For the phylogenetic reconstruction of ancestral chromosome numbers, we sampled 58 taxa of 57 species of Cuscuta, including C. indecora Choisy var. indecora and C. indecora var. neuropetala (Engelm.) Hitch. The subgenera Cuscuta, Grammica, Monogynella, and Pachystigma were represented by eight, 42, four, and three species, respectively. While the monophyly of Cuscuta was never seriously challenged, its outgroup relationships and the phylogenetic position within Convolvulaceae remain unresolved (Stefanović and Olmstead, 2004). For this reason, the interpretation of character evolution was based on the ingroup distribution of character states similar to other character evolution studies conducted recently in the genus (e.g., Ho and Costea, 2018, references therein). We used a total of 226 sequences of nuclear (nrITS and 26S) and plastid markers (rbcL and trnL-trnF) obtained by Stefanović and Costea (2008) and García et al. (2014) deposited in GenBank database (Benson et al., 2012; Supplementary Table 1). In addition, new ITS and trnL-trnF sequences were obtained for C. globosa Ridl., because this hexaploid was not included in previous phylogenetic works. The methods of DNA extraction, amplification, and sequencing were those detailed in Stefanović et al. (2007). Sequences were uploaded to GenBank with accession numbers OL362011 (ITS) and OL362010 (trnL-trnF). To align the sequences, the plugin MUSCLE was used in the Geneious v. 7.1.9 software (Kearse et al., 2012).

The phylogenetic relationships were reconstructed using Bayesian Inference (BI) analysis. jModelTest v.2.1.6 (Darriba et al., 2012) selected GTR + I + gamma as the best model of DNA substitution for all analyzed regions, except for trnL-trnF, which had GTR + gamma as best model. We used MrBayes v. 3.2.6. (Ronquist et al., 2012) to perform BI using the concatenated sequences and selected models with two independent runs with four Markov Chain Monte Carlo (MCMC), sampling every 1,000 generations in a total of 15,000,000 generations. Both BI runs were evaluated in Tracer v.1.6 (Rambaut et al., 2014) to verify if the estimated sample sizes (ESS) for each parameter were higher than 200. The consensus tree was generated in MrBayes with a burn-in of 25%. The consensus tree with the posterior probability (PP) was visualized and edited in FigTree v. 1.4.2. (Rambaut, 2014). The jModelTest and BI analysis were performed through the CIPRES Science Gateway (Miller et al., 2010).

New data on the number and position of rDNA sites for 10 species of Cuscuta were obtained for this study (Table 1, voucher information in García et al., 2019). Young shoot tips or flower buds were used for slide preparation according to Ibiapino et al. (2019). Double CMA/DAPI staining was performed as described in Ibiapino et al. (2022). The images were captured with a COHU CCD camera attached to a Leica DMLB fluorescence microscope equipped with Leica QFISH software. After image capture, slides were destained for 30 min in Carnoy and 1 h in absolute ethanol and stored for in situ hybridization at −20°C. The destained slides were subjected to fluorescent in situ hybridization (FISH) according to the protocol described in Pedrosa et al. (2002). Two rDNA probes were used as: the PCR amplified insert of D2 from Lotus japonicus (Regel) K. Larsen (5S rDNA; Pedrosa et al., 2002) and pTa71 from wheat (25-28S, 5.8S, and 18S rDNA; Gerlach and Bedbrook, 1979). Probes were labeled by nick translation with Cy3-dUTP (5S) and digoxigenin 11-dUTP (35S). The 5S was labeled in a reaction with total volume of 12.5 μl containing 1 μg of PCR amplified DNA, 1× Nick Translation buffer (0.5 M Tris HCl pH 7.5; 50 mM MgCl2), dNTP mix (0.016 mM each of dATP, dCTP, and dGTP), 0.08 mM Cy3-dUTP or Alexa-dUTP, 7.5 U of DNA Polymerase I, and 0.006 U of DNase I. The mixture was incubated at 15°C for 1 h or longer if needed, until most fragments were under 500 bp, and reactions were stopped using 0.5 M EDTA. The 35S was labeled with the Nick Translation kit (Invitrogen). The images were obtained as previously described.

A total of 11 species had their genome sizes estimated by flow cytometry, six of them here for the first time: C. glomerata Choisy, C. partita Choisy, C. purpurata Phil., C. racemosa Mart., C. sandwichiana Choisy, and C. globosa Ridl. A suspension of nuclei from shoot tips was prepared using WPB buffer (Loureiro et al., 2007). The nuclei were stained using propidium iodide and the amount of nuclear DNA was estimated using the CyFlow SL flow cytometer software (Partec, Görlitz, Germany). Raphanus sativus L. “Saxa” (1C = 0.53 Gbp), Solanum lycopersicum L. “Stupické polní rané” (1C = 0.94 Gbp), Glycine max (L.) Merr. “Polanka” (1C = 1.20 Gbp), and Zea mays L. “CE-777” (1C = 2.57 Gbp) were used as internal standards (Doležel et al., 2007). The final 2C value was based on three different measurements with 5,000 nuclei each sample, and using the equation “(Sample peak mean/Standard peak) × mean 2C DNA content of internal control (Gbp)” and the software FloMax (Partec) for data processing. The 1C value was obtained by dividing the 2C result by two.

Data on chromosome numbers are summarized in Table 2. Numbers were obtained from the Chromosome Count Database (Rice et al., 2015), to which we contributed numerous counts published in several articles on the cytogenetics of the genus (e.g., García et al., 2019 and references therein). As part of our concerted efforts, these data cover at least one species for all the four subgenera, as well as the 18 sections of subgenera Cuscuta and Grammica recognized by Costea et al. (2015a). Haploid chromosome numbers were used to infer the basic ancestral numbers for each clade and the genus using ChromEvol v. 2.0 (Glick and Mayrose, 2014). To choose the model that best applies to the data set, the first run was made considering all 10 possible models of the program. Then, the model with the smallest Akaike Information Criterion (AIC) value was selected, and this model was submitted to the model adequacy test for adjustment of each selected model parameter (Rice and Mayrose, 2021). The selected model, BASE_NUM_DUPL, considered the most common chromosome number, that is, the number that appears most frequently in the phylogeny, n = 15, and its multiples. The parameters included in this model are the rate of increase of a single chromosome (_gainConstR), the rate of decrease of a single chromosome (_lossConstR), the rate of whole-genome duplications (polyploidy; _duplConstR), rate of transitions per base number (_baseNumberR), and the specified number of chromosomes that characterize a phylogenetic group (_baseNumber), noting that this is not the chromosome number at the root of the phylogeny (Glick and Mayrose, 2014; Rice and Mayrose, 2021).

Due to the numerical chromosome variation reported in C. epithymum (subgenus Cuscuta, 2n = 14, 16, 28, 30, 32, and 34), C. planiflora (sugenus Cuscuta, 2n = 14, 26, 28, and 34), and C. denticulata (subgenus Grammica, 2n = 30 and 60), all the counts found were added to the ChromEvol analysis. First, we made a standard run, using the parameters given by the model adequacy test mentioned above. Then, we executed two more runs, one removing the holocentric clade (subgenus Cuscuta) from the analysis to test for the influence of holocentric chromosomes and intraspecific chromosome number variation. Considering the presence of holocentric and monocentric chromosomes in the genus (Fogelberg, 1938; Pazy and Plitmann, 1994), it is possible that different evolutionary models better apply for different clades (Márquez-Corro et al., 2019). In the second additional run, we fixed n = 15 to the root, because this is the basic number proposed for Cuscuta from cytogenetic data (Pazy and Plitmann, 2002) and was supported by the data compilation produced in this work. The results were plotted in R using the ChromEvol functions as described in Cusimano et al. (2012).

An additional reconstruction of the ancestral chromosome number was performed in Mesquite version 2.75, using maximum likelihood (Maddison and Maddison, 2018) to compare results with those originated from ChromEvol. However, in Mesquite, the haploid chromosome numbers were categorized into nine states: n = 4 (coded as 0), n = 5 (1), n = 7 (2), n = 10 (3), n = 13 (4), n = 14 (5), n = 15 (6), n = 16 (7), n = 19 (8), and all polyploids from n = 21 to n = 75 (9). Additionally, we compared the results to the inference of the ancestral state of this character made along the branches using PastML (Ishikawa et al., 2019).¹ We applied the JOINT (highest likelihood) method. As the model assume only one state per sample, we used for C. epithymum and C. planiflora the cytotypes analyzed in the present work, n = 14, and for C. denticulata, n = 15.

Genome size estimations for six species were newly obtained for this paper in addition to new assessments for five species with previously published data. The reconstruction was performed for 28 Cuscuta species in total (Table 1), three of subgenus Cuscuta, 20 of subgenus Grammica, and five of subgenus Monogynella. This trait was analyzed as a continuous character in the phytools package (Revell, 2012). Thirty-three taxa that lacked GS information were excluded from our original tree using the ape package (Paradis et al., 2004) and we included three additional species with known genome size but unknown chromosome numbers (C. rostrata Engelm. & A. Gray, C. polygonorum Engelm., and C. exaltata Engelm.). Both phytools and ape packages were implemented in R (R Core Team, 2020). For this analysis, we considered the value of 1C in Gbp, and for species with two or more genome sizes published, an average was made between the values. In addition, for comparative purposes, a reconstruction of the genome size was also performed in Mesquite using the implemented maximum parsimony analysis.

To address whether the inclusion of outgroups changed significantly the results of the previous analyses, we performed three additional reconstructions of the ancestral genome size of Cuscuta. Each one included as outgroup candidates of Convolvulaceae with genome size data available of the two sister groups resolved by Stefanović and Olmstead (2004) as the closest relatives of Cuscuta. One of the analyses used two species of Calystegia R.Br. and six of Convolvulus L. (Convolvuleae, clade 1), another one Dichondra repens J.R.Forst. & G.Forst. (Dichondreae, clade 2), and the third one all of them.

The reconstruction of ancestral number and positions of the 5S and 35S rDNA sites were performed using Mesquite version 2.75 (Maddison and Maddison, 2018) on the 18 species for which rDNA information was available, 10 of them newly generated for this paper. Both the number and position of sites were transformed into categorical data (discrete characters): centromeric/pericentromeric position, interstitial, terminal/subterminal, and “mix” (when more than one of the previous conditions occurs in the same karyotype) as proposed by García et al. (2017). For the number of 5S sites, the characters were categorized as 1, 2, 3, 5, or 18 pairs. The number of 35S sites was categorized as 1, 2, 5, or 15 pairs of sites. Ancestral character states were inferred using maximum likelihood (Vaio et al., 2013). Due to the inconclusive results obtained in the rDNA sites number reconstruction using Mesquite, a second reconstruction was conducted using the Bayesian Binary MCMC (BBM) tool (Ali et al., 2012) implemented in the software Reconstruct Ancestral State in Phylogenies—RASP 4.2 (Yu et al., 2015, 2020) using the default parameters.

In total, 57 species of Cuscuta with DNA sequences and cytogenetic data available were sampled for the phylogenetic reconstruction, representing approximately 30% of the ca. 200 known species of the genus. The four subgenera were recovered as monophyletic each. Subgenus Monogynella was represented by four out of 15 species (26.67%), Cuscuta by eight out of 22 (36.36%), Pachystigma by three out of five (60%), and Grammica by 42 out of 150 (28.6%). The phylogenetic relationships obtained in this study were consistent with those based on a larger dataset reported by García et al. (2014), in which each of the four subgenera was strongly supported as monophyletic, with subgenus Monogynella as sister to the rest. The trees resolved the same relationships between the sections of subgenera Cuscuta and Grammica as those obtained by García et al. (2014; Supplementary Figure 1). The phylogenetic position of C. globosa was resolved as a member of section Gracillimae (clade N). Based solely on the description of the type, this species had previously been included in section Racemosae (clade C; Costea et al., 2015a). Here, it is placed in section Gracillimae based on molecular data, in agreement with the morphological features studied in the type and our new collections of this species.

Most Cuscuta species with published chromosome number are diploids with up to 2n = 38 (40 species plus two varieties, Table 2). Another 11 species are polyploids, mostly with 2n = 60. Of the species included in this study, four of them are known to have diploid and tetraploid populations: C. planiflora, C. epithymum, C. cotijana (2n = 60, a new cytotype; Supplementary Figure 2), and C. denticulata. Two new counts were included in this work, C. partita (2n = 30) and C. globosa (2n = 90; Supplementary Figure 2). The majority of polyploids belongs to subgenus Grammica, with the exception of some tetraploids and hexaploids of subgenus Cuscuta, such as C. approximata (2n = 28) or C. epilinum Weihe (2n = 42), considering a lower basic number for this subgenus (see below). The smallest number found was 2n = 8 in C. babylonica Aucher ex Choisy (Cuscuta), while the largest number was 2n = 150 in C. sandwichiana (Grammica). Cuscuta epithymum (Cuscuta) presents numerical intraspecific variation with 2n = 14, 16, 28, 30, 32, and 34.

For seven species, 5S and 35S rDNA site number and location were previously published (Table 1; Guerra and García, 2004; Ibiapino et al., 2019, 2020, 2022). New rDNA data were obtained for 10 additional species, C. australis (2n = 30), C. campestris Yunck. (2n = 56), C. epithymum (cytotype with 2n = 30), C. howelliana P. Rubtzoff (2n = 30), C. partita (2n = 30), C. psorothamnensis Stefanović, M. A. García & Costea (2n = 60), C. racemosa (2n = 60), C. sandwichiana (2n = 150), C. globosa (2n = 90), and C. glomerata (2n = 30). All rDNA sites in Cuscuta were colocalized with CMA⁺ bands. Most species presented at least one pair of CMA⁺/DAPI⁻ bands colocalized with nucleolus organizer regions (NOR) in proximal regions. Interstitial bands, when present, were weaker and smaller. Only in the holocentric C. epithymum did these bands occur in the terminal regions and were present in most chromosomes (Figures 1, 2). Most species showed only one pair of 5S and one pair of 35S rDNA sites. Usually, 5S sites occurred in interstitial regions, while 35S sites in pericentromeric regions, such as in C. australis, C. howelliana, C. partita, and C. glomerata (Figure 1). When more than one pair of 5S rDNA was present, these sites were also in interstitial regions, whereas when there were more than one pair of 35S, the extra pairs were interstitial, as in C. campestris, C. racemosa, C. psorothamnensis, C. globosa, and C. sandwichiana (Figure 2), or terminally located, as in C. epithymum (Figure 1). The largest number of rDNA sites in Cuscuta species was observed in C. monogyna, with approximately 18 pairs of 5S and 15 pairs of 35S rDNA. Information on the number and distribution of the rDNA in Cuscuta can be found in Table 1.

As for genome size, Cuscuta species varied from 1C = 0.27 Gbp in C. australis (2n = 30) to 1C = 34.73 Gbp in C. reflexa (2n = 32), both diploids. This variation represents the lowest and highest genome sizes known for Convolvulaceae (Figure 3). Some species showed infraspecific variation in genome size, such as C. gronovii Wild. ex Roem. & Schult., with five different values reported, ranging from 1C = 2.14 Gbp to 1C = 6.75 Gbp (Table 1).

The chromosome number reconstruction performed in ChromEvol with the best model, BASE_NUM_DUPL, indicated n = 7 as the basic ancestral number in Cuscuta (Supplementary Figure 3). This model considers five parameters, the rates of gains and losses of single chromosomes, duplications, in addition to considering a specific chromosome number that characterizes a phylogenetically close group, and the number variation rate. Based on this model, the variation in chromosome number in Cuscuta is most often related to duplication events [with a probability of frequency (f = 9.2)], followed by chromosome gains (f = 8.3) and losses (f = 7). The number n = 15 was indicated as ancestor of the subgenera Grammica and Monogynella. The ancestral number of subgenus Cuscuta was n = 7, and Pachystigma had n = 14, with 50% probability, but n = 7 was also very likely, with 40% probability (Supplementary Figure 3).

The holocentric clade (subgenus Cuscuta) includes species with intraspecific numerical variation. To test if its holocentric nature and high chromosome number variation influenced the analysis, we ran ChromEvol without the holocentric clade, following the same parameters described above (Supplementary Figure 4). The basic number in this analysis was n = 15 (x = 15), again with chromosome duplication (f = 7), followed by chromosome gains (f = 5.6) and chromosome losses (f = 3.7) as the main evolutive events. The reconstructed ancestral number for the remaining three subgenera was conserved as n = 15 (Supplementary Figure 4). Therefore, we repeated ChromEvol analysis fixing n = 15 at the base of the genus (Figure 4). In this scenario, numerical changes were mainly due to chromosome losses (f = 14.9), followed by duplications (f = 7.6) and chromosome gains (f = 6.6). The reconstructed ancestral numbers for Grammica and Monogynella were also n = 15. In subgenus Cuscuta, the basic number was n = 7 and the basic number for Pachystigma was n = 14 (Figure 4).

Mesquite and PastML also recovered n = 15 as the basic chromosome number for the whole genus. Furthermore, the maximum likelihood analysis suggested n = 15 for the subgenera Monogynella and Grammica. For subgenus Pachystigma, the most likely basic number reconstructed was n = 14 (with 50% probability), but n = 15 was also likely (with 43% probability). Only subgenus Cuscuta had three possible basic numbers depending on the analysis. The Mesquite analysis retrieved n = 7 and 14 as the most likely, each with approximately 31% probability (Supplementary Figure 5). The PastML JOINT approach reconstructed n = 15 as the ancestral number for both the entire genus and each of its four subgenera (data not shown).

Considering the smallest genome of 1C = 0.27 Gbp and the largest of 1C = 34.73 Gbp, the reconstructions of ancestral genome size made with phytools suggested that the ancestral genome of Cuscuta would be of intermediate size, approximately 20 Gbp without outgroups (Supplementary Figure 6), and 1C = 12 Gbp, including available outgroups. The results obtained using Dichondra or Calystegia and Convolvulus or these three genera as outgroup were similar (14, 12, and 12 Gbp, respectively; Supplementary Figure 6). Genome size increased in subgenus Monogynella, while it decreased in the other subgenera. Only C. indecora showed a massive expansion of genome size within subgenus Grammica (Figure 5), what was also observed with Mesquite (data not shown). Although there is no estimation available for subgenus Pachystigma, the long chromosomes of their bimodal karyotypes also suggest an increase in genome size, especially for C. angulata (Ibiapino et al., 2022).

The reconstruction of the ancestral number of rDNA sites made by Mesquite was inconclusive (data not shown). The analysis performed using RASP reconstructed 18 pairs of sites as ancestral for 5S rDNA and 15 pairs for 35S rDNA, but it is probably due to the presence of numerous sites in C. monogyna (Supplementary Figure 7). The reconstruction of rDNA site positions indicated that the ancestral position of the 5S rDNA was likely interstitial. The reconstructed position of the 35S rDNA site was inconclusive, with C. monogyna showing terminal and interstitial sites. Interstitial 5S rDNA sites were maintained throughout the genus. Only in C. indecora additional sites in terminal positions appeared. The 35S rDNA was reconstructed as terminal in the subgenus Cuscuta and peri/centromeric in Grammica. In this latter subgenus, species with only a pair of 35S sites had them always in peri/centromeric position. When more than one pair of 35S rDNA was present, the extra sites were inferred to have originated in interstitial positions (Figure 6).

Chromosome number variation across the entire genus Cuscuta was almost 19-fold between 2n = 8 in C. babylonica and 2n = 150 in C. sandwichiana (Pazy and Plitmann, 2002; García et al., 2019). The subgenus Cuscuta alone showed a variation of over 5-fold (from 2n = 8 to 2n = 42). The variation within this subgenus is also associated with intraspecific numerical variation, found in C. epithymum (2n = 14, 16, 28, 30, 32, and 34) and C. planiflora (2n = 14, 26, 28, and 34; García and Castroviejo, 2003). Subgenus Grammica species diversity was less represented with only ca. 30% of its ca. 150 species (Stefanović et al., 2007; Costea et al., 2015a). However, the variation in chromosome numbers found was similar to that observed in subgenus Cuscuta, just over 5-fold (2n = 28 to 2n = 150). In this case, most of the variation is attributable to auto- or allopolyploidy and only a few species have a chromosome number that is not n = 15 or a multiple thereof. The additional numbers can be explained by ascending or descending dysploidy. The relative low proportion of species studied suggests that the variation in this subgenus may be underestimated. Hybrid speciation was frequent in subgenus Grammica (e.g., Stefanović and Costea, 2008; Costea and Stefanović, 2010; García et al., 2014; Costea et al., 2015b) and a more detailed sampling will probably reveal additional cases of both auto- and allopolyploidy, as was recently reported for the small section Denticulatae (clade E; García et al., 2018).

In subgenera Monogynella and Pachystigma, no large variation in chromosome number was found. In this work, Monogynella is represented by five of its 15 species and the most common chromosome numbers found were 2n = 28, 30, and 32. Two additional species of the subgenus without sequence data were reported to have 2n = 32 chromosomes, C. gigantea Griff. (Khatoon and Ali, 1993) and C. sharmanum Mujerjee & P. K. Bhattach. (Mukerjee and Bhattacharya, 1970). Only C. reflexa has been reported to have a great intraspecific variation of chromosome numbers with triploid, tetraploid, and dysploid or aneuploid individuals (Kaul and Bhan, 1977). The small subgenus Pachystigma, represented by three of its total of five species (60%), has 2n = 28 and 30. The species of this subgenus have bimodal karyotypes; however, both the number of large and small pairs and the distribution of heterochromatic bands are quite variable (Ibiapino et al., 2022).

We used ChromEvol to reconstruct the basic ancestral chromosome number, which indicated the BASE_NUM_DUPL model as the best for our dataset. With this model, the basic ancestral number for Cuscuta was x = 7, as previously suggested by Fogelberg (1938) and García and Castroviejo (2003). However, most of the chromosome numbers reported in Cuscuta are multiples of 15, because 32 species (55%) are diploids with 2n = 30 and, among polyploids, 2n = 60 is the most frequent number. The highest numbers reported are 2n = 90 and 150, which are also multiples of 15. Additional analyses of both Mesquite and Past ML corroborated the number x = 15 at the base of the genus Cuscuta.

According to the Chromosome Count Data Base (Rice et al., 2015), the most frequent chromosome number in other Convolvulaceae is n = 15. Because of the accelerated rates of sequence evolution in Cuscuta, the extant sister group of the genus within Convolvulaceae could not be ascertained, but at least two nonparasitic lineages diverged before Cuscuta (Stefanović and Olmstead, 2004). Chromosome numbers in these lineages are known for a couple of genera in tribe Cardiochlamyeae: n = 13 for Poranopsis Roberty and n = 14 for Dinetus Sweet. Among more closely related lineages chromosome numbers are mostly n = 14 or 15, but some genera show greater diversity in chromosome numbers such as Convolvulus L. (n = 9–30), Merremia Endl. (n = 7–29), or Ipomoea L. (n = 14–45). Altogether, it is unlikely that, in Cuscuta, the ancestral number has reduced to x = 7, followed by independent chromosome duplications and gains in Monogynella and Pachystigma + Grammica. Therefore, we consider x = 15 more likely for the genus Cuscuta.

One reason for the inference of x = 7 for the genus is probably the presence of lower numbers in the subgenus Cuscuta. However, this subgenus is exclusively holocentric, and this chromosome type may go through karyotypic changes that are different than those in the other subgenera. Chromosome fusion and fission events can be favored in this karyotype type, as they have a diffuse kinetochore, which facilitates these types of rearrangements (Mandrioli and Manicardi, 2020). In groups where different evolutionary dynamics occur, it is necessary to consider clade specific models, that is, different parts of the phylogeny evolving according to different transition patterns of changes in chromosome numbers (Mayrose and Lysak, 2021). Márquez-Corro et al. (2019) used different methodological approaches to identify diverse patterns of chromosomal evolution in some clades of Cyperaceae. In that case, both a complete tree and subtrees were analyzed, suggesting several evolutionary model transitions in the entire phylogeny of the family. This type of analysis is particularly relevant when applied to the study of clades containing species with holocentric chromosomes, whose karyotypes can exhibit heterogeneous evolution modes. Therefore, we removed the holocentric clade to evaluate the reconstruction. In this analysis, the basic number x = 15 was retrieved for both the genus as a whole, and in each of the remaining subgenera, with duplication events as the most frequent resulting in the formation of polyploids. This corroborated the idea that the evolutionary dynamics of holocentrics are indeed different, and it exerted a large influence on the reconstruction of chromosome numbers in Cuscuta.

We therefore considered n = 15 fixed at the base of the genus Cuscuta as the best model to explain the evolution of chromosome numbers of this genus. In this model, seven chromosome losses occurred leading to x = 7 after the transition from monocentric to holocentric chromosomes in the subgenus Cuscuta. For the holocentric species, the chromosome numbers found were 2n = 8, 10, 14, 16, 18, 20, 26, 28, 30, 32, 34, and 42. Among the three sections of subgenus Cuscuta recognized by Costea et al. (2015a), the lowest chromosome number, 2n = 8, was found in C. babylonica, the only species of the monotypic section Babylonicae. This section is generally recovered as sister to section Epistigma, a group of five, mostly Asian species with chromosome numbers known in C. pedicellata Ledeb. and C. pulchella Engelm., both with 2n = 10, and C. capitata Roxb., probably a tetraploid with 2n = 20. The lineage of sections Epistigma and Babylonicae may have reduced their chromosome number by descending dysploidy through chromosome fusions. In this lineage, the biggest chromosomes are in C. babylonica, which shows the lowest chromosome number.

In the sister lineage, section Cuscuta of the subgenus Cuscuta, diploid species have at least 2n = 14 chromosomes, several species such as C. approximata and C. palaestina Boiss. are tetraploids (2n = 28), and C. epilinum is a hexaploid (2n = 42). Other diploids, not included in our analyses, have higher chromosome numbers, such as 2n = 18 and 20 reported for C. nivea M.A. García (García, 2001). Two species, C. epithymum and C. planiflora, are known to have diploid and polyploid populations with further variation in chromosome numbers that may have been generated from duplication and ascending dysploidy events. In C. epithymum (2n = 14, 16, 28, 30, 32, and 34), there are diploid and tetraploid cytotypes with both bimodal and symmetrical karyotypes. Some cytotypes with 2n = 14 and 2n = 32 are bimodal, whereas others with 2n = 16 and 2n = 34 are symmetrical (García and Castroviejo, 2003; García et al., 2019), suggesting that chromosome fusion or fission events, together with polyploidy, engender this numerical variation. The cytotypes with asymmetrical karyotypes in C. epithymum and the other species of subgenus Cuscuta have the active NORs located in the longest chromosomes and with hematoxylin staining they are observed associated to the nucleoles. In C. planiflora (2n = 14, 26, 28, and 34), the populations with 2n = 34 have an asymmetrical karyotype that together with morphological features indicate that it is an allopolyploid (García, 2001). Both, C. epithymum and C. planiflora are taxonomically difficult as revealed by the high number of infraspecific taxa that have been described (García and Martín, 2007), and their variation in chromosome numbers and karyotypes may indicate cryptic diversity in these species’ complexes.

Genome size data for 28 species of Cuscuta (Table 1, Figure 5) revealed a 128-fold variation between the smallest (1C = 0.27 Gbp in C. australis) and the largest genome (1C = 34.73 Gbp in C. reflexa). This tremendous variation in Cuscuta does not seem to be mainly caused by polyploidy events, despite its high frequency in the genus (García et al., 2019; Ibiapino et al., 2019). Polyploids with higher chromosome numbers, such as C. globosa (2n = 90) and C. sandwichiana (2n = 150), both belonging to subgenus Grammica, had small genome sizes, 1C = 1.79 Gbp and 1C = 1.8 Gbp, respectively. Cuscuta, therefore, fits with the general trend of genome downsizing of polyploids observed in angiosperms (Leitch and Leitch, 2008). Subsequent genome downsizing in polyploids is also known from other parasitic lineages (e.g., in Orobanche; Weiss-Schneeweiss et al., 2006).

It is remarkable that the largest genome sizes are found in diploid species such as C. lupuliformis Krock. (2n = 28) and C. reflexa (2n = 32), both belonging to subgenus Monogynella, with 1C = 21.97 Gbp and 1C = 34.73 Gbp, respectively. Species such as C. monogyna (1C = 33.05 Gbp, subgenus Monogynella) and C. indecora (1C = 24.46 Gbp, subgenus Grammica), both diploids, have numerous heterochromatic bands along their chromosomes, all these bands co-localizing with repetitive DNA such as 5S and 35S rDNA or satellite DNAs, indicating that the amplification of repetitive sequences in heterochromatin is involved (Ibiapino et al., 2020; Naumann et al., 2020). However, there is also an accumulation of heterochromatic bands in subgenus Cuscuta, but there is no drastic increase in genome size probably because of the reduction in the number of chromosomes associated with the transition to holocentric chromosomes. In the diploid C. europaea L., for example, the genome size is, on average, 1C = 1.11 Gbp (McNeal et al., 2007; Neumann et al., 2020). This species has satellite DNAs that occupy a large extension of all 14 chromosomes, which varies in sizes from 2.76 to 6.70 μm, approximately (estimated measurements based on mitotic metaphases reported in Oliveira et al., 2020). In subgenus Pachystigma, the differential accumulation of repeats in just a few chromosomes led to the appearance of bimodal karyotypes. Cuscuta nitida E. Mey. ex Choisy (2n = 28), for example, has an accumulation of different classes of repetitive DNA in only two chromosome pairs, which are, in average, 12.34 and 8.19 μm long, compared to 2.67 μm of the smallest pairs which is not enriched with repetitive DNA (Ibiapino et al., 2022). Thus, the accumulation of repetitive DNA can lead to an increase in chromosomes and, consequently, to an increase in genome size, especially in subgenus Monogynella.

Although genome size has been estimated for just a few species, the relative chromosome size may be an indirect indicator of genome size if we compare species with the same ploidy level. The smallest genome size in the genus estimated for the diploid C. australis (2n = 30; 1C = 0.27 Gbp) correlates well with the small chromosome size observed in mitotic metaphases (Figures 1A,B). Other diploids with the same chromosome number and similar chromosome size probably have a similar genome size. Such is the case of C. desmouliniana Yunck. (García et al., 2019) for which no genome size has been measured, but the size of chromosomes is similar or even smaller than in C. australis.

In the phylogeny of Cuscuta, some clades include species with noticeable differences in chromosome size. In subgenus Pachystigma, the number of large chromosomes varies from two pairs in C. nitida to five pairs in C. angulata suggesting a great difference in genome size within the subgenus. Further analysis of the two unsampled species from Eastern South Africa might reveal an even greater variation. Subgenus Grammica, accounting for ca. 70% of the Cuscuta species diversity (Stefanović et al., 2007), shows several clades with significant differences in chromosome and genome size between closely related species. A remarkable example is the strongly supported clade encompassing sections Umbellatae, Indecorae, and Gracillimae (clades L, M, N; García et al., 2014; Costea et al., 2015a), which has species with very small (C. desmouliniana, clade L), intermediate (C. sidarum and C. globosa, clade M), and large chromosomes (C. coryli Engelm. and C. indecora, clade M). The branch leading to species in sect. Indecorae is noticeably longer compared to others in the subgenus, showing an increase in mutation rates together with the increase of the genome size.

Although not as evident as in subgenus Pachystigma, some sections of subgenus Grammica (Costea et al., 2015a) appear to have a prevalence of asymmetrical karyotypes suggesting differential accumulation of DNA is some chromosome pairs (García et al., 2019) and probably to an increase in genome size. In section Ceratophorae (clade K; Costea et al., 2011), from which only diploid species are known, C. costaricensis Yunck. has one pair of chromosomes noticeably longer than the others, whereas in C. bonafortunae Costea & I. García or C. erosa Yunck. about half are long and half are shorter. Similar karyotypes have been documented in allopolyploids such as C. veatchii, but in these cases, the asymmetry is a consequence of the subgenomes of the two diploid parents with differences in chromosome size (Ibiapino et al., 2019).

Whereas some sections of subgenus Grammica show a tendency to increase genome size, others have undergone a significant reduction. The expansion and diversification of the genus in North America resulted in two lineages (Stefanović et al., 2007; García et al., 2014) having opposing directions in the evolution of genome size. The lineage of sections Oxycarpae (clade D) and Denticulatae (clade E) shows genome upsizing, which is especially evident in the former, including species with the biggest genomes in subgenus Grammica except for sect. Indecorae. On the contrary, the lineage that includes sections Californicae (clade A) and Cleistogrammica (clade B) has the smallest genomes not only in Cuscuta but also in Convolvulaceae. The only species of section Racemosae (clade C) with known karyotype and genome is C. racemosa (2n = 60; 1C = 1.39), having a relatively small genome compared to other tetraploids. Further sampling in this section will reveal whether the dispersal to South America before the diversification of this clade was accompanied by a significant variation in genome size.

Table 1 also summarizes the great intraspecific variation in the genome size of species of sections Oxycarpae and Indecorae (Costea et al., 2015a). Within Oxycarpae, differences such as in C. compacta (1C = 3.24 and 7.67) suggest that there might be diploid and tetraploid populations. In C. gronovii, all the chromosome counts are 2n = 60; however, the genome size reported for the species ranges from 1C = 2.14 and 1C = 6.75 Gbp. Taxonomy of the section in general and of C. gronovii in particular is difficult and these differences might be explained by species identification mistakes or might reflect the morphological diversity of the species, for which several varieties have been described (Yuncker, 1932; Costea et al., 2006a). The strong support for the monophyly of the section contrasts with the very short internal branch lengths (Stefanović et al., 2007) suggesting a recent and rapid diversification accompanied by the accumulation of repetitive DNA with rates that might be different even at population level. A similar case is C. indecora in section Indecorae, for which several varieties have been described in addition to other two species, C. coryli and C. warneri Yunck. (Yuncker, 1932; Costea et al., 2006b).

Neumann et al. (2020) suggested that there is no correlation between genome size and the parasitic lifestyle of Cuscuta. However, according to the Plant DNA C-value Database, other species of Convolvulaceae have small genome sizes, with an average of approximately 1C = 0.97 Gbp and the largest value reported for the hexaploid Ipomoea batatas (L.) Lam. (2n = 90; 1C = 2.25 Gbp). Our estimation of the ancestral genome size of Cuscuta was ca. 1C = 12 Gbp, and 1C = 20 Gbp when including only Cuscuta species in the analyses. The phylogenetic position of subgenus Monogynella as sister to the rest of the genus, with such large genome sizes, may have resulted in an overestimation of the ancestral genome size for the genus as a whole. In spite of this possible bias, based on the limited Convolvulaceae data available, there was probably a genome expansion in subgenus Monogynella, and independent increases in other lineages, especially in sections Indecorae and Oxycarpae of subgenus Grammica.

The genome constraint hypothesis (Knight et al., 2005) suggests that the costs associated with the accumulation and replication of repetitive DNA reduce plant performance and negatively affects speciation and the distribution and abundance of species. Parasitic lifestyle eliminates the restrictions imposed by the growth rate of the meristem or the “genomic economy,” because they take resources from their hosts (Gruner et al., 2010; Piednoël et al., 2012). Thus, despite the genomic reductions associated with the loss of autotrophic functions (Revill et al., 2005; Banerjee and Stefanović, 2019), there is a tendency for parasitic plants to have larger and more complex genomes (reviewed by Lyko and Wicke, 2021). For example, in Orobanchaceae, the genomes of the autotrophic Lindenbergia philippensis (Cham. & Schltdl.) Benth. and the hemiparasite Schwalbea americana L. are much smaller than those of the holoparasite Orobanche L. and Phelipanche Pomel, which fits the hypothesis of larger genome sizes in parasitic plants (Gruner et al., 2010; Piednoël et al., 2012). Without the selective constraints imposed by the nutrient and energy economy (Gruner et al., 2010; Piednoël et al., 2012), genome size could vary via mechanisms such as mobile elements activation and ectopic recombination. The amplification and diversity of transposable elements can be influenced by the DNA transposition and elimination rates, population size, reproduction mode, host plant defense mechanism, and even horizontal gene transfer (Devos et al., 2002; Bourque et al., 2018; Biscotti et al., 2019; Nishihara, 2020). This variation may or may not be fixed by the action of genetic drift and natural selection and could possibly allow a wider range of variation including the upper limits of genome size in Cuscuta, which are usually selected against in green plants.

Unlike in Orobanchaceae, genome size increases in Cuscuta are, however, not strictly related to the evolution to holoparasitism within the genus. Subgenus Monogynella has the least reduced plastome and higher ability for carbon fixation than the rest of subgenera; however, it has the largest genomes. Different levels of plastome reduction have been documented for the genus (Banerjee and Stefanović, 2020), and fully holoparasitic species are known in section Ceratophorae (subgenus Grammica, clade K; Banerjee and Stefanović, 2019). Although no genome size estimations have been done for the section, the chromosomes are not significantly bigger than in other related sections except for an apparent trend toward asymmetrical karyotypes. Species of sect. Subulatae (clade O) are also holoparasitic (Braukmann et al., 2013) and the only known genome sizes for the section are those of C. chilensis Ker Gawl. and C. purpurata Phil., two diploids with intermediate genome size (1C = 2.80 and 2.96, respectively).

Smaller genomes theoretically facilitate faster cell divisions and therefore growth (Gruner et al., 2010) and cell division rates (Símova and Herben, 2012). In Cuscuta, however, there is no clear negative correlation between genome size and growth rate, possibly because it may also be influenced by other factors, such as cell elongation. Cuscuta indecora, with the largest known genome in subgenus Grammica, is an invasive weed and a seed contaminant of crop plants, with profuse and fast growth over its hosts (Cudney et al., 1992; Costea et al., 2006b). Although we have not performed specific comparative experiments, we did not observe any evident difference in growth rate between this species and, e.g., C. australis, the species with the smallest genome known in the genus, even though both were growing on the same host in the same greenhouse conditions. It is remarkable that C. indecora behaves as a fast-growing weed, despite theoretically having longer cell cycles in which the whole genome must be replicated. Cuscuta indecora and weedy species of section Oxycarpae, such as C. gronovii, might be model systems to study the correlation between higher metabolic activity necessary to maintain the growth rate and the accumulation of repetitive DNA and transposable elements contributing to genome upsizing.

Most Cuscuta species have a few rDNA sites: only one pair of 5S and one pair of 35S rDNA sites. Although the number of 5S and 35S rDNA loci is positively correlated with ploidy level (García et al., 2017), this does not hold true for this genus. Cuscuta sandwichiana, the highest polyploid reported, has one pair of 5S and two pairs of 35S rDNA sites, while phylogenetically close diploids, such as C. australis, have one pair of 5S and one pair of 35S rDNA sites. This may be due to the fact that some Cuscuta polyploids are interspecific hybrids, such as C. sandwichiana (Stefanović and Costea, 2008; García et al., 2014). The occurrence of recombination and gene conversion that results in the presence of rDNA copies from only one of the parents is often observed in hybrids. In allotetraploids of the Dilatata group of the genus Paspalum L. (Poaceae), for instance, the recovered ITS sequences show homogenization toward the paternal genome only (Vaio et al., 2019). In addition, the decrease in the number of expected sites may occur due to the elimination of some sites in terminal regions. The terminal position of the rDNA sites would be selectively favorable compared to the proximal ones, as it would reduce the chances of deleterious chromosomal rearrangements related to unequal recombination and recombination between non-homologous chromosomes (Roa and Guerra, 2012; García et al., 2017). Besides, the number of parental rDNA sites may be quite conserved in young, artificial allopolyploids. However, in natural allopolyploids, this number is often reduced, especially the 5S rDNA sites (Lee et al., 2011; Volkov, 2017). In C. veatchii, for example, there is a reduction in the rDNA sites in relation to its parents C. denticulata and C. nevadensis I.M. Johnst., indicating an old origin of this hybrid (Ibiapino et al., 2019).

It is common for the 5S and 35S rDNA sites to be found at separate locations in the genome, even on different chromosomes. This may be related to the fact that they are transcribed in different cellular compartments, by different enzymes (García and Kovařík, 2013). However, in Cuscuta, many species had rDNA sites located on the same chromosome, across all subgenera. In C. monogyna (subgenus Monogynella, sister to the rest of the genus), almost all its 30 chromosomes had 5S and 35S rDNA sites positioned closely, both on the same chromosome arm and on different arms. Co-occurrence of 5S and 35S rDNA sites on the same chromosome is higher in karyotypes with multiple sites and is frequently on the same arm (Roa and Guerra, 2015; García et al., 2017). However, all other Cuscuta species (e.g., C. veatchii and C. indecora) with sites on the same chromosome have these sites positioned on the same arm.

In plants, 5S rDNA usually occupies proximal and less frequently interstitial and terminal regions, while 35S rDNA tends to occupy terminal regions (Roa and Guerra, 2012, 2015). In Cuscuta, while 5S rDNA was more frequently found in interstitial regions, 35S is frequently found on peri/centromeric regions. In the only two holocentric species of Cuscuta whose rDNA sites are reported in this work, the positions of these sites also diverged from that found in other groups of plants. Generally, both 5S rDNA and 35S rDNA occupy terminal regions in holocentrics (Roa and Guerra, 2012, 2015). In holocentric Cuscuta species, only the 35S occupied a terminal position. The 5S rDNA sites were at more interstitial positions.

The ancestral character reconstruction suggested the interstitial position as ancestral for the 5S rDNA. This characteristic is present in all species, including C. monogyna (subgenus Monogynella) and C. indecora (subgenus Grammica), which also showed proximal and terminal sites, respectively. The 35S was more variable, and the “mix” condition, where rDNA sites were found in more than one location, was present at several clades throughout the phylogeny. The 35S rDNA is commonly pericentromeric; however, in species with more sites, additional sites are usually found in interstitial regions. In the subgenus Cuscuta, without a localized centromere, all 35S sites were terminal.

According to the Plant rDNA Database (García et al., 2017), within the Convolvulaceae, published rDNA site data are available only for seven species of Ipomoea, which is not closely related to Cuscuta and, thus, may not aid in resolving ancestral rDNA state. In the subgenus Monogynella, there is an increase in the diversity of positions in which the rDNA sites were found, suggesting that the “mix” condition is derived and, due to the amplification of these sites, the ribosomal DNAs began to occupy different positions along the chromosome. Recent studies show the possible influence of repetitive DNA amplification on genomic changes in the genus Cuscuta (Neumann et al., 2020; Ibiapino et al., 2022), indicating that the increase of these sites in Monogynella may be caused by the amplification of rDNA repeats as observed for other tandem repetitive sequences in the genus, and these are actually pseudogenes.

The tandem repetitive DNA in Cuscuta appears to be quite complex. For example, C. europaea have species-specific satellite DNA sequences such as the CUS-TR24, colocalized with centromeric proteins, also species-specific (Oliveira et al., 2020). Analysis using long reads showed a complex organization of CUS-TR24. The sequence of this satellite is interspersed with insertions mainly from LINE retrotransposons (Vondrak et al., 2021). In the subgenus Pachystigma, there is also evidence of these complex satellites. For example, the CnSat10-1,400 found in C. nitida in addition to being similar to a LINE type element, co-localizes with 35S signals in the chromosomes of this species. In addition, the most abundant SF1 family of C. nitida also co-locates with 35S signals (Ibiapino et al., 2022). This complex organization, with possible insertions of rDNA in other repetitive DNAs, could influence the diversity of number and position of rDNA sites in Cuscuta. In Allium cepa L., for example, it was shown that 35S rDNA is able to move from one locus to another in the genome. Furthermore, it is associated with telomeric DNA and other satellite DNAs, suggesting that the 35S of this species undergoes excision-reintegration mediated by these sequences (Mancia et al., 2015; Fu et al., 2019). Thus, the current data suggest that the evolution of rDNA in Cuscuta may be influenced by other tandem repeats or transposable elements.