Cytogenetics and cytotaxonomy of some Brazilian species of Cymbidioid orchids

mosome number of all orchids: 2n = 10 in Psygmorchis pusilla (Dodson, 1957a,b) to 2n = 168 in a horticultural variety of Oncidium varicosum (Sinotô, 1962). There are previous reports for approximately 495 species distributed throughout 60 genera, of which 47 species belonging to 39 genera are from Brazil, representing 9.93% of all species analyzed. Oncidium, Catasetum, Stanhopea, Brassia, Miltonia, and Zygopetalum are the best studied of these genera (Blumenschein, 1960a). Chromosome number variation in Cymbidioid phylad and orchids as a whole is intriguing because most of the genera have high ploidy levels and variable base numbers (Goldblatt, 1980; Ehrendorfer, 1980). The base number of the family is still uncertain, difficulting to estimate species ploidy level and to understand the karyological evolution of the family. Raven (1975) reviewed the angiosperms base number and considered it premature to suggest a base number for Orchidaceae. In the present study, chromosome number and interphase nuclear types were investigated relative to 44 species of 20 genera of Cymbidioid orchids occurring in Brazil. Besides, the variability in chromosome number within the phylad was reviewed, along with its compatibility with the taxonomic treatment proposed by Dressler (1993), and the most probable base number for each genus, subtribe and tribe of the group.

Cymbidioid phylad has the highest variation in chro-mitotic or meiotic analysis) were fixed in Carnoy 3:1 (ethanol/acetic acid) for a period varying from 3 to 24 h and later stored at -20 o C in the same solution.For slide preparation, the material was hydrolyzed in 5 N HCl for 20-30 min at room temperature and stained with Giemsa 2% (Guerra, 1983) or hematoxylin at 1% (Guerra, 1999).Photomicrographs were taken with Kodak Imagelink or Agfa Copex Pan films, using a Leica DMRB photomicroscope adjusted to 25 ASA.

Karyological variation
A total of 44 species belonging to 20 genera and two of the four tribes from Cymbidioid phylad were analyzed (Table I).Chromosome numbers varied from 2n = 12 in Psygmorchis pusilla to 2n = ca.168 in Oncidium aff.flexuosum.No interpopulational numeric variation was observed in species with more than one population analyzed (Bifrenaria magnicalcarata, Catasetum discolor, Cyrtopodium intermedium, C. paranaense, Notylia lyrata, On-cidium barbatum, O. cebolleta, Psygmorchis pusilla, Rodriguezia bahiensis and Trichocentrum cornucopiae).In Oeceoclades maculata, samples of four populations produced clumped cells with 2n = ca.52, but in other three populations, in which the best metaphase was obtained, 2n = 54 was always observed (Figure 1a), suggesting that they have the same number.
Figures 1 to 5 illustrate the karyotype of all species analyzed.Chromosome morphology, whenever observed, was very variable, with metacentric, submetacentric and acrocentric chromosomes in almost all species.Satellites were observed in a few species, and up to two satellites were found in Catasetum barbatum, Coryanthes speciosa, Trichocentrum cornucopiae, Oncidium pumillum and Notylia lyrata.
The interphase nuclei varied from the simple to complex chromocenter types, according to the classification of Tanaka (1971).In Dichaea panamensis, Catasetum barbatum, C. discolor, C. luridum, Dipteranthus duchii, Dipteranthus sp., Cyrtopodium blanchetii, Gongora quinquenervis, Oeceoclades maculata, Trigonidium acuminatum and T. obtusum, along with all the species of In some other families, analysis of the chromatin organization in interphase nuclei has contributed to an un-derstanding of the genomic diversification, independent of number and chromosome morphology (Morawetz, 1986;Röser, 1994).There is a general tendency toward the conservation of a single interphase nuclear type throughout a genus or a higher taxonomic category, as in Rutaceae, subfamily Aurantioideae (Guerra, 1987).In orchids, Tanaka (1971) described five different types of interphase nuclei based on observations in 115 species of 52 genera.However, the occurrence of more than one interphase nuclear type in a single genus has been described, as in Habenaria (Félix and Guerra, 1998) and Platanthera (Yokota, 1990).In Catasetum and Cyrtopodium, which present chromosome numbers and morphology relatively constant, two different types of interphase nuclei occur.Otherwise, the occurrence of simple chromocenter nuclei in nearly all Oncidiinae species seems to reflect the uniformity of this group (Chase, 1986).Therefore, the meaning of this variation in orchids needs to be better understood.
The chromosome number variation of Cymbidioid seems to be much more elucidative.In order to attempt to understand the chromosome numeric variation of the phylad, a complete review of the recorded chromosome numbers was made, based on the review of Tanaka and Kamemoto (1984), followed by the chromosome number indexes published by Fedorov (1969), Moore (1973Moore ( , 1974Moore ( , 1977)), Goldblatt (1984Goldblatt ( , 1985Goldblatt ( , 1988) ) and Goldblatt and Johnson (1990, 1991, 1994, 1996).Furthermore, the chromosome numbers were checked in many original papers, although it has not been possible to obtain copies of all of them, since some journals were very difficult to access.
Table II presents the complete list of cytologically known Cymbidioid species, including original data of the present work.These data are synthesized in Table III, which shows the chromosome numbers recorded within each genus in decreasing order of frequency.The most probable base number of each genus was also tentatively recognized.The base number was identified as one of the haploid number actually found in the genus that most parsimoniously explains the chromosome number variation found in the taxon and more related genera (Guerra, 2000).Based on this concept, it was possible to indicate the number that most probably represents the original haploid complement for each genus.The criterion of the "most frequent" chromosome number was accepted as an indicator of the base number only when it was well represented in the related genera.In many genera, such as Liparis, Eulophia and Odontoglossum, two or more numbers seemed equally probable and were provisorily maintained as base numbers, although only one of them should represent the primary base number of each genus.

Table II -Continued
Some chromosome numbers registered in the literature were not included in Tables II and III because they clearly differed from other records for the same species or were incompatible with the records for the genus.For example, Blumenschein (1957Blumenschein ( , 1960a) ) reported n = 28 in the pollen mitosis of four Catasetum species.However, in Jones and Daker's (1968) analysis of 21 taxa of this genus, including three of the four species reported by Blumenschein, none presented this number.Further in the present work, 2n = 54, the most common number in the genus, was observed in four species (Figures 1h,i and 2a-c) and 2n = ca.108 in two populations of Catasetum discolor.All the counts considered as probably wrong were presented in a separate table (Table IV) and were not included in the discussion.
Numerical variations related to a single species were excluded from Table II, wherever other references confirmed only one of these numbers.In Oncidium microchilum, for example, Sinotô (1962Sinotô ( , 1969) ) and Charanasri

Continued on the next page
Table III -Chromosome numbers and probable base numbers of tribes, subtribes and genera of Cymbidioid (sensu Dressler, 1993).
Chromosome numbers are ordered from the more to the less frequent.Numbers conected with a line have equal frequencies.
Tribes and subtribes with the number Genera with the number of Chromosome numbers reported and more of genera/species known species known/ analyzed probable base numbers (underlined)  II.The number 2n = 41 for Eulophia euglossa was also removed because it was described as an occasional trisomy besides the normal number 2n = 40 (ar-Rushdi, 1971).Similarly, numbers attributed to B chromosomes, like the reference of Aoyama and Tanaka (1988) for a single individual with 2n = 39 + 5Bs of Cymbidium javanicum and 2n = 38 + 1 in C. lancifolium, were excluded.Occasional triploids, like that referred to C. javanicum (2n = 57) by the same authors above, were not considered significant for the cytotaxonomic evaluation of the genus and were also excluded.All these counts were listed in Table IV for future evaluation.Some other seemingly incorrect counts were not excluded for a lack of documentation or a strong argument proving the error.Daker and Jones (1969), for example, suggested that counts with 2n = 42 in the subtribe Stanhopeinae are "largely the result of detached satellites", but they admit that at least Stanhopea peruviana has 2n = 42.In this case all the counts of 2n = 42 were excluded in only S. grandiflora, S. inodora, S. oculata and S. tigrina, because other counts are known that confirm 2n = 40 for these species.In S. wardii and Acineta superba, the only records known were conserved (2n = 41, 42 and 2n = 40, 42, respectively).This "cleaning", albeit partial, reduced the importance of those numbers in the identification of the base number of Stanhopea and Stanhopeinae.

Karyological evolution
The chromosome number variability observed in orchids is not only very extensive but also difficult to relate to a single base number.Cytotaxonomical analysis can be better understood in genera with great cytological diversity, which often correspond to the genera with the highest number of species in the tribe or family, like Boronia in the tribe Boroniae, Rutaceae (Stace, 1995), Carex in Cyperaceae (Luceño, 1994), and Passiflora in Passifloraceae (Snow and MacDougal, 1993).In Cymbidioid, the largest genera are Oncidium and Maxillaria with about 420 species in each one.Maxillaria is very poorly investigated (only six species), whereas Oncidium is the genus most extensively studied of the phylad (117 species).Chromosome number variability in Oncidium is also quite representative of the group.The known haploid numbers are n = 13,14,15,18,19,20,21,22,25,26,27,28,29,30,36,42,56,63,70,84.This variation is clearly dominated by the polyploid series n = 14,21,28,42,56,63,70,84.The great majority (64.8%) are ortoploid with n = 14, 21 or 28, of which 46% display n = 28.These data strongly suggest x 1 = 7 as the primary base number for the genus, al-though this number is hypothetical, since no species of the genus is known with n = 7.Thus, most Oncidium species should be tetraploid (n = 14), hexaploid (n = 21) or octoploid (n = 28).The diploids have not yet been found or were extinct, since the hexaploid n = 21 could only arise from a cross between tetraploids (n = 14) and putative diploids (n = 7) followed by polyploidization (Harlan and De Wet, 1975).Therefore, if the genus was originated from a tetraploid lineage, the hexaploid species could not belong to this same lineage and the genus would be artificial.The same may have occurred in Rodriguezia, with 2n = 28 (Sinotô, 1962) and 2n = 42 (Figure 4a,b, Table II).
When the subtribe Oncidiinae is considered as a whole, the variation of chromosome numbers seems very similar to that of the genus Oncidium (Figure 6), with the numbers n = 21 and n = 28 prevailing, suggesting that the other genera have a common ancestor with Oncidium.The subtribe also has the smallest chromosome numbers of the family: n = 7 in Lockartia and n = 5, 6 and 7 in Psygmorchis.
In three populations of P. pusilla studied in the present work, 2n = 12 and n = 6 were always found (Figure 3g), disagreeing with records of Dodson (1957a,b) and Kugust (1966, apud Tanaka andKamemoto, 1984).Further analyses in other Lockartia species would be important to verify whether the polyploid series observed in Oncidium is also repeated in this genus.The only Lockartia species analyzed in the present work exhibited 2n = 56 (Figure 4d), which coincides with the previous reports of Charanasri and Kamemoto (1975) for L. micrantha.These data support the inclusion of Lockartia in Oncidiinae, in opposition to the assumption of Freudenstein and Rasmussen (1999) based on the absence of leaf articulation in this genus.
Considering the polyploid series observed in Oncidium and Oncidiinae in general, it is reasonable to suppose that x = 7 would be the primary base number of the subtribe, as suggested by Charanasri and Kamemoto (1975).In this case, most Oncidiinae genera would have hexaploid (Comparettia, Notylia) or octoploid origin (Aspasia, Gomesa, Miltonia, Sigmatostalix, Trichopilia).The number n = 7 may represent the original haploid complement of Orchidaceae, found nowadays in very few species.Successive cycles of polyploidy would have originated tetraploid (n = 14), hexaploid (n = 21) and octoploid (n = 28) lineages, some of which gave origin to entirely polyploid genera (Table III).As polyploidy is quite a recurrent phenomenon in the evolution of angiosperms (Soltis and Soltis, 1995;Leitch and Bennett, 1997), it is very probable that higher polyploids arose de novo many times in a number of other genera.
The only cytologically known genera distant from the series n = 7, 14, 21, 28 in Oncidiinae are Ionopsis, Macradenia and Trichocentrum.In Ionopsis, there is only one record with n = 23, whereas in Macradenia there are data for one species with n = 26 and another with 2n = 48 chromosomes (Blumenschein, 1957;Sinotô, 1962).In Trichocentrum, there are records of five species with 2n = 28 and 2n = 24, besides the present count with 2n = 20 in T. cornucopiae (Figure 3h).Trichocentrum may have a dysploid series with n = 14, 12, 10, but the available data are still very fragmented.Chase (1986), based on a combination of floral, vegetative and chromosomal characters, suggested that Trichocentrum could represent an independent evolutionary lineage distinct from the other genera of Oncidiinae.
The present interpretation for the karyological evolution of Oncidium/Oncidiinae conflicts directly with that of Chase and collaborators (Chase, 1986, Chase andPippen, 1988;Chase and Olmstead, 1988;Chase and Palmer, 1992).These authors observed that the most primitive representatives of the subtribe had higher chromosome numbers, whereas Psygmorchis and Lockartia, with more derived morphological characters, like laterally flattened leaves, displayed the lowest chromosome numbers.Therefore they concluded that Oncidium and some Oncidiinae have the original chromosome numbers (x = 28, 30) which, through successive dysploidy, originated the low numbered species with n = 7-5.This conclusion was supported by isoenzymatic evidence from representatives of this group, which almost always exhibited a single locus for each isozyme (Chase and Olmstead, 1988), like dysploids.However, the isoenzymatic analysis of several other definitely polyploid taxa also displayed a similar pattern (Haufler, 1987), suggesting that it is not an accurate indicator of ploidy level (Soltis et al., 1992).
The present interpretation is that the original stock was diploid and had been progressively substituted by polyploids.As polyploids often have very slow evolution rates, they may conserve more primitive characters (Stebbins, 1971), as observed in many present day polyploids of Oncidiinae and other groups (Guerra, 2000).This same reasoning is also applied to other primitive and highly polyploid genera of orchids, such as Neuwiedia and Apostasia (Okada, 1988).On the other hand, diploids and recent polyploids exhibit more derived characters in different parallel evolutionary lines, as Dipteranthus in Ornithocephalinae (Williams et al., 1994) and Lockartia in Oncidiinae (Chase, 1986;Freudenstein and Rasmussen, 1999).
The chromosome analysis of Oncidiinae helps one to understand the seemingly unrelated numbers of the remaining members of tribe Maxillarieae (Table III).Thus, the genera of Lycastinae, Maxillariinae and Stanhopeinae, clearly based on n = 20, may be derived by descending dysploidy from a hexaploid lineage with n = 21.Ornithocephalinae, karyologically known only from two counts in the present work for the genus Dipteranthus with 2n = 56 (Figures 3c,d), coincides with the base number of most Oncidiinae genera, supporting its affinity with that subtribe (Chase and Pippen, 1988).Only the subtribe Zygopetalinae seems to be more diversified in the hexaploid-octoploid level (n = 26, 24/48, 23).
The data from In the other tribes of Cymbidioid the best represented chromosome numbers are n = 15, 21 in Malaxideae, n = 14, 21 in Calypsoeae, and n = 27 in Cymbidieae.In Malaxideae, although n = 15 is a very common number, n = 14 has also been found at least in Liparis and Malaxis.In Liparis, the cytotaxonomic interpretation is made difficult by an apparent secondary polyploid series based on x = 10 (n = 10, 20, 40).What is particularly impressive is the high frequency of species with n = 15 in the three genera of Malaxideae, a very rare haploid number in other Cymbidioid (see Table II).Although Malaxideae is the second largest Cymbidioid tribe, it is notably little known, with less than 10% of its species investigated cytologically.In Calypsoeae, n = 14 has only been found in Calypso, with n = 21 prevailing in the other genera.If these numbers have a evolutionary history similar to that observed in Oncidium, probably they also have or have had representatives with n = 7.
In the tribe Cymbidieae, there is a higher diversity of chromosome numbers, in agreement with the polyphylie observed on the basis of morphological (Freudenstein and Rasmussen, 1999) and molecular evidence (Cameron et al., 1999).The main haploid numbers are n = 27 and 23 in the subtribe Eulophiinae, n = 21, 20, n = 28, 27 in the subtribe Cyrtopodiinae, n = 20 in a single species of Acriopsidinae, and n = 27 and n = 34 in Catasetinae.In general the subtribe Eulophiinae is cytologically represented by Eulophia, which displays the second largest variation in chromosome numbers known in the phylad.In this genus, a polyploid series based on x = 7 (n = 14, 21, 28, 35, 56) is also represented, with the octoploid level (n = 28, 27) strongly dominant.In Oeceoclades, the only two species analyzed are also octoploids, while in Dipodium the only record (n = 23) is probably a hexaploid.Poggio et al. (1986), analyzing the meiotic behavior of several species of Eulophia with n = 21, observed the frequent secondary association of bivalent three-to-three, suggesting that it would be a remaining homeology of the hexaploid condition with x = 7.
In Cyrtopodiinae, the most studied genera are Cymbidium with x = 20 and Cyrtopodium with x = 23.In the present work original data are supplied for six species of Cyrtopodium, one with n = 22 (Figure 1b), four with n = 23 (Figure 1c-e) and one with n = 46 (Figure 1g), reinforc-ing the importance of x = 23 in the genus.Cyrtopodium eugenii with n = 22 is morphologically distinguished from other species of Cyrtopodium by the presence of an inflorescence in raceme, whereas others generally present inflorescence in panicle.The numbers n = 28, 27 and 26 are represented in six of ten genera studied of Cyrtopodiinae and n = 21, 20 dominate in another three, once again suggesting a polyploid series with base in x = 7, followed by descending dysploidy.The genus Cymbidium is notable for its constancy in chromosome number (n = 20), except the species of subgenus Jensoa (sensu Christopher and Cribb, 1984), with 2n = 38 (Aoyama and Tanaka, 1988).According to Freudenstein and Rasmussen (1999) Cymbidium is a member of the Vandoid phylad while Jensoa is part of the large epidendroid polytomy, since Jensoa shows later antera bending and lacks other features such as two pollinia or the presence of endocarpic trichomes.
In Catasetinae, of the three cytologically known genera, Catasetum and Mormodes show x = 27, whereas Cycnoches presents x = 34.Of the five species of Catasetum studied in the present work, four showed 2n = 54 and one 2n = 108 (Figures 1h,i, and 2a,c), confirming x = 27 for the genus.Although Catasetinae and the genus Cyrtopodium display the same pollination syndrome and form a monophyletic group based on cpDNA restriction sites (Chase and Hills, 1992), they are not clearly related karyologically.
As a whole, the great majority of Cymbidioid are ortoploids of the series n = 7, 14, 21, 28 35, 42, 56, 84, or dysploids involving simple reductions.Compared to other large families of angiosperms, such as Poaceae (Hunziker and Stebbins, 1986) or Asteraceae (Watanabe et al., 1995), Orchidaceae stands out for the scarcity of representative diploids, where the Cymbidioid phylad is a very good example.These data suggest that the phylad, and consequently the family, may be older than is generally admitted (Garay, 1972), there having been sufficient time for diploids to be widely substituted by polyploids.

Chromosome numbers and habitat variations
In plants, the conquest of new habitats is often related to the occurrence of polyploidy (Stebbins, 1966).Frequently, polyploid races are associated to more extreme environmental conditions (Ehrendorfer, 1970;De Wet, 1986).In the orchid Anacamptis pyramidalis (L.) Rich., for example, the polyploid cytotypes are more adapted to regions with geologic formation different from those of diploid populations occurring in the same regions (Del Prete et al., 1991).

Table III -
Continued

Table IV -
Cymbidioid species with uncertain chromosome numbers.
Table III suggest the existence of three groups: a larger group (Oncidiinae and Ornithocephalinae), -Chromosome number variation among Oncidium species compared to other Oncidiinae.