SEARCH

SEARCH BY CITATION

Keywords:

  • adaptation;
  • common garden experiment;
  • elevation;
  • molecular adaptation;
  • outlier scan;
  • phenotypic divergence

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature survey
  5. Conclusions
  6. Acknowledgments
  7. References
  8. Supporting Information

Altitudinal gradients offer valuable study systems to investigate how adaptive genetic diversity is distributed within and between natural populations and which factors promote or prevent adaptive differentiation. The environmental clines along altitudinal gradients tend to be steep relative to the dispersal distance of many organisms, providing an opportunity to study the joint effects of divergent natural selection and gene flow. Temperature is one variable showing consistent altitudinal changes, and altitudinal gradients can therefore provide spatial surrogates for some of the changes anticipated under climate change. Here, we investigate the extent and patterns of adaptive divergence in animal populations along altitudinal gradients by surveying the literature for (i) studies on phenotypic variation assessed under common garden or reciprocal transplant designs and (ii) studies looking for signatures of divergent selection at the molecular level. Phenotypic data show that significant between-population differences are common and taxonomically widespread, involving traits such as mass, wing size, tolerance to thermal extremes and melanization. Several lines of evidence suggest that some of the observed differences are adaptively relevant, but rigorous tests of local adaptation or the link between specific phenotypes and fitness are sorely lacking. Evidence for a role of altitudinal adaptation also exists for a number of candidate genes, most prominently haemoglobin, and for anonymous molecular markers. Novel genomic approaches may provide valuable tools for studying adaptive diversity, also in species that are not amenable to experimentation.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature survey
  5. Conclusions
  6. Acknowledgments
  7. References
  8. Supporting Information

The geographical distribution of many species is so broad that various characteristics of their environment vary either abruptly or in a clinal manner within their range. A common pattern observed in response to such environmental heterogeneity is local adaptation, where, at a given location, the fitness of local individuals is higher than that of immigrants from other environments (Kawecki & Ebert, 2004). Local adaptation is possible if populations contain ecologically relevant genetic variation and if divergent selection between different environments is strong relative to the rate of gene flow (Morjan & Rieseberg, 2004). The distribution of adaptive genetic diversity and the factors promoting or preventing adaptive divergence are of fundamental interest to evolutionary ecologists but remain poorly characterized in most natural populations (see Hereford, 2009a for recent review). It is also largely unclear how consistently different species respond to similar selection pressures. A better understanding of these issues has direct implications for conservation and management, for example, by improving our ability to assess the future of populations in rapidly changing environments or to anticipate the effect of changes to population connectivity.

In addition to the environmental changes observed in space, anthropogenic climate change is expected to lead to temporal changes, but its implications for local climatic conditions are likely to vary widely. Thermal environments at high latitudes, for example, may become more similar to the current thermal environments at lower latitudes. Other environmental variables, notably day length, are not expected to change, which may lead to completely novel environmental conditions. This suggests that range shifts (Parmesan & Yohe, 2003; Parmesan, 2006) may be insufficient for locally adapted populations to track their preferred (multidimensional) environment and additional responses are necessary. Phenotypic plasticity provides one mechanism to deal with environmental variability, but plastic responses may be possible only within certain limits, and evolutionary change may be necessary in the face of large and consistent environmental change (e.g. Gienapp et al., 2008). Indeed, a number of case studies report evidence of such microevolutionary changes in response to global warming in natural populations (reviewed in Bradshaw et al., 2006; Hoffmann & Sgrò, 2011).

Spatial gradients can serve as surrogates for at least some of the temporal changes anticipated under climate change (Reusch & Wood, 2007), providing an opportunity to investigate the historical and current responses of natural populations to climate-related selection pressures. Altitudinal gradients are particularly relevant in this context because they are also climate gradients. Some of the environmental changes along altitudinal gradients are specific to certain locations or biogeographical regions, whereas others, namely decreasing temperature, decreasing atmospheric pressure and increasing intensity of solar radiation (Körner, 2007), are physical properties shared by altitudinal gradients worldwide, allowing particular effects to be studied in numerous replicated systems.

Altitudinal gradients offer a valuable contrast to latitudinal gradients, especially with respect to geographical scale. Altitudinal gradients are typically steep, with environmental transitions occurring at spatial scales that are small relative to the dispersal distances of many species. This has several important implications. First, it means that the effects of divergent selection may often be opposed by gene flow, which, if strong enough, acts to homogenize allele frequencies between environments. Altitudinal gradients thus provide the opportunity to investigate whether, and under which conditions, adaptive divergence is possible in the face of gene flow. Second, the small geographical scale of altitudinal gradients also implies that confounding effects, such as distinct regional evolutionary histories, are less of an issue than in latitudinal surveys, which are often performed across thousands of kilometres (e.g. Balanyá et al., 2006).

Many examples exist of phenotypic transitions associated with the changing environment along altitudinal gradients. Plants often show conspicuous intraspecific differences in growth form or leaf morphology between high and low altitudes (Körner, 2003), whereas phenotypic differences in animals can involve body size clines (Chown & Klok, 2003) or, perhaps more conspicuously, transitions from one generation per year at high altitudes to two or more at lower altitudes (Hodkinson, 2005). What is often less clear, however, is whether these phenotypic differences have a genetic basis or result entirely from plastic responses to the environment.

As a step towards a better understanding of the distribution of adaptive genetic diversity along altitudinal gradients, we surveyed the literature for studies on genetic divergence in phenotypic traits and at functional loci in animal populations. Our particular goal was to identify general patterns emerging from these studies by asking (i) whether some taxa are more prone to differentiation than others, for example due to differences in specific species traits, (ii) whether adaptive divergence is apparent at all geographical scales or whether it tends to be rare between nearby populations where gene flow may be high and (iii) whether different species show similar responses to selection pressures associated with altitude.

Literature survey

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature survey
  5. Conclusions
  6. Acknowledgments
  7. References
  8. Supporting Information

We performed a literature survey to identify studies investigating genetic differentiation between animal populations (both aquatic and terrestrial) along altitudinal gradients. We used ISI Web of Knowledge to conduct a search for papers on (altitud* OR elevation*) AND (gradient OR transect OR cline) AND (genetic OR ‘common garden’ OR transplant) NOT plant to obtain a list of ca. 500 publications. Based on the abstracts, we retained two types of studies. The first consisted of papers reporting measurements of phenotypic traits for individuals from different altitudinal origins studied under common garden conditions or in a reciprocal transplant design. Data on phenotypic traits collected directly in the field were included only in three cases where additional information supported a genetic basis for the trait (wing melanization, Ellers & Boggs, 2002, 2004a; macroptery, Fairbairn & King, 2009; and heat-shock protein expression, Dahlhoff & Rank, 2000). The second group contained papers providing molecular evidence of adaptive genetic differences between populations at different altitudes. These studies typically investigated associations between genotypes and altitude or used outlier locus detection (e.g. Storz, 2005) to identify loci showing unusually high between-population differentiation. The focus was on intraspecific phenotypic or genetic diversity. Papers on incipient species pairs were included only if there was evidence of on-going gene flow between the two species. Additional studies were identified based on the bibliography in relevant papers as well as from thematically related review articles (Leinonen et al., 2008; Conover et al., 2009; Hereford, 2009; Nosil et al., 2009).

Phenotypic data

Data availability and methodological limitations

A total of 68 publications met our selection criteria for phenotypic data, and these contained data from 44 different species: 24 arthropods, 19 chordates and one mollusc. From 66 of these papers (Table S1), we were able to extract data for at least one phenotypic trait from tables or from figures using g3data (http://www.frantz.fi/software/g3data.php).

Among the arthropods, Diptera was the best-represented order with 14 different species, followed by Lepidoptera and Orthoptera with three species each. Among the chordates, two-thirds of the species were amphibians or reptiles. The maximum altitudinal distance between sampling locations ranged from 126 m (Eales et al., 2010) to 4000 m (León-Velarde et al., 1996). Most studies were performed at a relatively small scale with a median Euclidian distance of 135 km between the two most distant source populations (range: 5–3900 km, distances were estimated using Google Maps if not provided in the original publication).

The majority of papers reported data on phenotypic traits recorded from individuals reared under the same environmental conditions. Although such common garden experiments are valuable to investigate the genetic basis of traits (but see caveat below), the adaptive significance of between-population differences cannot be inferred. Such an assessment would require results from reciprocal transplants or at least from multiple common garden experiments, which try to mimic the range of natural conditions associated with changes in altitude. Such reciprocal transplant experiments have been performed only for three species (the frog Rana sylvatica, Berven, 1982a,b; the butterfly Colias philodice eriphyle, Ellers & Boggs, 2004ab; the lizard Psammodromus algirus, Iraeta et al., 2006), in all cases in addition to common garden experiments.

Further, to exclude effects of the native environment on phenotypes, experimental animals should be reared in a common environment for two or more generations prior to measurement (Kawecki & Ebert, 2004). This was the case in only about one-third of the studies, all of them on flies. Another 13% of the studies used wild-caught individuals (F0), whereas the majority used laboratory-reared offspring of wild-caught animals (F1; 38%). In these cases, the phenotype may still be affected by the native environment through maternal effects (Kawecki & Ebert, 2004).

Analysis and graphical overviews

We used the assembled data to investigate whether populations from different altitudes show genetically based differences in phenotypic traits, initially ignoring the adaptive significance of these differences. Further, we investigated whether the altitudinal trends observed for a given trait were similar across species. This second analysis was conducted for 14 trait categories measured in at least five different species, where each trait category included several related traits as indicated in Figs 1 and S1. Melanization was also included due to its potential relevance for thermoregulation, even though this trait was only studied in three species. To account for differences in the range of trait values, all observations were standardized within trait and study to mean 0 and variance 1. These standardized trait values were then regressed against the altitude of the source population assuming a linear relationship. Regressions were calculated separately for each trait, study, common garden environment and other subgroups (e.g. sex, age class or study year) if available. Figures 1 and S1 show the regression slopes for different trait categories, and Table 1 provides a summary of the main patterns.

Table 1. Number of animal species showing significant phenotypic differences between altitudinal populations reared in a common environment for different traits. Patterns are summarized based on the detailed figures. (A) An increase in trait value with altitude is supported by all statistically significant tests based on > 2 populations (i.e. all large blue circles > 0 and no large orange circles). (B) A decrease in trait value with altitude is supported by all statistically significant tests based on > 2 populations (i.e. all large blue circles < 0 and no large orange circles). (C) Statistically significant differences between populations are not or not consistently associated with altitude. This category includes only species in which altitudinal effects were formally tested. In particular, estimates based only on two populations (small circles) are not considered. In parentheses, we indicate the number of studies relying on animals reared in a common environment for two or more generations before the experiments. Traits are listed in the order in which they appear in the text
Trait categoryPutative agent(s) of selectionExpected altitudinal patternNumber of species showing# species# different taxonomic groupsaDetails
(A) Consistent increase(B) Consistent decrease(C) Pop differences not associated with alt
  1. HSP, heat-shock protein; T, temperature.

  2. a

    Number of different orders (for Arthropoda) or classes (for Chordata, Mollusca).

  3. b

    HSP expression is a very complex and general response to cellular stress, and predictions are difficult. The observed response may depend heavily on common garden conditions (e.g. if these are closer to high- or low-altitude conditions).

  4. c

    Water availability often changes with altitude, but the direction of the change may vary among regions.

  5. d

    If mortality risks increase with altitude due to increased environmental stochasticity, this could select for higher fecundity early in life. However, the data set contains only two estimates of fecundity early in life and no trends can be inferred.

Wing sizeAir densityIncrease3 (3)02 (2)93Fig. 1a
Heat toleranceTDecrease02 (1)2 (2)53Fig. S1a
Cold toleranceTIncrease3 (2)02 (2)63Fig. 1b
HSP expressionT, other stressorsUnclearb02 (0)3 (2)54Fig. S1b
Desiccation toleranceWater availabilityVariablec002 (2)51Fig. S1c
MassT, resource availabilityIncrease (?)3 (1)01 (1)76Fig. 1c
Body lengthT, resource availabilityIncrease (?)02 (0)074Fig. S1d
Development timeT, season lengthDecrease (?)03 (0)3 (2)135Fig. 1d
Growth rateT, season lengthIncrease (?)001 (0)63Fig. S1e
Longevity Decrease (?)01 (0)3 (2)62Fig. S1f
Viability Unclear002 (2)113Fig. S1g
Fecundity Uncleard1 (1)1 (1)3 (3)114Fig. S1h
image

Figure 1. Observed changes in phenotypic traits in animals along altitudinal gradients. Shown are slopes from linear regression of trait value against altitude of the source population, with trait values standardized to mean 0 and variance 1 within trait and study. Separate estimates are shown for each data set, where data sets can be different common garden environments, latitudes, age classes, sexes, etc. Each row represents a different species and summarizes data from one or several studies, as indicated in the ‘ref’ column. Several related traits were combined into each trait category as specified in the ‘traits’ column. Shading indicates if the original publication reported statistically significant effects of altitude (blue) or population (orange), a nonsignificant (n.s.) effect (red) or did not provide test results (empty circle). Note that we did not distinguish between main effects and interactions involving altitude or population. The inclusion of significant interaction terms explains why some slope estimates near zero are displayed as statistically significant. The size of the circles indicates the number of source populations available to estimate the slopes (small: 2 sources; large: > 2 sources). Asterisks indicate traits for which the sign of the slope was reversed to produce consistent patterns across traits. For the wing size traits, for instance, wing loading is the only trait where a smaller trait value would indicate a larger wing. Consequently, if unadjusted, an increase in wing size with altitude would result in a negative slope for wing loading but a positive slope for all remaining traits. (a) Traits: a, wing/thorax ratio; b, wing loading*; c, wing length; d, wing width; e, wing centroid size; f, wing area. Ref: 1 = Bears et al. (2008), 2 = Bubliy & Loeschke (2004), 3 = Dahlgaard et al. (2001), 4 = Norry et al. (2001), 5 = Sambucetti et al. (2006), 6 = Collinge et al. (2006) 7 = Pitchers et al. (2012), 8 = Stalker & Carson (1948), 9 = Tantowijoyo & Hoffmann (2011), 10 = Belen et al. (2004), 11 = Karan et al. (2000), 12 = Karl et al. (2008). *Sign of slope reversed. (b) Traits: a, chill coma recovery time*; b, cold shock survival; c, lower limiting T for embryonic development*. Ref: 1 = Beattie (1987), 2 = Bridle et al. (2009), 3 = Sarup et al. (2009), 4 = Sorensen et al. (2005), 5 = Collinge et al. (2006), 6 = Parkash et al. (2010), 7 = Karl et al. (2008). *Sign of slope reversed. (c) Traits: a, tadpole; b, at metamorphosis; c, pupa; d, at hatching; e, adult. Ref: 1 = Ficetola & De Bernardi (2005), 2 = Jasienski (2009), 3 = Sommer & Pearman (2003), 4 = Buckley et al. (2010), 5 = Bears et al. (2008), 6 = Stillwell & Fox (2009), 7 = Karan et al. (2000), 8 = Karl et al. (2008). (d) Traits: a, embryonic development time (dt); b, larval dt; c, postdiapause dt; d, pupal dt; e, egg-adult; f, hatching-adult. Ref: 1 = Beattie (1987), 2 = Jasienski (2009), 3 = Marquis & Miaud (2008), 4 = Tsuchiya et al. (2012), 5 = Bubliy & Loeschcke (2004), 6 = Folguera et al. (2008), 7 = Norry et al. (2001), 8 = Sambucetti et al. (2006), 9 = Collinge et al. (2006), 10 = Etges (1989), 11 = Belen & Alten (2006), 12 = Blanckenhorn (1997), 13 = Karl et al. (2008), 14 = Tanaka & Brookes (1983), 15 = Dingle et al. (1990), 16 = Berner et al. (2004).

Download figure to PowerPoint

Studies including only one high- and one low-altitude source population, where regression slopes were estimated based on only two data points, were distinguished from studies with multiple populations from each elevation. Additionally, we retained information about statistical significances as reported in the original publications, distinguishing between significant effects of altitude and of population. The latter category included (i) studies using only two source populations, in which case altitudinal and population effects could not be distinguished (e.g. Conover et al., 2009), (ii) reports that did not formally test for altitudinal effects or (iii) studies where significant population differences were not associated with altitude. We did not distinguish between significant main effects and significant two-way interactions involving altitude or population. This inclusion of significant interaction terms explains why slope estimates near zero are sometimes displayed as statistically significant in the figures.

Molecular data

The second focus of our literature survey was on studies looking for signatures of divergent altitudinal selection at the molecular level. Many of these studies made use of outlier locus detection (e.g. Storz, 2005) to determine whether between-population divergence at a given gene or anonymous marker was significantly higher than the genomic average. Additionally, we included studies showing altitudinal clines in the frequency of particular alleles at candidate genes or anonymous regions of the genome. Similar to the genome scan approaches, these patterns should ideally be compared to those at putatively neutral genetic markers to exclude the possibility that clines result from purely neutral processes (e.g. isolation by distance; Storz, 2002). Such data from neutral loci were, however, not always available (see comments in Table 2).

Table 2. Candidate loci and anonymous markers (outlier loci) showing evidence of adaptive differentiation along altitudinal gradients in animalsThumbnail image of

We identified 30 studies that present evidence of divergent selection under the criteria outlined above (Table S1). These studies investigate 22 different species, and the taxonomic focus was less biased towards particular groups (i.e. Diptera) than in the phenotypic data set.

Genetically based phenotypic variation along altitudinal gradients

In many studies, phenotypic traits measured in common garden environments varied significantly between populations from different altitudinal origins. Among the statistical tests performed in the original publications, 73% detected significant differences between source populations or altitudes (including main and interaction effects; Fig. 2). A very similar proportion of significant test results was observed when we considered only studies using experimental animals bred in a common environment for at least two generations (70%; based on 142 tests from 27 studies).

image

Figure 2. Number of significant (dark grey) and nonsignificant (light grey) results as reported in the original publications for different geographical scales in animal species. The results from studies performed at a scale of less than 100 km are plotted again at a higher resolution in the small inset and show that significant phenotypic differences between populations can be observed even at very local scales. The numbers above the bars indicate the number of independent studies contributing to each distance class. Significant effects include main effects or interactions involving population or altitude.

Download figure to PowerPoint

We then asked whether traits measured in multiple species tended to show the same changes along altitudinal gradients. For the trait categories measured in five or more species, clear predictions of the variation with altitude could be formulated for three (Table 1), whereas for several additional trait categories, the expected patterns were more difficult to predict (Table 1). In the following discussion, we particularly focus on species showing statistically significant associations with altitude for particular traits (as reported in the original study), especially if these are consistent across data subsets or studies (columns A and B in Table 1). We additionally report all species for which significant between-population differences in a given trait were detected, but for which trait values did not change linearly with altitude (column C in Table 1). This latter category includes only species for which altitudinal effects were formally tested. The results from additional studies in which the experimental design precluded testing altitudinal effects or where such effects were either not tested or not significant are shown in Figs 1 and S1 but not considered in Table 1 and the following discussion.

First, air density decreases with altitude, and in addition to its significance for respiration, this also implies that more power is needed for flight. A possible adaptive response includes an increase in wing size relative to body size (e.g. Dillon et al., 2006). In our data set, traits related to wing size were investigated in seven fly species, a butterfly and a bird (Fig. 1a), but only for two traits – wing to thorax ratio (a in Fig. 1a) and wing loading (b in Fig. 1a) – relative to body size. In all three studies reporting significant altitudinal effects, wing size increased consistently with altitude (Fig. 1a; Table 1).

Air temperature is a second environmental variable that shows consistent altitudinal clines, dropping an average 5.5°C per 1000 m (e.g. Körner, 2007). Not surprisingly, traits potentially relevant to thermal adaptation were well represented in our data set, including diverse morphological, physiological, developmental and behavioural traits. The most obvious prediction is that the average cold tolerance of individuals should increase with altitude, whereas heat tolerance should decrease (Table 1). For heat tolerance, some significant between-population differences were reported in all five species (Fig. S1a), but only in two cases were these differences related to altitude. In both, heat tolerance decreased with altitude as predicted. Cold tolerance was investigated in four Drosophila species, one frog and one butterfly (Fig. 1b). In three of these species, an increase in cold tolerance was observed in populations from higher altitudes, while in a further two significant population differences were reported that were not associated with altitude (Table 1). Interestingly, the latter studies were all conducted with populations from equatorial regions (< 30° north/south), whereas all studies with higher latitude populations did find a positive correlation of cold tolerance with altitude. Heat-shock proteins (Fig. S1b), which are involved in general responses to cellular stress (Morris et al., 2013), showed either decreasing expression levels with altitude (two species) or between-population differences that were not consistently associated with altitude (three species; Table 1).

Clinal change with altitude is also predicted for body melanization, a morphological trait that probably has thermoregulatory relevance, as darker bodies absorb more energy (Clusella Trullas et al., 2007). Body melanization along altitudinal gradients was studied in only three species, with the expected positive correlation with altitude being found in each case (butterfly Colias philodice eriphyle, Ellers & Boggs, 2002, 2004a; Drosophila melanogaster, Sub-Saharan Africa: Pool & Aquadro, 2007; India: Parkash et al., 2008, 2010; Drosophila americana, Wittkopp et al., 2011). In addition to the thermoregulatory advantages, darker individuals could be better protected against elevated UV radiation at higher altitudes, and in some Drosophila populations, body pigmentation also shows a strong positive correlation with desiccation tolerance (Parkash et al., 2008; but not Wittkopp et al., 2011; see Fig. S1c for results on desiccation tolerance).

The altitudinal patterns expected for traits related to body size are less clear (Table 1). According to Bergmann's rule (e.g. Gardner et al., 2011), endotherms tend to be larger in cooler environments due to the thermoregulatory advantages arising from a smaller surface to volume ratio. Some ectotherms also comply with Bergmann's rule, although the underlying mechanisms are likely to be different (Gardner et al., 2011); furthermore, the opposite pattern is also common (Blanckenhorn & Demont, 2004). Our data set, which contained information for seven ectotherms from different taxonomic groups, showed that in three species, body mass tended to increase with altitude (Fig. 1c). Body length changed in the opposite direction, with a significant decrease with altitude reported from two insects (Fig. S1d; Table 1).

In temperate regions, the period available for growth shortens with increasing altitude (Körner, 2007). At the same time, the completion of different developmental stages may take longer, especially in ectotherms, because lower ambient temperatures slow down physiological processes. An expected adaptation to these conditions involves a compensatory response, with high-altitude individuals developing or growing faster than low-altitude individuals under a given thermal regime (i.e. countergradient variation; e.g. Hodkinson, 2005). Such a pattern was indeed observed in all three species for which development time was found to be significantly associated with altitude (Fig. 1d; Table 1), although the same number of between-population differences was found that were unassociated with altitude. Similarly, none of the observed between-population differences in growth rate were associated with altitude (Fig. S1e; Table 1).

Viability, longevity and fecundity were additional life-history traits, which were repeatedly investigated, almost exclusively in flies. A possible expectation here could be that more variable and unpredictable high-altitude environments favour a faster pace of life, characterized by high investments in reproduction early in life (Tieleman, 2009) and, perhaps, reduced longevity (Table 1). Fecundity early in life has been estimated in only two species (indicated by asterisks in Fig. S1h). All three traits showed some significant between-population differences, although demonstrations of altitudinal patterns were rare (Fig. S1f–h; Table 1). Remarkably, all four studies using multiple rearing temperatures found that the decrease in longevity with altitude was strongest in the coldest environment, that is, the conditions most strongly resembling high-altitude conditions.

Overall, our literature survey provided clear evidence for significant, genetically based phenotypic differences between populations of different altitudinal origin. Comparisons across species identified several traits for which parallel clinal patterns were observed in two or three species, and for which the phenotypic changes consistently occurred in the predicted direction (melanization, wing size, cold and heat tolerance, mass, development time; A or B in Table 1). However, in all of these cases, several species also showed significant between-population differences that were not, or at least not consistently, associated with altitude (C in Table 1). Overall, the available data are clearly limited. The number of species for which data on a given trait category were available was typically small (≤ 13) and skewed towards particular taxonomic groups; furthermore, for several species, only two source populations had been studied, making it impossible to distinguish population effects from altitudinal effects.

Evidence of adaptive genetic divergence from molecular studies

Almost all vertebrates rely on haemoglobin (Hb) for the transport of oxygen. Hb is therefore an obvious candidate gene for adaptation to the changing O2 partial pressure along altitudinal gradients, and the genes coding for different Hb subunits have been studied in a number of species, most prominently birds (Table 2). All of these studies found evidence of divergent selection for at least some of these genes. In deer mice, Storz et al. (2009) further demonstrated that the β-globin variant common in high-altitude populations has indeed a higher O2 affinity.

Additional genes with a potential role for adaptation to low O2 partial pressure have been detected in humans through genome scans (Table 2). For example, one enzyme from the hypoxia-inducible factor pathway (EGLN1) has been identified as a potential target of selection in both Tibetans and Andeans, but the haplotypes common at high altitudes differ between the two regions. Other genes have been implicated in high-altitude adaptation in only one of the two populations (Table 2).

Evidence of divergent selection along altitudinal gradients is also available for other candidate genes, including mitochondrial loci and several allozymes. For many of these loci, a role in adapting to thermal conditions is very plausible (Table 2), and in some cases, a direct link between genotype and phenotype of relevance for altitudinal adaptation has been demonstrated. For example, Fontanillas et al. (2005) found that nonshivering thermogenesis (i.e. mitochondrial heat production in brown fat cells) in white-toothed shrews was influenced by an interaction between sex and mitochondrial haplotype. Copper butterflies from low-altitude sites but with high-altitude-like genotypes at an allozyme locus (phosphoglucose isomerase) resembled high-altitude individuals with respect to development rates and chill-coma recovery time (Karl et al., 2008). And finally, in D. melanogaster, four candidate genes on chromosome 2 underlie altitudinal clines in developmental time (Mensch et al., 2010), and mutations in the cis regulatory elements of the ebony locus underlie altitudinal pigmentation clines (Rebeiz et al., 2009).

Chromosomal inversion polymorphisms have also been repeatedly found to show altitudinal clines in flies, where the presence of large polytene chromosomes has made chromosomal rearrangements more amenable to study (Table 2). Several lines of evidence suggest that climatic variables may play an important role in maintaining spatial patterns in the frequency of particular inversion genotypes. For instance, inversion polymorphisms in Drosophila subobscura show similar latitudinal clines on three continents (Balanyá et al., 2006). The position of the clines has shifted in recent decades, probably due to rising global temperatures (Balanyá et al., 2006), and similar temporal changes have been observed for a D. melanogaster inversion polymorphism in Australia (Umina et al., 2005). Some inversion polymorphisms also show consistent clines with altitude and latitude (Etges et al., 2006 and references therein) or recurrent seasonal fluctuations (Dobzhansky, 1943). Finally, several studies have demonstrated a link between particular inversion polymorphisms and resistance to extreme temperatures (reviewed in Hoffmann et al., 2004).

Six studies have also screened panels of anonymous markers (e.g. AFLPs, microsatellite loci; Table 2) and identified loci showing patterns consistent with divergent selection along altitudinal transects (i.e. elevated differentiation and/or genotype-altitude associations). In the four AFLP-based studies, between 0.8% and 8.8% of all polymorphic loci showed evidence of adaptive differentiation (Table 2). It should be noted, however, that these studies used varying approaches for identifying outliers. Furthermore, rather than being the actual targets of selection, the anonymous markers detected using these approaches are more likely to be in linkage disequilibrium with unknown divergently selected loci.

Synthesis

Pervasive evidence for genetically based phenotypic differentiation

Our survey of the literature provides evidence for genetically based phenotypic divergence along altitudinal gradients for a wide range of species and traits, including wing size, cold tolerance, mass and development time (Fig. 1). This finding suggests that phenotypic divergence between populations is not rare, even if its prevalence may be overestimated in our data set due to a possible bias towards publishing significant results. Our analyses also show that genetically based phenotypic differentiation is taxonomically widespread, with some significant differences between populations being detected in all the groups studied.

Furthermore, in some cases, significant phenotypic divergence occurred at local geographical scales (Fig. 2). For example, eight studies detected significant divergence between populations separated by ten kilometres or less (Fig. 2, inset), which in several cases was well within the dispersal range of the species. In Anolis lizards, the morphological divergence was larger than neutral genetic divergence (FST < QST; Eales et al., 2010) and, similarly, Sarup et al. (2009) detected significant phenotypic divergence between Drosophila buzzatii and D. simulans populations that were undifferentiated at neutral molecular markers.

Are the observed differences adaptively relevant?

The finding that phenotypic differentiation can be maintained – at least sometimes – in the face of gene flow strongly suggests that some between-population differences are maintained by strong divergent natural selection. And this conclusion is supported by additional lines of evidence from both the phenotypic and molecular data sets. First, several phenotypic traits showed consistent clines in multiple species in the direction predicted from known environmental gradients, and secondly, many of the molecular studies identify loci showing signatures of divergent selection (Table 2).

We can also predict that the level of genetic differentiation at loci with a putative role in altitudinal adaptation should increase with the altitudinal distance between sites, assuming that the latter provides a rough proxy for the intensity of divergent selection. Neutral loci, on the other hand, should not show such an association unless gene flow between different altitudinal environments is reduced across the entire genome, for example due to dispersal barriers or immigrant inviability becoming stronger as the altitudinal contrast increases. Consistent with these predictions, we found that FST increased with maximum altitudinal distance at candidate loci, but not at neutral loci (Fig. 3). However, these analyses were based on a subset of only 15 studies, and the data set did not lend itself to statistical analysis. First, altitudinal distance in these data was highly correlated with geographical distance (Pearson correlation coefficient 0.77), making it impossible to distinguish between the effects of the two variables. Secondly, the representation of different taxonomic groups was very uneven across altitudinal distance classes (e.g. all studies at > 4000 m are from birds). Despite these limitations, it will be interesting for future studies to investigate whether genetic differentiation increases with altitudinal distance, and whether this increase is genome-wide or limited to genomic regions with a direct role in altitudinal adaptation.

image

Figure 3. Left panel: Genetic differentiation between animal populations (FST) increases with the maximum altitudinal distance between sampling sites at candidate loci (black diamonds), but not at neutral loci (grey squares). Each point represents a study, and in the majority of cases (12 of 15 studies), the type of molecular marker was the same for candidate and neutral loci. Note that we did not assess the statistical significance of the observed patterns because of limitations of the data set. First, different taxonomic groups were unevenly represented in the different altitudinal distance classes. Secondly, altitudinal distance was highly correlated with geographical distance making it impossible to disentangle the effects of the two variables: there was no longer an association between altitudinal distance and FST after removing the effect of geographical distance (right panel: residuals from a linear regression of FST against geographical distance, plotted against altitudinal distance).

Download figure to PowerPoint

Most of the phenotypic traits discussed above were specifically selected by researchers because their adaptive relevance seemed likely and, for some of the traits, clear expectations as to how they should respond to the environmental change associated with altitude could be formulated. As discussed above, a number of traits showed patterns consistent with these predictions (Table 1). However, in all cases, some species did not conform to our expectations, for example showing no significant between-population differences or differences that were unassociated with altitude. There are many possible explanations for these conflicting results. First, a given phenotypic trait may simply not be relevant for fitness in populations that are diverging due to random drift. Alternatively, if trait differences are adaptively relevant, the link between a phenotypic trait and fitness in a given altitudinal environment may have been misjudged; for example, the optimal phenotype may be different from what we expect. Also, selection pressures may not change consistently along an altitudinal gradient. For instance, small-scale topographical features influence ambient temperatures and can produce local patterns that oppose large-scale gradients (Scherrer & Körner, 2011). Finally, the trait mean in a population can deviate from the local optimum due to, for example, indirect selection resulting from genetic correlations with other traits, lack of additive genetic variation or immigration from other environments (Lenormand, 2002; Hoffmann & Willi, 2008).

Open questions and directions for future research

Are phenotypic differences between populations adaptively relevant and how does mean population fitness change along altitudinal gradients?

Although the available studies provide some evidence of adaptive differences between populations, explicit tests of local adaptation along altitudinal gradients, and the ecological relevance of the observed interpopulation differences, are clearly needed. Reciprocal transplant experiments along altitudinal gradients would be particularly valuable in this context, although we are aware of only three animal species for which such studies have been performed. All of these found that some trait differences persisted also in common natural environments (body size of frogs: Berven, 1982a; age and size at first reproduction in frogs: Berven, 1982b; flight activity of butterflies: Ellers & Boggs, 2004b; size and growth rate in lizards: Iraeta et al., 2006), but none actually demonstrated that fitness (or any fitness proxy) was indeed higher for local than nonlocal individuals.

More thorough studies of local adaptation will also provide insights into how the relative and absolute fitness of populations change along altitudinal gradients, which is largely unknown. If most populations are indeed adapted to their local environment, we might expect little variation in fitness along the gradient; however, most species have a restricted altitudinal distribution, suggesting that there must be limits to adaptation (e.g. Bridle & Vines, 2007). It is also possible that the environment imposes constraints on the maximum fitness that cannot be overcome by adaptation. For instance, fundamental thermodynamic constraints may lead to lower population growth rates in cold-adapted than warm-adapted species, even when both are tested at their thermal optimum (Frazier et al., 2006).

Unfortunately, in many cases, it will remain difficult to perform well-designed experiments that speak directly to the extent and patterns of local adaptation along altitudinal gradients, as well as to the link between particular phenotypes and fitness in a given environment. These are not easy questions to address, even in species that are experimentally tractable, and their study becomes particularly problematic in animals that cannot be reliably followed through time. However, recent methodological advances have opened up exciting new opportunities to investigate potential adaptation in a more diverse array of species using molecular approaches (see Box 1).

Box 1. The promise of ecological genomics for testing the genetic basis of altitudinal adaptation

At present, genome scans and outlier locus detection are a commonly used approach to detect signals consistent with the action of divergent selection between populations (Schoville et al., 2012). Perhaps the most serious limitation of this approach is that – by definition – it detects loci showing elevated between-population differences. Adaptive divergence, however, does not always involve large allele frequency changes, especially for quantitative traits which can be influenced by many loci and where interactions between loci can be more important than additive effects (e.g. McKay & Latta, 2002; Le Corre & Kremer, 2012). A second hurdle in nonmodel organisms is that the actual target of selection will (mostly) not be the outlier locus itself but rather a locus linked to it. Without a well-annotated reference genome, it will be difficult to identify nearby candidate genes of known function. Still, the detection of outlier loci, even if they remain anonymous, may provide a relatively cost-effective and tractable way to gauge the extent of putatively adaptive differentiation between populations that may be relevant for designing conservation and management strategies (but see Allendorf et al., 2010 for additional limitations and caveats).

In recent years, genome-scale analyses have become increasingly possible also in nonmodel species. Using next-generation sequencing of reduced representation libraries (e.g. restriction site associated DNA; Baird et al., 2008), for example, tens of thousands of single nucleotide polymorphisms (SNPs) can be identified and genotyped at moderate cost and without the need for a reference genome (Stapley et al., 2010). These data offer promising new opportunities to investigate the genetic basis of particular phenotypic traits, for example, using association mapping in natural populations without known pedigrees (Slate et al., 2010; Stapley et al., 2010). Once adaptively relevant variation has been identified, the frequency of particular variants can be monitored in space and/or time and related to environmental changes. Ideally, such surveys should be replicated to distinguish between general and local effects and to follow the fate of genetic variants in different genomic backgrounds. Here, altitudinal gradients may be particularly valuable because similar gradients are replicated across the globe. Strong barriers to dispersal may exist also within a mountain range, subdividing species into units that follow largely independent evolutionary trajectories (e.g. aquatic organisms in different drainages; Keller et al., 2012).

A second feature of altitudinal gradients, namely the small spatial scale at which environmental changes occur, makes them particularly suited to investigate whether specific genomic architectures are overrepresented in cases where adaptive divergence occurs in the face of gene flow. The chromosomal location of loci involved in divergent adaptation can most easily be studied if a reference genome from a closely related species is available, but genetic maps can also be constructed by following the segregation of variants in a pedigree (Slate et al., 2010; Stapley et al., 2010). Of particular interest might be a comparison of the genomic architectures underlying adaptation along altitudinal vs. latitudinal gradients. The two types of gradients share some similarities with respect to the observed environmental transitions (e.g. temperature), but these changes occur across much larger spatial scales with latitude and divergence may consequently be less constrained by gene flow.

What is the evolutionary potential of populations along altitudinal gradients?

The available evidence suggests that there is some adaptively relevant genetic divergence between populations and implies that adaptation has occurred in the past. Whether adaptive change will be possible in the future will depend upon the availability of relevant genetic diversity within populations and the rate at which environmental conditions change (Bridle et al., 2008). Evidence from the molecular studies surveyed here suggests that populations may in fact often be genetically variable at loci with a potential role in adaptation to different environments; thus, the loci identified as outliers (Table 2) – whereas showing large allele frequency differences between populations – are rarely fixed for alternative alleles. Often the alleles thought to be advantageous at low altitudes are observed at lower frequencies also in high-altitude populations and vice versa (data not shown). Similarly, for the phenotypic data, an average coefficient of variation (CV) of 10.7 was estimated across 651 observations for which this calculation was possible, suggesting some variation between individuals of a population reared in the same environment.

Novel genetic variation can be introduced into a population not only through mutation but, perhaps more relevant for rapid adaptation (e.g. Abbott et al., 2013), also through gene flow. Altitudinal gradients tend to be steep relative to the dispersal distance of organisms, which means that immigrants will often come from different, but nearby environments. In such situations, the selection coefficients of variants in the different environments and the rate and symmetry of gene flow will determine whether between-population differences are maintained or lost (Lenormand, 2002). Although gene flow can hinder local adaptation by eroding allele frequency differences, it can sometimes also facilitate it by introducing novel and potentially beneficial variants (Garant et al., 2007). Gene flow from lower towards higher altitudes, for example, could introduce genetic variants that have been ‘pretested’ under warmer conditions. Consequently, if conditions at high-altitude sites indeed tend to become more similar to current low-altitude conditions under global warming, we might predict that contemporary gene flow is usually asymmetrical, occurring mainly from low into high-altitude populations. The symmetry of gene flow has been assessed along latitudinal gradients (Paul et al., 2011; Fedorka et al., 2012), but we are unaware of similar studies along altitude. A mark–recapture study in a butterfly, however, found that dispersal was indeed more common from low- to high-altitude populations, probably because, at higher altitudes, host plants became available later in the season (Peterson, 1997).

Does the extent of adaptive differentiation vary between species and, if so, what factors underlie these differences?

The available data show that genetic differences of potential adaptive relevance exist in a wide range of species, including highly mobile groups such as birds where population divergence is potentially maintained in the face of extensive gene flow. Still, it is important to keep in mind that the species covered in this review are probably a nonrepresentative sample and that the extent of intraspecific adaptive diversity may be low in many other species. This may be particularly true for species with narrow altitudinal distributions where opportunities for adaptive divergence might be more limited. To understand why local adaptations evolve in some cases, but not in others, it will be critical to study a diverse array of species with both narrow and broad altitudinal distributions. Of particular interest will be whether some species are somehow predisposed to evolving and maintaining adaptive differences between populations. Such a predisposition could involve a genomic architecture where adaptive traits are shaped by few loci of large effect and/or clusters of multiple loci in tight physical linkage, which is expected to facilitate adaptation in the face of gene flow (e.g. along altitudinal gradients; Yeaman & Whitlock, 2011).

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature survey
  5. Conclusions
  6. Acknowledgments
  7. References
  8. Supporting Information

Our literature survey on local adaptation to altitude detected extensive phenotypic and genetic diversity among animal populations sampled along altitudinal gradients, with several lines of evidence suggesting that these differences were, in part, adaptively relevant. Although these conclusions are based upon rather limited data, we remain convinced that altitudinal gradients provide very suitable model systems for investigating local adaptation, albeit systems that have not yet been used to their full advantage. Furthermore, we anticipate that methodological advances will enable future studies to address these phenomena in species that are not easily tractable experimentally. In the meantime, however, it seems prudent to assume that most populations show some adaptive differentiation along altitudinal gradients, sometimes at very local scales, and that these adaptively relevant differences should be considered in conservation and management efforts.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature survey
  5. Conclusions
  6. Acknowledgments
  7. References
  8. Supporting Information

This study was carried out in the framework of GeneReach, a project lead by J. Bolliger (WSL) and funded by the Competence Center Environment and Sustainability, ETH Zürich, Switzerland. IK would like to thank O. Seehausen, M. Haesler, K. Lucek, D. Marques and J. Meier for support and helpful discussion.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature survey
  5. Conclusions
  6. Acknowledgments
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature survey
  5. Conclusions
  6. Acknowledgments
  7. References
  8. Supporting Information
FilenameFormatSizeDescription
jeb12255-sup-0001-FigS1_TabS1.docxWord document686K

Figure S1 Genetically based phenotypic changes along altitudinal gradients for (a) heat tolerance, (b) heat-shock protein expression, (c) desiccation tolerance, (d) body length, (e) growth rate, (f) longevity, (g) viability, and (h) fecundity.

Table S1 List of publications included in the review.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.