Influence of phylogeny and ploidy on species ranges of North American angiosperms

Authors


*Correspondence author. E-mail: slwmartin@gmail.com

Summary

1. Species differ widely in the size and ecological characteristics of their geographic ranges. This variation reflects, in part, differences in ecological tolerance among taxa, but the influence of phylogeny and evolutionary divergence through genome duplication remains controversial.

2. Here, we examine the evolutionary sources of variation in species range area, latitudinal position and overlap, and six range climatic variables (maximum, minimum and breadth of temperature and precipitation) for North American angiosperms. We sampled two diploid and one polyploid species from each of 144 genera and examined sources of variation in a nested taxonomic analysis and congeneric species pairs comparison.

3. For all range characteristics studied, a significant portion of the variance among species (mean 53%, range 37–60%) was attributable to taxonomic membership above species (i.e. genus, family and order). Four of the seven range attributes were significantly correlated between diploid congeners (e.g. range area rs = 0.65) and between heteroploid (diploid and polyploid) congeners (e.g. range area rs = 0.62). The strength of correlations between heteroploid relatives was not consistently different than correlations between diploids.

4. No differences in mean range breadth (range area, temperature breadth, precipitation breadth) or climatological position (minimum and maximum temperature and precipitation) were observed between diploid and polyploid congeners. Polyploids were not more likely to occupy northern latitudinal positions and were no more displaced from diploids than other congeneric diploids.

5.Synthesis. Our results suggest that, at the scale of North America, the geographical and ecological aspects of species ranges are influenced by phylogenetic history. Furthermore, genome duplication does not cause or allow greater divergence in species range characteristics than occurs between related diploid species.

Introduction

Understanding the determinants of the species range, the area within which members of a given species occur for at least part of their lives, is a fundamental goal of modern ecology and the basis for management of rare and invasive species (Mayr 1963; Hoffman & Blows 1994; Kirkpatrick & Barton 1997; Holt & Keitt 2005). Interspecific variation in range size is large, spanning 12 orders of magnitude (Brown, Stevens & Kaufman 1996). The factors contributing to this variation are complex and include historical influences such as species age and biogeographic limits to expansion, as well as contemporary interactions between the environment and species’ ecological tolerances. Ecological tolerances, which govern the position and breadth of occurrence of species along environmental gradients, are the products of the physiological, morphological and developmental attributes of its members (Good 1974; Holt 2003) and can be heritable. Therefore, by extension, range position and breadth are potentially shaped by evolution and the limits to adaptation (Antonovics 1976; Kirkpatrick & Barton 1997).

To the extent that ecological tolerance has a genetic basis, species range size and position will be the product of two contrasting evolutionary forces: phylogenetic history, which results in resemblance between related species, and selection or drift in contrasting environments, which causes divergence. There is little consensus about whether phylogeny influences a species’ range. Some authors have suggested that the geographic range shows a phylogenetic signal because of the inheritance of ecological tolerances (Peat & Fitter 1994; Kelly & Woodward 1996; Waldron 2007), whereas others maintain that small differences in biology will lead to large differences in range attributes, and therefore, the effect of phylogenetic history will be negligible (Brown, Stevens & Kaufman 1996; Webb & Gaston 2003). Research on birds indicates that most variation in range area occurs at the species level (Gaston 1998), suggesting little phylogenetic signal. However, a study of fossil mollusks found significant correlations in range size between sister species (Jablonski 1987).

In plants, sources of variation in range area have been examined in only two studies, both of which focused on native or naturalized floras of islands, Crete and Great Britain (Peat & Fitter 1994; Kelly & Woodward 1996). As with birds, most of the variation occurred among species. However, the small geographic scale of these studies may have artificially constrained range size variation and limited the ability to detect a phylogenetic signal (Jones, Sechrest & Gittleman 2005). Furthermore, the ecological attributes of species ranges (e.g. breadth and position with respect to climate) have rarely been studied in this context, although they may also be conserved or constrained by the inheritance of ecological tolerances. To our knowledge, only one study has examined variation in ecological dimensions of plant ranges (Prinzing et al. 2001). In their study, Prinzing et al. (2001) found that for half of the climatic variables (temperature, light and soil conditions) they measured, more than 50% of the variance was explained above the species level. Thus, the results regarding phylogenetic influences on species range size and position are heterogeneous and conflicting. In plants, this can be addressed by studying both geographical and ecological attributes of species ranges on a larger spatial scale.

Among extant species, polyploidy is considered an important genetic cause of changes in the geographical and ecological range of species (Levin 2002). Genetic divergence among species of the same ploidy (homoploid divergence) via mutation, recombination, gene duplication and rearrangement, will influence ecological tolerance, but these changes are generally considered to have smaller effects than the transgressive influences of genome multiplication (Levin 2002). Two types of polyploids are recognized: autopolyploids, which form by duplicating the genome of a single species, and allopolyploids, formed by duplicating the genome of interspecific hybrids. The difference between the effects of these two types of polyploidy are not well explored, however, both types of polyploidization can potentially affect species ranges in two major ways (Muntzing 1936; Stebbins 1950). First, polyploidy can cause an immediate shift in phenotype and physiological tolerances directly as a result of a larger genome and cell size. Secondly, polyploidy may affect the evolution of a species’ range by increasing the capacity for genetic variation, the probability of beneficial mutations and the potential for novel adaptive responses to selection (Stebbins 1985; Levin 2000; Otto & Whitton 2000; Soltis & Soltis 2002). Indeed, the potential effects of polyploidization have led biologists to expect that both allo- and autopolyploidy will cause changes to the position of a species’ range along an ecological gradient, specifically towards more extreme environments and more northern latitudes, and to the breadth of environments occupied compared to diploids (Stebbins 1950; reviewed by Levin 2000). The distinction between these two effects is not well examined and the empirical evidence for greater breadth is conflicting. Early work by Muntzing (1936) and Stebbins (1950) corroborated these expectations, and more recently, a study of Clarkia found that polyploids had significantly larger ranges than diploids within the genus (Lowry & Lester 2006). In contrast, broader taxonomic surveys have found no evidence for this pattern (Stebbins & Dawe 1987; Petit & Thompson 1999). These broader studies, however, have not been conducted within a phylogenetic framework. This could obscure the effect of genome duplication on range attributes if the phylogenetic history of a species also has a significant effect on range size (Levin 2000).

In this study, we examined the evolutionary basis of species ranges in flowering plant species of North America. We estimated both geographical (area, overlap) and ecological attributes (precipitation and temperature) of species ranges. These attributes were used to characterize range breadth (geographic area, range of temperatures and precipitation) and position (latitude, mean precipitation and temperature) along ecological gradients. To account for phylogenetic history, we collected data for two diploid and one polyploid species from each of 144 genera. This sampling design allowed us to compare the effects of genome duplication (i.e. divergence between species with different ploidy – heteroploid species) to differences between homoploid species (species with the same ploidy). Because diploids are considered the ancestral state (Stebbins 1971), we used the differences between two related diploids as the evolutionary baseline against which the effects of polyploidy were evaluated. Specifically, we explored the following questions: (i) What proportion of the variation in each of the geographic and climatic range attributes is explained by taxonomic membership above the species level (i.e. order, family and genus)? (ii) Are range attributes correlated among relatives, and are heteroploid relatives less correlated than diploid relatives? (iii) Are polyploids different from their diploid relatives with respect to range size, overlap, climate position and climate breadth, and are these differences greater than those between diploid relatives?

Materials and methods

Data collection

We collected data on range attributes for two diploid and one polyploid species from each of 144 genera (representing 46 families and 23 orders) in the North American flora (see Table S1). Since each set of three congeners share a recent common ancestor, evolutionary divergence in range attributes within genera is treated as independent of divergence among other genera.

Information on range boundaries and ploidy for each species was primarily gathered from five sources: The Flora of North America Editorial Committee (1993, onwards), Vols 3–5, 19–23, 26); the Intermountain Flora (Cronquist 1972); Flora of Alberta (Moss 1983); Manual of Vascular Plants of Northeastern United States and Adjacent Canada (Gleason & Cronquist 1991); and Utah State University's online grass manual (2004). Information for Chamerion was taken from Mosquin (1967). Although the sources differed, information on geographic range and chromosome number was always acquired from the same source for all three species in a genus. The Flora of North America Online (1993, onwards), Utah’s online grass manual (2004) and Mosquin (1967) provided geographic range maps for 97 genera and we constructed range maps from verbal descriptions for the remaining 47 genera. In cases where the information from floras conflicted, information from the Flora of North America was used.

Species were picked randomly with respect to ploidy and range characteristics. Generally, diploid species were identified as those with twice the base chromosome number for the genus (x) as listed by the Flora of North America. For example, if the base number is = 5, the first two species listed in the flora with 2n = 2x = 10 chromosomes were designated as diploids, and the first species with 2n = 4x = 20 was designated as polyploid. Species that were identified as introduced to North America were excluded. There was insufficient information on the mode of origin of the polyploidy (autopolyploid versus allopolyploid) or of the polyploidy age (neopolyploid versus paleopolyploid) to treat these categories separately in the analysis.

It is difficult to determine the absolute ploidy of any given species, especially given the various genetic mechanisms by which plants can become ‘diploidized’ after polyploidization (Otto & Whitton 2000). However, our interest is in the relative effect of polyploidization and not in absolute states of ploidy (i.e. diploid, tetraploid, hexaploid etc.). Therefore, ploidy designations in this study reflect the relative differences in chromosome complement, with ‘polyploids’ in a given genus always having at least twice the number of chromosomes as the ‘diploids’. For 17 genera (Agropyron, Arenaria, Callisia, Cirsium, Cuscuta, Digitaria, Draba, Eriophyllum, Gaillardia, Helenium, Krigia, Muhlenbergia, Parthenium, Saccharum, Prunus, Stellaria and Saxifraga), it was not possible to find two species with twice the base number so we used the lowest ploidy shared by at least two species to represent the diploid state. For 17 genera (Arenaria, Balsamorhiza, Betula, Calamagrostis, Cardamine, Carduus, Conyza, Corydalis, Echinochola, Fimbristylis, Krigia, Liatris, Lorandersonia, Rubus, Saxifraga, Trifolium and Zigadenus) polyploids had more than twice the number of chromosome sets than the chosen diploids in the same genus. In nine genera (Agave, Boltonia, Chamerion, Delphinium, Hymenopappus, Isocoma, Lygodesmia, Pityopsis and Rudbeckia), we included diploid and polyploid subspecies. Autopolyploids are frequently described as subspecific taxa, but often meet the criteria for species designation (Soltis et al. 2006).

All range maps were redrawn onto a base map in the geographical projection using ArcMap 9.1 (Environmental Systems Research Institute, Redlands, CA, USA). These ranges served as the basis for calculating range area, range overlap and range centroid, and for collecting climatic data. Range area, expressed as square kilometres on the surface of North America, was determined by reprojecting the maps into the equal-area cylindrical projection with arcview 3.3 (Environmental Systems Research Institute, Redlands, CA, USA) and then calculating area using a CalcArea script (Jan Mersey, Personal communication). The overlap between the ranges of the two diploid species and the diploid–polyploid pairs within each genus was determined by extracting the intersection of the ranges and calculating the area of the extracted polygon. The location of the ranges’ centroid was based on the centre of mass as determined by ArcMap.

To characterize the climate for a given geographic range, we imported data on temperature and precipitation from the New, Hulme & Jones (1999) global climate dataset as rasters (numerical data in a grid) into ArcMap. These variables were chosen because of their importance for growth, reproduction and survival in plants (Stiling 1999). This data has a resolution of 0.5° latitude and 0.5° longitude and contains the average values for climatic variables for each month as measured from 1961 to 1990. We used the masking tools in arcmap to extract the values for each climatic variable within each species’ range. Temperature limits for each species range were measured as the mean monthly maximum July temperature and mean monthly minimum January temperature. Temperature breadth was calculated as the difference between the July maximum and January minimum temperature. For precipitation, we calculated the average monthly values for each point on the grid and recorded both the mean maximum and mean minimum values for each species. To calculate the average monthly values for precipitation, we used data from all 12 months and raster math tools in arcmap to calculate an average. Precipitation breadth was calculated by subtracting the minimum amount of precipitation received within the range from the maximum received. With range area and these climatic measures (maximum, minimum and breadth for temperature and precipitation), there were a total of seven range variables. Range area, temperature breadth and precipitation breadth were used as measures of the breadth of the species range, whereas maxima and minima provided a measure of range position along the temperature and precipitation gradients.

Statistical analysis

Effects of phylogeny

The influence of phylogeny on species range variation was examined by testing for similarities in geographical and ecological attributes of the range among relatives. To test this, we partitioned the variation in range area, July mean maximum temperature, January mean minimum temperature, temperature breath, precipitation monthly mean maximum, monthly mean minimum precipitation and precipitation breadth among orders, families within orders and genera within families (all random effects) using a nested analysis of variance (Peat & Fitter 1994; Kelly & Woodward 1996). These anovas were performed with jmp in 5.1 (SAS Institute, Cary, NC, USA) using parametric tests of significance. Variance components were estimated and those that were negative were truncated to zero for reporting. Because the data did not conform to assumptions of normality, we also assessed significance by comparing the variance components to a null expectation generated by randomly assigning species to the 144 genera 1000 times and repeating the nested anova (see Prinzing et al. 2001). This null model represents the expected amount of variation in range attributes explained just by the taxonomic structure of the data set alone (Prinzing et al. 2001). The upper 99% confidence limit of the variance explained by taxonomic groups above species (i.e. order, family and genus) in the null model was always lower than 2% of the total variance (mean 0.82%, range −0.04% to 1.66%). As the significance levels computed parametrically were congruent with the null-model analysis but more conservative, only parametric significance levels are reported.

The resolution of the geographic range maps varied widely among literature sources, from boundaries based on detailed herbarium records to crude distributions based on political boundaries. To evaluate whether the quality of the range maps was having a large effect on our results, we repeated the anova using only data from genera for which the most detailed maps were available (n = 42).

The influence of phylogeny on species range was also examined by testing for a correlation between congeners, for each range attribute, across all genera. If family or order accounted for a significant amount of the variance in the nested anova, then the correlation was conducted between randomly chosen taxa at this level, rather than among congeners, to account for non-independence of the data. For example, if order was significant in the anova, two diploids were chosen at random from all diploid species within each order for the homoploid correlation. Correlations were estimated using the nonparametric Spearman’s correlation coefficient (rs), because the data were not normally distributed, and significance was assessed using bootstrapped estimates with a correction for ties (Higgins 2004). The significance of the multiple correlations was corrected to α = 0.05 using a sequential Bonferroni test (Rice 1989). As with the anova, this correlation analysis was repeated using a subset of the data with the most detailed range maps (these estimates were calculated in JMP and not bootstrapped).

Effects of polyploidy

To compare the geographic position of diploid and polyploid species’ ranges, we scored the relative position of the range centroid (centre of mass) for each polyploid as north, south or between the centroids of its two diploid congeners. We tallied the total number of genera in each category and compared these counts to the random expectation of an equal number appearing in each category (north, south and intermediate) with a chi-squared test (contingency analysis). Additionally, we compared the mean area of overlap between the ranges of the two diploid congeners with the mean overlap between diploid and polyploid pairs, across all genera, using a Wilcoxon-signed rank test. Based on previous arguments (Muntzing 1936; reviewed in Levin 2000), we expected polyploids to be positioned further north and to have less overlap with their diploid progenitors than that between two diploid species.

The effect of polyploidy on other range attributes was assessed using paired species comparisons. To compare the relative difference (sign) between diploid and polyploid species pairs with respect to range breadth (area, breadth of temperature and precipitation) and position (minimum January temperature, maximum January temperature, minimum and maximum precipitation) we used the nonparametric Wilcoxon-signed rank test (jmp in 5.1) because of non-normality in the data (SAS Institute, Cary, NC, USA). For each range variable and genus, we compared the polyploid value to the average diploid value. To assess the robustness of the Wilcoxon test, we also compared the magnitude of the difference between diploid and polyploid species pairs using a paired t-test (two-tailed test; n = 144 congener pairs). The mean difference among heteroploid pairs was compared to the mean difference among diploid species pairs using 95% confidence limits.

Finally, we tested whether divergence in range area between species pairs, across all genera, was correlated with changes in range position or breadth (i.e. are differences in maximum, minimum and breadth for temperature or precipitation related to differences in range area) using independent contrasts. To do this we calculated the correlation between the difference in range area between congeners and the difference for each climatic characteristic using Spearman’s rank correlation. This was repeated for homoploid and heteroploid species pairs separately.

Results

Geographic range area for the 432 species in this study had a median of 1 128 210 km2 and ranged from 3483 to 13 920 761 km2 among species (Table 1). For the species sampled, range area was smallest for the order Malvales and largest for the Brassicales (Table S1), both eudicots. The monocot families Liliaceae and Agavaceae had the lowest family means (17 065 and 10 536 km2, respectively), whereas Brassicaceae and Saxifragaceae had the highest means (8 738 677 and 8 752 198 km2, respectively). The distribution of range area was skewed to the right, indicating a predominance of species with relatively small ranges. The median range overlap between diploid and polyploid pairs (407 350 km2) was significantly larger than between diploid congeners (296 490 km2) (Wilcoxon-signed rank = 1285.5; two-tailed test, P = 0.002). We also found no pattern in the north-south position of ranges when polyploids were compared to diploids in the same genus (observed: north – 44, intermediate – 51, south – 39; expected values: 44.6; χ2 = 2.44, P > 0.05). There was also wide variation in all climatic variables examined (Table 1).

Table 1.   The median value and range for species range characteristics for all 432 species of North American angiosperms included in this study
VariableMedianRange of values
  1. The median is presented here as representative of central tendency because of the skew in the distribution of the data.

Range area (km²)1 128 2103438–13 920 761
Maximum July temperature (°C)37.6  8.4–42.5
Minimum January temperature (°C)−18.2−45.3–8.0
Temperature breadth (°C)55 11.3–85.2
Maximum monthly precipitation (mm)477–187
Minimum monthly precipitation (mm)50–35
Precipitation breadth (mm)362–102

Effects of Phylogeny

Based on the nested anova, there was significant variation above the species level for all range variables. The percentage of variation above species averaged 53% across all range variables and spanned from 37% to 60% (Table 2). For range area, 59% of the total variance occurred above the species level. Variation above species for maximum, minimum and breadth of temperature was 37%, 60% and 52%, respectively. Similarly, values were 60%, 45% and 59% for maximum, minimum and breadth of precipitation, respectively. Variation among genera was significant for all seven range variables (average 20%, range 8–29%). Variation among families (average 12%, range 0–24%) was significant for all variables except maximum July temperature. Variation among orders (average 21%, range 9–27%) was significant for all variables except minimum monthly precipitation. Using the null model established by randomly assigning species to genera, the 99% confidence intervals for variance above the species level was <2% for all range variables. Using this as a baseline for comparison, variance components at all taxonomic levels and for all seven traits were significant except for July maximum temperature.

Table 2.   Variance components (%) for seven species range attributes of North American angiosperms
Range attributeOrderFamilyGenusSpecies
  1. Values are distributed among four taxonomic levels (order, family, genus and species). The significance of the variance component for species is not given, as this term is represented by the residual in a nested anova with order, family (order) and genus (family). ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001.

Range area (km2)27*24***8**41
July mean maximum temperature (°C)16*0ns21***63
January mean minimum temperature (°C)22*20***18***40
Temperature breadth (°C)24*16***12***48
Precipitation mean monthly maximum (mm)25**6**29***40
Precipitation mean monthly minimum (mm)9ns10**26***55
Precipitation breadth (mm)27**7*25***41

The analysis of the 42 genera for which more detailed maps were available showed very similar results to the full analysis (results not all shown). For example, for range area, 24% of the variation was explained by order, 29% by family and 12% by genus. In comparison, in the full analysis for range area, order explained 29%, family explained 23% and genus explained 8% of the variation. In total, 17 of the 21 variance components (three taxonomic levels by seven variables) were significant in the reduced data set, compared to 19 of 21 in the full analysis.

Correlations between diploid species pairs and diploid–polyploid pairs were significant with respect to two of the seven range variables: range area and minimum January temperature (Fig. 1, Table 3). Temperature breadth was significant for the diploid comparison only while minimum monthly precipitation was significant for only the diploid–polyploid comparison. Overall, two of three variables related to range breadth showed significant correlations, compared to two of four variables related to range position. There was no consistency as to which sets of congeners were most strongly correlated. Diploid–polyploid congeners were more strongly correlated than diploid–diploid congeners in four of eight variables.

Figure 1.

 Correlation between relatives in log range size (km2, panels a, b) and minimum January temperature (°C, panels c, d). Species pairs were drawn from the 23 different taxonomic orders as most of the variance in these traits existed among orders. Panels a and c represent the correlation for two randomly drawn diploid species. Panels b and d show the heteroploid correlation. The mean and 99% confidence limits for Spearman’s rank correlations (rs) are shown for each data set. *P < 0.05, **P ≤0.01 and ***P < 0.001.

Table 3.   Spearman rank correlations (rs) (99% confidence limits) between homoploid and heteroploid congeners for seven species range attributes
 Taxonomic levelDiploid–DiploidDiploid–Polyploidn
Mean and 99% CLMean and 99% CL
  1. Species were drawn at random at the taxonomic level listed, which is the highest taxonomic level in the nested anova that explained a significant portion of the variation. Each correlation value is the mean after re-sampling the data 10 000 times. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001.

Range areaOrder0.65 (0.643–0.649)*0.62 (0.618–0.624)*23
Maximum July temperatureOrder0.28 (0.271–0.279) n.s. 0.35 (0.341–0.350) n.s. 23
Minimum January temperatureOrder0.59 (0.585–0.591)**0.62 (0.618–0.623)*23
Temperature breadthOrder0.64 (0.636–0.642)*0.56 (0.562–0.568) n.s.23
Maximum precipitationOrder0.35 (0.342–0.349) n.s.0.43 (0.428–0.435) n.s. 23
Minimum precipitationFamily0.41 (0.408–0.412) n.s.0.49 (0.491–0.496)*45
Precipitation breadthOrder0.53 (0.521–0.528) n.s.0.50 (0.499–0.505) n.s.23

Analysis of the 42 genera with more detailed maps produced similar results to that of the full analysis. For example, the correlation between diploid congeners for range area was = 0.83 compared with = 0.65 in the full analysis; the correlation between heteroploids was = 0.66, compared with = 0.62 in the full analysis. In total, using only genera with more detailed maps yielded significant correlations between diploid congeners for five of seven range variables and four of seven variables for heteroploid pairs. In the full analysis, there were significant correlations between diploid species and between heteroploid species in four of seven range variables.

Effect of polyploidy

Based on Wilcoxon-signed rank tests, corrected for multiple comparisons, none of the range variables that were related to either position or breadth of range were significantly different between diploid congeners (in all cases, P > 0.05). Diploid and polyploid congeners showed no significant differences in range area (Wilcoxon-signed rank = −126, P = 0.80), maximum monthly July temperature (sign rank = 311, P = 0.466), minimum monthly January temperature (sign rank = 231, P = 0.63), temperature breadth (sign rank = −269, P = 0.58), maximum monthly precipitation (sign rank = −343, P = 0.43) or temperature breadth (sign rank = −55, P = 0.91). However, ranges of polyploid species had significantly lower mean minimum precipitation than those of diploids (signed rank = 1044, P = 0.01 versus diploid–diploid comparison signed rank = −392, P = 0.32). While the minimum precipitation level was statistically different between diploid and polyploid species pairs, the 95% confidence limit of the difference (−1.6 mm 95% confidence interval −3.1 to −0.1) overlapped with the difference observed between diploid pairs (1.5 mm 95% confidence interval −0.2 to 3.2). In addition, there were no differences between diploids and polyploids using a parametric paired t-test for six of the seven range variables (P > 0.35; e.g. range area; t = 0.52, P > 0.50). Similar to the nonparametric test, the exception was minimum precipitation, where polyploid ranges had significantly lower values (t = −2.13, P = 0.035). Diploid species pairs were not different for any of the variables (P > 0.05).

Differences in range area between species pairs are not correlated with differences in any climatic variables (results not shown). Correlations were close to zero for both diploid and heteroploid species pair comparisons. For example, the correlations between divergence in range area and divergence in maximum July temperature were rs = 0.01 (P > 0.05) and 0.04 (P > 0.05) for diploid and heteroploid pairs, respectively. Correlations between range area and maximum precipitation were rs = 0.06 (P > 0.05) and 0.00 (P > 0.05) for diploid and heteroploid pairs, respectively.

Discussion

Ecologists have long recognized that plant distributions may have a genetic basis (Good 1974), but few studies have actually tested for and quantified the evolutionary sources of variation in range size and ecology. Moreover, studies do not agree on the importance of phylogenetic history, and the effects of genome duplication have not been tested within a phylogenetic context. In this study, we quantified the sources of variation in species ranges using diploid and polyploid congeners from 144 genera of flowering plants. We found that for most range variables, >50% of the variance was attributable to differences among genera, families and orders, and that for many species range attributes were correlated among relatives. Both results indicated that the geographical and climatic breadth and position of species ranges show a phylogenetic signal. While, we might expect reduced correlation because of divergence through polyploidization, the correlations between relatives of different ploidy were as strong as those between diploid relatives. Finally, at this scale of investigation, polyploids were not consistently significantly different from related diploids with respect to species range area or ecological attributes. These patterns contrast markedly with conventional views about the consequence of genome duplication for ecological tolerance and geographical distributions.

The variable resolution of species range information likely affected the accuracy of our range estimates for any given species. For 102 of 144 genera, the boundaries of the species range were defined at the scale of states and provinces, rather than on actual observations of occurrence. While this information is reasonable for an analysis of distributions on a continental spatial scale, using political boundaries to delineate ranges is not ideal and may increase variation in the estimates of range variables and hence reduce our power to detect differences among taxonomic groups or between ploidies. In this sense, our results should be considered conservative. However, this limitation in species range information likely did not obscure our conclusions for several reasons. First, the fact that we observed a significant phylogenetic signal suggests that this limitation did not significantly restrict the power of our analysis. Second, we ensured that the same quality of range information was used for all species within a given genus; thus, this source of error should be uniform across taxa and should not lead to bias within the species paired comparisons. Finally, we evaluated the robustness of our conclusions with an additional analysis based on the 42 genera with the most detailed range maps. In all cases, these results showed the same general patterns as the complete data set.

The role of phylogeny

For all range variables, there was a strong influence of taxonomy above the species level (i.e. genus, family and order). The phylogenetic component of geographic and climatic variation may reflect the vestiges of common descent and possibly the fact that relatives retain similarities because they occur and experience selection in similar habitats (niche conservatism, Losos 2008). Identifying a causal mechanism is beyond this study, but with common descent alone, one would expect the phylogenetic effect to weaken with more distant relations. In contrast, taxonomic membership at the level of order explained as much of the variation membership at the level of genus, suggesting that evolutionary forces are maintaining geographic and climatic similarities among distantly related taxa.

Regardless of mechanism, it appears that emergent ecological properties of species have a heritable component (Harvey & Pagel 1991). This seems reasonable given that a species’ range is the summation of its constituents’ ecological tolerances, and that ecological tolerances are influenced by physiological, morphological and life-history traits, which are often found to be heritable in field studies (Geber & Griffen 2003). Furthermore, the importance of order and family in explaining range variation suggests that heritable traits that differentiate and even define particular clades, such as woody versus herbaceous life histories or C4 versus C3 metabolism may play an important role in dictating the limits to ecological differentiation. These may also be traits that are least variable within species and therefore most limiting to range expansion or adaptation at the range edge. Because our sample sizes for each family and order are relatively small, we refrain from seeking generalities about the traits associated with particular taxonomic groups in this study. Nevertheless, the implications of a phylogenetic signal are that taxa (families, orders) will differ in their tendency to be widespread or restricted in distribution, and that species can be good predictors of the ecological attributes of their relatives. This predictive ability may aid in the identification and mitigation of risks associated with invasive or rare species (Reichard & Hamilton 1997). For example, native ranges of plants (Broennimann & Guisan 2008) and animals (Hayes & Barry 2008; Bomford et al. 2009) are frequently used to predict the potential introduction success and distribution of alien or invasive species. The results of our study suggest that at least at the continental scale, the ranges of related species in the same family or order may be informative. This may be particularly helpful in situations where the range of the potential invader is poorly understood in comparison to relatives.

Although weak, there was a greater tendency for range variables related to breadth of distribution, compared to range position, to show a phylogenetic signal (based on correlations between species pairs). This pattern may be a matter of scale and reflect the fact that relatives can share some global life-history traits that dictate or limit the rate of extent of range expansion (e.g. woodiness), but may differ phenotypically in other respects enough to sort out at different positions along an ecological gradient. This pattern may also reflect the effects of competition for resources and potential for hybridization that may result in the displacement of species with respect to their position on ecological gradients. Since these interactions operate at a local scale, they may be less likely to influence the total breadth of distribution. This interpretation may also explain the greater degree of overlap between the ranges of polyploid and diploid congeners compared to diploid congeners. Polyploidization represents a strong reproductive barrier that largely isolates new polyploid species from their progenitors, so divergence in range position may be less strongly selected.

Previous studies on the phylogenetic signal of range size in animals have produced mixed results (Gaston 1998, 2003; Hunt, Roy & Jablonski 2005; Webb & Gaston 2005; Waldron 2007). Some studies involving bird taxa conclude that range size variation occurs primarily at the species level and, therefore, range size shows little phylogenetic signal (Gaston & Blackburn 1997; Böhing-Gaese & Oberrath 1999; Webb and Gaston 2003). However, others conclude that range can show a phylogenetic signal (Duncan et al. 2001; Blackburn et al. 2004; Hunt, Roy & Jablonski 2005; Waldron 2007). Similarly, one study of British butterflies found a significant effect of phylogeny (Hodgson 1993) while another did not (Freckleton, Harvey & Pagel 2002), and while two groups of fish, Cyprinella and suckers, showed a significant effect of phylogeny on range size (Taylor & Gotelli 1994; Freckleton, Harvey & Pagel 2002), sunfish did not (Freckleton, Harvey & Pagel 2002). Little or no evidence of phylogenetic signal was found in studies of geographic ranges for neotropical bats (Arita 1993), new world carnivores (Diniz-Filho & Torres 2002), new world venomous snakes (Reed 2003) or South-East Asian hawkmoths (Beck, Kitching & Linsenmair 2006), but those investigating fossil mollusks (Jablonski 1987; Jablonski & Hunt 2006), Australian marsupials (Freckleton, Harvey & Pagel 2002) and mammals (Jones, Sechrest & Gittleman 2005) did. This heterogeneity in phylogenetic signal across taxa has been attributed to poor resolution of range data, the taxonomic scope of the study and the geographic scale of the study (Jones, Sechrest & Gittleman 2005). Alternatively, this variation may suggest that there are important differences in the underlying determinants of geographic range for different taxa.

Our results contrast with previous research on plants by Peat & Fitter (1994) and Kelly & Woodward (1996), who found no evidence of a phylogenetic signal. The discrepancy may be explained by the differences in geographic scale of the research. The former studies were conducted on island floras (Great Britain, Crete). At this scale, the total variance in range size is limited, which may reduce the ability to detect a phylogenetic signal (Jones, Sechrest & Gittleman 2005). We chose to examine the North American range of angiosperms, which allows for large variation in range size, temperature and moisture. Interestingly, the results of our correlation analysis are similar to those from other studies based on large geographic areas such as Prinzing et al. (2001) on ecological range characteristics and Qian & Ricklefs (2004), who tested for correlation in range sizes between disjunct Asian – North American plant genera. These data are consistent with the idea that ecological tolerances governing broad scale distributions may be more conserved than smaller scale tolerances (Silvertown et al. 2006a,b).

Similarity of range attributes between relatives may be the result of not only shared genetic history, but could also reflect similarity in selective environments (Harvey & Pagel 1991). Sister taxa are more likely to arise in close proximity (if not in sympatry) and therefore experience selection on ecological tolerances that is more similar than that of unrelated species. Two factors suggest that this is unlikely to be the entire explanation. First, the species we chose were not restricted to phylogenetic sisters, but to congeners, which are less likely to have shared the same selective environments. Secondly, the signal of parallel evolution should diminish with increased taxonomic level. This is not the case for the species included in this study as order often explains a significant portion of the variance and, in some cases, explains more variation than family or genus. Therefore, based on significant variation at higher taxonomic levels and significant correlations among sister taxa, we suggest that genetic history is an important determinant shaping the ranges of North American angiosperms (Ricklefs & Latham 1992; Qian & Rickelfs 2004; Peat & Fitter 1994; Kelly & Woodward 1996).

Heteroploid versus homoploid divergence

Our results are not concordant with assertions in the literature that polyploid species have larger, more northerly ranges and broader ecological tolerances than diploids. These expectations have been based on the observation that genome duplication has direct effects on the physiology and morphology of plants and on the argument that traits in polyploids should be more plastic, owing to greater heterozygosity. In addition, it is widely believed that the adaptive potential of polyploids will be greater because of their increased capacity for genetic diversity and the potential for functional differentiation among gene replicates (Levin 1983; Otto & Whitton 2000; Wendel 2000; Ramsey & Schemske 1998;Crow & Wagner 2005). While, there is some evidence that polyploidy is associated with larger ranges in specific taxa, such as Clarkia (Lowry & Lester 2006), the evidence supporting this hypothesis as a general effect is largely anecdotal (Levin 2002) and has often been based on comparisons of the incidence of polyploidy in different floras rather than of comparisons of diploid and polyploid relatives that share recent evolutionary history.

We cannot exclude the possibility that differences between diploids and polyploids occur at a finer scale of resolution than our data allows; however, our results suggest this is not the case at a broad spatial scale. Results from our comparison of diploid–polyploid species pairs are consistent with those of two previous studies that contrasted the distributions of diploid and polyploid plants (Stebbins & Dawe 1987; Petit & Thompson 1999). Similarly, our result that polyploid ranges are not generally shifted north of the ranges of their diploid relatives concur with the results of Bowden (1940), who planted 100 species of diploid and polyploid plants and found no consistent differences in their degree of cold tolerance.

The polyploids included in this study are unlikely to be a homogeneous group. Rather they vary with respect to the mechanism of formation (autopolyploids versus allopolyploids) and the time since genome duplication (neopolyploid versus paleopolyploid). Indeed, since all angiosperms are estimated to have paleopolyploid origins, all plants included in this study likely have experienced multiple rounds of polyploidization in their evolutionary histories (Otto & Whitton 2000; Otto 2007). Unfortunately, the mode of origin and time since genome doubling is unknown for most of the polyploid species included here. However, by choosing plants that differ in the number of copies of the base chromosome number we have isolated the effects of genome duplication. While, we know of no theory that predicts how range and ecological characteristics vary with kind and age of polyploid, it is possible that such variation occurs (Levin 2002). Because they are the result of interspecific or intergeneric hybridization, we might expect allopolyploids to have broader ecological tolerances or a greater shift in range position in comparison to autopolyploids. Comparing the contribution of allopolyploidy versus autopolyploidy to species range characteristics would be an interesting undertaking once the mode of polyploidy origin is known for more species.

Our results do not imply that polyploidization has no effect on ecological tolerances or species range. Indeed, there is a rich literature corroborating the effect of genome duplication on habit, phenotype and physiology. Polyploids frequently differ in geographic range from diploid relatives. However, our study shows that, in general, the differences among polyploids and their congeners are not greater than the differences among homoploid congeners. This suggests that although the mechanism of divergence among heteroploid and homoploid species (genic change or genomic duplication) may differ, the magnitude of divergence between species does not. Indeed, the differences between heteroploid species may have been established rapidly through a single duplication event, while homoploid divergence occurred more gradually. Additionally, the significantly greater degree of overlap between diploid and polyploid congeners than between diploid congeners suggests that divergence between heteroploids may be different, leading to a greater potential for coexistence than does the divergence between homoploids. However, these ideas remain to be tested. Our results provide insights into the predictability of genome duplication effects. Our data reaffirm that polyploidization has no systematic effects on range attributes (position, size), just as it has no systematic effects on other phenotypic characters (Clausen, Keck & Hiesey 1940; Eigsti 1957; Levin 2002). However, our data suggest that the range attributes of a given polyploid depend, in part, on the range characteristics of its progenitor(s). Further work, using synthetic and naturally occurring polyploids, may help to understand this relationship and its impact on other aspects of phenotype.

Acknowledgements

We thank Lisa Pollock and Mary Jane Clark for help with data collection, Christopher Martin for help with programming and Dr Janet Mersey for help with ArcGIS and for providing the script for the calculation of area in ArcView. We also acknowledge financial support from Ontario Graduate Scholarship and Natural Science and Engineering Research Council of Canada to SLM, and from Premier’s Research Excellence Award, Natural Sciences and Engineering Research Council of Canada, Canada Research Chair Award and Canadian Foundation for Innovation to BCH.

Ancillary

Advertisement