Supporting Information
Appendix S1
Table S1. Threatened/near threatened species and populations where gene flow has been augmented for conservation purposes to alleviate genetic problems (known or suspected).
Common name Taxa References
Mammals
African elephant Loxodonta africana Frankham (2009a)
African lion Felis leo Trinkel et al. (2008);
Frankham (2009a)
African wild dog Lycaon pictus Davies-Mostert et al.
(2009)
Black rhinoceros Diceros bicornis Frankham (2009a)
Bighorn sheep Ovis canadensis Hogg et al. (2006)
Columbia Basin pygmy rabbit Brachylagus idahoensis Goodall et al. (2009)
Florida panther Puma concolor coryi Hedrick and Frederickson
(2010)
Golden lion tamarin Leontopithecus rosalia Frankham et al. (2010)
Mexican wolf Canis lupus baileyi Hedrick and Frederickson
(2010)
Northern quoll Dasyurus hallucatus Cardoso et al. (2009)
Birds
Greater prairie chicken Tymphanuchus cupido Westemeier et al. (1998)
in Illinois pinnatus
Red-cockaded woodpecker Piciodes borealis U.S. Fish and Wildlife
Service (2003)
Reptile
Swedish adder Vipera berus Madsen et al. (2004)
Plants
Button wrinklewort Rutidosis Pickup and Young (2007)
(within ploidy levels) leptorrhynchoides
Lakeside daisy population Hymenoxys acaulis Demauro (1993)
in Illinois var. glabra
Marsh grass of Parnassus Parnassia palustris Bossuyt (2007)
Mauna Kea silversword Argyroxiphium Robichaux et al. (1997)
sandwicense
ssp. sandwicense)
Brown’s banksia Banksia brownii Barrett and Jackson
(2008)
Round leafed honeysuckle Lambertia orbifolia Coates et al. (1998)
Species that would Potentially Benefit from Augmentation of Gene Flow
All species with fragmented distributions and at least one isolated population (no gene flow) can eventually benefit genetically from augmented gene flow (Frankham et al. 2010). The smaller the effective population sizes the sooner populations need augmented gene flow. For sexual species the benefits are increased genetic diversity, reduced inbreeding and increased reproductive fitness, whilst asexual species benefit from augmented genetic diversity. Metapopulations are especially susceptible to extinction from inbreeding (Saccheri et al. 1998; Tallmon et al. 2004). Island populations typically have reduced genetic diversity and are inbred compared to mainland populations and would benefit from re-establishment of gene flow (Frankham 1997, 1998). Many of the species on the planet fall into the above categories, including many threatened species of animals and plants in the IUCN Red List, most of which have fragmented distributions (World Conservation Monitoring Centre 1992; IUCN 2010). Frankham et al. (2010) alone listed the following taxa that fall into the category: black-footed rock wallabies on islands and fragmented mainland populations in Australia, koalas on islands and the mainland in Victoria and South Australia, giant pandas in China, black rhinoceros in Kenya, leopards in South Africa, ghost bats, Cunningham’s skink and matchstick banksias in Australia, Indian rhinoceroses, black-footed ferrets, grizzly bears, wolves, desert topminnow fish, and plants scarlet gilia, spreading avens and swamp pink in North America, tuataras on New Zealand offshore islands, Glanville fritillary butterflies in Finland, several species of birds in New Zealand, especially those with several island populations, and several species in the Tumut fragmentation study in Australia (in addition to the cases in Table 1).
Merging of Previously Long-Isolated Populations in the Wild
Many mammal, bird, fish, lizard and plant species in Australia, Europe, North and South America and Martinique show evidence of the merging of previously isolated and differentiated populations following climatic cycles (Soltis et al. 1997; Avise 2000; Hewitt 2000; Taylor et al. 2006; Antunes et al. 2008; Arnold et al. 2008; Grant & Grant 2008; Schwenk et al. 2008; Hu et al. 2009; McDevitt et al. 2009; De Carvalho et al. 2010; Thorpe et al. 2010). For example, several Australian rainforest vertebrate species have gone through cycles of isolation and rejoining (Joseph et al. 1995).
Many animal and plant species have become invasive following merging of genetically differentiated colonists (Ellstrand & Schierenbeck 2000; Kolbe et al. 2007). For example, many populations of Australian brushtail possums in New Zealand represent “successful” merging of mainland Australia and Tasmania sub-species (Taylor et al. 2004).
Occurrence of Inbreeding and Outbreeding Depression
There are many fewer published studies reporting outbreeding depression than inbreeding depression: the Web of Science from 1945-May 2010 revealed 379 references when “outbreeding depression” was a key word compared to 3,369 references for when “inbreeding depression” was used. However, this might reflect different research efforts devoted to the two phenomena. In a sample of 13 species, Edmands (2007) found that inbreeding depression was more frequent than F1 outbreeding depression, but that F2 outbreeding depression occurred with a similar frequency to inbreeding depression (based on data from only eight species). However, the largest data set on the issue, a fish meta-analysis involving 576 F1 and 94 F2 crosses by McClelland and Naish (2007), found that the average effect of crossing across all traits and all environments was significantly beneficial, indicating heterotic effects (the converse of inbreeding depression) are more common than outbreeding depression. Further, F1 and F2 effects were not significantly different.
Impacts on Fitness in Crosses of Fixed Chromosomal Differences
The impacts on reproductive fitness of crossing population that differ by centric fusions may be small or undetectable. Introduction of house mice from the Orkney Island of Eday into the genetically depauperate Isle of May population was successful in spite of the populations differing by three fixed centric fusions (Scriven 1992).
The deleterious effects of inversion polymorphisms arise because single (and other odd-numbered) crossovers result in unbalanced gametes (duplicated and deficient gene contents). The probability of a single crossover increases with the size of the inversion, and thus the deleterious effects increase correspondingly, as has been observed empirically (Coyne et al. 1993). As recombination rates vary along chromosomes, the increase in crossover rate with inversion size will not necessarily be proportional, but it will be an increasing probability function. This effect is not observed for all inversions, presumably due to lack of chromosomal pairing and crossing over in the inverted region (Coyne et al. 1993), meaning that our predictions of risks due to fixed inversions differences between populations will sometimes be overly conservative.
Additional References on the Role of Differential Adaptation in the Evolution of Reproductive Isolation
The following references also conclude that genetic adaptation to different environments drives most cases of rapid development of reproductive isolation (Rice & Hostert 1993; Schluter, 2001; Gavrilets 2004; Funk et al. 2006; Langerhans et al. 2007; Price 2008; Rolshausen et al. 2009; Stelkens & Seehausen 2009; Sobel et al. 2010; Thorpe et al. 2010; Wang & Summers 2010).
Coadapted Gene Complexes and Genetic Drift
Recent molecular evidence indicates that for large populations in the same environment the same loci and alleles are repeatedly involved in evolutionary change in large isolated replicates. For example, Hohenlohe et al. (2010) found rather similar patterns in 45,000 SNPS in parallel adaptations in three-spined sticklebacks and others have previously found that independent cases of benthic versus limnetic adaptations in this species repeatedly involve the same allele at the same locus (Colosimo et al. 2005). Further, the melanocortin-1 receptor (Mc1r) is repeatedly (but not exclusively) involved in adaptation in external pigmentation across mammals, birds and reptiles (Hoekstra 2006). Further, the same insecticide resistant mutations are often observed across populations and species (McKenzie & Batterham 1994; Hartley et al. 2006). For dieldrin-resistance, amino acid substitutions at the same site may be responsible for resistance in three insect orders. These studies indicate that the evolution of different coadapted gene complexes in replicate populations in the same environment is improbable unless there is a significant drift component. However, when drift is significant the populations will also become inbred, have reduced genetic diversity and have lowered adaptation and show genetic rescue effects upon crossing that would to a degree mask any outbreeding depression (OD) due to coadapted gene complexes. Different coadapted gene complexes are more likely to evolve in population in different, rather than similar environments as the selective forces are likely to be partially different and so favor different genetic variants (Pickup 2008).
Postzygotic Reproductive Isolation
Postzygotic reproductive isolation that is not a direct consequence of adaptation to different environments is widely viewed as arising primarily from Dobzhansky-Muller incompatibilities (Dobzhansky 1937; Muller 1940; see Coyne & Orr 2004; Presgraves 2010). However, these may often arise later in the speciation process (Muller 1940; Presgraves 2010; but see Coyne & Orr 2004) beyond the 500 years we have defined for the occurrence of OD in recently fragmented populations. In brief, Dobzhansky-Muller incompatibilities involve a two (or more) loci system, with initial homozygous genotype aabb. Independent mutations in the A and B loci occur in two allopatric populations and rise to fixation, yielding AAbb and aaBB genotypes in the respective populations. Crossing of these genotypes is difficult or impossible, due to incompatibilities of the A and B alleles, resulting in reproductive isolation. Fixation of the new mutations may occur through mutation pressure, genetic drift or natural selection, with the former two typically being slower than for prezygotic isolation. The probability of natural selection being involved will be greater when the two populations are in different environments. For example, Christie and McNair (1984) have described plant populations with complementary lethal factors, one being a copper tolerant allele that results in lethality in populations from non-polluted environments, but it confers normal viability in a copper tolerant population on copper polluted mine tailing. Cases involving natural selection are encompassed within our predictions (Predicting the Probability of Outbreeding Depression).
The development of Dobzhansky-Muller incompatibilities involve the waiting time for the occurrence of mutant alleles where the mutation rate is approximately 10-5 per locus per generation. Most such mutant alleles are lost by chance. The probability of eventual fixation of a new neutral mutation is its initial frequency 1/(2Ne), whilst the probability is 2sNe/N if the mutant is favored by selection (where s is the additive selection coefficient favoring it in heterozygotes) (Kimura 1983). A neutral mutation whose frequency reaches fixation takes on average of 4Ne generations to do so. A favored mutant allele with a selective advantage of 1%, has a probability of fixation of only 0.2%, given that Ne/N is approximately 0.1 (Frankham 1995). On average 500 such mutations have to occur before one is fixed, so the process will be slow unless the population size is large. Times to develop Dobzhansky-Muller incompatibilities will be shorter if they involve pre-existing genetic variants, but this was not what was envisaged by the initiators of the hypothesis.
One version of Dobzhansky-Muller incompatibilities that has been observed empirically involves silencing of alternative duplicated loci in different isolated populations (Lynch & Conery 2000). The rise to fixation of a non-functional (null) allele is expected to be a neutral or near-neutral process (Lynch & Conery 2000). For a duplicated gene pair, loss of function of one copy takes on average a few million years (Lynch & Conery 2000), well beyond our time frame. Incompatibilities due to duplicated histidine locus silencing were found in Arabidopsis thaliana (Bikard et al. 2009). Coyne and Orr (2004) doubt this is a common scenario, but to date it is one of the only two known case of Dobzhansky-Muller incompatibilities identified between populations within species that have been studied at the molecular level (Presgraves 2010).
An alternative version of the duplicated gene situation involves functional divergence of the two loci, driven by selection. This has occurred in many circumstances, including for duplicated hemoglobin loci in vertebrates. For examples, hemoglobins often differentiate in response to different environments, such as altitude (Storz 2007). In general, cases where natural selection drives development of incompatibilities are more probable when populations inhabit different environments than when they inhabit similar ones. Molecular analyses typically reveal that similar loci are involved in adaptations to similar environments and different ones in disparate environments (see Coadapted Gene Complexes and Genetic Drift above), as encompassed in our predictions (Predicting the Probability of Outbreeding Depression).
Based upon a review of molecular analyses on hybrid dysfunction loci, Presgraves (2010) concluded that “the first steps in the evolution of hybrid dysfunction are not necessarily adaptive”, but this conflicted with the view of Maheshwari et al. (2008) who concluded that “most HI (hybrid inviability) genes identified to date show evidence of positive selection.” The proposed evolutionary bases of changes described by Presgraves (2010) were duplicate gene silencing (as above), mutation pressure, host-pathogen genetic conflicts, and genetic conflicts. Of 14 loci, only three involved intraspecific effects, two involved duplicate gene silencing (see above) and one host-pathogen conflict and all were in Adabidopsis thaliana, a selfing species where drift effects are likely to be large (Probabilities of Inbreeding and Outbreeding Depression in species with Different Breeding Systems: Selfing Species below). Changes due to mutation pressure alone will be extremely slow and outside our time scale of relevance (Frankham et al. 2010). Host-pathogen conflicts involve either different evolutionary paths of adaptation to the same pathogen, with initial divergence due to genetic drift followed by selection, or more likely adaptation to different pathogens which falls into the realm of different environments. Selfish genetic elements (transposons, meiotic drive elements and gamete killing segregation distorters) have evolutionary interests that conflict with those of their hosts. Hosts typically evolve means to suppress the deleterious impacts of selfish elements. Postzygotic OD from selfish elements will arise either from crossing populations with and without the selfish elements, or from incompatibilities from different host loci evolved to suppress the deleterious impacts of the elements (with initial divergence due to genetic drift). It is unclear how rapidly such differences will evolve in populations that have only recently been isolated. Those requiring genetic drift to generate initial differences will typically be associated with genetic rescue effects that will tend to mask them (Frankham et al. 2010).
Overall, the evolution of OD via postzygotic isolation will often occur beyond our time frame, or involve differential adaptation to disparate environments.
Prediction Equations
Equations 1 and 2 have simplifying assumptions as detailed below, but have been experimentally validated and are widely used in quantitative genetics and animal breeding. The first assumption for equation 1 is that the drift term S(1 – 1/[2Ne])t – 1 is not unduly affected by selection. This assumption was first made by Alan Robertson, on the basis that many loci are likely to be affecting quantitative traits so that selection per locus is weak, and is widely used in quantitative genetics and animal breeding (Robertson 1960, 1970; James 1972; Bulmer 1980; Hill 1982a, b; Hill & Caballero 1992). The second assumption is that disequilibrium generated by selection (Bulmer 1980) has minor impacts. The drift term has been shown to be a very good approximation over 50 generations from experiments in invertebrates, mammals and plants (Weber 2004), and for populations subjected to short severe bottlenecks in population size and subsequently subjected to many generations of artificial selection (see Frankham 1980). Similar assumptions are inherent in equation 2 (Hill 1982a, b) and again empirical tests have shown them to be adequate for the current purposes (Hill 1982b; Frankham 1983).