1Small and isolated populations of species are susceptible to loss of genetic diversity, owing to random genetic drift and inbreeding. This loss of diversity may reduce the evolutionary potential to adapt to changing environments, and may cause immediate loss of fitness (cf. inbreeding depression). Together with other population size-dependent stochastic processes, this may lead to increased probabilities of population extinction.
2This set of processes and theories forms the core of conservation genetics and has developed into the conservation genetics paradigm. Many empirical studies have concentrated on the relationship between population size and genetic diversity, and in many cases evidence was found that small populations of plants do indeed have lower levels of genetic diversity and increased homozygosity. Although less empirical attention has been given to the relationship between low genetic diversity, fitness and, in particular, evolutionary potential, the paradigm is now widely accepted.
3Here we present five areas of the paradigm which could be refined, i.e. the ‘rough’ edges of the conservation genetics paradigm.
4Treating population size and isolation not as interchangeable parameters but as separate parameters affecting population genetics in different ways could allow more accurate predictions of the effects of landscape fragmentation on the genetic diversity and viability of populations.
5There is evidence that inbreeding depression may be a genotype-specific phenomenon, rather than a population parameter. This sheds new light on the link between population inbreeding depression and the expected increased probability of extinction.
6Modern eco-genomics offers the opportunity to study the population genetics of functional genes, to the extent that the role of selection can be distinguished from the effects of drift, and allowing improved insights into the effects of loss of genetic diversity on evolutionary potential.
7Incorporating multispecies considerations may result in the generally accepted notion that small populations are at peril being called into question. For instance, small populations may be less capable of sustaining parasites or herbivores.
8Comparative studies of endangered, common and invasive species may be a valuable approach to developing conservation biology from a phenomenological case study discipline into one investigating the general principles of what sustains biodiversity.
9The issues discussed set an agenda for further research within conservation genetics and may lead to a further refinement of our understanding and prediction of the genetic effects of habitat fragmentation. They also underline the need to integrate ecological and genetic approaches to the conservation of biodiversity, rather than regarding them as opposites.
Ecology is the scientific study of the interactions that determine the distribution and abundance of organisms (Krebs 1972), and ecologists try to understand the processes that influence biodiversity. In the modern landscape, biodiversity is endangered by the deteriorating quality of the environment and by a continuous decrease in the amount of space that is available for the natural environment. These two elements of biodiversity threat form the basis of two models of nature, which could be called paradigms in Kuhn's (1996) terminology. These paradigms underlie the interpretation and appreciation of biodiversity problems and their possible solutions.
The first paradigm, the ‘habitat quality’ paradigm, emphasizes that biodiversity problems are the consequence of habitat quality changes, leading to loss of species or populations that are not able to adapt to these changes. The solution to biodiversity problems within this paradigm lies in the control of habitat quality: management strategies to preserve the high quality of habitats where it is still present or measures to restore this quality where the habitat has deteriorated.
The second paradigm is the ‘conservation biology’ or ‘conservation genetics’ paradigm (as we concentrate mainly on the genetic part of conservation biology, we use the latter term throughout the rest of this paper). In this paradigm, biodiversity problems are appreciated in the light of characteristics of populations (or species), rather than in that of abiotic habitat characteristics: populations of species are endangered as a result of their small size and high degree of isolation from conspecific populations. Within this paradigm, solutions to biodiversity problems are defined in terms of population reinforcement (for instance by artificially increasing the number of individuals in small populations) or reduction of isolation (for instance by establishing corridors between existing conspecific populations).
Although these two paradigms do present two alternative interpretations of nature, they are not incompatible or mutually exclusive. The habitat quality paradigm does not deny that small populations may be more prone to extinction, but regards these problems as secondary to the problems imposed by habitat deterioration. By contrast, the conservation genetics paradigm fully acknowledges the importance of habitat quality, but emphasizes that even in optimal habitats, populations may be threatened with extinction. Shifting from the habitat quality paradigm to the conservation genetics paradigm implies shifting from a habitat-orientated to a genetics-orientated approach, from an ecosystem-orientated to a population-orientated view of nature and from a deterministic to a stochastic treatment of biological processes. Moreover, the distinction between these paradigms meets at least one criterion for a paradigm shift, as defined by Kuhn (1996): each paradigm is represented by its own community of scientists, united in their own societies (the Society for Restoration Ecology and Society for Conservation Biology, respectively), which tend to publish in their own journals.
In the rest of this paper, we concentrate on the conservation genetics paradigm. First, we outline the main model of nature in this paradigm, and then discuss five areas of the paradigm (‘the rough edges’) that could be refined. These discussions contribute to our strong conviction that optimal understanding of how biodiversity is threatened, and how it can be conserved, is most efficiently achieved when the two paradigms are integrated, rather than being regarded as opposites.
The conservation genetics paradigm
In their pivotal 1981 book Conservation and Evolution, marking the birth of the discipline of conservation biology, Frankel and Soulé presented a scientific framework showing how evolution and the dynamics of genetic variation within and among populations are important for the conservation of endangered species. Their point of departure was the fact that many endangered species only occur in small, isolated populations, resulting in the increased impact of a series of stochastic processes that negatively influence the chance of survival of these populations. More specifically, when a population becomes small, its growth rate is increasingly affected by four types of stochastic processes (Fig. 1).
First, demographic stochasticity – the random fluctuation in demographic parameters among individuals within a population – will lead to stochastic variation of the population growth rate in small or very small populations (Engen et al. 1998; Fox & Kendall 2002; Drake 2005; Fig. 1-1). Second, stochastic variation in environmental parameters over time and space will lead to stochastic variation of key demographic parameters, a phenomenon known as environmental stochasticity (Fig. 1-2). This type of variation will have significant impact on the population growth rate, even in average-size populations (Lande 1993; Boyce et al. 2006). Third, unpredictable catastrophes with a major impact on population size, such as hurricanes, flooding and forest fires, are more likely to lead to the extinction of small than of large populations (cf. catastrophic stochasticity; Goodman 1987; Lande 1993; Fig. 1-3). The fourth type, genetic stochasticity, is the impact of random population-genetic processes (i.e. genetic drift) on demographic parameters and population growth rate, an impact that increases with decreasing population size and has attracted much attention over recent decades (Soulé & Wilcox 1980; Schonewald-Cox et al. 1983; Soulé 1986; Ellstrand & Elam 1993; Young & Clarke 2000; Frankham et al. 2002).
The way in which genetic stochasticity, or genetic erosion as it is frequently referred to (e.g. Cole 2003; Oostermeijer et al. 2003), may affect extinction probabilities of populations is as follows (schematically outlined in Fig. 1). Small populations will be highly affected by genetic drift, as the extent of drift is an inverse function of effective population size. Genetic drift will ultimately lead to a loss of alleles from the population, and will also bring about an increase in homozygosity, averaged over loci and individuals (Hedrick 2005a). Moreover, in small populations, individuals will become increasingly related to each other over time (Frankham et al. 2002), leading to increased bi-parental inbreeding and increased homozygosity. Thus, small populations are expected to lose genetic variation over time, which is expressed as a decrease in the average number of alleles per locus and an increase in the average homozygosity.
This may have three interrelated consequences. First, increased homozygosity may be associated with reduced fitness. Negative correlations between fitness-related characters and homozygosity have frequently been reported (e.g. Mitton & Grant 1984; Allendorf & Leary 1986; Mitton 1997). Second, small populations are frequently associated with inbreeding depression. Both dominance and overdominance models of inbreeding depression predict that increased homozygosity may lead to reduced fitness (Charlesworth & Charlesworth 1999; Carr & Dudash 2003). Third, the loss of genetic variation from a population may reduce its potential to adapt to changing environments (Nunney & Campbell 1993; Frankham 1995). Rapid environmental changes may therefore raise the probability of extinction in small populations that lack the genetic variation needed for an adequate evolutionary response (Bijlsma & Loeschcke 2005).
These fitness aspects may increase the probability that a population becomes extinct. This may be either an immediate effect, when increased homozygosity and inbreeding depression affect those life stages with a high relative contribution to the population growth rate (i.e. high elasticity; De Kroon et al. 1986), or a postponed effect, when evolutionary potential is lost. In addition to this increased extinction probability, the reduced growth rate may result in a further decrease in population size, making it a self-reinforcing process. Once a population enters this process, its reinforcing nature results in a vortex, ultimately leading to extinction (e.g. Dennis 2002; Fagan & Holmes 2006; Hedrick et al. 2006).
Isolation – the lack of interpopulation dispersal – may play a role at both the demographic and the genetic levels. Migration may lead to the demographic rescue of small populations (Brown & Kodric-Brown 1977), with influx of individuals increasing the population growth rate. Gene flow, the genetic counterpart of dispersal, may lead to the ‘genetic rescue’ of genetically eroded populations, by reducing inbreeding depression and increasing the level of genetic variation (Richards 2000; Ingvarsson 2001).
In spite of this wide acceptance, there is an ongoing debate about the relevance of genetic stochasticity, compared with the other types of stochasticity, for extinction. It has been contended that populations go extinct owing to demographic and environmental stochasticity before genetic stochasticity becomes important (Lande 1988; Schemske et al. 1994; Holsinger et al. 1999; but see Frankham & Ralls 1998; Spielman et al. 2004). This debate partly reflects the ‘controversy’ between the habitat quality and conservation genetics paradigms. Several reviews have compiled and evaluated the empirical evidence for the importance of genetic effects (e.g. Young et al. 1996; Booy et al. 2000), including the very recent review by Leimu et al. (2006). In this paper we follow a different approach. Instead of evaluating the support for the conservation genetics paradigm, we discuss five areas where we feel that refinement of the paradigm is possible and needed (Fig. 1A–E). As such, we do not aim to criticize the core of the conservation genetics paradigm, but rather try to refine parts of it, the parts we call the rough edges of the paradigm. We do not claim that our discussion will resolve the ecology-vs.-genetics debate; instead, we argue that this debate is in fact asking the wrong question: what matters is not the question of what is more important for conserving biodiversity – ecology or genetics – but how ecology and genetics interact in affecting the viability of populations and species.
Population size vs. isolation
Habitat fragmentation and its effects on the persistence of populations and species is of major concern to conservation biology. When landscapes become fragmented, this may result in a decrease in the number of populations, a decrease in the average size of remaining populations and an increase in the average interpopulation distance. The conservation genetics paradigm includes these general effects by emphasizing the impact of reduced population size on the persistence of local populations, and by incorporating dispersal (or its genetic counterpart gene flow) as a means to mitigate these effects of reduced population size. In studying the effects of habitat fragmentation, many empirical studies have focused on the impact of reduced population size. Although in most cases it was acknowledged that isolation may play a role, isolation or dispersal were not explicitly quantified or investigated. This may be due to the inherent difficulties of defining isolation (Moilanen & Nieminen 2002; Soons et al. 2005) or measuring long-distance dispersal (e.g. Ouborg et al. 1999; Cain et al. 2000). Nevertheless, there are reasons to be more precise and use population size and isolation as two separate parameters, each with their own expected effects (Fig. 1A).
Fahrig (2003) pointed out that even though at the landscape level fragmentation is expected to lead to reduced population size and increased interpopulation distances, this is just one of several possible scenarios. She argued that, in fact, the number of remaining patches, the average size of remaining patches and the mean patch isolation may each either increase or decrease. Moreover, these three parameters will respond independently to landscape fragmentation and should therefore not be treated as interchangeable when studying the impact of habitat fragmentation.
Reduced population size and increased isolation will affect the persistence of local populations in different ways. From a genetic point of view, reduced population size results in increased genetic drift within local populations, leading to random loss of alleles from these populations. At the same time, the average degree of relatedness among individuals in a small population increases faster over time than in a large population, leading to increased levels of bi-parental inbreeding, increased homozygosity and, often, a reduction in average fitness (Fig. 1).
By contrast, increased isolation, i.e. increased average interpopulation distance, results in a decrease in among-population dispersal and gene flow. This may have several consequences. The demography and evolutionary dynamics of populations may become more independent of each other. The balance between genetic drift and gene flow may shift towards the former and the genetic differentiation between populations may increase (Wright 1969). The likelihood of genetic rescue (Ingvarsson 2001; Hedrick 2005b) of populations (i.e. the influx of ‘new’ genetic diversity into genetically impoverished populations) may decrease, as may the likelihood of demographic rescue (Brown & Kodric-Brown 1977). In a heterogeneous landscape, where isolated patches differ in environmental parameters, local selection may result in local adaptation of plant populations (Hufford & Mazer 2003; see below). Thus, not only do the landscape characteristics lead to reduced exchange of individuals (cf. seeds, pollen) but selection enhances differentiation in genetic quality between populations (as a consequence of local adaptation), leading to a further increase in effective isolation. If the situation involves a metapopulation, its genetic and demographic structure changes (Ouborg & Eriksson 2004). Based on a meta-analysis of 25 studies on the effects of habitat loss on species diversity in animals, Bender et al. (1998) concluded that there is good evidence that migration plays an independent role in the persistence of species. Whether a similar rescue effect could be prominent in plants, which generally have much lower migration rates and smaller migration distances than animals, is less clear, but several studies have demonstrated that isolation may affect extinction regardless of local population size (e.g. Dupre & Ehrlén 2002; Hooftman & Diemer 2002; Krauss et al. 2004; Matthies et al. 2004; Ozinga et al. 2005).
The current conservation genetics paradigm is strongly population-centred and treats dispersal and gene flow as forces from outside local populations, not affected by the genetic erosion process within the populations. Here we present three reasons to refine this approach towards a more realistic incorporation of dispersal and isolation.
First, dispersal is not just affected by landscape-level processes, but may be affected by local genetic erosion processes as well. Long-distance seed dispersal results from the integration of a number of demographic processes, starting with seed production per individual and followed by movement of seeds through space, germination and establishment. Each of these components might be affected by genetic erosion. Inbreeding depression in fecundity, germination and establishment has often been demonstrated in plants (e.g. Husband & Schemske 1996; for an overview see Keller & Waller 2002). However, the movement of seeds might also be affected by local population processes. Inbreeding effects on seed size, seed weight and the structure of dispersal-aiding appendages (plumes, bristles, etc.) are complex (Donohue 1999). This is the consequence of the complex genetic make-up of seeds: seed coats are entirely maternal, the embryo is entirely a diploid assembly of maternal and paternal genes, while the endosperm is frequently a triploid mix of maternal and paternal contributions. This makes it rather difficult to investigate and demonstrate inbreeding effects on seed size. Nevertheless, a few examples of such studies do exist (Schmitt et al. 1985; Pico et al. 2004a; Mix et al. 2006)
In a study of the effects of habitat fragmentation in a Dutch grassland landscape on the metapopulation structure and gene flow of two plant species, Mix et al. (2006) found that for the wind-dispersed Hypochaeris radicata, terminal velocity of seeds – a characteristic determining the dispersal potential of individual seeds (Soons & Heil 2002; Tackenberg 2003; Tackenberg et al. 2003; Soons et al. 2004) – was correlated with population size. The smaller populations had lower terminal velocities, indicating a higher dispersal potential (Fig. 2a). Although it was unclear whether these field effects could be fully accounted for by direct inbreeding effects on seed size (Mix et al. 2006), a separate glasshouse experiment with H. radicata demonstrated a significant effect of inbreeding on terminal velocity (Fig. 2b). Higher inbreeding levels resulted in higher terminal velocities, and thus in a reduced dispersal potential. Although field effects and glasshouse results thus pointed in different directions, these results at least indicate that the dispersal of seeds, a regional process, may be affected by local within-population processes. More experiments are needed in other landscapes and with other species to investigate the general applicability of these results. At the very least, however, these results suggest that isolation may be at least partly affected by local genetic erosion, rather than by a regional process independent of local processes.
Second, the conservation genetics paradigm places strong emphasis on the effects of genetic drift and inbreeding. However, demographically important traits will also be influenced by natural selection. Although it is generally assumed that in small populations drift might override the effects of selection (note, however, that experimental evidence for this is very scant), a trait such as dispersal potential is known to be selected for or against in a regional population or metapopulation context (McPeek & Holt 1992; Olivieri et al. 1995; Leimar & Norberg 1997; Olivieri & Gouyon 1997; Travis & Dytham 1999; Ronce et al. 2000, 2005). As the average interpopulation distance increases, dispersing seeds will more often be lost to the inhospitable surrounding landscape. This implies that at regional scales, there will be increased selection against dispersal. Evidence for this was presented in a study by Cody & Overton (1996), who found reduced dispersal potential in relatively isolated plant populations on Vancouver Island. Colas et al. (1997) demonstrated that fragmented populations of Centaurea corymbosa had very low dispersal capacity, and argued that even though unoccupied but suitable sites were present nearby, the low dispersal, and the resulting selection against dispersal capacity, would eventually drive the species to extinction in this area. Mix et al. (2006) found a correlation between the terminal velocity of seeds of Succisa pratensis and the distance to the nearest neighbouring population (Fig. 3), indicating that less isolated populations had higher dispersal potential. The final outcome of selection for dispersal will be a balance between costs (i.e. loss of individuals due to dispersal) and benefits (i.e. genetic rescue) at both local and regional scales. This balance between costs and benefits of dispersal deserves more empirical attention within the conservation genetics paradigm.
Third, dispersal also plays a crucial role in determining the potential for differentiation between populations. In a heterogeneous environment, selection may result in local adaptation of genotypes (Linhart & Grant 1996). The extent of local adaptation, a process of great importance for plants with their sedentary habit, depends on the level of exchange of genes between the various habitats. The current conservation genetics paradigm tends to ignore the impact of selection and local adaptation. Dispersal is regarded as a beneficial process, both because the paradigm focuses on the persistence of local populations and because it emphasizes the effects of drift and inbreeding. From this perspective, dispersal leads to an influx of genetic material, reversing the loss of alleles, promoting heterozygosity and possibly leading to heterosis (i.e. the increased fitness of offspring of crosses between parents from different populations). But if different populations are differently adapted in a heterogeneous landscape, dispersal might disrupt this local adaptation, resulting in a decrease in local average fitness (i.e. outbreeding depression) rather than an increase. Examples of both heterosis and outbreeding depression have been found (see the overviews by Hufford & Mazer 2003 and Tallmon et al. 2004). The topic of local adaptation is starting to draw the attention of applied conservation biologists, who are considering management practices such as introduction or reintroduction and ecological restoration (Hufford & Mazer 2003; Tallmon et al. 2004; Vergeer et al. 2004). The topic is also highly relevant to the discussion of what to preserve, as preserving several smaller, isolated populations might maintain more ecologically and evolutionarily important genetic variation at a regional scale than preserving one large population (see, for example, the SLOSS debate: McCarthy et al. 2005).
In summary, the conservation genetics paradigm should be refined by incorporating dispersal in a more realistic way. Investigating the specific impact of dispersal on local and regional population and metapopulation persistence, by adopting an evolutionary perspective on dispersal and by incorporating the effects of selection in heterogeneous environments, might significantly improve the adequacy of the paradigm. Studies are needed to evaluate the adequacy of various measures of population isolation, and to study the balance between genetic rescue and outbreeding depression in plants.
Inbreeding depression: population parameter or genotypic trait?
Inbreeding depression is defined as the reduction in fitness in inbred individuals as compared with their outbred relatives (Charlesworth & Charlesworth 1987). Genetic drift and bi-parental inbreeding will be higher within small populations than within large populations. Both processes result in an increased average homozygosity, and increased homozygosity allows the expression of recessive deleterious alleles, resulting in inbreeding depression (Charlesworth & Charlesworth 1999). Therefore, inbreeding depression plays an important role in the conservation genetics paradigm (Fig. 1B). Within the current paradigm, the degree of inbreeding depression is treated as a population parameter. In general, research to establish this parameter involves individuals from a particular population being both selfed and outcrossed. The offspring of both crossing types is then raised and the difference in fitness between the two offspring types is summarized in a parameter, δ, the population inbreeding depression. Alternatively, crosses can be made to create a number of inbreeding levels (e.g. selfings, sib-mating and unrelated crosses), after which inbreeding level is regressed on the fitness of the resulting offspring and the population inbreeding depression is expressed as the (standardized) regression coefficient. These methods of measuring inbreeding depression yield the potential decrease in fitness if the inbreeding level in the population increases over time as a consequence of continued small population size. The current inbreeding depression, or reduced fitness, is often measured by regressing population size on average fitness per population.
This method of incorporating inbreeding depression in conservation biology, as a single value per population, deserves re-evaluation and further refinement. Evidence is accumulating that inbreeding depression may be a trait that is variable among families within a population (e.g. Mutikainen & Delph 1998; Pico et al. 2004b; Bailey & McCauley 2006). A case study on Scabiosa columbaria (Van Treuren et al. 1993; Pico et al. 2004b) explained the difference between a population perspective and a genotype (cf. family) perspective on inbreeding depression. Scabiosa columbaria has been used as a model to investigate the effects of small population size on genetic variation within populations (Ouborg et al. 1991; Van Treuren et al. 1991) and the corresponding average fitness of individuals (Ouborg 1993a; Van Treuren et al. 1993). Inbreeding depression as expressed in germination, seedling survival, juvenile survival, adult survival, flowering percentage and reproductive output was estimated for a large population of this species in the Netherlands. The performance of selfed offspring compared with outcrossed offspring ranged from 1.08 (i.e. inbreeding enhancement; juvenile survival) to 0.65 (i.e. inbreeding depression; adult survival) (Van Treuren et al. 1993). In 2003, the same population was re-sampled, with a design that specifically aimed to quantify variance in inbreeding depression among families. Although the average inbreeding depression values in this second study were very similar to those in the first study, high variance in inbreeding depression was found among families for all life-history traits (Pico et al. 2004b). Some families showed strong inbreeding depression, as was expected based on the findings of the first study. At the same time, however, approximately as many families showed significant inbreeding enhancement, meaning that inbred individuals outperformed outcrossed individuals. A third category of families showed no inbreeding depression at all. While the findings of the first study would lead to the conclusion that inbreeding depression could be a serious problem to this population, the second study would result in the conclusion that inbreeding depression would only be a temporary threat. Over time, selection would lead to an increase in the frequency of families that show no inbreeding depression or even benefit from inbreeding.
A further refinement of the manner in which inbreeding depression is incorporated in the conservation genetics paradigm is stimulated by the observation that inbreeding depression is variable between stages (Husband & Schemske 1996). The study by Pico et al. (2004b) found no correlation between the inbreeding depression values that were found for different life-history stages. Furthermore, some evidence suggests that inbreeding depression is not a linear function of inbreeding level (Pray & Goodnight 1995; Ouborg et al. 2000), which further complicates matters.
The origin of the observed variation in inbreeding depression is not yet completely understood. Models have been constructed to investigate the role of differences in inbreeding history, of mutation and of genetic variation per se (Schultz & Willis 1995; Fowler & Whitlock 1999; Kelly 2004; Fox 2005; Moorad & Wade 2005). Thus, although considerable theoretical attention has been given to variance in inbreeding depression and its consequences for the purging of genetic load (Whitlock 2000; Glemin 2003), there is a great need for empirical studies that try to quantify this variance in the context of the conservation genetics paradigm.
In summary, inbreeding depression values not only seem to be dependent on population size, but seem to vary between life-history stages, families and inbreeding levels. This calls for a more sophisticated form of incorporation of inbreeding depression in conservation genetics. We need to know the extent of variation in inbreeding depression in each particular situation. Given the dependence of extinction probabilities on stochastic variation, the research agenda of conservation genetics should include studies searching for the effects of population size on the variance rather than the mean of inbreeding depression. Besides being important for the evaluation of extinction risks, variance in inbreeding depression might also play an important role in a regional population or metapopulation context. The future dynamics and genetics of a newly founded population, and of the metapopulation as a whole, might be very different when a site is colonized by a genotype that experiences inbreeding enhancement rather than inbreeding depression. There is a need for field studies that measure the development of fitness over several generations in experimental populations started with different genotypes, in order to evaluate the impact of variance in inbreeding depression on the probability of extinction.
Neutral markers vs. functional genes
As explained in the section on the conservation genetics paradigm, the effects of loss of genetic diversity are two-fold: there are the immediate effects on fitness (e.g. inbreeding depression) discussed in the previous section (Fig. 1B), and there is the postponed effect on extinction probability via the loss of evolutionary potential (Fig. 1C). The effect of small population size on genetic diversity and its loss has traditionally been investigated using genetic markers. From the early 1980s onwards, allozyme markers were used, but DNA markers, such as microsatellites, AFLPs and ISSR markers, have since replaced them, mainly because they are easy to use when scoring many markers for many individuals and because of their higher level of resolution compared with allozymes.
Whether an observed reduction of genetic marker variation in a small population necessarily indicates a loss of evolutionary potential is debatable. Although markers are excellent tools for studying the effects of non-selective processes such as drift and inbreeding, they are not necessarily suitable to quantify changes in selectively important variation. In general, what researchers did in the past was to measure neutral variation with many markers, and assume that selectively important markers would follow the same pattern (e.g. O’Brien et al. 1985; Allendorf & Leary 1986). The correlation between neutral marker variation and the variation in quantitative traits that are important for fitness is, however, not undisputed, with Lynch (1996) arguing that this correlation is likely to be low. Some meta-analysis reviews have indeed illustrated this weak or absent correlation (Butlin & Tregenza 1998; Reed & Frankham 2001). In a meta-analysis of 34 studies, Reed & Frankham (2003) found a significant correlation between fitness and genetic diversity in molecular markers, but genetic diversity measured in this way could only explain 19% of the total variation in fitness. It is thus dangerous to draw conclusions about the viability (and potential for evolutionary adaptation) of populations exclusively based on observed levels of molecular marker variation, and such conclusions are better avoided. In general, our understanding of the role of selection in small populations is limited, and requires more theoretical and empirical research.
Whatever the outcome of the interplay between drift and inbreeding on the one hand and selection on the other, very few tools were until recently available to get an empirical grip on selectively important genetic variation. With the recent emergence of the field of eco-genomics, this situation is changing rapidly, as several techniques from the genomics toolbox will allow detailed studies of non-neutral genetic variation in populations (Van Tienderen et al. 2002).
The screening of sequence variation in functional genes is developing rapidly, as a result of the availability of complete genome data of model species, the most prominent of which is Arabidopsis thaliana. Although each new species will require some work to fine-tune the Arabidopsis technique for the species of interest, this is becoming more and more common practice (Jackson et al. 2002; Stearns & Magwene 2003). There have already been many other examples besides Arabidopsis of the use of PCR-based methods to study variation in genes.
The genes involved in flowering in plants form a useful example for illustrating the potential of an eco-genomic approach. The regulatory gene pathways connecting environmental cues (light, day length and temperature) to the timing of flowering and the development of floral structures have been described in considerable detail (Blazquez 2000). Although these pathways are based on studies of A. thaliana, homologues of many genes in these pathways have been identified in other plant species. For instance, the LEAFY gene, which controls the identity of meristems that may develop into flowers, has been sequenced in species from the Leguminosae family (Archambault & Bruneau 2004) and the Poaceae family (Bomblies & Doebley 2005), as well as in Eucalyptus species (Southerton et al. 1998), apple (Wada et al. 2002), maize (Bomblies et al. 2003), orchids (Montieri et al. 2004) and Pinus (Mouradov et al. 1998).
Flowering time is controlled by two interacting genes from this pathway, FRIGIDA (FRI) and FLOWERING LOCUS C (FLC). A large number of variants of the FRI gene were detected when sequencing field ecotypes of A. thaliana (Le Corre et al. 2002). Haplotypes were categorized by the presence or absence of major deletions in the gene. A latitudinal cline in flowering time was found in Northern European and North American ecotypes, and proved to be correlated with variations in the FRI haplotype group, which lack the major deletions and are therefore fully functional (Stinchcombe et al. 2004). Variation within and among 12 French populations of A. thaliana was surveyed and compared with variation at neutral microsatellite loci (Le Corre 2005). A striking result was that the differentiation among populations in FRI was significantly higher than that in microsatellites, suggesting that local selection is affecting FRI variation.
While FRI and flowering time are among the more fully researched examples, many other traits are about to be investigated in the same way (Shimizu & Purugganan 2005). This opens up the exciting possibility of investigating how sequence variation in functional genes subjected to selection is correlated with population size and population isolation. It opens up new avenues to improve our insight into the role of selection within the context of the conservation genetics paradigm.
Another promising technique is the study of gene expression, by means of real time PCR or microarrays (Gibson 2002; Thomas & Klaper 2004). Although the latter method will at this point in time have to build on the availability of microarrays from well-studied species (A. thaliana, Medicago truncatula, maize, rice, etc.), the number of labs developing microarrays for their own species is rapidly increasing (e.g. Hegarty et al. 2005).
An example of the gene expression approach is given by Lai et al. (2006), who performed microarray analysis of hybrid sunflower species. They screened the expression of around 3000 genes, and discovered that 12.8% of them showed significant differences in expression between the two parental species Helianthius annuus and H. petiolaris, and the hybrid H. deserticola. A growth study in the environment of H. deserticola evaluated the adaptive significance of differential expression of five of these genes. A significant association with fitness was found for one gene. Another example is a recent field study by Schmidt & Baldwin (2006), who, among other things, evaluated the transcriptional responses of Solanum nigrum in a field study with competition as factor. They demonstrated, not unexpectedly, that competition reduced the reproduction of individual plants. At the same time they analysed the expression differences, using a microarray with 568 known genes, between competitive and control plants, and identified many genes that were either up- or down-regulated in the competitive plants. These examples show that gene expression studies can provide insights into the adaptive potential of populations (Whitehead & Crawford 2006).
In conclusion, the field of eco-genomics offers great opportunities for conservation genetics. First, it offers the tools to investigate the relationship between population size and population isolation and the extent of adaptively important genetic variation. Second, it also offers approaches to investigate in detail the relative roles of genetic drift and selection within the context of the conservation genetics paradigm. To profit fully from these possibilities, it is required that for a number of species, which could serve as conservation genetic model species, a sequence programme is started and an expressed sequence tag (EST) library is made and used to design microarrays.
From single-species to multi-species approaches
The conservation genetics paradigm implicitly focuses on single species. Although environmental impact on genetics and demography are of course not ignored, the problem is basically treated as involving a single species (or population) in an influencing but non-responsive world. That is, whereas the environment does affect the demography and genetics of the species, the species does not affect its environment. This is equally true for abiotic and biotic environments.
Obviously, the functioning of a plant population in its natural environment involves complex real interactions. Plants interact with their abiotic (e.g. nutrients, moisture level) and biotic environment (e.g. pollinators, herbivores, pathogens, mycorhiza). Neither the abiotic nor the biotic environment is an unchangeable, inert agent; they are responding reagents (Fig. 1D).
Several studies have shown a decrease in pollinator richness and abundance in response to habitat fragmentation (e.g. Kearns et al. 1998; Ashworth et al. 2004; Duncan et al. 2004). In recent years, studies of the relationship between plant–pollinator interactions and habitat fragmentation have been further extended to include plant–herbivore as well as plant–pathogen interactions (e.g. Ouborg et al. 2000; Carr & Eubanks 2002; Ivey et al. 2004). Tscharntke and co-workers have demonstrated that habitat fragmentation may alter the species web dramatically, by affecting different trophic levels in different ways (e.g. Kruess & Tscharntke 1994, 2000; Zabel & Tscharntke 1998). The approach is characterized by a multi-species community approach that until now, however, has largely ignored the genetic effects of habitat fragmentation.
Effects of habitat fragmentation on community structure may come about in several ways (Ouborg & Biere 2003). First, fragmentation may directly affect the presence and abundance of organisms at higher trophic levels. For instance, smaller plant populations offer smaller amounts of resources, which at some threshold level become insufficient to sustain viable herbivore populations (Anderson & May 1979; Getz & Pickering 1983). Thus, small populations may be protected to some extent from specialist herbivore and disease pressure. Examples of this have been described for natural plant–pathogen systems (Jennersten et al. 1983; Alexander 1990; Carlsson & Elmqvist 1992). An example of a threshold effect in plant–herbivore interactions was found in Salvia pratensis populations in the Netherlands. Salvia produces flowers which may each produce up to four seeds. Seeds are infected by the gall wasp Aylax salviae, which lays eggs in the developing seed. It thereby induces the formation of seed galls, and the reproductive output of an affected individual is reduced, sometimes by as much as 60%. In a survey of 20 populations, only populations larger than 300 individuals were found to be infected (Fig. 4). (N.J.O., unpublished data).
More isolated populations may be less likely to attract herbivores and diseases (Simberloff 1988). Positive correlations have been found between disease incidence in local populations and the disease status of neighbouring populations (Burdon et al. 1995; Antonovics et al. 1997; Ericson et al. 1999). While increasing connectivity in a fragmented landscape has been used as a method to alleviate the effects of habitat fragmentation on population viability (e.g. Harrison & Fahrig 1995), increased connectivity could also result in increased disease transmission. The net effect of increased connectivity on population viability therefore depends on the balance between these positive and negative effects (Hess 1996).
The second way in which habitat fragmentation may affect community structure is via the effects of loss of genetic diversity, which modifies the genotypic diversity in local populations. If the genotypic composition of a plant population affects the dynamics of the organisms at the second trophic level, as is expected when attractiveness, resistance, tolerance and/or avoidance have a genetic basis, the loss of resistance and other alleles from the plant population may change the entire food web structure, including third, fourth, etc., trophic levels.
Third, genetic erosion effects may also be evident from an increase in the average inbreeding level in small populations. If inbreeding results in changes in the physical, chemical and nutritional traits of individual plants, it will alter the interaction between plants and their pollinators, pathogens and herbivores. (Ajala 1992; Strauss & Karban 1994; Nunez-Farfán et al. 1996; Olff et al. 1999) As an indirect consequence, the interaction between herbivores and their natural enemies may be changed (e.g. Price et al. 1980; Hare 1992). Clear examples of this effect have been found in studies investigating the effect of monocultures or low biodiversity (e.g. Palmer & Maurer 1997; Knops 1999; Dukes 2002).
Although these effects have been studied to some extent, no coherent framework has yet been developed to link the conservation genetics paradigm with community structure effects. Much more research should be devoted to attempts to connect the genetic effects of habitat fragmentation on plant populations with the influence these effects may have on community structure. A more dynamic perspective should be developed, in which the biotic environment of the focal species is an ecologically and evolutionarily responding agent, in continuous interaction with the focal species. Some conclusions of conservation genetics might change once the community structure is incorporated in a dynamic approach. For instance, the general view that small populations are at peril may be reconsidered to some degree if we find that on the one hand small populations are protected against herbivores and diseases, while on the other they may lose resistance alleles owing to genetic erosion.
From phenomenology to general principles
Most empirical studies performed within the conservation genetics paradigm have concentrated on rare and endangered species. Although the conservation of what is threatened is in many cases the main and legitimate motivator for the research, this may also introduce a bias in the conclusions about the general applicability of the results. Rare species occupy habitats that also feature very common species. Even though the species are all subjected to the same degree of landscape fragmentation, some species are strongly affected whereas others do not respond at all. Exclusively studying rare species may lead to results that are specific for the set of traits that cause species to be unable to cope with fragmentation. If conservation genetics is to develop from a phenomenological crisis discipline, which concentrates on case studies, into a multidisciplinary science searching for the general principles that are involved in determining extinction probabilities of plant populations, it is also necessary to study common species within the context of the conservation genetics paradigm (Fig. 1E).
There are three approaches available to use common species to increase our understanding of the relationships that make up the conservation genetics paradigm, as outlined in Fig. 1.
The first is a comparative approach, in which the correlations in the diagram (Fig. 1) are determined for common species and compared with the correlation structure of rare or endangered species, which may reveal which processes render a species endangered. Some studies have indeed concentrated on common species. Hooftman et al. (2004) studied common species in fragmented Swiss fens and found some evidence that genetic drift does indeed affect genetic diversity in small populations of these species. A similar result was obtained in a study on the effects of fragmentation on the common plant species Succisa pratensis (Vergeer et al. 2003). Jules (1998) investigated the effects of fragmentation on the demography of a common understory herb of forests in the western USA, Trillium ovatum. He found evidence that fragmentation would lead to a decline in population persistence even in a common species. By contrast, a study of the conservation genetics of the co-dominant grass species Andropogon gerardii and Sorghastrum nutans from North American prairies by Gustafson et al. (2004a) found no evidence for a relationship between either population size or population density and genetic diversity. However, a second study by the same researchers (Gustafson et al. 2004b) did find evidence for the effect of local adaptation on the competitive ability of A. gerardii.
Thus, common species have received some attention within the context of the conservation genetics paradigm, including well-known work on Eichhornia paniculata (e.g. Husband & Barrett 1992) and Lythrum salicaria (e.g. Eckert et al. 1996). However, few studies have deliberately used a comparative approach. Karron (1989) compared inbreeding depression between rare and widespread species of Astragalus. Cole (2003) attempted to find general patterns in a comparison of genetic variation in a total of 247 rare vs. common plant species. He concluded that rare plant species show more significant reductions in genetic variation than common species. Although the results of that comparison were most likely confounded by the very different design of studies on common species as compared with rare species, they underline the importance of incorporating research on common species in a conservation biology context.
Second, the manipulation of common species offers a way of examining components of the conservation genetics paradigm in detail. Common species could be used as a model to investigate processes that are difficult or impossible to investigate for rare species, because of their susceptibility and low numbers. This approach has been used for several species, including Arabidopsis thaliana (e.g. Cahill et al. 2005), Brassica rapa (e.g. Wise et al. 2002) and Clarkia pulchella (Newman & Pilson 1997).
A third promising approach is the use of invasive plant species for conservation genetics research. These species invade new areas, at which stage they must temporarily be subjected to the same small and isolated population effects. In other words, they go through a bottleneck, and significant genetic drift effects can be expected. Yet, invasive species are apparently fully capable of dealing with these effects and developing into thriving species. Petit et al. (2004) adopted this approach for plants, by comparing species of oak (Quercus sp.) to elucidate the difference between invasive and non-invasive tree species. Other examples include work on Butomus umbellatus (Kliber & Eckert 2005), Capsella bursa-pastoris (Neuffer & Hurka 1999), Alliaria petiolata (Meekins et al. 2001) and introduced populations of Epipactis helleborine (Squirrell et al. 2001). The focus of most of these studies was on understanding the causes of the invasiveness of the species. Yet, studies of invasive species could provide reference data for conservation genetics studies of rare and endangered, conspecific, species. This comparative approach may offer a road towards understanding the general principles in conservation genetics.
In conclusion, the comparative study of rare, common and invasive species will allow evaluation of the conservation genetics paradigm, beyond the possibilities of case studies on endangered species alone.
The conservation genetics paradigm is currently widely accepted in ecology, genetics and conservation biology. With the paradigm becoming widely embraced, attention has shifted from the large and deterministic to the small and stochastic. The paradigm has emphasized the limits of theoretical models that implicitly assume (near) infinite population sizes, it has focused attention on the role of stochasticity in population dynamics and population genetics, and it has been an invaluable guideline for assessing the viability of endangered species and ways of improving this.
This paper has tried to identify aspects of the paradigm (‘rough edges’) that could, or should, be refined. Although conservation genetics research has provided ample evidence for the importance of genetics in conservation, we have discussed five areas where theoretical concepts and empirical approaches could be improved, to achieve a more thorough understanding of the role of genetics. These areas are also relevant to the ‘habitat quality’ vs. ‘conservation genetics’ controversy. For instance, although dispersal obviously has genetic consequences, the rate of dispersal has ecological determinants; inbreeding depression, though clearly having genetic causes, would be a meaningless concept without referring to environmental conditions; species interactions may prove to be as important as local loss of genetic diversity. These are not only areas where further refinement of the conservation genetic paradigm can be established, but research in these areas will also help in the further reconciliation of the ‘habitat quality’ and ‘conservation genetics’ paradigms. Instead of viewing ecology and genetics as opposite concepts, much can be gained by assessing how ecology and genetics interact in determining the viability of populations. The areas discussed may define a research agenda to come to a more complete understanding of this interaction between ecology and genetics, and thereby a more thorough understanding of the complex forces that determine biodiversity.
The present paper is the reflection of many years of research in the field of conservation biology. Although the views are entirely the responsibility of the authors, their development has benefited from discussions with numerous people. We would specifically like to thank Arjen Biere, Merel Soons, Eelke Jongejans, Ove Erikson, Mike Hutchings, Diethard Matthies, Tomas Herben and Jan van Groenendael. We would also like to thank David Gibson and two anonymous referees for their valuable comments on an earlier version of this paper. N.J.O. was supported by the EU TRANSPLANT grant (EVK2-1999-00042), P.V. was supported by a grant from the Dutch Technology Foundation (STW) and C.M. was supported by the Netherlands Organization for Scientific Research (NWO project 805-33-451).