Septet Theory in Rosa and Rubus

The genus Rubus is closely allied to Rosa; both have a basic set of 7 chromosomes. The inheritance of sectional differences as units is what led Hurst to propose his Septet Theory.

Hurst assumed that the distinguishing characteristics of diploid species were distributed among the septet of chromosomes, which required that whole septets be distributed to all viable and fertile offspring. Eileen Erlanson (Chromosome Organization in Rosa, Cytology 2:256-282, 1931) attempted to refute this assumption by direct cytological observation, but undermined her efforts by choosing "sterile" specimens: Orleans Rose with 45% bad pollen, a seedling R. blanda with 43.4%, and specimen of R. relicta Erlanson with 56-88% bad pollen. Hurst used "sterile" in the Darwinian sense of "not fully fertile", which certainly applies to Erlanson's selections.

Hurst's "highly artificial system" (Erlanson, Chromosome Organization in Rosa, Cytology 2:257) was based on his observation that groups of species characteristics are conserved in polyploids and hybrids. That these groups are maintained as supergenes rather than as whole septets is irrelevant to Hurst's basic system (the supergene concept was introduced by Darlington and Mather in Elements of Genetics, 1949). The septets of chromosomes do not always hold together, but the traits do.

In the following paper, Darlington concluded that in Rubus "it is at this (diploid) level that the long-term evolutionary processes have been at work, the polyploids being a very recent, indeed perhaps a post-glacial, novelty." Not so for Rosa, since the arctic octoploid R. acicularis is presumably pre-glacial, and 35 million year old fossils of a rose resembling hexaploid R. nutkana have been found.

Hurst supposed that the evolution of Rosa has been up to higher ploidy and down to lower, in which case the genotypically controlled isolation of the sectional distinctions would be expected to function more successfully in Rosa polyploids than in those of Rubus. As Darlington has noted, "if sexual fertility is important for a new tetraploid, selection in meiotic behaviour should readily improve it." (The Evolution of Genetic Systems, 1958, p154) That is, even in autotetraploids selection for fertility will reduce quadrivalent formation and enhance differential pairing of similar chromosomes. This will be even easier in allotetraploids (such as may be derived from hybrids of diploid species) and will tend to favor differential pairing and whole septet segregation in old polyploids and their hybrids. In addition, Darlington has noted (Elements of Genetics) that it is easier to establish linkage between genes on non-homologous chromosomes than at opposite ends of the same chromosome. Thus, there is no real problem in supposing that supergenes and linkage groups may involve multiple chromosomes even in the absence of segmental interchanges.

Darlington supposed that the supergene differences between raspberries and blackberries are ancient. So must be the differences among the 5 basic divisions of Rosa described by Hurst. Hurst excluded R. persica from Rosa, referring it to the genus Hulthemia. Even so, the existence of hybrids of this "rose" with proper roses forces us to consider its 7 chromosome set as a sixth division or septet-type (or supergene-type).

Hurst also excluded R. roxburghii (Microphylla) and R. stellata (Hesperhodos), so his Coryana (R. roxburghii x R. macrophylla) and Schoener's R. gigantea x R. stellata hybrids introduce further complexity—and septets/supergenes.

Furthermore, recent genetic research (Using RAPDs to Study Phylogenetic Relationships in Rosa; Millan, Osuna, Cobos, Torres; Theor. Appl. Genet., 1996, 92:273-277) suggests that R. cymosa and R. banksiae (both commonly grouped in the Banksianae) are as distant from each other as both are from the rest of the genus. Hurst included them in the A-septet group, which seems to be confirmed by Mansuino's crosses of R. banksiae with A-septet varieties. Time will tell whether the R. banksiae and R. cymosa supergenes and chromosomes constitute yet more types.

The supergenes (presumably) that distinguish Rosa from Rubus, Rubus from Fragaria, Fragaria from Potentilla, etc. must be even more ancient, and of considerable interest when intergeneric hybrids are raised. E.g. the strawberry Pink Panda gains its pink flower color from a Potentilla, though it is otherwise an apparently normal strawberry.

Though the origin of generic distinctions is assumed to be more ancient than the origin of sub-generic and specific distinctions, this is not necessarily the case. Where some initial population exhibits a balanced polymorphism of various traits, some of the components of the polymorphism may be lost in some of the derived populations, retained in others. I.e., numerous small super-genes may become linked as fewer, larger super-genes that become the basis for higher order distinctions, and yet the same super-genes may provide sectional and specific distinctions in different genera. E.g., few broad thorns vs. numerous small prickles (and smooth variants) may be found in both Rubus and Rosa. There is no need to suppose that there has been independent evolution of the traits in the two genera, since they may have been inherent in the ancestral group.

Intersectional or intergeneric (even interfamilial) hybrids may survive long enough to transmit variants of the ancient supergenes between distant relatives. In fact, such transient hybrids may even introduce new super-genes—chromosome segments that did not behave as supergenes in the original species, but which do when introduced to a new species. Such was the origin of some supergenes in Zea (corn) derived from Tripsacum.

In polyploids, such transfer of segments may occur between ordinarily non-pairing chromosomes by infrequent multivalents occurring during meiosis. If the modified chromosomes are then returned to the diploid population, the new chromosome segments may behave as super-genes and be retained if they confer some advantage—even if only of nonspecific heterozygous advantage (hybrid vigor). The apparent phylogenetic closeness of R. chinensis to R. gallica on the basis of RAPDs research may be the result of ancient triploid hybrids between the species giving modified diploids among the garden-raised progeny. "Let us moreover note that the fruits of Sanguinea as well as those of all the maroon-colored ones (Chinas) very much resemble those of the Gallica." Jean-Pierre Vibert, Essai sur les Roses, 1824-1830; quoted by Brent Dickerson in The Old Rose Advisor, 1992, p 26.

Heredity 3:103 (1949)
Summary from Reproductive Versatility in Rubus
M. B. Crane and P. T. Thomas

1. The Veitchberry 4x=28 was derived from R. rusticanus 2x=14 and a tetraploid form of R. idaeus 4x=28. Upon selfing there is no approach to either parental forms, it breeds true to its intermediate character.

2. Upon crossing with R. idaeus the behaviour of the Veitchberry is again that of a species, but exceptional diploid forms have appeared in the progeny: four blackberry-like, two intermediate but slender and one raspberry-like.

3. The chromosomes of the blackberry and the raspberry are differentiated and normally pair among themselves when two sets of each are present in the hybrid.

4. With less opportunity for differential pairing as in the Mahdiberry 3x=21 the chromosomes of the two types can pair effectively with one another.

5. The allotetraploid Veitchberry behaves exceptionally in that there is considerable failure of pairing. This failure appears to be due to genotypic control since no structural differences were observed between the raspberry and blackberry chromosomes.

6. One of the intermediate diploid plants from the Veitchberry similarly behaved like an asynaptic diploid.

7. It is concluded that seven diploids in the progeny from Veitchberry 4x=28 x raspberry 2x=14 have originated parthenogenetically from the embryo sacs of the Veitchberry.

Heredity 3:103-106 (1949)
On An Integrated Species Difference
C. D. Darlington

The families examined by Crane and Thomas from the cross of Veitchberry by raspberry fall into two classes, the sexual and the asexual [parthenogenic]. If these classes are separated and compared two apparently contradictory conclusions follow.

In the first place there is a segregation of species differences in the diploid asexual progeny which is evidently genuine. But the three types of diploid are those expected only if the raspberry-blackberry difference is behaving as a unit in inheritance. Now this difference, as the diagram shows, is an elaborate one. It must obviously depend on numerous mutually adapted gene changes. The recombination of these in diploids is prevented in the European flora, both wild and cultivated, by the elimination of all diploid hybrids. We now see that it is also prevented in polyploids by some other, even more fundamental, condition. This condition must be the integration of all the differences as a block within which no crossing-over occurs.

  Diagram to show the contrasted effects of segregation on the raspberry-blackberry difference in sexual and asexual families of the hybrid.  
R. rusticanus

Permanent brambles, unlimited growth
Few broad thorns
Palmate leaves
Black fruits—fixed plug

R. idaeus

Short-lived canes of limited growth
Many narrow prickles
Pinate leaves
Red fruits—free plug


* Genotypic failure of pairing at meiosis in pollen mother cells. VEITCHBERRY RR BB*

Intermediate in:
stems, prickles and leaves;
colour and flavour of fruit;
freedom of plug

Breeds true except for
intra-group differences
4 BB infertile
2 RB* infertile
1 RR fertile
237 RRB

In the second place the eggs of the Veitchberry in sexual families (selfed as well as crossed) show no segregation of the differences between its parental species. If such segregation occurs the homozygotes are eliminated. The eggs in asexual families on the contrary show, not merely segregation, but an excess of homozygotes: it is the heterozygotes that are eliminated.

This apparent contradiction is resolved by the cytological evidence. The frequency with which any one chromosome of the seven types enters into a quadrivalent in the Veitchberry is 0.39/7 or 5.5% of the cells. A quadrivalent will give RR—BB segregation in about half these cells and of them half will suffer loss of laggards and die, leaving 0.7% of RR gametes and 0.7% of BB, or one in 140.

Thus one in 1402, or about 20,000, selfed seedlings of Veitchberry should show segregation of each of the pure types, a possibility which has not been tested.

On the other hand the 0.7% segregation will be enhanced in the asexual families by the elimination of the diploid homozygotes. This elimination is independantly shown by the failure of crosses between the diploid species—except in the production of polyploids such as the Veitchberry and the Mahdi. Thus the diploid elimination conceals the heterozygotes and the tetraploid recombination conceals the homozygotes.

The physiological integration of the raspberry and blackberry types in the Rubus cell must be combined with the basis of genetic isolation between the two groups. This isolation appears to be twofold: it is shown by the non-viability of direct diploid hybrids and also by the sterility of the indirect diploid arising from parthenogenesis of the tetraploid. This sterility strongly expressed appears as morphological male-sterility, and weakly expressed as reduced pairing at meiosis; in all gradations it is genotypically controlled.

The non-viability of the diploid hybrid is also clarified by the present experiment. It seems not to be inherent since the diploid RB heterozygote can be raised in this indirect way. It probably depends on an error in the embryo-endosperm-ovary relationship in the diploid first cross which is no longer completely effective as a means of eliminating the diploid progeny of the tetraploid hybrid.

One consequence of this unitary difference is in the interpretation of the Mahdi sport. The pentaploid has evidently arisen by doubling (vegetative or sexual) in the triploid followed by the loss of one chromosome set, i.e. seven complementary chromosomes taken at random from the 42.

The maintenance of the same external form with such a change of chromosome complement was preposterous in terms of a dispersed difference: it becomes intelligible in terms of an integrated one. To suppose that the 2 to 1 raspberry-blackberry proportion of the original Mahdi might persist in its giant mutant as a 3 to 2 proportion, avoiding any external qualitative variation by a precise regulation for all seven chromosomes of the set was too much. But to suppose that such a change could take place for a single member of the set although astonishing, is not utterly unreasonable. It now therefore becomes necessary to suppose that extensive changes hitherto known only in the "vegetative regulation" of the mosses (Wettstein, 1924) may take place in mitotic chromosome numbers, subject to mechanical loss of chromosomes and physiological selection of cells.

How is the raspberry-blackberry difference constituted?

In Oenothera we know how interchange makes it possible to hold together in the interchange hybrids large or complex differences lying in the seven different chromosomes. But here there is no evidence of interchange. All the differences must therefore lie in one chromosome.

We have to visualise a single super-gene consisting of many parts with mutually adjusted effects. These parts must be recombinable within each group as has been shown both for the prickles of the blackberry by Crane and Darlington (1927) and of the raspberry by Lewis (1940). It is in this way that an enormous range of species has been produced in each group, limited in number indeed only by the supply of materials and names that will enable us to preserve and describe them. On the other hand in crosses between the two groups the ultimate differences are held together. The diploid blackberry segregates arising from the Veitchberry have not merely the general blackberry character; they have the very prickles of their rusticanus grandparent.

Such a system can arise by the inversion of a chromosome segment. In most natural populations there are inversions floating which occasionally will chance to include (or, shall we say, collide with) groups of mutually adapted genes capable of becoming a focus of fruitful discontinuity. And the coincidence having occurred, the resulting super-gene will evolve in the way that has been described from its successive stages by Darlington and Mather (1949).

Thus, in a variety of ways, we have evidence of the basis of the chief genetic divarication within the genus Rubus. The fact that such an elaborate divarication is effective only at the diploid level shows that it is at this level that the long-term evolutionary processes have been at work, the polyploids being a very recent, indeed perhaps a post-glacial, novelty.

In the present super-gene difference we can already see two stages of development. The integration of a block by suppression of crossing-over, and the non-viability of the diploid hybrid, both of which prevent recombination. But the integration of the block must have come first since the non-viability of the hybrid between two groups automatically brings to an end this organised divergence.

The closest analogy to the present situation is in the speltoid and fatuoid complexes of the cereals. The difference is that these complexes have arisen in cultivated plants and in hexaploids. Moreover they distinguish smaller systematic groups and lead to no inviability in the hybrid. The Rubus super-gene is therefore presumably a much older complex.

The problem that now arises is thus to compare the character and scope of the discontinuity between different sections of the main raspberry and blackberry groups in different parts of the world.


Hurst Bibliography