Hereditas 134: 1-13 (2001)

Skewed distribution of morphological character scores and molecular markers in three interspecific crosses in Rosa section Caninae GUN WERLEMARK and HILDE NYBOM Balsgdrd - Department of Horticultural Plant Breeding, Swedish University of Agricultural Sciences, Kristianstad. Sweden

Werlemark, G. and Nybom, H. 2001. Skewed distribution of morphological character scores and molecular markers in three interspecific crosses in Rosa section Caninae. -Hereditas 134: 1 ~ 13. Lund, Sweden. ISSN 0018-0661. Received October 11, 2000. Accepted March 12, 2001

The dogroses, Rosa section Caninae, are all polyploid and characterised by their unbalanced meiosis; the pollen parent contributes one genome, whereas the seed parent contributes 3-5 genomes depending upon ploidy level of the species. As a result, genetically determined traits are expected to be matroclinally inherited. In the present study, the transmittal of genetic material was studied using manually scored reproductive characters (ovary and sepals), automated image analysis of leaflet shape (Fourier coefficients), and molecular markers (RAPD). The plant material consisted of a pair of reciprocal crosses between R. sherardii and R. villosa, a R. rubiginosu x R. sherardii cross and offspring obtained from selfing or within-population crosses of the parental species. All but one of the maternal markers were transmitted to all the offspring plants, whereas only 41 'YO of the paternal markers were transmitted to all of them, 23 '!h were never transmitted and 36 Yn reached only one or two of the offspring plants. Canonical variates analyses (CVA) based on the vegetative characters could easily separate the offspring groups representing parental species and also all three hybridogenous offspring groups from each other, whereas CVA based on reproductive characters failed to separate R. sherardii x R. villosa from its seed parent, but otherwise distinguished all offspring groups. The study shows that the expression of characters as well as molecular marker inheritance is dependent upon the direction of the cross, and on the species involved. It also demonstrates the importance of employing several different types of character sets for an improved comprehension of the effects of the peculiar canina meiosis.

Gun Werlemark, Balsg2rd - Depurtment of Horticultural Plunt Breeding, Swedish University of Agricultural Sciences, Fjiilkestudsviigen 123- 1, SE-291 94 Kristianstud, Sweden. E-mail: Gun.Werlemark@hvf.slu.se

Species in the genus Rosa exhibit a polyploid series with the basic chromosome number of 7. Euploids range from 2n = 14 to 2n = 56 and aneuploids are rare (ROWLEY 1960). The section Caninae, dogroses, contains only polyploid species with 2n = 28, 35 or 42 and with pentaploidy as the most common level. At the peculiar meiosis only 7 bivalents are formed while the rest of the chromosomes occur as univalents. The egg cells obtain 7 chromosomes from the bivalents plus the univalents, whereas the viable pollen cells only obtain 7 chromosomes from the bivalent forma- tion (T~CKHOLM 1920; BLACKBURN and HESLOP- HARRISON 1921). The male meiosis often fails, and the pollen quality within this section is reported to be very poor (JICINSKA et al. 1976). This unequal segre- gation of meiotic chromosomes is expected to result in a skewed distribution of inherited characteristics. Matroclinal inheritance of morphological characters has indeed been mentioned in several papers (GUSTAFSSON 1944; MELVILLE 1975; GRAHAM and PRIMAVESI 1993), but also intermediate inheritance for some characters (JICINSKA 1976). Apomixis has also been reported which further strengthens the ma- ternal influence (KROON and ZEILINGA 1974; WISSE- MANN and HELLWIG 1997). This low recombination

of genetic material appears to have led to the evolu- tion of numerous more or less homogeneous entities, which, in the beginning of the century, were allowed species rank (ALMQUIST 1919), but now have been merged to form fewer but more variable species (NILSSON 1967; ZIELINSKI 1985).

Morphological characters, on which earlier days rhodologists based their observations, are still used to define the rose species we have today and are usually scored with manual inspection of live or pressed material (NILSSON 1967; MALMGREN 1986). Lately, the scoring of shape variation with automated image analysis has proven to be a rapid and cost-effective way to measure morphological shape characters in several plant species. Automated image analysis of leaf material has been shown to discriminate between species (OLSSON et al. 2000; MCLELLAN 2000), be- tween intraspecific populations (SELIN 2000) and, to a more limited extent, between clones within the same vegetatively reproducing population (PERSSON and GUSTAVSSON 2001).

Among molecular markers, RAPDs have only been used for the last decade but have become increasingly important in genetic analyses. They have been suc- cessfully used for studies in interspecific relatedness in

2 G. Werlemark and H. Nybom Hereditas 134 (2001)

many plant groups, e.g. wild rose species (DEBENER et al. 1996; MILLAN et al. 1996) and also for studies of inheritance in e.g. Paspalum notatum (ORTIZ et al. 1997), and Nuphar x rubrodisca (PADGETT et al. 1998).

Studies on the morphological variation among and within Nordic dogrose species have shown that al- though closely related, these species could still be distinguished (NYBOM et al. 1996, 1997). Moreover, they differed considerably in level of intraspecific variation with Rosa rubiginosa L. being the most homogeneous both within and between populations, whereas R. dumalis Bechst. instead was the most heterogeneous and R. villosa Kell and Gams. showed heterogeneity among populations, but homogeneity within populations (NYBOM et al. 1997). A study using molecular RAPD markers placed five of the Nordic dogroses into three distinct groups; the R. canina group containing R. canina L. and R. dumalis, the R. villosa group containing R. sherardii Davies and R. villosa, and the R. rubiginosa group with R. rubiginosa (OLSSON et al. 2000). An automated image analysis using Fourier coefficients of the same mate- rial showed a similar pattern, but with some interspe- cific resolution also within the three groups.

The transmittal of morphological characters and molecular markers has been studied in detail using two Nordic dogrose species, R. dumalis and R. rubig- inosa (WERLEMARK et al. 1999). When these two species were crossed, pronounced matroclinal inheri- tance of both morphological characters and molecu- lar markers was evident in all the progeny plants. In the same study, about 10 % of the progenies did not inherit any of the molecular markers from their re- spective pollen parent, and they were therefore inter- preted to be of apomictic origin. Most of these plants also had the same medium-high pollen viability as their respective seed parents compared to a consider- ably lower pollen viability in the undisputably hybri- dogenous plants (WERLEMARK 2000).

In the present study, we wanted to more closely study the effects of the canina meiosis on the trans- mittal of different characters. Therefore, we investi- gated both morphological characters and molecular markers in progenies from three interspecific crosses; R. rubiginosa x R. sherardii, R. sherardii x R . villosa and R. villosa x R. sherardii. For comparison, proge- nies were analysed also from self-pollinated R. rubigi- nosa and R. villosa and from within-population pollinated R. sherardii. Of these species, R. rubiginosa is pentaploid and R . villosa is tetraploid whereas R . sherardii has been reported to be tetra-, penta- and hexaploid. The three progeny groups were studied with classical morphological measurements of repro- ductive characters, automated image analysis of

leaflet shape using Fourier coefficients, and with RAPD markers. This is to our knowledge the first time these three types of data have been scored on the same plant material.

MATERIALS AND METHODS

Plant material

Root suckers were collected in wild dogrose popula- tions in the south of Sweden in 1987 and 1988 (Table l), and the resulting plants were brought into a greenhouse in 1990. Numerous flowers were emascu- lated and a series of intra- and interspecific pollina- tions were performed. The obtained seeds were sown and stratified according to WERLEMARK et al. (1995) and the resulting seedlings planted into a randomised field in 1992. The interspecific crosses analysed in the present study include: R . rubiginosa x R. sherardii (1401 x 1402 and 1408 x 1402), R. sherardii x R. vil- losa (0315 x 0105) and R. villosa x R. sherardii (0105 x 0315) (Table 2). The pair of reciprocal crosses were made on two plants, where one plant was used as both seed parent and pollen parent, respectively. For comparison, offspring from selfing of R. rubiginosa (1401) and R. villosa (OlOl), and within-population pollination of R. sherardii (031 5 x 03 14) were also analysed. Unfortunately, progeny representing R . sherardii was thus available only for population 03 (used also in the crossing with R. villosa) and not for population 14 (used in the cross- ing with R. rubiginosa). Morphological measurements of reproductive characters were performed in the field in 1994. In the same year, three leaves from the current year’s shoots were collected and pressed for automated image analyses. Newly developed leaves from the parental plants and the progeny plants of the interspecific crosses were collected in 1997 and kept in -80°C for DNA extractions.

Table 1. Collection sites of dogrose populations. The first two figures of the ‘blunt no. ’’ denote the collec- tion site and the second two figures refer to individual plants within the collection site

Species Plant no. Collection site

R. rubiginosa 1401 Lemmestro, Skine R. rubiginosa 1408 Lemmestro, Skine R. sherardii 0314 Falkenberg, Halland R. sherardii 0315 Falkenberg, Halland R. sherardii 1402 Lemmestro, Skine R. villosa 0101 Tjornedala, Skine R. villosa 0105 Tjornedala, Skine

Hereditas 134 (2001) Transmittal of genetic inaterial in Rosa 3

Table 2. Number of plants used in the different analyses. In the RAPD study, DNA samples from the parental plants (0105, 0315, 1401, 1402, 1408) were also analysed

Crosses No. of analysed plants

Reproductive characters Vegetative characters RAPD

R. rubiginosa 1401 x R. rubiginosa 1401 R. sherardii 0315 x R. sherardii 0314 R. rubiginosa 1401 x R. sherardii 1402

R. sherardii 0315 x R. sherardii 0314 R. villosa 0101 x R. villosa 0101 R. sherardii 0315 x R. villosa 0105 R. villosa 0105 x R. sherardii 0315

16 27 17 18

27 27 53 6

R. rubiginosa 1408 x R. sherardii 1402

16 27 17 18

28 27 56 5

16 17

20 6

Chromosome determination

Since R. sherardii was reported to be tetra-, penta- or hexaploid, plant material from population 03 and 14 was used for determination of ploidy level with flow cytometry (Plant Cytometry Services). Plant material from population 03 proved to be tetraploid, whereas plant material from population 14 was pentaploid. As standards for this assessment R. rubiginosa (penta- ploid) from population 14 and R. wichuraiana (diploid) were used.

Morphological measurements of reproductive characters

The choice of morphological characters was based on previous investigations where they had proved to be useful in species discrimination (NYBOM et al. 1996, 1997). When the plants were in full bloom, three flowers (preferably the apical flower from three differ- ent inflorescences) were collected, the petals were removed and the length (1) and width (2) of the ovary was measured. Also the length of the pedicel (3) and the number of glandular hairs on one side of the pedicel and the ovary (4) was counted. The flowers were then transported to the laboratory where the sepals were photocopied. On the copies we counted total amount of sepal lobation (5) and length (6) and width (7) of one of the sepals, avoiding the largest and the smallest. The mean values of the three mea- surements from each plant were used in further calcu- lations. Thereafter the ratios of (2)/( 1 ) and (7)/(6) were calculated. See Table 2 for number of plants in each progeny group.

Pollen viability

In 1998, pollen viability was measured in the form of cotton blue (aniline blue lactophenol) stainability, on all 35 progeny plants of R. rubiginosa x R. sherardii (1401 x 1402 and 1408 x 1402) and on the three

parental plants (1401, 1408 and 1402). Three flowers from each plant were used. Each flower was pressed to a slide, cotton blue was applied with a cover slip on top and the slides were stored in a refrigerator. A total of 200 pollen grains were counted on each slide and the percentage of viable pollen grains estimated. Stained and non-stained pollen grains were easily separated as the stained grains were large and dark blue whereas the non-stained grains were small, miss- shapen and transparent. The mean percentage of viable pollen cells in each plant was calculated.

Automated image analysis

Leaflet characters were assessed with automated im- age analysis. Different types of shape descriptors can be used in this analysis, and here we chose elliptic Fourier coefficients to describe the shapes of the stored outlines. One leaflet from each compound leaf was used for automated image analysis and each plant was represented by the mean Fourier coeffi- cients of three leaflets. The outline of the leaflet was recorded with a video camera linked to a computer via an analogue-to-digital converter and the shape of the image was then described by coordinates of the image points (WHITE et al. 1988). The first ten elliptic Fourier harmonics, giving a total of 40 coefficients, were used as characters in the subsequent multivari- ate analyses (WHITE et al. 1988; WHITE and PREN- TICE 1988). The same plants which were used in the morphological measurements of reproductive charac- ters were also used for automated image analysis, except for one plant of the R. villosa x R. sherardii cross which was too malformed from powdery mildew to be used for measuring (see Table 2 for number of plants). The automated image acqusition and shape description procedures were carried out using the program ARB0 written by R. J. White (WHITE et al. 1988).

4 G. Werlemark and H. Nybom Hereditas 134 (2001)

Molecular markers

DNA was extracted according to HOLM (1995) and stored in 1 x TE buffer in + 4°C. Amplifications were carried out in 25 yl reaction mixture containing 1 x buffer solution no. 4 (Advanced Biotechnologies), 2.5 mM MgCI, (Advanced Biotechnologies), 0.5 yM primer (Operon Technologies), 0.2 yM nucleotide mix (Roche Diagnostics), approximately 30 ng ge- nomic DNA and IU Taq polymerase (Advanced Biotechnologies) overlaid with 25 pl of mineral oil. A Hybaid Omnigene Thermocycler was programmed for 5 min of initial denaturation at 94"C, then 40 cycles of 1 min at 94"C, 1 min at 36°C and 2 min at 72"C, followed by an extension time for 7 min at 72°C. The fragments were separated on a 1.8% agarose gel using TPE buffer stained with ethidium bromide and photographed under UV light. DNA from the parental plants for each cross and a molecu- lar weight marker (Roche Diagnostics MWM VI) were present on each gel. See Table 2 for number of analysed plants.

One hundred and thirty primers were screened for polymorphic markers between the parental plants, R. rubiginosa (1401 and 1408), R. sherardii (1402 and 315), and R. villosa (0105). For the analyses of R. rubiginosa x R . sherardii (1401 x 1402 and 1408 x 1402) 16 polymorphic markers were chosen, six spe- cific for R. rubiginosa and ten for R. sherardii, and for the analyses of R. sherardii x R. villosa and its recip- rocal (03 15 x 01 03 , 12 polymorphic markers were chosen, nine specific for R. sherardii and three for R. villosa. While it was relatively easy to find primers which were polymorphic between R. rubiginosa and R. sherardii, it was more difficult to find good and reliable markers separating R. villosa from R. sher- ardii and especially to find markers specific for R. villosa. There were no markers which separated the two R. rubiginosa seed parent plants 1401 and 1408 from each other, so the two progeny groups, 1401 x 1402 and 1408 x 1402, will henceforth be considered as a single group.

Statistics

The SPSS Data Analysis Package (NORUSIS 1990a,b) was used for calculations. A set of univariate analyses of variance were performed on the reproductive char- acters for all three progeny groups to ascertain sig- nificant differences among the groups. Pairwise comparisons within each of two sets of progeny groups 1) R. rubiginosa, R. sherardii, R . rubiginosa x R. sherardii and 2) R. sherardii, R. villosa, R . sher- ardii x R . villosa, R. villosa x R. sherardii, were carried out for each of the five reproductive charac- ters using a Scheffk a posteriori test which compen-

sates for groups with unequal sample sizes but tends to underestimate deviation from the null-hypothesis, i.e. that groups do not differ.

For obtaining maximum separation between the different progeny groups, canonical variates analyses (CVA) were conducted, followed by re-classification tests in which the plants are assigned to the progeny groups defined by the CVA scores. The canonical correlation showed how much of the total variation that was attributable to the between-group variation in each function, and Wilks' lambda was used as a measure of variability within the groups. A lambda value close to 0 indicates that most of the variability resides between groups, whereas a value close to 1 indicates that the variation resides mostly within groups. Two CVAs were carried out on each of the two sets of progeny groups 1) R. rubiginosa, R. sherardii, R. rubiginosa x R. sherardii and 2) R. sher- ardii, R. villosa, R. sherardii x R. villosa, R. villosa x R. sherardii using the five reproductive characters and the 40 elliptic Fourier coefficients, respectively.

RESULTS

Pollen viability

The two R. rubiginosa seed parents (1401 and 1408) and the R. sherardii (1402) pollen parent had a pollen viability of 26.0 YO, 19.0 YO and 23.8 YO, respectively. The pollen viability in the progeny group R. rubigi- nosa x R. sherardii varied between 1.5 and 8.7 YO ex- cept for one plant which had a viability of 31.8 %. The very low pollen viability in the majority of the progeny plants indicated that the meiosis was heavily disturbed and that the plants were of hybridogenous origin. A Student's t-test revealed that the pollen viability in the plant with a much higher pollen viability was significantly different from values ob- tained in the other 34 hybrid plants (p<O.OOl), thereby suggesting that the deviant plant had a regu- lar canina meiosis and might have arisen by apomixis.

Reproductive characters

R. rubiginosa x R. sherardii Comparison between the seedling groups obtained by selfing of R. rubiginosa (1401) and within-population pollination of R. sherardii (0315 x 0314) revealed sig- nificant differences in all reproductive characters ex- cept sepal lobation (Table 3). The hybrid progenies differed significantly from the seed parent progenies in ovary and sepal shape and in pedicel length. Moreover, they differed from both parental groups by having more glandular hairs. The CVA separated the three groups from one another (Fig. l), and the subsequent re-classification test placed 88 % of the

Hereditas 134 (2001) Transmittal o f aenetic inaterial in Rosa 5

Table 3. Mean values of the five chosen reproductive characters, and significant variation between progeny groups. The letters express a pair-wise Schefli comparison @ < 0.05)

Crosses Ovary width/ Pedicel No. of Sepal Sepal width/ length length glandular hairs lobation length

R. rubiginosa 1401 x R. rubiginosa 1401 0.73 a R. sherardii 03 15 x R. sherardii 03 14 0.65 b R. rubiginosa 1401 x R. sherardii 1402 0.63 b R. rubiginosa 1408 x R. sherardii 1402 0.61 b

R. sherardii 03 15 x R. sherardii 03 14 0.65 b R. villosa 0101 x R. villosa 0101 0.71 a R. sherardii 0315 x R. villosa 0105 0.65 b R. villosa 0105 x R. sherardii 0315 0.72 ab

10.27 a 8.83 a 12.59 b 22.94 b 12.61 b 36.04 c 11.96 b 37.85 c

12.59 b 22.94 b 5.91 a 14.43 a

12.12 b 30.45 c 7.72 a 40.94 c

12.21 a 0.17 a 13.14 a 0.19 b 12.35 a 0.19 b 11.26 a 0.20 b

13.14 b 0.19 bc 1.49 a 0.14 a

13.58 b 0.19 bc 7.72 a 0.17 ab

plants into the correct progeny groups. Most of the variation in this material was due to differences be- tween the groups, and comparatively little was due to within-group variation as shown by the high canoni- cal correlation in both functions as well as a Wilks’ lambda value of only 0.16 (Table 4). Surprisingly, the R. rubiginosa x R. sherardii hybrid group was placed closer to the progeny group representing the pollen parent than to the progeny group representing the seed parent (Fig. l), and not the other way around which could have been expected for matroclinally inherited characters. The single R. rubiginosa x R. sherardii progeny plant with a high pollen viability was assigned to the R. rubiginosa group, again indi- cating an apomictic origin.

R. sherardii x R. villosa and its reciprocal All the reproductive characters were able to discrimi- nate between the R. sherardii (0315 x 0314) and the R. villosa (0101 x 0101) progeny groups (Table 3). By contrast, the interspecific hybrid groups could not be separated from the progeny groups representing the seed parent in any of the characters except for num- ber of glandular hairs. The R. villosa x R. sherardii progeny group was intermediate between (although not significantly different from) the two progeny groups representing the parental species in ovary and sepal shape. By contrast, the R. sherardii x R. villosa group was significantly different from the group rep- resenting the pollen parent in all measured characters (Table 3). The two hybrid groups also differed signifi- cantly from each other in pedicel length and sepal lobation, suggesting an unequal parental contribution to character inheritance.

The low value for Wilks’ lambda together with the high canonical correlation value in the CVA, indi- cated that most of the variation was distributed among and comparatively little within groups (Table 4). The first function separated both the R . villosa and the R. villosa x R. sherardii groups from the

indistinguishable R. sherardii and R. sherardii x R. villosa groups (Fig. 2). In the second function, the R. villosa x R. sherardii group had somewhat higher scores than either parental group. The re-classifica- tion test assigned 72% of the plants to the correct groups. Most of the mistakes were made between the R. sherardii and R. sherardii x R . villosa groups, which is suggestive of matroclinal inheritance. The six R. villosa x R. sherardii plants appear to be more intermediate between the two parental groups, al- though three of the seedlings were quite close to the group representing the seed parent.

4 1

3 1

0 = R. sherardii X R. sherardii

0 = R. rubiginosa X R. sherardii

:; 0

E ‘t o + 0 0

8 0

0 0

0 0

-3 t I I I I I I I I

4 -3 -2 -1 0 1 2 3

Discriminant function 1

Fig. 1. Plot from CVA based on five reproductive charac- ters scored in progeny groups derived from R. rubiginosa (1401 x 1401), R. sherardii (0315 x 0314) and R. rubigi- nosa x R. sherardii (1401, 1408 x 1402). Each symbol repre- sents one progeny plant. Ir denotes the R. rubiginosa x R. sherardii plant lacking pollen-specific RAPD markers.

6 G. Werlemark and H. Nvbom Hereditas 134 (2001)

Table 4. Canonical variates analyses used to dijferentiate among dogrose progeny groups using j ive reproductive characters and 40 Fourier coefficients f r o m automated image analyses on leaflet shape

Group Discriminant Relative Canonical Wilks' p function percentage correlation lambda

Reproductive characters

R. rubiginosa (1401 x 1401), R. sherardii (03 15 x 03 14) and

1 2

R. rubiginosa x R. sherardii (1401, 1408 x 1402)

R. sherardii (0315 x 0314), R. villosa (0101 x OlOl),

R. villosa x R. sherardii (0105 x 0315)

R. rubiginosa (1401 x 1401), R. sherardii (03 15 x 03 14) and

1 2

3 R. sherardii x R. villosa (0315 x 0105) and

Vegetative characters 1 2

R. rubiginosa x R. sherardii (1401, 1408 x 1402)

R. sherardii (0315 x 0314), R. villosa (0101 x OlOl),

R. villosa x R. sherardii (0105 x 0315)

1 2

3 R. sherardii x R. villosa (03 15 x 0 105) and

1 <0.001

<0.001

88.0 0.88 0.16

12.0 0.56 0.68

94.3 0.93 0.09 i <0.001

<0.001 0.675

5.5 0.2

0.53 0.12

0.71 0.99

1 < 0.00 1

0.040

87.9

12.1

0.96

0.80

0.03

0.37

<0.001 I 62.5 0.94 0.01

21.2 16.3

0.84 0.81

0.10 0.35

0.246 0.471

o= R. sherardii X R. sherardii

= R. villosa X R. villosa

B= R. sherardii X R. villosa

= R. villosa X R. sherardii

A

A 3

N e 2 t

-2 l a 0

-3 t I

I I I I 1 I I I I

- 5 -4 -3 -2 -1 0 1 2 3

Discriminant function 1

Fig. 2. Plot from CVA based on five reproductive characters scored in progeny groups derived from R. sherardii (0315 x 0314), R. villosa (0101 x OlOl), R. sherardii x R. villosa (0315 x 0105) and R. villosa x R. sherardii (0105 x 0315). Each symbol represents one progeny plant. Ir denotes the R. villosa x R. sherardii plant lacking pollen-specific RAPD markers.

Hereditas 134 (2001) Transmittal of genetic inaterial in Rosa 7

Vegetative characters

R. rubiginosa x R. sherardii A CVA based on the 40 Fourier coefficients for leaflet shape clearly separated the two pure species progeny groups from one another in the first function (Fig. 3). The hybrid progeny group was placed very close to the R. rubiginosa group but had somewhat higher scores in the second function. The close posi- tioning of the hybrid seedlings and those representing the seed parent species is again indicative of matro- clinal inheritance. The R. rubiginosa x R. sherardii plant with the high pollen viability could not be distinguished from its siblings. In the re-classification test 99Y0 of the plants were assigned to the correct progeny groups. Most of the variation in this mate- rial occurred between groups since Wilks' lambda was only 0.03 (Table 4).

R. sherardii x R. villosa and its reciprocal The CVA based on the 40 Fourier coefficients sepa- rated the four progeny groups R. sherardii, R. villosa, R. sherardii x R. villosa and R. villosa x R. sherardii very efficiently (Fig. 4), with 99Y0 of the plants assigned to the correct groups in a re-classification test. Just as previously with the reproductive charac- ters, the R. villosa x R. sherardii plants took a posi- tion intermediate between their parental groups, whereas the R. sherardii x R. villosa plants showed a more expected matroclinal inheritance of characters and were placed adjacent to their seed parent group. Wilks' lambda in this CVA was very low, 0.01, indicating that almost all of the variation resided between and very little within groups (Table 4). The first function separated the parental species groups and placed the hybrid groups between them. The second function then clearly separated the two hybrid groups.

Molecular markers

One hundred and thirty primers were screened for markers that occurred in one parental plant but not in the other, for the species pairs R. rubiginosa-R. sherardii and R. sherardii-R. villosa. The two R. sherardii plants used for the interspecific crosses (03 15 and 1402), originated from two different populations, which had previously shown to differ in some mor- phological characters (NYBOM et al. 1996). Therefore, the polymorphism in molecular markers between these two R. sherardii plants was checked. A total of 671 markers was obtained but only 21 of these, i.e. 3 YO, differed between the two representatives of R. sherardii.

For the first species pair (involving the R. rubigi- nosa plants 1401, 1408 and the R. sherardii plant

0 = R. rubiginosa X R. rubiginosa

0 = R. sherardii X R. sherardii

0 = R. rubiginosa X R. sherardii 3 41-

I I I I I I I I I I I I I I

-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5

Discriminant function 1

Fig. 3. Plot from CVA based on elliptic Fourier coefficients of leaflet shape scored in progeny groups derived from R. rubiginosa (1401 x 1401), R. sherardii (0315 x 0314) and R. rubiginosa x R. sherardii (1401, 1408 x 1402). Each symbol represents one progeny plant. * denotes the R. rubigi- nosa x R. sherardii plant lacking pollen-specific RAPD markers.

1402), 800 markers were obtained and 160 of them were polymorphic, i.e. 20 YO, with none of the mark- ers showing any differences between the two R. rubig-

4 .i A

P -3 U 0

A *

= R. villosa X R. villosa

= R. sherardii X R. villosa

= R. villosa X R. sherardii

A -6

-7 A

A

I I I I I I I I I I I I

- 4 - 3 - 2 - 1 0 1 2 3 4 5 6 7 Discriminant function 1

Fig. 4. Plot from CVA based on elliptic Fourier coefficients of leaflet shape scored in progeny derived from R. sherardii (0315 x 0314), R. villosa (0101 x OlOI), R. sherardii x R. villosa (0315 x 01015) and R. villosa x R. sherardii (0105 x 0315). Each symbol represents one progeny plant.

8 G. Werlemark and H. Nybom Hereditas 134 (2001)

inosa plants. However, the majority of the polymor- phic markers were not clear and reproducible when DNA samples from the progeny plants were am-

Each of the chosen six markers specific for the seed parent R. rubiginosa (1401 and 1408), were transmitted to all the hybrid plants (Fig. 5). Of the ten markers specific for the pollen parent R. sher-

rdii (1402), five were transmitted to all the hybrid lants, three were not transmitted to any of the

E13-500

D15-300

D10-450

E2-1100

G12-650

I I I

t Fig. 5. Distribution of RAPD markers in progenies of R. rubiginosa x R. sherardii. The markers above the horizontal line are specific for the seed parent and the markers below the line are specific for the pollen parent. The rectangle represents 29 progeny plants with exactly the same RAPD markers and the four vertical lines represent the four deviating plants.

Fig. 6A and B. Distribution of RAPD markers in progenies of (A) R. sherardii x R. villosa and (B) R. villosa x R. sherardii. The markers above the horizontal line are specific for R. sherardii and the markers below the line are specific for R. villosa. The rectangle in (A) represents 19 R. sher- ardii x R. villosa progeny plants with exactly the same RAPD markers and the vertical line represent the deviating progeny. The six vertical lines in (B) represent the six R. villosa x R. sherardii progeny plants.

hybrids and two markers appeared in one-and two hybrid plants, respectively (Fig. 5). The progeny plant with high pollen viability did not inherit any of the pollen-specific markers.

For the second species pair, R. sherardii-R. vil- losa (involving plants 0315 and 0105), 710 markers were obtained and 54 of these were polymorphic, i.e. 7.6 %. As with the first species pair, most of these polymorphic markers were not reproducible when DNA samples from the two progeny groups were amplified. Of the chosen nine markers specific for R. sherardii, eight were transmitted to all the plants in the R. sherardii x R. villosa group and the ninth was transmitted to all but one of these plants (Fig. 6). In the reciprocal cross, R. villosa x R. sher- ardii, one of the six obtained plants lacked all R. sherardii specific markers and was therefore as- sumed to be of apomictic origin (Fig. 6). Of the remaining five plants, one received all but one of the nine R. sherardii specific markers, one plant re- ceived three markers and three plants received two markers. Only one of the R. sherardii specific mark- ers was never transmitted to any of the progeny plants (Fig. 6).

In total, all but one of the maternal markers were transmitted to all the offspring plants, whereas only 41 % of the paternal markers were transmitted to all but the two plants of presumed apomictic origin. Twenty-three percent of the paternal markers were never transmitted to any of the offspring plants and 36% reached only one or two of them. The ex- pected differences in transmittal of maternal and paternal genetical material, brought about by the canina meiosis, are clearly demonstrated by these molecular markers.

DISCUSSION

Interspecijic variation

Previous studies of the Nordic dogroses have demonstrated that the three species involved in the present study; R. rubiginosa, R. sherardii and R. villosa, are well separated with morphological char- acters (NYBOM et al. 1996, 1997). In the present

Hereditas 134 (2001) Transmittal of zenetic inaterial in Rosa 9

study, progeny groups from the two species pairs R. rubiginosa- R. sherardii, and R. sherardii- R. villosa, could be discriminated with most of the chosen mor- phological characters as demonstrated both by the univariate analyses of variance and the CVAs of both reproductive and vegetative characters.

In a previous MDS (multidimensional scaling) analysis based on RAPD markers, OLSSON et al. (2000) found R. rubiginosa to be well separated from both R. sherardii and R. villosa, whereas the latter two species were largely overlapping. In the present study, screening for DNA polymorphism between R. rubiginosa (1401 and 1408, these gave identical profi- les) and R. sherardii (1402) yielded only 20 % poly- morphic markers, with an even smaller number that appeared sufficiently clear and reproducible for inclu- sion in the actual analysis. Polymorphism between R. sherardii (0315) and R. villosa (0105) was even lower, with only 7.6 'YO of the initially scored markers being polymorphic. This is in accordance with previous reports of low levels of RAPD differentiation among wild dogrose species collected in Spain (MILLAN et al. 1996). Usually, interspecific polymorphism is much higher: MAGDALITA et al. (1 997) report overall 64 'YO polymorphic RAPD markers between two dieocious Carica species, and KO et al. (1998), using pre-selected primers, obtained 95 YO polymorphic markers among different outcrossing Viola species. Similar values are reported for intraspecific compari- sons in outcrossing species; KESKITALO et al. (1998) found 85 % polymorphism in the herbaceous Tanace- turn vulgare, SYDES and PEAKALL (1998) 58 'YO be- tween populations of the subtropical shrub Haloragodendron lucasii, and ROSSETTO et al. (1 995) reported 98 'YO in the subtropical shrub Grevillea scapigera. The level of interspecific RAPD polymor- phism found in our study is actually more similar to values reported for intraspecific comparisons among populations in selfing species: using pre-selected primers, BLACK-SAMUELSSON et al. (1997) found only 7.2 % polymorphic markers between two popu- lations of the rare and selfing Vicia pisiformis. Plants from the two populations of R. sherardii of different ploidy levels in the present study, showed an even lower RAPD differentiation, i.e. 3 'YO. Most likely, the situation with almost clonal species created by low levels of recombination is responsible for the high levels of homogeneity in dogroses. Highly homoge- neous taxa are often encountered also in apomictic species complexes (NYBOM et al. 1997), where recom- bination is virtually absent.

A recently performed microsatellite DNA analysis has shown that R. rubiginosa (from Fjalkestad, province of Skine) and R. sherardii (0315) differed considerably in four out of the six investigated loci

(NYBOM et al. 2000). By contrast, R. villosa (0105) had exactly the same alleles as R. sherardii (0315) in all six investigated loci whereas the latter had several additional alleles as well. This suggests that most of the R. villosa genome is contained within R. sherardii, and may also explain the difficulties in finding R. uillosa-specific RAPD markers. It may also indicate that R. sherardii has originated from a hybridisation between R. villosa (or a very close relative) as seed parent and an unknown pollen parent.

Interspecijic hybrids

R. rubiginosa x R. sherardii Rosa rubiginosa has previously proved to be very homogeneous in Scandinavia, with almost negligible morphological variability between and within popula- tions (NYBOM et al. 1997). Similarly, R. rubiginosa populations overlapped completely when analysed with RAPD, and showed almost nil within-popula- tion variation (OLSSON et al. 2000). The two R. rubiginosa plants (both from population 14) used in the present crosses, were therefore, in all likelihood, close to being genetically identical, as also suggested by their identical RAPD profiles.

Two morphologically different varieties of R. sher- ardii occur in the Nordic countries, var. urnbellifera and var. uenusta (NILSSON 1967). In the present study, only var. venusta was represented, with one plant from the pentaploid population 14, which was used in the interspecific hybridisation and two plants from the tetraploid population 03, which were cross- pollinated to provide seedling progeny of the pure species. A previous investigation showed, however, that these two populations differed significantly in several morphological characters (NYBOM et al. 1996). Great care must therefore be taken when comparing the morphology of the hybrid progeny in the present study with the progeny representing pure R. sherardii, even though the RAPD polymorphism between the two R. sherardii plants 0315 and 1402, was only 3 %.

The CVA based on five reproductive characters separated the R. rubiginosa x R. sherardii hybrid group from each of the pure species groups. More- over, the hybrids appear to be closer to the pollen parent R. sherardii than to the seed parent R. rubigi- nosa but with higher scores on the first discriminant function than either of the pure species groups. In the univariate analyses, four of the morphological char- acters similarly place the hybrids closer to R. sher- ardii than to R. rubiginosa, which is contrary to what was expected according to the matroclinal inheritance model. However, ovary width/length, sepal width/ length and number of glandular hairs may have varied between the two involved R. sherardii popula-

10 G. Werlemark and H. Nvbom Hereditas 134 (2001

tions as already mentioned (NYBOM et al. 1996). Only the fourth discriminating character, i.e. pedicel length, which was identical between the two popula- tions (NYBOM et al. 1996), can thus be taken as evidence that the paternal influence is as strong as, or even stronger, than the maternal. There are very few reports of paternal inheritance within section Can- inae. However, JICINSKA (1 976) reported paternally inherited features in shape and quality of the prickles in intersectional crosses with R. rugosa (sect. Cin- namomeae, diploid) as pollen parent. In a previous study on hybrid offspring obtained in the cross R. dumalis x R. rubiginosa and its reciprocal, a similar set of reproductive characters instead showed pro- nounced matroclinal inheritance in both crossing di- rections (WERLEMARK et al. 1999).

Analysis of vegetative characters showed that the hybrid progeny in the present study were consider- ably closer to the R. rubiginosa progeny than to the R. sherardii progeny, but with somewhat higher scores on the second discriminant function. Obvi- ously, these characters indicate the previously re- ported mode of inheritance; JICINSKA (1976) noted a maternal influence of leaf characters in her crosses using R. rubiginosa as seed parent and R. rugosa as pollen parent. GUSTAFSSON (1944) reached the same conclusion from his interspecific crosses. Unfortu- nately, neither of these authors reported any statisti- cally analysed data to confirm their observations. In studies of Forsythia, Prunus and Rhododendron hy- brids, MELVILLE (1960) could not find any evidence for leaf shape being inherited from either parent. The hybrids instead displayed a mixture of parental traits, depending upon dominance effects. However, due to the skewed chromosomal distribution in dogroses, we expect the R. rubiginosa x R. sherardii hybrids to contain four maternally-derived genomes and only one paternally-derived, and therefore also dominantly inherited traits would show mostly matroclinal inheritance.

When studied with RAPDs, the expected matro- clinal inheritance is clearly evident in the R. rubigi- nosa x R. sherardii progeny since all maternal markers were transmitted to the hybrids in accor- dance with previously investigated reciprocal crosses between R. dumalis and R. rubiginosa (WERLEMARK et al. 1999, WERLEMARK 2000). Only seven out of ten pollen-specific markers were transferred, but five of these then occurred in everyone of the hybrid plants which is in contrast to the previous study (WERLEMARK et al. 1999) in which none of the pollen-specific markers occurred in all the hybrid plants.

R. sherardii x R. villosa and its reciprocal The same R. sherardii plant was used both as seed and pollen parent in the two crosses, and as seed parent in the within-population cross to provide a pure R. sherardii seedling progeny. Two different R. villosa plants were used, one for interspecific hybridis- ation and one for selfing. However, these plants came from the same population and should be genetically very similar since previous investigations have demonstrated very low levels of within-population variation in this species (NYBOM et al. 1997).

In a CVA based on reproductive characters, the R. sherardii x R. villosa hybrids were indistinguishable from the R. sherardii progeny, at first leading us to believe that R. sherardii reproduced only by apomixis. By contrast, the R. villosa x R. sherardii hybrids were relatively well separated from the two pure species progenies on both the first and the second discriminant functions. In a CVA based on vegetative characters instead, all four progeny groups were separated from one another, with the R. sher- ardii x R. villosa group adjacent to the R. sherardii group, again suggesting a strong maternal influence on leaf characters.

All maternal RAPD markers were transmitted to all offspring plants in both crosses with the exception of one R. sherardii x R. villosa seedling, which lacked one R. sherardii-specific marker. This is highly indica- tive of a pronounced matroclinal inheritance. By contrast, only two of the three paternal markers were transmitted to all the seedlings in the R. sherardii x R. villosa group, whereas in the R. villosa x R. sher- ardii group, all pollen-specific markers, except one, were transmitted to at least one of the progenies. This is rather different from the previously studied crosses between R. dumalis and R. rubiginosa (WERLEMARK et al. 1999), where only about half of all the pollen- specific markers were transmitted at all. Presumably variation in genetic relatedness among the four, five or six basic genomes in each polyploid species, and among genomes present in the different species, is responsible for the differences in number of markers being transmitted, and the proportion of seedlings that inherit them.

Apom ixis

Two of the progeny plants in the present investiga- tion lacked everyone of the pollen-specific RAPD markers. One of them, a seedling in the R. rubigi- nosa x R. sherardii progeny, had a medium-high pol- len viability like the parental plants. This plant also displayed reproductive characters typical of R. rubigi- nosa, whereas leaflet shape was uninformative. Apomixis was suggested in a previous study on recip- rocal crosses between R. dumalis and R. rubiginosa,

Hereditas 134 (2001) Transmittal of genetic inaterial in Rosa 11

where plants lacking pollen-specific RAPD markers had the same pollen viability as pure species (WER- LEMARK 2000). These plants were also separated by two reproductive characters (sepal length and ovary width) from their siblings which received pollen-spe- cific markers (WERLEMARK et al. 1999). The apomic- tic origin of some of these plants has later been corroborated with microsatellite DNA analysis (NY-

The second plant lacking everyone of the pollen- specific RAPD markers in the present study was a seedling in the R. villosa x R. sherardii progeny. In the CVA based on reproductive characters, this seedling was close to the pure R. villosa seedlings. Unfortunately, it was not subjected to automated image analysis due to malformation from powdery mildew, and it then died before it could be checked for pollen viability.

Transgressive characters

All three hybrid groups showed an enhanced amount of glandular hairs on the ovaries and pedicels com- pared to the pure seedling progenies with which they were compared. Whereas this may be artefactual for the R. rubiginosa x R. sherardii hybrid (the R. sher- ardii seedling progeny was not derived from the same population as the R. sherardii plant actually used in the crossing), transgressive inheritance is definitely indicated in the R. sherardii x R. villosa cross and its reciprocal. The increased amount of glandular hair is in agreement with BLACKHURST (1948), who also reported heavier and more dense armature in hybrids between R. rubiginosa and various Rosa species (in- cluding four from sect. Caninae) as pollen parents. JICINSKA (1976) noted bigger flowers with a deeper colour than either parent as well as bigger hips in her intersectional crosses with plants from section Can- inae as seed parent, which corroborates the findings of GUSTAFSSON and SCHRODERHEIM (1944). It is not unusual for transgressive characters to appear in first generation hybrids. RIESEBERG and ELLSTRAND (1993) reported in a review paper that 64% of the first generation hybrids exhibited extreme characters in which they included both transgressive and novel characters. The cause for the extreme characters could be either (1) increased mutation rate in hybrids, ( 2 ) complementary action of normal alleles, (3) reces- sive genes present in heterozygous form in the parents becoming homozygous in the hybrid or (4) a reduced developmental stability, or it could be any combina- tion of all four (RIESEBERG and ELLSTRAND 1993). In our crosses, alternatives (2) and (3) appear to be the most likely causes. Since the genomes are so similar to each other, reduced developmental stability or increased mutation rate seems rather improbable.

BOM et al. 2000).

Comparisons between morphological and RAPD data

Different studies have shown that the expression of morphological characters may be either uni-parental or intermediate depending upon what character is being studied (MELVILLE 1960; RIESEBERG and ELL- STRAND 1993). Characters governed by one or two genes, e.g. presence vs. absence and major changes in structure and architecture, can appear as either uni- parental or intermediate in first-generation hybrids depending upon dominant vs. co-dominant expres- sion. In contrast, characters under polygenic control are predicted to be more intermediate (RIESEBERG and ELLSTRAND 1993). Reproductive traits are gen- erally considered to be of higher taxonomic value and less influenced by the environment than vegetative ones, probably because of a higher selectional pres- sure on the reproductive organs (see review in LERCETEAU et al. 1997). Especially the ovary and sepal characters are highly heritable, which would make them very useful in investigations like the present. In a previous investigation comparing differ- ent morphological characters among dogrose species, the reproductive characters (ovary and sepal charac- ters) showed more than twice as much interspecific differentiation as did the vegetative characters (manu- ally scored leaflet shape) (NYBOM et al. 1996).

RAPD markers represent a random sample of the genome and have generally been considered as neu- tral and located in both non-coding and coding DNA. This results in a large number of independent markers, which are dominantly inherited in a Mende- lian fashion although deviations from the expected 3:l or 1:l ratio are often reported (GOMEZ et al. 1996, JENCZEWSKI et al. 1997). The segregation dis- tortion found in dogroses is very large as expected considering the canina meiosis (WERLEMARK et al. 1999).

The reproductive characters used in this investiga- tion discriminated the R. rubiginosa x R. sherardii hybrid plants in the CVA, and clustered the pre- sumably apomictically derived plant with its maternal parent. Contrary to expectation, the hybrid plants were instead placed closer to the pollen parent than to the seed parent. The reproductive characters also distinguished the R. villosa x R. sherardii offspring plants, which were rather intermediate between the parent species representatives. However, plants from the reciprocal cross R. sherardii x R villosa, were inseparable from the plants representing the seed parent. By contrast, the vegetative characters ob- tained with automated image analysis separated all three hybrid groups from their respective seed par- ents. Both R. rubiginosa x R. sherardii and R. sher- ardii x R. villosa were placed adjacent to their

12 G. Werlemark and H. Nvbom Hereditas 134 (2001)

respective seed parents, whereas the R. villosa x R. sherardii group was once again placed between its two parental groups. Vegetative characters thus yielded better separation between groups, and a much higher re-classification percentage in the CVAs com- pared to the reproductive characters. However, these characters, analysed by elliptic Fourier coefficients, resulted in 40 variables in contrast to the five chosen reproductive variables, which may to a large extent explain the difference in resolution between the two character sets.

In contrast to the variable inheritance of parental morphological characters, the molecular markers demonstrated a pronounced matroclinal inheritance in all three crosses; all but one of the 59 progeny plants inherited all seed parent-specific markers. This result corresponds to the previous investigation of the reciprocal crosses between R. dumalis and R. rubigi- nosa (WERLEMARK et al. 1999). Only 7 7 % of the pollen parent-specific markers were transmitted to any of the progenies except the two plants of pre- sumed apomictic origin; 41 YO were transmitted to all but the two above-mentioned plants, and 36% reached only one or two plants. This differs from the transmittal of pollen parent-specific markers in the previously mentioned crosses between R. dumalis and R. rubiginosa (WERLEMARK et al. 1999). The trans- mittal of the molecular markers from the pollen parent to the hybrid progeny plants appears to de- pend upon the genome homogeneity among the parental species. But since the major part of the genetic material comes from the seed parent, the resulting hybrid progenies would be expected to show mostly maternally inherited morphological charac- ters. However, certain dominance or co-dominance effects could sometimes demonstrate a paternal inher- itance in some characters and thereby give raise to the differences in morphological character inheritance seen in our crosses.

ACKNOWLEDGEMENTS

We thank H. C. Prentice for allowing us the use of her image analysis equipment and R. J. White for allowing us the use of his computer program ARB0 for the automated image analyses. Financial support was given by the Philip Sorensson foundation and the Swedish Research Council for Forestry and Agriculture (SJFR).

REFERENCES

Almquist S, (1919). Sveriges Rosae. Idun, Stockholm. Black-Samuelsson S, Eriksson G, Gustafsson L and

Gustafsson P, (1997). RAPD and morphological analy- sis of the rare plant species Vicia pisiformis (Fabaceae). Biol. J. Linn. SOC. 61: 325-343.

Blackburn K and Heslop-Harrison JW, (1921). The status of the British rose forms as determined by their cytolog- ical behaviour. Ann. Bot. 35: 159-188.

Blackhurst HT, (1948). Cytogenetic studies on Rosa rubigi- nosa and its hybrids. Proc. Am. SOC. Hort. Sci. 52:

Debener T, Bartels C and Mattiesch L, (1996). RAPD analysis of genetic variation between a group of rose cultivars and selected wild rose species. Mol. Breeding 2:

Gomez R, Angel F, Bonierbale M, Rodriguez F, Tohme J and Roca WM, (1996). Inheritance of random amplified polymorphic DNA markers in cassava (Manihot escu- lenta Crantz). Genome 39: 1039-1043.

Graham GG and Primavesi AL, (1993). Roses of Great Britain and Ireland. BSBI Handbook No 7, Botanical Society of the British Isles, London.

Gustafsson A, (1944). The constitution of the Rosa canina complex. Hereditas 30: 405-428.

Gustafsson A and Schroderheim J, (1944). Ascorbic acid and hip fertility in Rosa species. Nature 153: 196-197.

Holm S, (1995). Unexpectedly high levels of genetic varia- tion in Potentilla argentea L. (s.1.) in southern Sweden. Hereditas 123: 127-139.

Jenczewski E, Gherdrdi M, Bonnin I, Prosperi JM, Oliviero I and Huguet T, (1997). Insight on segregation distor- tions in two intraspecific crosses between annual species of Medicago (Leguminosae). Theor. Appl. Genet. 9: 682-691.

Jicinska D, (1976). Morphological features of F, generation in Rosa hybrids 1. Hybrids of some species of the sect. Caninae with Rosa rugosa. Folia Geobot. Phytotax. Prdha 11: 301-311.

Jicinska D, Koncalova MN and Sykorova 0, (1976). Stud- ies in rose pollen 111. Pollen viability and germinability in eight Czechoslovak Rosa species, Preslia. Praha 48: 347-353.

Keskitalo M, LindCn A and Valkonen JPT, (1998). Genetic and morphological diversity of Finnish tansy (Tanace- tum vulgare L., Asteraceae). Theor. Appl. Genet. 96: 1141 - 1150.

KO MK, Yang J, Jin YH, Lee CH and Oh BJ, (1998). Genetic relationships of Viola species evaluated by ran- dom amplified polymorphic DNA analysis. J. Horticult. Sci. & Biotech 73: 601-605.

Kroon GH and Zeilinga AE, (1974). Apomixis and het- erogamy in rose rootstocks (Rosa canina L.). Euphytica 23: 345-352.

Lerceteau E, Quiros J, Soria J, Flipo S, Petiard V and Crouzilat D, (1997). Genetic differentiation among Ecuadorian Theobroma cacao L. accessions using DNA and morphological analysis. Euphytica 95: 77-87.

Magdalita PM, Drew RA, Adkins SW and Godwin ID, (1 997). Morphological, molecular and cytological analy- ses of Carica papaya x C. cauliflora interspecific hy- brids. Theor. Appl. Genet. 95: 224-229.

Malmgren U, (1986). Slaktet Rosa i Sverige. Sv. Bot. Tidskr. 80: 209-227.

McLellan T, (2000). Geographic variation and plasticity of leaf shape and size in Begonia dregei and B. homonyma (Begoniaceae). Bot. J. Linn. SOC. 132: 79-95.

Melville R, (1960). A metric study of leaf shape in hybrids: I. The leaf-shapes of some F, hybrids and their parents. Kew Bull. 14: 88-102.

Melville R, (1975). Rosa L. In: Hybridization and the flora of the British Isles. (ed. CA Stace) A. P., London, p. 2 12-227.

510-516.

321-327.

Hereditas 134 (2001) Transmittal of genetic inaterial in Rosa 13

Millan T, Osuna F, Torres AM and Cubero JI, (1996). Using RAPDs to study phylogenetic relationships in Rosa..Theor. Appl. Genet. 92: 273-277.

Nilsson 0, (1967). Drawings of Scandinavian plants 1 -8., Rosa L. Bot. Not. 120: 1-8, 137-143, 249-254, 393- 408

Norusis MJ, (1990a). SPSSa Base systems user’s guide. SPSS Inc, Chicago.

Norusis MJ, (1990b). SPSS@ Advanced statisticsTM user’s guide. SPSS Inc,*Chicago.

Nybom H, Olsson A and Werlemark G, (1996). Morphol- metric variation in Nordic dogroses (Rosa sect. Caninae, Rosaceae). Acta Univ. Ups. Symb. Bot Ups. 31: 59-68.

Nybom H, Carlson-Nilsson U, Werlemark G and Uggla M, (1997). Different levels of morphometric variation in three heterogamous dogrose species (Rosa sect. Caninae, Rosaceae). Plant. Syst. Evol. 204: 207-224.

Nybom H, Esselink D, Werlemark G and Vosman B, (2000). Inheritance of microsatellite DNA markers in hemisexual dogroses, Rosa L. sect. Caninae. In: Wer- lemark G, Genetic variability and reproductive strategies in Nordic dogroses, Rosa section Caninae. Doctoral thesis, Dept. of Horticultural Plant Breeding, Swedish university of agricultural sciences, Alnarp.

Olsson A, Nybom H and Prentice H, (2000). Relationships between Nordic dogroses (Rosa L. Sect. Caninae, Rosaceae) assessed by RAPDs and Fourier analysis of leaflet shape. Syst. Bot. 25: 511-521.

Ortiz JPA, Pessino SC, Leblanc 0, Hayward MD and Quarin CL, (1997). Genetic fingerprinting for determin- ing the mode of reproduction in Paspalum notatum, a subtropical apomictic forage grass. Theor. Appl. Genet. 95: 850-856.

Padgett DJ, Les DH and Crow GE, (1998). Evidence for the hybrid origin of Nuphar x rubrodisca (Nymphaeaceae). Am. J. Bot. 85: 1468- 1476.

Persson HA and Gustavsson BA, (2001). The extent of clonality and genetic diversity in lingonberry (Vaccinium vitis-idaea L.) revealed by RAPDs and leaf-shape analy- sis. Mol. Ecol. (in press).

Rieseberg L and Ellstrand NC, (1993). What can molecular and morphological markers tell us about plant hy- bridization? Crit. Rev. Plant Sci. 12: 213-241.

Rossetto M, Weaver PK and Dixon KW, (1995). Use of the RAPD analysis in devising conservation strategies for the rare and endangered Grevillea scapigera (Proteaceae). Mol. Ecol. 4: 321-329.

Rowley G, (1960). Aneuploidy in the genus Rosa. J. Genet- ics 57: 253-268.

Selin E, (2000). Morphometric differentiation between populations of Papaver radicatum (Papaveraceae) in northern Scandinavia. Bot. J. Linn. SOC. 133: 263- 284.

Sydes MA and Peakall R, (1998). Extensive clonality in the endangered shrub Haloragodendron lucasii (Halorago- ceae) revealed by allozymes and RAPDs. Mol. Ecol. 7:

Tackholm G, (1920). On the cytology of the genus Rosa. Sv. Bot. Tidskr. 14: 300-311.

Werlemark G, (2000). Evidence of apomixis in hemisexual dogroses, Rosa section Caninae. Sex. Plant. Repr. 12: 353-359.

Werlemark G, Carlson-Nilsson U, Uggla M and Nybom H, (1995). Effects of temperature treatments on seedling emergence in dogroses, Rosa Sect. Caninae L. Acta Agric. Scand., Sect. B, Soil P1. Sci. 45: 278-282.

Werlemark G, Uggla M and Nybom H, (1999). Morpho- logical and RAPD markers show a highly skewed distri- bution in a pair of reciprocal crosses between hemisexual dogrose species, Rosa sect. Caninae. Theor. Appl. Genet. 98: 557-563.

White RJ and Prentice HC, (1988). Comparison of shape description methods for biological outlines. In: Classifi- cation and related methods of data analysis, (ed. HH Bock), Elsevier Science Publishers, Amsterdam, p. 395- 402.

White RJ, Prentice HC and Verwijst T, (1988). Automated image acquisition and morphometric description. Can. J. Bot. 66: 450-459.

Wissemann V and Hellwig FH, (1997). Reproduction and hybridisation in the genus Rosa Section Caninae (Ser.). Rehd. Bot. Acta 110: 251-256.

Zielinski J, (1985). Studia nad radzajem Rosa L. System- atyka sekcji Caninae DC. em. Christ. Arbor Kornickie 30: 3-109.

87-93.