The IUCN Red List reports 11 defined threats that affect the viability of populations, and it has these threats recorded for over 140,000 species (IUCN Red List 2022-1). At least five of these threat categories (Threats 1, 2, 3, 6 and 7) directly reduce the amount of suitable habitat that is available to species. Other threats cause population sizes to decline (e.g., Threats 5 and 9), whilst yet others reduce the population connectivity (e.g., Threat 4) (see IUCN Red List 2022-1). These threats to species survival can have immediate effects by causing mortality or failed reproduction. However, these threats also have more long-lasting effects because they result in genomic erosion, but the impacts of genomic erosion might not be immediately visible due to a time-lag between cause and effect. Conservation scientists are aware that we may underestimate the long-term threat to species survival, which is acknowledged by terms such as the “extinction debt” (Tilman et al., 1994; Kuussaari et al., 2009) and “drift debt” (Gilroy et al., 2017). Here we examine the time-lag effect of habitat loss and population fragmentation on genomic erosion.
We build a spatially explicit computer model to simulate the impacts of habitat loss and fragmentation, analysing their impacts on the genetic variation and the composition of the genetic load. Recently, Sachdeva, Olusanya, and Barton (2022) analysed the impact of migration, drift, and demographic stochasticity on the genetic load in peripheral populations, highlighting the importance of genetic Allee effects. We simulate a very low genetic load of less than 0.1 lethal equivalents. The simulated load is >30 times smaller than values observed in captive mammals (Ralls, Ballou and Templeton, 1988), ~120 times smaller than the average in wild animals (O’Grady et al., 2006), and ~150 times smaller than the genetic load estimated for the pink pigeon in the island of Mauritius (Jackson et al., 2022). This avoids genetic Allee effects and extinctions. Nevertheless, the present simulations are relevant for conservation genetics by improving our understanding of the spatiotemporal dynamics of genomic erosion, but there are a number of caveats.
First, we simulated a genetic load that is under hard selection (i.e., resulting in death), whereas in nature, a proportion of the genetic load experiences soft selection, which only reduces the relative fitness (Wallace, 1975). Second, we did not simulate pre- or post-zygotic selection, which is likely to play an important role in how populations cope with their genetic load (van Oosterhout et al. 2022b). Third, simulating a much higher genetic load will lead to rapid extinction of the modelled populations, yet we know that the natural population of the pink pigeon has not gone extinct despite harbouring a high genetic load of circa 15 lethal equivalents (Jackson et al., 2022). Altogether, this implies that a substantial part of the genetic load is under soft selection, or that pre- and/or post-zygotic selection are a vital missing component of our model.
Simulating an (unrealistic) low genetic load may bring some further technical complications. However, theoretically, the rate of load conversion due to inbreeding is independent of the magnitude of the genetic load (see equation in Box 2, and Bertorelle et al. 2022). In addition, the loss of load due to purging is also relative and unaffected by the size of the load. This assumes free recombination between deleterious alleles and a sufficient excess of offspring that can be scrutinised by selection. Given that we simulated an entire (albeit small) chromosome with realistic recombination landscape, we believe that the patterns we observed in our simulations are likely to hold true for more realistic load scenarios. Of course, real genomes and evolutionary processes are considerably more complex than those simulated here, and further research in the dynamics of the genetic load in spatially explicit populations is certainly needed. Ideally, more advanced computer models should include both soft and hard selection, and such models should be able to simulate selection acting on (large numbers of) gametes and zygotes. Despite these various caveats, our aim was to investigate the spatiotemporal dynamics of the components of genetic diversity in response to habitat loss and fragmentation. Accordingly, we believe we have successfully replicated the predicted pattern of stochastic loss of rare deleterious alleles by drift, genetic load purging by selection, and the conversion of the masked load into a realised load (e.g., Bertorelle et al. 2022).
Our results indicate that declining populations in fragmented habitats might be particularly vulnerable to genomic erosion. There is, however, a substantial time-lag between genomic erosion and habitat fragmentation (or population size decline). The time-lag was substantial for neutral variation, and not until circa 75% of the native undisturbed forest had been destroyed, did we observe a decline in neutral diversity. In the simulated Mauritius population, this was around the year 1900, when the census population size had dropped to approximately N=2000 individual. Assuming a similar Ne/N ratio as in the ancestral population (Ne/N≈0.5), this suggests that decline in genetic diversity started to be noticeable once the effective population size had dropped to around 1000 individuals (Ne=1000).
Remarkably, the impact of habitat fragmentation and population size decline was even more delayed for the genetic load. (As mentioned earlier, this is not an artefact of simulating a low genetic load because the load conversion is a function of the inbreeding coefficient and not the absolute magnitude of the load components.) Around the year 2000, we started to notice an increase in the realised load that would have reduced fitness and population viability. Because we simulated an extremely low genetic load of <0.1 lethal equivalents, the fitness impacts were negligible in our simulations. In natural populations with a genetic load one or two orders of magnitude higher, we suspect genetic Allee effects would exacerbate the rate of population size collapse, and possibly lead to extinction (see also Sachdeva, Olusanya. and Barton, 2022). In other words, the inferences we make about the impact of habitat fragmentation and decline on the genetic load are conservative, and in all likelihood, the situation is worse in natural populations. Moreover, is also important to note that we parametrised our model based on the highly mobile pink pigeon and that less-mobile taxa are likely to suffer more from the effects of localised inbreeding (Sachdeva, Olusanya and Barton, 2022).
The genetic load of unconditional deleterious variation is increasingly being recognised as a pervasive, long-term threat to the viability of declining populations (van Oosterhout, 2020; Mathur and DeWoody, 2021; Ochoa and Gibbs, 2021; Bertorelle et al., 2022; van Oosterhout et al. 2022a; Kyriazis, Robinson and Lohmueller, 2022). Our simulations show that during population decline and fragmentation, genetic drift and inbreeding increase the frequency of initially rare mutations, making them homozygous. Given that these mutations are present at low frequency in the ancestral population, this process takes time and can remain unnoticed. This can explain the observed time-lag between habitat deterioration and its impact on the genetic load. This implies that genetic load in declining populations poses very much a long-term threat, and that it needs to be accounted for in conservation and species recovery programs (van Oosterhout et al. 2022a).
One final point to make is that selective purging of deleterious alleles only occurs at the expense of fitness. However, if soft selection operates on the deleterious variation, extinction may be prevented (Wallis 1975; van Oosterhout et al. 2022a). With soft selection, fitness is relative, and even individuals with low fitness can survive and reproduce, as long as they are the fittest individuals in the population. Hence, soft selection does not reduce the population size, but rather, it determines which individuals survive and reproduce. In contrast, if hard selection operates against the realised load, individuals die or fail to reproduce irrespective of the fitness of others in the population. Hence, hard selection reduces the population size, which makes the population more vulnerable to extinction (van Oosterhout et al. 2022a). To better predict the impact of genomic erosion on the genetic load and population viability, we need to better understand the type of selection that is acting on deleterious mutations.
In summary, our simulation study shows that genomic erosion should be considered a long-term threat to population viability, alongside the immediate threats that are presently recorded on the IUCN Red List. Although efficient purging reduced the genetic load, a significant proportion of the deleterious mutations became expressed in our declining metapopulation. This elevation of the realised load occurred relatively late after the onset of habitat loss. However, it continued to pose a threat to the population also in future generations, even without any further habitat loss. In addition, the metapopulation continued to lose neutral genetic variation. Assuming that neutral diversity is a reasonable indicator for adaptive genetic variation (García-Dorado and Caballero, 2021; Willi et al., 2022), the evolutionary potential of the population is also likely to become increasingly undermined (Kardos et al., 2021; van Oosterhout et al., 2022a). The fact that these processes continued to cause genomic erosion emphasises the urgent need of habitat and ecosystem restoration. We are in the UN’s Decade on Ecosystem Restoration, and our study shows that we need to urgently restore habitat, which will help save species from future extinction.