Towards genetic markers in animal populations as biomonitors for human-induced environmental change




Genetic markers provide potentially sensitive indicators of changes in environmental conditions because the genetic constitution of populations is normally altered well before populations become extinct. Genetic indicators in populations include overall genetic diversity, genetic changes in traits measured at the phenotypic level, and evolution at specific loci under selection. While overall genetic diversity has rarely been successfully related to environmental conditions, genetically based changes in traits have now been linked to the presence of toxins and both local and global temperature shifts. Candidate loci for monitoring stressors are emerging from information on how specific genes influence traits, and from screens of random loci across environmental gradients. Drosophila research suggests that chromosomal regions under recent intense selection can be identified from patterns of molecular variation and a high frequency of transposable element insertions. Allele frequency changes at candidate loci have been linked to pesticides, pollutants and climate change. Nevertheless, there are challenges in interpreting allele frequencies in populations, particularly when a large number of loci control a trait and when interactions between alleles influence trait expression. To meet these challenges, population samples should be collected for longitudinal studies, and experimental programmes should be undertaken to link variation at candidate genes to ecological processes.

Introduction: rapid adaptation and genetic markers

Evolutionary changes in animal populations can be rapid, particularly as a consequence of human-induced environmental disturbances (Hoffmann & Parsons 1997; Kinnison & Hendry 2001). These include the evolution of resistance to chemicals applied to control weeds, pests and diseases, adaptation in plants and invertebrates to heavy metals, adaptation to chemical and thermal emissions from factories, responses to salinity, evolutionary responses to overfishing, and adaptation to global temperature changes (Hoffmann & Parsons 1997; Kinnison & Hendry 2001; Stockwell et al. 2003).

Large phenotypic shifts can evolve in populations over a short time period. For example, large and rapid evolutionary changes are evident from population responses to pollutants and chemical stresses. Strains of insects resistant and susceptible to pesticides can differ in the lethal dose by several hundred-fold (McKenzie 1996), and the survival of adapted invertebrate populations can be unaffected by contaminants even when all non-adapted populations die (Klerks & Levinton 1989). Evolved life-history responses to climatic stresses can also translate into large-fitness differences between populations (Mitrovski & Hoffmann 2001; Bradshaw et al. 2004). A measure of the rate of evolution in populations is the haldane, representing standard deviations of change per generation and defined as [(x2/sp) − (x1/sp)]/g, where sp is the pooled standard deviation of trait values of the population (x) and g is the number of generations since separation (Hendry & Kinnison 1999). Whilst rates of evolutionary change correspond to a median of 0.0058 haldanes, rates can exceed 0.20 haldanes over short periods (Kinnison & Hendry 2001). Therefore, traits have the potential to evolve several phenotypic standard deviations over a few generations.

Underlying these evolutionary changes are shifts in allele frequencies at loci. Allele frequency changes have long been considered as having potential for monitoring environmental stress (Luoma 1977). As populations adapt, alleles that are initially rare in populations increase to a high frequency. Studies of insecticide resistance in particular show how the evolution of resistance in response to pesticides in the environment involves allelic changes at a limited number of loci (Raymond et al. 2001). Although insecticide-resistance research has focused on understanding the mechanisms of resistance and on the development of strategies to minimize its evolution, it has also produced markers for monitoring resistance development in populations, as in the case of resistance to a pyrethroid insecticide in field populations of the Colorado potato beetle (Kim et al. 2005) and to Bacillus thuringiensis (Bt) toxins in the pink bollworm (Morin et al. 2004).

Genes involved in rapid evolution to a range of environmental stresses are likely to be identified over the next few years. Whereas previously researchers have focused on testing whether an evolutionary response had a genetic basis, they have now moved to dissecting responses at the physiological and molecular levels. By analogy, genetic research in the health area a few decades ago centred on the issue of whether illnesses such as heart disease or breast cancer had a genetic basis. Large quantitative genetic analyses were completed to test if a genetic basis was present. Now research has moved on to identifying specific marker genes that underlie these genetic variants (Kleyn & Vesell 1998). Banks of markers are emerging that can be used in the treatment of human diseases, indicating the likely vulnerability of individuals to different diseases and aiding in drug applications (Hiratsuka et al. 2006). An understanding of this genetic variability among individuals is providing a revolution in healthcare delivery, and a similar breakthrough seems possible in the environmental area.

There is abundant evidence that genetic changes occur in response to anthropogenic stresses, and that these can be rapid and involve a variety of physiological, morphological and life-history traits. Can evolutionary changes and the genes underlying them be used to understand and detect environmental changes?

Linking evolution to environmental change – the phenotypic level

Evolutionary shifts are traditionally detected at the phenotypic level. Even the evolution of chemical resistance in pests is normally first detected by screening organisms for their level of susceptibility. Typically a sensitive strain provides a baseline and resistance is identified by comparing field individuals or their offspring to this strain (e.g. Shelton et al. 1993). Resistance can also occur in organisms that are not the targets of insecticide applications, suggesting that the particular chemical is having non-specific biological effects. For instance, Drosophila melanogaster has evolved resistance to several chemicals despite never being targeted directly as a pest species (Wilson 2005). The presence of resistance may be indicative of population-level effects of the chemical. Phenotypic studies indicate that resistance has also developed in animal populations to other chemical changes including heavy metals (Klerks & Levinton 1989), polychlorinated biphenyl (PCB) (Nacci et al. 2002) and acidification (Meriläet al. 2004).

Evolutionary shifts have also been detected in response to thermal changes in the environment. Populations of copepods living in warm water produced by power plants are locally adapted to thermal stress (Bradley 1978) and there is evidence of local adaptation to thermal stress in Drosophila populations (Hoffmann et al. 2003). On a wider geographical scale, Bradshaw & Holzapfel (2001) and Bradshaw et al. (2004) have documented evolution in diapause induction in the pitcher plant mosquito, Wyeomyia smithii, as a consequence of warmer conditions over the last few decades. Diapause induction shows a latitudinal cline along the east coast of the USA, with southern populations from warmer areas entering a diapause at shorter daylengths than northern populations from colder regions. This allows southern populations to take advantage of the extended favourable conditions in their environment. The cline has shifted towards a more southern shorter daylength form over the last 30 years, particularly in populations at the northern end of the distribution of this species (Bradshaw & Holzapfel 2001). This clinal shift is likely to reflect the large-fitness benefit arising from entering diapause at the appropriate stage. For instance, when a southern population was exposed to mid-latitude daylengths, the southern populations entered diapause too late and suffered a 74% decline in fitness, while the northern populations entered diapause too early and experienced an 88% decline (Bradshaw et al. 2004). The rapid shift in this cline suggests strong selection allowing populations to take advantage of changing climatic conditions (Bradshaw & Holzapfel 2001).

Body size evolution has been related to climate change in several organisms. Smith et al. (1998) followed changes in pellet size (a measure of body size) in middens of woodrats over a period of 25 000 years. Pellet size decreased as temperatures increased during a series of temperature fluctuations. Size changes have now also been linked to climate change over a period of decades. Yom-Tov et al. (2006) tested the prediction that global warming was associated with decreases in body weight in 14 species of passerine birds at two localities in England: they found linear decreases in size in four species, nonlinear decreases in two other species, but an increase in size in one species that may have reflected changes in food supply. Therefore, predictions were met in the majority of species, although these were phenotypic changes and a genetic basis for the changes can be inferred only indirectly.

Genetic changes in size and other characteristics have been linked to recent climate change in Darwin's finches from the Galapagos. By estimating genetic parameters in finch populations, including interactions among traits, Grant & Grant (1995) accurately predicted changes in body size and beak characteristics as a consequence of a shifting food supply during drought conditions. As the incidence of drought events increased, selection was expected to favour a decrease in body size and blunter beaks. These shifts were documented over 30 years in one species of finch, although they were not found in a second species (Grant & Grant 2002).

One problem in predicting changes in body size due to shifts in environmental conditions is that size is influenced by other selective factors, including competitive interactions, mating success and resource acquisition. Most animals follow Bergmann's rule, where the size increases away from the equator. In mammals and other warm-blooded animals, this rule is thought to be associated with thermoregulation of internal body temperature, because of changes in surface/volume ratio. However, there are exceptions to Bergmann's rule. For instance, body size in Alaskan shrews shows the opposite pattern to Bergmann's rule, with size increasing with latitude (Yom-Tov & Yom-Tov 2005). Moreover, size has increased in these shrews in the last half of the 20th century, despite temperatures increasing. This pattern may be related to food availability; as more food becomes available, the shrews are likely to have sufficient resources to develop a larger body size. While body size changes can therefore occur rapidly in populations, they represent an indirect response to climate change.

Changes in morphological traits other than size might signal climate change. In barn swallows, Møller & Szep (2005) have found changes in the length of the outermost tail feathers that provide a mating advantage to males. In male birds from a Danish population length had increased more than 1 SD over a 20-year period. This increase was associated with selection on survival and breeding date of the birds. The swallows in Denmark originate from Algeria, where deterioration in the vegetation has occurred as a result of unfavourable climatic conditions. Perhaps, males with relatively longer outer tail feathers have an advantage under unfavourable conditions for vegetation growth, although this connection remains to be established.

In several species of birds migratory behaviour seems to have evolved due to climate change. The heritability of traits associated with migration in several bird species is high (Pulido & Berthold 2004), reflecting the potential for rapid evolution. In blackcaps, common garden experiments indicate that there has been a heritable decrease in migratory activity over 13 years, reflecting a change of more than 1 SD (Pulido & Berthold 2004). Changes in the spring arrival dates of barn swallows also appear to have evolved suggesting a shift in migration patterns (Møller & Merilä 2004).

Finally, rapid evolutionary changes can occur because of direct human biotic effects. Evolution can take place when humans act as predators and exploit resources. For instance, Atlantic cod off Newfoundland and Labrador declined precipitously in the 1970s and collapsed by the 1990s. Exploitation of mature fish was expected to select for early maturation, and Olsen et al. (2005) found evidence for an evolved phenotypic change in maturation time. They showed that reaction norms in populations had shifted towards maturation at earlier ages and at smaller sizes at the time the fishery collapsed. This shift to early maturation was likely to have been a consequence of selection imposed by fishing, particularly as maturation rate in fish has a genetic basis (Conover et al. 2005). Human activities can also indirectly generate evolution by altering natural environments, such as through the introduction of predators that in turn select for changes in morphology and physiology of prey species (Phillips et al. 2004).

These examples illustrate that phenotypic studies can be used to detect rapid evolution in response to human activities, including the effects of environmental contaminants and climate change. However, trait changes are not necessarily easily related to environmental factors, as illustrated by problems in making links between body size and climate change. Genetic changes in populations can be inferred only when familial relationships are established in the field or controlled breeding experiments are undertaken in the laboratory, which is not possible for many groups. Nevertheless, longitudinal monitoring of traits at the phenotypic level might serve as a way of identifying the biological impact of some environmental stressors. We return to this issue below.

Linking evolution to environmental change – allele markers

The notion that genetic markers can be used to monitor environmental changes has been around for a number of years and several types of markers have been associated with environmental changes (Table 1). Early studies in aquatic environments compared frequencies of genes in polluted and unpolluted environments to isolate genetic markers underlying tolerance variation. Attempts to link changes in genes encoding enzyme polymorphisms (allozymes) with pollutants were successful in some cases (Nevo et al. 1978, 1984) but not in others. Various studies have also tested if overall levels of genetic diversity are reduced under pollution stress (genetic erosion) using randomly selected sets of markers. However, these have produced unequivocal results because factors other than selection influence overall genetic diversity (van Straalen & Timmermans 2002) and this approach is not considered further here.

Table 1.   Types of genetic markers available for investigating environmental change associations
AllozymesProtein products provide targets under selectionLimited number of loci, requires well-preserved tissue for protein isolationVariation in GPI (glucose phosphate isomerase) and mercury pollution in mosquitofish (Tatara et al. 2002)
Chromosome inversionsRapid responses to environmental change due to number of loci in disequilibrium with inversions, can be scored cytologically or by polymerase chain reactionOnly described for limited number of insects, different gene combinations may be involvedAssociations between inversion polymorphism and climatic warming in several Drosophila species (Umina et al. 2005; Etges et al. 2006)
MicrosatellitesLarge numbers of microsatellites can be identified from all speciesMarkers are often neutral, can only be related to selection through identified disequilibrium with other lociDrosophila surveys linking microsatellite markers with environmental gradients (Kennington et al. 2006)
Transposable elementsScattered throughout genome, can provide signatures of selection at genomic levelsDepends on elements being identified for loci under selectionAssociations between transposable elements and insecticide resistance (Daborn et al. 2002)
Other DNA polymorphismsEPICS (exon primed intron crossing), single nucleotide polymorph-isms, and others
Heterozygotes readily detected
No limit to number of markers that can be identifiedAssociations between markers and insecticide resistance (Morin et al. 2004)
Clonal organisms (can be distinguished by any of the above marker systems)Clones can be identified with any type of genetic marker, rapid changes in clonal composition expected due to strong selection on entire genomeDepends on clonal organisms being present across environmental gradientDaphnia clones associated with thermal conditions (Lopes et al. 2004)

The evolution of insecticide resistance provides the best-known examples where specific genetic changes have been linked to environmental toxins. Precise changes at the DNA and protein levels have been identified in several cases. In D. melanogaster, resistance to the once commonly used cyclodiene insecticide, dieldrin, is due to a single mutation in the Rdl gene (ffrench-Constant et al. 1993). Rdl encodes the molecular target of dieldrin, a receptor for the γ-aminobutyric acid (GABA) neurotransmitter (ffrench-Constant et al. 1991). The point mutation in Rdl results in an amino acid substitution in the GABA receptor, reducing the ability of dieldrin to bind to the receptor, disrupt neurotransmission, and thereby kill the insect. The same amino acid substitution has now been found in natural populations of several insect species that are resistant to dieldrin (Thompson et al. 1993). Another example is resistance to pyrethroid insecticides associated with the kdr gene that encodes a protein involved in movement of compounds across membranes. A particular mutation in kdr, leading to a replacement of the amino acid leucine with phenylalanine in this protein, has been isolated from a number of insect species that have developed resistance (Soderlund & Knipple 2003).

The highly predictable nature of insecticide-resistance evolution not only carries across species, but is also capable of being replicated in the laboratory. McKenzie & Batterham (1998) reviewed evidence that resistance to organophosphate in the sheep blowfly, Lucilia cuprina, involves allelic substitution at the Rop-1 locus. Rop-1 encodes a carboxylesterase enzyme, which is involved in the detoxification of organophosphates. The allelic substitution at Rop-1 leads to an amino acid change that alters the substrate specificity of the enzyme and effective detoxification (Newcomb et al. 1997). This change has been observed not only in the field, but has also been generated in the laboratory. Flies from a susceptible laboratory population of L. cuprina were exposed to chemicals that cause mutations (mutagenesis), and this was followed by selection for diazinon resistance, which led to the same resistance outcomes as seen in resistant field populations (McKenzie et al. 1992). Similar experiments in L. cuprina involving mutagenesis and selection have also generated resistance to dieldrin, with the resistance mechanism involving a mutation in Rdl that matched the change in Rdl in field populations discussed above (Smyth et al. 1992).

There is little information on the genetic basis of evolutionary responses to other types of environmental toxins, although an exception is the evolution of mercury resistance in mosquito fish. The time taken for fish to die under mercury exposure has been related to genotypes at a locus encoding the enzyme glucose phosphate isomerase. The distribution of allozyme genotypes at this locus has been related to mercury levels in a contaminated environment (Heagler et al. 1993). A series of studies which included long-term mesocosm experiments on polymorphic populations have also implicated the same locus in mercury responses (Tatara et al. 2002).

While these types of studies can identify genes under selection, there is an additional approach available to detect selection which focuses on molecular variation around the genes rather than directly on trait associations. This approach can be illustrated by the evolution of resistance to insecticides in D. melanogaster involving the gene Cyp6g1. This gene encodes one of the many types of enzymes known as cytochrome P450s that detoxify toxins in animals. Cyp6g1 is overexpressed in D. melanogaster lines from natural populations that exhibit resistance due to the insertion of a transposable element known as Accord into the promoter region that regulates expression of the gene (Daborn et al. 2002). Analysis of the DNA region around Cyp6g1 shows that there is a sharp reduction in the level of molecular variation in this region. This reduced variation is a molecular signature of the recent spread of the allele containing Accord in D. melanogaster populations (Catania et al. 2004) as outlined in Fig. 1. When the allele with the transposable element first arose, it would have existed within a particular region of DNA and associated with other alleles at adjacent loci. As the new allele was favoured by selection due to the overexpression of the detoxification mechanism, alleles at adjacent loci would also have spread along with it, decreasing genetic variation in this genomic region in the population (a ‘selective sweep’). This loss of variation is only gradually restored because of recombination and mutation. In the case of Cyp6g1, there is also evidence of a selective sweep around this gene in a related species, Drosophila simulans. However, in this case the sweep is associated with a different transposable element known as Doc, but also inserted in the promoter region of Cyp6g1 (Schlenke & Begun 2004). In both species these changes are likely to reflect a recent history of exposure to pesticides from agricultural activities. Patterns of molecular variation can therefore be used to recognize areas of the genome under recent selection around genes that are being selected because of particular environmental changes.

Figure 1.

 A beneficial mutation and the resulting selective sweep, as exemplified by the Drosophila melanogaster insecticide resistance locus Cyp6g1. (a) A beneficial mutation occurs in a particular genetic background, in this example the insertion of the Accord transposable element (triangle) upstream of Cyp6g1 leading to insecticide resistance via Cyp6g1 overexpression. (b) A selective sweep occurs when the Accord carrying allele of Cyp6g1 is selected. In individuals without the beneficial mutation neutral polymorphisms (circles) are evenly, but randomly distributed across the chromosomal region because of recombination and mutation. In lines carrying the Accord insertion mutation, selection of the beneficial mutation (in this example the Accord insertion) results in a reduction of neutral polymorphisms at sites closely linked to Cyp6g1. Both the presence of the Accord insertion and the reduction in polymorphisms are signatures of selection.

What about evolutionary responses to climatic variables? Several longitudinal studies indicate that genetic markers can be linked to climate change. One case involving allozyme markers (Table 1) is selection on the alcohol dehydrogenase (Adh) gene in D. melanogaster. There are two common alleles at the Adh locus, AdhF and AdhS. These alleles show a strong cline along the east coast of Australia, with the S allele at a relatively higher frequency in the tropics (Oakeshott et al. 1982). This Adh cline is also present on other continents around the world suggesting that the AdhS allele is favoured in tropical conditions rather than temperate conditions, consistent with the results of laboratory experiments (Oakeshott et al. 1982). Clinal variation in Adh in eastern Australia was first established based on collections from 1979, and an identical pattern was observed in flies collected a few years later (Anderson et al. 1987). However, when the cline was resampled in 2000 and 2002, a different pattern emerged (Umina et al. 2005). While the slope of the association between latitude and Adh had not changed over this time, there had been a marked shift in the intercept of the relationship. This corresponded to a shift of several 100 km in latitude. In contrast, there was no change in the latitudinal association of the Gpdh gene, which showed a much weaker clinal pattern.

Several chromosomal regions in D. melanogaster and many other Diptera may exist in two forms: a non-inverted (standard) form, and an inverted form where the arrangement of genes in the chromosomal region is reversed. Chromosomal inversions lead to a reduction in viable recombinants in the inverted region of DNA causing alleles of genes within the inversion to be inherited together (Schaeffer et al. 2003; Hoffmann et al. 2004; Kennington et al. 2006). Drosophila melanogaster has several of these inversions that are cosmopolitan and populations can be polymorphic for the standard or inverted region. Changes in latitudinal patterns over time have now been established for these inversion polymorphisms (Table 1). One of these cosmopolitan inversions is known as In(3R)P and is located on the right arm of chromosome 3. The In(3R)P inverted chromosome arrangement increases sharply in frequency with latitude in Australia and on other continents (Knibb 1983), from near fixation for the standard arrangement at high latitudes to near fixation of the inverted arrangement towards the equator. On the east coast of Australia, the cline in this inversion has changed sharply over the last 20 years (Umina et al. 2005). While the slope of the cline has not changed, the intercept has moved such that populations now have inversion frequencies resembling those of populations 700 km nearer the equator 20 years ago.

Levitan & Etges (2005) and Etges et al. (2006) have related changes in the frequency of inversion polymorphisms in Drosophila robusta to temperature shifts. These inversions have been studied in North American populations since the 1940s and have exhibited latitudinal and altitudinal clines over several decades. The clines have changed over the last few decades. Since the 1970s, the frequency of several inversion arrangements have increased, matching increases in minimum temperature at several sites (Levitan & Etges 2005). Along an altitudinal transect in the Smoky Mountains, the frequency of high-altitude inversion arrangements has increased from the 1940s to the 1980s, matching locally cooler temperatures, but these trends have reversed recently as temperatures have warmed again. Changes in inversion frequencies in D. robusta are therefore rapid and closely track temperature shifts.

Finally, the frequency of inversions in the O chromosome of Drosophila subobscura in Spain have changed over the last 15 years (Rodriguez-Trelles & Rodriguez 1998). Two of the chromosomal arrangements that normally increase in frequency in summer have become more common in populations. These changes are associated with temperature, which has increased at a rate of 0.081 °C per year since the mid-1970s. Chromosomal diversity has decreased at the same time (Rodriguez-Trelles & Rodriguez 1998). In southwestern Europe, latitudinal patterns in four arrangements have changed over 30 years; arrangements common in warmer southern areas are now relatively more common in all populations even though latitudinal patterns have been maintained (Sole et al. 2002).

The chromosomal arrangements of D. subobscura have also been studied in the Americas which were recently invaded by this species. Following invasion, latitudinal clines in inversions were rapidly established in populations and converged on clines in the Old World where the species originated (Prevosti et al. 1990). However, this convergence was not supported by later sampling (Balanya et al. 2003) which suggested that, while latitudinal patterns in the New and Old Worlds were similar, convergence stopped for many of the polymorphisms. Thus the inversions may have developed a different dynamic following invasion, perhaps dependent upon the genetic makeup of the invading population.

Genes within inverted chromosomal regions that are responsible for these patterns have not yet been isolated. As mentioned above, inversion polymorphisms lead to regions of association between alleles along the inverted region, particularly near points where the inverted region starts and finishes, but also extending away from these points (Schaeffer et al. 2003; Kennington et al. 2006). Therefore, they help to lock together a number of alleles, which singly or together might influence fitness under different climatic conditions. Interactions among alleles within inversions might influence climatic adaptation. Under the classical explanation of inversion polymorphisms advanced by Dobzhansky, inversions in a population help to lock up alleles that are co-adapted. However, alleles within an inversion that are coadapted in one population may not necessarily be coadapted in a different population. If this explanation is correct (and there is presently only limited evidence for it), then changes in latitudinal patterns of inversions might be difficult to interpret because the allele contents of the inversion varies with latitude.

What about climatic variation in other species? In the Sierra willow beetle, Chrysomela aeneicollis, genetic variation at the locus coding for the enzyme phosphoglucose isomerase has been related by Rank & Dahlhoff (2002) to climatic variation. This species experiences large daily fluctuations in temperature from −5 to >32 °C. One of the alleles (Pgi-1) increased by 11% from 1988 to 1996, while there were no changes in allele frequency at two other allozyme loci. This increase was related to cooler conditions at the study site prevailing prior to 1996. Physiological studies indicated that female beetles homozygous for the Pgi-1 allele survived a laboratory cold shock better than beetles with other genotypes, suggesting that the change in allele frequency was associated with shifting climatic conditions. In addition, transplant studies (McMillian et al. 2005) were used to demonstrate that the Pgi-1 allele was at a disadvantage in terms of survival and development time when beetles were moved to a warm location. These effects may be partly mediated through an effect of Pgi genotypes on expression of HSP70, one of the heat-shock proteins that protects cells against heat-shock damage.

There are numerous other polymorphisms that could form the basis of longitudinal studies in future attempts to link environmental shifts to genetic changes. For instance, in deermice there is a well-known altitudinal pattern for polymorphism in the α-chain of haemoglobin (Chappell & Snyder 1984) and this might change as conditions become warmer. There are also other polymorphisms that can be easily scored and could be worth considering, such as snail-banding polymorphisms that are affected by heat stress (Richardson 1984), and melanism patterns in ladybirds that have been related to climatic selection (Honek et al. 2005) and pollution (Brakefield & Liebert 2000). Any polymorphism that shows a clear latitudinal and/or altitudinal pattern could be re-examined to test for associations with climate change. Surveys of clinal markers (e.g. Sezgin et al. 2004; Kennington et al. 2006) can be used to identify genes or genomic regions for monitoring in longitudinal studies. In fact, any screen of markers across an environmental gradient could be used to develop candidates. Comparisons of expression patterns of genes across gradients using microarray technology (that allows changes in expression of genes throughout the genome to be characterized) provide an additional approach for isolating candidates (Whitehead & Crawford 2006).

A hazard in all this work is that any associations between markers and environmental changes could be spurious. Identification of candidate loci needs to encompass multiple lines of evidence. For instance, if an association between variation in a gene and an environmental change is established in one organism, the impact of mutants in the gene could be examined in that organism or a related model species to help establish a causal association. For this reason, the examples discussed above involve accumulated knowledge based on diverse data sources that include mutant studies, expression patterns, fitness experiments and monitoring genetic changes in experimental systems.

Finally, changes in the frequency of clonal lineages in parthenogenetic organisms may be useful for environmental monitoring. Selection on these lineages can be extremely strong as demonstrated by rapid shifts in the clonal composition of field populations of parthenogenetic populations over just a few generations (Weeks & Hoffmann 1998; Vorburger 2006). This reflects the fact that, in the absence of recombination, the entire genome of a clonal lineage is under selection. Once associations between clones and particular environmental conditions are established, changes in the clonal composition of populations could be used to monitor for these conditions. For instance, in Daphnia, clonal composition has been linked to seasonal variation in climate (Carvalho & Crisp 1987) and levels of acid pollution (Lopes et al. 2004).

Marker lists and marker signatures

The potential of markers as monitors is illustrated by their growing application in the area of insecticide resistance. Traditionally, insecticide-resistance bioassays have been used to detect the frequency of resistant individuals. Molecular markers offer the advantage of being able to detect heterozygous individuals, otherwise undetectable for recessive traits, and also allow higher numbers of field individuals to be assayed. Several markers have now been developed to monitor the frequency of resistance in natural populations. For example, in the pink bollworm, Pectinophora gossypiella, polymerase chain reaction (PCR)-based assays have been developed to detect mutant alleles which are involved in resistance to Bt toxins (Morin et al. 2004). These toxins are produced by B. thuringiensis and bind to receptor proteins including cadherin proteins. However, in resistant mutants, alleles in the BtR cadherin gene fail to produce a cadherin protein that would normally bind to one of the Bt toxins, producing resistance. PCR-based technologies capable of allele discrimination have also been adapted to detect specific point mutations in the kdr gene that lead to pyrethroid resistance in aphids (Anstead et al. 2004), horn fly (Li et al. 2003) and mosquitoes (Tripet et al. 2006), as well as the detection of the single mutation in Rdl that is associated with resistance to the chemicals fipronil and dieldrin in the cat flea (Daborn et al. 2004) and the German cockroach (Hansen et al. 2005).

One advantage of using genetic markers in this context is that they provide a method of early detection of resistance, so that management programmes can be implemented prior to a high level of resistance developing. The assays described above are extremely specific, monitoring specific resistance alleles of particular genes. For resistance markers to work in both the context of management and monitoring environmental change, all possible mechanisms of resistance need to be monitored. Assays would need to have the flexibility to detect new resistance-conferring mutations in the genes currently being monitored, and to detect new types of resistance mechanisms as these evolve. The increasing understanding of potential resistance mechanisms makes it more likely that reliable markers for resistance detection can be developed.

The molecular marker approach can be applied to other toxin responses and traits as information accumulates on genes that may influence variation in particular traits (‘candidate’ genes). For some types of stresses candidate genes are already emerging, even when organisms are not well-known genetically. In earthworms, the mechanisms underlying resistance to cadmium pollution have been largely identified, as reviewed in Sturzenbaum et al. (2004). Metallothionein proteins are used to bind and compartmentalize the cadmium, and these have been isolated and the genes controlling them have been identified. Metallothionein expression has also been linked to cadmium tolerance in springtails (Timmermans et al. 2005), although in this case additional mechanisms of cadmium resistance are likely to be involved. Where genes in organisms of interest to a researcher are not known information from model organisms can be used to derive candidates. For instance, heavy metal resistance in D. melanogaster has been associated with metallothioneins, based on molecular studies of lines and natural variants with different levels of heavy metal resistance (Maroni et al. 1987). Drosophila is already used to detect candidates for insecticide resistance based on the knowledge that the identified genes are good candidates for resistance to the same stresses in non-model organisms (Wilson 2005).

For complex traits there is rapid progress in the identification of genes underlying natural variation. Currently, there is a large focus on understanding the genetic basis of variation in a range of traits by using model organisms such as mice, Drosophila and Arabidopsis. These efforts include a focus on traits likely to be of interest for environmental monitoring. In D. melanogaster a great deal of progress has been made in understanding the genetic basis of responses to stresses such as temperature extremes (Hoffmann et al. 2003; Norry et al. 2004; Morgan & Mackay 2006) and starvation (Harbison et al. 2005). Selection experiments indicate that different sets of genes tend to be involved in adaptive responses to stresses like desiccation, heat and cold resistance (Bubliy & Loeschcke 2005). Traits like heat resistance can be broken down further into genetically independent set of traits (Hoffmann et al. 1997; Bubliy & Loeschcke 2005). Mapping and tests on candidate genes have yielded lists of candidate genes likely to underlie these independent traits. For instance, phenotypic analysis of laboratory lines has indicated that high-temperature resistance in larvae is associated with levels of one of the heat-shock proteins, HSP70 (Sorensen et al. 2003). Natural variation in HSP70 levels have also been linked to variation in heat resistance within populations and across thermal gradients (Sorensen et al. 2003). However, while HSP70 levels and induction are involved in resistance, the exact nature of genetic variation underlying changes in these levels is not understood. Crosses between strains have been used to map areas of the genome where loci influencing heat resistance [quantitative trait loci (QTL)] are located. This has indicated that variation in heat resistance to a knockdown stress maps to QTL around Hsp70 (Norry et al. 2004), but variation in the Hsp70 gene itself has not been convincingly linked to heat resistance (Weeks et al. 2002). Eventually, a combination of mapping, candidate gene identification and correlated responses in selection lines, as well as genetic dissection of latitudinal patterns, can help isolate the contribution of specific genes to genetic variation in thermal resistance.

Identifying suitable candidates will be aided by molecular analysis that determines if genes have been exposed to recent selective sweeps (Fig. 1). Because recombination and genetic drift remove evidence for selective sweeps relatively rapidly (Przeworski 2002), selective sweeps can provide evidence of recent positive selection on genes, as in the case of Cyp6g1 (Catania et al. 2004) and Pgm (Verrelli & Eanes 2000). Hsp70 genes also show evidence of selective sweeps in populations (Bettencourt & Feder 2002). More examples are emerging as variation in and surrounding candidate genes are examined further (Pool et al. 2006).

As well as being useful for verifying candidates under selection, monitoring for areas of the genome experiencing selective sweeps can itself detect genetic change, without a priori knowledge of candidate genes. The presence of a low level of variation can indicate that selection has taken place on a particular allele of a gene recently. Monitoring the intensity of the sweep over time can be a means of assessing the intensity of selective forces acting in the population. Beisswanger et al. (2006) scanned for selective sweeps to identify candidates of selection in a European population of D. melanogaster. They initially found regions with decreased variation and then undertook finer level analyses to isolate small regions of the genome with particularly low variation. One of these regions, on the X chromosome, completely lacked variation. One of the seven genes in this region is likely to be responsible for this sweep, although the mechanism remains to be identified. In another selective sweep study on the X chromosome of D. melanogaster, a single gene with unknown function was identified to have been under selection (DuMont & Aquadro 2005).

Both mutations directly influencing gene function and mutations affecting gene expression can potentially affect traits involved in environmental responses. Most of these are single base mutations and therefore difficult to detect. However, one group of mutations that is easier to detect involves transposable element insertions, discussed above in the context of the Cyp6g1 gene (Fig. 1). Transposable elements play an important role in driving and shaping genome evolution (Kazazian 2004). They can cause mutation by inserting into the coding regions of genes and disrupting or altering gene function, or by inserting into the regulatory regions of genes and disrupting or altering gene expression. They often contain DNA sequences that can alter gene expression (Marino-Ramirez et al. 2005). Some transposable element insertions increase to a high frequency in populations because the mutant phenotype provides a selective advantage. Detecting these events and the genes under positive selection can be as simple as detecting transposable element insertions present at high frequencies in populations. For example, in many populations of D. melanogaster a Doc element insertion into the CHKov1 gene has been observed. This insertion results in a truncated form of the protein and is under recent strong selection, sweeping to high frequencies in D. melanogaster populations (Aminetzach et al. 2005). A Bari-1 transposable element inserted in the Cyp12a4 gene also increases expression of this gene. This insertion allele is fixed in natural D. melanogaster populations (Marsano et al. 2005). Monitoring the frequency of transposable element insertions throughout the entire genome may be a feasible approach in the future to identify parts of the genome that are involved in adapting to environmental changes (Franchini et al. 2004), although detailed follow-up work will be required to establish causal links.

Genetic markers: potential or lost cause?

Genetic markers have two advantages as environmental monitors. First, gene frequency shifts are likely to occur well before population extinction. As the majority of traits are thought to be heritable, there is the potential for evolution in response to almost all types of environmental changes. Allele frequencies at loci that underlie directional trait shifts can be monitored as long as enough is known about the underlying genetic basis of trait variation. Second, gene frequency shifts are likely to be highly specific depending on the nature of the environmental variable driving selection. This means that the genetic changes can indicate a particular type of environmental change.

Despite these advantages, there has been limited use of genetic markers to monitor environmental changes, even in ecotoxicology. One reason is that adaptive shifts may not always occur when an environment changes. Klerks & Weis (1987) reviewed the early literature and found numerous cases where animals appeared to show physiological differences between polluted and unpolluted sites, but often it was difficult to distinguish between acclimation and genetic change. In cases where genetic factors could be isolated, it appears that many species did not successfully adapt to pollutants like heavy metals (Klerks 2002). Animals may also fail to adapt to climate change. The most common response of bird populations to climate change involves a shift in breeding date; however, while there is heritable variation for breeding date in bird populations, there is no evidence that this trait has yet evolved, and instead shifts appear to be environmentally based (Pulido & Berthold 2004). If genetic approaches are to be successful, they need to focus on organisms likely to adapt and traits with a high evolvability.

Another reason is that adaptation can involve subtle changes that are difficult to detect and characterize. Adaptation to pollutants can involve life-history changes rather than physiological tolerance (Postma et al. 1995). These can be more difficult to detect than tolerance, requiring comparisons of an organism's development and reproductive allocation. Adaptation to climatic change can also involve life-history shifts, as is the case of clinal variation in Drosophila (Mitrovski & Hoffmann 2001) and latitudinal adaptation in mosquitoes (Bradshaw et al. 2004).

The genetic basis underlying the adaptive shift can be difficult to identify even when there are candidate genes. In the mummichog, Fundulus heteroclitus, an estuarine fish found along the east coast of the USA, several populations have evolved resistance to pollutants including PCBs and the resistance mechanism acts toxicologically through the aryl-hydrocarbon receptor (reviewed in Nacci et al. 2002). However, attempts to isolate genes underlying resistance have proved difficult, despite a considerable effort focusing on a variety of candidates and some positive associations being detected (Roark et al. 2005).

The task of isolating genes underlying an adaptive response can be onerous. Genomic regions identified via mapping of traits may depend on the exact nature of environmental conditions being tested, as well as sex, and the nature of the trait itself (Mackay 2004). When genomic regions are identified, it can be difficult to isolate specific genes because the phenotype depends on interactions among the genes. However, very rapid progress is now being made in identifying genes and pathways underlying traits. An increased understanding of the relationship between the physiological makeup of organisms, and the underlying genetics, will enhance our ability to select candidate genes, leading to numerous targets that can indicate the effects of selection.

Yet another reason why genetic markers have had limited use to date is that the same selection targets might be reached by multiple pathways, making it difficult to predict the specific genes involved in selection responses across populations and species. In the case of insecticide resistance, there is clearly a high degree of predictability across populations and even across unrelated species. Whether selection responses in complex traits involve predictable sets of genes remains to be seen. However, there is evidence from bacterial studies that selection to thermal extremes involves predictable genetic changes (Counago et al. 2006). In animal populations, independent selection responses often map to the same genomic regions. For instance, for heat resistance in D. melanogaster, two mapping exercises have been undertaken so far and produced consistent results. One of these (Norry et al. 2004) involved crosses between heat selected and control lines, while the other (Morgan & Mackay 2006) was based on crosses between two inbred lines that differed in heat resistance. Despite this difference in the genetic background of the stocks, there was a high degree of overlap in the areas of the genome identified as controlling heat resistance (Morgan & Mackay 2006). Both studies also identified regions where genes known to influence variation in heat resistance were located. A high degree of overlap is also evident in the mapping of size genes from the ends of different D. melanogaster clines: size genes have been localized to the right arm of chromosome 3 in mapping experiments involving lines from Australia and South America (Calboli et al. 2003); in males, this region accounts for 60% of the difference in size between cline ends (Rako et al. 2006). It remains to be seen if lists of candidate genes for complex traits show substantial overlap between model and related species.

What about monitoring using quantitative traits? This approach is feasible if traits can be assessed repeatedly. In the case of photoperiodic responses studied by Bradshaw & Holzapfel (2001), the responses were highly repeatable across assays so that results obtained in 1 year can be related to those from a different year. Body size and other morphological assessments are also likely to be highly repeatable. However, physiological traits are often variable across assays, making it more difficult to detect phenotypic changes across years. In the case of insecticide resistance, estimates of resistance levels can also vary across assays, and a common solution is to compare resistance levels to those from a sensitive laboratory culture. Nevertheless, there is a danger that resistance will change in this culture due to inbreeding or laboratory adaptation. This problem will also arise when testing for changes in physiological responses or life-history traits.

Concluding remarks

There is potential for identifying alleles associated with specific environmental changes. The examples discussed above illustrate that genetic changes can occur over a short time frame, that there are molecular signatures of such changes having occurred in the past and that highly specific information can often be obtained by monitoring such changes. A number of marker systems are available for linking environmental changes to genetic markers and each of these systems has advantages and disadvantages in relation to the number of markers available, ease of scoring and difficulties in linking markers to phenotypes (Table 1).

Genetic markers need to be compared with other methods for detecting environmental change and used in conjunction with them. For the detection of pollutants, genetic markers need to be compared with a range of biomarkers including biochemical and physiological parameters, changes in population growth rate and shifts in species distributions. Genetic markers are likely to provide a useful adjunct to these other approaches once the range of genes that respond to specific pollutants have been identified, and once there is information on the likelihood of adaptive responses in groups of species.

An experimental programme to further develop links between environmental change and genetic markers might include some of the following components.

  • 1Identification of candidate markers from model organisms. Lists of markers for any likely environmental response should become available from these organisms. This includes markers for specific chemical stresses including pesticides, heavy metals, hydrocarbon pollutants and specific toxins-like PCBs and PAHs (polycyclic aromatic hydrocarbons). They also include markers for traits underlying temperature extremes, desiccation and starvation resistance, at least some life-history shifts and responses to other stressors like radiation.
  • 2Comparative genomic data to indicate which markers have been under recent selection. Achieving this aim will depend on comparisons of genomic data from populations or closely related species that have been exposed to different conditions, or genomic data from longitudinal samples.
  • 3Development of screening techniques for comparing genetic variation at numerous candidate loci. This technology is under development for screening human genetic variation. Microarrays can be used for comparing levels of gene expression where connections between expression and particular phenotypes have been made.
  • 4Information about the response of clonal organisms to environmental variables, along with the development of genetic markers for distinguishing the clones.
  • 5Longitudinal samples for testing hypotheses about genetic responses to environmental changes. These samples already exist for some organisms. For instance, Daphnia egg banks exist in the sediment of lakes and DNA can be extracted from these eggs for genetic analysis and links to environmental variables (Reid et al. 2002). For other model organisms, efforts can now be made to collect and preserve samples from environmental gradients.


The authors were supported by the Australian Research Council whilst preparing this review through their Special Research Centre, Federation Fellowship and Linkage Postdoctoral Fellowship schemes. We thank Bill Bradshaw and Volker Loeschcke for discussions about issues raised in this article.