| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
2 Pacific Southwest Research Station, USDA Forest Service, 4955 Canyon Crest Drive, Riverside, California 92507 USA; 3 Department of Botany and Plant Sciences and Center for Conservation Biology, University of California, Riverside, California 92521-0124 USA
Received for publication November 9, 1999. Accepted for publication May 2, 2000.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: coastal sage scrub • common garden • cumulative fitness • ecological genetics • environmental distance • geographic distance • genetic distance • local adaptation • restoration • seed source
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Millar and Libby, 1989
; Montalvo et al., 1997
). Several problems arise from poorly chosen transplantation material.
The most immediate problem is that transplants may be maladapted to local conditions. Early studies by 19th century forest scientists (reviewed in Langlet, 1971
), and continuing through to classic work of Turesson (1922, 1925, 1930)
, and the Carnegie group (Clausen, Keck, and Hiesey, 1940, 1947, 1948
; Clausen and Hiesey, 1958
), repeatedly demonstrated population differentiation for local adaptation through "common garden" experiments. In these experiments, plants of the same species were collected from different habitats and reciprocally grown together at those locations. Numerous additional studies have demonstrated a fitness advantage of local transplants relative to transplants taken from distant sites (Bradshaw, 1965, 1984
; Langlet, 1971
; and Levin, 1984
, for review; Schmidt and Levin, 1985
; Silander, 1985
; Wang et al., 1997
; Montalvo and Ellstrand, 2000a, b
). This home site advantage implies that if transplants are introduced from a distance, they may be poorly adapted to the transplant site.
Longer term problems can arise if nonlocal transplants and their progeny survive to flower and hybridize with native local inhabitants. Reproductive success and performance of such hybrid progeny can be affected by the genetic similarity of mates (e.g., Price and Waser, 1979
; Millar and Libby, 1989
; Guerrant, 1992, 1996
; Fenster and Dudash, 1994
; Knapp and Rice, 1994
; Montalvo et al., 1997
). Many studies of wild plant species have shown that mating between geographically distant populations of the same species results in reduced seed production and reduced progeny fitness relative to within-population crosses (reviewed by Levin, 1978
; Waser, 1993
). This fitness reduction is called outbreeding depression (Price and Waser, 1979
; Templeton, 1986
; Lynch, 1991
). If transplanted populations are genetically differentiated from the resident population or each other, outbreeding depression may result, compromising the short and possibly long-term success of restorations. It can take many generations for a population to recover the fitness deficits caused by wide outcrossing (Emlen, 1991
). The importance of outbreeding depression has been increasingly recognized for animal restoration projects (Emlen, 1991
; Gharrett and Smoker, 1991
; Rhymer and Simberloff, 1996
). Few experimental studies have examined the importance of outbreeding depression to the conservation or restoration of plant populations, and these have concentrated on rare species (Gerard et al., 1995
; Byers, 1998
). Given the prevalence of local adaptation detected by reciprocal transplant experiments, we expect that outbreeding depression may occur when geographically differentiated populations are placed in contact.
Two notable alternative mechanisms have been described as causing outbreeding depression observed in intraspecific F1 hybrids (Price and Waser, 1979
; Lynch, 1991
; Waser, 1993
; Schierup and Christiansen, 1996
). The "ecological" mechanism is predicted if parental lines are each adapted to different environments and hybridization leads to a 50% dilution of the locally adapted genome. The ecological mechanism is revealed if: (1) parental populations are shown to be adapted to different environments (i.e., when there is genotype x environment interaction); and (2) F1 hybrid progeny perform worse than the mean of the parental lines when tested together in the parental environments. For interspecific and among-subspecies crosses, this scenario is parallel to environment-dependent hybrid fitness (e.g., Anderson, 1948
; Wang et al., 1997
; Hatfield and Schluter, 1999
).
In contrast, the "genetic" mechanisms [physiological mechanisms in Roff (1997)
] derive from divergence in the genetic architecture of individuals in different populations by any combination of evolutionary processes, including selection and genetic drift. When hybridization disrupts these harmonious architectures, progeny suffer reduced fitness even if the parental populations are ecologically adapted to identical habitats and performance of parental lines in different environments is parallel (i.e., there is no genotype x environment interaction (Lynch, 1991
; Roff, 1997
). Even if there is no outbreeding depression detected in the F1 generation, or if heterosis occurs, this mechanism leads to outbreeding depression in the F2 or subsequent generations when recombination disrupts favorable epistatic or dominance interactions. For example, F2 breakdown has been observed after intraspecific and interspecific F1 heterosis independent of environment (reviewed by Levin, 1978
). Thus, failure to detect outbreeding depression in the F1 generation does not preclude its subsequent expression.
These two mechanisms of outbreeding depression are neither mutually exclusive (Lynch, 1991
) nor necessarily exhaustive, and their relative roles are hard to determine empirically. Nonetheless, reciprocal common garden experiments of progeny from controlled crosses can illustrate the involvement of an ecological mechanism. For example, Emlen (1991)
gives a model by which ecological outbreeding depression at a particular site can be predicted by the relative performance of parental sources at the same site. In addition, the magnitude of outbreeding depression from an ecological mechanism should be related to the difference in environments of any pair of crossed populations compared to the transplant environment. Given the importance of the environment as a selective force (Endler, 1977, 1986
), we expect that an index of environmental distance between a source and local population might be more representative of adaptive genetic distance than their physical separation or genetic distance. On the other hand, both geographic and genetic distances have been used successfully in past studies as proxies for evolutionary divergence of the parents (e.g., Price and Waser, 1979
; Fenster, 1991
; Edmands, 1999
). Genetic distance based on allozyme analysis is a measure of genomic differences among populations, including both neutral and adaptive differences (Hamrick et al., 1991
). The use of genetic distance to characterize population differentiation is more appropriate for detecting a genetic mechanism than geographic distance; however, geographic distance between sources and transplantation site is much easier to measure and may be compared with genetic distance to evaluate isolation by distance. We have no clear expectation of the degree of correlation between genetic or environmental distance and geographic distance because gene flow, selection, and the spatial arrangement of environments will affect the correlations on a case-by-case basis.
While several studies have addressed the relationship between fitness of progeny relative to spatial distance between parents (Price and Waser, 1979
; Schmidt and Levin, 1985
; Waser and Price, 1989, 1994
; Fenster, 1991
), most are either at spatial scales too small or involve too few distance classes to be relevant to restoration with widespread, geographically variable species. Studies on a small geographic scale that include hybridization within local and nonlocal demes are more relevant to detecting optimal outcrossing and the restoration of rare species. The appropriate scale of study for restoration of more common, widespread species is the interpopulation level. Thus, study populations will cover a larger geographic scale than in optimal outcrossing studies, and matings among short distance classes should be avoided to minimize confounding effects of inbreeding depression (e.g., Edmands, 1999
).
Following hybridization, different life-cycle components may vary in their susceptibility to the effects of outbreeding just as occurs following inbred matings (e.g., Waser, 1993
). The effect of wide outcrossing on different fitness components and whether fitness deficits accumulate through the life cycle are poorly known. Important fitness components include the success of a cross as measured by seeds per flower, the fraction of seedlings emerging from those seeds, and the survival and reproductive success of hybrids in different environments.
In this paper we describe experimental transplant and hybridization studies with two varieties of Lotus scoparius (Nutt.) Ottley (Fabaceae). The two varieties in southern California, L. s. var. brevialatus Ottley and L. s. var. scoparius, have been used indiscriminately in many restoration and seeding projects (Montalvo, personal observation) even though they differ in floral characters and primary distribution. Lotus s. var. scoparius occurs primarily in the coastal regions of California and north of Los Angeles, while L. s. var. brevialatus occurs primarily in the interior regions of Riverside, Los Angeles, San Bernardino, and San Diego Counties (Isely, 1981
; Steppan, 1991
). We predict "ecological" outbreeding depression in the F1 generation because (1) there is habitat correlated geographic variation in L. scoparius (Steppan, 1991
; Montalvo and Weaver, unpublished data), and (2) Montalvo and Ellstrand (2000b) found evidence for local adaptation that could potentially influence outbreeding depression if populations were hybridized. Here we evaluate the fitness of inter-population crosses in the greenhouse and the performance of F1 progeny under field conditions. We test the prediction that the performance of crosses and progeny depends on the genetic distances of the parents. Given the evidence for local adaptation, we predict that progeny fitness will be a decreasing function of the environmental distances of each parental population to the location of transplantation. We also examine the effects of hybridization on different life cycle components and whether geographic distance can be accurately used as a proxy variable for the other distance measures.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). The seeds typically germinate after scarification by fire, but some germinate in disturbed sites. One of the dominant species after fire, its abundance decreases in 5–10 yr. Plants are facultatively drought-deciduous, self-compatible, and insect-pollinated (Moldenke, 1976
; Hickman, 1993
). Bumble bees are among the primary flower visitors (Montalvo, personal observation; Jones and Cruzan, 1999
). Both varieties have 2 n = 14 chromosomes (Isely, 1981
).
Study sites
Our study populations occur in coastal sage scrub vegetation, an endangered habitat under constant pressure of urbanization and associated disturbances in densely populated southern California. The community is characterized by frequent fire and facultatively drought-deciduous shrubs (Kirkpatrick and Hutchinson, 1977
). Many of these shrub species regenerate after fire, either from the seed bank or by resprouting. Because of increased fire frequency near urbanizations and anthropogenic disturbances, the habitat undergoes frequent restoration and post-fire seeding.
Coastal sage scrub occurs in a heterogeneous Mediterranean climatic zone characterized by summer drought and generally equable temperatures. However, both precipitation and summer temperatures vary within this zone. The different floristic associations within this plant community vary with climatic, soil, and topographic variation, (Westman, 1983
), allowing recognition of four major floristic units in southern California: the Diablan, Venturan, Riversidian, and Diegan formations (Westman, 1983
). The variation in environmental factors and plant associations occurring in coastal sage scrub suggests that the more wide-ranging species may develop adaptive differences over a relatively small region.
We used six populations, distributed from Santa Barbara County south to San Diego County, California (Fig. 1). Two populations were from each of the three major types of coastal sage scrub vegetation (sensu Westman, 1983
: Venturan—E and SR; Diegan—AW and F; and Riversidian—LS and MR, Fig. 1). All sites had substantial, natural L. scoparius populations that served as sources of seeds for experiments and had not been seeded previously. Two of the six populations (MR and LS) were L. s. var. brevialatus and the remaining populations were L. s. var. scoparius. To provide seeds for experiments and leaf tissue for population genetic surveys, we sampled 60 individuals from every population (20 from each of three subpopulations per population) (Montalvo and Ellstrand, 2000a, b). Plants were randomly selected at 
10-m intervals along arbitrarily arranged transects within each subpopulation. Fruits were collected in summer 1994; leaves were collected in spring 1995. Leaf samples were placed immediately in plastic bags and kept on ice or refrigerated until processing for use in allozyme electrophoresis.
|
) and provided data for 13 allozyme loci, 12 of which were polymorphic. The enzyme systems included: malic dehydrogenase, MDH; isocitrate dehydrogenase, IDH; phosphoglucose isomerase, PGI; phosphoglucomutase, PGM; shikimate dehydrogenase, SKDH; aspartate alpha-ketogluconate aminotransaminase, AAT; aconitase, ACO; alcohol dehydrogenase, ADH; triose phosphate isomerase, TPI, and menadione reductase, MNR. We calculated Nei's (1978)
genetic distances among all source populations using BIOSYS-1 for PC (cf. Swofford and Selander, 1981
).
Environmental distance
We used soil, climate, and elevation to characterize the physical environment of each source and common garden site and to calculate environmental distances. Soil variables included pH, particle size percentages, % combustion carbon (C), % combustion nitrogen (N), cation exchange capacity (CEC), and Ca, Na, Mg, total P, and K concentrations (ppm). Nutrient, C, and pH measures were analyzed at the USDA Forest Service, Pacific Southwest Research Station's chemical analysis laboratory in Riverside, California (RFL) from samples of nine soil cores collected from each population, including three from each subpopulation, and from samples of five soil cores from each common garden plot. We bulked subsamples of cores from each subpopulation and garden plot and assessed particle size using a standard pipette method for soil texture analysis (Gee and Bauder, 1986
). We used the mean values for each soil variable for each site and common garden plot in analyses. Climatic variables from the weather station closest to each site were based on means for hydroyears 1972–1992 (1 July–30 June) and included total rainfall, among-year CV (coefficient of variation) of total rainfall, mean minimum temperature for January, and mean maximum temperature for June and July. We corrected temperature variables for differences in elevation between source site and station site by the adiabatic rate of cooling (-6°C/100 m rise in elevation). Using the 20-yr means, we calculated an index of effective precipitation that estimates water available to plant growth by the end of the growing season in June as [(rainfall from November to April)/(June maximum temperature)(%clay + %silt)].
We reduced the variables used to calculate an environmental distance matrix by eliminating all but the five least correlated soil variables (all r < 0.5), and using effective precipitation in place of its component variables. For this exclusion procedure, we included data from an additional six source sites used in our companion study of home site advantage (Montalvo and Ellstrand, 2000b
). Consequently, we used ten variables including: effective precipitation, elevation, CV of total rainfall, January minimum temperature, July maximum temperature, pH, Mg, P, N, and K. We standardized variables with Gower's ranging technique (Sneath and Sokal, 1973
) and calculated the matrix of environmental distances (EDIST) among all pairs of sites and common garden plots using SIMINT, option DIST in NTSYSpc (Rohlf, 1997
), which employs the average "taxonomic" distance of Sneath and Sokal (1973)
, Eij = 
, including all n variables for each pair of populations, i, j.
We calculated the mean environmental distance of both parental populations of a cross relative to the common garden where the progeny were planted (GardenEDIST) as the mean of (EDIST between parental population x and relevant garden plot); and (EDIST between parental population y and same garden plot). We used GardenEDIST to test for ecological outbreeding depression (see below) and to predict progeny performance when planted in environments foreign to both parental sources.
Geographic distance
We measured the linear distances in kilometres between each pair of source populations for the geographic distance matrix (GEODIST). Measures were the shortest map distance between locations. To examine the usefulness of GEODIST as a proxy variable for GENDIST or EDIST, we examined correlations between their distance matrices using the Mantel test in module MXCOMP of NTSYSpc. We also calculated mean geographic distances (GardenGEODIST) of parents to transplant sites as done for GardenEDIST and compared these two matrices to each other and GENDIST for the six crossed populations.
Crosses and common garden experiment
To produce parents for crossing experiments, we planted equal numbers of wild collected seeds from each of 60 maternal parents, by source population (20 per subpopulation), into plug flats in a randomized blocks design in the greenhouse at RFL on 8–9 February 1995. Boiling water was poured over the planted seeds to stimulate germination. We originally collected seed from each of three subpopulations at the source populations and grew 20 random seedlings from each one to maturity in an insect excluded greenhouse at RFL. We randomly selected 25 of the 60 plants from each source population to serve as maternal parents (pollen recipients); the remaining plants served as paternal parents (pollen donors). The crossing design was a population diallel that included crosses between individuals from different populations as well as from the same population (Fig. 2A). Pollen donors were randomly selected from the appropriate pool of paternal parents. All within-population crosses were done among individuals from different subpopulations to minimize inbred matings and potential inbreeding depression. Each day, the sequence of pollen donors was randomized, and every pollination type on a particular pollen recipient was done within 1 h. Order of maternal parents was also random. Each maternal parent was pollinated on 3–5 separate days, each time achieving a full set of crosses. A cross set incorporated one inflorescence on each of seven separate branches, each of five assigned a different pollen donor population with two assigned to within-population crosses (Fig. 2B). All flowers in an inflorescence (usually 2–5/inflorescence) were pollinated with freshly collected pollen from one pollen donor. This crossing design resulted in 36 different types of crosses, including reciprocals (Fig. 2A). Undisturbed flowers in our insect excluded greenhouse or caged flowers in the field did not set fruit (Montalvo, personal observation; also Jones and Cruzan, 1999
). In L. scoparius and numerous related species, pollen will not germinate until the stigmatic cuticle is ruptured by floral visitors (e.g., Bubar, 1958
; Lord and Heslop-Harrison, 1984
). To prepare flowers for pollination, we removed the keel and androecium from each recipient flower and examined stigmas through a magnifying lens for ruptured cuticles and presence of self-pollen. Flowers with ruptured cuticles or obvious pollen were discarded. The cuticle of each remaining flower was then ruptured with the tip of a clean forceps and immediately saturated with cross pollen. A total of 13 973 flowers were pollinated.
|
8 wk). Fruits and seeds were scored as mature once fruits were brown and seeds filled. To examine the effect of cross type on seed germination, we planted seeds in partitioned plug flats in a randomized block design in the greenhouse at RFL on 18–19 December 1996. To break seed dormancy, immediately before planting, we poured boiling water over seeds and soaked them for 10 min. We planted six replicates of 15 seeds per row for each of the 36 cross types, except for within population crosses in which we planted 12 replicate rows (90 seeds/cross and a total of 3240 seeds). Flats were placed in the greenhouse at ambient temperature except that night and daytime extreme temperatures were kept to within a range of 4.4°-32.2°C, respectively. We rotated flats and scored seedling emergence "blind" every other day until 12 February; that is, cross identity of individual plants was unknown to persons collecting data. To examine performance of F1 progeny under field conditions, we transplanted randomized seedlings into common gardens established in native coastal sage scrub vegetation at two of the six source population sites (Fig. 1). One plot (Motte garden) was at the University of California, Motte Rimrock Reserve (MR) in western Riverside County, California and is home to L. scoparius var. brevialatus. The other plot (Fallbrook garden) was at the Fallbrook Naval Weapons Station (F) in northwestern San Diego County, California and is home to L. scoparius var. scoparius. Site MR is hotter in summer, receives less rainfall, and experiences colder winters than the relatively coastal Fallbrook site. Experimental plots were fenced to exclude large vertebrates, and gophers were removed routinely.
We planted the young seedlings into randomized blocks on 13–14 February 1997. Vegetation was trimmed to near ground level to allow planting without tilling. Because we were short of seedlings from particular cross types, we planted more seedlings of the reciprocal of any short crosses to obtain roughly equal numbers of seedlings for each pairwise population combination. We planted three complete randomized blocks each at Motte garden and Fallbrook garden. Each complete block was a 6 x 21 grid of 2 dm x 2 dm cells. We planted an additional incomplete block at Motte garden where survival of other transplants had been low (Montalvo and Ellstrand, 2000b
), in a grid of 6 x 17 cells. One seedling was transplanted in the center of every cell by removing 3 cm diameter soil plugs and replacing them with plugs containing single seedlings. All seedlings were marked with a plastic toothpick and an aluminum tag showing grid coordinates marked each cell. We planted a total of 378 seedlings at site Fallbrook garden (18 seedlings per population combination with six replicates per block) and 480 seedlings at site Motte garden (at least 18 seedlings per combination).
We monitored survival and size within a day of transplantation into the field, 6 wk after transplantation (25–28 April), and at the end of the growing season in June 1997 and 1998. At each census, we measured plant height (from top of basal rosette to tip of longest branch) and canopy diameter in two perpendicular planes. All monitoring was done "blind." We used plant height as one measure of plant performance because of its correlation with different fitness components. Plant height was the best correlate of aboveground biomass for juveniles (Pearson's r = 0.84; P < 0.0001; N = 99) and log adult height and log adult volume were good estimators of flower production (r = 0.88 and r = 0.90, respectively, P < 0.0001, N = 105) (Montalvo and Ellstrand, 2000b
, and unpublished data). We allowed open pollination of flowers. Because plants had different flowering phenologies, we harvested them between 12 June and 23 July 1998 on four dates, each time removing only those reproductive individuals that were within their last week of flowering (few if any flowers remained) and before mature fruits began dispersing to minimize contamination of the local gene pool. We dried the plants to constant mass at 65°C, weighed them, and counted all fruits (mature and immature) on small plants. Total fruit number on large plants was estimated using fruit counts for a portion of the plant and multiplying by the number of portions. We examined this estimate's accuracy by counting all fruits on 23 of the subdivided plants and examining the correlation between the observed and estimated values. The two values were highly correlated (Pearson's r = 0.93, P < 0.0001).
Testing for outbreeding depression
To test for outbreeding depression resulting from crossing of individuals from genetically differentiated populations, we ran linear regression models with fitness components as response variables and distance measures as independent variables. Any significant decrease in a response variable with increasing distance was considered evidence of outbreeding depression. Greenhouse-phase components of fitness were first analyzed separately from field-phase components of fitness because only the latter could be affected by GardenEDIST since this is the average environmental distance of crossed parents to a common garden field site. For each of the 36 cross types, seed set was the number of filled seeds/flower averaged over usually 20 mothers. In our first model, mean SEED SET was the dependent variable, and genetic distance (GENDIST) of the crossed populations was the independent variable. The analysis was weighted by the number of mothers represented in each mean. Fruits of L. scoparius produce 1–2 seeds so seed set yields essentially identical results as fruits/flower (analyses not shown). In a second model, we examined the effect of GENDIST on mean fruit initiation (FRUIT INITIATION), an early component of seed and fruit set, to examine whether any non-random effects of GENDIST occurred up to fruit initiation prior to any seed abortion. In the third model, the dependent variable was the proportion of seeds in each replicate planting that produced seedlings (EMERGENCE), transformed with an arcsine of the square root to improve normality of residuals. We used Proc Reg (SAS for Windows, version 6.12, SAS, 1986).
Fitness components measured under field conditions were analyzed separately and as part of cumulative fitness for each of the 36 cross types grown at each common garden. For progeny planted at Fallbrook garden, we calculated cumulative fitness of each cross type as (mean filled seeds/flower) x (seedlings/filled seed) x (proportion surviving to harvest) x (mean fruits/plant). There was high dry season mortality at Motte garden and few plants survived to February 1998. For progeny planted at Motte garden, we calculated cumulative fitness as proportion surviving (herein survival) times mean height of survivors as of June 1997 (herein height).
To test whether fitness is a decreasing function of GENDIST and GardenEDIST, we ran a multiple regression analysis with the square root of cumulative fitness as the dependent variable and with the distance measures, GardenEDIST and GENDIST, as independent variables in a simultaneous multiple regression procedure (Proc Reg, SAS). We found no evidence of collinearity. The correlation of GardenEDIST and GENDIST was low in both the Fallbrook garden and Motte garden models (r = 0.15 and r = 0.27, respectively). We also ran alternative analyses with GardenGEODIST or GEODIST as the independent variable. Parallel analyses were run on field cumulative fitness (proportion surviving to June 1997 x square root of height of survivors; or proportion surviving to June 1998 x log total fruits on survivors). Variables were transformed to improve normality of residuals in regression analyses. Our test for ecological outbreeding depression predicts a decreasing relationship of fitness relative to genetic distance or ecological distance and is one-tailed. We conservatively report two-tailed tests throughout.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
|
The probability of seedlings emerging from the "hybrid" seeds also decreased significantly with increasing genetic distance of the crossed parental populations (Fig. 3B). GENDIST explained a low but significant 12% of the total variance in emergence (F[1, 35] = 5.58; P = 0.0240). The emergence of intervariety hybrids was 75.4% of that for intravariety hybrids (mean seedlings/seed = 0.259 for the 16 intervariety cross types; 0.349 for the 14 among-population intravariety cross types; and 0.331 for the six within-population intravariety cross types). Again, in an alternative ANOVA model that included a main effect of cross group, we found that seeds of intervariety crosses produced significantly fewer seedlings than seeds from intravariety crosses (analysis on % emergence averaged over replicate plantings of each cross type weighted by the number of 15-seed replicates; cross group F[2, 33] = 3.15, P = 0.0558; a priori contrasts: among-variety crosses vs. others, df = 1, F = 5.97, P = 0.0201; within-population vs. among-population within variety, df = 1, F = 0.21, P = 0.6462).
Field phase fitness
The effect of distance measures on cumulative fitness after outplanting was strongest at Fallbrook garden, possibly because mortality at Motte was very high. At Motte garden, by June 1997 there were 167 survivors out of the original 480 transplanted progeny (35%). By February 1998, after 1 yr in the garden, all but 18 had died and by June 1998 there were no survivors. In June 1997 there was no relationship between cumulative fitness (survival x height) and either distance measure (N = 36 cross types; GENDIST: ß = -2.739, t = -1.251, P = 0.2199; GardenEDIST: ß = -0.010, t = -0.135, P = 0.8935). In February 1998, there was also no difference between intra- vs. intervariety hybrids in the proportion surviving (3.8 vs. 3.6% survival; 
2 = 0.02, df = 1, ns). At Fallbrook garden, there were 208 survivors at harvest in July 1998, or 55% of the original 378 transplants. Cumulative fitness to June 1997 (survival x height) and to July 1998 (survival x fruits) decreased significantly with increasing GardenEDIST but not with GENDIST (1997: GENDIST: ß = -0.257, t = -0.127, P = 0.8999; GardenEDIST: ß = -0.275, t = -2.69, P = 0.0112; 1998: GENDIST: ß = -29.084, t = -0.84, P = 0.4071; GardenEDIST: ß = -7.785 t = -4.467, P < 0.0001). This nonsignificant effect of GENDIST is consistent with the nonsignificant difference between intra- vs. intervariety hybrids in the proportion surviving (57.7 vs. 50.4% survival; 2 = 0.57, df = 1, ns).
We also examined whether size early in the life cycle is useful in predicting future size and survival. At the Fallbrook garden, plants that were largest after 4 mo in the field had increased survival to the last census (height of plants in June 1997 relative to survival to June 1998 at Fallbrook garden: mean[survived] = 22.8 mm, mean[died] = 18.6 mm; F[1, 258] = 15.26; P < 0.0001). Furthermore, at Fallbrook height at all first-year surveys was significantly correlated with both height and fruit production of plants surviving to the final census, but correlations weakened somewhat over time (Table 2). In contrast, at Motte there was no difference in mean height of plants in June 1997 that survived compared to those that died by February 1998 (mean[survived] = 16.3 mm, mean[died] = 15.9 mm; F[1, 166] = 0.07; P = 0.7913). Moreover, height was significantly correlated between adjacent surveys, but not over larger periods (Table 3). Together, these data suggest that size traits early in the life cycle have some use in predicting future fitness but are of limited use when mortality is catastrophic.
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). Here, we show how the effect continues into the next generation. We used the same two common garden sites and six of the 12 source populations as in the previous study and found substantial outbreeding depression upon distant crossing.
Most of the variance in crossing distance as measured by allozymes was due to within- vs. among-variety crosses. In Montalvo and Ellstrand (2000b)
, 12 populations of the two varieties of L. scoparius separated distinctly by genetic distance. Considering all 12 study populations, the mean genetic distance among paired L. brevialatus x L. scoparius populations was 0.104, among paired L. scoparius populations it was 0.017, and among L. brevialatus populations it was 0.001. These genetic distances for populations of the same variety and for populations of different subspecific designation are not unusual for plants (Levin, 1978
; Gottlieb, 1981
). For the six populations in the present study, we cannot distinguish between the possibility of a threshold effect due to the discrete genetic distance between varieties, from a continuous effect due to increasing genetic distance. We do not know if there are other populations of L. scoparius of more intermediate genetic distance that could be used to distinguish these hypotheses. However, the regression approach is retained here because it can be generalized to studies of other taxa where distinct breaks in genetic distance do not occur. Furthermore, this model allowed inclusion of the continuously distributed independent variable, environmental distance, and comparison of the relative contribution of both independent variables to variance in cumulative fitness.
We observed significant levels of outbreeding depression in fitness components both prior to outplanting, as measured by seed production and seedling emergence, and in success of F1 progeny after outplanting. The severity and persistence of outbreeding depression into future generations may depend on the cause of genetic divergence, the genetic basis of depressions in fitness, and on the initial relative frequencies of nonlocal to local parents (cf. Emlen, 1991
; Fenster and Dudash, 1994
; Schierup and Christiansen, 1996
; Edmands, 1999
). Revealing the genetic basis of outbreeding depression was not the goal of our study, but our data are consistent with a genetic mechanism occurring before outplanting and an ecological mechanism occurring after outplanting. It is known that several mechanisms can reduce crossing success in the F1 generation. These include structural differences in parental chromosomes (Stebbins, 1958
) and nonadditive gene effects, including underdominance at multiple loci (Schierup and Christiansen, 1996
). In addition, a breakup of favorable additive x additive epistatic effects can also result in F1 outbreeding depression (Campbell and Waser, 1987
), but detection of epistatic gene action can only be achieved by line-cross analysis through the F2 generation (Lynch and Walsh, 1998
). Overdominance or epistatic interactions among multilocus gene complexes may result in F1 heterosis, followed by hybrid breakdown in subsequent generations once favorable gene interactions are broken up by recombination (Stebbins, 1958
; Waser, 1993
; Templeton, 1986
; Lynch, 1991
).
Our combined experimental results suggest that at least two proximal mechanisms are causing the observed cumulative outbreeding depression. First, differential seed set following cross pollination was caused more by differential fruit abortion than by any prefertilization incompatibility because differences in fruit set did not occur until after fruit initiation and initial expansion of presumably fertilized ovules. Genetic distance effects on seed maturation were apparent in the greenhouse where plants were removed from influences of the field environment. This suggests a lack of genetic congruence among genetically distant mates that could involve a variety of genetic mechanisms, including chromosomal divergence and intragene interactions such as underdominance, and breakup of favorable additive x additive epistasis (Stebbins, 1958
; Waser, 1993
; Schierup and Christiansen, 1996
; Lynch and Walsh, 1998
).
Outbreeding depression was highest when we crossed the most genetically distant groups of populations (the two varieties). Mean seed set of among variety crosses was 24.8% less than that of within variety crosses. Similarly, seedling emergence from among variety seeds was 24.6% less than that of within variety seeds. The cumulative fitness of among relative to within variety crosses prior to outplanting (seeds/flower x seedlings/seed) was 0.59, representing a substantial level of outbreeding depression.
Second, after outplanting and by the time plants reached reproductive maturity at Fallbrook garden, intervariety crosses were only 40% as fit as intravariety crosses. There was a consistent decrease in cumulative fitness with increasing genetic distance (GENDIST) at both Motte garden and Fallbrook garden. The depression was likely caused by genetic differences among parental populations given this association of GENDIST with fitness.
Third, fitness suffered a strong and significant decrease with increasing environmental distance of parents to the common garden site (GardenEDIST) at Fallbrook garden but not at Motte garden. The effect of this environmental distance also increased between the first and second year (Fig. 5D, F). Differentiation caused by adaptation to different environments can occur even over relatively small spatial scales if habitats are heterogeneous. Given the observed home site advantage in L. scoparius (Montalvo and Ellstrand, 2000b
), together with the strongly significant effect of GardenEDIST on cumulative fitness at Fallbrook garden, our data are consistent with adaptive differences among parents causing much of the observed outbreeding depression at Fallbrook garden where plants survived to fruiting. If additional nonadditive genetic mechanisms are involved, we would expect fertility to be lower in the more highly outcrossed progeny, especially if chromosome pairing is a problem, and that outbreeding depression would be even greater in the F2 or later generations.
The degree of outbreeding depression is likely dependent on the environment experienced by the progeny independent of mechanism (Lynch, 1991
), so the most realistic measures will be those done in the wild (Waser, 1993
). For example, outbreeding depression caused by either local adaptation of parents or breakup of certain gene interactions should be detectable in the F1 generation, but its occurrence may be dependent on specific environmental factors experienced by the progeny. We detected qualitative differences in the strength of outbreeding depression at the two common gardens during the first year of growth even though random progeny from the same crosses were planted in both places. The lack of effect of GardenEDIST on first-year fitness at Motte garden in contrast to results at Fallbrook suggest that the overall high mortality at Motte could have swamped any differential culling based on adaptation to specific conditions. High mortality may have influenced failure to detect significant population effects in some plots in other experiments (Kindell, Winn, and Miller, 1996
; Montalvo and Ellstrand, 2000b
).
Other studies have detected differences in F1 outbreeding depression in different years (Waser, Price, and Shaw, 2000)
or in different environments (Wang et al., 1997
) that are consistent with ecological mechanisms. For example, two subspecies of the common shrub Artemisia tridentata typically grow in different environments. In a reciprocal transplant experiment, Wang et al. (1997)
grew progeny and parental individuals of two crossed subspecies together in common gardens at three locations, including the reciprocal parental habitats and a third habitat that was home only to natural hybrids. The parents grew best in their home environment, and the hybrids performed poorly in both parental environments. However, the hybrids performed better than the parental genotypes in the third habitat, a phenomenon often associated with hybrid zones.
Because the geographic distance between sites (GEODIST) was not highly correlated with GENDIST or EDIST, it could not serve as a good predictor of outbreeding depression in this study. EDIST and GEODIST were not significantly correlated because sites represented a mosaic of environments rather than clines (N = 6, Mantel's test r = 0.08, P = 0.60; r even weaker for 12 populations; Montalvo and Ellstrand, 2000b
). Some L. s. var. scoparius populations were closer to L. s. var. brevialatus populations than to other L. s. var. scoparius populations (Fig. 1). Even though the mean environmental and geographic distances of parents relative to the Fallbrook garden site were moderately correlated (Table 1), GardenGEODIST was not significantly associated with cumulative fitness. Had there been a geographic cline in allozyme variation resulting in a strong correlation between EDIST and GENDIST, or if GENDIST had been tightly correlated with GEODIST, then the predictive power of geographic distance may have increased.
We consider our measures of outbreeding depression conservative for the following reasons. First, crossing success and seedling emergence were tested under relatively benign greenhouse conditions. We might have seen some effect of environmental distance (GardenEDIST) if the seed germination phase had been in the field. Habitat correlated germination response has not been examined for L. scoparius but differences that can affect transplantation success have been detected in other plants (Meyer, McArthur, and Jorgensen, 1989
; Meyer, Monsen, and McArthur, 1990
; Meyer, 1992
; Meyer and Monsen, 1991, 1992
; Meyer and Kitchen, 1994
).
Second, examining fitness components over the lifetime of plants or examining seed set under more natural or stressful conditions might result in increased outbreeding depression. When overall mortality is not so extreme as to swamp genetic effects, experiments under natural field conditions frequently result in larger fitness differences among crosses than under greenhouse conditions and show an accumulation of fitness differences over time (e.g., Dudash, 1990
; Montalvo, 1994
). Furthermore, these plants potentially live to reproduce for 5–10 yr, but we harvested them in just over 1.5 yr. As in our study of home site advantage (Montalvo and Ellstrand, 2000b
), cumulative fitness in the first year at Fallbrook garden was useful in predicting relative performance through reproductive maturity, but differential success of crosses may increase over time in iteroparous species. We expect fitness differences among populations to accumulate and become even stronger for plants exposed to the transplant location over their lifetimes as seen in long-term provenance tests (Conkle, 1973
).
Third, it is also possible that more than one causal mechanism of outbreeding depression is operating on survival and reproduction. We have not yet measured the effects of backcrossing and hybridization among F2 and future generations. It is possible that ecological outbreeding depression swamped detection of heterosis in the F1 generation. Under epistatic or dominance models, the negative consequences of distant matings would be revealed in later generations and may cause even more severe outbreeding depression than already observed (cf. Fenster and Dudash, 1994
; Edmands, 1999
).
Implications
Our results suggest that fitness of L. scoparius populations can be adversely affected when genetically or environmentally distant populations are used for seeding and restoration, even when all source populations are from similar plant communities. The effect of genetic distance on seed set and emergence was to markedly reduce fitness when varieties were crossed and the effect of increased environmental differences between parental and transplant habitats was a reduction in survival and reproduction. We have observed several restoration projects that mismatched L. scoparius varieties. Given the large genetic and environmental distances between populations of the two different varieties, we hypothesize that such mismatches or mixing seeds from diverse populations would result in substantial outbreeding depression.
We discourage use of genetically distant populations in restoration even though diverse populations may come into secondary contact unaided by humans. Naturally occurring putative hybrids between the two varieties of L. scoparius have been observed in the natural transition between Diegan and Riversidiian coastal sage scrub vegetation (Steppan, 1991
; A. Montalvo, unpublished data). Studies of two subspecies of Artemisia and of two species of stickleback fish indicate the importance of the natural environment in maintaining hybrid zones (Wang et al., 1997
; Hatfield and Schluter, 1999
). In our study species the two varieties may be maintained, in part, by their differing ecological specialization which would result in ecological outbreeding depression. In addition, poor seed set and seedling emergence point to additional unspecified genetic mechanisms. In zones of natural secondary contact, hybrids may benefit from the presence of "hybrid" habitat and persist (Anderson, 1948
). For this reason and because environmental distances of parental populations are likely more extreme in mismatched restorations than in natural transition zones, we predict that hybrid fitness in manmade zones of secondary contact will be lower than in areas of natural secondary contact.
Local adaptation and outbreeding depression support the rationale for seed transfer guidelines. The U.S. Department of Agriculture adopted a tree seed transfer policy in the 1930s (McCall, 1939
; Kitzmiller, 1990
) recommending that seeds used for reforestation be collected only within specified "seed collection zones." More elaborate ecological and genetic criteria have been proposed for management of tree populations (Millar and Libby, 1991
). Seed transfer policies have been expanded to include other life forms and offer guidelines designed to increase the initial success of restored populations and also to preserve their "evolutionary potential" (e.g., Anonymous, 1992
; USDA Forest Service, Pacific Southwest Region, unpublished memo, 1994; Knapp and Rice, 1994
; Montalvo, 1995
; Montalvo et al., 1997
).
Seed transfer guidelines recognize that too much inbreeding can be as problematic as too much outbreeding. Augmentation with new genetic material may sometimes afford the best chance of increasing the long-term viability of a population that has suffered a severe drop in gene flow (and subsequent erosion of genetic diversity) due to fragmentation (Ellstrand and Elam, 1993
). Increased population fitness has been obtained by introducing new genes into populations of animals that have become small and fragmented (Westemeier et al., 1998
). Small plant populations that have a deficit of incompatibility alleles may especially benefit from augmentation (Les, Reinartz, and Esselman, 1991
; DeMauro, 1993
). However, augmentation can also end in failure if outbreeding depression results from genetic dissimilarities between local and introduced populations (Templeton, 1986
; Rhymer and Simberloff, 1996
).
Our data suggest when selecting source populations, at a minimum, one should consider carefully the geographic location together with ecological zones encompassed by a vegetation type, and keep subspecific taxa separate whenever they actually occur in the same ecological zone. However, geographic distance alone did not prove to be a useful tool for predicting home site advantage or outbreeding depression. Preferably, seed collection zones based on ecologically similar areas in combination with careful consideration of taxonomic and floristic literature can be used as a meaningful guide in the absence of detailed population genetic or environmental data. Other researchers have come to similar conclusions (e.g., Knapp and Rice, 1994, 1998
).
Choosing relatively local seed collections from appropriate habitats can guard against using incorrect taxa or genetically differentiated populations. Seed companies should be encouraged to keep careful records of seed source location. Often, the only way to obtain appropriate material is to carefully match habitat within specific geographic zones. In California and Utah, a number of native seed companies are having locations of native seed collections certified through special new programs (Chip Sundstrom, California Crop Improvement Association, personal communication; Young et al., 1995
).
|
FOOTNOTES |
|---|
4 Author for correspondence (montalvo{at}citrus.ucr.edu
), current address: Department of Botany and Plant Sciences, University of California, Riverside, CA 92521-0124 USA.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Anonymous. 1992 Guidelines for Management of Genetic Resources in the California State Park System. Prepared by Sycamore Associates for the California Department of Parks and Recreation, Lafayette, California
Bradshaw, A. D. 1965 Evolutionary significance of phenotypic plasticity in plants. Advances in Genetics 13: 115–155
———. 1984 Ecological significance of genetic variation between populations. In R. Dirzo and J. Sarukhan [eds.], Perspectives on plant population ecology, 213–228. Sinauer, Sunderland, Massachusetts, USA
Bubar, J. S. 1958 An association between variability in ovule development within ovaries and self-incompatibility in Lotus (Leguminosae). Canadian Journal of Botany 36: 65–72
Byers, D. L. 1998 Effect of cross proximity on progeny fitness in a rare and a common species of Eupatorium (Asteraceae). American Journal of Botany 85: 644–653[Abstract]
Campbell, D., and N. M. Waser. 1987 The evolution of plant mating systems: multilocus simulations of pollen dispersal. American Naturalist 129: 593–609[CrossRef][ISI]
Clausen, J., and W. M. Hiesey. 1958 Experimental studies on the nature of species, IV. Genetic structure of ecological races. Carnegie Institute of Washington Publication 615, Washington, D.C., USA
Clausen, J., D. D. Keck, and W. M. Hiesey. 1940 Experimental studies on the nature of species. I. Effect of varied environments on we
