How can inbreeding be used for selecting mutations?

How can inbreeding be used for selecting mutations?

We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

I understand that inbreeding, after a number of generations of crossing genetically related individuals eventually yields homozygotes, however I can't seem to understand how it can be used for selecting mutations.

Consider a sibling-sibling mating system, the simplest inbreeding situation, with a simple Mendelian trait.

Step 0: Let's start with heterozygotes, Aa x Aa.

Step 1: Their offspring will be a collection of AA, Aa, Aa, aa, in roughly those ratios assuming equal survival.

Step 2: Now we breed siblings. If we happen to breed Aa x Aa, we are back to the original starting point. Return to Step 1.

If we happen to breed AA x AA or aa x aa, and then continue to breed those offspring with their siblings, all of the following generations will be homozygous unless there is a mutation. Stop here, you have homozygotes.

If we happen to breed AA x Aa or Aa x aa, the next offspring will match those parents: AA, Aa or Aa, aa. If we breed Aa x Aa, return to Step 1. If we breed AA x AA or aa x aa, stop here, you have homozygotes.

Of course, theoretically you could end up with heterozygotes doing this forever, but the odds get lower and lower the more generations you breed.

For selecting a particular mutation, it is often simpler to genotype for presence of an allele rather than zygosity (i.e., just use simple RT-PCR). For example, if you want to eventually produce aa, you can test for the 'a' allele, and not breed any offspring that do not have any 'a', therefore they are AA. If you do this, you will always eventually end in a path with aa homozygotes.

If you have an easily produced model organism, you can also do this simply by phenotype (depending on the trait): once a strain settles into a consistent phenotype, you know you have a homozygote of one type or the other. If your model organism is easy to produce, it's okay if some don't end up carrying the mutation.

For selecting a random mutation, the same is true, except you have many genes instead of just gene "A". Any random mutation in just one individual early in the breeding could potentially be fixed in the inbred population somewhere down the line, following the steps above. If you breed BB x Bb, where Bb has a random mutation, you have a chance to breed a Bb x Bb pair in the next generation, which starts you at Step 0 for that particular random mutation.


When populations are partially inbred due to the population structure or to a mixed mating system like partial self-fertilization, some individuals will be more inbred than others. This heterogeneity among individuals in the history of inbreeding can greatly complicate the interpretation of measures of quantitative genetic variability when the traits studied exhibit inbreeding depression. Partial inbreeding can also bias measures of phenotypic selection toward the detection of strong directional and stabilizing selection. In this paper, data are presented from several inbreeding experiments conducted on two partially selfing, annual populations of the monkeyflower Mimulus guttatus that show that the means of many of the morphological and phenological traits measured were affected by inbreeding. These findings imply that estimates of heritabilities and additive genetic covariances would not reflect the potential for these populations to respond to selection. Phenotypic selection analyses conducted on naturally occurring plants, involving linear regressions of relative seed production on the traits, revealed significant directional selection on many of the same quantitative traits measured in the inbreeding studies. However, when the same selection analyses were performed on plants with known histories of inbreeding, part of the statistical relationship between relative seed number and the traits was found to be due to the mating system: inbred individuals had both lower seed production and different mean values for the traits than outcrossed individuals. It is also shown, with a hypothetical example, that partial inbreeding can bias measures of stabilizing selection toward the detection of strong stabilizing selection. Partial inbreeding therefore tends to make directional and stabilizing selection appear stronger than it is, and it may be that natural selection in the wild is actually weaker than many studies of partially inbred species suggest.


Inbreeding, the mating between relatives, leads to increased homozygosity of alleles that are identical by descent. Deleterious consequences of inbreeding on fitness, known as inbreeding depression, have been studied extensively and are found in almost all normally outbreeding organisms (for review see Charlesworth & Charlesworth, 1987 ). Typically, it has been found that inbreeding depression increases with an increase in the inbreeding coefficient F ( Willis, 1993 Dudash et al., 1997 Koelewijn, 1998 Ouborg et al., 2000 ), which is a measure of the degree of inbreeding of an individual and gives the probability of identity by descent of two alleles at a locus ( Wright, 1922 ). Apart from changes in mean genotypic values, inbreeding may also increase genetic variance for fitness ( Willis & Orr, 1993 Deng & Lynch, 1997 ) and thus influence natural selection. However, inbreeding depression and its variance may not only be influenced by the degree of homozygosity but may also be influenced by the interactions among loci and/or by interactions with the environment.

Deleterious mutations act independently when the effect of a mutation does not depend on the existence of other mutations in the genome. If their effects are not independent, they may interact with each other. This might be either through antagonistic epistasis, where each additional mutation has a smaller effect than previous mutations, or through synergistic epistasis, where each deleterious mutation added to the genome has a greater effect than the preceding mutations ( Peters, 1999 ). Synergistic epistasis leads to a decrease of the logarithm of fitness, with the slope declining more steeply as the number of mutations increases.

The presence of synergistic epistasis has important consequences for selection against deleterious mutations. As selection acts against individuals with low fitness, deleterious mutations are purged more efficiently with synergistic epistasis than without. The strongest effect is reached in truncation selection, where selection acts against individuals with more than a threshold number of deleterious mutations. The maintenance of sexual reproduction, despite its two-fold cost compared with asexual reproduction, may be explained if mutations interact synergistically (deterministic mutation hypothesis Kondrashov, 1988 ).

Although it is difficult to study epistasis directly, the relationship between the number of mutations and their effect on fitness can be examined indirectly by monitoring the effect of different levels of inbreeding on fitness or fitness components. If it holds that inbreeding depression is mainly due to recessive deleterious mutations, a nonlinear relationship between the logarithm of fitness and the inbreeding coefficient indicates epistasis of harmful mutations ( Crow & Kimura, 1970 ).

Recently it has been suggested that synergistic epistasis could even be stronger when individuals with large numbers of deleterious mutations suffer disproportionately more from parasitic diseases ( Howard & Lively, 1994, 1998 Lively & Howard, 1994 West et al., 1999 ). Again, this is particularly relevant for the maintenance of sex. So far the maintenance of sexual reproduction has been explained by either environmental or mutation-based models. Environmental models suggest that sex accelerates adaptations to a changing environment (for instance coevolutionary interactions with parasites) by creating new gene combinations (Red Queen hypothesis, Hamilton, 1980 Bell, 1982 ). Mutation-based models argue that sexual reproduction is advantageous because it allows populations to purge deleterious mutations more efficiently (deterministic mutation hypothesis, Kondrashov, 1988 ). The assumptions of these models may not always hold (e.g. highly virulent parasites, synergistic epistasis and high mutations rates). However, if an interaction between parasitic diseases and mutations exists, which would strengthen synergistic epistasis, it could render sex evolutionarily stable against repeated invasion by clones with strongly relaxed assumptions ( Howard & Lively, 1994 ).

The present study consists of a theoretical and an experimental part. In the theoretical part we use a mathematical model to examine how inbreeding depression, epistasis and variance of fitness are influenced, when experiments are conducted with individuals from a population, which vary in their degree of inbreeding. Studies that deal with inbreeding and its consequences sometimes work with base populations that are assumed to be outbred or at least have low variance in the inbreeding coefficient across individuals (e.g. Willis, 1993 Dudash et al., 1997 Koelewijn, 1998 Ouborg et al., 2000 ). This assumption is often fulfilled, but we show here that if within-population variation in the inbreeding coefficient is high, it may lead to substantial deviations from expectations. Therefore, in the present experiment great care was taken to avoid this problem.

In the experimental part, we assess and compare the fitness of six sets of increasingly inbred Daphnia magna clones in a competition experiment vs. a tester clone. With three different levels of inbreeding, epistasis can be inferred from the shape of the curve. The experiment is designed so that the inbreeding coefficients of the three inbred clones differ from the lower to the next higher inbred clone exactly by 0.25, independent of the inbreeding coefficient of their ancestors. Thus, the results are not influenced by the unknown history of the genetic material used. To test for interactions between parasites and inbreeding, each line is tested with and without a microparasitic infection. Testing for epistasis across different levels of inbreeding only includes the natural variation in mutational load, rather than artificially elevated mutation levels as they occur when mutagens are used to create lines with different genetic loads.

Materials and methods

Stocks and culture conditions

The focal population was the Ives (IV) population of D. melanogaster. IV was isolated from a wild-caught sample of 200 males and 200 females in Amherst, Massachusetts (Rose and Charlesworth, 1981). At the time, the south Amherst population was thought to be continuous and overwintering, and had been monitored since at least 1931 (Ives, 1970). From 1981 onwards, the IV laboratory population has been maintained as a large outbred stock at a minimum population size of 1000 individuals at 25 °C, 50% relative humidity, on a 14-day, discrete generation cycle with moderate densities of 60–120 individuals per 10 ml of banana/agar/killed-yeast medium (Rose, 1984). On day 14/day 0, the population is placed under CO2 anesthesia, mixed and transferred to new vials to oviposit until ∼ 100 eggs are laid in each vial. This usually takes ∼ 30 min and represents the only opportunity for offspring production.

In 2004, a replica of the IV population was created by backcrossing the IV population to a population bearing the recessive bw 1 (brown eyes) allele and denoted as IVbw. The IVbw population serves as an outbred, genetically similar population for use as competitors against IV-derived individuals in measurements of fitness. This marker has few deleterious side effects and the IVbw stock is vigorous. Periodically, the IVs are introgressed into the IVbw to prevent drift between the focal (IV) and competitor (IVbw) populations.

Inbred lines

The 18 inbred lines used for this study were generated using a novel application of the clone-generator system of markers and chromosome rearrangements (Chippindale et al., 2001). A selection of haploid genomes derived from the IV population and known to possess significant genetic variation for fitness in each sex provided the raw material for the creation of the inbred lines. These genomic haplotypes were made homozygous by mating males bearing the haplotype of interest in the heterozygous state along with a marked translocation (T(2:3)rdgc st in ri p P bw) to wild-type females, discarding the marked translocated autosomes in the female progeny. These females, having received a genomic haplotype from their father, were collected as virgins and again crossed to males with the haplotype of interest, along with the marked translocation, for 10 generations (Figure 1). With each successive cross, the proportion of genes identical to the founding line's haplotype increases by 50%. Subsequently, these lines were maintained at population sizes of 10 females and 6 males to minimize genetic variation. A reduced sex ratio was employed to reduce mate harm to females, minimizing the effective population size while maintaining productivity. Periodically, they were backcrossed again to their founding lines. As the final result of these crosses is the homozygous version of a specific genomic haplotype, we named this method directed inbreeding (DI).

Breeding schemes used in this experiment. (a) Generation of hemiclone lines. A single male from the IV population possessing an unknown genotype (black) was crossed to virgin clone-generator females possessing a compound X (C(1)DX, y, f, depicted by DX symbol), a Y chromosome and a marked translocation of chromosomes II and III (T(2:3)rdgc st in ri p P bw, depicted by solid bar spanning the two major autosomes because these chromosomes cosegregate in surviving offspring). The resulting male offspring possesses one of four possible genotypes, consisting of a set of X plus wild-type autosomes inherited paternally, with a Y and marked translocation inherited maternally. (b) A single male is then randomly selected to again cross to clone-generator females, thus fixing a genomic paternal haplotype (white) within a line. The resulting hemiclone line is propagated by crossing males heterozygous for the marked translocation with clone-generator females. (c) DI. Males from a given hemiclone line are first crossed to virgin wild-type IV females with an unknown genotype (black). Virgin females are collected from this cross, which are now heterozygous for the founding hemiclonal haplotype. These females are again crossed to hemiclone males, and the process is repeated 10 times to yield lines inbred for the founding haplotype. (d) Generation of experimental flies. Inbred females from a given line are crossed to either inbred males (white) or outbred males (black) to yield outbred and inbred flies of both sexes.

The DI method has several advantages over traditional inbreeding methods, such as brother–sister mating. First, DI is less susceptible to the stochastic loss of mutations resulting from the inbreeding process. Brother–sister mating exposes mutations to both drift and selection during the inbreeding process, which could result in genomes purged of a fraction of their deleterious mutation load. DI reintroduces all of the mutations present in the founding genome in each generation of inbreeding, preventing the loss of deleterious mutations. The exception, common to all inbreeding approaches, is that one cannot fix sterile or lethal mutations within a line these are thought to contribute to approximately half of the total mutation load in D. melanogaster (Lynch and Walsh, 1997). Second, the DI approach has the advantage of capturing all three of D. melanogaster's major chromosomes in the same experiment. With this species, recombination-suppressing balancer chromosomes are restricted to capturing and manipulating at most two of the major chromosomes simultaneously. Manipulating the whole genome should give a more accurate picture of the population consequences of genome-wide processes, while leaving open the potential for later deconstruction and analysis at the chromosomal level using balancer techniques.

Each inbred line was used to experimentally generate both inbred and outbred flies for use in measurements of fitness (Figure 1). Outbred individuals were created by crossing females from the line of interest with randomly selected wild-type males from the IV population. Inbred individuals were created by crossing these females with males from the inbred line. In this manner, both outbred and inbred individuals have inbred mothers with the same genotype, eliminating differences between outbred and inbred experimental flies because of maternal effects. Although the potential for ID exists along the entire genome for females, it was restricted to the autosomes for males, as both outbred and inbred males of a given line express the same X chromosome hemizygously.

Viability assay

Juvenile viability was assessed by placing 50 eggs from a given genotype in a vial along with 50 eggs from a standard competitor (IVbw), mimicking standard culture densities. After 12 days, sufficient for virtually all adults to emerge (confirmed by visual inspection), egg-to-adult viability was assessed by counting and scoring progeny for both sex and eye color. Each genotype/treatment combination was replicated five times.

Adult fitness assay

Inbred or outbred individuals of the sex/genotype of interest were collected during peak emergence (day 9 from date of oviposition) and singly transferred to a vial containing an age-synchronized culture of IVbw that was reared at standard culture densities ( ∼ 100 eggs/vial). After 5 days (day 14), experimental vials were anaesthetized with CO2 for 2.5 min and transferred to fresh vials to allow for oviposition (30 min), mimicking standard culture conditions. The adults were then removed from the vials and the sex/number of progeny from the target individuals (distinguishable by their red eyes) was scored 12–14 days later. Each genotype/treatment/sex combination was replicated 30 times.

Mutation and Conservation

Mutation can critically affect the viability of small populations by causing inbreeding depression, by maintaining potentially adaptive genetic variation in quantitative characters, and through the erosion of fitness by accumulation of mildly detrimental mutations. I review and integrate recent empirical and theoretical work on spontaneous mutation and its role in population viability and conservation planning. I analyze both the maintenance of potentially adaptive genetic variation in quantitative characters and the role of detrimental mutations in increasing the extinction risk of small populations. Recent experiments indicate that the rate of production of quasineutral, potentially adaptive genetic variance in quantitative characters is an order of magnitude smaller than the total mutational variance because mutations with large phenotypic effects tend to be strongly detrimental. This implies that, to maintain normal adaptive potential in quantitative characters under a balance between mutation and random genetic drift (or among mutation, drift, and stabilizing natural selection), the effective population size should be about 5000 rather than 500 (the Franklin-Soulé number). Recent theoretical results suggest that the risk of extinction due to the fixation of mildly detrimental mutations may be comparable in importance to environmental stochasticity and could substantially decrease the long-term viability of populations with effective sizes as large as a few thousand. These findings suggest that current recovery goals for many threatened and endangered species are inadequate to ensure long-term population viability.


La mutación puede afectar criticamente la viabilidad de poblaciones pequeñas al causar la depresión de endocría, mantener la variación genética potencialmente adaptativa en caracteres cuantitativos, y por medio de la erosión de la condicìn por acumulación de mutaciones levemente perjudiciales. En el presente estudio revisé e integré trabajos empíricos y teóricos recientes sobre mutaciones espontáneas y su papel en la viabilidad de las poblaciones y la planificación para la conservación. Se analizó tanto el mantenimiento de la variabilidad genética potencialmente adaptativa en los caracteres cuantitativos como el papel de las mutaciones perjudiciales en el incremento de riesgo de extinción de las poblaciones pequeñas. Experimentos recientes indican que la tasa de producción de varianza genética cuasineutral y potencialmente adaptativa en los caracteres cuantitativos es de un orden de magnitud menor que la varianza mutacional total debido a que las mutaciones con efectos fenotípicos pronunciados tienden a ser fuertemente perjudiciales. Esto implica que a efecto de mantener el potencial adaptativo normal en los caracteres cuantitativos bajo un balance entre mutación y deriva génica al azar (o entre mutación, deriva génica y selección natural estabilizadora), el tamaño poblacional efectivo debe ser de aproximadamente 5000 y no 500 (numero de Franklin-Soulé). Resultados teóricos recientes sugieren que el riesgo de extinción debido a la fijación de mutaciones levemente perjudiciales podría ser comparable en importancia a la estocasticidad ambiental y podría reducir substancialmente la viabilidad a largo plazo de las poblaciones con un tamaño poblacional efectivo de solo unos pocos miles. Estos descubrimientos sugieren que las metas de recuperación para muchas especies en peligro y amenazadas son inadecuadas para asegurar la viabilidad poblacional a largo plazo.

Gorillas Are Suffering From Bizarre Genetic Mutations Due To Inbreeding

This year, we learned that wildlife populations around the world have plummeted by 60 percent in just 40 years. Largely thanks to human activity, iconic creatures from vast elephants to teeny tiny bees are being lost at a troubling rate. But what effects do dwindling populations have on the animals themselves?

A new study, published in the journal Current Biology, has found that the critically endangered Grauer’s gorilla has lost so much genetic diversity in recent years that the apes are suffering from harmful genetic mutations. Essentially, fewer gorillas mean less diversity within a population and more inbreeding, leading to genetic problems in the next generation.

Also known as the eastern lowland gorilla, the Grauer’s gorilla (Gorilla beringei graueri) is a subspecies of eastern gorilla found only in the mountainous forests in the east of the Democratic Republic of Congo. It is currently listed as critically endangered by the International Union for Conservation of Nature (IUCN) thanks to human pressures like habitat destruction and poaching, which have caused populations to plummet by 80 percent in recent decades.

To see how this decline has affected the genetic health of the gorillas, a team of researchers from Uppsala University in Sweden and the Swedish Museum of Natural History analyzed the genomes of gorillas that lived up to 100 years ago and compared them with the genomes of those alive today.

Historical gorilla collections used in the study. Katerina Guschanski

They found that over the decades, Grauer’s gorillas have become more inbred and lost a significant amount of genetic diversity. Harmful genetic mutations that have increased in frequency include those linked to reduced disease resistance and male fertility. The team also found signs of mutations linked to a loss of function in the genes involved in the healthy development of fingers and toes, explaining why some of today’s gorillas appear to have webbed hands and feet as their digits are fused together.

“This recent increase in harmful mutations really emphasises the need to reverse the ongoing population decline in Grauer’s gorillas,” said Love Dalén of the Swedish Museum of Natural History in a statement.

The team also looked at the genomes of mountain gorillas, which are closely related to Grauer’s gorillas. However, they found that these gorillas haven’t experienced the same loss of genetic diversity and increase in harmful mutations. The authors posit that this might be because mountain gorillas have been very rare for thousands of years and have therefore lost harmful mutations thanks to natural selection, whereas Grauer’s gorilla numbers dramatically increased between 5,000 and 10,000 years ago.

Allowing the number of Grauer’s gorillas to expand is key to their conservation as it will allow their genetic diversity to increase once more. And there’s something you can do to help. Illegal mining for metals used in cell phones in the gorillas’ habitat is adding to their demise, so be sure to always recycle your old cell phones to lessen the demand.

3 Answers 3

Yes, "genetic engineering" can solve the problem.

First and most simply, you could just flip a coin and clone either the mother or the father (or make the gender their choice), thus zero mutations from the parent stock.

Second and more complex, the problem is mutation due to miscopying of a genome, or a poor choice of inheritance. With genetic engineering (even today) we can specify every letter of the genome and simply manufacture it that way.

For your purposes, an inventory of the genes of the ruling family can act as a library of allowable alleles (variants of a given gene) and gene segments, and a random (or intentional) selection of each part, given the founding library, could be the only allowable "persons" to be born.

In the event such a person turns out to have an undesirable trait due to some untested combination of allowable genes, Their combinations can be flagged as suspect, or if a combo appears multiple times as suspect, flagged as prohibited in the library.

The way to do this most efficiently is to bank sperm from the Most Excellent Ruler, and arrange that he be the father for serial generations, each with his own daughter / granddaughter.

Consider that 1st generation daughters are 50% most excellent, 2cond generation 75%, 3d is 87.5% and by 6th generation you are 98.44% Most Excellent Ruler genes. Of course a little scrambling might happen by meiosis etc but there will not be truly new genetic material introduced. These 6th generation descendants are like an inbred strain of lab mice, free of genetic disease and they may now breed freely with each other. Mutational events remain possible, of course and it might be safest to stick to the frozen sperm of the Most Excellent Ruler, thus reducing the chance for mutation by half. Do not worry about running out of banked sperm - he froze away an enormous amount.

A benefit from this method is that there can be many 1st generation daughters (of different mothers) and so the project can be done in parallel with many generations growing up at the same time. If the original mother has a disease causing recessive gene it will not be important because there is no chance for brother-sister matches and it will be diluted out along with her DNA.

In the (extremely unlikely!) event that the Most Excellent Ruler has a disease-causing recessive gene, the progeny carrying double recessives will become evident along the way. The dangerous gene will be identified and double recessives culled. Once you know that dangerous gene is present in the banked sperm one could correct it with genetic engineering by inserting the non-disease dominant copy of that gene from the Most Excellent Ruler. But the gene would persist in the banked sperm. Better and cheaper would be sorting out and disposing of sperm that carry that recessive gene.

I am not confident one can nondestructively identify various genes in sperm. Once the recessive genes of interest were known it might be necessary to grow embryos in vitro until a cell could be samples without destroying the embryo, to certify that dangerous genes were not present. In the unfortunate instance that the Ruler carried multiple recessive genes this would be cumbersome and would need to be done with each successive generation to avoid reintroducing those genes from the banked sperm. If that were the case (multiple recessives in the sperm) once one achieved near purity in the 6th or subsequent generation, one could allow a male to reach adulthood and be the new sperm donor for all subsequent generations. His genome (and sperm) would represent the genome of the Ruler purged of recessive genes.

Environmental Variance

Figure 5. The sex of the American alligator (Alligator mississippiensis) is determined by the temperature at which the eggs are incubated. Eggs incubated at 30°C produce females, and eggs incubated at 33°C produce males. (credit: Steve Hillebrand, USFWS)

Genes are not the only players involved in determining population variation. Other factors, such as the environment (Figure 5) also influence phenotypes. A beachgoer is likely to have darker skin than a city dweller, for example, due to regular exposure to the sun, an environmental factor. For some species, the environment determines some major characteristics, such as gender. For example, some turtles and other reptiles have temperature-dependent sex determination (TSD). TSD means that individuals develop into males if their eggs are incubated within a certain temperature range, or females at a different temperature range.

Geographic separation between populations can lead to differences in the phenotypic variation between those populations. We see such geographical variation between most populations and it can be significant. We can observe one type of geographic variation, a cline , as given species’ populations vary gradually across an ecological gradient. Species of warm-blooded animals, for example, tend to have larger bodies in the cooler climates closer to the earth’s poles, allowing them to better conserve heat. This is a latitudinal cline. Alternatively, flowering plants tend to bloom at different times depending on where they are along a mountain slope. This is an altitudinal cline.

If there is gene flow between the populations, the individuals will likely show gradual differences in phenotype along the cline. Restricted gene flow, alternatively can lead to abrupt differences, even speciation.

Selection Under Inbreeding

When dominance is presence, the selection response equations under inbreeding can become rather complex, they require additional variance components beyond the additive-genetic variance. Further, both transient and permanent components of response can arise. This chapter examines the theory of both the covariance of relatives under general inbreeding, as well as the expected selection response under inbreeding. Due to the decrease in the effective recombination rate under selfing, the Bulmer effect can be rather dramatic, as any linkage disequilibrium generated by selection is only weakly removed by recombination. Finally, this chapter also examines the evolutionary forces that interact to determine the selfing rate for a given population.

Oxford Scholarship Online requires a subscription or purchase to access the full text of books within the service. Public users can however freely search the site and view the abstracts and keywords for each book and chapter.

Please, subscribe or login to access full text content.

If you think you should have access to this title, please contact your librarian.

To troubleshoot, please check our FAQs , and if you can't find the answer there, please contact us .

Results and discussion

I) Potential advantages of inbreeding

We will hence start our reflection by asking ourselves what the advantages of inbreeding could be. If one carries out a simple literature search for the single keyword "inbreeding" on a server such as Google scholar, one can rapidly identify tens of thousands of citations. Upon rapid examination, it is actually striking to find that, in over 90% of those, the word inbreeding is systematically associated with either depression, cost or avoidance, compared to only a handful of papers where the potential benefits of inbreeding have actually been objectively considered. One important point to make here is that inbreeding is different from incest. Incest is the mating of extremely closely related individuals, usually sharing half of their genome (such as parent-child or brother-sister), or at least a quarter (such as grand-parent with grand-child). On the other hand, inbreeding results from the pairing of individuals that are more closely related than if they were picked at random from the surrounding population. What many studies have labelled as 'inbreeding avoidance' actually corresponded to 'incest avoidance', and we will see that, in many natural populations, although there are numerous examples of mechanisms to prevent selfing or incest, multiple strategies also exist that promote some degree of inbreeding.

I have actually identified so few papers that have constructively considered the positive aspects of inbreeding that it is possible to summarise them in just a few sentences. The notion that "selfing" is potentially advantageous can be traced back to R. Fisher in 1941 [11]. Around the same time, the works of S. Wright underlined that natural populations are seldom panmictic, but usually structured in partially subdivided, and more inbred demes. These divisions not only help to maintain more allelic and phenotypic diversity, but can also favour evolution and promote speciation [3, 4, 12]. In 1959, H. Carson put forward a model whereby speciation is promoted in small (marginal) inbred populations, whilst large, more outbred populations, will senesce, i.e. increasingly rely on heterosis, and progressively lose their capacity to evolve and to give rise to new 'young' species [5]. Many of the ideas developed in that article are very closely related to the ones I am presenting here. Because he adopted the view that speciation most often occurred through allopatry, later works by Carson focused on founder events, for which he is nowadays better known and this particular paper actually received surprisingly little attention from people trying to establish models of speciation (for example, it is not even cited in the book Speciation by C&O). Some twenty years later, based on the observation that quails mated preferentially with their cousins, P. Bateson produced the concept of optimal outbreeding [13–15], supported the following year by the work of Price and Waser on a wildflower [16]. Very soon afterwards, W. Shields put forward the theory that philopatry, i.e. the tendency of individuals of many species to breed near their birthplace, was related to the advantages conveyed by inbreeding, and in particular the capacity of inbreeding to maintain successful gene combinations [2]. Outside of the concept of crisis inbreeding developed by C. Grobbelaar in 1989 [17], and more recent works on the somewhat unexpected long term reproductive success of consanguineous marriages [18–20], I have so far failed to identify other works exploring the benefits of inbreeding that would contribute significantly to the ideas developed here (More recent but less directly relevant papers on the subject of inbreeding can be found in the 1993 book of collected works entitled 'Natural History of Inbreeding' [[21]] or in a 2006 paper by Kokko and Ots [[22]]). In the following pages, I will thus try to present and summarize the various advantages which can be found to inbreeding.

1) Inbreeding is necessary for the expression of advantageous recessive phenotypes

This undisputable advantage of inbreeding is the one which is most central to the model presented. In the first place, I thus felt that it was important to clearly define what is meant by dominant, recessive, co-recessive and co-dominant phenotypes. The laws of genetics initially discovered by Gregor Mendl at the end of the 19 th century concerned the transmission of characters in diploid organisms. Starting from homogenous stocks of peas, what he established was that all F1 had homogenous phenotypes (first law), but that those segregated in F2 generations, according to the well known 1/4 - 3/4 ratios for recessive versus dominant phenotypes (second law). A further observation was that different characters segregated independently from one another (third law). The considerations of linkage between genes and of genetic distance would be discovered by others, at the beginning of the 20 th century, after the 're-discovery' of Mendel's results.

Conversely to Mendelian genetics, which concern genes that remain identical through successive generations, the process of evolution involves mutations, which correspond to changes occurring in the DNA. Thus, starting from an ancestral genome, a new mutation will occur one day in the cellular lineage comprising the germline of an individual, and will only affect one strand of DNA. That mutation will thus be transmitted to some of the offspring (half at the most) in which it will be heterozygous. If the new mutation leads to a new phenotype, this new trait will only surface in the first generation of offspring having inherited the mutated DNA if it corresponds to a dominant character. If it is a recessive character, some degree of inbreeding between the descendants of that individual will be necessary for it to come to light.

Evolutionary "progress" is often perceived as the acquisition of new functions, resulting from mutations driving the appearance of new genes, or at least new functions in existing genes. Since the process of evolution is blind, however, new mutations will, much more often have a detrimental effect, if only because it is much easier to brake something that works than to create a new function from scratch. As early as 1930, Fisher had indeed realised that most new mutations are detrimental [23]. But this was even before the structure of DNA was known, and we now know that this is not quite true: most mutations actually occur in the silent DNA that surrounds genes, and thus have no detectable phenotype. Today, it is commonly acknowledged that, in humans, something of the order of 100 mutations take place every generation. Of those, the vast majority will occur in silent DNA, but somewhere between 3% and 1‰, i.e. between 3 and 0.1 per individual per generation will result in a detrimental phenotype, most of them through the inactivation of genes [24]. From work on laboratory strains of knock-out animals such as drosophila or transgenic mice, at least a third, and possibly as much as 50% of the mutations that result in the invalidation of genes, such as those interrupting an open reading frame, would actually be expected to be directly lethal in homozygotes, or to have such serious consequences that the homozygous bearer of such mutations would probably not go on to breed under natural conditions of selection.

Regarding mutations that actually result in new function, the proportion of those is difficult to evaluate precisely, but textbooks classically tell us that somewhere between one for every 10 4 and 10 5 new mutations will lead to new or different functions, i.e. one in every one hundred to one thousand individuals.

There is, however, a very important difference between mutations that inactivate genes, and those that result in new or different functions: in diploid individuals, having just one functional copy of a gene is very often sufficient, and most mutations that inactivate genes will thus be recessive, and thus have no detectable phenotype in heterozygous individuals. In a similar proportion of cases, however, there will be an effect of gene dosage, whereby individuals having lost one copy of the gene harbour an intermediate phenotype, and those mutations are then called co-recessive. Conversely, mutations that result in a gain of function will usually be dominant. The term co-dominant does not, however apply to mutations resulting in a gain of function with an effect of gene dosage (those are still co-recessive), but to mutations resulting in a change of function of a gene, where heterozygotes will thus express both functions, but homozygotes can only express one or the other. To clarify things, I have summarized those considerations in table 1.

Even if most recessive mutations correspond to alterations in the DNA that will result in the loss of a function, there are many cases, however, where losing a function can be advantageous for individuals. For example, losing certain patterns of colours can bring definite advantages to escape predators, such as the stripes of the African ancestor of zebras and horses. Those stripes were presumably very advantageous for remaining inconspicuous to predators in the savannah, but probably had the reverse effect for the early equidae that colonised more northern and greener latitudes and would later evolve into horses. As could already be suspected from the observations reported by Darwin in the 'Analogous Variations' section of The Origin, and later elegantly recounted by Stephen Jay Gould [25], crosses between various species of equidae, and more specifically between zebras and horses, reveal that the stripy phenotype is the dominant one. For the ancestors of horses to loose their stripes, significant inbreeding must therefore have occurred to express that recessive stripe-less phenotype, and similar reasoning could be applied for the loss of any dominant character that may have been selected for in ancestors, but was no longer beneficial, for whatever reason (climate modification, colonisation, evasion of an extinct predator or pathogen, sexual character that is no longer attractive. ).

Outside of the visible external phenotypes such as those considered in the previous paragraph, the capacity to resist infections by pathogens is another type of recessive trait which I perceive as particularly likely to play a major role in the selection of relatively inbred sub-populations. Most pathogens, and in particular viruses, do show high degrees of specificity for their hosts. This is due to the fact that pathogens use particular receptors to penetrate the body and/or the cells of their hosts. Infections by harmful pathogens will therefore eliminate individuals expressing that receptor, and select for organisms able to resist invasion because they carry mutated receptors to which that pathogen can no longer bind. Such characters of natural resistance are, however, usually recessive because heterozygous individuals will still carry one gene for a functional receptor, which will suffice to render those individuals susceptible to invasion by that pathogen. One particularly relevant example of this is the case of humans carrying the CCR5-Δ32 mutation, which, when homozygous, provides complete resistance to HIV infection, and an increased survival of a couple of years when heterozygous [26]. This delayed sickness would, incidentally, favour the spreading of HIV rather than be beneficial to the population, and thus bring a further advantage to the homozygotes for the CCR5-Δ32 allele. The geographic distribution of the mutant CCR5-Δ32 allele does suggest that this mutation arose several hundred years ago in northern Europe, and it is hypothesized that it was probably selected for because it provided resistance to a pathogen different from HIV, because the HIV epidemic only arose much later, in Africa [26].

Although the pressure of a particular pathogen can provide a very definite advantage to those individuals that can resist infection by that pathogen, the fact that this resistance will only be found in homozygotes would be a major hindrance for the spreading of that resistant allelic form to a whole population (something often referred to as Haldane's sieve), but would hugely favour particular subgroups where that allele would be homozygous, which could only occur through inbreeding. In addition to the fact that natural populations tend to be fragmented [4, 12], increased inbreeding will also result from increased selective pressures such as abrupt environmental changes or epidemics caused by very virulent pathogens, via a reduction in the effective size of populations. Under such conditions of increased strain, the individuals issued from groups harbouring advantageous recessive mutations will be endowed with a massive selective advantage. But the recessive nature of the characters that would be selected for under those conditions would provide the grounds for reinforcing breeding within the group rather than with members of the ancestral stock. Pushing this concept even further, Chris Grobbelaar actually proposed, over twenty years ago, the interesting idea that a mechanism of crisis inbreeding would be advantageous, whereby situations of stress would result in a shift from sexual preferences towards inbreeding [17].

2) Reducing the recombination load

One important concept in evolutionary genetics is that the fitness of individuals is not the result of a simple sum of functions harboured by each one of their genes, or loci, but that complex relationships exist between these different loci. For example, many phenotypes are epistatic: they result from particular associations of alleles carried by different genes. One of the major advantages of sex is that it will favour the shuffling of alleles between individuals, and thus promote the formation of such functional allelic combinations. It is commonly accepted that, if such associations of alleles from different genes are particularly advantageous, this can lead to the selection of co-adapted genomes. But, as outlined by S. Wright, "in a panmictic population, combinations are formed in one generation only to be broken up in the next" [3]. This dissociation of functional gene combinations is what is called the recombination load. And inbreeding is the only strategy that will reduce it, by allowing the maintenance of particular allelic combinations, albeit in only a portion of the offspring.

These aspects have been extensively developed and thoroughly documented by W. Shields in his book on the relationship between philopatry and inbreeding [2]: "One potential advantage of inbreeding, then, is that its genomic consequence of maintaining interlocus allele associations may permit more faithful transmission of coadapted genomes than would be possible with wider outbreeding".

From the point of view of the ideas developed here, advantageous allele associations are actually quite similar to recessive phenotypes, even if they are based on the association of dominant phenotypes. Indeed, once they have become fixed in a population, their fate will be threatened by hybridisation with an outside population that would not harbour those particular alleles. The threat would be less direct because, contrarily to recessive phenotypes, the advantageous association of two dominant alleles would still be present in all F1 individuals, but it would only be maintained in 9/16 of an F2 offspring, and in just 25% if the F1 matted with an individual from the outside population. On the other hand, the advantageous epistatic combination will be maintained in all future generations if the hybrid offspring backcrossed with the isolated population. If 'invaders' were rare, this would represent a very effective way for the introgression of genetic diversity into the isolated group, but under a more sustained presence of outsiders, we can see how the recombination load could promote the selection of reproductive barriers.

Alterations in the chromosomal structure also contribute very significantly to the recombination load (for example the case of a reciprocal translocation which will be depicted later (see Figure 2 and text relating to it). For such translocations, the general rule is basically the same as for epistatic combinations, with healthy F1 offspring. And the reduced fertility of those F1 effectively correspond to an extremely reduced fitness of those F2 individuals that do not inherit the right genetic combination. And similarly to advantageous gene combinations, once a particular chromosomal rearrangement has become fixed in a population, usually through inbreeding, the most effective way for the descendants of hybrid offspring to recover complete fertility will be by backcrossing with the isolated group. In cases where populations differ by several chromosomal rearrangements, however, hybridisation would become a real threat because the fertility of hybrid would be dramatically affected.

3) Fighting Muller's ratchet

A third advantage of inbreeding is that, for diploid organisms, it is the only effective way to fight off the accumulation of recessive deleterious mutations in their genomes. The notion that mutations accumulate inexorably in genomes over the course of generations is commonly referred to as Muller's ratchet [27]. Muller advocated that a major reason for the prominence of sexual reproduction among all animal species was due to the need to eliminate these mutations through genomic recombination. Following the views initially expressed by Fisher [23], Muller, in his early work on Drosophila, had documented himself that most new mutations tended to have recessive phenotypes. When it came to persistence of those in the genome over generations, however, he considered that all mutations were partially dominant (i.e. co-recessive, see table 1), and that even the most recessive deleterious mutations must have some slight effect (2 to 5%) on reproductive fitness [28]. Those weakly deleterious mutations would therefore be eliminated progressively over successive generations. Muller, however, carried out all of his work before the discovery of the structure of DNA and of how genes worked. Although his arguments were clearly valid for weakly deleterious co-recessive mutations, we now know that a very large proportion of deleterious mutations will be perfectly recessive and that mut/WT heterozygotes will show very little, if any, reduction in fitness compared to WT homozygotes [29, 30]. At any rate, even if inactivation of a fair portion of genes leads to co-recessive phenotypes through an effect of gene dosage, the frequency of deleterious mutations giving rise to completely recessive phenotypes will still be much higher than those leading to dominant, or co-dominant traits. Inbreeding, by promoting the conditions whereby recessive mutations can find themselves in a homozygous state, will hence allow the expression of those deleterious effects resulting from recessive mutations.

We will now attempt to compare the effects of accumulation of recessive deleterious mutations in populations undergoing various degrees of inbreeding and with a theoretical completely outbred population. As depicted in the upper panel of Figure 1, when breeding takes place between two individuals each carrying one copy of a defective essential gene, one quarter of their offspring will be either non viable, or very unfit because they will be homozygous for the deleterious mutation. If the mutation is truly recessive, the other three quarters will be perfectly viable, and two out of three among that viable offspring will be heterozygous for the mutation. The allelic frequency of that deleterious allele will hence pass from 0.5 in the parents to 0.33 in the offspring, and the mutation load from 1 to 0.66 mutations per individual.

Comparing the effects of accumulation of recessive deleterious mutations in populations undergoing various degrees of inbreeding, and with a theoretical completely outbred population. Panel A: Mendelian laws predict that when a crossing occurs between two individuals heterozygous for a recessive deleterious mutation, allelic frequency for that mutation drops from 0.5 in the parents to 0.33 in the offspring. Panel B: Evaluation of the fertility as a function of mutation loads and inbreeding coefficients. The thick red curve corresponds to the fertility predicted in a completely outbred population. It was drawn with the equation F = (1- M 2 .10 -8 ) 10,000 (see text). The thinner curves of different colours correspond to the fertility of crosses with a certain degree of inbreeding, as indicated on the figure. Those were calculated as F = (1- I) M , where F is the predicted fertility, M the average mutation load in the population, and I the inbreeding coefficient. In natural populations, the actual fertility would be a factor of those two theoretical degrees of fertility.

As a rough estimate based on the simplistic case of a single gene, one could therefore say that a rate of spontaneous mutation of 0.17 per generation (0.5 - 0.33) will be compensated by a reduction of 0.25 in fertility. This value of 0.17 is rather compatible with the various estimates of the rate of spontaneous mutations, which are, for humans, between 0.1 and 3 new deleterious mutation per genome per generation [24]. Although I realise that those figures are probably inaccurate for the additive effect of multiple genes, it was beyond my limited mathematical capacities to perform more precise calculations. I am confident, however, that others will later find such calculations rather straightforward, and it will then be particularly interesting to evaluate what types of equilibriums are reached for various mutations loads, various rates of mutations, and various effective sizes of population (i.e. various degrees of inbreeding).

In panel B of Figure 1, I have tried to evaluate the fertility of breeding pairs as a function of mutation loads and inbreeding coefficients. The theoretical fertility of breeding pairs in a population can be calculated as a function of M, the average number of recessive mutations per individual (i.e. the mutation load) and of I, the overall average inbreeding coefficient in that population (i.e. the probability that a locus taken at random in the genome will be homozygous by descent, corresponding to half the average degree of consanguinity of parents). The average fertility will then be (1- I) M . The different coloured curves were calculated for the indicated inbreeding coefficients, and we can see that fertilities only start to be significantly affected for populations with inbreeding coefficients > 0.01, corresponding to parents with degrees of consanguinity of 0.02, i.e. roughly that of third or fourth cousins.

In parallel, one can also evaluate the fertility of breeding pairs in an panmictic infinitely large population, where there is effectively no inbreeding (in real populations, the average fertility would actually be a factor of the two degrees of fertility). For mammals, if one estimates that roughly one third of genes are essential, this would amount to a total of approximately 10 4 essential genes. If the mutation load in the population is M, the probability of any locus being mutated will be M/10 4 , and the probability of carrying two mutated alleles of any given gene will be (M/10 4 ) 2 = M 2 .10 -8 and hence the effect on fertility would be (1- M 2 .10 -8 ) 10,000 overall since the threat applies for every single one of the 10,000 essential genes. This is represented as the thick red curve on panel B. We can see that, whilst the chance of carrying two inactivated copies of the same gene remains extremely low for mutation loads below 20, it starts becoming quite significant for mutation loads over 30, and fertility will drop below 75% when genomes have accumulated, on average, over 50 recessive mutations. For populations harbouring levels of consanguinity superior to 0.02, the reduction in fertility is, as could be expected, much more sensitive to mutation load, and for a population with an inbreeding coefficient of 0.06, a drop of fertility to 75% will occur with a mutation load between 5 and 6, but this figure climbs to nearly 30 mutations for an inbreeding coefficient of 0.01.

We have seen in the previous paragraph that, based on calculations for a single gene, a drop of 0.25 in fertility would keep up with the rate of 0.17 new deleterious recessive mutation, i.e. one per genome every six generations. The figures would possibly be slightly different if one considered the additive effect of multiples recessive deleterious mutations affecting different genes, each with lower allelic frequencies, and clearly different with different mutation rates. The mutation rate of 0.17 per generation cannot, however, be very far from reality since the most extreme estimates go from 0.1 to 3, and a decrease in fertility of 0.25 does not seem a completely unrealistic figure to keep up with new recessive mutations occurring once every six generations. For humans, this does not, however, mean that one in four newborn babies would come to the world with mental retardation or grievous physical defects. Indeed, most recessive mutations that touch essential genes would be expected to cause spontaneous premature abortions at very early stages of pregnancy, and many even before they would be recognised as miscarriages. From this point of view, it is actually rather striking to note that, in modern humans, miscarriages occur at a rate of somewhere between 10 and 40%. Whilst the occurrence of these miscarriages is clearly also related to other factors such as the age and the health of the mother, these figures suggest that it is not unreasonable to envisage that the price to pay to fight Muller's ratchet is that a fair proportion of the zygotes (say 20 to 30%) will have to be lost to compensate for the occurrence of one new recessive mutation every six generations. And these figures also seem compatible with what one sees in mice. Indeed, although mice can have as many as 10 to 12 pups in a litter, inbred strains are much less prolific, with litters often limited to 4 to 6 pups. When I have had to sacrifice pregnant female mice for experiments on embryonic tissues, I have often been struck by the proportion of aborted foetuses one can find in the uterus of a gestating female mouse, which is often near 50%. Thus, even in inbred mice in which the inherited mutation load must be close to zero, the rate of abortions suggests that de novo recessive mutations occur at a rate that is probably superior to one in six zygotes, or one in six generations.

The adjective "inbred" has clear derogatory connotations when referring to human beings and the commonly held perception about inbreeding is that it promotes degeneracy of the genome. Somewhat ironically, inbreeding actually results in "improving" the genome, and the fact that inbreeding results in elimination of recessive deleterious mutations from the population is actually well known, at least by animal or plant breeders and scientists: the extent of inbreeding depression decreases over successive generations of inbreeding [31] 1 (throughout the manuscript, subscript numbers refer the reader to footnotes which can be found in addendum 4, at the end of the manuscript). Via this type of phenomenon, the consequence of inbreeding will be that the allelic frequency of recessive mutations will be lower in the offspring than in their parents. For each mutation, the efficiency of the process is, however, remarkably low. Indeed, in the case of a heterozygous breeding pair, the allelic frequency for the mutated copy of the gene would only pass from 0.5 in the parents (each heterozygote for the deleterious allele), to 0.33 in the offspring (see Figure 1A). But inbreeding is the only practical way for the members of a species with an obligatory diploid genome to cleanse their genomes off the recessive mutations that will otherwise inexorably accumulate over successive generations until they reach an equilibrium, when the average number of recessive mutations in the genomes of individuals is sufficiently high that the rate at which they accumulate in the genome is balanced by a rate of elimination by random chance rather than by consanguineous descent (see Figure 1 and above text). The reason why I have used the word "practical" in the previous sentence is because of the bdelloid rotifers, the one undisputed example of asexual diploid organisms, that seem to have adopted an alternative strategy to sex to cleanse their diploid genomes from recessive mutations, but as discussed in addendum 1, it calls upon such extremes that it would be impractical for most other organisms. Haploid organisms such as prokaryotes do not have this problem of keeping their genome from accumulating deleterious mutations, because in haploids, all mutations are dominant, and deleterious ones will hence be eliminated very rapidly. Multiple cases exist in nature of the use of a haploid state by otherwise diploid eukaryotes, and in addendum 2, I have developed three such examples that I find particularly eloquent i.e. the cases of organisms that go through haploid stages, of the sexual chromosomes and of the endosymbiotic organelles.

Diploid genomes must have contributed greatly to the adaptive 'explosion' which took place among eukaryotes 1,5 billion years ago. The most important factor for this must have been the robustness of organisms, i.e. their newfound tolerance to new mutations that would have been instantaneously deleterious in haploid organisms. Conceivably, this may even have allowed the diploid organisms to "lower their guard", i.e. to reduce the fidelity of the replication of their DNA, and favour mechanisms of recombination [32], thereby favouring the appearance of novel adaptive mutations, helping them in particular to combat pathogens more efficiently, or to adapt to new environments. This view is supported by the fact that the vast majority of metazoans of today are obligatory diploids. The drawback of relying only on diploid genomes is that this also gives rise to the insidious type of Muller's ratchet I have just discussed, whereby recessive deleterious mutations can start accumulating silently in the genome of outbred individuals. Without sex, the benefits of a diploid genome would, thus, be very short lived, especially on the evolutionary time scale, and genomes would ultimately reach a mutational meltdown [33]. But sex without inbreeding is fraught with even more insidious, and thus far greater dangers that, as we will see, can ultimately lead to species extinction.

DNA replication is far from being a perfectly faithful process, and the rate of appearance of mutations in the genomes of vertebrates is commonly recognised to be of the order of 2.10 -8 per nucleotide for every generation, although the complete sequencing of the whole genomes of a family of four suggests it may be half as high [34]. For mammals, since their haploid genomes comprises roughly 3.10 9 base pairs, each diploid newborn will thus carry, on average, around 100 nucleotides that will differ from those it should have inherited from its parents if DNA replication was perfectly faithful, and if DNA was perfectly stable and completely resistant to damages by radiation and chemicals. Among those mutations, the vast majority will be silent, but, as developed on table 1 some will modify or inactivate gene functions, and most of those will be deleterious, but recessive.

In the long run, the phenomenon of evolution will be based mostly on the acquisition of new characters, corresponding to dominant mutations. But this can very easily be obscured by the much higher prevalence of recessive mutations. This can be ascertained by the repeated observations that the particular characters selected for in domestic species prove almost systematically to be recessive against the phenotype of the wild stock 2 . Even if DNA replication could be selected to become completely faithful, this would not be a solution, because, as famously underlined by Leigh van Valen [35], organisms have no choice but to evolve continuously in the face of natural selection, just like Lewis Carol's Red Queen, who needs to keep running just to stay in the same place.

But because evolution is blind, and occurs only by random mutations, in order to have a chance to see adaptive mutations arise, be they new functions or the advantageous loss of existing ones, there will be no avoiding the hundred fold excess of deleterious mutations, which will need to be eliminated by natural selection. As alluded to earlier, most of those deleterious mutations will, however, be perfectly recessive, i.e. they will have no phenotype in heterozygotes. Hence, within a large out-breeding population, the chance that one individual will carry two copies of an inactivated gene will be very low. But those will consequently be transmitted to half of the offspring, and over successive generations, since such mutations will keep accumulating, the mutation load will inexorably increase. Even at the lowest rate of the range envisaged above, i.e. one additional recessive mutation every ten generations, the mutation load will thus still increase rather rapidly until, as proposed by Muller [28], it reaches an equilibrium where as many mutations are eliminated at every generation than arise due to new spontaneous mutations. This process of elimination, which correlates directly with infertility, will, obviously, be greatly dependant on the inbreeding coefficient, i.e. on the effective size of the population. Above, I have tried to evaluate how the accumulation of recessive mutations in a population can affect the fertility of individuals as a function of the inbreeding coefficient in that population. From rather simplistic calculations, I conclude that, if the rate of accumulation of recessive mutations is of the order of one every six generations, this will be compensated by a drop in fertility of the order of 0.25. These figures, although rather speculative, seem to be compatible with the rates of spontaneous abortions one sees in human and mice, of which a fair proportion (I would guess between one and two thirds) are probably due to genetic causes. As already underlined by Muller 60 years ago [28], the proportion of miscarriages due to genetic defects necessary to keep the mutation load in a steady state will be principally dependent on the rate with which new mutations appear in the genome at every generation. The process of outbreeding will indeed reduce the initial frequency at which recessive mutations are found on both copies of a gene, but this advantage will only last for a while, until the mutation load has increased to levels where the decrease in fertility due to mutations once again compensates for the rate at which they appear. The advantage of outbreeding is thus very short lived on the evolutionary time scale. And, as mentioned earlier, I contend that it opens the door to a much greater threat. Indeed, if a large population undergoes extensive outbreeding for hundreds of generations, the equilibrium will only be reached when each individual carries, on average, several dozens of recessive mutations in its genome. If that population undergoes a sudden increase in selective pressures, for example because of a novel pathogen, of competition with another species, of a recrudescence in predators or of abrupt changes in the natural environment, the effective size of that population will shrink, and the inbreeding coefficient among the survivors will consequently become very significant 3 . If we imagine that the mutation load in such a large population had reached 40, and that the reduced numbers of individuals causes the inbreeding coefficient to rise to 0.03 in the remaining population, this will result in only 30% of viable zygotes. If we consider that this would happen under conditions where natural selection would be particularly harsh, the delayed cost of having avoided inbreeding for the short term benefits provided by outbreeding may well, in the long run, play a major role in the rapid extinction of that species, as well as reducing their capacity to colonise new environments (in the section 'convergence of character' of The Origin, Darwin himself remarked that 'When any species becomes very rare, close interbreeding will help to exterminate it'). In the face of Muller's ratchet, as Muller himself very rightly stated 60 years ago, "We cannot eat our cake today and have it tomorrow" [28] p150.

In cases where there is a relatively sudden shift in the pressures of natural selection, such as those caused by natural catastrophes (volcano, meteorites. ), or by a global change in the earth's temperature, the resulting shrinkage in effective populations sizes would thus be expected to be less well tolerated by the more prominent populations, i.e. probably those having taken full advantage of extensive outbreeding. Incidentally, such a mechanism would provide an explanation for the phenomenon of punctuated equilibrium proposed by Gould and Eldredge [36, 37]. Indeed, over periods of stability, the individuals of the most successful species will proliferate and colonise ever increasing territories. They will thus be the ones most likely to be found in the fossil record. But with this increase in effective sizes of populations will come the insidious consequence of increased mutation loads, and consequently the least chances to survive when unrest arises, causing dramatic reduction in the sizes of the populations. From this point of view, it is thus not surprising that, during periods when the natural scene changes, it should be the most numerous species, those found in the fossil record, that would struggle the most in the face of imposed inbreeding caused by population shrinkage, and become extinct with an apparent simultaneity.

4) Reducing the cost of sex

Another advantage of inbreeding is that it reduces the cost of sex. Indeed, in sexual reproduction, each parent passes only half of its genome to each of its offspring, which is directly related to the consideration that the cost of sex is two-fold [38], as compared to asexual reproduction, where each offspring inherits all of the parent's genome. But this factor of two is not quite a completely accurate measurement, if only because for most metazoans, sexual reproduction is obligatory and not an option. Furthermore, if we consider a hypothetical species with the most outbred population possible, each individual of that species will still be more genetically closely related to all the other individuals of the same species than to any other individual of a closely related species. In other words, all individuals of a given species share more common ancestors than they do with those of a closely related species. Hence when they breed within their own species, individuals do share some significant level of relatedness with their sexual partner compared with that of an individual of another species. So, even in a completely outbred population, because individuals of the same species will necessarily share some common ancestors, the cost of sex is never quite as high as two. And the more closely related an individual is to it's partner, the less that cost will be, for both of them 4 . Consequently, any evolutionary step that will favour inbreeding rather than producing offspring with more distantly related individuals, even of the same species, will thus reduce the cost of sex.

5) Inbreeding promotes population fragmentation, which can, in turn, promote collaborative or altruistic behaviour

From the point of view of the 'selfish gene' hypothesis [39], individuals should always favour their own interests, or at least those of closely related individuals [40, 41]. On the other hand, mathematical modelling has led certain population biologists to conclude that group level selection cannot work, and that for any behavioural trait to be selected, that trait must have a direct selective advantage for the individual. Such views are, however, much less prominent today, and anyone who is not convinced that group-level selection can play a major role in evolution should read the excellent recent review by Wilson and Wilson [42].

The type of reasoning which led to the rejection of group-selection was always based on the assumption that populations consist of large numbers of individuals breeding freely with the rest of the population. But, as underlined by Wright himself [12], natural populations are not like that. If we only look at the human population, although all individuals can theoretically breed with all those of the opposite sex with apparently equivalent efficiencies, we can see that the total human population is structured in ethnic groups, races, types, families. and that certain characters are more prominent in certain groups of individuals than in the rest of the population. In addition to the well recognised and very significant advantage of slowing down the spread of pathogens, and of favouring the maintenance of genetic diversity [4], population fragmentation has the other, much less direct and less obvious benefit of favouring the evolution of altruistic behaviours, by making group-level selection possible [42]. On the subject of group selection, I choose to adopt the view that, in fragmented populations, each group effectively becomes equivalent to a multi-cellular organism (see [43] for recent views on organismality). In metazoans, the fact that all the cells share the very same genetic makeup makes it possible for the vast majority of cells to sacrifice themselves either directly by apoptosis, or by differentiating into somatic cells that have absolutely no hope of generating offspring, for the benefit of the very few that will be destined to the germ line. Similarly, if a population is comprised of many small groups of individuals that are more closely related to one another than to the rest of the population, I firmly believe that it then becomes possible for natural selection to favour the evolution of collaborative or altruistic behaviours, because, in the end, even if those behaviours do not directly benefit the individuals that undertake those altruistic behaviours, the members of that group, and hence, on average, all the genes of the gene pool of that group, will fare better than those of the "group next door" that may have stuck with strictly selfish behaviours. On this subject, in 1871, Darwin himself made the following statement in his book "The Descent of Man":

It must not be forgotten that although a high standard of morality gives but a slight or no advantage to each individual man and his children over the other men of the same tribe. an increase in the number of well-endowed men and an advancement in the standard of morality will certainly give an immense advantage to one tribe over another.

Although, when one looks at natural populations, scores of examples can be found in all the kingdoms of life where altruistic, or at least collaborative behaviours have apparently been selected for, the questions linked to group level selection remain very contentious issues today. I know of no better example of cooperative altruistic behaviour than that of the lowly slime mould, Dictyostelium discoideum, and I contend that it is promoted by the ability of single cells to colonise new niches, resulting in fragmented populations. One of the reasons for which I find the example of Dictyostelium particularly telling is that it is not complicated by the intervention of sexual reproduction (see addendum 3 for more details).

In some cases, speciation could conceptually correspond to the need for populations having developed cooperative/altruistic strategies to fend off more selfish invaders. The issue of altruism is, however, really a side issue to the main focus of this essay. All I wish to say here is that, from an admittedly ultra-Darwinian point of view, the only realistic way to explain the evolution of cooperativity and altruism in natural populations is via group level selection, and this selection can only occur in populations that are fragmented into small groups of genetically inter-related individuals, or in other words, by natural selection acting on groups undergoing more inbreeding than if the population was considered as a whole. The fact that inbreeding can have the additional characteristic of providing a selective advantage at the levels of populations simply reinforces the view that inbreeding can and will occur and will not always be avoided. This will result in structured populations, which will, in turn contribute to the phenomenon of speciation.

6) Disadvantages of inbreeding

For the sake of fairness of argument, it seems necessary to counterbalance our arguments here, and underline that inbreeding also has several very significant disadvantages. Indeed, when starting from an outbred population, inbreeding depression will result in a high proportion of completely unfit offspring, and in most of the offspring being less fit than those from outbred breeding pairs. Another consequence of excessive inbreeding is that, by reducing the gene pool available for generating varied combinations of genotypes, it will result in less diversity, and thus in a more limited adaptability of the populations. Hence populations that undergo excessive inbreeding will be less likely to develop new functions than large populations undergoing outbreeding, where new functions bringing selective advantages can rapidly spread to the whole population, and can further combine with other advantageous functions that will have arisen independently in other individuals. Inbreeding may thus result in a slower rate of evolution.

This last argument does, however, need to be balanced by several counter-arguments. First, as we have seen previously, advantageous traits are not necessarily dominant, and those that are recessive can only come to light under some level of inbreeding. Thus, although inbreeding will reduce the probability of dominant traits spreading to whole populations, it will increase the frequency at which recessive traits appear, and since the mutations causing such traits are much more frequent than those causing novel functions this may balance the effect of inbreeding on slowing evolution. Second, when it comes to epistatic phenotypes resulting from advantageous gene combinations, we have seen that inbreeding is, once again, the only way to maintain them. Finally, as has been recognised for a long time, the rate at which characters can become fixed in populations is inversely correlated to the size of those populations [44]. By reducing the effective size of populations, the slower rate of evolution caused by inbreeding may thus also be compensated. As we will see later on, I actually contend that excessive inbreeding, leading to excessive speciation, will consequently result in the shorter lifespan of individual species, and thus in an accelerated rate of the species' turnover, which is not the same thing as the rate of evolution, although the two are too often considered equivalent.

Another potential disadvantage of excessive inbreeding is that it could result is reductions in the levels of polymorphism in a population, by provoking what would effectively amount to repetitive bottlenecks. For jawed vertebrates, which rely on polymorphism at the level of the MHC (Major Histocompatibility Complex) for fighting and eliminating infectious pathogens, this would be expected to have particularly nefarious consequences. As we will see later, however, comparing MHC polymorphism between related species reveals that inbreeding, and speciation, can apparently take place without losing healthy levels of polymorphism over the MHC region [45, 46], and presumably over most of the genome.

II) Focusing our reflections on what the ORIGIN of species could be

Or how can it sometimes be beneficial for a few individuals to breed preferentially among themselves rather than with the rest of the population, in others words with the ancestral stock?

We have thus underlined how inbreeding can have numerous advantages, and how systematic outbreeding is actually a strategy which has mostly short-term advantages, but that can lead to great drawbacks in the long run. I now propose to follow the path laid out by Darwin in the title of his book, and to focus on the very origin of species, i.e. to try to imagine what initial genetic event could eventually lead to the separation of a subgroup of individuals that will breed preferentially with one another rather than with the rest of the population.

Outside of the rather anecdotic cases of one step speciation via polyploidy (see C&O, p321), for the vast majority of metazoans, successive steps of progressive separation appear as more likely scenarios to reach speciation. But even if it does not result in instant speciation, an initial mutation must occur at some stage which will eventually result in promoting the interbreeding between individuals carrying that mutation rather than with the rest of the population. I have chosen to call such a process 'saeptation', from the latin word saeptum: barrier, envelope. In other words, saeptation will be the consequence of a mutation that will promote increased inbreeding within a group inheriting that mutation, and thus in a reduction of the gene flow between this new group and its immediate ancestral stock.

Lets us now envisage what type of mutation could eventually lead to saeptation. This new mutation will, one day, occur on one strand of DNA of one cell belonging to the germline, and hence be present in up to half of its gametes, and go on to be present on one chromosome of all the cells of some of its offspring.

1) Saeptation scenarios caused by a recessive mutation

As alluded to repeatedly in the previous paragraphs, I think the most likely scenario involves a recessive mutation as the very first step, i.e. the initial saeptation, which will end up promoting partial reproductive isolation of its bearers. The first reason for this is that, as outlined in table 1 outside of silent mutations, new mutations will most frequently lead to loss of functions, and will usually be recessive. But, as we have seen in the previous section, a loss of function does not necessarily mean a selective disadvantage.

Let us go back to the example of the horse precursors, and how they could have lost the stripes carried by their zebra-like ancestors. In the first place, to reveal the non-striped recessive phenotype, some significant inbreeding must have taken place. That inbreeding could actually have been promoted by the very fact that the group for which the stripe-less phenotype was advantageous was in the process of colonising more northern latitudes. Colonising populations, having smaller effective sizes, have consequently higher inbreeding coefficients [47], and we will see later that this is particularly relevant for the situations of island colonisation. Another consequence of the small size of such a group is that it will greatly facilitate the fixation of an advantageous recessive phenotype [44]. This isolation of a small relatively inbred group would hence result in reduction of the gene flow with the ancestral group because the adapted group would occupy a different territory. This would not, however, really represent a step of biological speciation, i.e. bona fide reproductive isolation, because if one individual of that adapted group ended up among individuals of the ancestral stock, it would probably breed with them very happily and efficiently, and the defining stripe-less phenotype would be diluted and only surface on very rare occasions (as we will see later on, this type of phenomenon actually happens in sticklebacks, which gain a selective advantage by losing their armour plates when they colonise freshwater environments). This type of situation would, however, lay the grounds for the evolution of further isolating characters because, in the context of their isolated group, it would be very disadvantageous for individuals to breed with stripy partners from the ancestral pool since all off their offspring would then end up with the dreaded stripes on their back, and thus be much more susceptible to becoming eliminated by predators.

Consequently, if an additional mutation took place in a member of that adapted group that led to more effective reproduction with kin than with individuals not carrying that second mutation, the inherent disadvantage of such a mutation due to the reduction of fertility with the rest of the adapted group would be balanced by a very significant advantage to its bearers because it would help prevent that sub-group of individuals from being re-invaded by the dominant but disadvantageous trait. This preferential mating with kin would also amount to promoting further inbreeding. This may be further facilitated by the fact that, when populations have previously gone through stages of significant inbreeding, the cost of inbreeding depression is very much reduced because most recessive deleterious mutations will have already been cleansed from the genome. Hence, from the above reasoning, we see that, in the context of an outbred population, a mutation that simply results in promoting inbreeding will struggle to become established because it would have many disadvantages to weigh against the advantage of reducing the cost of sex. But in the context of a group having undergone significant inbreeding, the safeguard of the mutation load against further inbreeding will have become much weaker, and under the selective pressure of the persistent threat posed by invasion by the ancestral stock, the probability of additional steps of saeptation within that group would thus be much higher.

2) Scenarios involving two mutations (Dobzhansky-Muller model)

To explain how mutations promoting reproductive isolation could ever appear in natural populations, Bateson (1909), Dobzhansky (1936) and Muller (1942) all came up with a similar hypothetical model, which is nowadays unjustly referred to as the Dobzhansky-Muller model (see C&O, p 269). This model calls upon the existence of two completely separated groups (allopatry), where two separate mutations take place that would each have no effect on the reproductive fitness in the group in which they arise, but that would result in incompatibility between the groups if and when those two groups are brought back in contact with one another. Such models are, however, not in line with Darwin's views that each step along the very long path of an evolutionary process must carry its own selective advantage. In the context of a group carrying a recessive advantageous mutation, however, we can see how the pressure of the outside populations, carrying dominant but disadvantageous alleles, could promote the selection of a mutation favouring reproductive isolation from the ancestral stock. At the end of the previous paragraph, I have argued that this selective pressure may be sufficient to promote further steps of saeptation, i.e. isolation from the other members within the adapted group, because the disadvantages of this mutation promoting inbreeding would be overcome by the advantage of resisting invasion by the dominant disadvantageous phenotype. And this modified tilt of the balance would be further favoured by the reduced inbreeding depression resulting from the relatively high level of inbreeding already present within that group.

Another scenario is, however, possible, which is to a certain degree related to the Dobzhansky-Muller model in that it would involve multiple steps, but those would occur in sequence, and not independently: the secondary steps of isolation would target traits specific to the saeptated population which could quite possibly be the one having driven the saeptation, but not necessarily. Indeed, during the initial phases of saeptation, inbreeding among a limited number of individuals would result in a high proportion of other genes becoming homozygous, and could thus reveal additional recessive phenotypes only rarely encountered in the ancestral population. In addition, in other genes than the one having driven the saeptation, certain alleles would have become much more frequent, either because they were genetically linked to the advantageous mutation, or simply because the smaller size of the population had favoured their drift towards fixation. For these three types of genes (additional recessive phenotype, genetically linked to the advantageous recessive mutation, gene having reached fixation by chance), the allelic frequencies would therefore be very different in the saeptated inbred population and in the ancestral one. And those would then represent as many potential targets for the selection of isolating mechanisms that would prevent the individuals of the saeptated group from mating back with the ancestral group. Technically speaking, this would, however, not represent saeptation, but reinforcement, because the mechanism of isolation would specifically target the outsiders, and not the direct ancestral stock, i.e. the isolated group. This type of scenario would thus involve two or more steps like the Dobzhansky-Muller model, but the fundamental difference with the Dobzhansky-Muller model is that selective pressures would be driving the isolation, rather than rely on chance for the separate evolution of two traits that will, at a later stage, turn out to be incompatible. One of the predictions inferred from the Dobzhansky-Muller model is that the rate of accumulation of reproductive barriers should increase with time, the so called "snowball effect" [48–50]. But this prediction does not actually allow to discriminate with the sympatric scenario described above. Indeed, if the threat of hybridisation is maintained throughout the speciation process, one would expect a similar snowball effect: once some degree of reproductive isolation has started accumulating between the two populations, resulting in reduced inclusive fitness of the hybrids further than the simple initial loss of the recessive advantageous phenotype (for various reasons including reduced fertility, intermediate maladaptive phenotypes, poor health, increased recombination load or even lethality), the cost of mating and/or breeding with the ancestral stock will have increased even more. Consequently, the pressure for selecting further mechanisms of reproductive isolation will also be increased, and one would thus expect the rate at which such traits are selected to go up, until such times when the two populations are sufficiently isolated that neither represents a significant threat for the other one.

3) Scenarios involving a dominant mutation

Lets us now consider whether a scenario can be envisaged whereby a dominant mutation would promote saeptation. The most obvious type of such a mutation would seem to be one that modifies the actual niche of the population, a phenomenon often referred to as ecological speciation. Indeed, if individuals carrying a novel mutation can start occupying new territories (geographical, seasonal, nutritional. ) they will, in this new territory, naturally find themselves in the presence of those other individuals carrying the same mutation, which will, by definition, be descended from the same ancestor, and will therefore be their close relatives (sibs or cousins). Since we are now talking about a dominant mutation, to allow the first individuals with the new mutation to find mates to reproduce, the initial separation between the adapted subgroup and the ancestral stock can, however, only be partial, and the possibility of hybridisation between the two groups must therefore be preserved. Although inbreeding among colonisers may carry an initial cost because of inbreeding depression, this could easily be offset by the advantage of the lack of competition in the new territory, and the inbreeding depression would only be transient, and recede after a few generations. Although, as we will see later, dominant mutations could play important roles in further steps of the speciation process, i.e. in reinforcement, it is thus hard to envisage how they could, on their own, promote the selection of reproductive barriers with the ancestral stock. In the case of a dominant mutation leading to the colonisation of a new niche, the increased inbreeding among the individuals carrying the mutation would, however, greatly increase the probability of revealing some additional recessive characters, of which some may turn out to be adaptive to the newly colonised environment. And those recessive mutations could, in turn, provide the grounds for a selective advantage to stop breeding with the ancestral stock.

4) The special cases of co-recessive characters, chromosomal translocations and reinforcement

4a) Co-recessive characters

Within the frame of the analyses carried out in the previous paragraphs, mutations that lead to hybrids harbouring intermediate co-recessive phenotypes (see table 1) would seem particularly prone to promoting speciation. Indeed, if such a mutation brings about an adaptive phenotype, such that the partial gain or the partial loss of a function makes it possible to colonise a new niche (warmer or colder climates, higher altitude, different food, different breeding time. ), the heterozygotes of the first few generations would be closely related to one another, but would be expressing intermediate phenotypes that would not separate them too much from the ancestral stock, and hence allow for the generation of multiple individuals. Crossing of those semi-adapted individuals with one another would be favoured by the fact that they would occupy that new niche. This would result in a quarter of their offspring becoming homozygous for the adaptive trait, which they would hence express more strongly, and would possibly be restricted to occupying only the newly colonised niche, with little or no possibility of contact with the ancestral one. The intermediate phenotype of the heterozygotes could thus be likened to some sort of stepping stone for the assembly of an isolated, necessarily more inbred group of individuals homozygous for the adaptive trait. Once that group has been constituted, in addition to the fact that the cost of sex would be higher with the outside group than within the group, a further advantage would be that additional adaptations to the new niche would probably be selected for quite rapidly, and the phenotype of the offspring that would result from encounters with the ancestral stock would very possibly make them unfit for either environment. This would thus provide the grounds for the Wallace effect, i.e. for the selection of further mutations reinforcing the reproductive isolation between the two populations. We can thus see how co-recessive traits could conceptually promote reproductive isolation even more rapidly than completely recessive ones.

Importantly, whether the mutation driving the saeptation is completely recessive or co-recessive could have significant consequences on the size of founder populations. Indeed, in the case of completely recessive mutations, those could stay completely silent for long periods of time within a population, and hence surface when crossings occur between individuals that are not necessarily very closely related to one another. In the case of a co-recessive mutation, however, the new intermediate character will be expressed in half the offspring of the founding individual, and the founding population will thus necessarily be comprised mostly by brother-sister matings, or close cousins at best. We will come back later to considerations regarding the size of founder populations and preservation of heterogeneity in the population.

4b) Chromosomal translocations

Chromosomes can be either circular, as in most bacteria and in endosymbiont organelles, or linear, as in all eukaryotes and a few bacteria. As far as I know, there are no known organisms with circular chromosomes that can carry out meiotic sexual reproduction, and all eukaryotes also have multiple chromosomes. Multiple linear chromosomes thus appear as a prerequisite to meiosis, with three chromosomes being the smallest number documented, in the fission yeast Schizosaccharomyces pombe (most species have several dozens, and up to several hundreds, or even over one thousand in certain ferns). One of the main reasons having driven the arrangement of the genetic information on such multiple and linear structures is almost certainly to promote one of the main purposes of sex, i.e. to achieve an efficient shuffling of the genes between individuals having evolved in parallel, via both inter- and intra-chromosomal recombination. Another commonly recognised advantage of this arrangement in metazoans is that the maintenance of telomeres provides a certain level of safeguard against the rogue selfish multiplication of cells that will lead to cancer. Outside of these two obvious advantages, I perceive that the arrangement of genomes on multiple linear chromosomes is also likely to play a central role in the phenomenon of speciation. Indeed, in line with the observation that even closely related species almost always differ in their chromosomal architecture, the role of chromosomal rearrangements in speciation has long been hypothesized (see C&O p 256-267, citing White 1978). One hurdle to this hypothesis, however, is that a chromosomal rearrangement such as the textbook example of a whole arm reciprocal translocation pictured in Figure 2 will result in a significant decrease of the fertility of the individuals in which this translocation occurs in the first place, with half of the zygotes predicted to be non viable when mating occurs with individuals of the rest of the population, which would not carry this translocation. Once the translocation has become fixed within a group, complete fertility will be restored to all individuals of that group. But for this to happen, heterozygous individuals carrying the same mutation will first have to mate with one another, and under such circumstances, the proportion of viable offspring is predicted to drop even a little bit more, from 1/2 to 3/8 (Figure 2), and this is without accounting for the inbreeding depression that would necessarily occur since those individuals would, logically, have to be closely related to one another. Furthermore, the translocation would then become homozygous in only 1/6 of their viable offspring (corresponding to 1/16 of the zygotes). Although other types of chromosomal remodelling, such as inversions or centromeric fusions, may not affect the proportion of viable offspring to the same extent as reciprocal translocations, some effect on the proportion of viable gametes would still be expected since such modifications are known to disturb the phenomenon of chromosomal pairing that takes place during meiosis [51].

Predicted chromosomal structures in zygotes issued from individuals carrying a whole arm reciprocal chromosomal translocation. In an individual carrying a reciprocal chromosomal translocation, only 50% of the offspring is viable (first line). If the cross takes place between two heterozygotes, the proportion of viable offspring drops to 6/16 (= 3/8). Once the translocation has become fixed in a population, crosses with the ancestral stock will generate a first generation (F1) that will be 100% viable, but those F1 individuals will be back to the situation of reduced fertility faced by the individuals who first carried the translocation, and this will be true whether they cross to individuals from the ancestral stock, or to individuals homozygous for the translocation.

Given the above considerations, it is difficult to see how chromosomal translocations could ever take hold in any population and reach fixation unless they were directly associated with a phenotype endowed with a very significant selective advantage. If that advantage corresponded to a dominant phenotype, the remodelled chromosomes could spread to the whole population. Many phenotypes associated to chromosomal remodelling would, however, be expected to get fixed via inbreeding rather than through a selective sweep. For example, a chromosomal modification could bring loci corresponding to an advantageous gene combination near to one another on the same DNA strand, and thus reduce the recombitional load. Many such genetically linked sets of genes can actually be found in the genome, for example in the MHC [52]. This genomic architecture can only have been the fruit of successive events of genomic remodelling, and the fixation of most of those must have required very significant inbreeding. Alternatively, one of the breakpoints may disrupt a gene, and this would be expected to lead to a recessive phenotype, which, once again, would only be expressed in the context on inbreeding.

In addition to the argument that even very closely related species usually do show significant differences in their chromosomal architecture, the view that chromosomal remodelling plays a significant role in speciation is also supported by the relatively high frequency at which chromosomal rearrangements do occur, and could thus conceivably be sufficiently frequent to occur even in small isolated groups undergoing saeptation. Indeed, systematic studies of human karyotypes have revealed that detectable neo-rearrangements occur at a frequency of approximately one in a thousand [53]. Whilst many of such rearrangements may result in spontaneous abortions (as many as 50% of human reproductive failures could be due to chromosomal abnormalities), many others will be viable, as testified by the fact that as many as one in 625 phenotypically healthy human beings carries a reciprocal chromosomal translocation [54]. Because those translocations do provoke significantly reduced fertility, unless they are linked to an advantageous phenotype, they are expected to get progressively eliminated from large outbreeding populations over successive generations. But finding them at such a sizeable frequency vouches for the fact that individuals carrying chromosomal rearrangements will occur quite often in humans, and hence probably in all species.

Another possibility to consider is that chromosomal rearrangements could be selected for as secondary saeptation steps, i.e. simply because they would reduce fertility of a saeptated group when they breed with the ancestral group, even if it would initially also involve some reduced fertility with the rest of that founder group. Conceptually, this decrease in fertility may sometimes represent a sufficient advantage to be selected for its own sake, as suggested by the observation that chromosomal rearrangements are more frequent between sympatric than between allopatric species of drosophila [55]. The recessive beneficial advantage would then be one of maintaining optimised fertility, but the process would certainly be much more direct, and thus favoured if the chromosomal translocation was directly associated to a mutated gene leading to an advantageous phenotype.

4c) The Wallace effect: Secondary steps towards speciation, i.e. reinforcement

Once a small group of individuals has 'sprouted' from the ancestral stock, if they have to keep expressing the recessive advantageous traits that drove the constitution of that group, breeding with the ancestral stock will represent a permanent threat for the welfare of their offspring, and the different sizes of the two groups will be a factor that greatly increases the weight of this threat (see [56]). If the initial mutation was directly linked to a chromosomal rearrangement, this would limit the gene flow between the two groups, but would actually further increase the threat because the hybrid offspring would be viable, but less fertile.

After an initial step of saeptation, further steps of reproductive isolation from the ancestral stock would therefore be clearly advantageous for that new, but much smaller group. Within the saeptated group, any further mutation that would increase reproductive isolation from the ancestral stock would therefore be expected to carry a very significant advantage, and could thus rapidly spread to the whole group, which the small size of the saeptated group would further favour.

We can now ask ourselves what sort of mutations and/or traits could intervene in the progressive establishment of completely separated populations, i.e. undetectable gene flow, such as what one witnesses between closely related groups recognised as separate species, although living side by side in natural environments. And I contend that, once a saeptated group has been constituted, in which individuals are all more closely related to one another than to the rest of the ancestral group, further steps of reproductive isolation will not necessarily have to rely on recessive mutations. In the previous paragraphs, I have argued that, in some circumstances, the selective pressure from the ancestral stock may be sufficient to promote further steps of saeptation within the isolated group, based on additional recessive mutations, which would be favoured by the increased inbreeding coefficient, and consequent low mutation load within that saeptated group. On the other hand, dominant traits would presumably spread to the group very rapidly, and would have the added advantage that the process would not require the elimination of the rest of the group. In the long run, as long as hybridisation with the ancestral stock remains a threat, any additional trait that significantly reduces the chance of producing offspring with members of that ancestral population could bring on a sufficient advantage to be selected for. As such, mechanisms that prevent either mating or the formation of zygotes (and hence called prezygotic isolation) such as sexual preference, occupation of niches more remote from the ancestor, gamete incompatibility or even culturally acquired traits could all contribute to protecting the newly formed group from the threat of breeding with the ancestral population. This type of reasoning, which assumes an asymmetric relationship between a newly formed group and a more numerous ancestral stock, provides an explanation for the observation first underlined by Muller in 1942 that incompatibilities between closely related species are very often asymmetric (C&O, p274).

When prezygotic isolation is not complete, and closely related species can still mate and produce zygotes, those hybrids are often found to be either non-viable, or fit, but sterile. Scenarios for the development of this type of barrier between species, which is called postzygotic isolation, are slightly more difficult to envisage because one needs to explain how, although mating has occurred and gametes used to generate zygotes, it can still be more advantageous not to produce offspring at all than to produce hybrids. For explaining this, however, I find one observation particularly useful: whilst problems of viability usually affect offspring of both sexes, problems of sterility usually follow Haldane's rule, and almost always affect only the heterogametic sex (C&O, p311-312). We can thus consider the problems of explaining hybrid lethality and hybrid sterility as completely separate cases of postzygotic isolation.

Regarding hybrid lethality, I can see two obvious reasons whereby it would be better not to produce offspring at all than to produce hybrids. First, if there is a significant cost to one or both parents for the rearing of offspring that will ultimately be unfit, it will be advantageous to save those resources for the subsequent rearing of "purebred" offspring. And second, if the hybrid offspring occupies a niche that overlaps with that of the purebred offspring, those two types of offspring would then be competing with one another. Sometimes, a further threat for the more inbred offspring could lie with the fact that the hybrids would be particularly fierce competitors for the occupation of the niche because they would benefit from hybrid vigour, and it would thus be best not to produce that hybrid offspring at all.

Regarding the phenomenon of hybrid sterility, I can see three ways whereby it can be promoted, which are not mutually exclusive.

1) The first one lies with chromosomal rearrangements. As already mentioned in the previous pages, chromosomal rearrangements are very often associated to phenomena of speciation, and even closely related species are often found to diverge by several chromosomal structural differences. Although hybrids carrying a single chromosomal translocation such as the one depicted on Figure 2 will only see their fertility drop by 50% when they mate with homozygous individuals of either type, this proportion will drop further for every additional chromosomal rearrangement and soon reach figures approaching zero. A factor further contributing to sterility is the observation that chromosome pairing has been found to be a necessary step for the proper completion of meiosis, at least in eutherian mammals (C&O p 262-264, citing Searle1993). As we have seen in the previous pages, the fixation of such rearrangements would be most likely to occur when they are directly linked to an advantageous phenotype. The observation that there are more differences in chromosomal architecture between drosophila species living in sympatry that in allopatry [55] does, however, suggest that the reduced fertility provided by such rearrangements may sometimes represent a sufficient advantage per se.

2) The second reason lies with the haploid nature of the sex chromosomes in the heterogametic sex (see addendum 2). As already discussed earlier (section II-3), following a process of saeptation, the allelic frequencies of many genes in the newly formed group would be expected to be significantly different from that in the ancestral population. Similarly to what was discussed above, those genes, whether carried by autosomes or sexual chromosomes, would thus represent potential targets for the selection of new mutations carried by the sexual chromosomes: newly mutated genes would still function well with the genotypes frequently present in the isolated group, but would no longer work in combination with the genotypes prominent in the ancestral stock. This would be particularly likely for the heterogametic sex because any mutation carried by one or the other of the sex chromosomes, even those corresponding to a loss of function, would be immediately dominant, as already underlined by Muller in 1940, and formalised as the dominance theory put forward by Turelli and Orr [57]. Since sexual chromosomes are, necessarily, endowed with many genes related to sexual reproduction, a likely phenotype resulting from such a selective process would be one affecting the sexual capacities, and hence result in the sterility of the heterogametic sex. Alternatively, the genes involved in the reproductive isolation may be part of the large number of genes carried by the chromosomes which are diploid in half the individuals (X in mammals and flies or Z in certain insects, fish, reptiles and birds. For the sake of clarity and simplicity, I will use X as an example for the rest of this paragraph, but I could just as well have used Z). Lets us now envisage that a mutation takes place on a gene carried by the X chromosome, such that the gene product will still function well with the allelic form of some other gene found at high frequency in the saeptated group threatened by hybridisation, but will no longer function with the allelic form(s) found in the ancestral group. As long as the individuals of the group breed among one another, that mutation would have no detectable effect, and would thus not really have any reason to spread to the whole group. But if hybridisation with the ancestral stock took place, because this mutation corresponds to a loss of function, it will most of the time result in a recessive phenotype, and it would thus have the typical characteristics of X-linked deficiencies, i.e. be silent in diploid female offspring, and expressed in the hemizygous males. The X chromosome carries many genes involved in vital functions, and disabling of those would presumably result in lethal phenotypes. Under the threat of generating hybrid offspring with an outside group, the individuals carrying such mutations would then be endowed with a definite advantage that would explain how, although neutral within the saeptated population, such mutations could be driven to fixation in the group undergoing speciation. The above scenarios would thus explain why phenotypes of reproductive isolation are often asymmetric, why they are often stronger in situations of sympatry, and provide potential explanations for Haldane's rule, i.e. why, when inter-species crosses take place, if only one sex is affected, it is usually the heterogametic one that is either non-viable [58], which I contend could often occur by recessive mutations of vital genes on the X chromosome, or sterile, by mutations of genes involved in sexual reproduction carried either by the Y or the X chromosome.

3) The third reason for which hybrid sterility may be selected for lies with the fact that sexual reproduction is usually much more costly for females than for males, with the latter having the capacity to produce virtually unlimited numbers of offspring. In the case where a population undergoing speciation competes with the ancestral stock for the occupation of a niche, I contend that the generation of hybrids where females are fit and fertile, but males are unfit can represent an extremely advantageous strategy. These aspects will be developed further in section IV.

III) There is probably seldom such a thing as truly allopatric speciation

In the previous section, we have seen how advantageous recessive traits could promote the formation of small saeptated groups within large populations, and how the need to keep expressing those recessive phenotypes could subsequently drive reinforcement, i.e. further steps of reproductive isolation, based on a whole array of different mechanisms. The recurring theme of the reasoning developed in the previous pages is that reproductive isolation would not arise as a bystander effect of divergent evolution, but would be directly selected for under the pressure of an outside group, most frequently the immediate ancestral population. Even if today, the majority of evolutionary scientists believe that most events of speciation must have occurred in allopatry, I do actually believe that if truly allopatric speciation ever happens, i.e. for whole populations to drift apart sufficiently to become infertile with one another, it must be an extremely slow process, and consequently a very rare occurrence. Indeed, if populations of individuals are completely separated, there will be no selective pressure for evolving features that will further reduce gene flow between the two groups, because the gene flow will already be non extant. If the geographical barrier is later lifted, the features of the individuals in each group will almost certainly be quite different because they will have adapted to their respective environment. Some mechanisms of preference between similar phenotypes may favour reproduction among the individuals having co-evolved, but since there will have been no selective pressure, I contend that there would be no reason why the individuals from either group should have become infertile with those of the other group. This is in fact in complete agreement with what has been very recently described for Caribbean Anoles lizards. Those have evolved independently for millions of years on separate islands that only joined relatively recently to form the large island of Martinique, and more reproductive barriers appear to have been selected for between populations that have evolved side by side to adapt to coastal or mountainous conditions than between those that have evolved on separate islands [59].

This is also exactly what happens with domesticated species. Under conditions of domestication, species can diverge to become very noticeably different, and reproduce for scores of generations under very divergent conditions of selection, yet they do not become infertile with one another. In this regard, I find the example of dog breeds particularly telling. Upon comparing the skeletons of a great Dane and of a Chihuahua, or of a Dachshund and a Saint-Bernard, no taxonomist in their right mind would ever place them as belonging to the same species. Yet, when my steps take me to public parks or other places where people go to let their four legged friends relieve their natural needs, I am often struck (and amused) to see how dogs of very different sizes and appearances can still recognise one another as potential sexual partners. And we do know that they do indeed belong to the same species. They all share exactly the same chromosomal architecture as wild wolves. In fact, if all these dogs of different sizes were placed in a giant enclosure and fed regularly, some sexual preferences between certain types may surface (see long citation of Wallace's book in section V), pregnancies between small females and large males may turn out to be fatal for the mothers, and the smaller males would probably not fare too well in fights with larger ones, but in the end, all those dogs would produce extremely fit offspring that would certainly be much more homogenous than the starting population, and would almost certainly contain genes inherited both from the Chihuahuas and the great Danes. I contend that, if domesticated species do not undergo speciation, it is because the process of selection is carried out by the breeders, and not by natural selection. Under natural conditions, individuals, and groups of individuals, compete directly with one another for the production of offspring and the occupation of a niche, and loosing this competition means dying with no offspring.

In settings of domestication, even if most characters that are selected by the breeders are recessive, and could even sometimes be associated to chromosomal rearrangements, there is never any direct pressure for individuals to stop breeding with the ancestral stock, and there can thus be no selection for either saeptation, or reinforcement. The fact that different domestic breeds, including dogs and pigeons, have now been maintained in effective allopatry, i.e. in complete separation from one another for hundreds of generations without any discernible sign of speciation ever being witnessed is, in my eyes, one of the stronger arguments against the possibility that allopatric speciation, resulting from divergent selection and/or genetic drift, could play a significant role in the phenomena of speciation that are clearly taking place continuously in the natural world.

Another argument against the role of intrinsic genetic incompatibility resulting from a random process in the evolution of reproductive isolation can be found in comparing the estimations of lifetime of species, and of the time it takes for such incompatibilities to develop. Indeed, for both mammals and birds, the fossil record tells us that the average time of existence of a species is around one million years [35], whereas the time it takes for the genomes of mammals to diverge sufficiently to become genetically incompatible is estimated to be around 2-4 million years [60], and well over 10 million years for birds [61]. Given those numbers, one can note that there is a flagrant inconsistency between the biological data and the fossil record since one would have to envisage that most taxonomic species would become extinct before they would have a chance of evolving into genetically incompatible species. I perceive this as a strong argument against the idea that allopatric (and hence passive) genetic divergence could be the main factor responsible for speciation.

Detractors of the views expressed in this essay would not fail to point out that there are many documented examples of allopatric speciation, i.e. where groups of individuals that were geographically separated have become "good species", i.e. completely infertile with one another. But to counter this argument, we only need to think back to the ancestral species, the one which is presumed to have occupied the ancestral territory, and colonised the new one (or, as proposed by Darwin, become split in two by a rising mountain range). If the two modern species cannot breed with one another, then we can safely assume that at least one of the two would also have been infertile with the ancestral species. But, by definition, individuals of that ancestral species were initially present on the two territories, and that species cannot have disappeared before the appearance of a subgroup of individuals that were less fertile with the ancestral individuals, and would eventually lead to the modern species. The logical consequence of this point of view is that, when allopatric speciation appears to have occurred, it actually probably corresponds to several successive steps of 'sympatric' saeptation, with the new, better adapted group replacing the ancestral intermediate.

The most striking examples of speciation often occur on islands, and when Charles Darwin visited the Galapagos in the course of his voyage on The Beagle, the observation of all the very unusual specimen found on those remote islands would later on help him greatly to formulate his theory of evolution, as well as to consider the idea that geographic isolation could contribute to speciation because of the independent evolution of populations that would progressively become infertile with one another.

Let us now consider the phenomenon of island speciation from the point of view developed in the previous paragraphs, i.e. that speciation occurs mostly as a consequence of natural selection, in other words in a context where it is advantageous for subgroups of individuals to stop breeding with the ancestral stock. Colonisation of islands are, inherently, very rare events, and even more so for an obligatory sexual species because this implies that at least two individuals from opposite sexes find themselves on the same island at the same time, which could, quite often, be brothers and sisters descended from a single pregnant female. The initial population will, consequently, go through a very tight bottleneck, with extreme degrees of inbreeding. The resulting reduced fitness of the individuals may, however, be well tolerated because, in the newly colonised territories, those few individuals will have no competition from kin, and presumably very few predators and pathogens adapted to them. Because of this initial episode of inbreeding, however, the cost of subsequent inbreeding will be expected to become much reduced after just a few generations, and this population of colonisers would then presumably multiply quite rapidly to occupy its newfound niche. But the characters of the ancestral stock would probably not be best adapted to their new environment, and conditions would thus seem very favourable for the selection of new characters allowing them to adapt. As we have seen before, mutations leading to recessive characters are much more frequent than dominant ones. And these would be even more likely to come to light in the envisaged conditions, where inbreeding would be favoured both by the small size of the population, and by the fact that inbreeding depression would be minimal. Hence, if a recessive mutation occurred that brought on an adaptive advantage to the new environment of the colonised island, there would be a very significant advantage for the individuals carrying the adapted, recessive, phenotype, to reduce their breeding with the rest of the colonising group. Any mutation coming to reinforce that saeptation would thus be advantageous, and would not necessarily have to be recessive itself. Hence, mechanisms reinforcing the isolation of the adapted group from the rest of the population, such as traits of genetic or post-natally inherited sexual preference, gametic incompatibility, genomic incompatibility or chromosomal rearrangements could evolve within that group, whereas the initial selection of such traits is normally not favoured in larger, more outbred populations, where inbreeding depression is high.

The picture we get from the above scenario is one where, when a secluded niche, such as an island, is initially invaded by very few individuals, successive steps of saeptation and/or reinforcement among a few adapted individuals will be greatly favoured by the initial inbreeding episode. And at every step, the better-adapted descendants of that group would most probably wipe out the less-well adapted stock of their immediate ancestors. For every one of these steps, the reduction of gene flow with the immediate ancestors would not necessarily be very high but, although that ancestral stock would have long been eliminated from the island, each one of those steps would reduce the fertility between the population of adapted individuals and their immediate ancestors, and consequently would be expected to have a cumulative effect on the fertility between the adapted population and the ancestral stock. Hence, if the population of individuals that have adapted to the island through successive steps of saeptation and/or reinforcement was ever brought back in contact with the more numerous, outbreeding population which stayed on the continent, individuals from those two groups would very probably be completely infertile with one another, even if the latter one had not evolved away much from the ancestral stock. The speciation process so witnessed would, however, not really have occurred in allopatry, but as a succession of sympatric steps which can only occur under the selective pressure of the immediate ancestral stock. An argument that supports the validity of this type of reasoning is the recurrent observation that events of speciation seem especially prone to occur in the context of small populations, such as those promoted by small islands. The size of the niche itself (for example a small island, or a small lake) could indeed be the main factor contributing to the maintenance of a relatively high degree of inbreeding, and hence to the reduced level of inbreeding depression that can promote speciation. Thus, even in the context of islands that are not completely isolated from the regular invasion by individuals from the mainland (such as the Baleares, the Caribbean or the Canaries), or from other nearby islands (such as the Galapagos), small islands have been found to be particularly propitious to speciation in all sorts of genera (birds, lizards, mammals, insects. ).

To conclude this section, I would say that, for most cases considered as undisputable examples of allopatric speciation, the times of separation are often much longer than the expected lifetime of the species considered. Also, since in most cases ancestor and speciating groups probably co-exist for much less time than the lifetime of species, it is not surprising that so few cases of speciation appear sympatric. But it is not because we do not see it happen that sympatric speciation does not happen. Thus, contrarily to the stance proposed by Coyne and Orr, I contend that allopatric speciation should not be considered as the default mode (C&O, p84). Rather, to prove that truly allopatric speciation has ever taken place, I advocate that one would have to demonstrate that no step of saeptation has taken place during the evolutionary process, whereby one sub-population would have become reproductively isolated from its immediate sympatric ancestor, and subsequently eliminated it.

IV) What relationship can be expected between the different modes of speciation, the mechanisms of reproductive isolation that are being selected for, and the diversity of the newly separated population? 5

Despite the arguments presented in the previous section, there is no denying that the conditions under which speciation occurs (sympatry, parapatry, allopatry) would be likely to play important roles on both what types of reproductive isolation mechanisms are being selected for, and on the size and diversity of the founding population that will ultimately result from the speciation process. In Figure 3, I have drawn simplistic sketches that would correspond to scenarios of speciation occurring in those three conditions. In this drawing, the shapes represent the niche occupied by a population. I feel that an important point to underline regarding the nature of niches is that they are not solely linked to geographical constraints, but to many other factors such as the nature of the nutrients, the timing of the life cycle, the identity of other partner species such as pollinators for plants, or hosts for parasites, etc. All in all, I perceive that the defining point between parapatry and sympatry is whether the niches of two populations undergoing speciation are sufficiently non overlapping that neither could ever wipe out the other one. On the other hand, even if two groups have such different life styles or life cycles that they seldom breed with one another, but still compete for the very same food, or for the same territory, one could fully expect that one of the two protagonists will, sooner or later, inherit a new character allowing it to eliminate the other one completely. In short, when occupation of the niche equates to competition for survival, I will call this sympatry if the two populations can exist side by side without one ever being wiped out by the other one, I will call this parapatry and when the two populations have so few interactions that neither is a threat for the other one, I will call this allopatry.

Watch the video: Disruption - Day 2 - Part 2 ENG (January 2023).