The selfish gene revisited:

reconciliation of Williams-Dawkins and conventional definitions

MIT Press allows self-archiving of pre-print versions. This is the submitted manuscript version, which was subject to editorial copy-editing prior to final publication, without change in substance. This is the stylistically most authentic and, I believe, clearest, version.

Donald R. Forsdyke  (2011) Biological Theory 5, 246-255















ABSTRACT  Sightings of the revolutionary comet that appeared in the skies of evolutionary biology in the 1976 – the selfish gene – date back to the nineteenth and early twentieth century. It became generally recognized that genes were located on chromosomes and compete with each other in a manner consistent with the later appellation “selfish.” Chromosomes were seen as disruptable by the apparently random “cut and paste” process known as recombination. But each gene was only a small part of its chromosome. On a statistical basis a gene should escape disruption for many generations. This led George Williams and Richard Dawkins to a new definition of the gene differing from conventional biochemical definitions in that there were no consistent genic boundaries. There had been no previous sightings of another revolutionary, albeit less verbally spectacular, comet that appeared in 1975 – the homostability principle of Akiyoshi Wada. Each gene has a base composition “accent” that distinguishes it from its neighbours. We now see that recombination can be triggered by the shift in base composition at genic boundaries. Hence, the Williams-Dawkins definition approaches the conventional.

Keywords: Base composition, Chromosome, Crossing-over, Gene definition, Homostability, Meiosis, Recombination

While it is true that an egg is only a hen’s way of making another hen, it is also true, as pointed out by Samuel Butler (1878), that a hen is only an egg’s way of making another egg. Whether Butler considered one as more selfish than the other, we do not know. But Richard Dawkins (1976) left little doubt that, when he wrote of a body being only a gene’s way of making another gene, it was to the gene that selfishness should be attributed. By this he meant that, when weighing the benefits conferred by an evolutionary process, these could often be seen as having been conferred on genes, rather than on an individual organism or group of organisms. The powerful impact of this concept on Dawkin’s contemporaries has recently been related (Grafen and Ridley 2007). The gene was newly defined in a way that appeared to distinguish it both from the classical gene of Mendel and the molecular gene of biochemists (Griffiths and Stotz 2006; Forsdyke 2009; Stotz 2009). The present paper attempts to reconcile these definitions through new interpretations of historical antecedents and of the mass of DNA sequence information that has emerged from various genome projects. The lines of development that preceded and followed Dawkins’ remarkable work are here considered for three periods: the period when Mendel’s ideas were available but unrecognized (1866-1899), the period when the ideas gained general acceptance (1900-1926), and the period when genes were biochemically characterized (1927 onwards).


Darwin missed Mendel, not Sanderson and Beale

Although Mendel’s famous paper was published in 1866, its importance was not recognized until 1900 when the Mendelian revolution began (Cock and Forsdyke 2008). But another 1866 publication was brought to Charles Darwin’s attention by an article in Gardeners’ Weekly (Beale 1866; Darwin 1868: 378). This was a Royal Commission Report on the cattle plague – now known to be caused by a virus – that was then ravaging Great Britain (Spencer et al. 1866; Romano 2002). The physician John Burdon Sanderson related how he had transmitted the disease to healthy cattle by injecting minute quantities of blood from infected cattle. The power of transmission was not decreased by diluting the blood.

    Another contributor, the pathologist Lionel Beale, concluded: “With regard to the nature of the contagium itself, evidence has been adduced to show that it consists of very minute particles of matter in a living state, each capable of growing and multiplying rapidly when placed under favourable conditions.” In earlier testimony to the Commission, the Medical Officer to the Privy Council had declared (Simon 1865):


Hourly observation tells us that the contagium of smallpox will breed smallpox, that the contagium of typhus will breed typhus, that the contagium of syphilis will breed syphilis, and so forth, - that the process is as regular as that by which dog breeds dog, and cat cat, as exclusive as that by which dog never breeds cat, nor cat dog.

These observations, which extended Louis Pasteur’s work on the, then highly controversial, germ theory of disease (O’Malley 2009), led to Darwin’s “provisional hypothesis of pangenesis” (Darwin 1868: 357-404; Forsdyke 2001). He proposed that the individual characters of an organism could be transmitted through the generations in the form of minute “gemmules” – now best equated with genes. These were “capable of largely multiplying themselves by self-division, like independent organisms” and were part of the “formative matter which is contained within the spermatozoa.” Thus, “the child, strictly speaking, does not grow into the man, but includes germs which slowly and successively become developed and form the man.”

    Given that a “contagium” could spread from animal to animal, it was easy to believe that a gemmule, after appropriate somatic education, might leave its cell of origin and transfer to the germ line, so that parental experience could be transmitted to offspring. While this Lamarckian aspect of pangenesis gained little support, the idea that the “gemmules,” or “pangens,” could correspond to individual characters, was deemed sound by Hugo de Vries (1889).

    Darwin’s mathematician son, George, had calculated the minute size of the inorganic molecules that chemists were then studying and Darwin noted (Darwin 1875: 373-374): “No doubt the [organic] molecules of which an organism are formed are larger, from being more complex, than those of an inorganic substance, and probably many molecules go to the formation of a gemmule; but … we can see what a vast number of gemmules … [an organism] might contain.” He concluded that: “Each living creature must be looked at as a microcosm – a little universe, formed of a host of self-propagating organisms, inconceivably minute and numerous as the stars in heaven” (Darwin 1868: 404).


Competition between gemmules

Although unaware of Mendel, many biologists at that time knew that in crosses between dissimilar types certain characters (e.g. tallness in peas) were dominant (“prepotent”) over others (e.g. smallness in peas; Roberts 1929: 170). When considering prepotency, Darwin (1868: 386, 395) thought that the gemmules of one parent might “have some advantage in number, affinity, or vigour over those derived from the other parent.” He also thought that new gemmules might preferentially multiply “until at last they become sufficiently numerous to overpower and supplant the old gemmules,” thus being selectively transferred to future generations. Thomas Huxley noted in 1869 that, just as organisms within a common environment might compete to bring about Darwinian natural selection, so organic molecules within an organism might compete likewise (Huxley 1893: 115):


It is a probable hypothesis, that what the world is to organisms in general, each organism is to the molecules of which it is composed. Multitudes of these, having diverse tendencies, are competing with each other for opportunities to exist and multiply; and the organism, as a whole, is as much a product of the molecules which are victorious as the Fauna or Flora of a country is the product of the victorious beings in it. On this hypothesis, heredity transmission is the result of the victory of particular molecules contained in the impregnated germ. Adaptation to conditions is the result of the favouring of the multiplication of those molecules whose organizing tendencies are most in harmony with such conditions.

 More colourfully, Butler (1880) commented on the protoplasm within cells that, to the microscopes then available, appeared relatively structureless, yet contained warring elements somewhat like those in a region that approximates to modern-day Afghanistan:


Protoplasm is, for all its structurelessness, composed of an infinite number of living molecules, each of them with hopes and fears of its own, and all dwelling together like Tekke Turcomans, of whom we read that they live for plunder only, and that each man of them is entirely independent, acknowledging no constituted authority.

The case was made more formally by the embryologist Wilhelm Roux (1881). To the end of his life, Huxley (1893: vi) persisted in his belief in a “struggle for existence within the organism.”

    Although not agreeing with the Lamarckian aspects of his cousin’s pangenesis hypothesis, Francis Galton accepted that an organism could be viewed as a constellation of characters for each of which there could be a corresponding gemmule, which he referred to as an “element.” In Germany, August Weismann used the term “id.” Galton supposed that, on average, a child would receive half of its elements from each parent, who in turn would each have received half of their elements from each of their parents. But what determined which half of a parent’s set of elements would be transmitted? He envisioned a competition among elements for a limited number of places in the gametes that would move forward to the next generation (Galton 1894):


As we know nothing about the arrangements and movements of the ultimate living units or germs, we can answer only by analogies. The exact answer would require a knowledge of the cause of what, in the nomenclature of Weismann, would be called the architecture of the id [structure of the gene], and of which he assumes the existence, but does not attempt to account for. We know that the germ contains the seeds of a vast number of ancestral potentialities, only a very few of which can be simultaneously developed, being to a great extent mutually exclusive. It may therefore be inferred with confidence, that organisation is reached through a succession of struggles for place among competing elements, the successful ones owing their success through position, through superiority in vigour, and so on (Galton’s italics).

His case was broadly framed in abstract terms:

However vague such an explanation may be, it is far from being an inefficient one, for it defines the general character of a process though avowedly incapable of dealing with the details. It applies, moreover, to every theory of heredity which is of a ‘particulate’ character; – that is to say, wherever the theory is based on the supposition of a vast number of partly independent biological particles … . Theories that have this general idea for their foundation seem to be the only ones that are in any way defensible (Galton’s italics).

    From this we see that, in apparent ignorance of Mendel’s work, by 1899 the idea of what we now refer to as “genes,” with a degree of competitiveness that we can now refer to as “selfish,” was established in some influential quarters. Although when considering sex ratios in 1871 Darwin may have hinted at it (Edwards 1998), at this time there was no appreciation of potential hierarchies of competitiveness – namely that the demands of a gene, of the individual containing that gene, and of the species containing that individual, might conflict.


Gene, not individual, as unit of heredity 

Following the discovery of Mendel’s work in 1900, the zoologist William Bateson and the horticulturalist Charles Hurst became its major advocates in the English-speaking world (Cock and Forsdyke 2008). Their elliptical writings, especially those of Bateson, made heavy demands on readers. For example, the word “gene” having not been yet coined, rather than Galton’s “elements” or Weismann’s “ids” they referred to “factors” – a term that had to be carefully interpreted in context. Parental gametes united to form a zygote that grew into an adult that then produced fresh gametes. This process was fundamentally the same in plants and animals. It was now recognized that, during gametogenesis, each pair of genes (“allelomorphs” corresponding to a particular character) that had been separately introduced into the zygote with each parental gamete, were again separated (“segregated”).

    Having noted Michael Guyer’s description of the cytology of spermatogenesis (Bungener and Buscaglia 2003), on October 1st 1902 Bateson (1904) declared that there was “reason to believe that the chromosomes of the father plant and mother plant, side by side, represent blocks of parental characters”. But it took the next two decades to clarify the cytological details, which were elaborated as the “Sutton-Bovari” hypothesis by Edmund Wilson (1925). During meiotic division in the gonads, homologous chromosomes were seen to pair and randomly exchange opposing segments (“recombination”). There was “crossing over” – a breaking and joining ­– between the chromosomes. This meant that, while order was usually unchanged, the set of the segments in each chromosome of an emerging gamete (today referred to as its “haplotype”) was different from that of the corresponding chromosomes of the parent.

    In an address to the Leicester Literary and Philosophical Society Hurst (1904) referred to germ cells (gametes) as containing a “factor of” or “factor for” a character,25 and added:


Seed shape and seed colour are distinct characters with an independent inheritance. Mendel’s discovery of this elementary fact was one of the secrets of his success, and his experimental demonstration of the existence of single hereditable characters proved once and for all that the true unit of heredity is not the individual [organism], but the [factor for a] single character (Hurst’s italics).

Guided by the interpolations within square brackets, the modern reader can see that, if by “the true unit of heredity” Hurst meant, not the single character, but that which determines the single character, then he was declaring that what we now call a gene is a more fundamental unit of heredity than the individual containing that gene.

    Bateson (1919) expanded on this in an address to the Yorkshire Natural Science Association on the “mongrel” composition of the populations of modern nation states:


The fact however, which I wish more to emphasize, is that by the workings of the phenomenon of genetic segregation a man’s children may possess few of the transferable ingredients [genes] which characterized him, his grandchildren may possess none at all, and of his collaterals it is practically certain that few will contain so much of his that he need feel any personal satisfaction or humiliation in their performances. … Looked at coldly in the light of physiological knowledge, what is called the tie of blood is therefore in modern times exceedingly slender, and in all likelihood many of us contain no more of the elements that went to the making of Shakespeare and our heroes than the modern Greek contains of Zeus or Phoebus, despite the frequent alliances which those deities contracted with the daughters of man.

    That a man might transfer no genes to his grandchildren, although consistent with some of the observations of Guyer, somewhat overstated the case. In “some remarks about units of heredity” Wilhelm Johannsen (1923), whose first language was not English, spelled out the biological detail:


Gametogenesis with chromosome-reductions, accompanied by reformations and, as it were, partial rejuvenescence of cell-structures, must in some way act as if especially organized for obliterating the individual’s personally ‘acquired characters,’ which as a rule totally disappear in sexual reproduction. … Continuity in inheritance, the cardinal idea of Aristotle, is – as applied to Mendelian heredity – represented by the continuity of chromosomes in the forthcoming generations – but [is] greatly complicated by disjunctions and recombinations of chromosome pairs. This hereditary continuity is … dissolved into … regular periodic discontinuities: Mendelian heredity always operates with discreet genotypical elements [genes]. Hence differences are here always discontinuous … chemical constitutional differences. Phenotypes however may show discontinuous as well as all degrees of continuous variation (Johanssen’s italics).

    Bateson’s death in 1926 marked the end of an era. By then the word “gene,” introduced by Johanssen around 1909, was well established, as was the notion that genes were distributed along chromosomes in linear order. Sexual reproduction, through recombination between homologous chromosomes, was seen as disrupting that order so that, seeming to be the least disruptable, it was the gene that appeared to be the most stable unit of heredity (Morgan 1926). Although the idea of competition was around, the extension of this, to the idea that a gene might have its own agenda, was for the future. But it was the close future, not the distant future.


Agendas, trade-offs, and kin-selection 

 In the 1920s work on pneumococcal transformation initiated studies leading to the biochemical characterization of genes (Griffith 1928; Olby 1974). Shortly thereafter, the polymath J. B. S. Haldane (his initials being derived from his above-mentioned uncle, John Burdon Sanderson) united two apparently disparate concepts. First, since plant fertilization involved transfer of the male germ by way of the pollen tube to the ovule, then if multiple pollen grains alighted on the female stigma, success would go to the grain whose tube grew fastest. Second, a single gene expressed at two different stages of a life cycle might produce different, stage-specific, outcomes (i.e. it could be pleiotrophic); ideally, it would be beneficial at both stages, but a mutant form of the gene might benefit one stage, not the other (antagonistic pleiotrophy). Haldane (1932) wrote:


A higher plant species is at the mercy of its pollen grains. A [mutant] gene which greatly accelerates pollen tube growth will spread through a species [enhance its own survival] even if it causes moderately disadvantageous changes in the adult plant. [Conversely] a gene producing changes which would be valuable in the adult would be unable to spread through a community if it slows down pollen tube growth.

    The ability of a gene to influence fertilization is likely to dominate, since without fertilization there can be no adult. On the other hand, a less than perfect adult might still be able to reproduce sufficiently to ensure the gene’s passage to future generations where it would again promote fertilization at the expense of adult fitness. Thus, individual plants can be at the mercy of a mutant gene in pollen grains and, in the absence of any countervailing influence, eventually the gene should spread to the entire species. In the apparently ascending hierarchy – gene, organism, species – it would be the “agenda” of the lowest member that was followed.

    Haldane also noted that genes that promoted altruistic conduct would be included in this category: “For in so far as it makes for survival of one's descendants and near relations, altruistic behaviour is a kind of Darwinian fitness, and may be expected to spread as the result of natural selection.” This idea was mathematically elaborated in terms of trade-offs and preferential kin-selection by William Hamilton (1964). At that time George C. Williams was composing his seminal text Adaptation and Natural Selection (Williams 1966). These two – Hamilton and Williams – were the main antecedents acknowledged by Dawkins (1976).


Definition of the gene 

It follows from the mechanics of recombination that genes, rather than the chromosomes that contain them, are most likely to remain intact through the generations. Thus children differ from their parents because they have different combinations of genes, but a given gene in a child is usually identical to that of the parent it was inherited from. This theme, set out by Bateson and Johannsen above, was echoed by Williams. However, he went further in proposing that genes should be defined by this characteristic (Williams 1966: 22-25):


Socrates’ genes may be with us yet, but not his genotype, because meiosis and recombination destroy genotypes as surely as death. It is only the meiotically dissociated fragments of the genotype that are transmitted in sexual reproduction, and these fragments are further fragmented by meiosis in the next generation. If there is an ultimate indivisible fragment it is, by definition, ‘the gene’ that is treated in the abstract discussions of population genetics. … I use the term gene to mean ‘that which segregates and recombines with appreciable frequency.’ Such genes are potentially immortal (William’s italics).

Dawkins (1976) agreed:


Sexual reproduction has the effect of mixing and shuffling genes. This means that any one individual body is just a temporary vehicle for a short-lived combination of genes. The combination of genes which is any one individual may be short-lived, but the genes themselves are potentially very long-lived. Their paths constantly cross and re-cross down the generations. One gene may be regarded as a unit which survives through a large number of successive individual bodies (Dawkins’ italics).

      We must here digress to note that for present purposes we do not follow Williams (1985) in his view that “the gene is not the [DNA] molecule, but the information coded by the molecule”. This was later elaborated in Natural Selection: Domains, Levels, and Challenges  where he noted (Williams 1992: 10-13) that: “Information can exist only as a material pattern, but the same information can be recorded in a variety of patterns in many different kinds of material. A message is always coded in some medium, but the medium really is not the message.” Here Williams conflates the words “message” and “information.” This is easily done. For example, “Did you get his message?” can enquire either whether you received the medium (e.g. paper in an envelope), or whether you had understood the information contained in that medium (i.e. whether information flowed from the medium to your head). Here we take the medium to be the message (McLuhan 1964). A DNA sequence (gene), if subjected to an appropriate reading process, can be seen to contain information (e.g. for the amino acid sequence of a protein). But the information itself is an abstract entity. The medium – a sequence of bases in DNA – is the message. This message contains information, but is not itself that information. Genes are DNA! Typically, information flows from DNA (gene), by way of messenger RNA (not a gene, even though containing genic information), to protein. This particular form of information – protein-encoding information – is but one of many forms of information that can be contained simultaneously in a given DNA sequence (Forsdyke 2006: 183-224). These forms include “accent” or “dialect” (see below).


The boundary problem

Although certain parts of genomes were known to be more prone to recombination than others (“hotspots”), it was not thought that recombination would respect genic boundaries. Recombination occurred both between and within genes. While averring that “a body is the genes’ way of preserving the genes unaltered,” Dawkins (1976) recognized that William’s “ultimate indivisible fragment” might need some qualification:


Even a cistron [gene] can be cut in two by crossing over. The gene is defined as a piece of a chromosome which is sufficiently short for it to last, potentially, for long enough for it to function as a significant unit of natural selection (Dawkins’ italics).

    Thus, to get round the boundary problem, Dawkins advanced a statistical conception of the gene. Genes were small parts of genomes, so the probability was high that a small chromosome segment containing a gene would escape meiotic fragmentation for many generations. While eventually the gene would succumb to the recombination mill, he implied that this would be a rare event:


Individuals are not stable things, they are fleeting. Chromosomes too are shuffled to oblivion, like hands of cards soon after they are dealt. But the cards themselves survive the shuffling. The cards are the genes. The genes are not destroyed by crossing-over, they merely change partners and march on.

    The perspective of Williams and Dawkins was very different from that of many biochemists who had to deal with real DNA sequences, and had to associate gene names with distinct segments for which start and stop signals had to be precisely assigned (Forsdyke 2009; Griffiths and Stotz 2006; Stotz 2009). In extreme form, the Williams-Dawkins definition would include an entire chromosome as a “gene” if that chromosome happened to be excluded from recombination (as in the case of chromosomes in the male fruit fly germ line). So should we shrug off the Williams-Dawkins gene definition in terms of recombination as rhetorical, designed to emphasize their new perspective, rather than as a serious attempt to delineate the gene? Or could they, unknowingly, have been pointing to a hidden genic boundary, somehow related to recombination? Here we turn to the homostability principle and its role in recombination.


The homostability principle 

The revolutionary selfish gene “comet” that appeared in the skies of evolutionary biology in 1976 was widely observed. It followed a no less important, but largely unobserved, homostability principle “comet” that had appeared in 1975. Each gene, while adhering to the general trend in base composition of the genome that contains it, has its own distinctive base composition – it has a distinctive proportion of the four DNA bases that are either G or C (i.e. a distinctive “GC%”). The biophysicist Akiyoshi Wada, whose first language was not English, described this as a “homostabilizing propensity” that might relate to recombination (Wada et al. 1975):


Genetic information is ‘written’ by a variation in sequence on the one hand, and the physical stability of the double-stranded structure is determined by the base composition on the other hand. … DNA is found to consist of a number of homostability regions which come from homogenous base sequences consisting of 500 base pairs or more. … Biologically, it is hard … to believe that such regional homostability originates in a fundamental characteristic of the genetic code itself. It is quite plausible … that the homostability region plays an important part somewhere in the biological process within which the DNA is closely related. If so, then the evolutionary selective force can be considered to have fixed such regions of DNA. From the size of the homostability region, recombination might be one possible process which is aided by it. In any cases, the wobble bases [in codons] must give the necessary redundancy to make a homostability region without spoiling the biological meaning of the genetic code.

    There had already been theoretical pronouncements by Holliday (1968) and Schaap (1971) that there were at least two hierarchical levels of information in DNA, one of which might specially relate to recombination. But, to this theory, Wada and coworkers added hard data. They distinguished base order-dependent “genetic information” that, in the tenor of their times, largely meant protein-encoding information, and base composition-dependent “genetic information.” They recognized that “to make a homostability region” other functions might be threatened, but potential conflicts could be minimized “without spoiling the biological meaning” through the “necessary redundancy” of the genetic code (e.g. to encode glycine, a low GC% gene uses GGT or GGA, and a high GC% gene uses GGC or GGG).

    The emergence of sequencing technologies in the 1970s permitted the boundaries of the Wada homostability regions to be approximated to those of genes (Bibb et al. 1984; Wada and Suyama 1986), and the term “microisochore” was suggested (Forsdyke 2004). This indicated a region of uniform base composition less extensive than the large sub-genomic segments of uniform base composition (“isochores”) that Georgio Bernardi and his associates had described (Filipski et al. 1973).

    The work of Erwin Chargaff in the 1950s had shown that biological species tended to differ in their overall genomic GC% values (Chargaff 1963). Indeed, in 1980 Richard Grantham advanced his “genome hypothesis,” noting that a distinctive base composition could be viewed as the “dialect” or “accent” of a genome that would be imposed on its genes (Grantham 1980; Grantham et al. 1986; Paz et al. 2006). Yet, within this genome framework there can be further “accent” differentiation. Thus, today we can distinguish three levels of homostability (i.e. GC% uniformity) – microisochores (genes), “macroisochores” (isochores), and whole genomes. We here omit consideration of macroisochores.


Structure-mediated homology recognition 

Central to the issue is the homology search process by which nucleic acid molecules recognize each other prior to recombination, a process that can occur both in somatic cells and when homologous chromosomes pair during meiosis in germ-line cells. Despite decades of research on recombination and its associated enzymes, it is lamented that “the mechanism by which homologs uniquely pair with each other is poorly understood” (Blumenstiel et al. 2008). We still have little idea how two “homologous needles find each other in the genomic haystack” (Barzel and Kupiec 2008). We are urged “to ‘branch out’ in our thinking about meiotic recombination” (Cromie and Smith 2007).

    Homologous recombination occurs between sequences which are identical, or very closely so. It is easy to imagine – and many textbook diagrams support the idea – that first one strand of a DNA duplex is cut and then a free single-strand seeks a complementary strand in another DNA duplex. However, certainly in fruit fly and nematode worms, homology recognition does not require initial strand breakage (Moore and Shaw 2009). Indeed, homology recognition can occur between intact DNA duplexes in a simple salt solution in the absence of proteins (Kornyshev and Wynveen 2009; Danilovitz et al. 2009).

    As to the mechanism, there is evidence that “the DNA duplex is labile on a millisecond time scale, allowing local bubble formation” (i.e. strand separation without breakage), so that “alignment of DNA molecules of identical sequence and length could be stabilized by such perturbations occurring simultaneously (and transiently) at identical positions in the sequence” (Inoue et al. 2007). This has been referred to as “structure-mediated homology recognition” (Baldwin et al. 2008; Kornyshev 2010). There is some, albeit indirect, evidence consistent with this occurring in vivo without participation of gene products (i.e. RNA and proteins; Bateman and Wu 2008; Blumenstiel et al. 2008).

    But the “bubbles” would be unlikely to remain single stranded. If not prevented by single-strand binding proteins or unusual sequence characteristics, the extruded single-strands should collapse on themselves and adopt folded stem-loop configurations. In studies of pairing between single-stranded RNA molecules, Tomizawa showed that complementary molecules must first recognize each other, prior to hybridization, by exploratory “kissing” interactions (base pairing) between the loops of stem-loop structures (Eguchi et al. 1991). By analogy, Kleckner and Weiner (1993) proposed that this might apply to the pairing of complementary DNA sequences. Initial loop-loop interactions would progress to the annealing of complementary strands (Figure 1), thus setting the stage for strand breakage and recombination.

     Recombination being a genome-wide activity, this proposal led to predictions that could be subjected to bioinformatic analysis: (i) the potential to extrude stem-loops would be widely distributed; (ii) a regularity in DNA base composition, known as Chargaff’s second parity rule (“PR2”), would tend to apply generally to single-strands of DNA. Both predictions were confirmed (Forsdyke 1996). For example, consider the following two stranded DNA sequence where a T in the top strand pairs with an A in the bottom strand (and vice versa), and a C in the top strand pairs with a G in the bottom strand (and vice versa). Stare at the sequence as long as you like and nothing remarkable is likely to emerge:



 Yet, the sequence has a special property found widely distributed in biological DNA sequences. If we peel away the top strand it can then fold into the following form with two major elements, a stem and a loop:


                TGCGACGC      G

               ACGCTGCG      TA


 For this structure to form there have to be appropriately arranged (“palindromic”) sets of matching (complementary) bases. The stem consists of paired bases (T matching A, and C matching G). Only the bases in the loop (CGATA) are unpaired. If you count the bases, A=4, C=7, G=7 and T=3. Numerically T approximates to A, and C approximates to G, so PR2 applies.

     Some thought that PR2 would be explained in terms of “mutational biases” that were of little biological significance (Sueoka 1999). However, it is now agreed that genomes contain sequences that “may be under selective pressure to preserve their palindromic character and therefore follow PR2 (as pure palindromic sequences are effectively base paired).” (Lobry and Sueoka 2002). Indeed, Bultrini et al. (2003), noting a “symmetrical trend” in DNA sequences, invoked “formation of stem-loop structures.”

    From this perspective, some long-standing recombination models, classified as “paranemic” (no initial strand breakage), can be seen as having much to commend them (Crick 1971; Sobell 1972; Wagner and Radman 1975; Doyle 1978; Wilson 1979). Evidence for “paranemic crossover DNA” is mounting (Wang et al. 2010). Its importance for the present paper is that the rate-limiting step in recombination is likely to be, not the actual pairing of homologous DNA strands, but the extrusion of appropriate stem-loop structures from duplex DNA (Figure 1). The extrusion is likely to be symmetrical, affecting both strands of a duplex equally, and to be critically dependent on base composition (Forsdyke 2007). When the GC% values of two sequences are close, the shapes and tempos of stem-loop extrusions can be similar, a condition propitious for the strand pairing that precedes recombination. When GC% values differ, then, however similar sequences are in other respects, the pairing is disfavoured. Asymmetrical extrusion is considered elsewhere (Lao and Forsdyke 2000; Zhang et al. 2008).


Homology recognition between DNA duplexes

 Figure 1. Homology-recognition depends on prior formation of a stem-loop intermediate. A. Two DNA duplexes of either paternal (P) origin (in blue) or maternal (M) origin (in red). The base pairing between the two strands of each duplex is represented by the laddering (short vertical lines). The two gray boxes indicate sequence segments encoding two genes (microisochores) that differ in base composition (GC%) values. B. Transient, unstable, “bubble” intermediates with unpaired bases. For simplicity only one strand is shown bubbled. The forms shown in B quickly return to the forms shown in A (stable) or, if the sequences contain palindromic segments, can transition to the stable stem-loop conformations shown in C. Although stem-loop extrusions are symmetric, affecting both strands equally, for simplicity only one strand is stem-looped. The stem-loop conformations endure for sufficient time to allow a “kissing” homology search by way of the loops (D). If this succeeds (i.e. there is complementarity between the bases in the loops), then a stable intermediate is formed without strand breakage (E). The cross-over points are known as Holliday junctions.

Strand migration to the genic boundary 

So whether recombination will occur between two sequences depends on the extent to which their GC% values agree. What has this to do with the Williams-Dawkins definition of the gene? The crossover junctions within recombination complexes (“Holliday junctions”) can move along DNA to extend a region of paranemic pairing (Figure 2). This branch migration is affected by the sequence (Sun et al. 1998). Recent work suggests that migration proceeds when the junction unfolds, and stalls when the junction folds – a process that is sensitive to differences in GC% (Karymov et al. 2008). Thus, a shift in GC% at the genic boundary should suffice to stop migration. Should the initial cross-over have occurred within a gene, then the cross-over point would proceed along a region of relatively uniform GC% (corresponding to the gene) until it approached a region of different GC% (intergenic DNA, neighbouring gene or intron). The change in GC% would halt the migration. It is postulated that this would be sensed by recombinogenic enzymes, which would then initiate strand-breakage. (For genes with multiple exons, we should read exon instead of gene.) While not excluding intragenic recombination, this process would greatly decrease its probability. Thus the Williams-Dawkins gene would tend to remain intact through the generations.

Strand breakage at genic boundaries

 Figure 2. Migration of a Holliday junction to the genic boundary where the change in base composition triggers strand breakage. A. The cross-over intermediate shown in Figure 1. B. Branch migration leftward to the genic boundary without strand breakage. C. Strand breakage. D. Ligation so that a maternal strand (red) is joined to a paternal strand (blue). The other Holliday junction can migrate to the right. The complex is resolved by mechanisms described in standard texts.

    Since there is a positive association between palindromes (i.e. stem-loop potential) and recombinational crossing-over (Inagaki et al. 2009), then also working to protect genes from recombinational disruption would be the lower potential to extrude stem-loops in exons than in introns and flanking DNA (Forsdyke 2006: 207-224). Indeed, fine-scale maps show that recombination crossover points mainly occurs outside human genes (McVean et al. 2004; Coop et al. 2008).

    These considerations emphasize the positive role of GC% identity in the successful pairing of allelic genes, but of equal significance is the negative role of GC% non-identity in preventing the pairing of non-allelic genes, both somatically and in germ-line cells. This has implications for the successful duplication both of genes and of genomes (speciation). When GC% non-identity extends to entire meiotic chromosomes then gametogenesis fails, a condition propitious for divergence into two species. The case has been made from evidence, both historic and modern, that reproductive isolation that leads to speciation can initially be secured by virtue of differences in GC% (Forsdyke 2010).



Historical antecedents of the selfish gene concept date back to the nineteenth century. Williams and Dawkins went further in defining the gene in terms of its ability to resist recombinational disruption. But there need be no fundamental discrepancy between their definition and the conventional definitions of biochemists. Numerous studies, especially those of biophysicist Akiyoshi Wada and colleagues (Wada et al. 1975; Wada and Suyama 1986), have shown that, in addition to encoding a distinctive function dependent upon primary sequence (base order), within the same boundaries each gene has a distinctive, and to some extent independent, base composition. The latter (GC%) is a gene’s “accent” that has the potential to determine whether paranemic pairing (no strand breakage) between it and its allele can occur. The postulate that recombinational strand breakage can proceed when GC% uniformity is lost (i.e. at conventional genic boundaries) brings the two definitions into close correspondence.


Queen’s University hosts my web-pages which display some works of Michael Guyer and several of the cited references including the Third Report of the Commissioners on the Cattle Plague ( This paper is dedicated to the memory of George Williams, who died on 7th September 2010.



Baldwin GS, Brooks NJ, Robson RE, Wynveen A, Goldar A, Leikin S, Seddon JM, Kornyshev AA (2008) DNA double helices recognize mutual sequence homology in a protein free environment. Journal of Physical Chemistry B 112: 1060–1064.

Barzel A, Kupiec M (2008) Finding a match. How do homologous sequences get together for recombination? Nature Reviews Genetics 9: 27-37.

Bateman JR, Wu C ting (2008) A genomewide survey argues that every zygotic gene product is dispensable for the initiation of somatic homolog pairing in Drosophila. Genetics 180: 1329-1342.

Bateson W (1904) Practical aspects of the new discoveries in heredity. Memoirs of the Horticultural Society of New York 1: 1-9, 123.

Bateson W (1919) Science and nationality. Edinburgh Review 229: 123-138.

Beale LS (1866) Letter to Charles Darwin. In: The Correspondence of Charles Darwin (Burkhardt F, Porter DM, Dean SA, Evans S, Innes S, Sclater A, Pearn A, White P, eds.), Vol. 14, pp 50-51. Cambridge: Cambridge University Press.

Bibb MJ, Findlay PR, Johnson MW (1984) The relationship between base composition and codon usage in bacterial genes and its use for the simple and reliable identification of protein-coding sequences. Gene 30: 157-166.

Blumenstiel JP, Fu R, Theurkauf WE, Hawley RS (2008) Components of the RNAi machinery that mediate long-distance chromosomal associations are dispensable for meiotic and early somatic homolog pairing in Drosophila melanogaster. Genetics 180: 1355-1365.

Bultrini E, Pizzi E, Giudice PD, Frontali C (2003) Pentamer vocabularies characterizing introns and intron-like intergenic tracts from Caenorhabditis elegans and Drosophila melanogaster. Gene 304: 183-192.

Bungener P, Buscaglia M (2003) Early connection between cytology and Mendelism: Michael F. Guyer’s contribution. History and Philosophy of the Life Sciences 25: 27-50.

Butler S (1878) Life and Habit, p. 134. London: Trübner.

Butler S (1880) Unconscious Memory, p. 269. London: David Bogue.

Chargaff E (1963) Essays on Nucleic Acids. Amsterdam: Elsevier.

Cock AG, Forsdyke DR (2008) “Treasure your Exceptions.” The Science and Life of William Bateson. New York: Springer.

Coop G, Wen X, Ober C, Pritchard JK, Przeworski M (2008) High-resolution mapping of crossovers reveals extensive variation in fine-scale recombination patterns among humans. Science 319: 1395-1398.

Crick F (1971) General model for the chromosomes of higher organisms. Nature 234: 25-27.

Cromie GA, Smith GR (2007) Branching out: meiotic recombination and its regulation. Trends in Cell Biology 17: 448-455.

Danilowicz C, Lee CH, Kim K, Hatch K, Coljee VW, Kleckner N, Prentiss M (2009) Single molecule detection of direct, homologous, DNA/DNA pairing. Proceedings of the National Academy of Sciences USA 106: 19824-19829.

Darwin CR (1868) The Variation of Animals and Plants under Domestication. London: John Murray.

Darwin CR (1875) The Variation of Animals and Plants under Domestication, 2nd ed. London: John Murray.

Dawkins R (1976) The Selfish Gene, pp. 25-38. New York: Oxford University Press.

Doyle GG (1978) A general theory of chromosome pairing based on the palindromic DNA model of Sobell with modifications and amplification. Journal of Theoretical Biology 70: 171-184.

Edwards AWF (1998) Natural selection and the sex ratio: Fisher’s sources. American Naturalist 151: 564-569.

Eguchi Y, Itoh T, Tomizawa J-I (1991) Antisense RNA. Annual Review of Biochemistry 60: 631-652.

Filipski J, Thiery JP, Bernardi G (1973) An analysis of the bovine genome by Cs2SO4 Ag+ density centrifugation. Journal of Molecular Biology 80: 177–197.

Forsdyke DR (1996) Different biological species "broadcast" their DNAs at different (C+G)% "wavelengths." Journal of Theoretical Biology 178: 405-417.

Forsdyke DR (2001) The Origin of Species, Revisited. A Victorian who Anticipated Modern Developments in Darwin’s Theory. Montreal: McGill-Queen’s University Press.

Forsdyke DR (2004) Regions of relative GC% uniformity are recombinational isolators. Journal of Biological Systems 12: 261-271.

Forsdyke DR (2006) Evolutionary Bioinformatics. New York: Springer.

Forsdyke DR (2007) Molecular sex: the importance of base composition rather than homology when nucleic acids hybridize. Journal of Theoretical Biology 249: 325-330.

Forsdyke DR (2009) Scherrer and Josts’ symposium. The gene concept in 2008. Theory in Bioscience 128: 157-161.

Forsdyke DR (2010) George Romanes, William Bateson, and Darwin’s “weak point.” Notes and Records of the Royal Society 64: 139-154.

Galton F (1894) Discontinuity in evolution. Mind 3: 362–372.

Grafen A, Ridley M (2007) Richard Dawkins: How a Scientist Changed the Way We Think. Oxford: Oxford University Press.

Grantham R (1980) Workings of the genetic code. Trends in Biochemical Science 5: 327–331.

Grantham R, Perrin P, Mouchiroud D (1986) Patterns in codon usage of different kinds of species. Oxford Surveys of Evolutionary Biology 3: 48-81.

Griffith F (1928) The significance of pneumococcal types. Journal of Hygiene 27: 113-159.

Griffiths PE, Stotz K (2006) Genes in the post-genomic era. Theoretical Medicine and Bioethics 27: 499-521.

Haldane JBS (1932) The Causes of Evolution, pp. 123-131. London: Longman and Green.

Hamilton WD (1964) The genetical theory of social behaviour. Journal of Theoretical Biology 7: 1-32.

Holliday R (1968) Genetic recombination in fungi. In: Replication and Recombination of Genetic Material (Peacock WJ, Brock RD, eds.), pp. 157-174. Camberra: Australian Academy of Sciences.

Hurst CC (1904) Mendel’s discoveries in heredity. Transactions of the Leicester Literary and Philosophical Society 8: 121-134.

Huxley TH (1893) Darwiniana. Collected Essays. London: Macmillan.

Inagaki H, Ohye T, Kogo H, Kato T, Bolor H, Taniguchi M, Shaikh TH, Emanuel BS, Kurahashi H (2009) Chromosomal instability mediated by non-B DNA: Cruciform conformation and not DNA sequence is responsible for recurrent translocation in humans. Genome Research 19: 191-198.

Inoue S, Sugiyama S, Travers AA, Ohyama T (2007) Self-assembly of double-stranded DNA molecules at nanomolar concentrations. Biochemistry 46: 164-171.

Johannsen W (1923) Some remarks about units of heredity. Hereditas 4: 133-141.

Karymov MA, Boganov A, Lyubchenko YL (2008) Single molecule fluorescence analysis of branch migration of Holliday junctions: effect of DNA sequence. Biophysical Journal 95: 1239-1247.

Kleckner N, Weiner BM (1993) Potential advantages of unstable interactions for pairing of chromosomes in meiotic, somatic and premeiotic cells. Cold Spring Harbor Symposium on Quantitative Biology 58: 553-565.

Kornyshev AA (2010) Physics of DNA: unraveling hidden abilities encoded in the structure of ‘the most important molecule.’ Physical Chemistry Chemical Physics 12: 12352-12378.   

Kornyshev AA, Wynveen A (2009) The homology recognition well as an innate property of DNA structure. Proceedings of the National Academy of Sciences USA 106: 4742-4746.

Lao PJ, Forsdyke DR (2000) Crossover hotspot instigator (Chi) sequences in E. coli occupy distinct recombination/transcription islands. Gene 243: 47-57.

Lobry JR, Sueoka N (2002) Asymmetric directional mutation pressures in bacteria. Genome Biology 3 (10):research 0058.

McLuhan M (1964) Understanding Media: The Extensions of Man. London: Routledge.

McVean GAT, Myers SR, Hunt S, Deloukas P, Bentley DR, Donnelly P (2004) The fine-scale structure of recombination rate variation in the human genome. Science 304: 581-584.

Moore G, Shaw P (2009) Improving the chances of finding the right partner. Current Opinion in Genetics and Development 19: 99-104.

Morgan TH (1926) The Theory of the Gene. New Haven: Yale University Press.

Olby R (1974) The Path to the Double Helix. Seattle: University of Washington Press.

O’Malley MA (2009) What did Darwin say about microbes, and how did microbiology respond? Trends in Microbiology 17: 341-347.

Paz A, Kirzhner V, Nevo E, Korol A (2006) Coevolution of DNA-interacting proteins and genome "dialect." Molecular Biology and Evolution 23: 56-64.

Roberts HF (1929) Plant Hybridization before Mendel. Princeton: Princeton University Press.

Romano TM (2002) Making Medicine Scientific. John Burdon Sanderson and the Culture of Victorian Science, pp. 55-74. Baltimore: Johns Hopkins University Press.

Roux W (1881) Der Kampf der Theile im Organismus. Leipzig: W. Engelman.

Schaap T (1971) Dual information in DNA and the evolution of the genetic code. Journal of Theoretical Biology 32: 293-298. 

Sobell HM (1972) Molecular mechanism for genetic recombination. Proceedings of the National Academy of Sciences USA 69: 2483-2487.

Simon J (1865) Minutes of Evidence Taken Before the Cattle Plague Commissioners. London: Houses of Parliament.

Spencer JP, Cecil RATG, Lowe R, Playfair L, Read CS, Jones HB, Quain R, Parkes EA, Wormald T, Ceely R, Spooner C (1866) Third Report of the Commissioners Appointed to Inquire into the Origin and Nature, etc. of the Cattle Plague. London: Houses of Parliament.

Stotz K (2009) Experimental philosophy of biology: notes from the field. Studies in History and Philosophy of Science 40: 233-237.

Sueoka N (1999) Two aspects of DNA base composition: G+C content and translation-coupled deviation from intra-strand rule of A = T and G = C. Journal of Molecular Evolution 49: 49-62.

Sun W, Mao C, Liu F, Seeman NC (1998) Sequence dependence of branch migratory minima. Journal of Molecular Biology 282: 59-70.

Vries H de (1889) Intracellulare Pangenesis. Jena: Verlag von Gustav Fischer.

Wada A, Suyama A (1986) Local stability of DNA and RNA secondary structure and its relation to biological functions. Progress in Biophysics and Molecular Biology 47: 113-157.

Wada A, Tachibana H, Gotoh O, Takanami M (1975) Long range homogeneity of physical stability in double-stranded DNA. Nature 263: 439-440.

Wagner RE, Radman M (19750 A mechanism for initiation of genetic recombination. Proceedings of the National Academy of Sciences USA 72: 3619-3622.

Wang X, Zhang X, Mao C, Seeman NC (2010) Double-stranded DNA homology produces a physical signature. Proceedings of the National Academy of Sciences USA 107: 12547-12552.

Williams GC (1966) Adaptation and Natural Selection. A Critique of Some Current Evolutionary Thought, pp. 22-25. Princeton: Princeton University Press.

Williams GC (1985) A defense of reductionism in evolutionary biology. Oxford Surveys of Evolutionary Biology 2: 1-27.

Williams GC (1992) Natural Selection: Domains, Levels, and Challenges, pp. 10-13. New York: Oxford University Press.

Wilson EB (1925) The Cell in Development and Heredity, 3rd ed., p. 928. New York: Macmillan

Wilson JH (1979) Nick-free formation of reciprocal heteroduplexes: a simple solution to the topological problem. Proceedings of the National Academy of Sciences USA 76: 3641-3645.

Zhang C, Xu S, Wei J-F, Forsdyke DR (2008) Microsatellites that violate Chargaff's second parity rule have base order-dependent asymmetries in the folding energies of complementary DNA strands and may not drive speciation. Journal of Theoretical Biology 254: 168–177.

Journal Date submitted Identification code Date rejected (accepted)
Biological Reviews 15 April 2009 BRV-04-2009-0044 11 May 2009
Biology and Philosophy 17 May 2009 BIPH365 24 Aug 2009
Journal of Theoretical Biology 27 Aug 2009 JTB-D-09-00685 3 Sept 2009
Journal of Heredity 14 Sept 2009 JOH-2009-196 1 Dec 2009
Journal of Genetics 11 Jan 2010 JGEN-D-10-00010 6 Apr 2010
Notes & Records of the Royal Society 5 Sept 2010 RSNR-2010-0077 1 Nov 2010
Biological Theory 2 Nov 2010 BIOT-D-10-00019 15 Dec 2010 (accepted)

End Note on Hotspots (November 2011)

Although not extending to the above selfish gene scenario, geneticists approached the recombination problem through consideration of recombination hot spots, the location and DNA sequences of which were found to be highly species specific, and thus transient on an evolutionary time scale. So humans and chimpanzees did not have common hotspots. Some 25000 to 50000 hotspots were detected in the human genome by Myers et al. (2005), perhaps extending to 80000 (Khil & Camerini-Otero 2010). This number lay intriguingly close to the total number of exons.

   Using antibodies to a protein that localizes at putative recombination-initiated sites in mice, Smagulova et al. (2011) reported a central motif, flanked on the "top" strand by a purine-loaded segment to the left and a pyrimidine-loaded segment to the right. Thus, in extruded structures one would expect large purine-rich "kissing" loops to the left and large pyrimidine-rich "kissing" loops to the right (and the converse on the "bottom" strand). Hotspots were found mainly to localize to intergenic regions and introns (where the potential to form stem-loop structures is greater and Chargaff's second parity rule applies more strictly).

   As random mutations bring about changes in base sequence and composition (e.g. changes in GC%), the central canonical consensus motif would vary, and proteins interacting with that motif would be on a mutational "treadmill" (contrasting with proteins reacting with a more stable substrate, such as glucose; Paz et al. 2006; Forsdyke 2011). Thus, a most rapidly evolving positively-selected gene is that encoding the meiosis-specific "PR domain containing protein-9" (PRDM9), whose zinc finger segments bind to the DNA motif. Some 40% of currently identified human hotspots match the binding specificity of human PRDM9 (Ponting 2011). Mouse PRDM9 matches 73% of mouse hotspots (Smagulova et al 2011). Hailed as a "speciation gene" in vertebrates, the question arises as to whether the PRDM9 gene "tail" wags the genomic "dog", or vice-versa (Forsdyke 2011).

   Normally, within a species, identical motifs on homologous parental chromosomes would support strand pairing and meiotic recombination cross-over. However, since the motifs vary within a species, there is a chance of mismatch (i.e. heterozygosity). Recombination with gene conversion can follow, and the direction of the conversion is usually such as to eliminate the consensus in favour of the deviant. (Why this should be is another question.) However, although the consensus motif may disappear at a particular site, it will be reformed at other sites by mutations affecting cryptic motifs that differ only slightly from the consensus motif (Wahls and Davidson 2011).

   Thus, within the population, there will be a continuing flux of motifs around the consensus and, from time to time, individuals enriched in the same deviant motifs will meet and reproduce, hence tending to initiate a new line. When members of this line meet members of the consensus line, there will come a stage when the extent of the heterozygosity difference is such that recombination will fail. There will then be no gene conversion and a state of incipient reproductive isolation will exist (potential speciation).

Forsdyke DR (2011) Evolutionary Bioinformatics, 2nd Edition. Springer, New York, pp. 238-239.

Khil PP, Camerini-Otero RD (2010)Genetic crossovers are predicted accurately by the computed human recombination map. PLOS Genetics 6, e1000831.

Myers S, Bottolo L, Freeman C, McVean G, Donnelly P (2005) A fine scale map of recombination rates and hotspots across the human genome. Science 310, 321-324.

Ponting CP (2011) What are the genomic drivers of the rapid evolution of PRDM9? Trends in Genetics 27, 165-171.

Smagulova F, Gregoretti IV, Brick K, Camerini-Otero RD, Petukhova GV (2011) Genome-wide analysis reveals novel molecular features of mouse recombination hotspots. Nature 472, 375-378.

Wahls WP, Davidson MK (2011) DNA sequence-mediated, evolutionarily rapid, redistribution of meiotic recombination hotspots. Genetics 189, 685-694.  


End Note on Akiyoshi Wada (October 2014) 

Wada's pioneering studies mentioned above were but a small part of an illustrious career. As described by Cyranoski (2009 Nature 460, 171-172): "In the 1970s Wada ran into many sceptical biologists when he was one of the first to envision large scale automated genomic sequencing. But even as these technologies were ramping up elsewhere, Japan's bureaucrats stalled, its genomics fell behind, ... The unfolding of Wada's failed efforts are described in a book aptly titled A Defeat in the Genome Project." Much too late, in 1998 he became founding director of the RIKEN Genomic Sciences Center in Yokahama. One may speculate that, had the bureaucrats paid as much attention to Wada as to Kimura (the pioneer advocate of the "neutral theory" so loved of the biomathematicians, which is falling out of favour), the story might have been very different.

Bioinformatics Index (Click Here)

Videolectures Page (Click Here)

HomePage (Click Here)

This page was posted in Jan 2011, and last updated 27 Oct 2014, by D. R. Forsdyke