Karyotype characterization and evolution of chromosome number in Cactaceae with special emphasis on subfamily Cactoideae

Cactaceae species are karyotypically well-known with x = 11 and chromosome number variation due mainly to polyploidization. However, both assumptions are based on descriptive observations without taking an evolutionary framework of Cactaceae into account. Aiming to confirm these hypotheses in an evolutionary context, we obtained chromosome numbers for 20 species of Cactoideae, performed an extensive review of chromosome number for the family, and analyzed these data using a phylogenetic approach. The karyotypes presented here were characterized by CMA/DAPI banding, and for six species 5S and 45S rDNA sites were located. Our data, along with a survey of the literature, reinforce the long-standing hypothesis of a x = 11 as the base chromosome number for Cactaceae. They also reinforce the relevance of polyploidy in karyotype evolution of cacti, although polyploidy was important just after the diversification of subfamilies Maihuenioideae and Pereskioideae. Despite the homogeneous chromosome complements observed among cacti, chromosome banding and FISH techniques revealed informative characteristics, allowing the identification of chromosome synapomorphies, such as proximal CMA + bands in Melocactus and proximal 5S rDNA in Pilosocereus, indicating the taxonomic potential of chromosome characterization in cacti.

Most species of Cactaceae have showy flowers and exhibit adaptations to arid environments, the so-called succulent syndrome: shallow roots, thick and waxy cuticle and CAM photosynthesis (Anderson 2001;Arakaki et al. 2011).The family diversified at ≈35 Ma, but the major diversification is more recent, in the Miocene-Pliocene, ≈10-2.5 Ma, a period coinciding with the global evolutionary burst of C 4 photosynthesis (Arakaki et al. 2011).Cactaceae monophyly is supported by morphological traits -inferior ovary, short shoots into areoles and shoot apical meristems organized into four zones -and molecular phylogenetic results (Downie & Palmer 1994;Nyffeler 2002).Two subfamilies, Opuntioideae and Cactoideae, are well characterized by morphological traits and molecular synapomorphies (Nyffeler 2002), but relationships among these two subfamilies and the two remaining genera, Pereskia and Maihuenia, are challenging (Anderson 2001;Nyffeler 2002;Griffith 2004;Edwards et al. 2005;Hernández-Hernández et al. 2011).Among the previous taxonomic proposals, these genera were grouped in Pereskioideae based on the absence of morphological synapomorphies (Nyffeler 2002;Edwards et al. 2005).The two species of Maihuenia were also placed in their own subfamily, Maihueniodeae, which are sister to Cactoideae (Anderson 2001;Nyffeler 2002;Edwards et al. 2005) or sister to both Cactoideae and Opuntioideae (Hernández-Hernández et al. 2011).The best currently available phylogenetic hypothesis supports both genera, Pereskia and Maihuenia, as forming a two-or three-parted grade to the rest of Cactaceae (Hernández-Hernández et al. 2011).
Cactaceae species are well characterized in terms of chromosome numbers.In some specific genera, e.g.Opuntia, a large cytological database was used to understand species evolution and subgeneric delimitation (Pinkava & McLeod 1971;Pinkava et al. 1973;1977;1985;1992;Weedin & Powell 1978;Pinkava & Parfitt 1982;Weedin et al. 1989;Powell & Weedin 2001;2004).However, a complete survey of chromosome data for Cactaceae is still missing, preventing a more comprehensive identification of the main chromosome rearrangements in an evolutionary context.
Symmetrical karyotypes and 2n = 22 are conserved characteristics across the family (Stefano et al. 2011;Majure et al. 2012;Castro et al. 2013;Moreno et al. 2015).Nevertheless, the homogeneous karyotypes of cacti, when submitted to differential staining and rDNA sequence localization, reveal to be variable and typically informative.Pyrrhocactus species, e.g., could be characterized by its variation in pericentromeric CMA + bands (Peñas et al. 2008).The genera Opuntia and Brasiliopuntia also exhibit pericentromeric CMA + bands, whereas the related genus Tacinga is characterized by the absence of such pericentromeric CMA + bands (Castro et al. 2016).Duplicated 5S rDNA sites are characteristic of most Lepismium clades, but this is absent in the more derived clades (Moreno et al. 2015).
Polyploidy and x = 11 in Cactaceae have been hypothesized as the main mechanism and the base chromosome number of karyotype evolution, respectively.However, these two assumptions are based just on the high frequency of n = 11 and multiples of chromosome numbers among Cactaceae species and have never been examined in an evolutionary framework, a defficiency this paper aims to rectify, emphasizing subfamily Cactoideae.To complete our objectives, we first generate new chromosome data for Cactoideae from South America, using classical (chromosome number and CMA/DAPI banding) and molecular cytogenetic (5S and 45S rDNA localization by fluorescent in situ hybridization, FISH) approaches for 20 and six previously unexamined species, respectively.Finally, based on an extensive compilation of Cactaceae chromosome data and close species used as outgroup, we analyzed all data in a phylogenetic context aiming to confirm (1) the base chromosome number for Cactaceae and subfamilies and (2) the main chromosome modification that have occurred in Cactaceae.

Materials and methods
Plant Material -A total of 20 species (Tab. 1) were analyzed, representing ten genera from Cactoideae covering tribes Cereae, Trichocereae and Hylocereae (Anderson 2001).Vouchers have been deposited at EAN.

Cytogenetic analysis -Chromosome number, CMA/DAPI banding, and FISH
Chromosome number -Root tips were pre-treated with 0.002 M 8-hydroxyquinoline (8HQ) for 5 h at 18 °C and subsequently fixed in 3:1 (absolute ethanol: glacial acetic acid, v:v) for 2-24 h at room temperature (RT) and stored at −20 °C.For chromosome slide preparation, material was washed twice in distilled H 2 O (dH 2 O) for five minutes and digested in an enzymatic solution containing 2 % cellulase (Onozuka) and 20 % pectinase (Sigma) for 1 h at 37 °C.Subsequently, they were squashed in a drop of 45 % acetic acid and the coverslip was removed in liquid nitrogen.
CMA/DAPI banding -For each voucher, the three best slides (i.e., slides with at least five well-spread metaphases) were aged for three days and double-stained with the fluorochromes CMA (0.2 mg mL −1 ) for 1 h, and DAPI (2 μg mL −1 ) for 30 min (Schweizer 1976).Slides were mounted in glycerol/McIlvaine pH 7.0 (1:1/v:v) buffer medium.The metaphases were captured on a Zeiss epifluorescent microscope using an AxioCam MRC5 video camera and Axiovision 4.8 software.Images were processed in Adobe Photoshop CS6®.Usually, between 20 and 40 cells were analyzed per voucher.For species with more than one voucher analyzed, and presenting consistent banding pattern, the individual with best slides (i.e., at least ten good metaphases) were chosen, and two slides were destained in 3:1 absolute ethanol: glacial acetic acid (v:v) for 30 min at room temperature, followed by absolute ethanol for 2 h at 10 °C and stored for FISH.

Chromosome number evolution
Following standard conventions, throughout this article we refer to 'x' as the base chromosome number of a clade, taking in account closely related taxa; x 2 , x 3 . . .as the derived secondary, tertiary (. ..) base chromosome number of a clade; 2n as the chromosome number in somatic tissues and n as the haploid chromosome number observed in gametes (Stebbins 1966;1971;Guerra 2008;2012).
Applying the chromosome numbers obtained both here and in the literature and considering the phylogenetic hypothesis of Hernandez-Hernandez et al. (2011) downloaded from TreeBase (https://treebase.org/),we reconstructed the base chromosome numbers for the tree nodes with the ChromEvol v.2 software (http:// www.zoology.ubc.ca/prog/chromEvol.html)(Mayrose et al. 2010;Glick & Mayrose 2014).This software uses likelihood-based methods to evaluate the chromosome number changes along the phylogenetic branches.Such inference employs four parameters: polyploidy (chromosome number duplication, with rate ρ), demi-duplication (fusion of gametes of different ploidy, with rate μ), and dysploidy, which could be ascending (i.e., chromosome gain, with rate λ) or descending (i.e., chromosome loss, with rate δ).These parameters are evaluated with constant models and linear models, which use both constant (λl) and linear (δl) rates for gain and loss of chromosomes.The linear models estimate the chromosome number changes depending on the current chromosome number, whereas the constant models measure the rate of changes independently of the current chromosome number.Finally, eight models (four models considering just constant rates and other four models considering both constant (λl) and linear (δl) rates) were tested, allowing us to infer the main mechanism for chromosome number changes according to the best model.Both model sets include a null model that assumes no polyploidization events.The best-fitted model was selected using the Akaike's information criterion (Akaike 1974), and considering this best-fitting model we re-ran the analysis using 10,000 simulations to infer the expected number of changes along the branches, as well as to infer the ancestral base chromosome number for specific nodes throughout the phylogenetic hypothesis.The haploid chromosome numbers of 70 species were recorded and plotted into a file, representing 31 % of the 226 species present in the phylogenetic analysis published by Hernández-Hernández et al. (2011) (Tab.S2 in supplementary material; unknown chromosome numbers were coded by "x"].

Chromosome markers -FISH
The 5S and 45S rDNA sites were localized in six species, four diploids and two tetraploids.Although position of the 5S rDNA site was an informative chromosome marker, exhibiting variation among the analyzed species (Fig. 3), the position of 45S rDNA was conservative among species and between ploidies -two sites in diploid and four sites in polyploid species/cytotypes.
The diploid Discocactus zehntneri (Fig. 2L) had one pair of 5S rDNA site, and Pilosocereus pentaedrophorus (Fig. 2H) exhibited two pairs of 5S rDNA sites, one proximal and another adjacent to 45S rDNA, whereas the two other diploid species had 5S rDNA duplications (two sites on the same chromosome): (1) Cereus jamacaru (Fig. 1E) with four pairs of 5S rDNA loci -one proximal pair, one subterminal pair, plus one chromosome pair at a proximal and an interstitial site on one chromosome arm; (2) Pilosocereus chrysostele (Fig. 2B) with two pairs of 5S rDNA sites on the chromosome bearing the 45S rDNA site -one 5S rDNA pair is adjacent to the 45S rDNA and another is interstitial (detail in Fig. 2B), plus an additional proximal 5S rDNA pair observed on another chromosome pair (see asterisks in Fig. 2B).The two polyploid species, P. pachycladus.subsp.pernambucoensis (Fig. 2F) and M. oreas (Fig. 1K), also presented variation regarding the 5S rDNA sites; whereas the former had two pairs of proximal 5S rDNA sites plus two interstitial pairs adjacent to the 45S rDNA (four pairs in total), the latter exhibited only two proximal pairs that were co-localized with the CMA + / DAPI -heterochromatic bands -a typical characteristic of M. oreas (compare Fig. 1J and K, see Fig. 3).

Literature survey
The chromosome number survey showed polyploid counts for 31.32 % of Cactaceae species, whereas dysploidy was poorly represented, occurring in just 6.65 % of 677 available species (Tabs. 2 and S1 in supplementary material).Polyploidy was concentrated in Cactoideae (20 %) and Opuntioideae (65.5 %), absent in Pereskia and with a single record in Maihuenia (one species with a polyploid cytotype) (Tab.2).However, even in Cactoideae and Opuntioideae, polyploidy is not evenly spread and is especially concentrated in some tribes, such as Cereeae (19 of 42 species, 45.24 %) and Trichocereeae (40 of 99 species, 40.40 %) (Tab.2).The high frequency of polyploidy observed in Opuntioideae (65.5 %) is a consequence of polyploid records in Opuntia and Cylindropuntia, the two genera with the greatest number Table 2. Chromosome number survey in Cactaceae.Species are presented following the taxonomic proposal of Anderson (2001).
The number of species with available chromosome number is presented followed by the number of species with polyploidy (polyploid species plus species with polyploid cytotype) and dysploid counts.For each tribe/subfamily and for the whole Cactaceae, the number of taxonomic groups analyzed is provided, with the total number of polyploid and dyploid species.The percentage of polyploidy and dysploidy are presented into brackets.of species analyzed and also an elevated frequency of polyploidy (65 of 94 and 18 of 34 species, respectively) (Tab.S1 in supplementary material).

Chromosome number evolution and base chromosome number in Cactaceae
Based on the chromosome numbers obtained here in addition to those from the literature and an available phylogenetic hypothesis for Cactaceae (Hernandez-Hernandez et al. 2011), we performed an analysis for chromosome number evolution and determination of the base chromosome number for Cactaceae.Here, the ChromEvol analysis highlights the importance of polyploidy for cacti.Eight models were tested, and two of them did not consider polyploidy, working as a null hypothesis for this event.These two models, Constant rate no-dupl and Linear rate no-dupl, presented an elevated AIC value compared with the six remaining models that do consider polyploidy (compare the AIC for constant no-duplication, 510.9, and for linear rate no duplication, 467.7, with remainder models; see Tab. 3), reinforcing the importance of polyploidy.
The best model, estimated by the lowest AIC value, was 'Constant rate demi-est' (Tab.3), a model that considers (1) gain and loss of chromosomes (ascending and descending dysploidy), (2) duplication (polyploidy; i.e., a multiplication of the chromosome number by a factor of 2×) and (3) demiduplications (demi-polyploidy; i.e., a multiplication of the chromosome number by a factor of 1.5× caused by fusion of gametes with different ploidies) as causes of chromosome number variation.In this model, the rate for polyploidy and demi-polyploidy were considered as independent rates (ρ is independent of μ), and for cacti, the expectation values were different for ρ and μ (Tab.3).Constant rate models consider that all rates were constant, i.e., the changes were independent of current chromosome numbers.
Considering the base chromosome numbers, the same result was obtained in the six tested models that include polyploidy (data not shown): x = 11 was estimated for Cactaceae, as well as Cactoideae, Opuntioideae and Pereskia and Maihuenia (the non-monphyletic subfamily Pereskioideae) (Fig. 3).Both Pereskia-as ancestral and Pereskia-as-derived hypothesis would give the same result due to the stability of chromosome number in Cactaceae.
Gain and loss of chromosomes (ascending and descending dysploidies) are restricted to the outgroup -a gain in Talinum paniculatum and loss in Grahamia bracteata and Portulacaria afra, whereas polyploidy is widespread in Cactaceae (Fig. 3).

Chromosome number and the importance of polyploidy
For three taxa, chromosome numbers were described here for the first time -Arrojadoa dinae and Pilosocereus pachycladus subsp.pachycladus and subsp.pernambucoensis.For the other 18 species, our new chromosome numbers agree with previously published data (Darlington & Wylie 1955;Pedrosa et al. 1999;Goldblatt & Johnson 2003;2006;Castro et al. 2013;Peñas 2018; Tab.S1 in supplementary material), except for M. zehntneri and M. ernestii, which had polyploid karyotypes in our accessions, 2n = 44, but diploid Table 3. Summary of the eight ChromEvol models for the used phylogenetic tree.The Log-likelihood and the AIC values are presented, and the lowest values are indicated in bold and underlined.The best model, constant rate demi-polyploidy est, is indicated by the * and this model was re-run under Optimize Model option with 1,000 simulations.The expectation values for gain, loss, duplication, and demi-duplication under constant rate demi-polyploidy-est are presented, indicating that polyploidy is the more frequent form of chromosome number change.Karyotype characterization and evolution of chromosome number in Cactaceae with special emphasis on subfamily Cactoideae counts, 2n = 22, in a previous study (Assis et al. 2003).Such chromosome number variation represents polyploid populations in a single species, a frequent phenomenon in Cactaceae (Pinkava & McLeod 1971;Pinkava et al. 1973;Castro et al. 2013;2016).Polyploidy has been long considered one of the main evolutionary phenomena in plants (Stebbins 1971;Grant 1981), and this subject still being reviewed to evaluate its importance in angiosperm diversification (Peer et al. 2017;Rice et al. 2019;Magallón et al. 2019).Diversification of Cactaceae fits this model well, and polyploidy is more frequent after Maihuenioideae and Pereskioideae diverged, i.e., in the called 'core cacti'.This group diversified at ≈25 Ma, with subsequent increase in diversification rate (Arakaki et al. 2011), which seems to occur concomitant with an increased frequency of polyploidy.In a similar way, many families and genera display an association between species-richness and occurrence of polyploidy (Soltis et al. 2009), reinforcing the hypothesis that polyploidy has been important for angiosperm diversification.Among Cactaceae, the well-studied genus Opuntia (Opuntioideae) is one of the largest in the family, with 226 accepted species (following The Plant List 2019), and also present high frequency of polyploidy (70 % of species are polyploid or have some polyploid cytotypes).Nevertheless, this assumption could be biased by study effort more than a positive relationship between polyploidy and diversification.In this sense, the efforts should now be concentrated in the poorly studied groups with recent diversification, such as Cereinae (tribe Cereeae, diversification ca.7.5 -6.5 Ma; Arakaki et al. 2011), to achieve a more realistic scenario for the effect of polyploidy on cacti evolution and its role in diversification.
The recent polyploidy observed in Cactaceae has frequently been hypothesized to involve interspecific and intergeneric hybridization (Baker et al. 2009).For example, it is assumed that the morphological variation observed in Melocactus (Zappi 1994;Anderson 2001) could be related to hybridization and polyploidy, as observed in the sympatric zone of Melocactus paucispinus and M. concinnus where there have been hybrids detected with variable morphologies (Lambert et al. 2006).Here, polyploidy was detected also in sympatric populations of Pilosocereus pachycladus subsp.pachycladus (diploid and polyploid cytotypes) and subsp.pernambucoensis (all populations are polyploid).Both cytotypes of P. pachycladus subsp.pachycladus are terrestrial and occur in specific vegetation locally known as carrasco -arborescent semi-arid vegetation in northeastern Brazil.The polyploid P. pachycladus subsp.pernambucoensis occurs in Bahia State in populations sympatric with P. pachycladus subsp.pachycladus, where they could hybridize and expand to the remaining states of northern Brazil, always colonizing extremely xeric vegetation similar to carrasco (Zappi 1994).The wider geographic distribution of the tetraploid subsp.pernambucoensis reflects the positive effect of polyploidy, increasing the genetic variability and putative allowing it to colonize a wide range of extreme habitats (Leitch & Leitch 2008;Fawcett & Peer 2010;McIntyre 2012;Theodoridis et al. 2013;Ramsey & Ramsey 2014;Oberlander et al. 2016).

Chromosome markers -Heterochromatic bands and rDNA loci
The presence of both terminal and proximal CG-rich fluorescent bands detected here seems to be a common characteristic among Cactoideae species (Peñas et al. 2008;2009;2011;Moreno et al. 2015).Pereskia did not have proximal CMA + bands (Peñas et al. 2014;Castro et al. 2016), suggesting that the amplification of CG-rich heterochromatin occurred in the common ancestor of Opuntioideae and Cactoideae.
Variation in heterochromatic bands has long been used in karyotypic characterization among species, e.g.Orchidaceae (Koehler et al. 2008;Moraes et al. 2016;2017), as well as, among varieties, as in species of Citrus (Guerra 1993;Yamamoto & Tomiaga 2003;Moraes et al. 2007).Independent of the taxonomic group, there is a tendency for heterochromatin to be CG-rich and not ATrich.It is also true for Cactaceae, with only one report of AT-rich heterochromatin in Pyrrhocactus from Argentina (Peñas et al. 2008).The CMA + band pattern proved to be taxonomically informative in cacti, introducing useful chromosome markers in the stable karyotypes typical of Cactaceae.For example, in Melocactus two species, M. ernestii and M. oreas, have long spines but distinguishing these two species based on morphology is a challenging task.However, they can be differentiated by the number of CGrich heterochromatin blocks.In Arrojadoa, the proximal CMA + bands are potentially informative for this small genus endemic to northern Brazil, differentiating taxa with heterochromatic bands (here, A. dinae and A. penicillata) from taxa without proximal bands (A.rodantha).In this sense, when studying a group with stable chromosome numbers such as cacti, the use of additional chromosome markers is mandatory, allowing us to distinguish species and supporting taxonomy studies, in addition to improving the understanding of chromosome evolution.
Fluorescent in situ hybridization provides additional chromosome markers for these analyses.The the physical mapping of 45S rDNA exhibited a conserved pattern, with the number of sites strictly correlated with species ploidy: two sites in the diploid species, four sites in the tetraploid species (Peñas et al. 2009;Moreno et al. 2015;Castro et al. 2016; result presented here).All 45S rDNA sites were terminally localized and co-localized with CMA + bands, in agreement with the most common situation observed in plants (Lima-de-Faria 1980;Roa & Guerra 2012).In contrast, 5S rDNA sites were variable in number and position, occupying proximal and interstitial positions and occasionally adjacent to the 45S rDNA sites, as observed in other Cactaceae such as Lepismium (Moreno et al. 2015), or located on different chromosomes.As observed in M. oreas, 5S rDNA could also be detected as a heterochromatic block.The 5S consists of a transcriptional unit of 120 nucleotides separated from the next copy by a non-transcribed spacer (NTS).Although NTS varies in length and sequence, with a rapid rate of concerted evolution, the 5S exon is highly conserved (Rebordinos et al. 2013).Changes in the NTS region could involve shifts to AT-rich or CG-rich, and due to amplification it can be detected as a heterochromatic block by chromosome banding, whereas the remaining sites in the same karyotype and closely related species are not identified as AT-or CG-rich.In addition, 5S rDNA is known to be part of a multigene family that could be linked to different multigene families, e.g.transposable elements (TE) as demonstrated in animals (Rebordinos et al. 2013) and plants.For example, the Cassandra element, a retrotransposon associated with 5S rDNA in plants could facilitate the movement and creation of additional 5S rDNA sites throughout the genome (Kalendar et al. 2008).Hybridization also could favor TE-movement, a consequence of which is dynamism of 5S rDNA sites associated with TEs, as detected in Aegilopsis (Raskina et al. 2004) and some fish species (Fontdevila et al. 2005).It is possible that in Cactaceae the combination of hybridization and polyploidy favors the dynamic movement of retrotransposons, creating new sites with or without a relationship to ploidy, as observed with the 45S rDNA here.
Finally, the diversity of 5S rDNA sites also highlights the importance of structural chromosome rearrangement, such as inversions.This phenomenon could be responsible for the creation of two sites in the same chromosome arm, as observed in both Cereus jamacaru and Pilosocereus chrysostele.It is possible that a break point could have occurred inside the original 5S rDNA site (an event putatively favored by the TE activity), and after an inversion, inserted some copies of 5S rDNA, creating a new site while retaining some copies at the original site.Such an event has previously been suggested for some groups of plants, for example, different unrelated species of Orchidaceae (e.g., see Moraes et al. 2012;2017;Lee et al. 2017) and Nothorcordum (Amaryllidaceae; Souza et al. 2012).In this sense, evolution of 5S rDNA sites in cacti contrasts to evolution of 45S rDNA, with 5S being more variable than 45S, which is the opposite of the commonly accepted hypothesis that position and number of 5S rDNA loci in plants are usually more conserved than those of 45S rDNA loci (Roa & Guerra 2012;2015).

Conclusions
Except for Pereskia, in which there have been no records of polyploidy, this event seems to be an important evolutionary mechanism in Cactaceae, commonly associated with natural hybridization.The use of karyotype characterization in combination with phylogenetic analyses provides a useful tool in chromosome evolution studies.In our analyses, x = 11 was confirmed, and the important role of polyploidy in chromosome evolution in our results confirms the traditional assumptions about Cactaceae.Heterochromatin distribution and 5S rDNA positions were both considered as good cytotaxonomic markers in Cactoideae, mainly in Arrojadoa, Melocactus and Pilosocereus, and are likely to be useful in other genera as well.For the last two genera, both with many taxonomic problems due to the recurrent interspecific hybridization, karyotype characterization based on CMA + bands and 5S rDNA appears to be promising tools to aid in species delimitation.

Table 1 .
Cactoideae species collected for cytogenetic analysis, with respective voucher and provenance.Somatic chromosome numbers (2n), CMA + bands, and number of 5S and 45S rDNA site pairs are summarized.t = terminal position; st = subterminal position; p = proximal position; i = interstitial position.