Botanical Gazette 94: 551-566 (Mar. 1933)

1Paper from the Department of Botany of the University of Michigan, no. 376, representing
work done partially under a National Research Fellowship in the Biological Sciences.



The chromosomes of dicotyledons are small, and a true picture of their pairing behavior can be obtained only under optimum conditions of fixation, staining, and microscopy. Evidence in support of the chiasma theory of chromosome pairing (DARLINGTON 6) has been accumulated from widely separated genera, so that it is now indisputable that chromosome pairing at the first metaphase of meiosis in animals and plants is conditional upon the formation of chiasmata among the chromatids of homologous parts of chromosomes in prophase of meiosis.

Multiple associations of chromosomes may arise from: (1) the presence of more than two homologous whole chromosomes in a nucleus, as in polyploids and trisomic organisms; (2) the presence of more than two homologous chromosomal segments owing to reduplication in diploids and polyploids; (3) reciprocal translocation (STURTEVANT and DOBZHANSKY 18), which is also called segmental interchange between non-homologous chromosomes (ERLANSON 12). References to each type of multiple association have been compiled by DARLINGTON (8). Multivalent groups resulting from each of these causes have been found in Rosa.

Reciprocal translocation and reduplication

Reciprocal translocation appears to occur rather frequently in diploid roses, and is an important cause of sterility because of the consequent production of non-disjunctional gametes (BEADLE 1, GAIRDNER and DARLINGTON 13). The first rose in which it was observed was the garden hybrid polyantha rose Orleans (species complex Rosa multiflora Thunb.), a plant of which had a low percentage of trivalent, quadrivalent, and sexivalent chromosome associations at diakinesis and at first metaphase in the pollen mother cells (ERLANSON 12). A plant of R. blanda Ait. (diploid), with similar configurations, was found to be 3x (triploid) for a short chromosomal segment. Quadrivalent groups have since been observed in other individuals of R. blanda, as well as in plants belonging to the related diploid species R. woodsii Lindl. and R. pisocarpa A. Gray (individuals of wild origin). In no rose examined has there been a regular formation of quadrivalent groups in all pollen mother cells at meiosis. This shows, on the chiasma theory of chromosome pairing, that the translocations and reduplications concerned are short segments; they are consequently only rarely able to form chiasmata at diplotene. An analogous situation has recently been reported in Zea by BURNHAM (2) and BEADLE (1), and is thought by MÛNTZING (16) to be present in Galeopsis.

The diploid roses that show multivalent configurations may have small segments of some chromosomes reduplicated, but my earlier report of whole chromosome reduplication in R. blanda was in error (ERLANSON 10). The mistaken interpretation was due to multivalent configurations in conjunction with the presence of univalents. In Rosa the unpaired chromosomes may divide at both anaphases of meiosis and give rise to 8-chromosome gametes (n=7), which, however, seem to be non-functional.

True aneuploidy is very rare in Rosa. The two extra chromosomes of the 16-chromosome R. pyrifera from Utah (10) are smaller than the others. Their pairing behavior, which is discussed in detail later, shows that they are reduplicated fragments.

2 Serial accession number in the Botanical
Garden of the University of Michigan.

An aberrant sterile plant (6424/1)2, which was raised from seed of the tetraploid R. arkansana Porter, obtained from a nursery, was found to have three extra chromosomes which frequently formed trivalent groups at diakinesis (fig. 1). The first anaphase was irregular, and lagging chromosomes sometimes caused a restitution nucleus to be formed at first telophase (fig. 2).

Chromosomes and fragments in Rosa pyrifera

A plant of Rosa pyrifera Rydberg (6610B) from Utah was found to have 16 pairs of chromosomes (10). One pair was distinctly smaller than the other seven. This plant is not entirely hardy in Michigan and produces only a few flower buds. Some young anthers fixed in acetic alcohol were cut at 20 µ and stained with Newton's gentian violet method. A sufficient number of whole nuclei at diakinesis were obtained for a detailed study of chromosome pairing at this stage. Although very few nuclei have been observed at earlier or later stages of meiosis, the conditions at diakinesis are similar to those in other roses at the same stage, and are therefore believed to give a reliable idea of the chromosome organization in this plant.

FIGS. 1-8.*—Fig. 1, diakinesis in atypical offspring of R. arkansana with two trivalents, 12 bivalents, and one univalent. Fig. 2, restitution nucleus at first telophase in atypical R. arkansana with 31 chromosomes. Figs. 3-8, diakinesis nuclei in R. pyrifera (6610B), 2n=14+f+f´. Figs. 3, 4, f and f´ paired with different pairs of chromosomes. Fig. 5, 1III+6II+1I; f and f´ in the trivalent group. Fig. 6, f and f´ paired with each other, also 6II+2I. Fig. 7, 7II; f and f´ unpaired. Fig. 8, f and f´ paired with different members of the same pair, also 6II.
* Figures drawn with a camera lucida at a magnification of approximately 6000 and reduced it, two-thirds in reproduction; except figs. 9-12 and fig. 22, which were drawn at approximately 4500X and reduced to three-fourths in reproduction.


No. of divisions
No. of potential pairs of chromosomes Total no.
of Xta
No. of Xta
No. of Xta
per bivalent
Highest no. of
Xta in any pair
25 175 221 127 1.27 0.57 3

*Proportion of total chiasmata terminal at any stage (13).

The two small extra chromosomes in the rose 6610B pair not only with each other, but also with two different pairs among the normal set of fourteen. Only homologous parts of chromosomes pair. To interpret chromosome behavior we must assume homologies that are in accordance both with the parts capable of pairing and with the frequencies of pairing. The members of the small extra pair are unequal. They must evidently be considered as fragments rather than as whole chromosomes, since the larger of the two is about half as long as the smallest of the normal chromosomes. The larger is here designated as fragment f´; the other, slightly smaller, is called f.

Chiasma formation was analyzed in 25 whole nuclei at diakinesis in the pollen mother cells (figs. 3-8). As shown in table I, the mean number of chiasmata per bivalent is only 1.27, calculated for the seven normal pairs. Table II gives the frequencies of the same chiasmata (from 1 to 3) in bivalents (in uncut nuclei). As in other roses, there is a failure of chiasma formation in one pair of chromosomes in a small number of nuclei. In table III the individual 6610B of R. pyrifera is compared in its chromosome behavior at late diakinesis with the other two roses (12) in which chiasma formation has been studied. The R. pyrifera plant has a slightly lower chiasma frequency than the other two, which is perhaps due to terminalization having proceeded further. The percentage of bivalents lacking chiasmata is similar in all three, yet the R. pyrifera has only one chiasma in 70.3 per cent of the bivalents. This is a higher proportion than was shown even at first metaphase by the other two roses. The terminalization coefficient is higher in this rose. The fragments fail to pair more frequently than the major chromosomes, as would be predicted on the chiasma theory. In the nuclei studied they have been paired by terminal chiasmata only. As in Fritillaria imperialis (DARLINGTON 5) all the chiasmata are near the ends of the fragments, and terminalization is complete at diakinesis in these though not in the longer chromosomes.


  No. of chiasmata Total potential bivalents Total
Mean no of
Xta per bivalent
0 1 2 3
No. of bivalents 4 123 46 2 175 221  
Percentage of bivalents 2.3 70.3 26.3 1.1      
Total no. of chiasmata observed 0 123 92 6 175 221 1.27


Name of Rose Percentage of bivalents failing
to pair
Mean no. of
Xta per bivalent
Terminalization coefficient Percentage of bivalents with only one Xma Highest no. of Xta observed in any bivalent
Rose Orleans 2.6 1.65 0.50 33.12 3
Rosa blanda, 7N29 1.3 1.47 0.47 52.6 3
Rosa pyrifera, B. 2.3 1.27 0.57 70.3 3

To analyze the pairing behavior of fragments, each of them has to be treated as a potential half-bivalent; where they form a chiasma with each other, or with a chromosome, the individual fragment is credited with a half chiasma (5, 12). Table IV shows that the shorter fragment, f, has a higher pairing frequency than its longer mate, f´. The mean number of half chiasmata for f is 1.1, which is a little less than half the mean for each chromosome of the normal set (2.54); while the mean chiasma frequency of f´ is almost a third less than that of the other chromosomes. In 25 nuclei, fragment f´ was free three times, while f was free only once. The reason for this is shown in an analysis of the types of pairing undergone by the fragments (table V). The small fragment, f, pairs with the smallest pair of chromosomes (figs. 3, 4), which I have designated as AA, and f´ pairs with a medium sized pair, BB. The pairing arrangements 1-5, table V, are very similar; the fragments are paired with different whole pairs. Since the homologies of chromosomes are specific to their parts, the constitution of the parts involved in these pairing arrangements can be represented thus: AA=ab/ab, f=b, BB=cd/cd, f´=c. Arrangements 6, 7, 8, and 9, table V, show that f and f´ also pair with each other and with the same pair, AA (figs. 5, 6, 8); therefore one of them consists of two segments. We may assume that this is the larger f´, because it has the lower pairing frequency. The constitutions of the chromosomes and fragments involved are therefore ab/ab, b, cd/cd, cb.


No. of 1/2 chiasmata 0 1 2 Total no. of
1/2 chasmata
Mean no. of 1/2
Xta per fragment
No. of 1/2
Xta terminal
Fragment f 1 20 4 28 1.1 28 1
Fragment f´ 3 22 0 22 0.88 22 I


Pairing arrangement No. of nuclei
1. AA+f, BB+f´ 12
2. AA+f, B+f´ 1
3. A+f+A, BB+f´ 1
4. A+f+A, f´ free 1
5. AA+f, f´ free 1
6. f+f´ 5
7. AA+f+f´ 1
8. A+f+f´ 1
9. f+AA+f´ 1
10. f and f´ free 1
Total nuclei 25

In prophase, f´ would have difficulty in using both ends at once, and this discontinuity in pairing homology reduces its chiasma frequency (DOBZHANSKY 9).

The pairing behavior of these fragments shows that the plant is quadrivalent for a short segment (h above), and trivalent for another small segment (c above). DARLINGTON (NEWTON and DARLINGTON 17) has pointed out that short fragments are better testers of homologies because they are not carried away by long continuities of homology. Because f´ consists of two segments, cb, both fragmentation and translocation must have taken place. The configuration ab/ab+b+bc+cd/cd has not been observed; it would be expected to occur infrequently if the translocation c be rather short. Changes of homology arrest terminalization (12), and it is possible that when they are present the process of terminalization causes a strain which might cause fragmentation; but in Rosa and Oenothera (DARLINGTON 7) such arrest does not lead to breakage of chromatids. There is no evidence to indicate that translocations tend to occur at the same locus in particular chromosomes (2).

This is the only instance of fragments that has been found in Rosa. DARLINGTON (3, 4) predicted that effective fragmentation would be rare in the great plant genera Rubus, Ribes, Rosa, Prunus, Avena, and Triticum. Fragmentation is not absent, but when it occurs it is usually eliminated in the course of sexual reproduction. It may become fixed by apomixis, however, as in Potentilla (MÜNTZING 15). This plant of R. pyrifera has 40 per cent of the pollen grains either dwarfed or shriveled.

Rosa pyrifera belongs to the species complex of R. woodsii, from which it differs chiefly in having pyriform hips and pubescent foliage.

Pairing in higher polyploids

Multiple chromosome association occurs in all the balanced tetraploid, hexaploid, and octoploid roses I have examined. Tetraploids frequently show one or two quadrivalent groups in 50 per cent or more of the nuclei at diakinesis and first metaphase (10, 12). Non-disjunction is frequently associated with rings and chains of chromosomes, and this is probably one of the causes of the relatively high sterility found in tetraploid species of Rosa (11), as in autopolyploids generally (8). Multivalent chromosome groups are less frequent in polyploids higher than tetraploid. This may be due to a difficulty in obtaining maximum association for all pairing possibilities among sexivalent homologous segments in prophase, on a basis of random chiasma formation in these short chromosomes (13). In hexaploids there are 21 pairs in a majority of nuclei at diakinesis (fig. 9); occasionally 20II+2I appear (fig. 10). More than one chain of five or six has not been observed in one nucleus (figs. 11, 12), although as many as four quadrivalent groups have been found (ERLANSON 10, fig. 63).

FIGS. 9-12.— Diakinesis in F1 of R. engelmanii x nutkana, 2n=42; nuclei with 21II, 20II+2I, 1VI+18II, and 1V+18II+1I respectively.

Figures 9-12 are diakinesis configurations in an F1 individual from R. engelmanni (6x) X R. nutkana (6x). The male parent was obtained from the Priest River region of northern Idaho, and the female parent from the Medicine Bow Mountains of Wyoming. The offspring are hexaploid, have only 10 per cent of the pollen empty, set good fruit, and would be classified as R. engelmanni by a taxonomist.

The multivalent configurations considered in this section are presumably due to the presence of more than two homologs in the nuclei of the polyploids. If reciprocal translocations are present and are of sufficient dimensions to form chiasmata, then associations of more than four chromosomes appear in tetraploids, as in R. relicta (12). Associations of more than six chromosomes have not been found in hexaploids.

Chromosome pairing in unbalanced polyploids

TRIPLOIDS.— The first wild triploid roses that I reported (10) were spontaneous hybrids between diploid and tetraploid species. They usually had seven paired and seven univalent chromosomes at diakinesis. A triploid has since been found in a culture of diploid R. blanda, from Michigan. This plant probably originated in the fusion of a diploid with a normal haploid gamete of the same species. It had a high percentage of trivalent groups at diakinesis and first metaphase of meiosis; different nuclei showed from three to seven trivalent groups. This variation is characteristic of non-hybrid triploids, and is expected on the chiasma theory of pairing (8). Figure 13 shows a first. metaphase with five trivalents and two univalents, Figure 14 shows the end of first anaphase with six chromosomes at one pole, eleven at the other; one pair lagging and two univalents dividing at the equator.

In a culture of diploid R. macounii Greene (species complex R. woodsii) from Reno, Nevada (no. 12205/15), which was grown from seed of a single individual and cultivated at the California Institute of Technology, one plant among 21 was found to have 90 per cent of the pollen shriveled. Cytological examination showed that it was a triploid with a high proportion of trivalents at diakinesis. Among ten whole nuclei, three had seven trivalents (fig. 15), five had 6III+1II+1I, and two had 5III+2II+2I. This plant also probably arose from an unreduced gamete.

A seedling of R. pisocarpa Gray (12259) with sterile pollen and ovules was also triploid with 21 somatic chromosomes, and had a high percentage of trivalents at diakinesis. The parent plant grew near Jackson, Oregon, and was laden with fruit; it was presumably a normal diploid.


7II+14I 1 3 1
1III+7II+11I I 3 1
1III+6II+13I   1  
2III+5II+12I 1    
2III+6II+10I     1
8II+12I   6  
9II+10I   1  

A TETRAPLOID IN SECTION CANINAE.— Chromosome pairing was analyzed in a few whole nuclei of the unbalanced tetraploid R. villosa L. This rose belongs to the section Caninae, all the members of which have been reported by TÄCKHOLM, HURST, HARRISON, and BLACKBURN to be unique in never having more than seven bivalents at diakinesis and first metaphase. On the chiasma theory of chromosome pairing, one would expect occasional variations in this arrangement owing to the presence of more than two homologous chromosomes. Table VI shows the types of chromosome pairing actually found in 20 nuclei at early and late diakinesis and at first metaphase. The chromosomes are small and it is difficult to analyze pairing conditions, especially at early diakinesis and metaphase. These few figures, however, show clearly that 7II+14I appear only in a minority of the nuclei. In these 20 nuclei, five had 7II+14I (fig. 17), six had 8II+12I (fig. 16), and five had 1III+7II+11I (fig. 19). One cell had 9II+10I. Trivalents were found in eight nuclei.

FIGS. 13-21.—Fig. 13, R. blanda (7N44), 2n=21, first metaphase with 5III+2II+2I. Fig. 14, R. blanda (7 N44), 2n=21, firm anaphase, two univalents dividing at equator, one pair lagging. Fig, 15, R. macounii, 2n=21, diakinesis with 7III. Figs. 16, 18, nuclei of R. villosa, 2n=28, in early diakinesis with 8III+12I, 7II+14I, and 2III+5II+12I respectively. Fig. 19, R. villosa, late diakinesis with 1III+7II+11I. Fig. 20, metaphase complement in R. villosa, with 2III+6II+10I. Fig. 21, metaphase complement in R. rubrifolia, 2n=28, with 1III+8II+9I.

In the early diakinesis nuclei shown in figures 16, 17, and 18 there are clearly 14 chromosomes with a high chiasma frequency, and 14 with a very low chiasma frequency which may be associated with each other or with the regular pairs by terminal chiasmata only. The contrast between the chiasma frequencies of these two sets of chromosomes is shown graphically in figure 23. fit three nuclei at early diakinesis there was a total of 67 chiasmata among the seven normal pairs, and only 22 of them were terminal (table VII). The mean chiasma frequency of 3.2 per bivalent is the highest yet found at this stage of prophase in a rose. In eight cells in the same loculus at late diakinesis, the seven regularly pairing bivalents were again distinguished (fig. 19), and the mean number of chiasmata per bivalent was found to be 1.98. In table VII it is shown that the process of terminalization of chiasmata in R. villosa goes on as in other roses (12). As meiosis proceeds, the mean number of chiasmata per bivalent decreases but the proportion of total chiasmata that are terminal increases. At first metaphase terminalization is practically complete (fig. 20), and the terminalization coefficient is 0.89 (table VII). This may be due to a genetic property or to the absence of structural hybridity, as between the normally pairing chromosomes among the modern roses of the Caninae.


STAGE No. of
No. of
Total no.
of Xta
Mean no. of
Xta per bivalent
No. of
terminal Xta
Highest no. of Xta
observed in any pair
Early diakinesis 3 21 67 3.2 22 0.33 4
    slide 1
6 42 110 2.6 46 0.42 4
    slide 2 8 56 111 1.98 58 0.52 3
Metaphase 3 21 37 1.76 33 089 2

A metaphase plate with 1III+8II+9I in R. rubrifolia Vill., another member of the Caninae with 2n=28, is shown in figure 22.

Figures 16, 17, and 18 show that the univalent chromosomes at diakinesis are noticeably less condensed than their paired fellows. It has always been difficult to understand why hybridity should prevent the pairing of one-half of the complement while not affecting the other. On the precocity theory of meiosis, and the theory of the origin of sex chromosomes which follows from it (DARLINGTON 6, 8), this is intelligible; for it may be supposed that the 14 unpaired chromosomes have a genetic property of differential precocity producing the same effect as that in the sex chromosomes in animals. Thus the possibilities of genetically caused abnormal zygotene pairing are as follows:

1. Incomplete precocity of the prophase (for example, Matthiola). The prophase is intermediate in precocity between a normal mitotic and normal meiotic condition; this is expressed in long chromosomes at first metaphase, that is, reduced condensation.

2. Differential precocity, the two known types of which are: (a) In sex chromosomes there is a differential precocity of the chromosomes themselves. The unpaired sex chromosome is itself precocious, as is the prophase, and so restores normal mitotic conditions for itself. This is expressed in precocious condensation and precocious splitting as in ordinary mitosis (DARLINGTON 6). (b) In the unpaired chromosomes of the Caninae the prophase contraction begins later than in their paired fellows. They presumably have divided before they condense and therefore cannot pair; a condition that was suspected by HUSKINS (14) to be a cause of asynapsis in dwarf oats and in sorghum.

FIG. 22.— Somatic metaphase from root tip of Rosa sp., 2n=35, section Caninae


FIG. 23.— Percentage frequency polygons of numbers of chiasmata in bivalents (or their equivalents in unpaired chromosomes and multivalents) in the two sets of chromosomes of R. villosa; p, paired set, un, unpaired set. Data from 14 nuclei at late diakinesis.

Somatic chromosomes of Caninae

A somatic metaphase plate from a root tip of a member of the Caninae is shown in figure 22. This plant, still a seedling, was raised from seed collected in Persia by C. D. DARLINGTON. It is a pentaploid with 35 somatic chromosomes. These show both primary and secondary constrictions, as in diploid roses, and resemble the chromosomes of Oenothera (DARLINGTON 7, fig. 22).


1. Multiple pairing of chromosomes clue to reduplication and reciprocal translocation of segments has been found in diploid roses belonging to the species complexes Rosa multiflora, R. blanda, R. woodsii, and R. pisocarpa. Gametes with eight chromosomes are sometimes produced, but no diploid has been found possessing a reduplicated whole chromosome.

2. A plant of R. pyrifera Rydb. (group of R. woodsii), previously reported to have 2n=16, has a small unequal pair which are really reduplicated homologous fragments of approximately half a chromosome. They pair with each other and with two different pairs.

3. Both fragments fail to pair in a proportion of cases. The larger, f´, consists of two segments, homologous with two different pairs. This reduces its pairing frequency (in agreement with the chiasma theory of chromosome pairing) to less than that of the shorter fragment, f. The plant is 4x for one short segment and 3x for another. This is the only example of fragments yet found in Rosa.

4. Multivalent associations of chromosomes are frequent among tetraploid roses, but are relatively rare in hexaploids and octoploids.

5. Triploids with a high proportion of trivalents at diakinesis and first metaphase have been found in diploid cultures of R. blanda, R. macounii (group of R. woodsii), and R. pisocarpa. These probably originated from unreduced gametes and arose without hybridization.

6. The unbalanced tetraploid R. villosa (section Caninae) shows a high chiasma frequency among 14 chromosomes and a very low frequency among the other 14, which usually occur as univalents. In 75 per cent of the nuclei examined, from one to four of the latter were paired terminally with each other or with one of the regular pairs. Trivalents were found in eight nuclei out of 20. Failure of pairing of the chromosomes is probably due to differential precocity in prophase development, not to hybridization.

7. The chiasma frequency of the fourteen regularly pairing chromosomes in R. villosa is 3.2 chiasmata per bivalent at early diakinesis, and terminalization is almost complete at first metaphase.

8. The types of structural change and the different kinds of polyploids found in Rosa are described. The exceptional types of pairing found are in conformity with the chiasma theory of pairing and are in three instances such as would be predicted on that theory.



  1. BEADLE, G. W., The relation of crossing-over to chromosome association in Zea-Euchlaena hybrids. Genetics 17: 481-501. 1932.
  2. BURNHAM, C. R., An Interchange in maize giving low sterility and chain configurations. Proc. Nat, Acad. Sci. 18: 434-440. 1932.
  3. DARLINGTON, C. D., A comparative study of the chromosome complement in Ribes. Genetica 11: 267-272. 1929.
  4. ————, Chromosome behaviour and structural hybridity in the Tradescantiae. Jour. Gen. 21: 207-286. 1929.
  5. DARLINGTON, C. D., Chromosome studies in Fritillaria. III. Chiasma formation and chromosome pairing in Fritillaria imperialis. Cytologia 2: 37-55. 1930.
  6. ————, Meiosis. Biol. Rev. 6: 221-264. 1931.
  7. ————, The cytological theory of inheritance in Oenothera. Jour. Gen. 24: 405-474. 1931.
  8. ————, Recent advances in cytology. Philadelphia. 1932.
  9. DOBZHANSKY, T., The decrease in crossing-over observed in translocations and its probable explanation. Amer. Nat. 65: 214-232. 1931.
  10. ERLANSON, EILEEN W., Cytological conditions and evidences for hybridity in North American wild roses. BOT. GAZ. 87: 443-506. 1929.
  11. ————, Sterility in wild roses and in some species hybrids. Genetics 56: 75-96. 1931.
  12. ————, Chromosome organization in Rosa. Cytologia 2: 256-282. 1931.
  13. GAIRDNER, A. E., and DARLINGTON, C. D., Ring-formation in diploid and polyploid Campanula persicifolia. Genetica 13: 113-150. 1931.
  14. HUSKINS, C. L., Factors affecting chromosome structure and pairing. Trans. Roy. Soc. Canada. Sec. 5. 1-12. 1932.
  15. MÜNTZING, ARNE, Note on the cytology of some apomictical Potentilla species. Hereditas 15: 166-178. 1931.
  16. ————, Disturbed segregation ratios in Galeopsis caused by intra-specific sterility. Hereditas 16: 73-104. 1932.
  17. NEWTON, W. C. F., and DARLINGTON, C. D., Fritillaria meleagris: chiasma formation and distribution. Jour. Gen. 22: 1-14. 1930.
  18. STURTEVANT, A. H., and DOBZHANSKY, T., Reciprocal translocations in Drosophila and their bearing on Oenothera cytology and genetics. Proc. Nat. Acad. Sci. 16: 533-536. 1930.