Introduction

Chromosomes represent the final organizational stage of the nuclear genome and have relative stability throughout the reproductive process. Thus, changes in chromosome numbers serve as key mechanisms for understanding the evolutionary history of a group (King 1995; Graham and Cavalcanti 2001; Carta et al. 2020). Polyploidy, characterized by the whole-genome duplication (WGD), is a widespread phenomenon in flowering plants (Meyers and Levin 2006; Wood et al. 2009). Polyploids are commonly classified into two major categories based on their genotypic composition: autopolyploids, formed by the duplication of homologous chromosomes, and allopolyploids, resulting from interspecific hybridization followed by chromosome set duplication or fusion of unreduced gametes. In both cases, the nuclear content comprises non-homologous chromosomes (Parisod et al. 2010). Together with WGD, dysploidy events, which encompass variations in chromosome number occasionally induced by chromosomal fissions and/or fusions, represent major karyological change phenomena in plant evolutionary history (Escudero et al. 2014).

Considered intrinsic to plant evolution, WGD events are accompanied by gene loss, differential expression, or neofunctionalization, promoting increased genetic and genomic variation (Blanc and Wolfe 2004; Soltis et al. 2009; Ramsey and Ramsey 2014). Polyploid individuals when compared to their diploid ancestors, are known for their enhanced adaptability to new habitats, with a greater capacity to occupy new ecological niches (Baniaga et al. 2020; Otto and Whitton 2000). Although these phenomena typically yield individuals that initially have lower adaptive value than their progenitors, they are subject to natural selection that may favor a small fraction that can better adapt to a distinct environment (Stebbins 1971). Consequently, even though the phenotypic impact may be subtle, the resulting differences are of utmost importance for the evolutionary success of new polyploid lineages (Otto and Whitton 2000).

One possible consequence of polyploidy in plants is the fixation of mutations that can generate morphological variation. This variation is primarily characterized by an increase in cell size (Otto and Whitton 2000), including changes in stomatal size and density (Balao et al. 2011; Beaulieu et al. 2008; Munzbergová 2009), and also affecting various vegetative and reproductive traits, such as larger flowers and seeds (Chansler et al. 2016; Griesbach 1985). Understanding the relationship between morphological diversity and WGD events is crucial, as the selection of novel morphological characteristics can facilitate the adaptation to new environments, leading to evolutionary and ecological consequences (Garbutt and Bazzaz 1983; Li et al. 2010; Chansler et al. 2016; Soltis et al. 2016; Van de Peer et al. 2017; Gloria 2020; Moraes et al. 2022).

Numerous angiosperm families display extensive karyotypic variation, and several polyploidy events, including Asteraceae, Fabaceae, and Orchidaceae (Goldblatt 1980; Bairiganjan and Patnaik 1989; Weiss-Schneeweiss et al. 2008; Doyle 2012; Hedrén et al. 2018; Senderowicz et al. 2021). Lythraceae, a eudicot family of Myrtales (Chase et al. 2016), encompasses 28 genera (Facco and Cavalcanti 2023; Inglis et al. 2023). Among them, Cuphea is the most diverse, comprising around 260 species (Facco and Cavalcanti 2023). Cuphea species are herbaceous or small perennial shrubs known for their substantial morphological variation in vegetative and reproductive organs (Fig. 1). The native geographical distribution of Cuphea is limited to the Neotropics, with accidental introductions in Africa naturalized and possibly in China, spanning diverse ecosystems such as wetlands, forests, savannas, and prairies, ranging from coastal regions to the central continent, and from sea level to mountainous areas (Facco and Cavalcanti 2023). Cuphea represents the most karyotypically diverse genus within Lythraceae, with haploid chromosome numbers ranging from 6 to 86 (Graham 1992; Graham and Cavalcanti 2001; Pozzobon et al. 2022).

Fig. 1
figure 1

Variation of interspecific flower and leaf characters in Cuphea. a Leaf of C. appendiculata; b leaf of C. thymoides; c flower of C. llavea; d flower of C. pascuorum. a Hartweg K000532885 (KEW); b Hatschbach K001073728 (KEW); c Davis Fl212408 (FLAS); d Harley et al. K001073612 (KEW)

Full size image

Detailed taxonomic studies on Cuphea have been ongoing since the late twentieth century, involving taxonomic revisions and the description of new species (Brauner and Cavalcanti 2018; Cavalcanti and Graham 2008; Facco 2015; Facco and Cavalcanti 2022, 2023; Graham 1988, 1989a, 2017, 2019a, 2019b; Graham and Cavalcanti 2013). The advancement of Cuphea phylogeny, combined with available taxonomic and cytogenetic knowledge, provides ample opportunities for investigating hypotheses concerning the evolution of chromosome numbers. Previous studies hypothesized a potential relationship between chromosomal diversity and the taxonomic and morphological richness of Cuphea, indicating that the genus diversity may be linked to polyploidy and aneuploidy events (Graham and Cavalcanti 2001; Barber et al. 2010).

Thus, we conducted an investigation using a molecular phylogeny to: 1. Explore the evolution of chromosome numbers in the genus by reconstructing ancestral chromosome numbers; 2. Assess the presence of phylogenetic signals for morphological characters and karyotypes; 3. Determine possible correlations between karyotypes and leaf and floral morphology. Considering that Cuphea represents a widely distributed genus across heterogeneous ecosystems, and exhibits diverse morphology with a broad range of karyotypes, we hypothesize that polyploidy events throughout its evolutionary history are associated with the vegetative and reproductive morphological variation observed in the genus.

Material and methods

Taxon sampling

The chromosome number data available for Cuphea species was gathered using the Chromosome Counts Database—CCDB (2021, continuously updated) and the IAPT/IOPB Chromosome Data (2021, continuously updated). These databases provide information from various sources, including Dollon and Hamel (1967), Graham (1968a, b, 1980, 1982, 1987, 1988, 1989a, 1989b, 1992), Graham and Cavalcanti (2001, 2007), Goldblatt (1981), Guha (1972), Molero et al. (2002), Ornduff (1967), and Pozzobon et al. (2022). A total of 104 species with reliable data were sampled, representing approximately 40% of the genus. In cases where different karyotypic counts were reported for the same species (3%), the highest recorded chromosome number was considered. Species with supernumerary chromosomes were included, although the B chromosomes were not counted (Online Resource 1).

A morphological database was established for the 104 Cuphea species with available chromosomal data. Morphological traits related with species environment adaptations, such as leaf length and width (Knight and Ackerly 2003; Ackerly 2004), along with reproductive biology traits, floral tube length was measured (Vichiato et al. 2014, Caruso et al. 2019). These measurements were obtained from taxonomic revisions of the genus and other published sources (Graham 1968a, b, 1988, 1989a, 1990, 1998, 2017; Cavalcanti and Graham 2008; Facco 2015; Brauner 2018) and, when necessary, supplemented with images from online botanical collections (eg. K, P, MO, NY, RB, R, UB, UFMT) utilizing the software ImageJ for accurate measurements (Rasband 2012).

Out of the 104 characterized species, sequences from 54 species were retrieved from the Genbank database for constructing a phylogenetic tree. This selection, although representing ca. 21% of the genus, encompassed nearly all sections of Cuphea, with the exception of C. sect. Amazoniana for which karyotype data were not available in the literature (Graham 2019b). Geographically, the sampling covered the entire distribution range of the genus (Graham 1968a, b, 1988, 1989a, 1990, 1998, 2017; Cavalcanti and Graham 2008; Facco 2015; Brauner 2018). Molecular markers, including the plastid spacer trnL-trnF, plastid intron rpl16, and nuclear spacer (rDNA) ITS (including 5.8S), were used for the phylogenetic analysis. Most species have at least one nuclear and one plastid marker. Pleurophora anomala (A.St.-Hil.) Koehne was chosen as the outgroup based in Barber et al. (2010).

Phylogenetic analyses

Sequences of each marker were aligned into independent matrices using the MUSCLE (Edgar 2004) on the Geneious platform (Biomatters, Auckland, New Zealand), and alignments were subsequently manually optimized. The best-fitting nucleotide substitution model for each data matrix (ITS, trnL-trnF, and rpl16) was selected using JModelTest v.2.1.5 (Darriba et al. 2012) under the Akaike information criterion (AIC). The GTR + G model was found to be the most appropriate for all three matrices.

The matrices were then concatenated, and Bayesian inference (BI) analyses were performed using Mr. Bayes v3.2 (Ronquist et al. 2012) implemented on CIPRES Science Gateway (Miller et al. 2009) for both each individual data matrix and the concatenated matrix. BI was performed with two simultaneous runs and four chains each. The Markov chain Monte Carlo (MCMC) parameters were set to 20 million generations, sampling every 1000 trees. The initial 25% of trees were discarded as burn-in. Convergence between the two independent runs was checked with Tracer 1.6 (Rambaut et al. 2018) based on the estimated sample size value (> 200). Clades with a posterior probability (PP) greater than or equal to 0.95 were considered to have strong support; while, clades with PP values between 0.94 and 0.90 were considered moderately supported (Cummings et al. 2003; Erixon et al. 2003).

The resulting phylogenetic trees were generated only for the purpose of assessing branch support, identifying well-supported incongruences, and validating the robustness of the inferred phylogenetic relationships. During this process, 10 species were excluded from the analyses due to strongly supported incongruences observed among the independent marker trees (ITS, trnL-trnF, and rpl16), which promoted the occurrence of polytomies in the concatenated data tree. Incongruent phylogenetic patterns can be the result of hybridization, technical issues like misidentification, contamination, uncertainty in phylogenetic reconstruction, inaccurate orthology assessment, analytical artefacts such as branch attraction, horizontal gene transfer, and incomplete lineage sorting (Guo et al. 2018; Davis and Xi 2015; Wendel and Doyle 1998). Thus, to scape these possibilities the final dataset included 44 species (approximately 17% of the genus) for subsequent analyses (Fig. 2).

Fig. 2
figure 2

Consensus phylogram derived from a Bayesian inference analysis using nuclear (ITS) and plastidial (trnL-trnF and rpl16) marker data for Cuphea species. Posterior probability (PP) values are displayed above the branches. The current classification of species sections is positioned next to the scientific name and follows the following acronyms: ARC (C. sect. Archocuphea); BRA (C. sect. Brachyandra); CUP (C. sect. Cuphea); DIP (C. sect. Diploptychia); EUA (C. sect. Euandra); HTD (C. sect. Heterodon); LEP (C. sect. Leptocalyx); MEL (C. sect. Melvilla); and TRI (C. sect. Trispermum). The five major clades are identified by the black bars

Full size image

Cytogenetics

To reconstruct changes in basic chromosomic numbers, an ultrametric tree including all 44 species was constructed using both ITS and plastid datasets, employing BEAST 1.8.4 (Drummond et al. 2012) on CIPRES Science Gateway (Miller et al. 2009). The BEAUti 1.8 program was used to create the input files, applying the same evolutionary models cited above. A lognormal relaxed molecular clock model was applied for this analysis, following the Yule process prior. The MCMC parameters were set to perform 150,000,000 generations, sampling every 1500 iterations. The resulting tree was evaluated using Tracer v1.6 (Rambaut et al. 2018), and the final tree was constructed by processing the post-burn-in trees using TreeAnnotator v1.10.4 (Bouckaert et al. 2019).

The reconstruction of ancestral states for chromosome number was performed using ChromEvol v.2 software (Mayrose et al. 2009; Glick and Mayrose 2014) implemented in the RASP platform (Yu et al. 2015). To accomplish this, we employed the ultrametric tree generated by our analysis and the available chromosome number data provided in Online Resource 1. All available models of chromosome number evolution in ChromEvol were evaluated to determine the best model to explain the observed variation in chromosome numbers. The optimal model, as determined by the Akaike information criterion (AIC), was identified as CONST_RATE. The analysis encompassed 10,000 simulations; while, the basic number considered for analysis was x = 8.

Phylogenetic signal

In order to examine potential associations between the evolutionary history of the genus and the distribution of morphological characteristics, as well as karyotypes, phylogenetic signal analyses were employed. To accomplish this, we used our ultrametric tree and the morphological traits investigated included the minimum and maximum measurements of leaf length and width, and floral tube length. Subsequently, Pagel’s λ coefficient (Pagel 1999) was estimated for each selected trait using the 'phylosig' function on the 'phytools' package v1.9-16 (Revell 2012). Pagel’s λ represents a measure of phylogenetic signal, indicating the extent to which the evolutionary history accounts for the observed distribution of traits among the terminal taxa. Furthermore, this function conducts a hypothesis test to assess the statistical significance of the obtained lambda values. That means, if closely related species resembles each other more closely than expected by chance alone. The lambda values range from 0 to 1, where 0 indicates that trait evolution is independent of phylogeny, while 1 signifies the maximum degree of phylogenetic signal, e.g., trait evolution follows the exact pattern of the phylogenetic tree (Münkemüller et al. 2012).

These analyses were instrumental in evaluating whether the potential correlations between morphological traits and karyotypes could be attributed to the evolutionary history of the group, rather than implying a causative relationship. If both the morphological and cytogenetic datasets exhibit significant phylogenetic signal, it becomes challenging to ascertain whether the observed correlation arose due to ancestral traits shared among descendants (indirect relationship, mere causality) or whether it is a consequence of recent independent events, such as genomic structural changes (e.g., polyploidy) in terminal nodes, thereby enhancing adaptive processes that favor the fixation of morphological traits. Conversely, if no phylogenetic signal is detected in one of the datasets, any putative correlation cannot be explained by indirect factors such as the conservation of ancestral traits and is more likely to possess a cause-and-effect relationship between morphological traits and karyotypes.

The mapping of morphological characters into the ultrametric tree was executed utilizing the 'contMap' function on the 'phytools' package v1.9-16 (Revell 2012). This function, proposed by Revell (2012), facilitates the visualization of ancestral estimations for continuous characters through maximum likelihood (Schluter et al. 1997), interpolating states along branches utilizing the equation outlined in Felsenstein (1985). In the context of this study, its utility lies in examining whether the characters demonstrating phylogenetic signal in Cuphea exhibit a congruent topological distribution when compared to ancestral genomic alteration events related to chromosome numbers. The analyses for ascertaining the phylogenetic signal and character mapping within morphological datasets were executed using R software (R Core Team 2023).

Statistical analyses

The normality of the variables was verified using the Shapiro–Wilk test. Non-normal distributions were observed across all attribute sets, thus, nonparametric tests were performed. The analyses were undertaken in two distinct manners to investigate potential correlations between ploidy levels and the observed variables. To this end, the species were examined separately within two groups: diploids alone (59 spp.) and diploids combined with polyploids (104 spp.), following a similar approach to that employed by Moraes et al. (2022). In this context, both ascending and descending dysploidies were taken into consideration. The diploid group encompassed species with n = 6 to n = 12; whereas, the polyploid group encompassed species with n = 13 to n = 56. Spearman test was used to proceed with the correlation analyses between the chromosome number and morphological variables. Mann–Whitney test was applied to evaluate the potential differences between two morphological datasets, namely diploids (D) and polyploids (P). The statistical analyses were conducted using the R software (R Core Team 2023).

Results

Phylogenetic analyses

Our final phylogram, based on three markers (ITS, trnL-trnF, and rpl16) and derived from Bayesian Inference (Fig. 2), provides good support for five primary clades within Cuphea (clade 1, PP = 1.0; clade 2, PP = 1.0; clade 3, PP = 0.98; clade 4, PP = 1.0; clade 5, PP = 1.0) that were found in previous phylogenetic studies conducted by Graham et al. (2006), Graham and Cavalcanti (2007). Despite our smaller sample size, our analysis demonstrated a good resolution (Fig. 2).

According to Koehne (1903) classification, the first divergent lineage corresponds to C. subg. Cuphea (clade 1, PP = 1.0), including a species from C. subg. Bracteolatae sect. Brachyandra (C. circaeoides Sm. ex Sims) nested among the other taxa. Consequently, C. subg. Cuphea may be found to be non-monophyletic, as well as its two sections, C. sect. Cupheaand C. sect. Archocuphea, which are also not monophyletic. Together, clades 2, 3 and 5 represent C. subg. Bracteolatae and display moderate support (PP = 0.93). However, the sections C. sect. Brachyandra, C. sect. Diploptychia, C. sect. Heterodon, C. sect. Leptocalix, and C. sect. Melvilla, are also non-monophyletic. Notably, only C. sect. Trispermum is confirmed to be monophyletic as part of clade 3. A significant observation is the presence of a clade composed solely of North American species (clade 5). Here, we contribute to this clade by including C. salicifolia Schltdl. & Cham.

Cytogenetics

The retrieved chromosomal numbers exhibit a range from n = 6 (C. lanceolata W.T.Aiton) to n = 56 (C. retroscabra S.Watson), being n = 8 and n = 16 the most frequent (18 and 19 species respectively) (Fig. 3). Within Cuphea, the majority of chromosomal numbers are multiples of eight, suggesting a basic chromosomal number of x = 8. However, ancestral state reconstruction of chromosomal numbers revealed that the ancestral chromosomal number for the genus is x = 11 (n = 11–87.09%; n = 8–7.37%; Online Resources 2, 3 and 4). This number is primarily supported by its prevalence as the common karyotype in clades C1 and C2, which represent the earliest diverging lineages in the genus phylogeny (Fig. 4). Conversely, chromosomal numbers multiples of eight are more frequently observed in clades C3, C4, and C5, which contain most species with available molecular and chromosomal data. Chromosome numbers close to n = 8 largely originate from a descending dysploidy event in clade 10 (Fig. 4; n = 8–100%, Online Resources 2, 3 and 4), which represents the most recent common ancestor of C3 + C4 + C5.

Fig. 3
figure 3

Frequency and distribution of the chromosomal numbers retrieved from the Chromosome Counts Database—CCDB and IAPT/IOPB Chromosome Data

Full size image
Fig. 4
figure 4

Reconstruction of chromosome numbers in the genus Cuphea. The colors filling the circles represent the most probable ancestral chromosome number at each node of the tree. The numbers inside the circles identify the nodes of the tree. The recorded chromosome number changes are represented by squares with P↑ polyploidy; Dip↓ diploidization; D↑ ascending Dysploidy; D↓ descending dysploidy, *Cuphea node. **North American clade node

Full size image

The examination of ancestral karyotypes in Cuphea reveals a high incidence of polyploidy in the genus, with at least 11 events of whole-genome duplication occurring in clades 23 (C4, Fig. 4; n = 8–35.46%; n = 14–32.82%, Online Resources 2, 3 and 4), 29 (C5, Fig. 4; n = 16–81.49%, n = 8–18.51%; Online Resources 2, 3 and 4), and 37 (C5, Fig. 4; n = 8–54.19%, n = 8–45.81%; Online Resources 2, 3 and 4). Additionally, polyploidy events are observed in terminal taxa such as C. pseudosilene Griseb., C. parsonsia (L.) R.Br. ex Steud., C. thymoides Cham. & Schltdl., C. hyssopifolia Kunth, C. spectabilis S.A.Graham, C. aequipetala Cav., C. appendiculata Benth., C. schumannii Koehne, and C. salicifolia Schltdl. & Cham (Fig. 4). The North American clade (C5) accounted for approximately 59% of the polyploidy events in the genus; while, C4 and C3 accounted for 41% of the events.

Dysploidy, characterized by changes in chromosomal numbers, also occurs recurrently in the group. Specifically, 10 events of ascending dysploidy were identified in clades 14 and 43, as well as in terminal taxa such as C. melvilla Lindl., C. pseudosilene Griseb., C. bustamanta Lex., C. spectabilis S.A.Graham, C. tolucana Peyr., C. appendiculata Benth., C. paucipetala S.A.Graham, and C. salicifolia Schltdl. & Cham. Furthermore, 10 occurrences of descending dysploidy were observed within the genus in clades 10, 20, 23, 24, 34, as well as in terminal taxa such as C. gaumeri Koehne, C. denticulata Kunth, C. avigera B.L.Rob. & Seaton, C. aequipetala Cav., and C. lutea Rose ex Koehne. Lastly, a diploidization event was noted in clade 34 (C5, Fig. 4; n = 7–100%, Online Resources 3 and 4).

Overall, it is apparent that alterations in chromosomal numbers in Cuphea primarily occur closer to the terminal branches of the phylogeny. The tree’s backbone consists of clades with the most recent common ancestor possessing n = 8 or n = 11 chromosomes. Additionally, these events are not clustered to specific clades but are dispersed across all major clades (Fig. 4).

Phylogenetic signal

The analyses demonstrated that, apart from maximum leaf width, all other morphological characteristics sampled in this study exhibited significant phylogenetic signal (λ = 0.73–0.45; p < 0.05) (Table 1). However, the chromosomal data exhibit no phylogenetic signal (λ < 0.05). Therefore, any observed correlation between morphology and karyotypes cannot be attributed to indirect factors such as the retention of ancestral characters. By observing the mapping of morphological traits into the phylogenetic tree topology, it is noted that species displaying the greatest lengths of floral tubes and the largest leaves are situated within the North American clade, as well as clade C2. Contrariwise, the smallest values are predominantly observed in clades C1 and C4 (Fig. 5).

Table 1 Phylogenetic signal for morphological characteristics and chromosome number in Cuphea
Full size table
Fig. 5
figure 5

Mapping of all leaf and flower morphological traits that exhibited phylogenetic signal in Cuphea. a Maximum floral tube length; b minimum floral tube length; c maximum leaf length; d minimum floral tube length; e minimum leaf width. The values a 4.5–40, b 3.5–25, c 8–140, d 2–70 and e 0.5–34 represent the range of sizes sampled in millimeters

Full size image

Statistical analyses

The results obtained from the correlation tests indicate a statistically significant relationship between chromosomal numbers and certain morphological characteristics in Cuphea (ρ = 0.22–0.32). Analysis of the data with all ploidy species revealed a positive correlation between the chromosomal number and the minimum leaf length, as well as the minimum and maximum lengths of the floral tube. However, when considering only diploid species, no correlation was observed between the floral traits and the chromosomal number (Table 2). Furthermore, our investigation revealed significant variations in the length of the floral tube between polyploid and diploid species. Specifically, polyploid species exhibited longer floral tubes when compared to diploid species (Table 3; Fig. 6).

Table 2 Analysis of correlations performed using the Spearman test
Full size table
Table 3 Results of the Mann–Whitney test comparing diploid and polyploid species
Full size table
Fig. 6
figure 6

Comparison of the datasets from maximum floral tube length (a) and minimum floral tube length (b) among diploid (D) and polyploid (P) species in Cuphea

Full size image

Discussion

Our research confirmed that the chromosomal evolution of the genus Cuphea exhibits multiple independent polyploidy events, and the pivotal role of WGD and dysploidy in the extensive karyotypic variation observed. We found phylogenetic signal for some morphological features but not for chromosome numbers, so the correlations found cannot be explained by conservation of ancestral traits and is more likely to possess a cause-and-effect relationship. There is compelling evidence suggesting that the evolution of floral tube and leaf sizes has been influenced by the occurrence of WGD events, as evidenced by a positive correlation with higher chromosomal numbers. The North American clade within Cuphea represents a noteworthy model for potential future deliberations regarding the contribution of polyploidy to adaptive radiations.

This study aimed to examine the potential relationship between karyotypic evolution and morphological character variation in Cuphea. The presence of different chromosome numbers within the genus can be attributed to events of WGD accompanied by chromosomal rearrangements, including descending and ascending dysploidy and subsequent diploidization (Lysák and Schubert 2013). Our findings demonstrate that polyploidy and dysploidy events played a crucial role in the karyotypic diversification of Cuphea, with 20% of the studied species exhibiting polyploidy events in terminal branches. This prevalence of polyploidy in terminal branches is an uncommon occurrence among angiosperms but has been observed in select genera such as Opuntia (Cactaceae, Majure et al. 2012), Epidendrum (Orchidaceae, Cordeiro et al. 2022), Turnera (Passifloraceae, Solis Nefa et al. 2022), Ulex (Fabaceae, Bellot et al. 2023), and Hieracium (Asteraceae, Krahulcová et al. 2009).

We were able to confirm several dysploidy events in the evolutionary history of Cuphea. Dysploidy has been proposed as one of the primary forces triggering karyotypic variation in plants (Carta et al. 2018), as observed in Asteraceae (Mas de Xaxars et al. 2016) and Marantaceae (Winterfeld et al. 2020). Changes in chromosomal numbers created by dysploidy events persist longer through evolutionary time compared to polyploid changes (Escudero et al. 2014), it may explain the high number of dysploidy events found in our analysis. We found at least nine ascending or descending events in terminal clades, indicating that they could be related to speciation as described to Artemisia L. (Asteraceae, Mas de Xaxars et al. 2016), Reichardia Roth (Asteraceae, Cho et al. 2022) Chamaecrista Moench (Fabaceae, Braz et al. 2024), and Passiflora L. (Passifloraceae, Sader et al. 2019), playing a crucial role in the diversification (De Storme and Mason 2014).

Our phylogenetic analysis exhibited higher resolution compared to that of Barber et al. (2010), who chose not to concatenate the matrices of plastidial and nuclear markers. By excluding taxa that caused significant incongruence between the matrices, we were able to produce a concatenated analysis with robust branch supports. The incongruences observed may have arisen from past hybridization and introgression events, retention of ancestral polymorphism (Pessoa et al. 2022), technical issues as misidentification, contamination, horizontal gene transfer, and incomplete lineage sorting (Wendel and Doyle 1998; Davis and Xi 2015; Guo et al. 2018).

Notably, we confirmed the presence of a clade comprising solely North American endemic species (Clade 5; Fig. 2). These species belong to the second center of diversification within the genus and are predominantly found in the dry subtropical forests of the southern Mexican mountain ranges, known for their remarkable biodiversity and richness of endemic species (Marshall and Liebherr 2001; Morrone 2014; Montano-Arias et al. 2018). This clade had been previously identified by Barber et al. (2010), who differentiated it based on its larger flowers and seeds compared to taxa originating from South America. Our analyses align with these previous observations as this clade encompasses species with larger floral and leaf measurements. Additionally, this clade accounts for 59% of the identified polyploidy events within the genus.

Although autopolyploidy was also previously recorded to Cuphea in natural conditions (Gathman and Ray 1987), one of the main phenomena related to the origin of polyploid species is hybridization (allopolyploidy) (Soltis and Soltis 2009). Hybrids have been described for Cuphea since the only completed taxonomic monograph of the genus (Koehne 1903), furthermore significant efforts have been made to produce hybrids with economic potential (Gathman and Ray 1987). However, most of these hybrids are not fertile due to meiotic irregularities, nevertheless, the uncommon fertile hybrids display greater vigor and can have morphological differences (Ray et al. 1988). Gathman and Ray (1987) indicates that when hybridization happens between closely related species with genomes that are little differentiated, with close chromosome homology, diploidization occurs quite rapidly. The occurrence of fertile hybrids in Cuphea suggests a potential mechanism underlying the remarkable presence of polyploids within the genus (Robertson et al. 2010; Bertier et al. 2013; Alix et al. 2017; Morales-Briones et al. 2018).

Overall, we did not detect significant phylogenetic signal in the variation of chromosome numbers (Table 1). This is consistent with the reconstructed karyotypic evolution (Fig. 3), which indicates recent and rapid diversification of chromosome numbers in Cuphea, with most dysploidy and polyploidy events occurring near the terminal branches. Our results stand against the common pattern observed in angiosperms, where karyotypes distribution exhibit phylogenetic signal (Carta et al. 2018). But, it is important to highlight that clade 5 include most of the WDG events and perhaps new analysis restricted to this clade may provide opposite results. WGD and dysploidy phenomena are known to induce meiotic failures (Bomblies et al. 2015). Meiotic adaptations can lead to high karyotypic diversity among closely related species (Haiduk and Baker 1982) which may provide a non-clear phylogenetic organization, as evidenced in Cuphea.

Conversely, we observed significant phylogenetic signal in the morphological characteristics of the genus (Table 1), indicating that closely related species share similar sizes of vegetative and reproductive organs. This pattern is expected as morphological characters tends to correspond to the phylogeny (Fougère-Danezan et al. 2010). Correlations between WGD events and specific morphological characters have been well-documented in the literature (Balao et al. 2011; Beaulieu et al. 2008; Munzbergová 2009). Previous studies in the genus also suggested a relationship between floral tube size and increased chromosome number in certain sections, such as C. sect. Melvilla (Graham 1980). However, a comprehensive analysis integrating phylogenetic, cytogenetic, and statistical approaches is not available for this taxon.

Our sampling effectively represented the discrepancy in floral morphologies within the genus (vide Graham 1990, 1998, 2017; Cavalcanti and Graham 2008; Facco 2015), as well as the variation in chromosome numbers. Based on our data, we were able to confirm that polyploid Cuphea species possess longer floral tubes (Fig. 6). Furthermore, there is a positive correlation between chromosome numbers and floral tube lengths (Table 2), supporting the hypothesis suggested by Graham (1980). This correlation is consistent with observations made by Carta et al. (2018) for angiosperms in general, where smaller chromosome numbers are associated with smaller flowers.

The production of flowers with longer floral tubes may entail trade-offs, such as increased water and nutrient demands and higher carbon allocation (Teixido and Valladares 2014). Bigger flowers are more expensive because of the high demand of energy needed for their production, development, and to keep then physiologically active (Patiño and Grace 2002; Galen 1999). In Cuphea some species of longer flower tubes are known to be pollinated by hummingbirds (Melazzo and Oliveira 2012; Feinsinger 1978). Ornithophilous flowers are characterized by low sugar concentration and high nectar volume (Reed Haisworth and Wolf 1976). Nevertheless, polyploids are known to exhibit enhanced ecological tolerance and adaptive potential (Balao et al. 2011), often accompanied by apomictic reproductive mechanisms that support further population expansion and establishment. Additionally, the fixation of larger floral characteristics throughout evolutionary processes may be linked to the ability of these flowers to attract more pollinators, suggesting a potential positive selective pressure (Galen 1999). However, environmental conditions can neutralize selective pressure of pollinators (Caruso et al. 2019).

WGD phenomena can also contribute to variations in vegetative organs, with polyploid plants often exhibiting larger leaf sizes and higher stomatal densities (Otto and Whitton 2000; Beaulieu et al. 2008). It is important to note that multiple selective pressures can influence the development of these characters (Givnish 1987). However, our study revealed a clear positive correlation between chromosome numbers and leaf length (represented by minimal leaf length) in the genus, similar to the relationship observed with reproductive characters (Table 2). As individuals can have leaves of different sizes, minimum length recognizes better species with bigger leaves from the ones with smaller leaves. Furthermore, the absence of phylogenetic signal for karyotypes (Table 1) excludes the possibility of these correlations being explained by the retention of ancestral characters.

Considering that WGD phenomena serve as facilitators of speciation and evolution in plants (Alix et al. 2017), the polyploidy events in Cuphea may have enabled and driven the emergence and fixation of longer floral tubes and larger leaves as crucial mechanisms during the adaptive radiation of the genus, when expanding its distribution northward on the American continent, given that Cuphea has likely a South America origin (Inglis et al. 2023). Therefore, the colonization of new North American niches may have been facilitated by WGD events and subsequent chromosomal rearrangements.

Conclusions

We found evidence to support a positive correlation between chromosome numbers and overall size of selected morphological traits with polyploid species showing longer floral tubes. Our investigations also revealed a potential association between polyploidy and the adaptive radiation of Cuphea during its distribution expansion to the North America. These results reinforce the importance of whole-genome duplication events in producing trait diversity and, consequently, speciation. We propose that further studies be undertaken, focusing on this clade, to gain deeper insights into the role of polyploidy in Cuphea diversification. A comprehensive analysis, encompassing genome size data, physiological and anatomical information, and an expanded phylogenetic framework, holds promise for elucidating the mechanisms by which polyploidy may have facilitated the diversification of Cuphea.