1. Rosa Sabini and R. Wilsoni are reciprocal crosses between R. pimpinellifolia and some Tomentosa microgene.

  2. Rosa pimpinellifolia is a balanced tetraploid, both the egg cell and the generative nucleus possessing 14 chromosomes.

  3. The Tomentosa microgenes are unbalanced pentaploids, the microspores, when functional, carrying 7 chromosomes in their nuclei, and the egg cells 28.

  4. Reciprocal crosses between R. pimpinellifolia and tomentosa forms should not therefore agree in chromosome complements. With pimpinellifolia as seed parent the cross should have 14 + 7 (= 21) in its somatic nuclei, and with the same plant as pollen parent the number should be 28 + 14 (= 42).

  5. R. Wilsoni undoubtedly has R. pimpinellifolia as seed parent, yet its chromosome number is 42.

  6. It has, therefore, like Primula Kewensis, doubled its original complement.

  7. In doing so, again like that hybrid, it has attained fertility.

In a former communication, in which the British roses were discussed, we showed that cytologically they were divisible into two groups ; one with members in which microspore formation proceeded normally, and a second in which that process exhibited certain definite abnormalities differing only in degree among the different microgenes involved. Moreover, the peculiarities by which the latter group was characterised exactly paralleled the phenomena observed by Rosenberg when investigating the cytology of the cross between Drosera rotundifolia and Drosera longifolia. From this and similar evidence we concluded that the forms in question could only be regarded as hybrids. This was an extraordinary state of affairs, for the plants concerned, except for two species, comprised the whole of the British rose flora. However, no escape from that conclusion seems possible as no other explanation suffices to account for the observed facts.

That the plants so marked off included forms admitted by every modern worker to be hybrids, of course, excited no surprise, but that, in addition, these were accompanied by no less than three of the five species of such an ultra-conservative worker as Bentham was indeed remarkable, and the wonder was accentuated when one penetrated a little deeper into the matter and took the Jordanian little species, or microgene, into account ; then not less than 98 per cent, of the forms listed by British rhodologists were revealed as hybrids.

Most botanists, from Linnaeus onwards, have marvelled at the extreme range of the variation in Rosa, but no one, until quite recently, ventured to look to hybridity as the responsible agent. However, in 1917, Miss Cole, examining the pollen of the rose forms accessible to her, found it in many cases in a very bad state. In fact, the conditions approximated so closely to those of known hybrids that she suggested a hybrid origin for most of the plants investigated. Clearly, however, their status was not on the same level as that of the recognised hybrids, for in the latter cases the hybridity was patent, and to assign parents on the basis of the collective species not difficult. To mark the difference she applied Jeffery’s term crypthybrid to them. Her solution of the problem was accepted and forthwith expanded on theoretical grounds by Matthews.

Nevertheless, neither of these workers tested the hypothesis cytologically; it remained for Täckholm and ourselves to prove that in this respect the course of events in the phenhybrids was exactly paralleled in the so-called crypthybrids which, therefore, presented themselves as F1 crosses of unknown parentage.

It is not our intention to discuss the latter type; we wish, on this occasion, to confine our remarks to the phenhybrids, and of these only that small section will be considered which have the one parent Rosa pimpinellifolia, either in its typical guise, or in the form of its hispid peduncled variety R. spinosissima, in common. Any of the multitudinous microgenes of the section Caninæ may supply the second parent.

Thus, to be more precise, using the classification adopted in our former paper, the possible British combinations are :—
formula

Although reference to most of these protean hybrids will be found unavoidable, it is to the second group, Pimpinellifoliæ × Tomentosæ, known when massed with the Pimpinellifoliæ × Villosæ fraternity as Rosa involuta, Smith, that attention is specially directed.

For the sake of convenience, this group may be divided into two, (1) including those forms more nearly approaching the Tomentosæ and (2) those presenting marked similarities to the Pimpinellifoliæ. So treated the group would break into plants clustered round Rosa Sabini, Woods, and a like series depending upon the hybrid to which the name R. occidentalis has been applied by Baker. With the latter, in spite of its acquisition of unusually large leaves towards the end of the season, must be included Rosa Wilsoni, Borr, since critical analysis, added to its obvious leanings in its spring conditions, only serves to bring out the more clearly its pimpinellifolia affinities.

Now many representatives of the first-named subgroup have come before us, and all, without exception, grew intermingled with our common hedgerow Caninæ ; on the contrary, Rosa Wilsoni* accompanies Rosa pimpinellifolia. From these observations the obvious inference must be drawn that the plants linked with R. Sabini have R. tomentosa (agg.) as the seed parent and R. pimpinellifolia as pollen parent, whilst the reverse must be the case with the R. occidentalis subgroup; in other words, the two subgroups represent reciprocal crosses. The full significance of this will be developed later.

When our first paper was produced Rosa Sabini alone had been submitted to cytological examination ; but now, owing to the kindness of Miss A. J. Davey, M.Sc., of University College, Bangor, we have been able to examine Rosa Wilsoni obtained from its type habitat near that town.

Our methods of the earlier work need not be given here, but we must mention that the R. Wilsoni material was fixed in acetic-alcohol and stained with Flemming’s triple stain.

a. Rosa Sabini (fig. 1)

As far as could be determined from field observations the plants providing our material had the parentage R. tomentosa var. sylvestris ♀ × R. pimpinellifolia ♂, and although a fairly detailed account of its cytology has already been published some consideration seems necessary here in view of subsequent comparisons. In this rose, as in most others, the style affords the best somatic mitoses for making chromosome counts, and Plate II., fig. 15, chosen from many similar prophases, shows the number to be 42. Granting then, as was discovered by the examination of Rosa arvensis, Huds., and R. rugosa, Thun., that the base number of the genus is 7, R. Sabini is clearly hexapioid.

In the early stages of the meiotic phase, from the reticulum onward through synapsis, little emerges to demand special comment except that, as in the case of Lactuca reported by Gates, binucleate pollen mother cells occur not infrequently. However, soon after that period abnormalities begin to appear and in good diakinetic figures, instead of the anticipated number of bivalents found quite typically at the same point in R. arvensis and R. pimpinellifolia, certain chromosome groups stand out clearly as bivalents, whilst others, just as assuredly, are univalents (Plate II, figs, 11a and 12b). In nuclei suitable for counting the number both of bivalents and univalents appears to be 14.

As the chromosomes pass to the equatorial plate for the heterotype division the procedure is very anomalous. In the first place, the paired univalents take up their stations in the middle of the plate followed, with less regularity, by the univalents which arrange themselves around them. If all succeed in doing so then in polar view 28 chromosomes are to be seen, as in Plate II., fig. 12. In the majority of cases, however, a varying number less than this reach the equatorial plate, the missing chromosomes in every case being univalents. Plate II., fig. 13, shows the exceedingly irregular metaphase typical of the plant.

At the anaphase first the bivalents move apart, then some of the univalents split equationally, the split halves following the former towards the poles. So quickly do the bivalents reach their goal that numbers of the univalents go astray to constitute themselves as micronuclei. Thus, at the interkinesis (Plate II., fig. 14), two major nuclei, with accessory micronuclei, are to be observed. The former, at the homotype division, give rise to primary spindles, whilst the micronuclei form abnormal spindles of a secondary nature.

That many halves of the split univalents are finally included in the major nuclei soon becomes evident, for in the primary homotype plate (Plate II., fig. 16), counts differing widely but still, often enough, approaching the average heterotype number are possible. At this division the phenomena of the heterotype division are repeated, save, of course, that the chromosomes representing the original bivalents also divide equationally. Their subsequent journey to the poles takes place with such rapidity that, for the most part, the nuclei to which they give rise are constituted before the derivatives of the univalents can approach them. In this way the major nuclei, destined to play the chief part in the construction of the tetrad, tend to be built up of the 14 favoured chromosomes descended from the original bivalents. Again, however, the laggard chromosomes manifest themselves as micronuclei so that, when due allowance is made for those produced at the heterotype division, a considerable number of such may be present. Thus, at the succeeding interkinesis very curious combinations may occur. Plate II, fig. 18, shows a peculiarly anomalous condition with only two major nuclei.

Finally, the major nuclei, with or without the varying numbers of micronuclei, as well as the micronuclei acting independently,. share amongst them the cytoplasm, and attempt to develop a tetrad. Naturally, under the circumstances, the microspores so produced are exceedingly variable in size, build, and composition but, no matter how built up, the end is the same; great or small, with many nuclei or few, the microspores collapse at once. Thus the plant generates no functional pollen.

b. Rosa Wilsoni (fig. 2)

Unmistakably a hybrid involving R. pimpinellifolia and one of the tomentosa-villosa allies, and therefore of similar origin to R. Sabini, Rosa Wilsoni might be expected to repeat, or at least resemble, that plant in its cytology. Much to our amazement, however, meiotic stages succeeded in their proper order with much of the smoothness and regularity noted in R. pimpinellifolia and R. rugosa. This, when compared with 98 per cent, of our roses during this phase is truly remarkable, and special attention will be devoted to it in the concluding remarks.

In its somatic plates it agrees with R. Sabini as will be seen by a glance at Plate I., fig 9, when 42 chromosomes are easily counted. Nor is there any great difference between the two forms until the heterotype prophase when, in diakinesis (Plate I., figs. 1a and 1b), many more paired chromosomes are visible than in R. Sabini and correspondingly less univalents. Indications even exist that all, on occasion, may find mates, although the force determining the pairing of some members of the complex, as noted by Gates in Œnoihera, may be so slight as to make it somewhat difficult to decide whether pairing has occurred or not. Be that as it may, on one heterotype plate after another (Plate I., fig. 5), the chromosomes present are certainly 21 bivalents although, as if to emphasise the looseness of the connection just noted, occasionally 22 (Plate I., fig. 6) are to be seen.

Fig. 1.

Rosa Sabini.

Fig. 1.

Rosa Sabini.

As the chromosomes pass to their position on the equatorial plate, since all show tendencies, in greater or less degree, to act as bivalents, on theoretical grounds one would expect no preferential assembly of a special group at the centre. Nevertheless that does occur, for 7 somewhat larger ones usually take up the central position (Plate I., fig. 5) again contrary to expectation, for on the analogy of most of the other roses one might have anticipated a central group of 14. Around these are massed the smaller paired univalents or bivalents producing an arrangement which strongly recalls the figure of the same stage in R. mollis in our previous paper (fig. 5g). At this point, comparisons might be made between Plate I., fig. 5, representing a heterotype plate of Rosa Wilsoni and Plate II., fig. 12, from R. Sabini, when marked differences in size distribution are noticeable; on the former the central set of 7 are obviously larger, whilst in the latter the reverse holds true, the central 14 appear distinctly the smaller. A comparison between Plate II., fig. 13, and Plate I., fig. 2, will make the reason for this clear. In R. Sabini the outer univalents lie tangentially, but the bivalents, being attached end to end, are foreshortened; hence the former appear falsely to be the larger, whereas in R. Wilsoni all or nearly all are bivalents end to end, so the difference in size on the metaphase plate is a real difference.

Fig. 2.

Rosa Wilsoni, from the type station at Bangor.

Fig. 2.

Rosa Wilsoni, from the type station at Bangor.

In the anaphase, the chromosomes lying on the circumference of the plate once more betray the feebleness of the attraction holding them together by diverging before the middle pairs separate (Plate I., fig. 2). Still, in spite of this, towards the close of the anaphase the members of the central combination of bivalents will be found at the poles before a few at least of their former companions. These laggards are always definitely in pairs one going to each pole and also, curiously, seem frequently to pass to the poles in couples (Plate I., figs. 3 and 4). In spite of their conspicuous lagging, these chromosomes seldom fail to be included in the daughter nuclei, though at times the other chromosomes become clumped before they arrive, as in Plate I., fig. 4. In consequence, instead of the interkinesis between heterotype and homotype division showing both major and micronuclei as in R. Sabini, only the two normal major nuclei appear, save in extremely rare cases. Therefore the chromosomes arrange themselves in preparation for the homotype division without interference from subsidiary spindles or isolated chromosomes and almost always with the utmost regularity (Plate I., fig. 7).

At this stage, in harmony with the heterotype determination, and with even more regularity, 21 chromosomes are made out. Then follows the anaphase and again the outstanding feature is its remarkable regularity (Plate I., fig. 8), though occasional anthers show irregularities throughout. Thus, with the construction of the daughter nuclei, we have generated, not an anomalous structure like that of R. Sabini to which the term “octad” is more applicable, but a genuine tetrad differing in no way from that found in Rosa pimpinellifolia. A careful study and comparison of Plate I., fig. 10, and Plate II, fig. 19, only adds emphasis to this observation.

Only in exceedingly rare cases, and then to a very minor degree, do the members of the tetrad include micronuclei; hence, and again in violent contrast to what obtains in R. Sabini, the pollen of R. Wilsoni rarely aborts.

To sum up our comparison of the two forms, cytologically speaking Rosa Sabini behaves as a hybrid, whilst Rosa Wilsoni, no matter what its real origin, distinctly recalls a pure species.

Upon the fundamental chromosome number 7, to which reference has been made above, there has been built up in Rosa an orthoploid series with diploid, triploid, tetraploid, pentaploid, hexaploid, and octoploid members. Of these the tetraploids occur in two types, one balanced and behaving quite normally during meiosis, and the other unbalanced and exhibiting at that stage remarkable peculiarities leading to the formation of microspores which, when functional, contain nuclei with a much smaller number of chromosomes than expected. To be exact, in the second type of tetraploid (and in this respect they agree with the pentaploid) this number is 7 as shown in our previous paper. Obviously, if any degree of constancy is to be maintained in the somatic cells of these forms, the nuclei of the egg cells in the tetraploids must contain 21 chromosomes and in the pentaploids 28 as Täckholm actually proved.

Thus the two gametes, male and female, of one and the same plant, possess widely different chromosome numbers so that, when such plants hybridise with any member of the balanced groups, the chromosome complements of the two possible reciprocal hybrids should not agree.

Now Rosa Sabini and Rosa Wilsoni are crosses of the type just considered, one parent being a balanced tetrapioid and the other some member of the pentapioid* Tomentosæ. If the pollen parent is derived from the first named section this necessitates a somatic count of 28+ 14 (= 42) chromosomes for the hybrid plant ; on the other hand, if the crossing takes place in the opposite direction, the chromosome complement should be 14 + 7 (= 21). However, we have discovered that both forms possess a somatic number of 42 which ought to indicate that the pollen parent, in both cases, is R. pimpinellifolia. That this is really correct in the case of R. Sabini there can be no doubt; the plants representing that form detected by us grow far away from any pimpinellifolia colony amongst a tangle of Caninæ microgenes of diverse affinities and, moreover, resemble very closely the adjacent Tomentosa forms. Rosa Wilsoni, on the contrary, occurs with R. pimpinellifolia, as its discoverer stated, and except for its large leaves, immediately recalls that species. The presumption is strong, therefore, that R. Sabini and R. Wilsoni do represent reciprocal crosses; which, as we have seen, should mean diversity, instead of similarity of chromosome complements.

This being the case, the discrepancy must be explained. Throughout the foregoing sketch of the cytology of the two plants R. Wilsoni has impressed us by the regularity of its meiotic phenomena. At the reduction division the correct reduced figure was freely encountered on heterotype metaphase plates, and this figure persisted to the end. Clearly, each chromosome in the somatic cells of R. Wilsoni had its homologue present with it and, just as clearly, this ought not, under ordinary circumstances with the parentage assigned, to be possible. However, as to its hybridity, and to its parents being tetraploid and pentaploid respectively, there can be no doubt; nor can there be any hesitation in fixing upon Rosa pimpinellifolia as the seed parent. Hence, in some way a doubling of the original chromosome number has taken place ; in other words, the plant falls into line with Digby’s Primula Kezvensis and Bremer’s Saccharum crosses, in both of which cases chromosome duplication followed hybridisation. Looked at from this point of view, some of the puzzling features of the cytology of Rosa Wilsoni become clear. Since the individual members of the pairs in the double set of chromosomes are actually derived from the same original chromosome, the seven large bivalents reveal themselves as the duplicated representatives of the larger seven introduced by the tomentosa parent; the remainder, therefore, are of pimpinellifolia origin. The pairing being then of actual homologues, it is no longer surprising that microspore development proceeds to perfect pollen just as in an ordinary pure species.

Comparison of cell size offers us evidence corroborating the truth of our explanation. Careful measurements made over a long series of pollen mother cells of the forms involved, i.e. Rosa, pimpinellifolia, R. Sabini, R. Wilsoni, and R. tomentosa, has demonstrated that, on the average, their relative volumes are 11.9, 16.2, 29.4, 13.45, and much the same relation holds in the volumes of their nuclei. Looking closely at these we observe that the cell size in R. Sabini is slightly greater than that of either of its parents,* that of R. Wilsoni being nearly twice that value. Almost certainly a doubling of original cell and nuclear volume has taken place.

Fig. 3.

R. pimpinellifolia. x. R. mollis.

Fig. 3.

R. pimpinellifolia. x. R. mollis.

Let us return for a moment to the fertility of R. Wilsoni. This calls for special comment in view of the persistent sterility of most of the recognised British rose hybrids. They uniformly flower well, but with the disappearance of the corolla the undeveloped fruit immediately falls. Only two other cases of fertile hybrids are known to us in nature; one a cross between R. pimpinellifolia and R. mollis (fig. 3) collected at Corbridge, Northumberland, and the second (fig. 4) described by Barclay from Auchterarder as originating in a cross between R. omissa and R. pimpinellifolia. In both instances the plants accompany the Villosa parent and, presumably, should be endowed with a chromosome number of 21 + 14 (= 35); nevertheless, if they attain fertility by the same path as pursued by Wilsoni, this number should have been augmented to 70—a matter we propose to investigate in due course. Incidentally, we may remark that some occurrence of the same order may explain the circumstances attending Täckholm’s extraordinary octoploid hybrid numbered by him 190.

Fig. 4.

R. pimpinellifolia x R. omissa..

Fig. 4.

R. pimpinellifolia x R. omissa..

From a theoretical standpoint the case of R. Wilsoni is of enormous interest, for it affords us an insight into the mechanism by which a fertile hexaploid form could arise from a cross between lower members of a polyploid series, and assures us that hybridity does play a part in the evolution of such a group. In particular, once the tetraploid appears (and its development is much more easily understood than that of the higher members), a cross between it and a diploid species render the triploid possible; next, by the workings of the process outlined above, the hexaploid appears, to be followed, similarly, by the octoploid.

With the development of the octoploid the way is opened for the Eucanine pentaploid, and thus the orthoploid chain found in Rosa is finally completed.

Our best thanks are due to Rev. Professor M. C. Potter for facilities provided in connection with this work.

Barclay
,
W.
(
1911
), “
Our Native Hybrid Roses
,”
Trans. Perthshire Soc. Nat. Sc.,
V., Part III.
Blackburn
,
K. B.
, and
Harrison
,
J. W. Heslop
(
1921
), “
The Status of the British Rose Forms as determined by their Cytological Behaviour
,”
Ann. Bot.,
35
.
Bremer
,
G.
(
1923
), “
A Cytological Investigation of Some Species and Species Hybrids within the Genus Saccharum
Genetica,
5
.
Cole
,
R.
(
1917
), “
Imperfection of Pollen and Mutability in the Genus Rosa
,”
Bot. Gaz.,
63
.
Digby
,
L.
(
1912
), “
Cytology of the Primula Hybrids
,”
Ann. Bot.,
26
.
Gates
,
R. R.
, and
Rees
,
E.
(
1921
), “
A Cytological Study of Pollen Development in Lactuca
,”
Ann. Bot.,
35
..
Gates
,
R. R.
(
1924
), “
Polyploidy
,”
Journ. Exp. Biology,
1
.
Harrison
,
J. W. Heslop
(
1921
), “
The Genus Rosa, its Hybridology, and Other Genetical Problems
,”
Trans. Nat. Hist. Soc., Northumberland and Durham, N.S
.,
5
.
Matthews
,
J. R.
(
1920
), “
Hybridism and Classification in the Genus Rosa
,”
New Phytologist,
19
.
Rosenberg
,
O.
(
1909
), “
Cytologische und Morph. Studien an Drosera longifolia × D. rotundifolia
,”
Kungl. Svenska Vet.-Akad. Handl.,
43
.
Täckholm
,
G.
(
1920
), “
On the Cytology of the Genus Rosa. A Preliminary Note
,”
Svensk. Bot. Tidskr.,
14
.
Täckholm
,
G.
(
1922
),
“Zytologische Studien über die Gattung Rosa
,”
Acta Hort.Bergiani,
7
.

All the figures were drawn with an Abbe Camera Lucida at a magnification of 4400, except numbers 10 and 19 which were × 1300. The drawings reduced to two-thirds in reproduction.

Plate I. Rosa Wilsoni

Figs. 1a and 1b are the two halves of a diakinetic figure showing 17 bivalents (one cut in half by the knife) and 8 single chromosomes.

Fig. 2. A typical heterotype metaphase showing 6 single chromosomes.

Figs. 3 and 4. Anaphase showing laggards which do, however, finally reach the poles. Note the curious association of these chromosomes in pairs.

Fig. 5. Heterotype metaphase showing 21 chromosomes. Note 7 central large ones.

Fig. 6. The same but showing 22 chromosomes.

Fig. 7. Homotype metaphase showing 21 chromosomes on each plate.

Fig. 8. Anaphase showing regular behaviour.

Fig. 9. Somatic plate showing 42 chromosomes.

Fig. 10. Group of perfect tetrads.

Figs. 1a and 1b are the two halves of a diakinetic figure showing 17 bivalents (one cut in half by the knife) and 8 single chromosomes.

Fig. 2. A typical heterotype metaphase showing 6 single chromosomes.

Figs. 3 and 4. Anaphase showing laggards which do, however, finally reach the poles. Note the curious association of these chromosomes in pairs.

Fig. 5. Heterotype metaphase showing 21 chromosomes. Note 7 central large ones.

Fig. 6. The same but showing 22 chromosomes.

Fig. 7. Homotype metaphase showing 21 chromosomes on each plate.

Fig. 8. Anaphase showing regular behaviour.

Fig. 9. Somatic plate showing 42 chromosomes.

Fig. 10. Group of perfect tetrads.

Plate II. Rosa Sabini

Figs. 11a and IIb are the two halves of a diakinetic figure showing 14 bivalents and 14 univalents.

Fig. 12. Metaphase plate showing univalents lying tangentially.

Fig. 13. Heterotype metaphase showing unusually little abnormality.

Fig. 14. Interkinesis before homotype division. Note micronuclei.

Fig. 15. Somatic plate showing 42 chromosomes.

Fig. 16. Homotype metaphase showing a third spindle and only 19 chromosomes on the plate.

Fig. 17. Homotype anaphase. Three chromosomes left out of heterotype division.

Fig. 18. Peculiarly abnormal “tetrad” with only two major nuclei.

Fig. 19. Group of “octads “giving rise to sterile pollen.

Figs. 11a and IIb are the two halves of a diakinetic figure showing 14 bivalents and 14 univalents.

Fig. 12. Metaphase plate showing univalents lying tangentially.

Fig. 13. Heterotype metaphase showing unusually little abnormality.

Fig. 14. Interkinesis before homotype division. Note micronuclei.

Fig. 15. Somatic plate showing 42 chromosomes.

Fig. 16. Homotype metaphase showing a third spindle and only 19 chromosomes on the plate.

Fig. 17. Homotype anaphase. Three chromosomes left out of heterotype division.

Fig. 18. Peculiarly abnormal “tetrad” with only two major nuclei.

Fig. 19. Group of “octads “giving rise to sterile pollen.

*

The same holds of a very pimpinellifolia-like hybrid, as yet without a name but with the parentage pimpinellifolia × (pimpinellifolia × coriifolia), detected by us on the sandhills of the Northumberland Coast

*

The argument would remain unaltered if the second parent were a pentaploid Eucanine rose.

*

The increase over the mean may be assignable to hybrid vigour.