Donald R. Forsdyke  

     In some quarters Gould's demise may be seen as marking the demise of the non-genic viewpoint (Kitcher, 2004). However, apparently unknown to Gould, in various laboratories including my own, bioinformatic analyses of the vast amounts of sequence information that emerged in the 1990s had yielded what seemed an independent affirmation. Some members of the genic community took notice and a "commentary" was provoked (Kliman, Rogers and Noor, 2001).

     The commentary by Kliman et al. in 2001 of "Forsdyke's chromosomal viewpoint" on the initiation of speciation (Forsdyke, 1996, 1999) would, I hoped, have been effectively dealt with by my book published later that year (Forsdyke, 2001). But a member of the Advisory Board of the Journal of Theoretical Biology has recently reviewed the book, citing with approval the paper of Kliman et al. (Charlesworth, 2003). Accordingly, a reply to the commentary with an updating of references may now be in order.

2. Non-Genic Factors are Sub-Microscopic

      The non-genic "chromosomal" viewpoint on speciation has a long history that, to my reading of the literature, can best be considered as beginning with George Romanes (1886). Indeed, the classic dual between Alfred Wallace and Romanes in the 1880s covered much the same ground as that between Dawkins and Gould (Forsdyke, 2004a; Smith, 2004). The non-genic torch was borne through the twentieth century by some of the most able biologists of our times - William Bateson, Nicolai Kholodkovskii, Richard Goldschmidt, Cyril Darlington, Addison Gulick, Gregory Bateson, Michael White, Max King, and Stephan Jay Gould. Romanes and Bateson were both convinced of the importance of what we would now call "non-genic" factors, but were reluctant to specify a chromosomal location (Forsdyke, 2001). The role of the chromosomes was made explicit by Kholodkovskii and the Russian school (Krementsov, 1994), a view adopted by all who followed.

      As first formulated, the chromosomal viewpoint was particularly vulnerable to criticism (Coyne and Orr, 1998) since it postulated that large, often microscopically visible, chromosomal differences ("macromutations") would be evident early in the speciation process. These would inhibit the meiotic pairing of homologous chromosomes so that recombination would be inhibited and the hybrid would produce fewer and/or maldeveloped gametes (hybrid sterility). Thus, the parents of the hybrid would be reproductively isolated form each other. In contrast, Goldschmidt (1940) postulated sub-microscopic changes ("micromutations") that we can now interpret at the level of individual bases. My bioinformatic analyses emphasized the role of differences in Chargaff's species-specific component of the base composition, GC% - an agency that earlier scholars could only refer to in abstract terms (Forsdyke, 2003).
 

3. Areas of Agreement and Disagreement

      Coyne and Orr (1998) considered "the question of whether postzygotic isolation in animals is based on chromosomal or genic differences" and listed five major criticisms of "chromosomal speciation," which I addressed in my 1999 paper. The paper of Kliman et al. (2001) that acknowledges the "helpful comments" of Coyne, and the review of Charlesworth (2003), seems to constitute the Drosophila community's latest response.

      We all agree that members of a species are so defined because they are reproductively isolated from members of other species. (When so inclined, even prokaryotes exchange nucleic acid best with their own kind; Radman and Wagner, 1993; Gratia and Thiry, 2003). We all agree that the question of the origin of species is essentially the question of the origin of reproductive isolation. We all agree that reproductive isolation can be achieved in a variety of ways, both genic and non-genic. However, we disagree on which of these ways is likely to have predominated over evolutionary time. This is an issue that cannot be settled by showing how reproductive isolation has been achieved in an individual, or even in several, case(s). The induction as to which form of reproductive isolation is the best candidate for having been the most usual origin of species must be made by looking at a large body of biological, genetical and biochemical evidence that relates to a variety of taxa.

      Related issues include the questions of the existence of hierarchical levels, and different units, of selection (Gould, 2002), and the question whether the mechanisms postulated for the type of group selection known as speciation are vulnerable to the same criticisms as are the mechanisms postulated for some other types of group selection (Wynne-Edwards, 1962; Maynard Smith, 1989). Not considered here is the question of the existence of some vital force not explicable by known laws of physics and chemistry (Driesch, 1914). Also not considered here is the argument that, despite disagreements, evolutionists should present a united front to the advocates of "creationism" and "intelligent design." Also not considered here is the argument that evolutionists should present a united front in order not to confuse other scientists who need to understand speciation although working in other fields. Sadly, these issues have been coloured by, sometimes ad hominem, attacks by members of the genic school on members of the non-genic school. Early targets were Romanes, Bateson and Goldschmidt (Forsdyke 2001, 2003). The most prominent modern target has been Gould (Morris, 2001; Sterelny and Turney, 2001; Forsdyke, 2004a).

4. Are GC% Differences Primary?

    Recently diverged bacterial species also show early GC% differences, predominantly at third codon positions (i.e. at the codon position that, when changed, is least likely to change the encoded amino acid; Bellgard and Gojobori, 1999; Bellgard et al., 2001; Gupta and Ghosh, 2003). The case that such differences are primary has been strengthened by the demonstration of a plausible mechanism relating GC% differences to reproductive (i.e. recombinational) isolation. There is no one definitive model for recombination (Holliday, 1990).  But all models regard similarity between sequences as necessary for legitimate recombination, since a successful similarity search initiates the pairing of complementary strands of potentially recombining duplexes. Thus, all models associate sequence non-similarity with recombinational suppression. Among the various models there is one that implicates GC%. This model postulates that the similarity search is between stem-loop structures that must be extruded from classical Watson-Crick duplexes so that a mutual loop-loop "kissing" exploration, which precedes extensive strand pairing, can begin. A slight degree of negative supercoiling of DNA (the usual situation) would favour such extrusion. Small differences in stem-loop configurations should suffice to misalign loops and prevent pairing, so inhibiting recombination (Sobell, 1972; Wagner and Radman, 1975; Doyle, 1978).

      Since the ability to recombine is a general property of DNA, this model predicts that (i) stem-loop potential should be widely dispersed in the DNA of all species, (ii) at least by virtue of the base pairing in stems, Chargaff's parity rule (%A = %T, %G = %C) should apply, to a close approximation, to single-strands of DNA of all species, and (iii) the quantity of the strongest bonding Watson-Crick base pair (G and C) should be a critical determinant of the configuration of extruded stem-loop structures. These predictions are fulfilled (Forsdyke, 2001). In particular, the energetics of stem-loop formation is very sensitive to extremely small fluctuations in GC% (Forsdyke, 1998). This implies that such small fluctuations would suffice to change stem-loop configuration and so prevent recombination.

      Thus, the present non-genic case rests on the facts that (i) GC% differences have been observed early in the speciation process, and (ii) there is a plausible mechanism by which small GC% differences inhibit recombination. Furthermore, although contested by Kliman et al. (2001), a variety of biological phenomena ("empirical data") lend themselves to interpretation in such terms ("post hoc evaluations of published data"). These include (i) large differences in the GC% of viruses that have a common host cell (Forsdyke, 1996, 1999), (ii) the cure of hybrid sterility by polyploidization (Winge, 1917; Goldschmidt, 1940), and (iii) Haldane's rule (Forsdyke, 2000, 2001).

      Of course, GC% differences can affect gene function, but this would be largely by affecting first and second codon positions. The non-genic hypothesis attaches most importance to changes in genomic regions that do not necessarily affect gene function (i.e. third codon positions, introns, extragenic DNA). Thirteen years before the discovery of DNA structure, this point was elegantly made by Goldschmidt (1940):

Thus, stem-loop potential should be best developed in introns and extragenic DNA as, indeed, the evidence suggests (Forsdyke, 1995; Barrette et al., 2001; Bultrini et al., 2003). To regard "G+C differences in --flanking DNA" as producing "the reproductive isolation effect via Forsdyke's compositional model," is not to be shrugged off as a "special pleading" (Kliman et al., 2001).

6. "The Problem of Simple Underdominance in Forsdyke's Model"

      By "underdominance" (sometimes called "negative heterosis"; King, 1993) is meant that heterozygotes are less fit than parental homozygotes (e.g. hybrid sterility). If gene A were incompatible with its allele a, then heterozygotes Aa would be underdominant (decreased fitness relative to their parents). Accordingly, homozygous forms AA could not productively cross with homozygous forms aa. The forms would be reproductively isolated by this "speciation gene," and so could further develop into distinct species. However, a single initial mutation in a member of a healthy homozygote population (say AA) to Aa would be lost, so that a could not begin to become established in the population and aa homozygotes would not arise.

      A solution of this underdominance problem (Coyne and Orr, 1998) is to invoke multiple non-allelic incompatibilities, the simplest case being that with two "speciation genes" a and b that are incompatible. A healthy homozygous form might be AABB.  In one individual there might be a mutation to AaBB. In another individual there might be a mutation to AABb. Both heterozygous forms would be viable and further crossing could lead to forms AAbb and aaBB. However, hybrids between these forms would be incompatible.

      From this it is seen that a single genic change (e.g. a single base change in a first or second codon position) constitutes a problem for the genic hypothesis that is surmounted by postulating polygenic changes (involving at least two base changes in non-allelic genes). This argument leaves the genic speciation hypotheses in contention (i.e. on the table, for serious consideration). The same argument is less necessary for the non-genic GC% hypothesis, since GC% begins by supposing that one base change, although it would slightly affect GC%, would probably be insufficient to impair a similarity search between chromosomes. Again, in complete ignorance of GC%, Goldschmidt (1940) may have been close to the mark:

 Thus, it can be supposed that two forms will progressively deviate in their respective GC% values (repattern their chromosomes) until some threshold is reached at which chromosomal mispairing is sufficient to activate "check-points" and disrupt meiosis (Page and Orr-Weaver, 1996).

      Kliman et al. (2001) recognize that "the general concept of underdominance can be applied to any unit of heredity in a diploid system," meaning genic or non-genic. Thus, it is stated that my "model may suffer from the same problems inherent in other models that require underdominance." They argue that the problem of "hybrid fitness reduction -- addressed at great length by population geneticists" is also a problem for my non-genic GC% model for hybrid fitness reduction. "The major theoretical problem" with my model is that it "requires deleterious mutations to rise in frequency in at least one of the incipient species." Thus, "the G+C model -- cannot easily get around the problem of underdominance. For there to be G+C divergence between species there must first be deleterious G+C variation within species."

      Without defining what they mean by "neutral," and seeming to imply that they are going some way towards solving the alleged underdominance problem of my model, they state that "heterozygozity within species could be nearly or effectively neutral as long as an arbitrary threshold of G+C divergence - which would be exceeded in interspecies heterozygotes - was not reached." It is not clear from this whether Kliman et al. are clearly distinguishing the conventional phenotype and the genome phenotype. They would not be alone if they are not. Agreeing with Romanes (1886), Ernst Mayr when referring in 1963 to what he called "the biological species concept" postulated that: "The most indispensable step in speciation is the acquisition of isolating mechanisms." However, Mayr continued: "Isolating mechanisms have no selective value as such until they are reasonably efficient and can prevent the breaking up of gene complexes. They are ad hoc mechanisms." 

         Here, Mayr seems to have been thinking only of natural selection acting on the conventional phenotype.  But it is the genome phenotype that is primarily affected by the postulated differences in GC%, and there may be no accompanying change in the conventional phenotype (i.e. no deleterious conventional mutations). At the "threshold of G+C divergence" there would be deleterious mutations affecting the genome phenotype (e.g. changed stem-loop architecture) to an extent sufficient to suppress recombination. To this extent the mutations would not be "neutral" (Forsdyke, 2002). The hybrid would immediately be selected against in that gametogenesis would be impaired. Accordingly, the parents of that hybrid could immediately be recognized as being partially or completely reproductively isolated from each other.

     So achieving underdominance is a problem neither with the genic model, nor with my model - less so with my model. Perhaps Kliman et al. (2001) have the same problem as Darwin (1875):

"Take the case of any two species which, when crossed, produce few and sterile offspring; now what is there which could favour the survival of those individuals which happened to be endowed in a slightly higher degree with mutual infertility, and which thus approached by one small step towards absolute sterility? Yet an advance of this kind, if the theory of natural selection be brought to bear, must have incessantly occurred with many species, for a multitude are mutually quite barren."

This is sometimes called the paradox of negative heterosis. Of course, Darwin recognized that absolutely sterile individuals themselves survive. He was writing about hybrid sterility, not hybrid inviability. He meant that "the survival" of the lineage of a pair of individuals endowed "with mutual infertility" would not be favoured since they produce offspring that are themselves unable to produce offspring. The line could not continue. Thus, the healthy fertile parents of a healthy, but sterile, individual are reproductively isolated from each other. Kliman et al. (2001) seem to regard those mutations most likely to cause reproductive isolation as least likely to fix in populations. Here the concept of hierarchical levels of selection, that Gould (2002) struggled so valiantly to communicate, may help.

7. GC% as an Agency of Group Selection

      In the usual situation, when the speciation process begins Darwin's "two species" are a main species with many members and an incipient species with a few variant members. The healthy hybrid offspring formed when a member of the main species crosses with a member of an incipient species are usually discarded from the point of view of maintaining either lineage. Although phenotypically healthy, the offspring are genetically unfit (genotypically unhealthy) by virtue of their partial or absolute sterility. Their discardment reflects the fact that they belong neither to the main species, nor to the incipient species. There is then nothing that "could favour the survival of individuals" (i.e. the survival of individual lineages corresponding to a potentially incipient species) that have undergone changes in GC% leading towards "absolute sterility" (when crossed with members of the main species), save finding a GC%-compatible mate (i.e. another member of the same incipient species lineage that has undergone similar changes in GC%).

      Because the hybrids are sterile, the main species can be viewed as constituting a "reproductive environment" that moulds the genome phenotype by negatively selecting (by denying reproductive success to) variant organisms that attempt (by mating and producing healthy, fertile, offspring) to recross the emerging interspecies boundary. Thus, a species can positively select itself by negatively selecting variants (i.e. sterile hybrids are discarded). Should these variants find compatible mates, then they might accumulate as a new species that, in turn, would positively select itself by negatively selecting further variants. This would be "species selection," a form of group selection that many biologists have found hard to imagine. Indeed, Dawkins (1986), having scorned the "argument from personal incredulity," was obliged to resort to it when confronted with the possibility of species selection: "It is hard to think of reasons why species survivability should be decoupled from the sum of the survivabilities of the individual members of the species."

        When the latter sentence is parsed its logic seems impeccable. Yet, "the species" is the established main species, members of which imperil themselves only marginally, if at all, by mating with (denying reproductive success to) members of a small potentially incipient species. Thus, in reproductive interactions between a main and an incipient species, survivability of the main species, as a species, is coupled negatively to the sum of the survivabilities of individual members of the incipient species much more than it is coupled positively to the survivabilities of individual members of the main species. In this sense, main species survivability is coupled to the former survivabilities and decoupled from the latter survivabilities. Again, by individual survivabilities is meant, not just mere survival, but survival permitting unimpeded production of fertile offspring. Survival of members of an incipient species occurs not only when classical Darwinian phenotypic interactions are favourable (e.g. escape from a tiger), but also when reproductive interactions are favourable (e.g. no attempted reproduction with members of the main species). Tigers are a phenotypic threat. Members of the main species are a "reprotypic" threat (Forsdyke, 2001). 

      There is no dispute that the differential survival of groups of individuals (species), or of genes within individuals, operate through the differential survival of individuals. The dispute is over how, in the general case, groups of individuals survive to form a new species. The non-genic viewpoint is that individual members of a main species, which are involved (when there is an attempted crossing) in the denial of reproductive success to individual members of potential incipient species (that differ significantly in GC%), are like individual stones in the walls of a species fortress against which the reproductive arrows of potential incipient species become blunted and fall to the ground. Alternatively, the main species can be viewed as a Gulliver who barely notices the individual Lilliputian incipients brushed off or trampled in his evolutionary path. 
 
     Just as individual cells acting in collective phenotypic harmony constitute a Gulliver, so individual members of a species acting in collective "reprotypic" harmony constitute a species. That harmony is threatened, not by its own members, but by deviants that, by definition, are no longer members of the main species (since a species is defined as consisting of individuals between which there is no reproductive isolation). These deviants, differing in GC%, constitute potential incipient species that might one day pose phenotypic threats to the main species.

      It is true that a member of a main species that becomes irretrievably pair-bonded with a member of an incipient species (e.g. pigeons) will leave fewer offspring, so that both members will suffer the same fate (have decreased survivability in terms of number of fertile offspring). But, in the general case, one such infertile reproductive encounter with a member of an incipient species will be followed by many fertile reproductive encounters with fellow members of the main species. Much more rarely, a member of an incipient species may encounter a fellow incipient species member with which it can successfully reproduce (i.e. they have similar deviations in GC%). This is an essential precondition for species divergence.

8. Quantitative Factors

Kliman et al. (2001) correctly state that "the essence" of my hypothesis "is that subtle differences in G+C content of genetically similar individuals have a role in the hybrid sterility leading to speciation." We currently have no quantitative information allowing us to state what "subtle" means in terms of a GC% difference sufficient to activate meiotic checkpoints and cause a partial or complete impairment of gametogenesis. Thus, the quantitative studies of Kliman et al. (2001) of crosses between various Drosophila species which, as they note, have already begun to differentiate prezygotically ("it would be improper to call them incipient"), are of questionable relevance.

     Needed are more detailed in vitro biochemical studies of the enzymes of recombination, of their interactions with DNAs from different sources, and of meiotic checkpoint mechanisms (Page and Orr-Weaver, 1996). There should also be more introgression studies as described by Naviera and Maside (1998). The latter found their results "unexpected" and suggested that "a new paradigm is emerging, which will force us -- to revise many conclusions of past studies" (Forsdyke, 2001).

9. Conclusions

While asserting that the "genic model -- is far from a theory in crisis," Kliman et al. (2001) dismiss my non-genic model (Forsdyke, 1996, 1999) as mere "conjecture" that provides "a misleading account of the state of the field," and does not further "our understanding of the genetics of post-zygotic isolation." Calling for "quantitative analysis" and conceding only that the "G+C model cannot be explicitly ruled out," they imply that a chromosomal model based on GC% differences does not deserve a prominent place at the table of evolutionary discourse. 

      On the contrary, I have argued that the model, with a history spanning over a century, has been greatly strengthened by recent bioinformatic studies (Forsdyke, 2001). It is time for Coyne and Charlesworth to call off their "bulldogs" and concede that, however elegant, mathematical modelling without a sufficient appreciation of the number of variables, and their underlying chemistry, may not be productive. The GC% model should be taken seriously. Researchers, both old and new, should be encouraged to study it.

Acknowledgements

Queen's University hosts my web-pages, where full text versions of some of the cited references may be found.

References