17 It has been said that the switch-mechanism operating in polymorphism may be a single major gene or else a super-gene. These situations may be difficult to distinguish. On the one hand, ebony body-colour and the effects associated with it can be polymorphic in laboratory stocks of Drosophila melanogaster, yet the condition has arisen several times as the result of mutation, so suggesting a change in one unit of a cistron. We are not here dealing with a super-gene nor do we appear to be when confronted with a series of non-complementary phases of similar type, indicating several intra-cistronic mutants. These, even when polymorphic, represent the multiple alleles of classical genetics: the OAB blood group series may belong here. On the other hand, complementary features of a nearly related kind may yet give evidence that the genes responsible for them are distinct though closely linked, so as to behave effectively as one unit or super-gene. These suggest that they have been developed as a duplication or a break through a single cistron (they are ex-cistronic), the resulting parts evolving into neighbouring major genes by selection acting on their effects and perhaps through chromosome reconstruction. It looks as if the two incompatibility genes present in various Basidiomycetes, Coprinus and others, have originated in this way (Crowe, 1964).
Yet the alternative or multiple phases of a polymorphism may involve the combined segregation of two or more features of a very distinct kind, such as colour-pattern together with the presence or absence of tails in Papilio memnon. The control of such conditions requires the interaction of distinct but co-adapted genes which have been brought together from different parts of the same chromosome or from non-homologous ones, as must often happen in the evolutionary adjustment of organisms in general. 18 Among the various instances of the kind already analysed, the heterostyle polymorphism of Primula (e.g. P. vulgaris, P. veris and many more) shows us, perhaps, the way in which such super-genes can evolve. Here a cross-over within the combined unit gives rise to the homostyle condition, while the situation in P. sinensis, in which other genes scattered elsewhere on the chromosomes can affect stigma-length and anther-height, indicates the type of material from which the normal heterostyle-homostyle locus has been constructed.
A "super-gene" was defined by Darlington and Mather, who introduced this important concept in 1949, as "a group of genes acting as a mechanical unit in particular allelic combinations". That is to say, the members of the group though consisting of major genes are so seldom separated by crossing-over that they can operate as a single genetic entity having diverse effects. That end may be achieved by close proximity, as already indicated, or else by inclusion within an inversion.
The multiple effects of a single gene appear often to be remarkably distinct. Thus that for white eyes (w) in Drosophila melanogaster involves not only the development of the pigment in the ommatidia but also changes in the shape of the spermatheca, the colour of the testis-sheath and the viability and fertility of the fly, all arising together as the result of a single mutation which has taken place on a number of occasions in laboratory stocks. Individual major genes, in fact, nearly always influence the general physiology of the organism owing to some change in enzyme-production, even if the visible structures which they control are quite trivial. Yet it is extremely unlikely that the multiple action of a single gene can often produce a sufficiently near approach to the diverse co-adapted characters required in a polymorphism to be effectively improved later by selection. We are faced with a similar improbability if we suppose that a super-gene controlling distinct characters can be built up by duplication or the breakage of one cistron.
It seems clear then that, in the majority of instances at least, 19 the genetic control of co-adapted characters must be vested in separate genes which have been collected together to form the switch-unit involved. Structural interchanges, or possibly translocations, bringing them on to the same chromosome would be favoured. Subsequently, selection could operate to reduce crossing-over between them, as it would if they chanced to be linked in the first instance. This could be achieved in several ways: (1) by the localization of chiasmata (whether terminal or near the centrosome), a process known to be under genetic control (Darlington, 1956, p36), their frequency being selectively reduced between the loci concerned, and (2) by means of cross-over inhibitors. (3) A similar effect could also be obtained by chromosome reconstruction, which could move the controlling genes closer together. (4) Another result of such breakage and reunion is the production of inversions capable of acting as super-genes, the effect of which, however, requires separate discussion. (5) In addition, polyploidy may be associated with reduced crossing-over, for the diminished fertility to which it gives rise can be, and generally is, mitigated by selective restriction of pachytene pairing or of chiasma-formation. It may here be remarked that such compensation for this particular defect in polyploids might reasonably be thought of as particularly requisite in annuals, which are wholly dependent upon propagation by seeds, compared with the opportunities for vegetative reproduction enjoyed by perennials. On the other hand, as Darlington (1956, p42) has pointed out, seed-production is generally so vastly in excess of what is required, that most plants can well tolerate some degree of infertility.
I have pointed out elsewhere (Ford, 1964) how little research on the selective adjustment of cross-over values has so far been undertaken. This is not because of its inherent difficulty: it has indeed been carried out several times and with striking success, while some of this work has been known for many years. For instance, Detlefsen and Roberts (1921) were able to reduce the recombination between the loci for white eyes (w) and miniature wings (m) in Drosophila melanogaster, normally 36 per cent, to 6 per cent in one line and 0.6 in another. 20 They could not obtain a significant effect by selecting in the reverse direction. However, this has been achieved by Parsons (1958) who raised the cross-over value between black body (b) and purple eye-colour (pr) significantly from 5 to 8 per cent. The fact is that such investigations are outside the tradition of research in Drosophila melanogaster, which has tended to treat the loci as fixed points upon which linkage-maps could be calculated, ignoring the evolution of the chromosomes. Yet the selective adjustment of crossing-over, a subject of great importance, can be deduced, since the distribution and frequency of chiasmata vary within and between species, while these qualities are genetically controlled and must therefore be open to adaptive modification.
The evolutionary effect of reducing cross-over values needs some brief comment here. On the one hand, only the closest linkage can itself maintain two co-adapted genes in association: for it is a platitude to say that when some distance apart on the same chromosome they will, with equal frequency, assort in their repulsion and coupling phases, in which they are held apart as often as they are kept together. On the other hand, as crossing-over between them decreases in frequency, so a smaller selection-pressure will suffice to maintain the more advantageous combination in the population or, with a fixed selective advantage, it will be maintained more efficiently. The end-product of such a process is reached when the genes concerned act as a single 'switch-unit': when, in fact, they have formed a super-gene. In this situation, linkage really does preserve the genetic co-adaptation that is required; in the sense that though rare cross-overs between the units concerned produce another, but ill-adjusted, super-gene, as stable as its 'allele', this can then be eliminated by counter-selection with approximately the same ease as a correspondingly undesirable mutant at a single locus: which, indeed, it simulates. [Editor's note: the double flowered Petunia originated as such a cross-over, rather than as a "gene mutation". Mini roses may be another example.]
When a super-gene, which is necessarily composite, is thus evolved by a selective reduction in crossing-over and more than two alternative sets of genic substitutions within it are favoured, these will behave as if they were multiple alleles since not more than two of them can be present in a diploid individual. 21 Each member of the series produces effects which can be modified by selection operating on the gene-complex. That process will be relatively quick in polymorphism, since the proportions even of the less common phases will be quite considerable.
It is partly for these reasons, and partly for those given on p. 25, that genetic polymorphism frequently appears to be multiple allelic in character. We find it so in the heterostyle-homostyle system of plants, in the human blood groups, in Batesian mimicry, and in numerous other instances where colour-patterns are involved (in Snails, Cepaea; Orthoptera, Apotettix, Paratettix and others; Coleoptera, Harmonia, and many more). In such instances, the switch-control is likely to be a multiple unit, and its true nature becomes apparent when rare cross-overs separate its constituent parts.
In addition to the situations just mentioned, the co-adapted genes determining polymorphic phases may be held together within an inversion. If this be 'short', the structurally heterozygous sections will rarely pair, and then only straight, by torsion when, as Darlington says, they probably fail to undergo crossing-over. If 'long' of sufficient length, that is, for the formation of counter-turned loops, even included chiasmata normally fail to break down the genetic isolation involved.
They lead indeed to a number of possible situations. Assuming extra-centric inversions, a single included chiasma produces one acentric and one dicentric chromatid, and the latter forms a bridge which destroys the dividing nucleus. The rarer condition of two included chiasmata, on the other hand, may give rise to two dicentrics and two acentrics or to one of each. It does so both if the pairs of chromatids respectively involved are different (when all the chromatids formed are abnormal), and if the arrangement is the diagonal one (when two of those formed are abnormal). If the chiasmata arise in the same two chromatids, a bridge does not result but destruction will generally follow owing to the duplication and deletion that then occurs. However, a pericentric inversion with one chiasma produces four centric chromatids and so does not provide a basis for super-gene formation.
22 It appears that inversions are widespread and abundant in the genetic material, taking their place among the essential features both of plant and animal cytology. They can of course be recognized easily in polytene nuclei, when it can be seen that the 'short' type is scattered in large numbers throughout the chromosomes. Most of these little reconstructions are without known visible effects. They are nearly indetectable in ordinary cells, for it is yet technically too difficult to distinguish the small unpaired regions to which they generally give rise during pachytene. 'Long' inversions, on the other hand, though undoubtedly difficult to demonstrate, have been identified rather widely among organisms, including Man (Koller, 1937), owing to the chromatid bridges which may result from included chiasmata.
An additional consideration deserves brief mention at this point. 23 The difficulty of building two or more genes into a co-adapted system must be increased when they start on different chromosomes. There is evidently some advantage in few and large linkage-groups here. Darlington (1956, pp. 82-3) has stressed that cytological mechanics ensure that changes in chromosome-number are predominantly upward, to higher values; for it is far more dangerous to lose genes than to multiply them. Even so, with a large chromosome set, one aspect of genetic plasticity can indeed be attained by means of small variations in the total number, even downwards by eliminating a chromosome or two, though upwards is, of course, much easier. With small chromosome-numbers, that source of variation is in general unidirectional; for though the genetic balance will be much disturbed by an increase, the proportionate genic loss when a chromosome drops out must almost inevitably prove fatal unless it be relatively a very small one. Thus, in these circumstances, chromosome reduction can only be achieved by a translocation (not by the balanced, and therefore more general, process of structural interchange), leaving a centric fragment that can then be eliminated. Such a step has undoubtedly been taken (as in Drosophila species, from the more primitive n=5 to the more specialized n=4) but it evidently necessitates a complex evolutionary process. We have here been considering polysomics but the tendency to chromosome increase is even more obvious in polyploidy, for this is virtually irreversible.
[Editor's note: a translocation occurs when a portion of one chromosome is transferred to a non-homologous chromosome. A structural interchange occurs when two non-homologous chromosomes exchange segments.]
On the other hand, as Darlington (l.c.) points out, some aspect of selection must resist these upward tendencies for yet, on the whole, the chromosome-numbers of organisms do remain small. He himself suggests that this may be due to the greater efficiency in the management of meiosis which that condition ensures, as well as to economy in cell-space. Possibly an additional factor opposing numerical increase in the chromosomes is the need, urgent and universal as it must be, to produce sets of co-adapted genes.
A further step in setting aside chromatin to control polymorphism has been taken in some organisms in which the necessary switch-mechanism consists of whole chromosomes; 24 for example, in Nicandra physaloides (distantly related to the tobacco plant, Nicotiana) a species in which most frequently 2n=20. The total comprises nine pairs of autosomes and one pair of isochromosomes. These latter are inherently bound to pair so as to give univalents or bivalents. When univalents are lost, the pollen dies but eggs can be formed. If fertilized, these produce seedlings in which 2n=19. They are subject to delayed germination (Darlington and Janaki-Ammal, 1945), and the variation so produced is of advantage in giving plasticity to meet environmental differences. Accordingly, the species is polymorphic for chromosome-number. Indeed Darlington (see Huxley 1955, p10) has suggested that the B-chromosomes, the number of which vary in many plants, may provide a switch-mechanism by which polymorphism in such features as germination-date and drought-resistance can be controlled. Moreover, in an animal, Cimex (Hemiptera), Darlington (1939) showed that an increase in the number of B-chromosomes, which are derived from X-chromosomes, enables the organism to pass from out-breeding to in-breeding, in which the B's increase, without loss of variability.
We may indeed think of the control of polymorphic forms as constituting a series in chromosome commitments. This ranges from a single major gene to the various forms of super-gene in an order depending on the relative amount of chromatin involved: that in which the association of co-adapted genes is maintained by close proximity only, by a short inversion, by a long inversion, by a whole chromosome or by a group of chromosomes as in the chain of 9 sex-chromosomes in the male Centipede Otocryptops (Ogawa, 1954). It must not be supposed, however, that this corresponds with an evolutionary sequence or that there is a tendency to pass from a single gene up to the other end of the scale. Indeed the control of polymorphism by means of a whole chromosome seems to be a specialized and exceptional situation, while there is surely a decided advantage in a short inversion, when attainable, compared with a long one: indeed the short inversion may well provide the most efficient switch-mechanism in polymorphism.
25 Finally, there does not appear to be sufficient evidence to determine with any certainty the origin of the numerous distinct genetic units within a cistron. The changes to which these give rise are non-complementary, while their recessive effects* show the cis-trans difference in heterozygous material. It looks therefore as if each intra-cistronic group has in process of time and at a remote period been evolved, possibly by duplication, to control the development of a given set of characters. In consequence, its origin appears to be distinct from the constituent members of a super-gene; for these, being complementary yet co-adapted in their action, have in some instances been constructed from one cistron but in others brought together from within the same chromosome or from non-homologous ones. In either event, the process must have taken place relatively recently, to meet the current needs of micro-evolution.
It seems therefore that there must be two main types of super-genes: those arising from one cistron and those whose members have been accumulated from different parts of the chromosomes; and the means by which the components of the latter are brought and held together are various. There must, then, be effectively three types of multiple alleles: those behaving as such because they are associated with the two kinds of super-genes, and those that are intra-cistronic.
*The erroneous term 'recessive gene' constantly recurs in the literature of molecular genetics.