key: cord-0010484-41jdx4cd authors: ROBERTS, A. V. title: Relationship between species in the genus Rosa, section Pimpinellifoliae date: 2008-06-28 journal: Bot J Linn Soc DOI: 10.1111/j.1095-8339.1977.tb01184.x sha: 449140a00869d892e9ef6e262eb39d1e4d175ad9 doc_id: 10484 cord_uid: 41jdx4cd The morphology of twelve species of Rosa is described and similarities between these species are assessed. Possible origins of the tetraploid species from diploid species are indicated on grounds of comparative morphology. The wild origins of living and herbarium specimens are given in order to supplement published data on geographical distribution. Meiosis in pollen mother cells, viability of pollen grains at anthesis and ability to set seed was studied in several F(1) hybrids: no indication of complete or even partial sterility was found. Reproductive isolation is therefore unlikely to be maintained by reduced fertility of interspecific hybrids. Three species are reduced to synonymy with three other species, being retained as subspecific taxa. Two species are transferred from section Pimpinellifoliae to section Cinnamomeae. The genus Rosa includes over 100 species, distributed throughout the temperate and subtropical regions of the Northern Hemisphere. Their nomenclature is extremely confused (Rowley, 1959) , but according t o the popularly accepted classification of Rehder (1940), the genus has 4 subgenera: 3 are monotypic and the fourth, Rosa, includes all the remaining species grouped into 10 sections. The genus exhibits a typical polyploid series with a basic chromosome number of 7. Euploids range from 2n = 2x = 14, to 2n = 8x = 56 (Darlington & Wylie, 1955) and aneuploids are rare (Rowley, 1960a) . The present investigation was initiated in order to obtain data on the comparative morphology and geographical distribution of species in the section Pirnpinellifoliue of the subgenus Rosa, and t o identify pre-or post-zygotic sterility barriers which might reproductively isolate the species from one another. I t was hoped that the information obtained would provide a fresh perspective for taxonomic revision of this group. The investigation was basetl on fresh material at the University of Reading, and herbarium specimens which were examined at the Royal Botanic Gardens, Kew, and the British Museum (Natural History). Tlie section Pirnpinellifoliae includes nine diploid species ( R . ornciensis Rolfe, R. sericea Lindl., R . hugonis Hemsl., R . xanthina Lindl., R graciliflora Rehd. & Wils., R. primulu Boulenger, R. ccae Aitch., R. koreana Kom., and R . furreri Stapf.) and three tetraploid species (R. spinosissima L., K. hemisphaerica Herrm., and R . foetidu Herrm). Boulenger (1934) considered the tetraploid species R. nanotliamnus Boulenger, t o have affiliations with the Pimpincllifoliae, but Roberts (1975) showed that this species should be placed in the section Caninae. No living accession of R. gruciliflora was obtainable, and since reference t o living material was considered essential it was decided not to consider this species further. R. ~ernisr~~~aericu is merely a double-flowered cultivar of R. rapinii Boiss., and no other character has been described by which they can be distinpished (Rowley, 1959) . Its flowers have no functional anthers or carpels as these parts have been replaced by petals and populations of this mutant could not perpetuate themselves by seed in the wild. For this reason these two taxa have here been treated as one species, under the earlier name of R. hemisphaerica. Rehder ( Fig. 3 . The distributions of R. sericea, R. omeiensis, and R. spinossisima are each based on over 100 records, and those of both R. foetidu and R. hugonislR. xanthina are based on over 50 records. As there were only nine records of R . ecae, two each of R. hemisphaerica, R. koreana and R. forrestiana, and one of R. primula, they may well be more widely distributed than the above stat emen ts imply. All the specimens which were examined of R . farreri, had been obtained from plants in cultivation whose wild origins were unknown. Tackholm (1922) and Hurst (1925) drew attention to a pattern of distribution of species of Rosa which is related t o levels of polypoidy. They showed that in both the new world and the old world, diploid species are nearest t o the equator, tetraploid species generally have a more northerly distribution, while hexaploids and octoploids are nearest to the North Pole. The ranges of each of the three tetraploid species which fall into the I'impinellifoliae extend further West than the diploid species, but only that of R. spinosissima extends significantly further North. In order t o study meiosis in pollen mother cells, young flower buds were first fixed in 3 : 1, ethanol : acetic acid, hydrolysed in N HC1 for 10 minutes at 60°C, and then stained in Feulgen Reagent. Acetocarmine squashes of anthers were then viewed by phase contrast microscopy. The characters which are discussed are those which provide effective discriminants between the species under consideration. Fortuitously all the character states were amenable to qualitative definition and this enabled an assessment of similarity between the species to be based on Coefficients of Asso ciation. Taxonomic discriminants are listed in Table 1 , and the character states assumed by each species are presented in Table 2 for purposes of comparison. In Table 1 14 black * The red colour of the main stems and branches of R . primula and R . ecae can be distinguished, even on herbarium sheets, from the dull brown colour in the other species investigated. Newly formed branches of these other species are sometimes red in early spring, but darken after 2 or 3 weeks. t The number of leaflets per leaf was counted on each of ten leaves, which were arbitrarily chosen on each herbarium specimen. The highest and lowest figure for each herbarium specimen was then recorded. In 4 species and the R. hugonislR. xanthina assemblage, the number of specimens examined was large enough for meaningful histograms to be drawn (Pig. 4) showing numbers of spccimens in cach class of leaflet number. Modal classcs for highest numbers of leaflets were the same in R . sericea, R . spinosissima, and R . hugonis/R. xanthina, as also werc the modal classes for the lowest numbers. In R . omeiensis, the divergence between the modal classes of highest and lowest numbcrs of leaflcts is greater than in any of the other species. This might be a conscquence of thc generally higher values exhibited allowing more flexibility in numbers of leaflets. The arbitrary designations of distributions 1, 2 and 3 are used in Table 1 to distinguish the three types of distribution which were found. $ Only one living plant of R . forrestiana was available for examination. Its flowers had broad discs, t o which the filaments were partly fused, and its stamcns were consequently incapable of inflexion. As this character could only be satisfactorily studied on living material and in the absence of further information, an entry of 'no comparison' was made against this species in Table 2. * * Oval hips and short styles werc found together in the same species, so usually were globose hips and long styles. Style length was recorded shortly after the flowers opened, while the receptacle was at an early stage of development, and the hip was not recognisably globose or oval. ThereEorc the 'length of style' is not geometrically dependent on hip shape, i.e. thcy are not 'neccssary corrclates' in the sense of Sokal & Sneath (1963) . They wcre therefore recorded as separate Characters. character are fully defined and entered in a column which is headed by symbols (0, + or x ) t o facilitate the construction of Table 2. Coefficients of Association were calculated between all pairs of species, and are shown in Table 3 , where they are expressed as percentages, t o the nearest whole number. Each coefficient is the ratio of the number of character states two species have in common to the total number of characters on which they were compared, and is therefore an estimate of the degree of similarity between the species. For a full discussion and presentation of the different possible coefficients, see Sokal & Sneath (1 963: 128 et seq.). Where the symbol 'nc' (no comparison) was entered in Table 1 against a particular species, the denominator (total number of characters compared) was correspondingly reduced when coefficients of association were calculated. When two or three states of the same character were found in a single species, coefficients of association were calculated separately for minimum and maximum similarity and the mean value was taken. When two species showed the same states of a particular character on which they both varied, they were said not to differ, and t o have maximum similarity on that character. The similarity between species and groups of species is expressed in the form of a dendrogram of affinities in Fig. 5 . In constructing this dendrogram, the Table 2 . Character states found in the species. (See Table 1 for the character states referred to by the symbols 0, species with the greatest similarity were the first t o be united. The group or groups so formed were linked progressively with other species or groups of species with which they showed the greatest similarity. The similarity between a species and a group is given as its average similarity with each member of the group. The similarity between two groups was calculated as the average similarity of each member of one group with each number of the other group This method is described and discussed by Sokal & Sneath (1963: 180 Figure 5 . Dendrogram of affinities. method of clustering by single linkage'. This method seemed well suited for this study as a relatively small number of taxa are involved, and it was desirable t o express affinities in two dimensional form. In attempting to detect barriers t o breeding which might serve to isolate species, it is clearly appropriate t o establish whether or not fertile inter-specific hybrids can be raised. Rowley (unpublished) raised vigorous hybrids listed above between some species of the Pirnpirzellijdiae and found no evidence to suggest that others could not be raised between further combinations of species. Meiosis, viability of pollen at anthesis, and ability to set seed was studied in these hybrids as indices of their fertility. In order to provide a standard of normality with which to compare the hybrids, these indices were also studied in seven diploid and two tetraploid species. No further hybridization was attempted in this investigation as mature flowering plants could not have been raised within the time scale of this investigation. Meiosis was studied in pollen mother cells, particular attention being paid t o diakinesis and metaphase I (Table 3) , as the regularity with which bivalents were formed would reflect the degree of homology between parental genomes and would indicate whether or not there was likely t o be balanced seyegations of homologous chromosomes. The viability of pollen at anthesis was assessed visually after staining in acetocarmine. Some grains failed to take up the stain and were shrivelled, indicating that they were moribund Others stained deep red and were plump indicating thay they were viable at this stage in their development. Estimates of the proportion of viable pollen grains ( Table 4 ) were based on samples of not less than 900 grains. Values obtained for different flowers on the same plant were generally consistent and infrequently differed by more than 7%, either within a single growing season or from year to year. However, estimates of (Table 4 ). There are many environmental factors which might affect the production of seeds and this value cannot be regarded as a sensitive index of a plant's fertility. However, an observation that an F, hybrid never, or only rarely produced seed might be taken as evidence that it is incapable of completing its cycle of sexual reproduction. Such failures might be due to irregularities in megasporogenesis, which might be strongly suspected if irregularities had been detected in microsporogenesis. Alternatively, it might be due either to the inability of stylar tissue to support normal growth of a pollen tube, or to an inability t o support normal development of a zygote. As a result of terminalization, the frequency of chiasmata at metaphase I was lower than at diakinesis in each of the plants which was studied at both stages. In all species and hybrids, most cells at diakinesis and at metaphase I of meiosis had seven bivalent associations. Univalent chromosomes and/or multivalent associations were found in some species and hybrids, but in relatively low frequencies. As the frequency of univalent chromosomes in the F, hybrids was no higher than in species, univalents provide no clear evidence of limited homology between parental genomes; it is more probable that these univalents resulted from random failure in the formation of chiasmata leading to a complete lapse of association between paired chromosomes at diplotene. Multivalent associations were seen in five of the diploid species and all seven of the diploid hybrids. These took the form of trivalent or quadrivalent associations indicating heterozygosity for at least one translocation, and hexavalent associations indicating heterozygosity for at least two translocations. In only two plants, R . sericea and R . omeiensis, were no multivalents recorded. Studies in the genus Oenothera, reviewed by Cleland (1962) , have shown that chromosomes in multivalent associations may be so oriented at metaphase I of meiosis, as to ensure an ordered segregation at anaphase I, leading t o the formation of balanced gametes. However, such a mechanism demands a regularity in the formation of multivalents and an alternate configuration of chromosomes at metaphase I, neither of which were evident in any of the plants discussed here. Thus it might be anticipated that in each of the plants with multivalent associations a proportion of segregants would be unbalanced and that this would result in the production of unbalanced gametes. However, this expectation was not borne out by studies of pollen and seed, as several plants in which multivalents were found compare favourably with R. sericca and R. omeiensis with respect t o viability of pollen and/or numbers of seed per hip (Table 4) . Furthermore, as multivalents were found in five out of the seven diploid species examined in this investigation and as Erlanson (1933) has reported multivalents in plants of wild origin of R. blanda, R. woodsii and R. pisocarpa it seems that natural selection does not rigorously eliminate translocation mutants in natural populations of roses. The only immediate conclusion that can be drawn from them is entirely negative: the appearance of multivalent associations in the hybrids constitutes neither proof that the genomes of the parent species have diverged significantly during evolution, nor evidence that gene flow between the parent species would be restricted. In those plants representino the seven diploid species, the highest percentaze of viable pollen was recordedfn R . farreri (97.4%) and the lowest was recorded in R. hugonis (62.2%). In the hybrids R . sericea x R. prinzula 7.56., R. xanthina x R sericea 18.57.E., R. primula x R. ecae 40.57.A., R. primula x R. hugonis 13.57.1., and R hugonis x R. xanthina 8B.56.) the percentage of viable pollen fell within this range and several seeds per hip were set. However, in the hybrid R. primula x R. ecae 40.57.B. there was only 48.6% viable pollen and in the hybrid R. sericea x R. xanthina 13.59.) there was only 22.2% viable pollen. In the case of R . prirnula x R. ecae 40.57.B., there were, on average, 6.9 seeds per hip and there is thus no question of there being a complete barrier t o gene flow between the parental species through F2 sterility. Its sib. R. primula x R. ecae 40.57.A., was apparently highly fertile, having 95.2% viable pollen and on average 4.7 seeds per hip, so there is no unequivocal evidence that hybrids between R. primula and R. ecae suffer even reduced fertility. The hybrid R. sericeu 0 x R. xanthina d 13.59.) had fewer viable pollen grains (22.2%) than any of the other species or hybrids and no seed developed to maturity during the three year period of the investigation. In the absence of further information it might have been concluded that this was indicative of a sterility barrier between R. sericea and R. xanthina. However, the reciprocal hybrid R. xanthina 0 x R. sericea d 18.57.E., had a high frequency of viable grains (94.4%) and, on average, there were 5.4 seeds per hip. The reason for the sterility of R sericea 9 x R. xanthina d 13.59., is obscure. The frequency cf chiasmata at metaphase I was higher than in the other plants examined but otherwise meiosis was unexceptional. The plant seemed healthy and there was no sign of any disease which might have reduced its vigour. Possibly the maternal cytoplasm of R . sericea differed in some important respect from that o€ R. xanthina. In order to assess whether or not sterility of Fl plants might, in some degree, reproductively isolate the parental species, it would be necessary to assess the fertility of a wider selection of F1 hybrids between them. Apart from the hybrid R. sericea x R. xanthina 13.59., only one diploid plant, R. koreana M87, failed to produce any seed. This accession is known to be of wild origin, meiosis in pollen mother cells was in no way exceptional and a high proportion (94.5%) of pollen grains were viable at anthesis. It therefore seems that some environmental factor, possibly climatic, is responsible for its failure to set seed. Nevertheless this observation serves to show that caution is needed in assessing the viability of a plant outside its natural environment. Univalent chromosomes and multivalent associations were found in both accessions of R. spinosissima, in R. foetida and in the hybrid R. spinosissima x R. foetida. Whereas multivalent associations in a diploid plant is indicative of heterozygosity for a translocation, in a tetraploid plant it might be due either to heterozygosity for a translocation or t o segmental allopolyploidy. In either case, the formation of multivalents might be expected to lead to a loss of fertility. However, as in the diploid species and hybrids which were studied, there was no clear evidence that appearance of univalents or multivalents was necessarily associated with loss of fertility, because in clone K1451 of R. spinosissirna, 84.5% of pollen was viable at anthesis and there were, on average, 10.7 seeds per hip. AIso, although only 50.6% of pollen was viable in clone S403 of R. spinosissima a high average of 13.9 seeds per hip was recorded. The fertility of clone SG of R. foetida was strikingly low. Only 27.5% of pollen was viable at anthesis and no seeds were found in the hips. This species is grown in many gardens in Britain, where it is well known for its low fertility. No successful attempt to use it in crosses as an ovular parent has been recorded in the literature. However, as a pollen parent it is not totally barren; it has been used in breeding several garden roses (Harms, 1956) and it was the pollen parent of the hybrid R. spinosissirnu x R. foetida 25.59. In this hybrid, the viability of pollen grains at anthesis was high (72%) and there were, on average 3.3 seeds per hip. This hybrid is clearly more fertile than R. ,foeticla and these observations underline the strangely low level of fertility in that species. The dendrogram of affinities (Fig. 5 ) , shows a primary dichotomy between one group formed by R. forrestium, R. furreri and R koreana and a second group formed by all the other species. It can also be seen that the three species of the first group are closely similar to one another (united at the 86% level of similarity). The evidence that R. farreri and R. koreuna are phenetically more closely related t o R. jorrestiuna than to the other Yimpirzetlifoliat! is therefore strong. Rosa f'orrestiunu, R. furreri and R. koreana, each of which is diploid, share 7 character states (represented in Table 2 by the symbol + for characters 9, 10, 19, 22, 26, 32 and 34) which are not found in any of the other diploid species studied here. The simplest interpretation of this phenomenon is that they form a monophyletic group which diverged from the other diploid species before or when differences in these morphological characters arose (Table 6) . As R. furreri and R. koreana are both phenetically and cladistically closely related to R . forrestiunu, and since they appear t o have no obvious affinity with any of the remaining 11 sections of the genus, it is proposed that they should be placed witii R. forrestiarzu in the section Cinnarnomeae. Each of the tetraploid species combine character states which are otherwise polarized in one or other of the two diploid groups which were presumed, above, t o be monophyletic, and may thus have arisen as amphidiploid hybrids between them. This theory could be tested by attempting t o resynthesize the tetraploid species experimentally through hybridization and doubling the number of chromosomes. Chromosome pairing in R. spinosissirnu was quite regular, which suggests that either the genomes of the putative parental groups were sufficiently distinct for pairing to occur preferentially between genomes of common origin, or that after the initial tetraploid was formed, genomic or genotypic changes took place which stabilized meiosis, ensuring regular formation of bivalents. A study of meiosis in an F, hybrid between the presumed monophyletic groups of diploid species would have helped t o clarify this issue and it is unfortunate that such a hybrid was not available in this investigation. The irregularity of meiosis and consequent sterility of R. foetida indicate that neither of the alternative assumptions made for the evolution of R. spinosissima applies in this species. In view of its sterility it is pertinent to record comments written on herbarium sheets by collectors of specimens kept at the herbarium of the Royal Botanic Gardens, Kew. Two collectors in Iraq stated, respectively, that this species was grown as a hedge plant and that petals were used in making jam. A collector in Turkey commented on its popularity in gardens and as a hedge plant and another stated that he had not seen it growing wild and doubted if it did. These observations indicate that R . foetida is a popular domestic plant and may be either rare or non-existent in the wild condition. It therefore seems possible that its propagation is principally clonal and that there is not a high selective premium on fertility. R. spinosissima extends further to the north and west than the diploid species, although it partially overlaps the ranges of R. hugonis/R. xanthina and R. koreana. If, as Vavilov (1951) has postulated, the centre of evolution of roses is in S.E. Asia, where most of the diploid species of the genus are concentrated, R. spinosissima may have exploited the more remote and possibly less hospitable territories of Mongolia and Europe and thereby avoided competition with the longer established diploid species. The two other tetraploid species, R. fbetida and R. hemisphaerica, have a more westerly distribution than the diploid species and the same principle may be invoked t o explain this. Although no hybrids between diploid and tetraploid species were studied, it is likely that they would be largely sterile, but it would be wrong to assume that they would necessarily be totally sterile. Wulff (1954) and Rowley ( 1 960b) have reported that although triploid roses are usually relatively infertile, they occasionally give rise to fertile diploid or tetraploid offspring. Work on other genera has indicated the possibility of introgression between diploids and tetraploids through partially fertile triploids. For example, Jones parental species, they are incomplete. R. sericea sensu amplo, R. crae sensu amplo and R. .rantkina sensu amplo were found to differ consistently in their morphology. If these species were not, by some means, reproductively isolated, it might be anticipated that these differences would be blurred by introgressive hybridization. As no evidence of effective sterility barriers was found in these studies of F, hybrids, it is postulated that reproductive isolation may be effected through spatial separation. Distributions of the three species overlap in China, but separation may be achieved in the region of overlap, perhaps because of differ en t ecological preferences. Rowley (1959) united R. sericea and R. omeiensis under the earlier name; R . sericea. These species were also found to be closely similar in this investigation (92%), but the similarity between R. primula and R. ecae is even greater (97%) and R. hugonis is so closely similar to R. xunthina that it is virtually impossible to distinguish herbarium specimens of the two species. As R. sericea and R. omeiensis have been united, it would be consistent t o unite R. primula with R . ecae and R. hugonis with R . xanthina. No evidence of sterility barriers between R. Jirimula and R. ecae emerged from studies on the two F, hybrids R. primula x R. ecae 40.57.A. and 40.57.B., or between R. hugonis and R. xanthina from studies of the F, hybrid R. hugonis x R. xanthina 8B.56., so no a priori argument can be advanced against uniting the species of either pair. Although the evidence presented accords with Rowley's decision to unite R. omeiensis with R . sericea omeiensis (Rolfe) Roberts comb Lindley in Rosarum Monographia Rolfe in Curtis's Botanical Magazine I t is proposed that R. primula and R. ecae should likewise be united and that the two subordinated taxa should also rank as subspecies for similar reasons, viz. Rosa ecae Aitch Bulletin du Jardin botanique cle I'Etat Bruxelles xanthina do not deserve separate status as species and it is proposed to unite them, retaining the former at the lowest rank of forma Hemsley in Curtis's Botanical Magazine representing a cross between R. ecae sensu amplo and R. hugonis sensu amplo nor R. sericea x R. primula 7.56., representing a cross between R. sericea sensu amplo and R . ecae sensu amplo, showed any sign of F, sterility. Furthermore, it was concluded from a study of R . xanthina x R . sericea 18.57.E. and R . sericea x R. xanthina 13.59., which represent crosses between R. sericeu sensu amplo and R Roses d'Asic. Bulletin du .lardin botanique de I'Etat The cytogenetics of Oenothera Chromosome atlas of flowering plants Chromosome pairing, structur-l hybridity and fragments in Rosa. Botunicul HARMS, F., 1956. The origin of Rosa pernetiana Experiments in genetics Chromosomal status, gene exchange in Dactylis, 3. The role of the REHDER, A., 1940. Manual of cultivated trees and shrubs Planrae Wilsonianae The nature and taxonomic significance of the system of Principles of numerical taxonomy Zytologische studien uber die gattung Rosa The origin, variation, immunity and breeding o f cultivated plants Arnold Arboretum No. 4. nanothamnus (Rosaceac)