transmission advantage compared to outcrossing (Fisher 1941) and the main selective force maintaining outcrossing is believed to be inbreeding depression, that is, the reduced performance of inbred compared to outbred progeny (B. Charlesworth and D. Charleswoth 1979). Basic models considering the balance between the two-fold transmission advantage of self-fertilization and inbreeding depression suggest that mating-system evolution should result in either of two stable endpoints, predominantly selfing or predominantly outcrossing populations (Lande and Schemske 1985). The magnitude of inbreeding depression is expected to evolve with the selfing rate, and the predicted dynamics depend on the genetic basis of inbreeding depression (D. Charlesworth and B. Charlesworth 1987; B. Charlesworth and D. Charlesworth 1999).

Inbreeding depression is caused by increased homozygosity, and the genetic mechanism underlying fitness reductions can be both increased exposure of recessive deleterious mutations and homozygosity at loci with heterozygote advantage (a.k.a. overdominance; B. Charlesworth and D. Charlesworth 1999; Crnokrak and Barrett 2002). Heterozygote advantage may contribute to inbreeding depression in some systems (e.g., Kärkkäinen et al. 1999; Crnokrak and Barrett 2002), but most studies suggest that it results primarily from exposure of recessive harmful alleles (reviewed by Carr and Dudash 2003; Charlesworth and Willis 2009). This can be inferred directly from experimental studies demonstrating purging of the genetic load (Willis 1999a; Fox et al. 2008), and indirectly from surveys documenting an association between population selfing rate and the magnitude and timing of inbreeding depression (Husband and Schemske 1996; Winn et al. 2011).

Self-fertilizing populations have less early-acting inbreeding depression, but similar late-acting inbreeding depression, compared to outcrossing and mixed-mating populations (Husband and Schemske 1996). Genetic load expressed as reduced survival during seed development, chlorophyll deficiencies in seedlings and reduced seedling survival is thus typically lower in selfing compared to outcrossing populations, whereas inbreeding depression expressed as reduced growth and fecundity of adults may be more similar. This is consistent with the hypothesis that inbreeding results in the purging of recessive lethal mutations expressed early in development, whereas inbreeding depression due to mildly deleterious mutations expressed late in development is difficult to purge even under considerable inbreeding.

The association between outcrossing rate and inbreeding depression has been a long-standing theme of both theoretical and empirical studies (e.g., Lloyd 1979; Lande and Schemske 1985; Schemske and Lande 1985; D. Charlesworth and B. Charlesworth 1987; Ågren and Schemske 1993; Willis 1993; Porcher et al. 2009), but recent reviews point out that current understanding is still severely limited by the amount and quality of empirical data available (Goodwillie et al. 2010; Winn et al. 2011). In particular, there are few estimates of inbreeding depression for highly selfing and highly outcrossing species, respectively, and few studies have quantified effects of inbreeding across all life-history stages of perennial plants (Igic et al. 2008; Winn et al. 2011).

Self-incompatibility, a genetically controlled mechanism that prevents self-fertilization, is widespread among flowering plants (Igic et al. 2008) and should be associated with strong inbreeding depression. Severe inbreeding depression is expected in self-incompatible species, because obligate outcrossing will effectively shield deleterious, recessive alleles in the heterozygous state, allowing accumulation of a large genetic load (D. Charlesworth and B. Charlesworth 1987). By comparison, highly outcrossing self-compatible populations may go through periods of increased selfing and purging, and can therefore be expected to harbor a lower genetic load. However, there are few empirical estimates of inbreeding depression in self-incompatible species that would allow an assessment of this hypothesis. This reflects the experimental problems associated with producing progeny by self-fertilization in self-incompatible taxa. Typically, bud-pollinations are used to circumvent the incompatibility reaction (Levin 1984; Kärkkäinen et al. 1999; Busch 2005), or alternatively, sib-crosses are used to generate related progeny (Nason and Ellstrand 1995; Wagenius et al. 2010).

Self-incompatibility should be associated with high inbreeding depression, but variation in the magnitude of inbreeding depression can still be expected both among populations, because population history and size will influence which deleterious mutations are present (Whitlock et al. 2000; Charlesworth and Willis 2009), and among traits, because of differences in the underlying genetic architecture. Moreover, if most inbreeding depression is caused by rare mutations that segregate at low frequencies (Charlesworth and Willis 2009), substantial among-family variation in inbreeding depression can be expected (e.g., Ågren and Schemske 1993; Dudash et al. 1997).

Here we quantify inbreeding depression across two growing seasons in two Scandinavian populations of Arabidopsis lyrata ssp. petraea, one alpine population from Norway and one coastal population from Sweden (populations N6 and S2 in Gaudeul et al. 2007). The species is a perennial herb with a well-characterized sporophytic self-incompatibility system (Charlesworth et al. 2003; Mable et al. 2003). European populations are self-incompatible, whereas several North American populations (A. lyrata ssp. lyrata) have evolved self-compatibility, which has been associated with a shift to inbreeding (i.e., with outcrossing rates < 0.5; Mable et al. 2005; Mable and Adam 2007; Hoebe et al. 2009; Foxe et al. 2010). The shift from self-incompatibility to self-compatibility is a frequent evolutionary transition (Igic et al. 2008), commonly associated with selection for reproductive assurance (Baker 1955), due to low pollinator visitation or reductions in population size (Reinartz and Les 1994; Cheptou et al. 2002; Mable and Adams 2007).

Estimates of fixation indices at 19 microsatellite loci indicate that the two study populations are each randomly mating, like the large majority of Scandinavian populations (Gaudeul et al. 2007), and hence should harbor a high mutational load. Consistent with this, an earlier study based on a single maternal line from a different Swedish population suggested high inbreeding depression (Kärkkäinen et al. 1999). We ask whether inbreeding depression is particularly strong at early life stages (i.e., during seed development, germination, and seedling survival), whether mutations causing chlorophyll deficiencies account for reduced survival of seedlings produced by selfing, and whether the magnitude or timing of inbreeding depression varies among populations and maternal lines. In addition, we compile available empirical data to examine whether in general self-incompatible plant species display stronger inbreeding depression than do highly outcrossing (t > 0.8) self-compatible species.

Materials and Methods


Arabidopsis lyrata ssp. petraea (L.) O'Kane and Al-Shehbaz (Brassicaceae) has a disjunct, patchy distribution in Europe (Jalas and Suominen 1994), and occurs across a wide altitudinal range in habitats characterized by moderate soil disturbance.

In 2007, we collected seeds from 50 plants in each of two large Scandinavian A. lyrata populations (>2000 individuals): an alpine population in Norway (Spiterstulen; 1100 m a.s.l., 61°41′N, 8°25′E), and a coastal population in Sweden (Stubbsand; <5 m a.s.l., 63°13′N, 18°58′E). The study populations are substantially differentiated at putatively neutral microsatellite loci (Gaudeul et al. 2007), as well as in phenotypic traits such as flowering phenology (Sandring et al. 2007), leaf trichome production (Løe et al. 2007), and tolerance to drought (Sletvold and Ågren 2012).


Progeny raised from field-collected seeds were grown for one generation and randomly outcrossed in the greenhouse. In 2008, we produced seeds through controlled self- and cross-pollination, respectively. To avoid the self-incompatibility reaction, we pollinated flowers in the bud stage with self-pollen (cf. Kärkkäinen et al. 1999). Bud pollination allows self-pollen tubes to grow prior to the activation of maternal proteins on the stigmatic surface that normally prevents germination of related pollen grains in self-incompatible species of the Brassicaceae (Cabin et al. 1996; Kachroo et al. 2002). Self-pollinations were conducted at the stage of bud development when petals first become visible between sepals. Preliminary crosses indicated that self-pollinations performed at earlier and later stages of bud development did not yield any seeds, which is consistent with observations of rapid developmental transition from self-compatibility to self-incompatibility in other members of the Brassicaceae (J. Nasrallah, pers. comm.). Because bud pollination avoids the incompatibility reaction, the estimate of inbreeding depression at the embryonic stage should not be confounded by any partial self-incompatibility. Each cross-pollinated flower received pollen from two random donors within the source population. We harvested fruits at maturity. For each parental plant and cross type, we counted the seeds and determined total mass of filled seeds to the closest 0.01 mg in each of three fruits. From these data we calculated mean seed size (mass) for each parental plant and cross type.

In a separate experiment, we assessed whether bud-pollination per se affected seed production (e.g., because of damage to the flower or the presence of immature ovules; cf. Cabin et al. 1996) by conducting cross-pollinations in both the bud and mature flower stage. Number of seeds per fruit resulting from cross-pollination in the bud versus mature flower stage did not differ significantly (one-way analysis of variance, F1,118 = 0.06, P = 0.81), suggesting that bud-pollination per se did not affect seed production.


To quantify inbreeding depression, we compared number of seeds per fruit and mean seed mass after controlled self- and cross-fertilization, and we documented the performance across two growing seasons of progeny resulting from controlled crosses. Because total failure of seed production could reflect a self-incompatibility reaction rather than inbreeding depression, we included only fruits that produced at least one viable seed in the calculation of inbreeding depression. As a consequence, our estimate of inbreeding depression expressed during embryo formation and seed development is likely to be conservative. In 2009, we planted nine replicates per crosstype of 24 maternal parents per population in 5.5 × 5.5 cm pots, yielding a total of 864 pots. The pots were filled with standard commercial potting soil (Yrkesplantjord, Weibulls Horto, Sweden) and a layer of 1-cm nutrient-poor topsoil (S-jord, Hasselfors Garden, Sweden). In each pot, we planted two seeds. The pots were arranged in randomized positions in 24 trays, and were then moved to a dark cold-room at 4°C for stratification of seeds for 4 days. Trays were subsequently transferred to a growth room and kept at 20°C 16-hour day (150 μEm−2s−1)/16°C 8-hour night. We rotated the trays every other week. After 20 weeks, the plants were transferred to a cold-room at 6°C 8-hour day (ca. 50 μEm−2s−1)/16-hour night, where they remained for 3 months. After this cold period, plants were moved to a greenhouse, where they were placed individually on greenhouse tables and were grown at 20°C 16-hour day/16°C 8-hour night for 3 months.

In the first growing season, we recorded germination, survival, growth, and flowering status (flowering or not flowering). Germination and seedling survival were scored after 1, 2, and 3 weeks, and seedlings were thinned to one per pot after 3 weeks in the growth room. After 12 weeks, we recorded the number of leaves and the maximum width and length of the leaf blade of the largest leaf on all plants. Rosette size was estimated as total number of leaves × leaf blade area (mm2) of the largest leaf, where the area of the leaf blade (A) was quantified as the area of an ellipse A = π(leaf blade length × leaf blade width/4). This estimate is highly correlated with aboveground dry mass in A. lyrata (r = 0.87, n = 184). Plant survival and flowering status were documented at 20 weeks, before transfer to the cold room.

In the second growing season, we documented plant survival, flowering status, fecundity, ovule and pollen production per flower, and final size. We recorded survival at the time of transfer to the greenhouse, and scored flowering status on a weekly basis after that. To estimate pollen production in flowering plants, we collected two buds per plant and dissected the anthers under a stereomicroscope. The mass of the pooled anthers was determined to the nearest 0.01 mg. To quantify ovule and seed production, we cross-pollinated two newly opened flowers per plant. Each flower received pollen from two randomly chosen pollen donors from the pool of plants that were the products of cross-pollination. Fruits were harvested at maturity, and we counted filled and empty seeds and determined mean seed mass of filled seeds to the closest 0.01 mg. As for parental plants, we only included fruits that produced at least one viable seed. We summed the number of filled and empty seeds to obtain number of ovules per flower, and we estimated seed set as the number of filled seeds divided by number of ovules. At the end of the experiment we recorded survival and total flower production, and harvested the aboveground parts of all plants. Plant parts were dried to constant mass at 70°C and weighed to the nearest milligram.

For each maternal parent and crosstype, we estimated viability as the number of established plants per seed planted (viability = proportion of seeds germinating × seedling survival × adult survival), fecundity as flower production per established plant (fecundity = proportion of plants flowering × mean number of flowers of reproductive plants), and total fitness as the product of parental mean number of seeds per fruit, viability, and mean biomass of offspring at harvest. Plant size is positively related both to flower production (Sandring and Ågren 2009) and plant survival (Løe 2006) in natural populations of A. lyrata, and is thus a good predictor of future fitness.


To verify the success of experimental selfing at the bud stage, we genotyped maternal plants and all surviving progeny in week 12 at 7 microsatellite loci (Table S1; Clauss et al. 2002; Woodhead et al. 2007). DNA was extracted from fresh leaves using Qiagen DNeasy 96 Plant Kit (Qiagen, Ltd., Hilden, Germany). Extractions were carried out according to the manufacturer's instructions using mechanical grinding of frozen tissue to homogenize the tissue before DNA extraction. The microsatellites were PCR-amplified in two touchdown multiplexes using Qiagen Multiplex PCR Kit (Table S1). The following PCR conditions were used: preheating for 95°C for 15 min, then 35 cycles at 94°C for 30 sec, annealing temperature for 90 sec, 72°C for 60 sec, followed by 60°C for 30 min, and an ambient hold temperature (multiplex specific annealing temperatures are given in Table S1). The PCR products from the two multiplexes were pooled and then separated based on length and primer labeling using an ABI Prism 3730 capillary Sequencer (Applied Biosystems). Genotypes were scored with GeneMapper 4.0 (Applied Biosystems, Foster City, CA).

The overall proportion of contaminants was 0.15, and ranged from 0 to 0.78 within maternal lines. We excluded from statistical analysis maternal lines with a proportion of contamination larger than 0.45 (three lines in the alpine and two lines in the coastal population), and individual contaminants in other cases (15 and 19 individuals in each of the two populations).


The effects of inbreeding on plant performance were examined by hierarchical mixed models including crosstype, population, and their interaction as fixed factors, and maternal parent nested within population and its interaction with crosstype as random factors. In analyses of first season response variables, tray was included as a random blocking factor when possible. Analyses of parental seed production per fruit, parental mean seed mass, germination, and seedling survival (3 weeks) were based on maternal parent means, and the interaction between crosstype and line could not be evaluated. Adult survival and proportion of plants flowering in the first and second season were analyzed with binomial errors and a logit link function (proc GLIMMIX; SAS 9.2, SAS Institute Inc., Cary, NC), whereas proportion of seeds germinating, proportion of seedlings surviving, rosette size (12 weeks), number of flowers, pollen mass, number of ovules per flower, proportion seed set, seed mass, and final biomass were analyzed with normal errors and identity link (proc MIXED). Fruit set among plants that were the product of selfing was low, and we could not evaluate the interaction between crosstype and maternal parent in analyses of progeny ovule production, seed set, and seed mass. To determine whether survival during the second season was related to rosette size at the end of the first season, we analyzed a model with binomial errors and a logit link function (proc GENMOD).

Inbreeding depression (δ) was calculated as relative performance for each maternal parent: δ = (wows)/max (wo, ws), where wo is performance of outcrossed progeny and ws is performance of selfed progeny (Ågren and Schemske 1993). Relative performance can take values between −1 and 1, where positive values indicate that progeny produced by outcrossing outperforms progeny produced by selfing. Population-level inbreeding depression was quantified as mean relative performance across families. We used the boot package in R (Canty and Ripley 2012; R version 2.15.1) to generate 1000 bootstrap samples of population-level inbreeding depression and calculated 95% confidence intervals by the bias corrected and accelerated (BCa) method. To determine whether loci causing inbreeding depression had similar effects across life stages, we calculated the correlations between maternal family inbreeding depression for seed production, seedling survival, adult survival, and biomass at harvest.


To examine whether inbreeding depression is stronger in self-incompatible species compared to self-compatible species with a high outcrossing rate, we compiled available published and unpublished estimates. Based on the recent review by Winn et al. (2011) and a literature search, we identified four self-incompatible species and 10 self-compatible species with a primary outcrossing rate > 0.8, for which estimates of inbreeding depression at four different life-history stages (seed development, germination, survival, and fecundity) were available (Table S2). The self-compatible species included seven angiosperms and three gymnosperms. We did not include studies that presented mean seed mass as the single estimate of inbreeding depression at the seed stage, because this estimate is likely to be confounded with effects quantified at later life-history stages. We also did not include studies where estimates of inbreeding depression were based on sib-crosses. We used one-tailed t-tests to examine the hypothesis that inbreeding depression for individual fitness components and for total fitness is stronger in self-incompatible than in highly outcrossing self-compatible species.



Parental plants produced fewer and smaller seeds per fruit following selfing compared to outcrossing and the magnitude of these effects did not differ significantly between populations (Table 1, Fig. 1). Inbreeding depression for seed production was 0.62 in the alpine and 0.60 in the coastal population, whereas inbreeding depression for seed mass was 0.07 in both populations (Table 1, Fig. 2). Seed mass varied among maternal parents (P < 0.001).

Table 1. Trait mean ± SD for the two crosstypes and mean and range of inbreeding depression (δ) based on maternal family means in two populations of Arabidopsis lyrata (Spiterstulen, n = 21 maternal families; Stubbsand, n = 22 maternal families). P-values associated with the fixed effects population, crosstype (selfing vs. outcrossing) and their interaction in mixed models are indicated (P-values < 0.05 in bold). Random effects (maternal line and its interaction with pollination treatment) are reported in the text
TraitSelfOutcrossδδ rangeSelfOutcrossδδ rangePopCrosstypePop × Cross
  1. a

    Arcsine square-root transformed prior to analysis.

  2. b

    Square-root transformed prior to analysis.

  3. c

    n = 16 maternal lines in the Spiterstulen population.

Parental no. seeds per fruit6.6±2.819.2±5.20.62−0.23 to 0.926.3±2.516.8±4.10.60−0.08 to 0.820.091<0.00010.20
Parental mean seed mass (mg)0.24±0.0370.26±0.0300.074−0.12 to 0.200.24±0.0350.26±0.0380.070−0.08 to 0.270.83<0.00010.98
Proportion germinatinga0.91±0.110.97±0.040.057−0.11 to 0.330.88±0.090.96±0.040.081−0.10 to 0.290.0980.00040.41
Seedling survival (3 weeks)a0.87±0.150.98±0.040.12−0.12 to 0.610.90±0.111.00± to 0.330.90<0.00010.78
Adult survival0.76±0.220.90±0.090.15−0.22 to 1.000.89±0.130.98±0.050.09−0.10 to 0.440.0009<0.00010.14
Rosette area 12 weeks (cm2)b121±44247±420.500.07 to 0.90130±63257±410.49−0.32 to 0.790.30<0.00010.95
Aboveground biomass (g)b0.62±0.290.94±0.260.32−0.17 to 0.911.05±0.351.56±0.390.31−0.10 to 0.69<0.0001<0.00010.18
Prop. flowering first season0.14±0.160.26±0.120.47−0.44 to 1.000.09±0.180.13±0.170.15−1.00 to 1.000.016<0.00010.41
Prop. flowering second season0.61±0.180.77±0.160.19−0.44 to 0.620.84±0.180.97±0.060.14−0.11 to 0.57<0.0001<0.00010.082
Number of flowers37±2858±270.36−0.53 to 0.95130±88157±490.24−0.63 to 0.85<0.0001<0.00010.72
Pollen mass per flower (mg)0.12±0.0350.14±0.0210.15−0.33 to 0.670.15±0.0330.16±0.0210.09−0.24 to 0.60<0.0001<0.00010.90
No. ovules per fruitc9.7±5.613.9±3.50.28−0.23 to 0.9410.1±4.015.3±2.90.34−0.16 to 0.820.66<0.00010.31
Seed set1,30.83±0.190.92±0.070.10−0.13 to 0.540.84±0.160.94±0.050.10−0.13 to 0.570.330.00580.59
Mean seed mass (mg)c0.17±0.0460.22±0.0400.22−0.11 to 0.660.17±0.0420.20±0.0200.15−0.16 to 0.510.98<0.00010.57
Total fitness (biomass)b2.6±2.015.3±6.20.810.49 to 1.004.8±2.925.1±11.20.780.39 to 0.98<0.0001<0.00010.15
Figure 1.

The effect of crosstype (self-pollination vs. cross-pollination) on fitness components in two populations of Arabidopsis lyrata: (A) number of seeds per fruit, (B) viability (proportion of seeds germinating × seedling survival × adult survival), (C) biomass at harvest, (D) fecundity (proportion of plants flowering × number of flowers per reproducing plant), (E) pollen mass per flower, and (F) number of ovules per fruit. Means + 1.96 SE are indicated. Inbreeding depression was statistically significant for all six fitness components (P < 0.0001; Table 1).

Figure 2.

Inbreeding depression for fitness components measured across two growing seasons in the Spiterstulen and Stubbsand populations of Arabidopsis lyrata. Bars indicate means with 95% confidence intervals. Total fitness was estimated as number of seeds per fruit × viability × biomass at the end of the experiment.


There was substantial inbreeding depression for viability in both populations, and effects were of similar magnitude at the seedling and adult stages. A high proportion of seeds germinated in both populations and treatments (range 0.88–0.97), and inbreeding depression for germination was 6–8% (Table 1). Survival at the seedling and adult stages were also high in both populations, with 9–15% reductions following selfing compared to outcrossing (Table 1). Almost all seedlings that were the product of outcrossing survived. Chlorophyll deficiency occurred in about 35% of the maternal lines (in 38% of 21 lines in the alpine population, and in 32% of 22 lines in the coastal population), accounted for 81% of the total seedling mortality in the selfing treatment in each of the two populations, but was not observed among offspring resulting from outcrossing. Adult survival was lower among plants from the alpine Spiterstulen population compared to plants from the coastal Stubbsand population (Table 1). Most of the adult mortality occurred during the second growing season, and most inbreeding depression for adult survival was due to differential mortality during the second growing season, and in the alpine population also during the simulated winter (Supplementary Table S3). Survival in the second season was positively related to rosette size in the first season (χ2 = 47.2, P < 0.0001). Inbreeding depression for viability components did not differ significantly between populations (Table 1, Fig. 1B); survival from the seed stage until the end of the experiment of selfed progeny was 27% lower than that of outcrossed progeny in the alpine Spiterstulen population, and 24% lower in the coastal Stubbsand population (Fig. 2).


There was strong inbreeding depression for size in both seasons. Inbreeding depression varied among maternal families in the first season (P = 0.041), but did not differ between populations in any season (Table 1). After 12 weeks of growth, rosette size of inbred plants was 50% smaller than that of outbred plants (Table 1), and varied among blocks (P = 0.0016). At the end of the experiment, plants from the alpine Spiterstulen population were smaller than plants from the coastal Stubbsand population (Table 1, Fig. 1C), and aboveground biomass of inbred plants was reduced by 32% and 31% compared to outbred plants in the two populations (Fig. 2). Biomass varied among maternal lines (P = 0.0068).


Fecundity was significantly reduced among inbred progeny compared to outbred progeny, and inbreeding depression for components of fecundity varied between 14% and 47% (Table 1). The proportion of plants that flowered in the first season was low, and flowering propensity was higher in the alpine Spiterstulen population than in the coastal Stubbsand population (Table 1). In the second season, the majority of plants from both populations flowered, and flowering propensity was higher in the coastal population (Table 1). Inbreeding reduced flowering propensity by 47% in the alpine and by 15% in the coastal population in the first season and by 19% and 14%, respectively, in the second season. Flowering propensity varied among maternal lines in both seasons (both P < 0.01). Flowering inbred plants produced fewer flowers compared to flowering outbred plants, and plants in the coastal population produced nearly three times as many flowers as plants in the alpine population did in the second season (Table 1, Fig. 1D). Inbreeding depression for fecundity components did not differ significantly between populations (Table 1) or maternal families (P > 0.09), and fecundity of selfed progeny was 41% lower than that of outcrossed progeny in the alpine Spiterstulen population, and 32% lower in the coastal Stubbsand population (Fig. 2).


Inbred plants produced less pollen per flower, fewer ovules per flower, and fewer and smaller seeds compared to outbred plants. Inbreeding depression ranged from 9% to 34%, and did not differ between populations (Table 1, Fig. 2). The coastal Stubbsand population produced more pollen per flower than did the alpine Spiterstulen population (Fig. 1E, Table 1). Ovule production varied among maternal lines (P = 0.014), but not among populations (Fig. 1F).


Inbreeding depression for total fitness was strong, and correlations among estimates of inbreeding depression across life stages were generally weak. The reduction in total fitness of selfed progeny relative to outbred progeny was on average 81% and 78% in the alpine and coastal population, respectively (Table 1, Fig. 2). Inbreeding depression for adult survival and biomass were significantly positively correlated in the alpine population (r = 0.52, P = 0.016, n = 21), but not in the coastal population (r = 0.21, P = 0.36, n = 22). Estimates of inbreeding depression for parental number of seeds per fruit and for seedling survival were not significantly correlated with each other or with estimates of inbreeding depression for adult survival and biomass (r = 0.03–0.32, all P > 0.30).


Inbreeding depression tended to be stronger for self-incompatible species compared to highly outcrossing self-compatible species across all four life stages (Table 2). Differences were statistically significant for seed formation, germination, and total fitness (with or without the seed stage included) when only angiosperms were considered (Table 2). Only the difference in inbreeding depression for germination was statistically significant when three gymnosperms (all self-compatible) were included in the comparison. The magnitude of inbreeding depression varied across life stages, with the strongest effects at the seed and fecundity stages in both self-incompatible and in highly outcrossing self-compatible species (Table 2).

Table 2. SI-status (self-incompatible or self-compatible), division (angiosperm or gymnosperm), primary selfing rate, inbreeding depression for four life-cycle stages (seed development, germination, survival, and fecundity), and cumulative inbreeding depression across four and three stages for self-incompatible and highly outcrossing (selfing rate < 0.20) self-compatible plant populations. Estimates of selfing rate are from Winn et al. (2011). Details about fitness components examined are given in Table S2
SpeciesSI-statusAngio/GymnoSelf-rateSeedGermSurvFecund4-Stage3-Stage (excl seed)References
  1. a

    P-values < 0.05 in bold.

Arabidopsis lyrataSIANA0.6050.0680.2050.3420.8070.512Sletvold et al., this study
Campanula rapunculoidesSIANA0.8770.2330.5540.8690.9940.955Vogler et al. 1999
Leavenworthia alabamicaSIANA0.4780.3670.1340.220.7770.572Busch 2005
Phlox drummondiiSIANA0.1820.228−0.2660.2130.370.231Levin 1984, Levin and Bulinska-Radomska 1988
Allium schoenoprasumSCA0.1980.5350.0850.0570.2920.7160.389Stevens and Bougourd 1988
Campanula americanaSCA0.0660.0490.0220.0940.2510.3690.336Galloway et al. 2003
Collinsia vernaSCA0.0540−0.020.0240.2760.280.279S. Kalisz, pers. comm.
Eichhornia paniculataSCA0.050.0950.0200.2550.340.27Toppings 1989
Lupinus perennisSCA0.1270.3440.0130.050.190.5020.241Michaels et al. 2008
Pachycereus pringleiSCA0.0160.5170.005−0.310.0080.375−0.293Molina-Freaner et al. 2003
Yucca filamentosaSCA0.1010.0290.3470.0830.1580.5110.496Pellmyr et al. 1997, Huth and Pellmyr 2000
Abies proceraSCG0.1350.2900.1170.7250.8280.757Sorensen 1999
Picea glaucaSCG0.1270.8740.0840.2580.3440.9440.554Fowler and Park 1983
Picea marianaSCG0.1620.5450.0740.2380.1890.740.428Park and Fowler 1984
Mean SI   0.5360.2240.1570.4110.7370.568 
Mean SC   0.3280.0630.0610.2690.5600.346 
Mean SC angiosperms   0.2240.0670.0000.2040.4420.245 
t-Test (effect of SI)a  t121.212.460.751.081.251.34 
t-Test (effect of SI) angiospermsa  t91.211.991.111.682.431.91 


In the two study populations of A. lyrata, selfing reduced total fitness after two growing seasons by about 80% compared to outcrossing. This reflected strong predispersal inbreeding depression, and moderate to strong effects on offspring viability and fecundity. The severe reduction in parental seed set and the occurrence of chlorophyll-deficient seedlings in the self-pollination treatment suggest that the populations harbor considerable numbers of early-acting lethals. Moreover, reductions in quantitative traits like biomass and flower production are likely to reflect effects of harmful mutations at a large number of loci. The results thus indicate that a mixture of alleles of large and small effect contributed to inbreeding depression, and suggest strong selection for the maintenance of outcrossing in these populations.

The magnitude and pattern of inbreeding depression in A. lyrata was consistent with expectations in self-incompatible species, with strong effects both on early viability and on quantitative traits expressed at later life-history stages. Estimates of predispersal inbreeding depression were particularly large. Strong effects on seed production may reflect accumulation of early-acting lethals, and the fact that numerous genes are first expressed during embryogenesis. Comparisons of seed set after cross-pollination in the bud stage and of mature flowers suggested that the estimate of inbreeding depression was not inflated by an inherent reduced seed set in the bud stage (cf. Cabin et al. 1996). Instead, the estimate obtained is likely to be conservative as only fruits with at least one viable seed were included when calculating the mean seed production of flowers pollinated with self-pollen. A strong reduction in number of seeds per fruit following inbreeding has also been documented in self-incompatible populations of Campanula rapunculoides (Vogler et al. 1999) and Leavenworthia alabamica (Busch 2005), but was not found after controlled crosses in Raphanus sativus (Nason and Ellstrand 1995), possibly reflecting among-species differences in historical inbreeding.

Inbreeding depression for germination and survival were of similar magnitudes as recorded for other outcrossing species (Husband and Schemske 1996; Winn et al. 2011). Mutations causing chlorophyll-deficiency could explain most of the reduction in seedling survival after selfing, and are known to contribute to inbreeding depression in a wide variety of cultivated and wild plant species (Willis 1992, and references therein). Mutations at many loci may cause chlorophyll-deficiency (Willis 1992), which means that relatively high frequencies of chlorophyll-deficient lethals can be maintained by mutation-selection balance in fully outcrossing populations. In this study, the proportion of families containing chlorophyll-deficient seedlings (32% and 38% in the two populations) was only about half of that observed after controlled selfing of 12 plants from another Swedish population of A. lyrata (Kärkkäinen et al. 1999), indicating among-population variation in the frequency of this category of lethals. Inbreeding depression for survival at the adult stage was mainly due to differential mortality during the second growing season in both populations, and also during simulated winter in the alpine population. Both the onset of winter conditions and reproduction require activation of new physiological processes and genetic pathways, and may therefore result in the expression of additional deleterious mutations. The magnitude of inbreeding depression for viability at the seedling and adult stages was similar, but estimates were only weakly correlated. This suggests that loci responsible for reduced survival differ at least partly between the seedling and adult stages.

There was strong inbreeding depression for plant size and fecundity, probably because these traits are influenced by many genes, each of which can be subject to deleterious recessive mutations. Inbreeding depression for growth and reproduction was stronger than the mean recorded for outcrossing species in the reviews of Husband and Schemske (1996) and Winn et al. (2011). Inbreeding depression for size was lower in the second growing season compared to the first (about 30% vs. 50%), which at least partly can be explained by size-dependent mortality after the size measurement in the first growing season. Mean inbreeding depression for flowering propensity (proportion of plants flowering) was lower than that reported for one family by Kärkkäinen et al. (1999), but the latter estimate was still within our recorded family-level range. In this study, effects on fecundity were mainly expressed as differences in flower production. In animal-pollinated species, it is well documented that floral display influences attractiveness to pollinators and thereby plant reproductive success (e.g., Sletvold and Ågren 2011). In both A. lyrata populations, outcrossed plants produced more flowers compared to selfed plants. Earlier experiments have demonstrated strong pollen-limitation and consistent pollinator-mediated selection for more flowers in the coastal Stubbsand population (Sandring and Ågren 2009). It is thus likely that the documented inbreeding depression for flower production would be aggravated by reduced attractiveness to pollinators of inbred plants in natural populations of A. lyrata.

There were considerable reductions in male and female gamete production following inbreeding. The reduction in pollen mass is likely to reflect fewer pollen grains, but may also be associated with reduced pollen quality. Male-sterility alleles are known to make significant contributions to inbreeding depression in several plant species (Willis 1999b; Busch 2005; Glaettli and Goudet 2006), and if inbreeding also affects pollen viability, the estimated effects on male fertility are conservative. The number of mature seeds per fruit was substantially reduced among inbred progeny, mainly reflecting a lower number of ovules per fruit, but also reductions in the proportion of ovules developing into seeds. In addition, mature seeds were smaller, which is likely to be associated with reduced progeny performance.

The similar magnitude of inbreeding depression for different fitness components in the two populations suggests that they harbor comparable genetic loads, reflecting a history of outcrossing in both populations. Within-population genetic diversity is lower in the alpine compared to the coastal population, indicating a stronger influence of genetic drift (Gaudeul et al. 2007). However, there was no evidence of any recent bottlenecks in the studied populations (Gaudeul et al. 2007), and the importance of genetic drift is likely to be restricted if the large population sizes seen today correspond to historical population sizes. Among-family variation in estimated inbreeding depression was statistically significant for plant size in the first season, indicating substantial variation in the segregation of harmful alleles for this trait.

The magnitude of inbreeding depression is context dependent (reviewed by Armbruster and Reed 2005; Fox and Reed 2010; Cheptou and Donohue 2011), most likely because both gene expression and selection vary among environments. In general, inbreeding depression has been found to be lower when estimated under benign greenhouse conditions compared to in the wild (e.g., Dudash 1990), suggesting that fitness reductions in A. lyrata would exceed the documented 80% if estimated at the sites of the source populations. If inbreeding depression approaches 100% and very few selfed progeny survive to reproduce, the efficacy of selection against deleterious alleles will be low (Lande et al. 1994). At high genomic mutation rates, selective interference among loci creates a threshold selfing rate for purging lethal alleles (Lande et al. 1994; Porcher and Lande 2005). The very strong inbreeding depression observed in Scandinavian populations of A. lyrata indicates that low levels of inbreeding would be inefficient at purging the genetic load.

A loss of self-incompatibility has occurred repeatedly among North American populations of A. lyrata (Mable 2008; Hoebe et al. 2009; Willi and Määttänen 2010), but has not been observed in the European range. It seems unlikely that current selection for reproductive assurance should be stronger in North America compared to Europe. Selfing A. lyrata populations in North America are neither particularly small, sparse nor situated at the edge of the current distribution (Mable and Adam 2007), whereas many Scandinavian populations are indeed small and scattered, and occur in alpine habitats that are likely to experience strong pollen limitation. It has been suggested that the colonization of North America led to a long-term population bottleneck that reduced genetic variation (Ross-Ibarra et al. 2008), potentially lowering the genetic load (Pujol et al. 2009) and the initial fitness reduction associated with a transition to inbreeding (Foxe et al. 2010). If so, inbreeding depression may be lower in the North American range compared to the Scandinavian, even in fully outcrossing populations. Controlled crosses (self-pollination vs. within-population outcrossing) of two plants from a self-incompatible North American population of A. lyrata (Stift et al. 2013) provided estimates of inbreeding depression for germination, early survival, and growth that were within the ranges recorded for individual maternal parents in this study. However, a recent study based on controlled crosses conducted on twelve plants per population in a total of 18 populations reports markedly lower inbreeding depression for germination and flower production in both selfing and outcrossing A. lyrata populations in the North American range (Willi 2013), consistent with the hypothesis of lower inbreeding depression in this part of the species range.

More generally, our review of available estimates of inbreeding depression suggested that self-incompatible species tend to display stronger inbreeding depression than highly outcrossing self-compatible species. This may be because self-incompatible populations historically and on average are characterized by lower levels of inbreeding and purging of deleterious alleles than are highly outcrossing self-compatible populations. Self-incompatibility is widespread in the angiosperms (Igic et al. 2008), and the results suggest that undersampling of self-incompatible taxa may bias estimates of the relationship between mating system and inbreeding depression in seed plants.

This study has documented strong selection against the evolution of self-fertilization in two Scandinavian populations of A. lyrata, and the results suggest that a mixture of large- and small-effect alleles underlie inbreeding depression in this species. It is now possible to focus on affected traits and use expression studies and QTL mapping to identify the number of genes involved, and their relative effect sizes (cf. Remington and O'Malley 2000; Charlesworth and Willis 2009). The combination of mating system variation (e.g., Mable 2008) and abundant genetic resources (Hu et al. 2011) renders A. lyrata an excellent system in which to address the causes and consequences of the evolutionary transition from outcrossing to selfing.

Associate Editor: J. Etterson


The authors thank L. Lehndal, S. Bolinder, T. Preuninger, C. Wimmergren, and J. Glans for greenhouse assistance, S. Barrett, E. Elle, S. Kalisz, K. Karoly, and A. Winn for help in retrieving data on inbreeding depression from published and unpublished sources, J. Nasrallah for discussion, and D. Schemske for helpful comments on the manuscript. Financial support from the Swedish Research Council to NS, BH, and JÅ is acknowledged. The authors declare no conflict of interest.