- Top of page
We tested for the occurrence of Bergmann's rule, the pattern of increasing body size with latitude, and Rapoport's rule, the positive relationship between geographical range size and latitude, in 34 lineages of Liolaemus lizards that occupy arid regions of the Andean foothills. We tested the climatic-variability hypothesis (CVH) by examining the relationship between thermal tolerance breadth and distribution. Each of these analyses was performed varying the level of phylogenetic inclusiveness. Bergmann's rule and the CVH were supported, but Rapoport's rule was not. More variance in the data for Bergmann's rule and the CVH was explained using species belonging to the L. boulengeri series rather than all species, and inclusion of multiple outgroups tended to obscure these macroecological patterns. Evidence for Bergmann's rule and the predicted patterns from the CVH remained after application of phylogenetic comparative methods, indicating a greater role of ecological processes rather than phylogeny in shaping the current species distributions of these lizards.
- Top of page
The recent groundswell of species and higher level phylogenies coupled with the advent of modern comparative methods has provided evolutionary biologists with the opportunity to re-examine many long-held adaptive hypotheses. One area that has experienced such revitalization is the evaluation of various biological ‘rules’. In particular, Bergmann's and Rapoport's rules have attracted considerable recent attention. Bergmann's rule was proposed to explain a general pattern of increasing body size with increasing latitude or elevation (Bergmann, 1847; Ashton et al., 2000; Blackburn & Ruggiero, 2001; Ashton & Feldman, 2003; Reed, 2003), whereas Rapoport's rule describes the positive relationship between the geographical range size of species with increasing latitude or elevation (Stevens, 1989, 1992; Price et al., 1997; Stephens & Wiens, 2003). Documenting the underlying pattern and a possible basis for these rules is crucial for understanding patterns of species richness and distributions (Reed, 2003).
In his seminal work, Bergmann (1847, translated in James, 1970) described a relationship between increasing body size with decreasing environmental temperature among closely related species. Nearly a century later Mayr (1956) revised the definition to describe latitudinal variation in size within a species. Recently, Blackburn et al. (1999) defined Bergmann's rule as a tendency for a positive relationship between body mass of species belonging to a monophyletic higher taxon and the latitude inhabited by these species. By requiring monophyly of examined taxa, Blackburn and coworkers identified the importance of phylogenetic scale for identification of Bergmann's rule. Blackburn & Ruggiero (2001) extended their definition to include elevation as well as latitude.
Four hypotheses have been offered to explain Bergmann patterns (Gaston & Blackburn, 2000; Blackburn & Ruggiero, 2001): (1) phylogenetic history, (2) interspecific variation in migration, (3) variation in resistance to starvation and (4) variation in the ability of species to conserve or dissipate heat. However, there is no evidence that a single variable is responsible for Bergmann patterns (Partridge & Coyne, 1997) and these variables may be manifested differently among or within species. Gay & Best (1996) showed that among potential climatic variables affecting latitudinal gradients in body size within a species temperature was a better predictor than rainfall, although seasonality can be a better predictor than either rainfall or temperature in some cases (Lindstet & Boyce, 1985). Recent intraspecific studies found a decrease in body size of vertebrates presumed to be associated with global warming (O'Brien et al., 2000; Portner, 2001).
Selection acting on variation in thermal tolerance may limit the extent of geographical species distribution, thereby acting as a mechanism underlying Rapoport's rule (Gaston & Blackburn, 2000). The climatic-variability hypothesis (CVH) predicts that species with broad thermal tolerances should be resistant to variable climatic conditions (Stevens, 1989; Addo-Bediako et al., 2000). Hence, thermal tolerance should correlate positively with geographical range size. Yet, studies of the thermal tolerance of species over latitudinal or elevational gradients are rare (Addo-Bediako et al., 2000). Other mechanisms proposed to explain Rapoport's rule are: (1) differential extinction rates as a consequence of glaciations, (2) less competition at higher latitudes due to lower species richness, (3) the land–area relationship, especially in the Northern Hemisphere and (4) hard biogeographical boundaries (Gaston & Blackburn, 2000). The first two hypotheses are difficult to test, whereas land area and boundary effects can be controlled through careful selection of the study system.
The relationship between body size, geographical range size and latitude is not straightforward. Diverse groups of organisms demonstrate a strong correlation between body size and geographical range with large species having a large geographical distribution and small species occupying both large and small land areas (Gaston & Blackburn, 1996a,b; Reed, 2003). If large-bodied species are found at higher latitudes in accordance with Bergmann's rule, these species also may occupy a larger geographical area, biasing the effect of Rapoport's rule (Reed, 2003). Therefore, the link between body size, latitude, and geographical range must be assessed when studying these rules. Ideally, a rigorous test of these patterns would include a monophyletic group whose range was extensive in latitude and elevation and for which related biological data are available.
Lizards of the Liolaemus boulengeri series and closely related clades (subgenera Eulaemus and Liolaemus) provide a unique opportunity to perform tests for Bergmann's and Rapoport's rules within a phylogenetic framework, as well as address the importance of body size and the CVH for interpreting these ecological rules. The species in the L. boulengeri series range from 16 to 50°S latitude (spanning 3700 km of latitude) and from sea level to >5000 m. This group includes 26 described and several undescribed species (Etheridge, 1995; Etheridge & Espinoza, 2000; Etheridge & Christie, 2003). The taxonomy used in this paper follows Schulte et al. (2000). We present data collected for 21 species of the L. boulengeri series, including three undescribed, genetically distinct forms. Species belonging to this clade inhabit similar types of desert habitat (Monte, Puna and Patagonian steppe) and use similar microhabitats (Halloy et al., 1997; Schulte et al., 2004). Thus, we assume that by using similar microhabitats along similar types of deserts, the thermal characteristics of these microhabitats are comparable. Furthermore, the desert habitat occupied by members of the L. boulengeri series does not suffer the effect of narrowing towards the south (as with southern forests; Ruggiero & Lawton, 1998), so land area or profound biogeographical boundary effects are not expected.
Here, we test whether patterns of body size follow Bergmann's rule in the L. boulengeri series, as well as Liolaemus species sampled from other clades within the subgenera Eulaemus and Liolaemus. We also test whether the geographical distributions of these species follow Rapoport's rule. To evaluate thermal tolerance breadth as a mechanism to explain Rapoport's rule, we test for a relationship between this variable and latitudinal extent of geographical distribution. There has been increased interest recently in testing both rules (e.g. Gaston & Chown, 1999a; Addo-Bediako et al., 2000; Ashton, 2001b; Blackburn & Ruggiero, 2001), and the validity of Bergmann's and Rapoport's rules has been called into question (Gaston et al., 1998; Ashton, 2001b). For this reason, some authors refer to the patterns as effects instead of rules (Gaston et al., 1998; Ruggiero & Lawton, 1998). Even though these patterns do describe effects, we retain the use of their traditional names to be consistent with previous use in the literature.
Although including taxa generally increases statistical power, addition of taxa also expands the phylogenetic scale over which ecological patterns can be evaluated. If species in different clades exhibit dissimilar macroevolutionary patterns, the variance attributed to the inclusion of more distantly related taxa may obscure the detection of such patterns even if they are present within some clades in the analysis (Fig. 1). Consequently, it is important to evaluate patterns within, as well as among clades. As noted above, previous authors have disagreed over whether these rules are more appropriately applied intra- vs. interspecifically. We tested the influence of phylogenetic scale in detecting Bergmann's rule, Rapoport's rule, and the CVH in Liolaemus lizards by testing for these patterns using phylogenetically independent contrasts for 34 taxa representing members of several related clades, and just within the monophyletic L. boulengeri series. The availability of a phylogeny for lizards of the L. boulengeri series and related clades (Fig. 2) allowed us to evaluate the importance of phylogenetic scale in detecting broad ecological patterns such as Bergmann's rule and Rapoport's rule. The CVH has not been tested in ectothermic vertebrates with phylogenetically based comparative methods, despite arguments suggesting that traits related to thermal tolerance may be inherited and are taxonomically constrained (Bogert, 1949).
Figure 1. Phylogenetic scale may obscure macroecological patterns such as Bergmann's or Rapoport's rules. When ecological patterns differ for clades 1 and 2, there may be no correlation when the two clades are plotted together.
Download figure to PowerPoint
Figure 2. Phylogeny of Liolaemus lizards sampled in the L. boulengeri series and several species from the subgenera Liolaemus (L. kriegi, L. elongatus, L. petrophilus, L. cf. capillitas, L. bibronii, L. robertmertensi) and Eulaemus (all other species). Numbers below branches represent branch lengths obtained from parsimony analysis (left) and likelihood analysis (right) of mtDNA sequence data.
Download figure to PowerPoint
- Top of page
We collected data from 333 specimens representing 34 lineages of Liolaemus and ranging from north-western to south-eastern Argentina and from sea level to over 3800 m (Fig. 3). Individuals of each species were collected from one locality except L. multimaculatus and L. cuyanus that were collected at two proximate localities, and L. darwinii, which was collected at three localities. No statistical differences in thermal tolerance were found among these populations (L. darwini F2,17 = 2.56, P > 0.107; L. multimaculatus t8 = 2.09, P > 0.83, L. cuyanus t9 = 0.031, P > 0.73); therefore, we pooled the data for each species, however we are aware of low sample size in these data sets. Twenty-four taxa are members of the L. boulengeri series, including four species in the L. wiegmannii subclade. The remaining 10 taxa are used as outgroups with six species in the subgenus Liolaemus, three members of the L. montanus series and one member of the L. lineomaculatus section (L. somuncurae). Temperature data were not available for L. somuncurae and sequences were not available for L. kingii. However, these taxa are more closely related to each other than any other taxa used in our analyses (see Etheridge, 1995), so they were subsequently used interchangeably. The 15 non-boulengeri-series species were used as outgroups and to improve statistical power of the bivariate analyses based on independent contrasts (Bonine & Garland, 1999).
Figure 3. Map of Argentina indicating collecting localities of Liolaemus species. The bold line indicates the extent of Patagonian shrub-steppe and Monte xeric habitats. The habitat of the single locality on the Atlantic coast is sand dunes.
Download figure to PowerPoint
Lizards were captured by noosing or by hand between November and early March (austral summer) 1999–2003. Lizards were housed in small groups of the same species in aquaria in the lab where room temperature was maintained near 29 °C and exposed to a (13L : 11D) photoperiod. Temperature experiments (described below) were conducted no more than 12 days after lizards were captured. Lizards were not fed during this period but water was available ad libitum. Only nonreproductive adults were used for thermal trials. Thus, our data are comparable to previous studies of critical temperatures (Lutterschmidt & Hutchison, 1997).
Measures of temperature tolerance
We measured the critical maximum and minimum temperature (CTMax and CTMin, respectively) for each specimen. The critical temperature was recorded as the temperature at which individuals lost the ability to right themselves after being placed in a supine position (Cowles & Bogert, 1944; Spellerberg, 1973; Carothers et al., 1997). To determine CTMax, lizards where placed individually in an aluminium cylinder (260 mm diameter × 340 mm deep), containing a 30 mm layer of sand to avoid thermal conductance from the aluminium bottom. Individuals were previously cooled to 18 °C, then placed in the cylinder and heated by a 100 W light bulb 70 mm above the container. A thermocouple was taped to the lizard cloaca and body temperature (Tb) was monitored every 30 s, and every 10 s following the onset of gaping. To determine CTMin, lizards (Tb 20 °C) were placed individually in plastic containers (lids had holes for temperature and air exchange), which were placed in a −10 °C freezer, and their Tb was monitored every 20 s. The thermal tolerance range of each species was calculated by subtracting the mean CTMin from the mean CTMax.
Data on maximum body size (snout–vent length; SVL) and distribution (degrees latitude and elevation in m) were taken from the literature (Cei, 1986, 1993; Etheridge, 1995; Espinoza et al., 2004, Table 1). SVL is a widely accepted measure of body size for squamate reptiles and is correlated with body mass and numerous ecological, morphological and life-history traits (Pough, 1973).
Table 1. Species of Liolaemus lizards analysed.
|Liolaemus species||n||SVL (mm)||Elevational range (m)||Latitudinal range (°S)||Thermal tolerance range CTMin–CTMax (°C)|
|L. andinus poecilochromus||6||72||3500–4400||24°00′–27°00′||9.35–44.16|
|L . cf. telsen||18||75||0–800||42°55′–46°29′||8.98–44.60|
|L. cuyanus ||18||102||400–2000||27°19′–33°00′||12.00–45.70|
|L. cf. elongatus ||4||85||700–3000||29°00′–46°00′||7.68–45.10|
|L. olongasta ||10||67||900–1770||28°38′–31°14′||11.85–44.58|
|L. riojanus ||2||62||500–1000||29°00′–32°00′||10.70–45.20|
|L . sp. 1||3||95||400–2000||33°00′–37°39′||7.30–45.00|
|L . sp. 2||9||90||0–400||44°00′–44°45′||8.06–44.70|
|L . sp. 3||8||68||3300–3400||23°00′–23°22′||9.67–43.85|
Recent work on the taxonomy and distribution of the L. boulengeri series has greatly improved our knowledge of this species-rich South American lizard clade (Etheridge & Espinoza, 2000; Etheridge & Christie, 2003). As noted above, there are several undescribed species included in our analyses, as well as other discovered, yet undescribed species. The taxa we sampled for this study are the most well studied members of the L. boulengeri series whose distribution is known with reasonably good accuracy. Undescribed species are expected to have a close phylogenetic relationship to all species sampled for this study (Etheridge & Christie, 2003), so expected to have similar body size, geographical distributions, and life history traits.
Two independent data sets are available that include most of the species in our analyses (Etheridge, 2000; Schulte et al., 2000). We chose data from Schulte et al. (2000) because all taxa in this study could be included allowing the estimation of branch lengths for comparative analyses. We present new sequence data for species not considered in Schulte et al. (2000) and include sequences presented in Harmon et al. (2003). New sequence data include L. cf. telsen, L. sp. 1, L. sp. 2, and L. sp. 3. New sequences have been deposited in GenBank (accession numbers to be included upon acceptance of the manuscript) and the alignment used for phylogenetic analyses is available from TreeBase (Study accession number S1281; Matrix accession number M2237). Harmon et al. (2003) include phylogenetic data for L. canqueli, L. kriegi, L. riojanus and L. xanthoviridis.
The phylogeny and branch lengths were obtained using PAUP* beta version 4.0b10 (Swofford, 2002) from maximum parsimony (MP) and maximum likelihood (ML) analysis of mtDNA sequences spanning the protein-coding genes ND1 to COI (Schulte et al., 2000) using only the species for which ecological data were available. Phylogenetic trees were estimated with 200 heuristic searches featuring random taxon addition under the MP optimality criterion. For ML analyses, the best fitting model parameters using the General Time Reversible substitution model with proportion of invariant sites and among site variation estimated from the sequence data were fixed then used in 25 heuristic searches with random addition of taxa to find the overall best likelihood topology. ModelTest v3.06 (Posada & Crandall, 1998) was used to find the best fitting model of sequence evolution for the tree from the unweighted parsimony analysis of these molecular data. The resulting phylogenies under MP and ML are shown in Fig. 2.
Temperature declines with increasing elevation as well as increasing latitude (Bergmann, 1847; Ashton, 2002a). Therefore, it is necessary to account for the effect of elevation on ambient temperature when investigating the temperature dependence of the latitudinal range of a species. Assuming equal conditions, temperature declines 0.65 °C for every 100-m increase in elevation (Lutgens & Tarbuck, 1998). To correct our dataset for latitudinal and elevational covariation in temperature, we plotted the reduction in temperature every 2° latitude from 20 to 55°S against the elevational temperature gradient (0.65 °C for every 100-m rise in elevation), and computed a correction factor by adding 1.752° latitude for every 200-m increase in elevation after a baseline of 600 m above sea level. We applied this constant to adjust the latitudinal range midpoint for each Liolaemus species according to its elevational midpoint occurrence. This value is referred hereafter as the adjusted latitudinal midpoint.
Bivariate linear regressions were calculated between each of the following variables: extent of geographical range (degrees latitude), latitudinal range midpoint, elevational range midpoint, CTMin, and adjusted latitudinal range of each species. To test for Bergmann's rule, we used SVL as the dependent variable. Extent of geographical range is the most common variable used when testing Rapoport's rule and tends to be highly correlated with area measures of range size (Gaston et al., 1998; Reed, 2003). Thermal tolerance range and CTMin were used as dependent variables in regressions against extent of geographical range and adjusted latitude in testing the CVH as a mechanism underlying Rapoport's rule. As mentioned above, SVL in squamate reptiles correlates strongly with numerous biological traits. This body size measure was regressed against thermal tolerance range and CTMin to explore its impact on our analyses of Bergmann's rule, as well as extent of geographical range (Gaston & Blackburn, 1996a,b; Reed, 2003) to examine for a potential bias in our test of Rapoport's rule.
A priori and a posteriori tests for serial independence (TFSI; Abouheif, 1999) were used to assess the adequacy of applying independent contrasts to our data. We applied the tests for all 34 taxa, and for species only in the L. boulengeri clade. The null hypotheses in TFSIs were that trait data are independent and not significantly phylogenetically autocorrelated. When we failed to reject the null hypothesis of no phylogenetic autocorrelation by a priori TFSIs, we used conventional methods of analysis. When the hypothesis was rejected in an a priori test, we proceeded with independent contrast analyses (Felsenstein, 1985) for that trait dataset. A posteriori TFSIs were then performed on the independent contrasts, and the procedure ended if we failed to reject the null hypothesis of serial independence. Abouheif (1999) recommended caution in interpreting results in cases when the null hypothesis was rejected in a posteriori tests, because phylogenetic signal could persist in the dataset even after independent contrasts were computed.
We used environmental traits, specifically elevational and latitudinal range, in comparative analyses. Individuals inherit adaptations from ancestors that may also have inhabited similar environments (Garland et al., 1992). For example, the ancestor of species capable of living in a certain elevational range may also have been capable of inhabiting the same elevational gradient. Hence, the heritability of these traits is complex because of heritability of traits at the individual level can scale up to heritabilities at the level of species (Jablonski, 1987). We computed phylogenetically independent contrasts (Felsenstein, 1985) using COMPARE v. 4.4 (Martins, 2001). Regressions of independent contrasts, with positivized horizontal axes, were forced through the origin to allow for meaningful interpretations (Garland et al., 1992).
Traits such as body size or thermal tolerance may have evolved in different ways among members of a clade, and may fit one evolutionary model better than another (Garland et al., 1992; Mooers et al., 1999). For this reason, we applied three evolutionary models to our data prior to computing independent contrasts: (a) a nonhistorical model using untransformed species data for the ecological and geographical traits (internal nodes were collapsed and tip branch lengths set equal to one), (b) gradual model using branch lengths derived from the phylogenetic analysis, (c) speciational model where branch lengths were set equal to 1.0.
- Top of page
R. Martori, P. Bellagamba, M. Kozykariski, L. Belver, L. Aun, N. Ibargüengoytía and A. Marcus helped in the field. M Angilletta, A. Arrington, I. Castro, G. Dayton, T. Hibbits, J. Losos, G. Perotti, W. Ryberg, J. Wiens, K. Winemiller and two anonymous reviewers provided helpful comments on the manuscript. T. Hibbitts helped with statistics. CRILaR-CONICET provided logistical support. Funding was provided by FONCYT PICT98-04-0867, PIP CONICET 2846, a CONICET postdoctoral fellowship and a Senior's Program Fulbright fellowship (FBC), American Society of Ichthyologists and Herpetologists, National Geographic Society, Society of Integrative and Comparative Biology, Chicago Herpetological Society, Upstate [NY] Herpetological Association, Department of Biology, Graduate School and Biological Resources Research Center at the University of Nevada, Reno, and a release-time grant from California State University, Northridge (REE), and by NSF DEB-9318642, DEB-0071337 and DEB-9982736 (JAS).