Chromosome Number and Genome Size Evolution in Brasolia and Sobralia (Sobralieae, Orchidaceae)

Despite the clear circumscription of tribe Sobralieae (Orchidaceae), its internal relationships are still dubious. The recently delimited genus Brasolia, based on previous Sobralia species, is now assumed to be paraphyletic, with a third genus, Elleanthus, nested in it. The morphology of these three genera is significantly different, indicating the necessity of new data for a better genera delimitation. Though morphology and molecular data are available, cytogenetics data for Sobralieae is restricted to two Sobralia and one Elleanthus species. Aiming to evaluate the potential of cytogenetic data for Brasolia-Elleanthus-Sobralia genera delimitation, we present chromosome number and genome size data for 21 and 20 species, respectively, and used such data to infer the pattern of karyotype evolution in these genera. The analysis allowed us to infer x = 24 as the base chromosome number and genome size of average 1C-value of 5.0 pg for the common ancestor of Brasolia-Elleanthus-Sobralia. The recurrent descending dysploidy in Sobralieae and the punctual genome upsize suggest a recent diversification in Sobralieae but did not allow differing between Brasolia and Sobralia. However, the basal position of tribe Sobralieae in the subfamily Epidendroideae makes this tribe of interest to further studies clarifying the internal delimitation and pattern of karyotype evolution.


Introduction
Since Ruiz and Pavón established the genus Sobralia (Orchidaceae) in 1794 [1], its species reached about 200 taxa distributed from southern Mexico to Bolivia and Brazil, with a remarkable species diversity in the Andes. The species are often terrestrial, rarely epiphytic, usually tall, or very tall (for example, stems of Sobralia luerorum Dodson can reach up to 3.5 m height), with slender, cane-like stems growing in clumps and strongly developed root systems. The inflorescence position and structure have served as a basis for the infrageneric classification of Sobralia [2], with five recognized sections [2,3].
The nominal section was characterized by lateral or rarely terminal inflorescences with branching, well-developed raceme and relatively small floral bracts compared to the size of the ovary [2]. The section Racemosae Brieger, despite terminal inflorescences, could differ from the former by its elongated and unbranched inflorescences with large floral bracts. Section Globosae Brieger is composed of small plants with narrow leaf blades, small flowers positioned in terminal and condensed inflorescences (shortened internodes differ from the former by its elongated and unbranched inflorescences with large floral bracts. Section Globosae Brieger is composed of small plants with narrow leaf blades, small flowers positioned in terminal and condensed inflorescences (shortened internodes hidden under the floral bracts) that successively produce a single flower at a time and elongate with successively produced floral bracts. Species from section Abbreviatae Brieger share terminal and condensed inflorescences with the previous section but, instead, present floral bracts forming a cone. The fifth section, Intermediae Brieger, was established for a single taxon-Sobralia fragrans Lindl.-to emphasize its elongated basal internode of the inflorescences. Dressler [3] enlarged this section, placing other species with small flowers and inflorescences ( Figure 1).  Molecular phylogenetic studies [4,5] revealed that the nominal section of Sobralia is paraphyletic, indicating the necessity of taxonomic reassignment. Based on the morphological distinctness and the mentioned phylogenetic relations, the section Sobralia was elevated to the status of a separate genus named Brasolia (Rchb.f.) Baranow, Dudek and Szlach [4]. Dressler et al. [6] already had predicted the necessity of such reassignments, proposing to conserve the name Sobralia with Sobralia biflora Ruiz and Pav. as a nominal species to avoid transferring most of the genus representatives into Cyathoglottis Poepp. & Endl., the second-oldest generic name available for Sobralia. Thus, the former nominal section of the genus Sobralia now constitutes a separate genus renamed Brasolia, with the former section Abbreviatae being considered the Sobralia nominal section [4]. Nevertheless, the genus Elleanthus C. Presl was nested in the new Brasolia [4], an association previously suggested by Neubig et al. [5], suggesting further taxonomic realignments are necessary.
Despite the availability of molecular and morphological data, the cytogenetic information, which is known to significantly help to understand the relationship and delimitation among taxa [7], is limited in the tribe Sobralieae, with only three chromosome counts available until now: two Sobralia species, S. liliastrum Lindl. and S. sessilis Lindl., with 2n = 2x = 48, and one Elleanthus, E. brasiliensis (Lindl.) Rchb.f. with 2n = 2x = 50 [8].
Regarding the genome size, no data is available. The karyotype evolutionary history results from the successive accumulation of numerical and structural chromosome rearrangements (for a review, see [9,10]), making it mandatory to evaluate the cytogenetic data in a phylogenetic background. The family Orchidaceae Juss. is known for its large variation in both chromosome numbers, which varies more than 20-fold (from 2n = 10 to 2n = 240; [11,12]), and genome size, which varies more than 168-fold, the second-largest variability in Angiosperms [13]. It makes Orchidaceae an interesting clade for studies in karyotype evolution, and by inferring the historical karyotype and genome size modifications, it also gives clues about taxonomic delimitations [14,15]. However, the scare cytogenetic data in tribe Sobralieae prevents any further discussion on Brasolia and Sobralia delimitation, as well as the Sobralieae karyotype evolution.
Based on the statement that the more taxonomical data sources, the more realistic the classification becomes, we have chosen the karyotype evolution of Sobralieae as the next step to better understand the relationship between genera Brasolia, Elleanthus and Sobralia. Based on chromosome number and genome size for these three genera, we aim to answer the following questions: (1) Which are the main evolutionary mechanisms shaping karyotype and genome size in tribe Sobralieae? and (2) Based on the ancestral base chromosome number (x) and genome size (1C-value) for these three genera and tribe Sobralieae, is it possible to separate the genera? We gathered the somatic chromosome number for 21 species and the genome size for 20 species to answer these questions and interpreted all the data in a phylogenetic approach, based on Maximum Likelihood (ML) and Bayesian Inference (BI) analyses, both using nuclear ribosomal DNA (nrITS) and maturase K (matK) regions.

Phylogenetic Analysis Suggested Brasolia as Paraphyletic
Both applied methodologies (ML and BI) revealed the same general results, with Sobralia as a monophyletic genus and Brasolia as paraphyletic with Elleanthus nested in it (Figures 2 and S1). Despite the general similarity between both reconstruction methods, some differences could be detected, mainly inside Brasolia regarding the position of B. portillae and B. mandonii in relationship to the remainder Brasolia species ( Figure S1). While this group was placed as a sister of remainder Brasolia and Elleanthus in ML, it was placed as a sister of B. flava and Elleanthus in BI. Considering the slight differences between ML and BI, we used both phylogenies to further steps. Both trees were pruned, removing species without chromosome number and genome size data, saving the tree topology for following analysis (Figure 2).        The DNA content was estimated for 20 species (six Brasolia, two Elleanthus, and 12 Sobralia), ranging from 1C = 2.04 pg in E. crinipes to 1C = 8.05 pg in S. luerorum, reflecting a four-fold difference in genome size values (Table 1). According to the categorization proposed by Soltis et al. [16], three species can be considered as having small genomes (1Cvalue varying from 2.04 to 3.43 pg) and 17 as having intermediate genome sizes (1C-value varying from 3.77-8.05 pg; Table 1). Table 1. Chromosome number and genome size in Sobralieae. The list of Brasolia, Elleanthus, and Sobralia species analyzed here and published on the literature is presented. For each species, the determined chromosome number is presented followed by its figure in text or reference on the literature. For each species analyzed here, the number of analyzed specimens in FCM is indicated in brackets. The genome size is presented in pg, followed by its standard deviation (SD), and Mega base-pairs (Mbp). The genome size (the 1C-value) is given in pg ± SD (Standard Deviation) and Mbp (Mega base-pairs). For each 1C-value, the sample Coefficient of Variation (CV) is presented and the genome size is classified according Soltis et al. [16]. The genome size among Brasolia species ranged from 1C = 6.37 pg in B. altissima to 1C = 3.30 pg in B. portillae, representing a two-fold variation. According to 1C-value, Brasolia species could be organized into four groups (F = 3.719, p < 2 × 10 −16 ; Figure 5a): the first represented by B. altissima with the largest genome size, the second group formed by B. cattleya-B. dichotoma-B. flava, the third composed by B. hirtzii, and the last group formed by B. portillae with the smallest genome size found in the genus (Figure 5a). The two Elleanthus species differ two-fold in their genome size, which enabled the identification of those two species based on the genome size (F = 5.622, p < 2 × 10 −16 ; Table 1). Among the 12 Sobralia species, the estimated 1C-value presented a significant difference among species (F = 650, p < 2 × 10 −16 ; Figure 5b), ranging from 1C = 3.43 pg in S. fragrans to 1C = 8.05 pg in S. luerorum (Table 1; Figure 5b). The three larger genome sizes were found in S. luerorum, S. macrantha, and S. rosea, while the two smaller ones were found in S. crocea and S. fragrans, while the remaining species formed a heterogeneous group without clear differentiation among species (Figure 5b).

No
The genome size among Brasolia species ranged from 1C = 6.37 pg in B. altissima to 1C = 3.30 pg in B. portillae, representing a two-fold variation. According to 1C-value, Brasolia species could be organized into four groups (F = 3.719, p < 2 × 10 −16 ; Figure 5a): the first represented by B. altissima with the largest genome size, the second group formed by B. cattleya-B. dichotoma-B. flava, the third composed by B. hirtzii, and the last group formed by B. portillae with the smallest genome size found in the genus (Figure 5a). The two Elleanthus species differ two-fold in their genome size, which enabled the identification of those two species based on the genome size (F = 5.622, p < 2 × 10 −16 ; Table 1). Among the 12 Sobralia species, the estimated 1C-value presented a significant difference among species (F = 650, p < 2 × 10 −16 ; Figure 5b), ranging from 1C = 3.43pg in S. fragrans to 1C = 8.05 pg in S. luerorum (Table 1; Figure 5b). The three larger genome sizes were found in S. luerorum, S. macrantha, and S. rosea, while the two smaller ones were found in S. crocea and S. fragrans, while the remaining species formed a heterogeneous group without clear differentiation among species (Figure 5b). Regarding the ancestral reconstruction for chromosome number, for both trees, the ChromEvol suggested the same model as the best one: constant rate without duplication

Ancestral Reconstruction of Chromosome Number and Genome Size Indicate x = 25 and Small Genome Size for the Common Ancestor of Brasolia-Elleanthus-Sobralia
Regarding the ancestral reconstruction for chromosome number, for both trees, the ChromEvol suggested the same model as the best one: constant rate without duplication (Table 2). However, considering the ∆AIC < 2, no differences were found among constant rate, constant rate with demi-polyploidy and the best model for ML and BI phylogenies ( Table 2). The Model Adequacy test confirmed the null model, constant rate without polyploidy, as the most adequate considering the variance and entropy tests (p = 0.13 and 0.21 for ML and p = 0.31 and 0.35 for IB, respectively) but not the parsimony (p = 0.00 for both ML and BI), suggesting that the chromosome number transitions are variable among clades. The best models highlight the importance of dysploidy variation as the main mechanism in chromosome evolution in Brasolia, Elleanthus and Sobralia. The best model applied in both phylogenetic trees revealed similar results for most phylogenetic nodes. The base chromosome number x = 24 was suggested for the tribe Sobralieae, as well as for the genera Brasolia and Sobralia, while x = 25 was suggested for genus Elleanthus (Figure 2). The descending dysploidy was more frequently detected than ascending dysploidy, while no duplication was observed (Table S2). Successive descending dysploidies were observed in both phylogenetic trees reaching x = 23 in the large group comprehending S. atropubescens, S. sessilis, S. yauaperyensis, S. decora, S. fimbriata, S. powellii in IB phylogeny, but also grouping S. bimaculata, S. vallecaucana, S. crocea, and S. fragrans in ML phylogeny (Figure 2). Further descending dysploidies could be observed on S. crocea and S. fragrans, reaching x = 22, and B. cattleya and B. dichotoma, reaching x = 23 (ML and IB). Only one ascending dysploidy was detected in both trees, on the node of Elleanthus genus with x = 25 ( Figure 2 Figure 2). Inside Brasolia, genome size increase could be observed in B. altissima (1C = 6.37 pg, 2n = 46; Figure 2). A genome size reduction could be observed in S. crocea (1C = 3.77 pg, 2n = 44) and S. fragrans (1C = 3.43 pg, 2n = 42) and in B. portillae (1C = 3.3 pg, 2n = 48). An additional important genome size reduction was also observed in E. crinipes that presented half of the genome size (1C = 2.04 pg) compared with E. brasiliensis (1C = 4.79 pg), but with the same chromosome number (2n = 50 for both species; Figure 2).
Chromosome number presents no phylogenetic signal considering both phylogenetic trees, ML and BI, and tested indexes, Pagel's lambda and Blomberg's K suggesting that phylogenetic history has a minimal effect on both traits (Table 3). Otherwise, genome size seems to present phylogenetic signal considering both phylogenetic trees, ML and BI, and tested indexes, Pagel's lambda (values very close to 1) and Blomberg's K (values equal or superior to 1; see Table 3). Following the concept of phylogenetic signal, i.e., related species seems to present similar traits more than they resemble species drawn at random in this group, the phylogenetic signals for genome size suggested that this trait could reflect the evolutionary history of Brasolia-Elleanthus-Sobralia, being a potential informative trait to discuss the genera delimitation.

Phylogenetic Reconstruction
Both methodologies, ML and BI, presented the same general topology and agreed with previously published literature [4,5]. Baranow et al. [4] and Neubig et al. [5] noted that Brasolia species (along with Elleanthus) and Sobralia were placed on separate clades. It explicitly confirms the Dressler [6] attempt, followed by Baranow et al. [4], to place these two groups in two genera. The relationship between Brasolia and Elleanthus is hard to explain, especially considering the morphological distinctness between these two genera (compare Figure 1a-c with Figure 1d-j). The morphological differences between Brasolia and Elleanthus are much more visible than Sobralia and Brasolia morphological differences (compare Figure 1a-c with Figure 1d-j). Despite this, Sobralia and Brasolia formed two clades, both with high support. However, Elleanthus, even presenting apparent morphological differences from Brasolia, was placed nested on it in both phylogenies. Taking into account the present and previous phylogenetic studies, e.g., [4][5][6], one can suggest that Brasolia should be transferred to Elleanthus. However, from the practical point of view, it is not supported as the taxa are morphologically distinct, and the Elleanthus sampling is low. The Elleanthus taxonomic circumscription remains an open question for future studies in which a broader Sobralieae sampling should be mandatory.

Ancestral Base Chromosome Number Reconstruction Indicates Dysploidy as the Main Mechanism of Karyotype Evolution
There were just three available data regarding Sobralieae chromosome number [8] and no data about genome size. However, such data is now enlarged to 23 chromosome numbers and 20 genome sizes. The chromosome number reported to Sobralia [8], 2n = 48, was recurrent among species analyzed here, present in 38.1% of the counts. Furthermore, we confirm 2n = 50 to Elleanthus brasiliensis [8].
Our data showed a sizeable dysploid variation in both Brasolia and Sobralia, confirmed by the ancestral reconstruction analysis that suggested the descending dysploidy as the main mechanism of chromosome change in Sobralieae. Nevertheless, one should keep in mind that here dysploidy refers to chromosome number variation by one or few chromosomes and, in this sense, it includes both aneuploidy and dysploidy (for review see [9,17]). The importance of dysploidy in orchid chromosome evolution was already suggested in different orchid groups as subtribe Pleurothaliidinae (tribe Cymbidieae [18]) and the genera Christensonella Szlach., Mytnik, Górniak andŚmiszek [19], Heterotaxis Lindl. [16], Bifrenaria Lindl. and close genera [20], all from subtribe Maxillariinae, tribe Cymbidieae. In Christensonella and Heterotaxis, it was possible to map the point involved in the fusion-fission rearrangement causing the dysploid variation, confirming the occurrence of fission-fusion (i.e., dysploidy variation) associated with DAPI + heterochromatic bands, reducing from 2n = 38 to 2n = 36 in the former genus [19] or with rDNA 45S, reducing from 2n = 42 to 2n = 40 in the last genus [15]. In tribe Sobralieae, we could confirm the importance of descending dysploidy with chromosome number reduction from 2n = 48 (n = 24) to 2n = 46, 44, 42 in Sobralia and Brasolia genera. However, based on the homogeneous chromosome morphology presented in the species analyzed here, it was impossible to infer the fissionfusion points or chromosome loss, causing the chromosome number reduction, making mandatory the use of in situ hybridization techniques.
The frequent occurrence of dysploidy in Sobralieae is in accordance with the diploidisation period following whole-genome duplication. Polyploidy is a central evolutionary mechanism in plants, but its occurrence is not random or constant throughout plant evolution [21][22][23]. Otherwise, it is suggested that polyploidy pulses occurred with high frequency at some geological periods as the Cretaceous-Paleogene transition (K-Pg; 66 Mya), Paleogene (Paleocene-Eocene transition; 56 Mya), and Neogene (at Late Miocene; 11 -5 Mya), configuring periods of major stress (reviewed by [24]). Such concentration of polyploidy events in periods of geological history is confirmed by a high concentration of polyploid basal nodes in the Angiosperms phylogeny, representing some large extant plant families [23][24][25][26], which could be true for Orchidaceae with an estimated age of 90 million years ago (Mya; [27][28][29][30]). Furthermore, the suggested base chromosome number in Orchidaceae is x = 7 [8] or x = 10 [31], which suggests that most extant orchid species are polyploids. Such an assumption could be applied to the tribe Sobralieae (x = 24).
The tribe Sobralieae is placed on the base of the subfamily Epidendroideae (44.50-60.27 Mya; [30]), the largest Orchidaceae subfamily [32]. Sobralieae possibly diverged at 45 Mya on the Neotropics, with Brasolia, Elleanthus, and Sobralia diverging more recently, at ca. 7-10 Mya [27,28,30]. Based on the previously assumed base chromosome number of Orchidaceae, the tribe Sobralieae (x = 24) should be considered a hexaploid followed by descending dysploidy (x = 7 -> 14 -> 28 -> 26 -> 24) or tetraploid followed by ascending dysploidy (x = 10 -> 20 -> 22 -> 24). The genomic redundancy generated by polyploidy were "accommodated" in a diploid genome by the diploidisation period with descending dysploidy as one of the main chromosomal rearrangements observed during diploidisation [33]. The identification of descending dysploidy as the main mechanism of karyotype evolution in tribe Sobralieae is in accordance with the diploidisation period following polyploid pulses, possibly placed on the ancestral node of Orchidaceae. Since Brasolia and Sobralia are subject to the same karyotype evolution mechanism and because chromosome number has no phylogenetic signal, it is hard to use the karyotype to differentiate these groups. However, along the descending dysploidy, the elimination of repetitive sequences is another frequent event during the diploidisation period [34], affecting the genome size, which presents a phylogenetic signal.

Genome Size Reconstruction: The Role of Genome Size in Species Evolution
The genera Brasolia and Sobralia did not differ in genome size, while Elleanthus differs from both genera, probably due to the reduced 1C-value of E. crinipes. The genome size seems not to have any association with chromosome number variation. For example, Sobralia species with 2n = 48 vary 2-fold in their genome size (1C-value from 4.34 pg to 8.08 pg), while Brasolia species with 2n = 46 vary from 4.97 to 6.37 pg, and between the two Elleanthus species, both with 2n = 50, the genome size varies more than 2-fold (from 2.04 to 4.79 pg). The difference observed in genome size is probably due to differences in repetitive sequences as transposable elements and satellite DNA families [34,35]. The homogenisation and elimination of such redundant sequences are characteristics of the diploidisation process [35] and probably shaped the genomes in alternative ways resulting in genome size differences, even with the same chromosome number [36].
The chromosome number and genome size are essential as reproductive barriers, ensuring the species' integrity [37]. Here, the most significant difference in genome size, conserving the same chromosome number, was observed in Elleanthus. The two studied Elleanthus species, besides differing in genome size, also differ in their distribution. While E. brasiliensis ranges from French Guiana to southern Brazil and in the Brazilian Atlantic Rain Forest occurring from the sea level until 1000 m, E. crinipes is endemic to the highland forest areas in Brazilian Atlantic Rain Forest, occurring only above 800 m [38,39]. Despite the region of sympatry between the two species that present simultaneous blooming and partially share pollinators, no hybrids were described between these two species, with no fruit being formed after manual interspecific crossings [38]. Despite the two species presenting the same chromosome number with similar karyotype morphologies, the twofold genome size difference should configure as a strong post-zygotic barrier preventing hybridisation, highlighting the importance of genome size as a reproductive barrier and its importance in speciation (reviewed [36]).

Genus Delimitation in Sobralieae: Future Perspectives
Despite the genera Brasolia and Sobralia being placed on two groups with high support in the molecular phylogeny [4,5], they present very similar morphology and could not be differentiated based on chromosome number and genome size, as suggested by the ancestral state reconstruction. However, Elleanthus, with a distinct morphology, was nested in Brasolia [4,5] and present study but also did not show a clear difference in chromosome number and genome size. Our limited sampling of Elleanthus prevents further discussion but highlights the importance of future studies in this genus.
The phylogenetic position of Sobralieae makes studies in this tribe of particular interest to better understand the evolutionary trends in the major subfamily of Orchidaceae. Furthermore, considering that Epidendroideae represents ca. 80% of orchids, better comprehending the diversification of this clade configures as one of the main objectives in orchid diversification and evolution studies [30,40] since such subfamily reflects the diversity of the whole family, and Sobralieae, placed at the base of subfamily Epidendroideae, is of particular interest.

Plant Material
Plant material for cytological and molecular studies was cultivated in two living orchids collections: of the University of Gdansk, Poland and Federal University of ABC (UFABC). All the experiments were performed in accordance with relevant guidelines and regulations.

Phylogenetic Analysis
For phylogenetic reconstruction we sampled 57 individuals from 53 species distributed in four genera: 11 Brasolia, four Elleanthus, 34 Sobralia, and four Neottia species (outgroup) ( Table S1). We sampled species favoring those with available material for cytogenetic data (chromosome number and genome size), but that also represent all groups detected in previous studies [4,5].
Genomic DNA extraction was performed in two ways, both AX Sherlock kit (A&A Biotechnology, Gdansk, Poland) and CTAB 2× protocol [41] excluding β-mercaptoethanol were used. For both protocols, a fragment of plant tissue (~30 mg) and a ceramic sphere were placed in FastPrep tubes and the samples were homogenized. The final pellet was resuspended in 50 µL TE buffer. Two sequences were amplified: nrITS and matK. The nrITS sequences were amplified using two different pairs of primers: 101F and 102R [42] and 17SE and 26SE [43], while, for the matK gene, the primers 19F [44] and 1326R [45] were used for all samples. The PCRs, for both of the markers, were performed in a total volume of 25 µL, containing 1 µL temple DNA (~10-100  The phylogenetic reconstruction was based on the two obtained matrices, from nrITS and matK, aligned in MAFFT v. 7.453 software using progressive alignment and iterative refinement methods [46]. The aligned sequences were trimmed using BioEdit v. 7.2 [47] and concatenated using Mesquite v 3.61 [48]. The generated matrix was analyzed under Maximum Likelihood (ML) and Bayesian Inference (BI).
The ML analysis was performed using IQ-TREE2 [49], searching for the best-scoring ML tree under the following nucleotide substitution models: TNe+G4 to nrITS and K3Pu + F + G4 to matK, both chosen by ModelFinder applying the Bayesian Information Criterion (BIC) [50]. We used 100 bootstrapping to assess branch supports. The BI was performed using MrBayes v. 3.2.7a [51] with GTR+I+G as the selected evolutive model for both partitions. Four independent runs, one hot and three cold chains, were started from different random trees to ensure that individual runs had converged to the same result. We used 20 million generations per chain with a sampling frequency of 2000. Split frequencies below 0.01 were used to check for convergence and the effective sample size (ESS) for each run was checked in the Tracer v. 1.7.1 [52]. In this case, 25% of trees were excluded as burn-in. Saved trees were summarized in a majority rule consensus tree created with nodal confidence assessed by posterior probabilities (PP), which were strongly supported when equal to or higher than 0.95 [53,54]

Cytogenetics Analysis
For chromosome counting, between one to four specimens per species were analyzed. For each specimen, at least three slides were analyzed, with, at least, five metaphase plates each. Fresh root tips were pretreated with 0.002 M 8-hydroxyquinoline at 4 • C or 10 • C for 12 h. The material was then fixed in a mixture of 99.8% ethanol and glacial acetic acid (3:1, v:v) and stained according to standard methods with aceto-orcein [55], Feulgen's stain [56] or DAPI (1 µg mL −1 ) for 30 min. Chromosomes in late prophase and/or metaphase were analyzed using a Nikon Eclipse E800 epifluorescence microscope equipped with differential interference contrast (DIC) optics and a Nikon DS-5Mc CCD camera (PRECOPTIC Co., Warszawa, Poland) or with a fully automated upright fluorescent microscope (Leica DM6000 B, KAWA.SKA Sp. z.o.o., Zalesie Górne, Poland). Photomicrographs were arranged and analyzed in Adobe Photoshop (Elements 11 and CS6; Adobe System Inc., San Jose, CA, USA).
The genome size estimation was measured using fresh and young leaves of the selected Brasolia, Elleanthus, and Sobralia species. Depending on plant material availability, between one to three specimens per species were analyzed (Table 1), and each speci-men was analyzed in triplicate (Table 1). Plant material was prepared as previously described [57], using 1 mL of Tris-MgCl 2 nucleus isolation buffer [58] supplemented with 1% (w/v) polyvinylpyrrolidone (PVP-10) for Brasolia and Sobralia and using Ebihara buffer for Elleanthus [59]. The buffers were supplemented with ribonuclease A (50 mg mL −1 ) and propidium iodide (PI, 50 mg mL −1 ). For 25 accessions of Brasolia and Sobralia, Secale cereale cv. Dankowskie (2C = 16.19 pg; [60]) was used as an internal standard, whereas Pisum sativum cv. Set (2C = 9.11 pg; [61]) was used for S. luerorum accessions and Vicia faba cv. Inovec (2C = 26.90 pg; [60]) was used for Elleanthus species, avoiding sample and standard peaks overlapping. Genome size was estimated using a CyFlow SL Green (Partec GmbH, Münster, Germany) flow cytometer and a FACS Canto II cytometer (Becton Dickinson, San Jose, CA, USA), kindly made available by the Microbiology and Immunology Department of IBB-UNESP/Botucatu, Brazil, both equipped with a high-grade solid-state laser with green light emission at 532 nm. The 2C DNA content was measured in at least 5000 events using linear amplification. Histograms were evaluated using a FloMax program (Partec GmbH, Münster, Germany) and Flowing Software 2.5.1 (http://www.flowingsoftware.com/ accessed on 12 October 2021). The coefficient of variation (CV) of the G0/G1 peak of the studied species ranged between 2.84 and 5.00%. The nuclear DNA content of each species was calculated using the linear relationship between the ratio of the target species and the internal standard 2C peak positions on the histogram of fluorescence intensities. The 2C DNA contents (pg) were expressed in Mbp using the equation 1 pg = 978 Mbp [62]. The results of FCM estimation were analyzed using a one-way analysis of variance (ANOVA) following by Duncan's test (p < 0.05) in R [63].

Chromosome Number and Genome Size Evolution
The phylogeny reconstructed here was pruned, removing species without information about chromosome number and genome size. The pruning used the function drop.tip from 'ape' package [64] in R [63]. The pruned phylogenetic tree was used to reconstruct the ancestral state for chromosome number, using ChromEvol v.2 [65,66] with posterior Adequacy test (Rice and Mayrose 2021) and genome size, using Trace Character History in Mesquite v 3.61 [60].
ChromEvol uses likelihood-based methods to evaluate the chromosome number changes along the phylogenetic branches, choosing the model that better performs among the eight available models. The inference considers four parameters: polyploidy (rate ρ), demi-duplication (i.e., a fusion of gametes of different ploidy, with rate µ), and dysploidy, which could be ascending (rate λ) or descending (rate δ). These parameters are combined and tested under two scenarios: constant and linear models. While the former estimates the rate of changes independently of the current chromosome number, the last considers the possibility of chromosome number change depending on the current chromosome number. Both model sets include a null model that assumes no polyploidization events. So, ultimately, the eight models (four constant models and four linear) were tested, allowing us to infer the main mechanism in chromosome number changes and the ancestral haploid number on the nodes of the phylogenetic tree. The best-fitted model was selected using the AIC and, considering the best-fitted model, we re-ran the analysis using 10,000 simulations. The difference among AIC values below two (∆AIC < 2) were considered not distinguishable [67,68]. The model adequacy test [69] was performed online (http://chromevol.tau.ac.il/ accessed on 12 October 2021) to evaluate the selected model's capability in generating the observed chromosome numbers.
Given the overall effect of phylogeny on a variation of chromosome number and genome size, we estimated the phylogenetic signal (i.e., the "tendency for related species to resemble each other more than they resemble species drawn at random from the tree"; [70,71] using Pagel's lambda (λ; [72]). The index was estimated using phylosig from 'phytools' package [73] in R [63].