Genome size estimation

Nuclear DNA C-values ( = holoploid genome sizes) and Cx-values ( = monoploid genome sizes) were estimated using propidium iodide flow cytometry (FCM). Sample preparation generally followed the two-step procedure originally described by Otto (1990). Plants were cultivated for at least 1 year under homogeneous conditions. About 1 cm² of young and intact fresh leaf tissue and internal standard was co-chopped in a sandwich-like arrangement with a sharp razor blade in 1 mL of ice-cold Otto I buffer (0·1 m citric acid, 0·5 % Tween 20). The nuclear suspension was filtered through a nylon mesh (42-μm pore size) and centrifuged at 150g for 5 min. The supernatant was then removed and nuclei were gently resuspended in 100 µL of fresh Otto I buffer. After incubation (15 min at room temperature), 1 mL of Otto II buffer (0·4 m Na₂HPO₄.12H₂O) supplemented with propidium iodide (at a final concentration 50 µL mL^–1), RNase IIA (50 µL mL^–1) and 2-mercaptoethanol (2 µL mL^–1) was added. The samples were incubated for 30 min at room temperature, after which fluorescence intensity of 5000 particles was recorded on a Partec Cyflow instrument (Partec GmbH, Münster, Germany) equipped with a 532-nm solid-state laser (Cobolt Samba 100 mW, Cobolt, Sweden). Each plant was re-analysed at least three times on different days and only histograms with peaks of approximately the same height were accepted. If between-day variation (max./min. value) exceeded 2 %, the outlying value was discarded and the sample re-measured. Glycine max ‘Polanka’ (2C = 2·50 pg; Doležel et al., 1994) was selected as a primary internal reference standard. Solanum lycopersicum L. ‘Stupické polní tyčkové rané’ (2C = 2·11 pg) and Bellis perennis L. (2C = 3·96 pg) were used as secondary reference standards for Curcuma samples with low and high genome sizes, respectively, in order to minimize standard-to-sample peak ratio and thus avoid potential non-linearity of FCM measurements. Genome sizes of secondary reference standards were calibrated against the primary one, based on nine replications on different days. The total number of flow cytometric measurements for Curcuma was 678.

Chromosome counts

Chromosome numbers were counted in actively growing root tips of the cultivated plants. Samples were pretreated with a saturated solution of p-dichlorbenzene (3 h, room temperature), fixed in a 3:1 mixture of ethanol and acetic acid (4 h, 4 °C), macerated in 1:1 hydrochloric acid/ethanol (30 s, room temperature) and immediately squashed in a drop of lactopropionic orceine. The number of chromosomes was determined in 5–10 complete well-spread mitotic plates using a Carl-Zeiss Jena NU microscope equipped with an Olympus Camedia C-2000 Z camera.

Basic chromosome number and ploidy level variation

Since the pioneering karyological surveys, there has been continuous dispute concerning the basic chromosome number in Curcuma. Investigating 25 Zingiberaceae species, including three curcumas, Raghavan and Venkatasubban (1943) were the first to report x = 21. Sato (1948) suggested x = 8 in his early work on C. longa, but later claimed that two basic numbers (x = 7 and x = 8) occur in the genus (Sato, 1960). Further support for x = 21 appeared in the study of Ramachandran (1961), who observed regular bivalent formation during meiosis in C. decipiens (2n = 42) and a high frequency of trivalents in C. longa (under the name C. domestica) with 2n = 63. In line with Venkatasubban (1946), Ramachandran considered x = 21 a secondary number possibly derived from a combination of x = 9 and x = 12, which are common in several genera of Zingiberaceae. Yet another basic chromosome number, x = 16, has been proposed and adopted by some researchers (e.g. Sharma and Bhattacharya, 1959). However, x = 21 became the most commonly accepted basic chromosome number in Curcuma (Ramachandran, 1961; Prana, 1977; Prana et al., 1978; Ardiyani, 2002; Islam, 2004).

It should be noted that the lower alternative, x = 7, is not in conflict with the large majority of published somatic counts. Actually, this value fits better with the range of chromosome numbers currently known (see our prime count 2n = 77). We therefore believe that x = 7 should be considered a primary basic chromosome number, at least for the majority of Indian Curcuma species (belonging to the subgenus Curcuma). Grant (1982) regarded x = 7 and 10 as the most common basic chromosome numbers in monocots so the present findings are in concordance with a large body of evidence. Following this calculation, species investigated here correspond to 6x, 9x, 11x and 15x cytotypes. In addition, dodecaploid (2n = 84) plants from subgenus Curcuma were reported by earlier researchers (Table 1).

It appears that there are some Curcuma species in which somatic chromosome numbers (2n = 20, 22, 24, 32, 34, 36 and 38) do not fit into the series based on x = 7. However, they all belong to the subgenus Hitcheniopsis, which encompasses mainly Thai species such as C. alismatifolia, C. gracillima, C. harmandii, C. parviflora, C. rhabdota and C. thorelii (see Table 1), and which differs markedly in morphological traits from the nominate subgenus. From the present species set, only C. vamana (2n = 22) belongs to subgenus Hitcheniopsis. We are convinced that this species is only distantly related to the other investigated taxa (subgenus Curcuma), and its basic chromosome number is most plausibly x = 11, as also documented for several other genera of Zingiberaceae (e.g. Kaempferia, Zingiber and Renealmia).

It is plausible that both subgenera Curcuma and Hitcheniopsis represent independent evolutionary units with distinct basic chromosome numbers. In addition, x = 7 may be considered a subgenus-specific cytological marker for the former.

Inter- and intraspecific variation in genome size

The taxa involved in the present study differed 2·87-fold in their holoploid genome sizes (2C = 1·66–4·76 pg). This range is slightly below the mean for other plant genera (3·28-fold based on data from the Plant DNA C-values database; Bennett and Leitch, 2005a). There was no clear gap in nuclear DNA C-values between hexaploid and nonaploid plants and a major discontinuity in these cytotypes actually occurred between two groups of hexaploids (Fig. 4). These results indicate that DNA content alone is often insufficient to distinguish between 6x and 9x cytotypes, and FCM measurements should always be accompanied by chromosome counts. The C-values of diploid and high polyploid taxa were more distinct, but again FCM measurements alone may not be an accurate indicator of ploidy level, as illustrated by 11-ploid C. oligantha and 15-ploid C. raktakanta with reverse total amounts of nuclear DNA (Fig. 4).

By contrast, homoploid genome sizes formed three well-defined clusters with significantly different values, namely a genome group I (1Cx = 0·30–0·33 pg; 29 taxa), a genome group II (1Cx = 0·36–0·39; 13 taxa) and a genome group III (1Cx = 0·41–0·43 pg; five taxa). We assume that these groups may have different evolutionary histories, as also supported by their distinct distribution patterns (see below).

Intraspecific genome size variation was low in most taxa in which multiple individuals were analysed. However, values above 4 % (i.e. beyond potential instrumental fluctuation) were encountered in five species. When C. longa with about 1·15-fold divergence in C-value is excluded, the extent of intraspecific variation was comparable in all cytotypes: hexaploids, 0·0–6·1 %; nonaploids, 0·5–4·9 %; and a 15-ploid, 5·1%. Intraspecific variation in nuclear DNA content has been controversial since the early studies on genome size. This debate has been fuelled by numerous early reports of intraspecific variation, many of which were dismissed in subsequent investigations using the best practice methodology (Greilhuber, 2005). Several sources of artefactual variation have been identified: (1) instrumental or methodological errors; (2) disturbing effects of secondary metabolites with potential seasonal fluctuation (e.g. Walker et al., 2006); (3) differences in measurements among different laboratories (Doležel et al., 1998); and (4) taxonomic heterogeneity of the material under investigation (Murray, 2005). As a result, the concept of genome size stability has gained broader support. Examples have nonetheless been accumulated in recent years of some genome size variation in FCM assays despite meticulous methodology (Obermayer and Greilhuber, 2005; Šmarda and Bureš, 2006). Several Curcuma species are known to contain, particularly in rhizomes, phenolic secondary metabolites such as curcumins (Lubis, 1968; Prana, 1977). Although phenolics may bias genome size estimates (Greilhuber, 1998; Walker et al., 2006), we believe that our FCM measurements are accurate for several reasons. Low coefficients of variation were achieved, which are not compatible with the presence of interfering metabolites. Nuclear suspensions were clear, lacking any coatings of debris that might indicate a negative effect of metabolites (Loureiro et al., 2006). Repeated measurements of the same sample yielded nearly identical genome size values (average difference 0·8 %) even if particular FCM analyses were performed in different years. Between-plant differences remained stable also in studies using DAPI (an AT-selective fluorochrome less sensitive to secondary metabolites than the intercalating propidium iodide). Also, a narrow species concept was adopted to guarantee taxonomic homogeneity of the plant material. Moreover, co-processed samples with different genome sizes always gave two distinct peaks, which is the most convincing evidence for genuine differences in DNA content (Greilhuber, 2005).

The reasons for intraspecific genome size variation in Curcuma remain unknown. Aneuploidy or presence of B-chromosomes may lead to heterogeneity in nuclear DNA amount, but this explanation seems rather unlikely as only euploid numbers were revealed in our study and, more importantly, two accessions of C. longa with more than 9 % genome size variation possessed the same number of chromosomes (see Table 2). Plausibly, intraspecific variation may be related to a long-term cultivation and targeted selection of desirable genotypes in several Curcuma species, C. longa in particular. India is responsible for around 90 % of turmeric production worldwide, and this species has been widely used in India since Vedic times. Perhaps the variation may have adaptive value, as previously documented in another crop, Zea mays (Rayburn and Auger, 1990). Murray (2005) argued that intraspecific genome size variation may indicate incipient speciation; the blurred species boundaries in several Curcuma alliances may support such a hypothesis. A third explanation takes into account heterochromatic polymorphism. In vegetatively propagating lines, strains with smaller and larger telomeric or centromeric heterochromatin regions may occur (J. Greilhuber, University of Vienna, Austria, pers. comm.) leading to greater genome sizes.

Comparison of genome sizes in Curcuma with those in other members of Zingiberaceae and other monocots

More than three-quarters of the taxa (39 out of 51) in the present study possessed very small genomes defined as 1C ≤ 1·40 pg (Leitch et al., 1998), whereas the remaining 12 taxa had small genomes (i.e. 1C = 1·41–3·50 pg). Low nuclear DNA content in Curcuma spp. is in line with previous records for other members of Zingiberaceae and related families. The Plant DNA C-values database (Bennett and Leitch, 2005a) includes measurements for three members of Zingiberaceae with 1C-values varying from 1·30 pg in Curcuma zantorrhiza (misspelled as C. xanthorrhiza) to 6·03 pg in Zingiber officinale. Very small genomes were also revealed in all but one (Lowiaceae with 1C = 3·55 pg) families from the order Zingiberales, including Marantaceae (14 species, 1C = 0·33–0·68 pg), Musaceae (six species, 1C = 0·58–0·61 pg), Heliconiaceae (1C = 0·45 pg), Strelitziaceae (1C = 0·58 pg), Cannaceae (1C = 0·72 pg) and Costaceae (1C = 1·00 pg).

Although only one Curcuma record is included in the Plant DNA C-values database, more genome size estimates have been published (Table 1). Das et al. (1999) used Feulgen densitometry to analyse three Curcuma spp.: C. caesia (2n = 22, 4C = 3·120 pg), C. amada (2n = 40, 4C = 4·234 pg) and C. longa (2n = 48, 4C = 5·100–5·263 pg). Curcuma longa was also investigated by Nayak et al. (2006) who reported 4C-values from 4·30 to 8·84 pg (i.e. 2·06-fold variation) in 17 cultivars. However, both these studies come from the same laboratory and we believe the results should be treated with caution. First, chromosome counts have not been confirmed by other researchers. Moreover, the authors used hot hydrolysis, which is strongly discouraged owing to the very short optimum treatment that is difficult to control (Greilhuber, 2005). No fixative solution was specified, and if a standard acetic acid/alcohol fixation was applied to plant samples rich in phenolics, stoichiometry of the Feulgen staining was likely to be distorted (the so-called ‘self-tanning’ effect; Greilhuber, 1986, 1988). Feulgen methodology is also not recommended for conventional chromosome counting in Curcuma as it does not allow good spreading of the root tips issue on slides (Islam, 2004) and gives rather pale staining. By contrast, the FCM data published for Curcuma spp. collected in Bangladesh by Islam (2004) seem to be accurate, at least from the methodological point of view. Unfortunately, genome size estimates were not linked to a particular herbarium voucher, which lessens their value. We had the opportunity to examine several well-prepared herbarium specimens prepared from Islam's living material and encountered some misidentifications (e.g. for C. aeruginosa and C. zedoaria).

In comparison with Islam's work (Islam, 2004), our estimates for ten species in common were on average 14 % lower, whereas we determined genome size to be about 11 % higher in C. zantorrhiza than that reported by Bharathan et al. (1994). Plausibly, different reference standards could be responsible for such divergences (i.e. Raphanus sativus – Islam, 2004; chicken red blood cells – Bharathan, 1994).

Species distribution and origin of high polyploids with respect to homoploid genome size

Excluding the distantly related C. vamana, species with the lowest number of chromosomes (2n = 6x = 42) clustered in three groups with significantly different Cx-values (Table 3), which also showed a distinct geographical pattern (Fig. 1). Group I (1Cx = 0·30–0·33 pg) contained ten taxa, the majority of which occurred in north-east India, group II (1Cx = 0·36–0·39 pg) contained ten taxa with a centre of distribution in the Western Ghats (west and south-west India) and Central India and group III (1Cx = 0·41–0·43 pg) contained four taxa occurring either in south-west or north-east India. A link between genome size and geographical location was also confirmed by correlation analysis, which revealed highly significant negative relationships between Cx-values and both latitude and longitude of the localities.

Homoploid genome sizes of all but one (C. oligantha) high-polyploid taxa matched perfectly the Cx-values of hexaploids from group I. This allows us to hypothesize that these hexaploids have played a key role in the evolution of polyploidy in Indian Curcuma spp. Nonaploid cytotypes probably originated by a fusion of reduced and unreduced gametes of hexaploids, either within or between species, giving rise to auto- or allopolyploids. Within Zingiberaceae, regular production of gametes with the somatic number of chromosomes has been reported in the genus Globba (Takano and Okada, 2002), which shows similar ploidy composition to Curcuma. Both nonaploid (unreduced gamete) and hexaploid (reduced and unreduced gamete) plants were probably involved in the genesis of rare dodecaploid and 15-ploid taxa. The vast majority of nonaploid species undergo asexual propagation via rhizome branching and produce a high percentage of aborted pollen grains (Prana, 1977; Nasir Uddin, 2000). It is plausible, however, that even a small fraction of viable pollen may eventually lead to the formation of higher polyploids. The origin of 11-ploid C. oligantha cannot be explained with certainty given the current stage of knowledge. Homoploid genome size (1Cx = 0·43 pg) of this species fits in the range of a genome group III.

Genome size and chromosome numbers as supportive markers for delimitation of Curcuma species

Chromosome numbers and/or ploidy levels have long been utilized as an efficient taxonomic marker, helping to delimit boundaries between various taxonomic categories or reveal cryptic taxa (Stace, 2000). In addition, the last decade has seen significant progress in the application of FCM to exploit differences in nuclear DNA content, which can in some cases be used as a supportive marker for distinguishing between closely related taxa at both heteroploid and homoploid levels (e.g. Mahelka et al., 2005).

As Curcuma is a taxonomically challenging group, the usefulness of genome size and chromosome counts as supportive markers for taxon determination was assessed. Although the low C-value variation (2·87-fold) and the small number of species-specific DNA amounts reduce their utility, some alliances with morphological similarities have been identified in which cytogenetic data may nonetheless support taxonomic decisions. For example, C. oligantha (described from Sri Lanka) and C. cannanorensis (described from Kerala, western coast of South India) have traditionally be merged as one taxon due to their overall phenotypic resemblance (Bhat, 1987; Mangaly and Sabu, 1993; Sabu, 2006). However, the present examination of living plants in natural conditions revealed constant, though minor, morphological divergences (see Fig. 5), along with differences in ecological requirements. Distinctiveness of both species was further corroborated by distinct chromosome numbers (2n = 77 in the former, 2n = 42 in the latter). Similarly, it was found that C. aromatica is a species complex comprising three distinct morphotypes with non-overlapping genome sizes, including hexaploids (2C = 1·86 pg) from south India, nonaploids (2C = 2·83 pg) from north-east India and nonaploids (2C = 2·68 pg) from eastern India. Taxon descriptions and clarification of the nomenclatural issues are in progress. Curcuma reclinata (2C = 2·29 pg), C. decipiens (2C = 2·35 pg) and C. inodora (2C = 2·29 pg) are difficult to define morphologically and have more or less overlapping distributions. Cytogenetic data do not provide reliable data for species-level determination and we believe that they should be merged as one taxon (i.e. C. reclinata Roxb.).

A long-term stay of J.L.-Š. at the Calicut University, Kerala, India, was supported by fellowships from the Indian Council of Cultural Relationships and the Ministry of Education, Youth and Sport of the Czech Republic. We thank the authorities of the Ministry of Environment and Forests, Government of India, and the Botanical Survey of India for granting collecting permits and permits to access Indian herbaria. We are also grateful to Johann Greilhuber (Vienna) for critical comments on intraspecific genome size variation, and Ilia Leitch (Kew) for providing some literature. We thank the handling editor Michael Fay (Kew) and two anonymous reviewers for useful remarks on the manuscript. This project was supported by the Grant Agency of the Academy of Sciences of the Czech Republic (project no. B6407401), the Department of Science and Technology, Government of India (Order No. SP/SO/A-20/99 dt. 9·11·2001), the European Commission's Research Infrastructure Action via the SYNTHESYS Project (GB-TAF-606), the Ministry of Education, Youth and Sport of the Czech Republic (MSM 0021620828) and the Academy of Sciences of the Czech Republic (AV0Z60050516).