Proc. Am. Soc. Hort. Sci. 52: 510-516 (1948)
Cytogenetic Studies on Rosa Rubiginosa and Its Hybrids
Homer T. Blackhurst
A. & M. College of Texas, College Station, Texas
Submitted to the Graduate School of the A. & M. College of Texas in partial fulfillment of the requirements for the degree of doctor of philosophy. Grateful acknowledgment is made to Dr. S. H. Yarnell for assistance and counsel during the course of this work.

The peculiar meiosis of the Caninae roses was worked out in detail and reported by Täckholm.8 In this particular group are found tetraploid 2n = 28, pentaploid 2n = 35 and hexaploid 2n = 42 forms but in all forms Täckholm found exactly seven pairs at meiosis, the remaining chromosomes behaving as univalents in the reduction division. In the first meiotic division the seven bivalents are arranged in a normal fashion on the equatorial plate with the univalents scattered at random over the spindle. According to Täckholm the bivalents separate normally at anaphase I and proceed to the poles, while the univalents remain behind on the spindle. Later the univalents divide and sister halves are distributed regularly to the two poles. Täckholm reported, however, that all chromosomes were included in the restitution nuclei. At second anaphase the seven chromosomes from the original bivalents divide and the sister halves are distributed to opposite poles. The univalents having divided previously at the first anaphase do not divide again and are left behind in the cytoplasm to form supernumerary microspores. Those microspores receiving exactly seven chromatics from the original bivalents are assumed to be the only functional ones, the others becoming abortive.

The reduction division in the macrospore mother cells is quite different. The bivalents behave as in the divisions of the microspore mother cell; but at anaphase I the univalents do not divide but are all oriented on the micropylar side of the equatorial plate and are included in the micropylar nucleus at the end of the first division. The egg cell finally formed is a derivative of this micropylar nucleus and contains seven less than the somatic number of chromosomes. Thus, if self pollination occurred the resulting progeny would contain the original somatic chromosome number of parent.

*In fact, Hurst observed that Rosa spp. from around the world expressed five sets of fifty characters in various combinations. He proposed his "septet scheme" as a working hypothesis to explain how so many traits could remain united.

To explain this peculiar meiosis in the Caninae roses Hurst4 advanced a genome theory in which he postulated* the existence in the genus Rosa of five differential diploid septets of chromosomes, which he designated as A, B, C, D, E. These septets are supposed to segregate essentially as complete units and the genome makeup of a species may supposedly be determined by a study of the morphological characters which are more or less specific for the various septets. It is further supposed that pairing at meiosis occurs between similar septets. According to Hurst the Caninae roses have resulted from hybridization between species having only one septet in common, resulting in well differentiated species having only seven pairs of homologous chromosomes.

To explain the origin of polymorphy in the Caninae group Täckholm advances a theory of hybridization and believes that the various forms arose from three original crosses, namely:

7 X 21 = 7" 14' (Rosa pomifera)
7 X 28 = 7" 21' (Rosa canina)
7 X 35 = 7" 28' (Rosa Jundzillii)

While there is no evidence in this or other families to suggest the origin of so many forms from only three crosses, he believes that these three crosses arising in the pre-Tertiary era could give rise to the many types through mutation and further crossing. Aneuploid types would arise but these are not assigned a survival value. To explain the survival of so many types he assumes that reproduction in the Caninae roses is principally apomictic, the species still remaining facultatively sexual. Assuming apomixis, he makes no effort to explain the possible survival value of the peculiar meiosis in the Caninae group. Hurst4 agrees with Täckholm in assuming an original decaploid species to account for Rosa Jundzillii Besser (7" 28'). He based his theory of five differential diploid septets on this ancestral decaploid.

CybeRose note: This is incorrect. Hurst (1925) wrote: "But one day when comparing the taxonomic characters of the species in the living collection at Kew, one was struck by the fact that the tetraploid species showed the combined characters of two distinct diploid species, while the hexaploid species showed the combined characters of three distinct diploid species and the octoploid species showed the combined characters of four distinct diploid species." Further study revealed a 5th. Finally, "In conclusion it may be useful to point out the value of the septet scheme of chromosomes and characters in Rosa as a working hypothesis."

Gustafsson3 studied the hybrids canina-rugosa, rubiginosa-rugosa, canina-rubiginosa and rubiginosa-canina. To explain the meiosis in Caninae roses he advanced the theory of internal autotriploidy, according to which these irregular species are assumed to contain three more or less homologous genomes which in pure forms pair only in two's. Such a theory does not explain the absence of trivalents in meiosis of these species.

CybeRose note: This is not a valid objection. Autotetraploids may be selected for fertility, which reduces quadrivalent formation. The same could occur with internal autotriploidy. It is just as likely, however, that differential pairing is not absolute, as Blackhurst apparently thought Gustafsson's theory required.

All investigators up to the present are in agreement in assuming a hybrid origin for the Caninae roses, but no entirely satisfactory theory has been advanced to adequately explain the peculiar type of meiosis found in the group.

The present study of Rosa rubiginosa L and its hybrids is an attempt to explain the interspecific relationships of roses, as indicated by the chromosome behavior of the available interspecific hybrids, reflected in meiosis of the pollen mother cells. In particular it was hoped that the results might throw some light on the validity of Hurst's genome theory, and if the theory appears inadequate to suggest an alternative theory. The Caninae species R. rubiginosa was used because it was well adapted to the locality and bloomed consistently with a good set of hips.


Sixteen hybrids of Rosa rubiginosa L are included in the present report. Buds were fixed in Allen's modification of Bouin's B15 solution, imbedded in paraffin, sectioned at 12 microns and stained in crystal violet.

Somatic chromosome counts were made from leaf smears according to the method of Baldwin1.

Taxonomic Characters of Hybrids: In most instances the hybrids fall into an intermediate class but in general resemble more closely the seed parent. There is one exception to this rule, that of Rosa rubiginosa x R. Täckholmii, in which the hybrid more nearly approaches the pollen parent in outward characteristics. However, this hybrid received 28 chromosomes from the pollen parent, whereas in the other hybrids 14 chromosomes was the highest carried in the pollen.

A majority of the hybrids have heavier and more dense armature than either parent and are more vigorous.

Leaves, stipules, sepals and bracts are intermediate. Color, shape and size of hips fall into an intermediate class with the exception of the cross with Rosa Täckholmii in which these characters resemble more nearly the pollen parent.

In the hybrids hip and peduncle are more or less glandular or glandular spiny regardless of the parentage.

The only character behaving in a clear cut fashion is adnation of the sepals. There are three types of behavior. In one the sepals shed soon after petal fall, in the intermediate type they are held for a time after petal fall and in the third are persistent to maturity. Deciduous x deciduous gives deciduous, deciduous x persistent gives persistent and deciduous x intermediate results in intermediate.

The absence of clear cut modes of inheritance makes classification difficult, and it is small wonder that species naming has gotten out of hand in the genus Rosa.


Crosses: All crosses made on Rosa rubiginosa are listed in Table I in order according to the septet theory of Hurst.

On the basis of Hurst's formulas as shown in the Table there are no septets used in the crosses that show complete incompatibility. The E septet was not available in the diploid form but is represented in the pollen of the regular tetraploid and irregular hexaploid forms. Since Hurst gives the formula ABBCD for Rosa rubiginosa the E septet is the only one that would not supposedly find a mate in the egg of the seed parent. From the Table it is obvious that such a foreign septet had little or no influence on compatibility of the parents.

Cytological Results: Cytological studies were made on 16 different hybrids. The average number of pairs of chromosomes was determined on a minimum of 10 plates. This number is not considered sufficient for statistical treatment, but should give a fair idea of the pairing behavior. The behavior of the chromosomes at meiosis is presented in Table I.

TABLE I — Crossing Results and Chromosome Pairing in Hybrids of Rosa rubiginosa L
Listed According to Hurst's Septet Formula
Chromosome Pairing
R. Rubiginosa L (ABBCD) 7.0 21.0
    x Helenae (AA) AABCD 1.2 8.4 13.8
    x multiflora Thunb (AA) 18 22.0AABCD 1.5 7.5 15.5
    x multiflora Cathayensis R&W (AA) 1921.0AABCD 7.3 9.6 12.8
    x odorata Sweet (A)A AABCD 1.5 7.5 15.5
    x Wichuraiana Crep (AA) AABCD 0.08 0.23 0.08 9.2 14.9
    x Xanthina Lindl. (BB) 21 33.3 ABBCD 1.0 0.17 9.2 12.2
    x rugosa Thunb (CC) 18 88.8 ABCCD 0.45 11.2 10.8
    x blanda Ait (DD) 35 22.8 ABCDD 0.1 0.5 1.5 8.0 9.4
    x damascena trigentipetala D (AACC) AABCCD 0.34 0.67 10.0 18.7
    x R. lucida alba (AADD)* 16 56.3 AABCDD
0.5 13.0 14.0
    x heliophila Greene (CCDD) ABCCDD 2.0 1.3 8.7 12.7
    x foetida bicolor Willm (CCDD)** ABCCDD 1.0 1.0 11.0 13.0
    x laxa Retz (DDEE) 18 66.6 ABCDDE 1.0 12.2 13.6
    x Canina L (AABCD) 23 70.0 AABCD Breakdown at diakenesis
    x Coriifolia froebeli Christ (ACDEE) ABCDE 0.6 6.4 20.4
    x Täckholmii Hurst (AABBCCDD) 15 87.0 AABBCCDD 0.4 0.8 0.2 1.0 14.4 4.0
* Hurst gives R. virginiana Mill. (R. lucida Ehr.) as CCDD.
** Hurst gives this as BBDD


The genus Rosa is a freely interbreeding and very polymorphic group. The taxonomic characters show no clear cut mode of inheritance, and environmental influence is so great that it is difficult to separate the various forms into satisfactory taxonomic groups.

The chaos existing in the taxonomic classification led Hurst in 1925 to propose his septet theory, and all species studied were assigned to the five basic diploid sets or a combination of these sets. He did not imply cross sterility among septets but stated only that any pairing in the species or hybrid would be between similar septets. Assuming the above theory to be well founded it might be supposed further that a differential compatibility would be found in crosses between the various septets. Such was not found to be the case.

In explaining the origin of the modern roses Täckholm8 proposed the theory of descent through hybridization and postulated the existence of a primitive decaploid to explain the origin of Rosa Jundzillii Besser with 7" + 28'. Hurst4 agreed in assuming the existence of the decaploid species and proposed the theory that evolution had occurred from this hypothetical decaploid by successive loss of entire septets.

The writer believes that in assuming the existence of such a decaploid the previously mentioned authors are merely avoiding the issue, because it does not seem logical to start with a higher, more specialized polyploid and move downward to the lower and less specialized forms. Further, unless we assume that the decaploid arose from lower perhaps diploid forms, there would be no basis for considering it a polyploid; instead, we ought to regard it as a primitive diploid n = 35.

Since the known octoploid roses are Arctic species, Hurst believes that to assume the origin of our modern forms through ascent would be contrary to our knowledge of the origin of the Arctic flora. It is difficult to find logical reasons for such an assumption. The writer sees no great fallacy in picturing the Arctic as a once temperate region maintaining a large number of rose forms. With the advent of a rigorous climate the diploids or lower polyploids would be less likely to survive than the higher polyploids due to their lower possible number of beneficial mutations. The higher polyploids would therefore remain and through hybridization and further mutations give rise to new forms. In addition there seems to be little evidence that polyploid roses are of primitive origin or that they are especially long lived. Darlington2 has pointed out that the polyploid, and more truly the apomict, is exploiting a momentary advantage at the expense of its long term adaptability. His basis for such an assumption is that ploidy distinguishes species within a genus more often than genera within a tribe or tribes within a family.

Hurst's reasons why evolution of Rosa could not have occurred by ascent from diploid forms are quite obscure, and he is even less convincing in his reasoning for the descent from a primitive decaploid through successive loss of entire septets.

CybeRose note: "From the geological and genetical evidence it is highly probable that the polyploid species of Rosa were originally built up from older diploid species, either by hybridisation followed by the doubling of the chromosomes or by mutations and crossings in polyploid varieties or by both processes." Hurst: Mechanism of Creative Evolution, 1932.

Regarding the Caninae group of roses it seems equally as difficult to explain the origin of such a specialized group through hybridization between species with only one septet in common. If the Caninae did originate in this manner, it is difficult to understand why none of the species show 14 pairs and seven or 14 univalents. In addition, cytological studies on diploid hybrids indicates that in Rosa the various septets are quite similar.

Another unwarranted assumption regarding the Caninae roses is that they are chiefly apomictic but facultatively sexual. [Hurst did not make this assumption.] Assuming that the irregular Caninae roses have survived through apomictic seed production, it is most difficult to understand how such a reciprocal meiosis as exists in egg and pollen could survive the rigors of selection with no selective advantage. Rather than assume that the Caninae roses have survived in such manner the writer prefers to think of them as a sexually reproducing group maintaining their individual identities through a preponderance of genetic characters inherited through the egg cell. Further it seems reasonable to assume that within the two pairing genomes there exists a high degree of homozygosity. Self fertilization would result in little if any segregation among the offspring, and crossing would perhaps give rise to a high degree of sterility in egg and pollen.

The present cytological studies of hybrids of Rosa rubiginosa reveal complete breakdown of the parental meiotic behavior, with a consequent formation of more than seven bivalents and in addition a frequent occurrence of multivalent configurations. This meiotic behavior indicates the existence of a more complicated mechanism controlling pairing in the species of this group. The frequent occurrence of multivalent configurations in the hybrids is probably due to homology between chromosomes of the different septets within the genus, either as more than two homologous chromosomes or more than two homologous segments. The presence of reciprocal translocations might lead to multivalent associations, but this is merely a possible explanation of how the homology between the different septets originated.

From the foregoing results and discussion it seems quite evident that Hurst's septet theory is totally inadequate to explain the type of pairing found among the chromosomes in Rosa species and hybrids in general, and in the Caninae group in particular. According to Hurst's theory a number of the hybrids have a formula identical with or comparable to that of Rosa rubiginosa, but they show breakdown of the meiotic behavior with a serious reduction in fertility. Nor is the explanation of Gustafsson,3 based on internal autotriploidy, a satisfactory explanation because such a condition would lead to the occurrence of trivalents in meiosis of the Caninae species. A more satisfactory explanation than simple chromosome homology must be sought to explain the meiotic behavior in this group of Rosa.


In the writer's opinion the most likely explanation of the peculiar meiosis in the Caninae roses is that of control through an allelomorphic series of meiosis-regulating genes. It is assumed that each established species is homozygous for a separate allele of the series and that incompatibility between the chromosomes of the different species is the result of heterozygosity at this locus. Such a theory would explain the presence of only seven pairs of chromosomes in this group of species and would also account for the constancy of each species within the group and the separation of the group from others within the genus. It is assumed that such an allelomorphic series has been built up over a long period of time. The several species could have arisen as homozygous types either through selfing or through hybridization between two heterozygous types. Such species once established would remain relatively constant, and if they did hybridize the hybrids would be completely or highly sterile in nature. Such rare hybrids, if they did produce F2 plants, would give no viable segregates except types homozygous for the meiosis regulating genes, and these would be almost identical with the original seed parent. Backcrosses occurring by chance would be expected to provide a small population of segregates but these can hardly be considered as significant.

In the writer's opinion the data presented here on the behavior of chromosomes of Rosa rubiginosa and in hybrids between it and other species are explained satisfactorily by the hypothesis of an allelomorphic series of meiosis-regulating genes, as outlined above, but not by any hypothesis proposed heretofore.


Cytogenetic studies are reported on a number of interspecific rose hybrids with Rosa rubiginosa as the seed parent.

Crossing results indicate a high degree of compatibility between Rosa rubiginosa and other species of the genus.

Taxonomic characters distinguishing the species are dependent for their expression on a large number of genes or are the result of segregation in large allelomorphic series. In addition their expression is so modified by environmental factors that classification is exceedingly difficult.

Cytological studies of pollen mother cells shows a complete breakdown of the meiotic system in the Caninae roses following hybridization. The various theories explaining the peculiar meiosis are discussed and their weaknesses pointed out. Gustafsson's explanation, based on internal autotriploidy, is rejected because it does not exclude the possible frequent occurrence of multivalent configurations in the established species, which is contrary to fact. Hurst's theory precludes this occurrence of multivalents but involves an unwarranted regimentation of the various genomes in Rosa. Separation of species and groups in Rosa presents a sufficiently complex problem without the further complications involved in segregation of chromosomes into strictly defined septets. Neither theory explains the reciprocal meiotic behavior in pollen and egg. However, it must be admitted that, while Hurst's theory does not explain the pairing behavior in Rosa in general and in the Caninae in particular it is not entirely without merit. The first step toward an understanding of relationships among the various species of the genus should be its division into a small number of interrelated groups. Hurst's septet theory, based on aggregates of taxonomic characters and genetic tests furnishes a starting point toward a better understanding of the species and of their phylogenetic relationships.

In conclusion, an alternate hypothesis is offered to explain the peculiarities of meiosis in the Caninae roses. The hypothesis of genetically controlled pairing is more sound because it precludes multivalent configurations, explains the constancy of the several species within the group, and permits continued sexual reproduction. Finally it is the only hypothesis that explains the reciprocal behavior of pollen and egg.


  1. BALDWIN, J. T. Chromosomes From Leaves. Science N. S. 90: 240. 1939.
  2. DARLINGTON, C. D., and JANAKI AMMAL, E. K. Chromosome Atlas of Cultivated Plants. George Allen and Unwin Ltd. p. 397. 1945.
  3. GUSTAFSSON, ÄKE. The Constitution of the Rosa canina complex. Hereditas 30: 405-428. 1944.
  4. HURST, C. C. Chromosomes and Characters in Rosa and Their Significance in the Origin of Species. Experiments in Genetics pp. 534- 550. Cambridge, 1925.
  5. HURST, C. C. Differential Polyploidy in the Genus Rosa L. Handl. V. Inter. Kong. Vereb. pp. 866-906. 1928.
  6. HURST, C. C. Embryo-Sac Formation in Diploid and Polyploid Species of Roseae. Proc. Roy. Soc. Lond. 109: 126 48. 1932.
  7. TÄCKHOLM, GUNNAR. On the Cytology of the Genus Rosa. Svensk. Bot. Tids. 14: 301-311. 1920.
  8. TÄCKHOLM, GUNNAR. Zytologische Studien uber die Gattung Rosa. Acta Hort. Bergiani, 7: 97-387. 1922.

Riley [Wheat Septets (1958)] examined meiotic behavior in wheat that is similar to that in roses. Triticum vulgare is a hexaploid, but behaves like a diploid with few bivalents (pairs of chromosomes). However, when a monosomic line (41 chromosomes instead of 42) was crossed with a normal form (all 42), some of the derived haploids lacked one chromosome, and "the mean bivalent frequency exceeded the maximum observed in euhaploids". That is, a gene or genes in the missing chromosome apparently restricted pairing of homeologous chromosomes in euhaploids. The loss allowed pairing of chromosomes that usually do not.

Heslop Harrison: Durham Wild Roses (1954/1955)

Hurst Bibliography