A review of chromosome cytology in Hyacinthaceae subfamilies Urgineoideae and Hyacinthoideae (tribes Hyacintheae, Massonieae, Pseudoprospereae) in sub-Saharan Africa

The chromosome cytology of subfamilies Hyacinthoideae and Urgineoideae of the monocot family Hyacinthaceae are reviewed for their centres of diversity in sub-Saharan Africa within the framework of a recent molecular-based classification. We also provide some new chromosome counts for genera or species that are unknown or poorly known cytologically. We conclude that the ancestral basic chromosome number for Hyacinthoideae is x=10 but tribe Pseudoprospereae evidently has x=9, the most likely base in Hyacintheae. Tribe Massonieae has five of the nine (out of 10) genera counted apparently ancestrally tetrapaloid with 2n=40 and we infer a basic chromosome number for the tribe of x=10 based on patterns within the tribe and by outgroup comparison. An extensive descending dysploid series is present in Lachenalia, ranging from a possible ancestral base of x=10 to n=5, and several species are polyploid or have diploid and polyploid populations. Basic number in Urgineoideae is also x=10 and the subfamily exhibits little divergence from that base among sub-Saharan species. Polyploidy at species rank is relatively rare among the sub-Saharan members of both subfamilies. Based on available data just 7% of species of Urgineoideae and 15% of sub-Saharan Hyacinthoideae are species level polyploids but several more have diploid and polyploid populations. This conforms to the pattern of low level of polyploidy in subfamily Ornithogaloideae of Hyacinthaceae and other sub-Saharan families of geophytic plants. © 2012 SAAB. Published by Elsevier B.V. All rights reserved.


Introduction
Recent, molecular-based classifications of Hyacinthaceae recognize four subfamilies: the monogeneric South American Oziroëoideae and the three much larger Old World Hyacinthoideae, Ornithogaloideae and Urgineoideae (Manning et al., 2004). Subfamily Hyacinthoideae is subdivided into three tribes, the species-rich Eurasian Hyacintheae, the monospecific southern African Pseudoprospereae, and the predominantly sub-Saharan African Massonieae with 10 genera and ± 104 species in Africa and southern Asia. Urgineoideae are largely sub-Saharan African with several species in Eurasia as far east as India. Massonieae has a similar distribution but with a marked secondary radiation in Eurasia. This review complements the similar study in Ornithogaloideae (Goldblatt and Manning, 2011) and completes our cytological review for the family in sub-Saharan Africa.
Chromosome counts for Hyacinthoideae and Urgineoideae are widely scattered in the literature, many of them published under genera and species that are now relegated to synonymy. We assemble here all published counts for the two subfamilies excluding Hyacintheae (none of which occur in sub-Saharan Africa), under their current names and arranged according to the most recent infrafamilial classification (Manning et al., 2004). We also provide some new chromosome counts for genera and species uncounted or poorly known. We analyze the patterns of variation in chromosome number in relation to the molecular phylogeny, infer basic numbers for genera and sections, and highlight important gaps in our knowledge of the cytology of the subfamilies. These gaps limit a deeper understanding of the chromosomal evolution in some lineages of Hyacinthoideae but we infer an ancestral base for both Hyacinthoideae and Urgineoideae of x = 10, which we also identified as the likely base number in the sister clade Ornithogaloideae (Goldblatt and Manning, 2011).

Published counts
Data on chromosome number in genera of Hyacinthoideae and Urgineoideae were extracted from indexes to plant chromosome numbers covering the years since chromosome numbers were first made available in accessible compilations (Bolkhovskikh, 1969;Goldblatt, 1981;Goldblatt and Johnson, 1990;Moore, 1973Moore, , 1974Moore, , 1977; see Tables 2 and 3). We tabulate counts according to the current subfamilial classification, with species names corrected to reflect current nomenclature and taxonomy (Manning et al., 2004). Original sources were consulted for most counts, especially those we had reason to question. We had hoped to check voucher specimens for questionable counts but those for the important contribution by De Wet (1957) could not be located at PRE where they had been deposited (C. Archer pers. comm. 2009). Most papers published before the 1980s do not list voucher specimens. There is no precedent for ignoring chromosome counts not linked to voucher specimens and we see no reason to do so here. Examples of doubtful identification are discussed in the text and noted in Table 3.
Patterns of change in chromosome number and karyotype are inferred using established hypotheses for these phenomena (Jones, 1970;Raven, 1975;Stebbins, 1950Stebbins, , 1971. Polyploid sequences are interpreted as proceeding from lower to higher numbers by doubling. Dysploid (aneupoloid) sequences, i.e. stepwise changes rather than doubling of base numbers, are widely believed to be largely descending in a process involving translocation of chromosome material to a second chromosome and loss of a centromere plus those genes associated with cell division. Descending dysploid reduction frequently results in translocation of a long arm of an acrocentric chromosome to the short arm of another and loss of the centromere of the donor chromosome, resulting in a large metacentric chromosome (and lower base number), a process often called chromosome fusion or Robertsonian translocation.
We do not list authorities for species in the text as these are included in Tables 2 and 3.

Original counts
Material for the original counts reported here ( Fig. 2; Table 1) was prepared according to the protocol described by . The vouchers are housed at the Missouri Botanical Garden Herbarium (MO) and Compton Herbarium (NBG). Counts are based on samples of three to four individuals and are assumed to represent entire populations, following widespread practice in plant cytology.

Results
The results of our review of the literature and our original counts are presented in Tables 1-3. Postulated ancestral numbers are plotted on Fig. 1. Karyotypes for three of the four species counted for this study are illustrated in Fig. 2.
Total DNA content per cell in Urgineoideae is established only for Bowiea volubilis, 1C = 4.63 pg (Bennett and Smith, 1976).
Although there are no formally published counts for the Tenicroa/Sypharissa group of Drimia (Table 2), Speta (1998) indicated that Tenicroa has 2n = 20, without references and without listing the species counted. This is consistent with Urgineoideae. Numbers cited by Speta (1998) for three other generic synonyms of Drimia, again without referencing sources or species counted, are more problematic, namely those for Rhadamanthopsis (2n = 18, 16 and 12), Rhadamanthus (2n = 18) and Urgineopsis (including only D. salteri (Compton) J.C. Manning & Goldblatt,2n = 14). In light of all other published chromosome numbers in Drimia, these counts are questionable, and require full documentation. In comparison, published counts for two species of Rhadamanthus (now included in Drimia) are both 2n = 20 (Table 2).
Urginea langii Brem., counted by De Wet (1957), is a synonym of Albuca seineri (Engl. & Krause) J.C. Manning & Goldblatt (Ornithogaloideae). De Wet's count of 2n = 20 for the species is consistent with our suggested base number of x = 10 for Albuca subg. Namibiogalum, to which A. seineri was referred by Goldblatt and Manning (2011). The only other published count for A. seineri is 2n = 24 (Vosa, 1980). When discussing the cytological evolution of Albuca we were unable to explain the significance of Vosa's (1980) count and the strongly bimodal karyotype, which was inconsistent with that of A. donaldsonii Rendle (2n = 20), only other member of subg. Namibiogalum counted. The karyotype of A. donaldsonii described by Stedje and Nordal (1984) [as Ornithogalum donaldsonii] is moderately asymmetric but not bimodal. The karyotype illustrated by De Wet (1957) accords broadly but not exactly with A. donaldsonii. We conclude that Vosa's (1980) count is more likely for another species, possibly for a member of Albuca subgen. Urophyllon in which A. seineri (as Ornithogalum) was included at the time of Vosa's count (Obermeyer, 1978) and in which karyotypes are consistently bimodal.
Based on the scenario above it seems reasonable to hypothesize that the basic, ancestral chromosome number in Bowiea and Drimia, and thus for Urgineoideae as a whole, is x = 10 ( Fig. 2), not x = 5 as suggested by De Wet (1957). On available data, just one of the 14 counted species of Drimia in sub-Saharan Africa is exclusively polyploid, representing 7% of those species that have been examined cytologically.

Hyacinthoideae
Based on the two available counts, Pseudoprospero firmifolium (Baker) Speta, only species of tribe Pseudoprospereae, has 2n =18 (De Wet, 1957;Jessop, 1970 [as Scilla firmifolia Baker]). The karyotype illustrated by De Wet (1957) consists of a graduated series of relatively large chromosomes, both acrocentric and metacentric. This suggests a base number for Pseudoprospereae of x = 9. The genus and tribe are thus evidently dysploid and derived from our hypothetical ancestral base of x = 10 as found in the outgroup Urgineoideae (Fig. 2).
We have not examined the cytology of the largely Eurasian Hyacintheae (12-20 genera, for which there are numerous counts) in detail but the following observations are relevant to our review. Wetschnig and Pfosser's (2003) phylogeny of Hyacintheae places Barnardia scilloides Lindl. (= Scilla scilloides (Lindl.) Druce) as sister to the remaining members of the tribe included in their study. B. scillaris may have x = 9, although the cytology of the species is complex, with numbers of 2n = 16, 18, 34 and 36, sometimes with B chromosomes (e.g. Araki, 1972;Bang and Choi, 1993;Haga, 1962;Haga and Noda, 1956). B. numidica (Poir.) Speta (= Scilla numidica Poir.), which has not been sequenced for phylogenetic study, has 2n = 18 (Cardona, 1991) but no other species of the genus appear to have been studied cytologically. We offer a preliminary hypothesis that x = 9 is ancestral for Hyacintheae, which seems at least plausible given the phylogenetic position of Barnardia in the tribe and the base numbers in related tribes but we refrain from further speculation (Fig. 2).
In Massonieae, the only counts for Merwilla, sister to the remainder of Massonieae, are 2n = 40 (  Schizocarphus also has 2n = 40 (again we question De Wet's report of 2n = 28 and 56 in this genus). The count of 2n = 38 for the genus by Chaudhuri and Sen (2001) may represent dysploid plants. Their calculation of total DNA per cell of 4C = 16.18 pg (1C = 4.03) appears inconsistent with polyploidy because Veltheimia, with the same diploid chromosome number, has 1C values of 9.99 and 10.73 pg (see below). Although cross genus comparisons of C value cannot always be relied to produce valid inferences of homology and ploidy level this merits mention here. Additional counts are needed to establish that there are no diploid populations in these two genera, both of which are relatively widespread.
All counts for both species of Veltheimia, sister to Massonia plus the uncounted Namophila (Fig. 2), are also 2n = 40 (Table 3). Total DNA for both species, determined by Zonneveld et al. (2005), are 1C = 9.99 and 10.73 pg (chromosome numbers not recorded for either sample), which is consistent with polyploidy when compared to 1C values for Massonia (mean value for three species 1C = 3.20 pg) and for Bowiea volubilis (1C = 4.63 pg; 2n = 20). Like Merwilla and Schizocarphus, Veltheimia must be inferred, on available data, to be ancestrally tetraploid.
In Massonia the only count for M. bifolia (= Whiteheadia bifolia), sister to the remaining species of Massonia, is 2n = 40, thus evidently tetraploid, but other species of Massonia (Fig. 2, Table 3) appear to be ancestrally diploid, with 2n = 18 and 22. The report of 2n = 26 for M. depressa by Johnson and Brandham (1997) may be for some other species. Karyotypes in the genus are moderately bimodal. In our sample of M. depressa we recorded two long and seven shorter chromosome pairs, and for M. echinata two long and nine short pairs (Fig. 2). Total DNA per cell is known for three species, M. depressa (1C = 3.36 pg), M. pustulata (1C = 3.19 pg) and M. sp. (1C = 3.05 pg) (Zonneveld et al., 2005). Although chromosome numbers were not recorded for these samples we provisionally assume that each was diploid as this is the only ploidy level recorded for the two named species. Despite the limitations of cross genus comparisons, we note that genome size in these Massonia species (mean 1C = 3.20 pg for the three species examined) is consistent with ancestral diploidy when compared with genome size in Bowiea (Urgineoideae) (1C = 4.63 pg). Additional counts in Massonia, in which only half the species have been counted, will be helpful in interpreting the cytological evolution of the genus. Based on available data, we hypothesize an ancestral base for Massonia of x = 10, with M. bifolia interpreted as tetraploid, although this is based on a single count and possibly a single plant ( Table 3). Records of 2n = 22 in Massonia may represent the presence of B chromosomes or ascending dysploidy.
Lachenalia, with ± 120 spp., is the largest of the sub-Saharan genera of Hyacinthaceae (Goldblatt and Manning, 2000), and sister to the Namophila-Massonia clade (Fig. 2). It is cytologically complex (e.g. Hamatani et al., 2004;Ornduff and Watters, 1978;Spies et al., 2009) and we do not list the numerous published counts for the genus, in which some 86 species have been counted. A descending dysploid series is evident, with diploid numbers of 2n = 20, 18, 16, 14, 12 and 10, with 2n = 14 the most common number (in 30 spp.). Notably L. mutabilis has populations with 2n = 7, 6 and 5. Some 18 species have 2n = 16 and nine species have 2n = 18. Two species, L. comptonii and L. undulata, have 2n = 20 and seven have 2n = 11. Of the species counted, 15 are exclusively polyploid, most on secondary base numbers of x = 14 or 11; and 14 more have diploid and polyploid populations. Just one species of the Polyxena/Periboea group, which is deeply nested in Lachenalia, has been counted, L. ensifolia (Thunb.) J.C. Manning & Goldblatt, with 2n = 24 and 26 (Johnson and Brandham, 1997). The species may be tetraploid on a secondary base. Most available cytological studies do not illustrate karyotypes (or at least not accurately enough), preventing comparisons of total chromosome length among species with different numbers as a crude estimate of ploidy level.
We hypothesize an ancestral base of x = 10 or 9 for Lachenalia as most likely in light of our inferred base number of x = 10 for the sister clade, Veltheimia and Massonia (plus the uncounted Namophila). Evolution and classification of Lachenalia should be viewed with this hypothesis in mind. Measurements of total chromosome length (or total DNA) would help refine our understanding of which species are polyploid, and hence establish more reliably the ancestral base for the genus. According to our hypothesis of an ancestral base of 10 or 9, species with base numbers lower than n = 10 would be derived and those with base numbers above n = 10 would be polyploid. Little more can be said of the cytological situation here until a molecular-based phylogeny of the genus is available but cytology appears likely to be useful in determining relationships and evolution within Lachenalia.
For the Spetaea/Daubenya clade (Fig. 2), the monospecific Spetaea has n = 10 (Table 3) and a remarkable bimodal karyotype consisting of one pair of very large chromosomes, one medium-sized pair, and the remainder very small chromosomes (Wetschnig and Pfosser, 2003). In Daubenya two of the four species counted have 2n = 32 and the other two have 2n = 34. As in Spetaea, karyotypes are bimodal. The karyotype in the population of D. aurea that we examined (Fig. 2) consisted of three long and 13 very short chromosome pairs (less than one third the length of the long chromosomes) and is clearly not directly polyploid.
Total chromosome length, a proxy for total DNA content, is 230 mm in our preparations in Daubenya aurea compared with 116 mm in Massonia species with diploid numbers of 2n = 18 or 22 (measured in millimetres on metaphase karyotypes using the same preparation method). Provisionally, based on available counts and total chromosome length measurements, it seems most reasonable to conclude that Daubenya is palaeotetraploid. Although this is the most parsimonious conclusion for ploidy level in Daubenya, other explanations cannot be excluded, although we are unaware of any that accord with the data. Accordingly we suggest as a possible hypothesis dysploid reduction in the ancestors of the Daubenya clade to n = 9 or 8, followed by polyploidization and subsequent secondary dysploid reduction to n = 17 and 16. Additional counts for the genus, in which only four of the 10 species have been examined cytologically, are needed to expand and refine our understanding of its cytological evolution.
The phylogenetically isolated genus Eucomis (11 spp.) has x = 15 (Table 3). [If E. autumnalis subsp. amaryllidifolia is recognized as a separate species on the basis of its diploid status in an otherwise tetraploid species, as suggested by Zonneveld and Duncan (2010), the genus has 12 species]. Four species (or five if E. amaryllidifolia is recognized) are exclusively diploid, 2n = 30, and four (or five) are exclusively tetraploid (depending on the status of E. autumnalis subsp. amaryllidifolia). Most counts for E. regia are diploid, 2n = 30, but there is one of 2n = 60 (Riley, 1962) suggesting a polyploid population. Records of 2n = 30-32 may indicate the presence of B chromosomes or merely difficulty in obtaining an exact count. Karyotypes are markedly bimodal, some with three pairs of long acrocentric (macro-) chromosome pairs in diploid species, but the karyotypes of E. bicolor and E. zambesiaca each have a prominent metacentric pair, possibly an indication of unequal reciprocal translocations in the populations examined (Reyneke and Liebenberg, 1980). In an extensive examination of genome size in Eucomis, Zonneveld and Duncan (2010) show that diploid species (chromosome numbers not determined) have genome sizes of between 1C = 10.2-15.1 pg, with E. grimshawii having the loweast values and E. regia the highest in the range. Tetraploid species (again chromosome numbers not determined) have 1C = 20.3-30.2 pg.
We offer the hypothesis that Eucomis is polyploid on a derived, dysploid base. Genome size measurements provide support for the hypothesis of ancestral polyploidy for the genus based on the following argument. Eucomis is nested in a clade with base number x = 10 and diploid species of Eucomis have almost the same 1C values as Veltheimia, also tetraploid (2n = 20). Genome size in these two genera is about three times that in ancestrally diploid Massonia species (mean value for three species examined for genome size is 1C = 3.20 pg). As noted earlier, cross genus comparisons of genome size may not always be reliable indicators of ploidy levels.
In the isolated Ledebouria clade (including Drimiopsis and Resnova) two species have 2n = 20 and three more have 2n = 20 plus other numbers. The L. hyacinthina/revoluta group, including the Madagascan L. nossibeensis, is unusual in having a range of numbers from n = 9-15. Karyotypes in the group are strongly bimodal, e.g. L. nossibeensis (2n = 30) has two large chromosome pairs with a balance of much smaller chromosome pairs, the same pattern reported for L. somaliensis by Stedje (1996). Despite the high chromosome number in these species, however, the karyotypes are not consistent with direct polyploidy. The karyotype of L. urceolata, n = 10, does not exhibit the bimodality evident in species with higher base numbers. According to current data L. humifusa (= Resnova humifusa) has 2n = 10 but there is also a count of 2n = 20 for the species under the synomym Drimiopsis saundersii (Table 3), with a karyotype of four long and six shorter pairs. No other members of the Resnova group have been counted (although L. nossibeennsis was at one time included in Resnova). Polyploidy and dysploidy are frequent in Ledebouria and if species identifications are correct then many are heteroploid. Meiotic studies by Jessop (1972b) are particularly confusingdifferent accessions of some species have a range of haploid numbers but are said to exhibit no meiotic abnormalities. Total DNA per cell (Zonneveld et al., 2005) has been determined for three species, L. cooperi (1C = 5.60 pg), L. petiolata (as Drimiopsis maculata) (1C = 3.75 pg) and L. socialis (1C = 5.85 pg) but these estimates do not include the chromosome number of the samples examined, rendering assessments of ploidy levels impossible.
The diploid count of 2n = 10 in Ledebouria humifusa defies easy explanation. Sometimes segregated in the genus Resnova with several other species (e.g. Lebatha et al., 2006), L. humifusa is deeply nested within the Ledebouria clade in molecular-based phylogenies (Ali et al., 2011;Wetschnig et al., 2007) and a second count of 2n = 20 for the species is consistent with many counts in the remainder of the genus. We cannot discount the possibility that L. humifusa is a dysploid derivative and that the count of 2n = 20 represents neopolyploidy (polyploidy at species rank or lower). The karyotype of the 2n = 20 plants examined by De Wet (1957) could be interpreted as tetraploid: there are 5 pairs of more or less like chromosomes in the haploid karyotype.
Based on the count of 2n = 10 for Ledebouria humifusa Wetschnig and Pfosser (2003) suggested a base number of x = 5 for Ledebouria and other genera of Massonieae. In this scenario counts based on n = 10 would be tetraploid and those with n = 15 hexaploid. Given the karyotypes this seems unlikely: few species, even within Ledebouria, show an expected four or six sets of like chromosomes. Indeed, the karyotype of the monospecific Spetaea (n = 10) includes one long, one medium, and one moderately short chromosome pairs (plus seven pairs of very short chromosome pairs), thus generally inconsistent with polyploidy. If the genus is in fact polyploid then considerable chromosome repatterning must have occurred.
Given the hypothetical base number of x = 10 for Massonieae, Urgineoideae and also Ornithogaloideae, the base in the phylogenetically isolated Ledebouria clade is most parsimoniously inferred to be x = 10, with some species polyploid, heteroploid or dysploid. We suggest that the typical vegetative reproduction common in populations of the alliance makes it possible for chromosomal aberrations (non-coding fragments, B chromosomes) to accumulate, perhaps tolerated in species or populations that reproduce mainly vegetatively. Hence the range of numbers mostly above n = 10, the most common number in the genus, that have been reported, notably by Jessop (1970Jessop ( , 1972aJessop ( , 1972b.

Summary
We infer an ancestral basic chromosome number of x = 10 for Urgineoideae. This is the only base number for almost all species of the two genera in the subfamily, the same base was postulated for Ornithogaloideae (Goldblatt and Manning, 2011), sister to Hyacinthoideae plus Urgineoideae (Manning et al., 2004). We infer the same ancestral base of x = 10 for Massonieae, in part by outgroup comparison (Fig. 1) and because the pattern within the tribe seems to us most consistent with this hypothesis. The striking, decreasing dysploid series in Lachenalia is notable for Massonieae, indeed for all sub-Saharan Hyacinthaceae, and merits detailed investigation in combination with systematic and molecular phylogenetic study. Similar extensive dysploid sequences are known in another geophytic southern African family, Iridaceae, especially in Lapeirousia , Romulea (De Vos, 1972) and Moraea (Goldblatt, 1971(Goldblatt, , 1976(Goldblatt, , 1986. In Hyacinthoideae, Pseudoprospereae evidently has x = 9. Originally placed as one element in a trichomy with Massonieae and Hyacintheae (Wetschnig and Pfosser, 2003;Manning et al., 2004), Pseudopropereae has more recently been resolved as basal in the subfamily, i.e. sister to Massonieae + Hyacintheae (Buerki et al., 2012). All topologies support a postulated ancestral base of x = 9 for Hyacintheae.
Neopolyploidy is relatively uncommon among sub-Saharan Hyacinthoideae. According to available counts, one species of Massonia out of four counted is tetraploid. For Lachenalia, 86 species or almost three quarters of the total in the genus have been counted. Available counts show that 15 species are exclusively polyploid and a further 14 have diploid and polyploid populations. Assuming a total of 11 species in Eucomis, four are tetraploid and two have diploid and tetraploid populations. Among these three genera, just 20 species out of 101 species counted are exclusively polyploid. In contrast, 16 species have diploid and polyploid populations.
The modest level of neopolyploidy in sub-Saharan Hyacinthoideae is consistent with the pattern in Ornithogaloideae (Goldblatt and Manning, 2011) in which just one of 24 species of sub-Saharan Ornithogalum and three of 23 species of Albuca subgen. Albuca are exclusively polyploid. Other families of sub-Saharan geophytes show the same pattern of low polyploid frequency. In Gladiolus (Iridaceae) just five of 70 (7%) sub-Saharan species sampled so far have polyploid populations whereas all six Eurasian species are exclusively polyploid . In Moraea (Iridaceae) the two Eurasian species are tetraploid, only nine of 164 (b 5%) species of sub-Saharan Africa counted are exclusively polyploid and 15 more species have diploid and polyploid populations (Goldblatt, 1976;Goldblatt and Manning, 2011). As in Ornithogalum, polyploidy is relatively frequent in Eurasian genera of Hyacinthoideae, judging from records in the cytological literature.
In contrast to the low frequency of neopolyploidy in African Hyacinthoideae, five of the nine genera of the subfamily in sub-Saharan Africa for which we have chromosome counts appear palaeopolyploid (polyploid at generic rank or higher). All counts for Merwilla, Schizocarphus and Veltheimia are tetraploid, 2n = 40. For Eucomis (x = 15) and Daubenya the best explanation we can offer, based on outgroup comparison and genome size estimates, is that both are hypotetraploid.
The situation in Ledebouria is less clear, given the infraspecific variation in chromosome number in almost all species. Only two counted species have exclusively 2n = 20 and several more have populations evidently with 2n = 20 as well as other numbers.
Numbers not multiples of 10 may represent the presence of B chromosomes but the frequency of plants or populations with dysploid numbers is remarkable. Several Ledebouria species frequently reproduce asexually and some exhibit meiotic abnormalities, rarely producing viable seeds (Jessop, 1972a(Jessop, , 1972b. We suggest that frequent asexual reproduction via bulbils may allow abnormal karyotypes to persist in the wild.