Euphytica 32 (1983) 205-216
Canada Agriculture, Research Branch, Research Station, Morden, Manitoba ROG 1J0, Canada Received 27 April 1982


In breeding for color and winter hardiness in Rosa, more than 1200 progeny from 47 families were analyzed for anthocyanin pigments. Cyanin, peonin and pelargonin were found in 99%, 52% and 31% respectively, of the seedlings. Each pigment was highly heritable from seed or pollen parents or both. All showed quantitative inheritance, particularly cyanin and peonin. A system is proposed to explain most of the synthetic pathways and controls for anthocyanin production in roses.


The rose breeding program at the Canada Agriculture Research Station at Morden, Manitoba has the objective of producing hardy attractive roses for the Canadian prairies. A range of pleasing colors has been sought as most existing hardy shrub roses bear pink flowers.

There is little doubt that flower color in roses has been studied for hundreds of years but genetic results from this group of hybrids have been hard to interpret due to complex interactions of pigments giving numerous shades of color (LAMMERTS, 1945, 1960). The identification by HARBORNE (1961) of peonin as a third anthocyanin in Rosa exposed a further level of complexity. The three main anthocyanin pigments found in Rosa cyanin, pelargonin and peonin, are 3-5 diglucosides of the respective cyanidins. They also occur as 3 glucosides. Surveys of the occurrence of cyanidin, peonidin and pelargonidin totalling many hundreds of cultivars and species (ARISUMI, 1963; YOKOI, 1974; DE VRIES et al., 1974; MARSHALL, 1975) have shown the distribution of these pigments in Rosa and the occurrence of cyanidin in all pink or red roses. DE VRIES et al. (1980) demonstrated that the inheritance of quercetin and kaempferol, the flavonol precursors of the anthocyanidins cyanidin and pelargonidin, was controlled by additive gene action.

HARBORNE (1961) refuted earlier claims for the presence of myricitin, a precursor of deiphinidin, and found only two flavonols, quercetin and kaempferol. These two were also observed by DE VRIES et al. (1974, 1980). HARRISON & STICKLAND (1978) showed that anthocyanin synthesis can be blocked at three stages in Antirrhinum and that methylation and the addition of the 5' hydroxyl appeared to occur at a late stage in the biosynthetic pathway. In Mathiola incana, FORKMAN (1980) found hydroxylation was achieved at the dihydroflavonol stage.

Information is presented here on the heritability of cyanin, peonin and pelargonin in roses and on the possible relationships between peonin and pelargonin.


In the thirty year period from 1950 to 1980, a large number of crosses were made involving several generations from Floribunda, Grandiflora and Hybrid Tea roses crossed with hardy species and cultivars such as R. arkansana PORTER and R. laxa RETZIUS through 'Prairie Princess' (intr. BUCK, 1972). Recently R. rugosa THUNB. was introduced to the gene pool through tetraploid RSM K1 (MARSHALL, 1980) and through R. kordesii WULFF. Controlled crosses were made and recorded in a way which facilitated tracing the parentage of any cross or selection.

Pigment analyses as previously described (MARSHALL, 1975) have been performed since 1973 on all parents currently in use and on those available from earlier work. Samples for analyses were usually taken from all plants blooming on a given day or in a few cases on a second day. Pigment separations on paper were rated from 0, no visible color to 5, intensely colored. An extract from 'Cuthbert Grant' (intr. Canada Agriculture, 1967) was included on every sheet of 15 extracts as the standard of rating 4 for cyanin. The methods used were considered suitable for screening fairly large numbers for exceptional colors but not particularly effective for observing small amounts of pigment, especially cyanin which does not fluoresce visibly in long wave UV.

Segregation ratios of progenies from families of greater than 10 seedlings and with known parents were compared to expected ratios for tetraploid segregation. Pedigrees for up to 6 generations were developed for many crosses and for some named cultivars based on Modern Roses 8 (MEIKLE, 1980). Correlations between pigment ratings were run among 1128 progeny to show interactions among the pigments. Regressions of parents on progeny were used to calculate heritability estimates for each anthocyanin, and to indicate genetic interactions between pigment ratings.

The parents used in these crosses, their parent numbers, anthocyanin ratings and the number of times each was used as seed or pollen parent are shown in Table 1. Each breeding number consists of eight characters, the first two are the year the cross was made, the second two represent the seed parent, the third pair identify the pollen parent and the last two are the selection number.


All families showed wide variation for both the type and amount of pigment present. Numbers of plants in the various pigment classes for each family are presented in Table 2. Cyanin was the most abundant anthocyanin found. It was present in at least 99% of all seedlings tested and in many more shades than indicated by the six class rating system used. Most of the remaining 1% of the seedlings contained some unidentified pink color, possibly cyanin, at too low a level to be identified by the method used. Cyanin was not identified or suspected in only 2 white flowered members of family 43, and 3 members containing peonin in families 39 and 47. It was the only red pigment observed in family 15. Peaks in segregation frequencies for cyanin varied with the mean of the parents but no family distribution ranged over the full range of ratings from 0 to 5 (unpublished data). Cyanin occurred as the only identified anthocyanin present or together with peonin and/or pelargonin.

Table 1. Parents of 36 progenies, their anthocyanin ratings and frequency of use as seed or pollen parents.

Breeding number or cultivar name Anthocyanin ratings1 Frequency of
cyanin peonin pelargonin seed pollen
09 54020101 Assiniboine 4 1 0 0 1
40 Independence 3 0 3 1 0
45 Fire King 2 0 3 1 0
62 64453402 4 0.5 0 3 1
65 64453405 (Adelaide Hoodless) 4 1 0 4 0
66 63401603 5 0 0 2 1
67 63480904 4 2 0 0 3
72 6734P001 4 0.5 0 0 4
84 64493407 3 1 0 1 2
89 6734P002 4 3 0 1 1
99 70934001 2 0 3 1 0
A4 Prairie Princess 2 0 0 2 0
A5 70666201 3 0 2 1 1
A7 6965RP02 3 1 2 0 4
BO Mount Shasta 0 0 0 0 0
C7 68656401 5 1 0 1 1
D7 70666203 (Morden Amorette) 4 1 1 0 7
E0 71A48401 2 2 0 3 1
F4 73E0D701 4 1 1 3 0
F5 73E0D702 4 0 3 1 1
GI 72A48401 0.5 0 0 3 0
HO 74D38901 3 3 0 1 0
JO 74EOE401 4 2 0 2 0
J1 74E0E402 4 1 3 2 0
J3 74A4D701 0.5 0 0 1 2
J4 74E0E403 (Morden Cardinette) 3 4 0 0 5
Q1 7799K501 3 0 0 1 0
R0 78J0J401 4 4 0 0 1
 1 Ratings: 0 = no pigment, 5 = intensely colored.

Peonin occurred frequently among the parents used in these crosses and was found in 586 or 52% of the seedlings in many visible shades of color. Peonin was identified in progenies where only one parent showed peonin and could be traced in an unbroken line through several generations (Figure 1). There were some unusual features of segregation in that only one family (47) had all peonin progeny but with many levels of color. No peonin was observed in families 15 and 16 where some would be expected. Attempts to fit segregation to tetraploid ratios were not usually successful.

Table 2. Segregation in progenies from 36 rose families grouped into five pigment classes (progenies of over 10 seedlings).

Cross Number of plant in five pigment classes
total with
with either
peonin or
with both
peonin and
5 706266 37 37 14 9 23 0
6 706662 43 43 10 7 17 0
7 72A484 64 61 33 0 33 0
8 718472 15 15 11 4 14 1
II 714572 20 20 5 11 12 4
13 714072 14 14 2 3 5 0
15 73B067 14 13 0 0 0 0
16 7299A7 28 27 0 27 27 0
17 726584 48 47 39 2 41 0
18 728967 33 33 26 0 26 0
19 72A489 39 39 20 3 23 0
20 73E0D7 27 27 10 2 12 0
21 7365A5 18 18 13 4 16 1
22 73A5A7 23 23 5 22 23 4
23 7366A7 33 33 21 13 28 6
26 7365A7 41 41 33 10 39 4
27 73C709 39 39 22 1 23 0
28 7362C7 40 40 22 11 32 1
29 736272 56 56 20 14 33 1
30 736567 13 13 8 1 8 1
32 76E0J3 14 14 3 1 3 1
33 76E0J4 28 28 23 7 27 3
34 76F4D7 14 13 1 8 9 0
35 76F4J4 14 14 9 6 13 2
36 76F5D7 14 14 2 13 14 1
37 76F5J4 28 28 18 18 28 8
38 76G1D7 28 28 6 20 23 3
39 76G1F5 14 13 4 7 11 0
40 76G1J3 28 21 2 10 12 0
41 76J0D7 28 28 18 12 27 3
42 76J1D7 28 28 13 25 28 10
43 76J3E0 28 19 6 4 10 0
44 78JOJ4 62 62 59 13 59 13
45 78J1D7 71 71 29 41 60 10
46 80Q1J4 39 39 34 22 39 17
47 80H0R0 45 43 45 0 45 0
Total   1128 1102 586 351 843 94

Pelargonin also appeared in 351 or 31% of these seedlings and at many levels of pigmentation. The progeny of original crosses with pelargonin parents, grown before 1970, which gave parents 62, 65 and 66 were not analyzed. The selections retained showed no pelargonin and none of the typical scarlet color indicative of pelargonin was seen among those discarded. Pelargonin had the recessive characteristics of missing generations (Figure 2) and of appearing from sources which did not exhibit this factor. Pelargonin segregation differed from that expected from a recessive character in families 16, 22, 36, 42 and 45 where both parents tested positive for pelargonin but it was not found in all progeny. It also appeared in two families with no known pelargonin parentage and in 8 other families where it was known in only one parent. Pelargonin segregation fit tetraploid ratios in some families. Pelargonin occurred with peonin and cyanin in 94 seedlings from 20 families (Table 2).

Fig. 1. Pedigree showing peonin transmitted through six generations of roses. Progeny on left, parents on right.

1 Parents of 78J0J401 are siblings.
2 Numbers beneath each identification indicate anthocyanin ratings in order cyanin, peonin, pelargonin.

Correlation of data for pigment ratings in progeny showed that there were statistically significant relationships among the pigment ratings, however, the correlation coefficients were low (Table 3). Cyanin ratings showed a significant but small positive correlation with peonin, pelargonin and peonin plus pelargonin ratings. A negative correlation was found between peonin and pelargonin ratings.

Heritability estimates derived from regression between parents and progeny showed that cyanin, peonin, pelargonin and total anthocyanin were each highly heritable (Table 4). These estimates for seed or pollen parent should not exceed 1.0 as they did in two instances possibly due to crossing taken place between relatives increasing the sample heritability above the population heritability or to seasonal effects on anthocyanin expression. With this limitation in mind, there seemed little difference between the effects of seed or pollen parent; both were high for each pigment.

Heritable interactions from regressions of one pigment in the parent on another pigment in the progeny were all much lower than when the pigments were the same (Table 5). Again results from the seed or pollen parents did not differ greatly from each other. Peonin parents by pelargonin progeny and the reciprocal both showed highly significant negative regressions. Regressions of cyanin parents by pelargonin progeny and the reciprocal were not significant, indicating little interaction. Cyanin parents by peonin progeny and the reciprocal gave a confusing mixture of positive and negative regressions.

Fig. 2. Pedigree showing pelargonin transmitted through four generations of roses. Progeny on left, parents on right.

1 Numbers beneath each identification indicate anthocyanin ratings in order cyanin, peonin, pelargonin.
2 Known sources of pelargonin.


Cyanin was by far the most common pigment found, accounting for 66% of all pigment ratings and being found in more than 99% of all seedlings. This high occurrence of cyanin agrees with ARISUMI (1963) and YOKOI (1974) who found it in all 670 roses analyzed. However, they did not analyze some common white cultivars which might not have contained cyanin. The high heritability estimate of 0.68 (Table 4) indicates that cyanin can be readily transferred in crosses and possibly is controlled by only a few genes. Its stability plus the wide and continuous ranges of pigmentation found here certainly suggests the action of more than one series of genes controlling cyanin production. It is possible that modifying genes controlling the total amount of anthocyanins produced also may be involved. Certain crosses have suggested a gene decreasing cyanin content (unpublished data).

Table 3. Correlation coefficients calculated from anthocyanin ratings from 1128 seedling roses.

Cyanin/peonin 0.21 **
Cyanin/pelargonin 0.07 *
Cyanin/peonin + pelargonin 0.24 **
Peonin/pelargonin -0.22 **
* significant at P = 0.05. ** significant at P = 0.01.

 Table 4. Heritability estimates for three anthocyanins in Rosa.

Seed parent/progeny 0.41 0.81 0.03 **
Pollen parent/progeny 0.57 1.14 0.04 **
Mean of parents/progeny 0.68 0.68 0.04 **
Seed parent/progeny 0.59 1.17 0.03 **
Pollen parent/progeny 0.44 0.88 0.02 **
Mean of parents/progeny 0.69 0.69 0.03 **
Seed parent/progeny 0.35 0.69 0.02 **
Pollen parent/progeny 0.44 0.89 0.03 **
Mean of parents/progeny 0.58 0.58 0.03 **
Total pigment        
Seed parent/progeny 0.33 0.66 0.03 **
Pollen parent/progeny 0.52 1.04 0.03 **
Mean of parents/progeny 0.61 0.61 0.03 **
** P = 0.01.

Table 5. Heritable interaction among anthocyanins in Rosa.

Cyanin parents/peonin progeny 0.12 0.12 0.04 **
Peonin parents/cyanin progeny -0.09 -0.09 0.04 **
Cyanin parents/pelargonin progeny 0.01 0.01 0.03 NS
Pelargonin parents/cyanin progeny 0.02 0.02 0.05 NS
Peonin parents/pelargonin progeny -0.14 -0.14 0.03 **
Pelargonin parents/peonin progeny -0.35 -0.35 0.04 **
Cyanin parents/peonin + pelargonin progeny 0.13 0.13 0.44 **
Peonin + pelargonin parents/cyanin progeny .07 .07 0.04 NS
**P = 0.01; NS = non significant at P = 0.05.

White flowers in tetraploid roses may be one extreme of the red series. The white of Floribunda and Hybrid Tea, the near white of R. arkansana and the white of R. rugosa alba (WARE) REHDER differ chemically (MARSHALL, 1975). White in diploid R. rugosa alba was fully recessive when crossed with red roses and colors in the progenies ranged around that of the colored parent. This could possibly resemble the nivea gene in Antirrhinum (HARRISON & STICKLAND, 1978). 'White Bouquet' x 'Assiniboine' gave a range of pink shades centered between the parents similar to families 15, 32 and 40 (Table 2). The near white form of R. arkansana, when crossed with the cardinal red 'Adelaide Hoodless', also produced dilute shades in its progeny in a similar action, or possibly due to genes restricting anthocyanin production. Due to tetraploid segregation and the large number of progeny that would be required to obtain most recombinants, the two extremes of the series would seldom occur, and thus both white and crimson would appear recessive to pink ((LAMMERTS, 1945). Pink or rose would be expected to be the most common color, a state that is obvious in wild and cultivated roses.

Peonin and pelargonin cannot be discussed separately due to apparent interactions. Analysis of segregation for these two factors (Table 2) was of limited value as the progenies were mainly too small for confident comparison with complex tetraploid ratios. Heritability estimates of 0.69 for peonin and 0.58 for pelargonin (Table 4) show that both are highly heritable and are possibly conditioned by few genes.

Peonin was introduced to these breeding lines from a deep pink form of R. arkansana and from 'J. W. Fargo', a medium pink double form of the same species. It can be followed without a break through six generations except for two parents which were not analyzed (Figure 1). This pedigree indicates dominant characteristics for peonin. As it also occurs in many shades, it indicates possible quantitative inheritance or modifying genes.

Segregation of peonin, however, (Table 2) agrees only in part with dominant gene action. Many progenies contained mostly peonin and it was found in all except two progenies (families 15 and 16) that had one or both peonin parents. It is possible that family 15, 'Mount Shasta' x ('Golden Scepter' x 'Assiniboine') may indicate peonin suppression where total anthocyanins are limited. Several families gave less than a 1:1 ratio, the least expected from a dominant tetraploid, but pelargonin or a pale pink color were often seen in these. Only family 47 showed peonin in all seedlings while several such families would be expected.

Fig. 3. Proposed pathways for anthocyanin synthesis in roses.

Pelargonin was introduced into the breeding program from 'Independence' and 'Fire King'. A four generation pelargonin pedigree is shown in Figure 2. Table 1 shows parents 62, 65 and 66 derived from 'Independence' and 'Fire King'. No pelargonin was observed in these F1 progenies but it reappeared in reciprocal progenies 5 and 6 (Table 2). Families 5, 6, 8, 28, 29 and 33 were similar in parentage and segregation and gave 167 without, to 52 seedlings with pelargonin, an acceptable fit for a tetraploid segregation range from simplex selfed. A further group of ten families (11, 23, 34, 35, 37, 38, 39, 40, 41, 46) gave 119 non to 127 with pelargonin, which may represent 1:1 segregation from simplex x homozygous recessive. These ratios were in agreement with recessive pelargonin as reported by DE VRIES et al. (1980). There were, however, a number of progeny families which did not fit any expected or theoretical patterns.

Firstly, there was a deficiency of pelargonin individuals in 6 families (16, 22, 34, 36, 42, 45) where both parents expressed pelargonin. If pelargonin were a simple recessive all seedlings should express this character when both parents express pelargonin. However, no family produced all pelargonin offspring.

Secondly, segregation for pelargonin occurred where it was not expected: two families (8, 19) had no known pelargonin ancestors, 3 families had only one parent expected to carry pelargonin (17, 40, 43) and 5 families had one parent which expressed pelargonin (11, 13, 20, 38, 39) but no known pelargonin in the other parent. Ratios in these families, however, were erratic. Apparently another factor or factors are involved in the production or expression of pelargonin.

Evidence linking many of the problems with peonin and pelargonin may be found in the negative regressions between peonin parents and pelargonin progeny and reciprocal combinations (Table 5). It is possible that these two pigments are in a competitive position, possibly with both being derived from a common precursor, kaempferol. DE VRIES et al. (1974) showed a positive relationship between pelargonin and kaempferol. This would allow methylation to occur at a late stage in the biosynthetic pathway as proposed by HARRISON (1978) (FORKMAN, 1980). A system is proposed in Figure 3 to explain most of the synthetic pathway and controls for anthocyanin production in roses.

The complex ratios found in peonin segregation could possibly be caused by methylation occurring somewhere in the pathway of kaempferol to anthocyanins. This could possibly deplete required substrate for the production of pelargonin when pelargonin x peonin segregated for pelargonin in the F1. It is possible that pelargonin was expressed due to quantitative changes occurring in the pathway from kaempferol which might limit transfer to peonin. This also would make pelargonin segregation more complex. The small positive correlation among the three anthocyanins and the suppression of both peonin and pelargonin where cyanin is low are all possibly due to the quantitative changes to the anthocyanin A ring. The deficiency of all peonin families was not explained but the absence of homozygous peonin parents could possibly be due to the action of a sub-lethal gene.

Peonin levels in hexaploid R. acicularis LINDL., tetraploid R. arkansana and diploid R. x dulcissima (LUNELL) W. H. LEWIS collected from native stands varied considerably (MARSHALL, 1975) and in the latter two species 30% and 78% contained cyanin only. Three seedlings, selected for intense peonin, grew poorly and were sterile. 'Quadroon' (intro. Wright) which has a high peonin level is reported to grow poorly (MEIKLE, 1980). The absence of homozygous peonin breeding lines, the heterozygous condition in wild populations and the presence of defective peonin selections suggest a sub-lethal peonin gene.

Three pigments as closely related as cyanin, peonin and pelargonin which differ only at the 3' position on the B ring of the anthocyanin molecule would be expected to interact, particularly when only two anthocyanin precursors, quercetin and kaempferol, have been found (DE VRIES et al., 1980; HARBORNE, 1961). The interactions seen (Tables 3, 5) were frequently significant but not large and did not preclude co-existence of pigments (Table 2).

This result agreed with the literature where neither peonin nor pelargonin was found without cyanin (ARISUMI, 1963; YOKOI, 1974; DE VRIES et al., 1974). Five exceptions, two previously noted (MARSHALL, 1975), and families 39 and 47 (Table 2), all involved only peonin but a very small amount of cyanin might have been undetected because of the screening method used. The weak qualitative interactions show that genes for these pigments are not alleles but probably interact by competition for substrate.

The commonly assumed origin of pelargonin by mutation in early scarlet cultivars such as 'Gloria Mundi', 'Independence', 'Little Darling' and 'Fire King' might possibly be explained more reasonably as a recombination of factors from a rose without anthocyanin or without peonin with one containing peonin. 'Austrian Copper' is a probable source of the genetic combination required to express pelargonin as it contains peonin (ARISUMI, 1963), is closely related to yellow roses and is an ancestor of the latter three cultivars (MEIKLE, 1980).

From this study is was concluded that:

  1. Color intensity and quality of parents should be as near as possible to the desired color of the progeny because of the quantitative factors influencing pigment intensity. It is difficult to recover pale pink from deep red parents or the reverse.
  2. Pelargonin and peonin may each exceed a moderate level of cyanin but it may be difficult to find either without cyanin as they tended to be suppressed when cyanin levels were low.
  3. Peonin and pelargonin can occur together but the small inverse correlation between them may be preclude high intensities of both. Together they have given shades similar to 'Morden Amorette' R.H.S. 52B to 57B (carmine to rose Bengal).
  4. Pelargonin may be derived from peonin parents by crossing with pelargonin and possibly yellow or others not showing peonin.
  5. Peonin has given cardinal red R.H.S. 53A, 53B similar to 'Europeana', 'Lilli Marlene' and 'Adelaide Hoodless' as well as the better known and undesirable purplish red of R. rugosa.
  6. The most intense peonin shades may occur in seedlings with poor vigor and low fertility.


CybeRose notes: The simple inheritance of Peonidin-3-glucoside involves a single gene encoding the AMT (anthocyanin methyltransferase) enzyme. As the following diagram shows, peonidin differs from cyanidin by a single methyl group.

This next \diagram is copied from Holton and Cornish: Genetics and Biochemistry of Anthocyanin Biosynthesis, The Plant Cell, 7: 1071-1083 (July 1995)

This diagram represents the biochemistry of Petunia, so it may be used with caution. The important thing to note is that there is no apparent biochemical connection between pelartonidin and peonidin. If 3GT attaches the glucose molecule before AMT attaches the methyl group, I don't see how there could be a pelargonidin-peonidin correlation. However, these two enxymes work in the reverse order, it might be possible that AMT has more affinity for peonidin than for pelargonidin, cyanidin being less favored than the other two. I'm not betting on either.