Mem. NYHS 3:23-31 (1927)
Int'l Conference on Flower and Fruit Fertility

*Presented to the Conference by Dr. Kathleen B. Blackburn.

Armstrong College, University of Durham, Newcastle-on-Tyne, England

During the past few years much of our research work has been planned with a view to elucidating the problems presented by so-called critical plant genera. Of these, to none have we devoted more time and study than to the genus Rosa. This genus, by the enormous spread of variation within its different groups, has practically defeated all attempts to classify it into species of the same value as those of other genera.

Indeed, so hopeless in some respects is the position that, in spite of the efforts of rhodologists of various countries, we have gone little beyond the position taken up by Linnaeus when he said: "Species rosarum difficillime distinguuntur, difficilius determinantur; mihi videtur naturam miscuisse plures vel lusu ex uno plures formasse; hinc qui paucas videt species facilius eas distinguit, quam qui plures examinavit."

To determine the causes of this uncertainty we have attacked the problem from various angles: that of the experimental breeder, of the field worker, of the parasitologist, of the cytologist, and so on.

In the present paper we propose more especially to deal with the results of our cytological investigations, although we shall not hesitate to utilize any useful facts yielded by our other work.

In our cytological examination of the Rosae we soon found that there existed in the roses a polyploid series based on the chromosome number of seven. More remarkable than this was the discovery that, in addition to forms carrying as their somatic chromosome complement an even multiple of the fundamental number, there existed others just as certainly endowed with odd multiples of that figure; of such forms we have recognized triploids, and pentaploids.

The behavior of roses of this latter type during meiosis could not be other than anomalous, and in the anomalies we see the explanation of much of the variability of the Caninae group of the Palaearctic roses. On the other hand, we naturally expected to find that the diploid, tetraploid, and hexaploid forms, or, as we prefer to call them, microgenes, would pass through all the stages of microspore formation on ordinary lines. This held true of the diploids, but, to our astonishment, we found that the tetraploids and hexaploids were of two types, one with pollen development of the usual type and the other in which it harmonized with the state of affairs in the triploids and pentaploids.

For instance, although Rosa pimpinellifolia, R. humilis, and R. lucida, as well as the whole of the Rosa mollis-omissa group were tetraploid, in their meiotic behavior the former set agreed with the quite ordinary diploid R. arvensis and R. rugosa, whilst the latter, except in chromosome number, could not be distinguished from the anomalous pentaploids of the sections Eucaninae, Tomentosae and Rubiginosae. Similarly, with the hexaploids, although R. acicularis was quite ordinary, the meiotic process in R. Sabini followed substantially the lines of the pentaploids.

To distinguish these two types we applied the term "balanced" to those displaying normal behavior in the reduction division in the pollen mother cells; on the contrary, those in which the course of events was irregular we regarded as "unbalanced." Thus, we spoke of "balanced" and "unbalanced" tetraploids, of "balanced" and "unbalanced" hexaploids.

Next, we shall be asked in what respect is pollen development abnormal?

To discuss this adequately we must digress a little and glance for a while at the classical case of Rosenberg's Drosera hybrids. In that work Rosenberg dealt with the two species Drosera longifolia and D. rotundifolia and with their hybrid D. obovata.

He found that the two parent forms differed in chromosome number, Drosera rotundifolia having 10 chromosomes as its reduced number and D. longifolia having 20; the hybrid showed 30 as its somatic number. Regarding 10 as the base, this makes the hybrid a triploid form, and it becomes significant from the standpoint of the orthoploid series in Rosa.

During meiosis little cause for remark existed in the pure species, but in the hybrid only 10 bivalents appeared on the heterotype spindle with 10 univalents more or less irregularly disposed. At the anaphase the bivalents separated normally, whilst the univalents wandered, split, disappeared in the cytoplasm and so on. Thus, when the pollen tetrads should have appeared in the usual way, owing to these irregularities, many of the grains carried an abnormal chromosome complement and therefore collapsed.

In the triploid, pentaploid and unbalanced tetraploid roses much the same phenomena are to be observed. On the heterotype plates we find seven pairs of bivalents with the other chromosomes scattered irregularly about, either on the plate or elsewhere on the spindle. These bivalents proceed to the anaphase long before the univalents, which, as in the Drosera hybrid, wander, lag on the spindle and show all possible degrees of abnormal behavior. Thus, in many cases, only seven chromosomes appear on the homotype plates, to result subsequently in microspores carrying a nucleus built up of 7 chromosomes, and usually only these pollen grains are fertile. Indeed, often enough, when many chromosomes fail to reach the pole in the heterotype division, supernumerary micronuclei are generated which, also dividing at the homotype stage, result in the appearance, not of pollen tetrads, but of "octads" and similar groups in which most of the grains fail to mature.

Comparing the occurrences in the Canine roses with the chromosome behavior of the recognized hybrid Drosera obovata, and coupling this with the fact that R. Sabini (the cytology of which resembles that of the pentaploids) is a patent hybrid between R. pimpinellifolia and some members of the Tomentosa group, we cannot escape from the conclusion that the Canine roses are themselves of hybrid origin. Recognizing this, and also that the group is dominant in Europe, we next ask how it maintains itself. In the first place let us look at the pollen conditions in a series of Canine microgenes. These yield the following percentages of good pollen:

AFZELIANAE: R. subcristata, Reuteri, subcanina, fugax
EUCANINAE: R. inconspicua, biserrata.
RUBIGINOSAE: R. echinocarpa.
AGRESTES: R. Borrerih.
VILLOSAE: R. coerulea, pseudorubiginosa.
TOMENTOSAE: R. tomentosa.
AFZELIANAE: R. frutetorum, coriifolia, venosa.
EUCANINAE: R. lutetiana, hemitricha, aciculata.
TOMENTOSAE: R. sylvestris, pseudocuspidata, foetida.
EUCANINAE: R. flexibilis.
RUBRIFOLIAE: R. rubrifolia.
VILLOSAE: R. submollis.
EUCANINAE: R. fallens.
VILLOSAE: R. omissa.
EUCANINAE: R. senticosa.
VILLOSAE: R. mollis.

Admittedly, whilst most of these show a high degree of pollen sterility, some of these microgenes produce enough sound pollen to be more or less effective; others, however, do not. How then is reproduction effected? That some method of securing it does exist is speedily proved. Take, for example, R. subcristata, R. fugax and R. coerulea, in all of which the whole of the pollen aborts. Visit the shrubs in September, and no rose makes a braver display of crimson globes than they. It looks clearly as if they relied on some process other than that of normal fertilization. This we tested experimentally by castrating many plants of the Eucaninae, Villosae and Rubiginosae and bagging them. Without exception, seeds were set; hence the Canine roses, to say the least, are facultatively apomictical, and this has a definite bearing on the question of their origin. In 1918 Ernst published a lengthy treatise, theoretical in the main, in which he advanced the view that hybridity and apogamy are linked in the way of cause and effect. Though the evidence adduced points in that direction, in the absence of indisputable proof this is little more than a pious expression of opinion.

Recently, however, we have been able, as a result of hybridity experiments with lepidoptera, to demonstrate that parthenogenesis does arise in the F1 hybrids between Tephrosia crepuscularia and T. bistortata (Harrison and Peacock, 1925). The occurrence of apimixis is thus another link in the chain of evidence connecting the conditions of hybrids with those of the Canine roses.

Assigning due weight to the meiotic and other facts in the triploid, pentaploid, and unbalanced tetraploid and hexaploid roses, we are inclined to think that they have arisen as hybrids. There would thus be a perfect agreement with the usually accepted theories as to the development of polyploid series in plant genera. But how do we account for the higher balanced members of the series?

Clearly, if seven be the base chromosome number in Rosa, and diploid species occur, some of them must be more or less primitive and must have participated in the events which gave rise to the polyploids; hence we must look about for some process in which they can play a part. The simple crossing of two diploids would yield a diploid F1, which, with a pairing of homologous chromosomes in the meiotic phase, brings us no further forward. However, we have direct evidence to offer as to the exact methods by which polyploids arise in Rosa. Rosa Wilsoni is a hybrid between R. pimpinellifolia and R. tomentosa, with the latter acting as pollen parent. Theoretically, therefore, it ought to have 14 + 7 chromosomes as its somatic number. Instead, direct cytological investigation shows it to be endowed with 42. Obviously, chromosome doubling by some means or other has occurred, but, what is most noteworthy, although the reciprocal hybrid is unbalanced and sterile, this is balanced and fertile. Thus we have generated before us, by the union of an egg with 14 chromosomes and a pollen grain with 7, a fully fertile hexaploid rose.

We, therefore, see not the slightest reason for rejecting in Rosa the usual views as to the hybrid origin of orthoploid series in plant genera.

If the fertile hexaploid arises thus, then the fertile tetraploid can be developed by a similar doubling in a cross between two diploids. In that case, once again, the supplying of homologous chromosomes would end in the attainment of fertility—and so with other members of the series.

Further, we believe that such happenings have taken place more than once in Rosa, for in the various sections the polyploid chains are quite independent. Take for example, the Cinnamomeae; starting with the diploid R. cinnamomea, we have the tetraploid R. pendulina and the hexaploid R. nutkana; in the Carolinae the diploid R. nitida and the tetraploid R. lucida and in the Pimpinellifoliae the diploid R. Hugonis and the tetraploid R. pimpinellifolia.

Recently, Hurst, arguing from the same series of observations as ours, has arrived at vastly different conclusions as to the development of such series. Instead of believing in the synthesis of the polyploids from lower members of the series, he regards all as having been derived from a theoretical decaploid polar species by the successive losses of chromosome septets. This view seems to us untenable and the reasons for our opinion are as follows:

Leaving on one side the parallelism existing between the conditions in Rubus, Crataegus and Rosa, we fail utterly to see why the same explanations offered and accepted as satisfactory for other plant groups like Chrysanthemum and Campanula should be rejected here.

One always regards, in matters numerical, the simple as the more primitive and the complex as the more advanced, yet here we are asked to believe that the decaploid existed first. This, as we have indicated, is in itself very improbable, but in addition is far from agreeing with chromosome numbers not only in primitive Rosaceae but also in the Leguminosae—an order, like the Rosaceae, a member of the cohort Rosales. There, likewise, a frequent chromosome number is 7 as in the genus Lathyrus, and we should rather imagine the two orders as making contact in their more primitive and fundamental numbers.

Again what is the most primitive species of Rosa? And what is its chromosome complement? The answer is Rosa persica—a diploid—primitive in respect to its simple leaves and the structure of its fruit. And with which other group does this primitive form link up? Unmistakably, it is closely allied morphologically to the Cinnamomeae, which are thus determined as forming a very early group indeed.

As we have seen already the Cinnamomeae possess in themselves a reasonably complete polyploid series and, furthermore, are a very homogeneous group of species well represented in Europe, Asia and America. In America we find only Cinnamomeae forms (with the doubtfully distinct Carolinae) and in the Palaearctic area, in addition to Cinnamomeae, our plethora of forms and groups.

In seems very unlikely that this distribution should have arisen by "defects" from a circumpolar decaploid species, but closer study of the circumstances attending the distribution of the Cinnamomeae still further lessens the plausibility of the view. Apart from the fossil evidence, which shows Rosa to have come into being in early Tertiary times, the geographical distribution of the Cinnamomeae is very significant. Comparing it with that of the less advanced members of the lepidopterous subfamily Bistoninae, which has been submitted to intensive study (Harrison, 1916), we find it to coincide with that of the twin genera Amphidasys and Biston, or, to take plant genera, with that of the Farinosae group of the genius Primula. We have, for very many reasons, assigned to Amphidasys and Biston a northeastern Asiatic origin, and most observers assume for Primula the same metropolis. Hence we imagine the Cinnamomeae to have originated in the same area and to have spread thence, giving rise in America to R. blanda, R. pratincola, R. arkansana, etc., and in the Palaearctic region to R. Sweginzowii, R. pendulina, R. setipoda, R. rugosa, R. cinnamomea, etc. Many rhodologists regard the other sections as having been evolved from this group by mutation, by response to environmental influences or by hybridity. We were probably the first to state the latter view in the case of the Eucaninae and Rubiginosae, but Boulenger has extended it to include the Rubrifoliae, Villosae and Tomentosae.

We are inclined to think that the American Cinnamomeae have had a dual origin, regarding R. nutkana, not as a form derived by chromosome loss from R. acicularis as it moved south, but as a very recent immigrant to Pacific America and therefore comparable with the butterflies Papilio machaon and Thecla (Callophrys) dumetorum. Hence its origin (and possibly that of other western Cinnamomeae) is quite independent of the eastern representatives of the Section. Of this we have rather neat independent proof.

Of the gall-making groups attacking Rosa none is more characteristic than the Cynipid genus Rhodites; wherever roses exist there we meet with that genus. But the American forms, except for one species, form a very homogeneous group. And what is this exception?—Rhodites bassetti closely allied to our common Palaearctic Rhodites rosae and forming a very similar gall, confined to R. nutkana!

[CybeRose note: Beutenmüller, W. New Species of Rhodites from Oregon.—Canadian Entomologist, London, Ont., 1(9): 305-309 (September 1918) The author, an entomologist, merely mentioned that the species was found on what he assumed to be Rosa nutkana, not that it avoided every other species.]

Hurst states in discussing this matter of geographical distribution, "In fact a general statement might be made that towards the Pole the number of septets of chromosomes and characters increases, while toward the equator the number decreases (Fig. 175)." Turning to this map, we see pictured a state of affairs, which, if accurate, would warrant the statement.

It does supply the distribution of Rosa as a whole and thereby confirms our opinion as to the Eastern Asiatic origin of Rosa. In addition it suggests an early migration of the Cinnamomean roses with later wanderings of other groups which took the usual routes of members of the Siberian and Oriental migrations passing into Europe and N. Africa.

However, on testing the accuracy of the various details of the map we find that they do not agree with the facts, the forms high in polyploid series all being placed much too far north and, on the contrary, the diploid species uniformly appearing too far south. A semblance of harmony is thus produced between the distribution and the main thesis.

To take actual details of the map, we find octoploid forms figured as occurring far to the north in America, Iceland, Europe and Siberia. This is erroneous in several respects. In the first place, it assumes that Rosa acicularis, the rose inhabiting the Holarctic region, is octoploid throughout. Admittedly Tãckholm found an octoploid R. acicularis in Europe, but this does not alter the fact that he also examined hexaploid forms of the same species. Apart from this, Penland and ourselves, the former from United States material and ourselves with material from the northern Manitoba-Saskatchewan border, found the American plant to be hexaploid. Moreover, the American distribution indicated is much too restricted, for R. acicularis has been collected by ourselves in Saskatchewan, Manitoba and Ontario, whilst its occurrence is reported from Michigan (Erlanson) N. Y. (Standley) and Texas (Schneider). Similarly, the figure "8" is placed in Iceland to indicate that R. acicularis grows there, but, as is well known, the only rose there is the tetraploid R. pimpinellifolia!

Turning next to the Mediterranean area, we note in Africa only the figure "2," the same figure labelling the rose flora of S. Spain, Italy and Sicily. Once more the map is not in accordance with the facts, the roses in the Canaries, Madeira, N. Africa, Italy, Sicily, S. Spain being predominantly pentaploid.

Submitting now the Asiatic distribution to critical examination, and testing it on the diploid Rosa rugosa, we find it represented as stopping short in Corea and S. Japan—although this rose is well known from Kamchatka more than a thousand miles to the north.

Although the list of mistakes like these could be lengthened almost indefinitely, we shall confine ourselves to one more example—a case from our own islands. Here we see situated in the north of England and Ulster the figure "6" which implies that we possess a hexaploid microgene. This we deny completely; we do certainly possess hexaploid forms, but these are obvious hybrids between tetraploids and unbalanced pentaploids. How casual hybrids, liable to be produced at any time, can have any bearing on the declared purpose of the map we utterly fail to see.

Let us now consider what support the cytology of the microgenes affords this theory of Hurst's.

To simplify the discussion, let us adopt that worker's symbolical representation of his theoretical circumpolar decaploid. To it he assigns the formula AABBCCDDEE, each capital denoting one of his postulated differential chromosome septets; in the same way his supposed derived diploids have the formulae AA, BB, CC, DD, and EE, respectively.

On this basis he represents Rosa "canina" as AABDE, thereby indicating (1) that it possesses only one set of the septets B, D and E, but (2) that the A set is duplicated. This, of course, is intended to be in agreement with the fact that on the heterotype spindle during pollen formation in this group of species there are seven bivalent and 21 univalent chromosomes. It, however, implies that the individual septets will only pair up with others of the same order as themselves; A with A, B with B and so on.

This cannot be correct, as is easily shown. For instance, in the pimpinellifolia X canina hybrids, where a species BBCC (on this notation) is crossed with a form AABDE, in pollen formation 14 chromosomes find mates. This, accepting as correct Hurst's position, would mean that, in addition to the two BB sets of the hybrid pairing as one would anticipate, the C introduced by R. pimpinellifolia had paired up with septet A, D or E, a very improbable thing if the "septets" exist, and are differential. Much the same kind of evidence is provided by the rugosa-cinnamomea, nutkana-pendulina, lucida-rugosa, cinnamomea=pendulina, and the multitudinous chinensis crosses.

[CybeRose note: In Hurst's notation, pimpinellifolia is BBDD. Thus, 14 pairs would be expected in the cross with canina, which is what was observed.]

Such occurrences seem much more in accord with a hybrid origin for the orthoploid series, and allow for varying shifts of value as the various diploid, and tetraploid species diverged physiologically from the species from which they were evolved.

Let us consider now Hurst's own cytological discussion as to the origin of Rosa species.

*see Malinowski (1927) on chromosome affinity,

He begins by formulating the cytological difficulty that the union of two of his differential diploid species AA and BB could only result in a sterile diploid AB. Alternatively, he assumes that, if fertile, such a hybrid could only give rise to plants of its own build and plants like its parents of composition AA and BB. We can only say that, not only is the sterility of such hybrids "not proven," but, on the contrary, in our experience, and in that of rose growers generally, the majority of such hybrids are fertile. Further, we ask for a single trace of proof that the chromosomes in diploids, instead of acting independently as in other plants, move in blocks of seven.* We have looked seriously for evidence of such behavior but have never been able to detect it.

[CybeRose note: I have not read of any AA x BB hybrid that was fully fertile. This would be something like Chinensis x Hugonis or Multiflora x Willmottiae. The Rosa huntii Hurst was reported to be AABB, the only species of the type.]

Moreover, the weightiest evidence we can adduce, i.e., the case of Rosa Wilsoni, demonstrates conclusively that the unlikely (in Hurst's opinion) event of chromosome duplication does occur, and does result in fertility. Hence we consider that the production of an AABB form from AA and BB parents a far from unlikely contingency. In addition, in our opinion gratuitously, he asserts that such a duplication gives rise to a homozygous tetraploid species whilst all the tetraploid species of Rosa are certainly heterozygous. Granting that the Rubrifoliae and Villosae are heterozygous tetraploids, what about R. pimpinellifolia, R. pendulina, R. humilis, R. lucida and the other balanced tetraploids, all necessarily homozygous? Even Hurst himself concedes the homozygous nature of R. spinosissima and R. altaica by writing their formulae as BBCC and BBDD, respectively.

[CybeRose note: This argument is a bit hard to follow. In fact, nearly-sterile diploids are known to produce, sometimes, fertile tetraploid offspring. As for heterozygous vs. heterozygous, Hurst only meant that the species are true-breeding for large sets of characteristics associated with the "septet" number. Furthermore, he distinguished the alpine R. pimpinellifolia (BBDD) from the R. spinosissima (BBCC) of beach and shore. The supposed "heterozygosity" exists only when the two are lumped into a single textbook species.]

To buttress up his case against the synthetic nature of rose polyploids he raises the geographical difficulty that the species necessary to generate polyploids are not present in the same stations to hybridize. Allowing for the immense age of the genus, we should indeed be surprised, with all the movement of plant species that has taken place, if such were so. This argument is much on the same level as supposing that Hurst's own theoretical decaploid, and all its possible "defect" species, should yet exist where they originated!

[CybeRose note: Another non-argument. Species that do not share a habitat today, may well have been together, at least in passing, during episodes of global climate change.]

Next he propounds the view that the evolution of the Arctic polyploids, by the hybridization of subtropical diploids, implies a distinct reversal of accepted opinions as to the Arctic origin of the Flora of the Northern Hemisphere. Once again we have two assumptions stated as facts: (1) that Rosa is Arctic in origin, and (2) that the diploids are subtropical; in other words Hurst is here arguing in a circle. We have already put forward our ideas as to the origin of Rosa whilst the second point concerning the subtropical nature of the diploids rests simply on his map and is not warranted by facts.

[CybeRose note: Hurst (1925) actually wrote: "The facts of distribution at once suggest two distinct possibilities concerning their evolution: either a line of descent from one primitive decaploid species by successive losses of septets of chromosomes and characters, or a line of ascent from the five simple diploid species by hybridisation and subsequent duplications of septets of chromosomes and characters. The fact that the polyploid species show the combined characters of the five diploid species seems to support the idea of origin by hybridisation and duplication." And I [CybeRose] will add that the two modes are not mutually exclusive. A line of ascent from diploids to polyploids, followed — after the Ice Age — by a descent from a hypothetical decaploid or an existing pair of differential octoploids (AACCDDEE and BBCCDDEE) to species with fewer chromosomes.]

We ourselves have collected diploid R. blanda far north in Canada; diploid R. rugosa goes up to Kamchatka in Asia, diploid R. arvensis reaches Scotland, diploid R. cinnamomea is more characteristic of Scandinavia than of Southern Europe and so on.

Again "an origin from five primitive diploids," he asserts, "implies a multiple origin with its consequent ulterior problems." We cannot agree with this. In species after species of Silene the reduced chromosome number is 12; in many Lepidoptera it is 31 and in Orthoptera 12. No inherent difficulty lies there to a multiplicity of diploid species; why then should we invent one for Rosa? The derivation of the diploids from a primitive polyploid is therefore not a more simple proposition.

[CybeRose note: One obvious problem would be getting five differently adapted species to merge into one hypothetical decaploid or two actual octoploids. Of course, this need not happen in one improbable event. For example, R. arkansana (CCDD) has made its home in the prairie states whether neither R. nitida (CC) nor R. blanda (DD) is found.]

We see a further objection to Hurst's hypothesis in the assumption that each individual septet of chromosomes in the original decaploid bears all the potentialities of the rose species. This seems an extremely improbable situation, and entirely at variance with all the pertinent facts, for, whilst we do know that each multiple of the base number does carry the whole powers in the case of a tetraploid Primula, we are likewise certain that such a Primula arose by a duplication of the pre‑existing complex.

[CybeRose note: The existence of an autotetraploid Primula does not exclude the possibility of an allotetraploid type, such as Primula kewensis.

To sum up: as a direct deduction from the pollen sterility and the course of its development in the unbalanced roses, we conclude that such roses originated in hybridity and, secondly, from allied considerations in the case of R. Wilsoni, we regard the balanced polyploids as of similar origin attaining fertility simultaneously with, and as a direct consequence of, a doubling of their chromosome complements. Thus we would seek to account for the orthoploid series developed in the various sections into which the roses are divided.

[CybeRose note: Please note that the authors have carefully ignored the basic facts that warranted Hurst's working hypothesis. The diploid species fall naturally into five groups that can be distinguished by sets of fifty (50) traits. The fact that the genus Rosa has made such remarkable use of differential polyploids is confusing to those people who prefer to look only at glands on the sepals and pubscence on the under sides of leaves.]

[I have to acknowledge a grant from the Armstrong College Research Endowment Fund Committee in aid of the work discussed above. J.W.H.H.]


It is amusing that Heslop-Harrison should be so dismissive of Hurst's efforts considering what he wrote in 1921.

"The similarity of the tables developed by both Almquist and myself to that drawn up by Mendeléeff to illustrate the Periodic Law in classifying the chemical elements cannot be too strongly emphasised. Almost to the most minute detail they agree; we have the vertical groups of the chemist agreeing with my section-species, the horizontal series or periods with our species-types. Furthermore, if we allow for the division of the six members of the Caninae into three pairs of more nearly related section-species, we have an immediate parallel to the divergence of the groups in Mendeléeffs table into two subgroups. For instance, the relationship between the Eucaninae and Afzelianae is much the same as that between the Lithium-Sodium-Potassium subgroup and the Copper-Silver-Gold subgroup. Again, the Semivillosan R. faroensis and R. orientalis, with R. pimpinellifolia, suggest in a striking way the transitional elements in the chemist's group eight. Lastly, it would not be a surprising thing should it turn out in the end that the parallel nature of the two tables is founded on chemical differences of the same type."

Hurst bibliography