Ploidy variation in Rhododendron subsection Maddenia and its implications for conservation

Abstract Polyploidy, which is common in plants, can confound taxon recognition and hence conservation assessments. In the taxonomically complex genus Rhododendron, 25 % of the over 1,300 taxa are considered under threat and 27 % Near Threatened or Data Deficient, with their taxonomy needing to be resolved urgently. Although ploidy levels of Rhododendron taxa range from diploid (2x) to dodecaploid (12x) according to previous reports, the extent of polyploidy across the genus has not been examined. We first summarized the taxonomic distribution of polyploids in the genus based on the literature. Then as a case study, we estimated ploidy levels of 47 taxa in subsection Maddenia (subgenus Rhododendron, section Rhododendron) using flow cytometry, together with verification of meiotic chromosome counts for representative taxa. The summary of reported ploidy in Rhododendron indicates that polyploidy is most common in subgenera Pentanthera and Rhododendron. In subsection Maddenia, all examined taxa are diploids except for the R. maddenii complex that shows a high ploidy variation (2–8x, 12x). We investigated ploidy level of 12 taxa in subsection Maddenia for the first time, and estimated genome sizes of two Rhododendron species. Knowledge of ploidy levels will inform phylogenetic analysis of unresolved species complexes. Overall, our study of subsection Maddenia provides a model for examining multiple issues including taxonomic complexity, ploidy variation and geographic distribution in relation to biodiversity conservation.


Introduction
Polyploidization, or whole-genome duplication (WGD), generates organisms containing multiple sets of chromosomes. This major mechanism of plant speciation results from either intraspecific genome duplication (autopolyploidy) or hybridization between different species and chromosome doubling (allopolyploidy) (Stebbins 1947; Van de Peer et al. 2017). Fertile polyploids can become new species when strong reproductive incompatibilities and distinct phenotypic differences occur, differentiating them from their diploid progenitors. Polyploidization, accompanied by corresponding morphological differences, has been considered as a characteristic for recognition of species which form conservation units (Soltis et al. 2007(Soltis et al. , 2010Laport and Ng 2017). Due to multiple copies of genes facilitating adaptive processes, polyploids may be more successful at adapting to new environments (Comai 2005; Van de Peer et al. 2017. As ploidy variation can be associated with regional biodiversity, it should be included in diversity measurements (e.g. phenotypic, inter-and/or intraspecific diversity) for the consideration of conservation, especially in temperate regions where polyploidization is frequently observed (Comai 2005;Laport and Ng 2017;Rice et al. 2019).
Rhododendron L. (Ericaceae) is a megadiverse genus with more than 1,300 taxa [species, subspecies (ssp.) and varieties (var.)] that typically grow in temperate regions (Gibbs et al. 2011). The wild distribution of Rhododendron covers a geographic range from the centres of diversity in the south-eastern Himalayas and Malay Archipelago to North America, Europe and northern Australia (Gibbs et al. 2011;Argent 2015;MacKay et al. 2018;Shrestha et al. 2018). The Himalayan region is characterized by a rich biotic assembly (Ming and Fang 1979;Yan et al. 2015;Hughes 2017;Shrestha et al. 2018), where polyploids are likely to diversify under environmental stress (Rice et al. 2019; Van de Peer et al. 2021). Extensive hybridization, due to weak reproductive barriers within Rhododendron, is also a possible cause of rapid speciation (Frodin 2004;Zhang et al. 2007;Zha et al. 2008;Soltis and Soltis 2009;Ma et al. 2010;Qiu et al. 2020). However, hybridization and polyploidy and their influence on speciation rate in Rhododendron are still under investigation (Milne et al. 2010;Schwery et al. 2015;Shrestha et al. 2018;Khan et al. 2021). Taxa produced from introgression of sympatric species often show morphological similarity to their parents, making correct taxon identification challenging (Darlington et al. 1955;Milne et al. 2010;Zhang et al. 2020).
Effective decisions and strategies for species conservation require distinct taxonomy to assess the risk of extinction of species. It has been reported that 25 % of Rhododendron taxa are under threat (Critically Endangered, Endangered and Vulnerable), and 27 % Near Threatened or Data Deficient . However, problems of taxon identification due to taxonomic complexity and lack of cytogenetic knowledge of particular accessions still need to be resolved to inform conservation strategies (Mehra 1976;Jones et al. 2007;Gibbs et al. 2011;Mao et al. 2017;Khan et al. 2021). Molecular phylogenetic techniques can assist to resolve taxonomic uncertainties (Gibbs et al. 2011;Gardiner et al. 2019). However, the presence of polyploids can confound analyses due to duplicated genomes that are often derived from multiple species (Yan et al. 2015;Rothfels 2021). For plant genera that include polyploids, such as Rhododendron, any investigation of phylogeny should be preceded by an examination of ploidy levels in the taxa under consideration (Khan et al. 2021).
Cytological studies of Rhododendron species began in 1930 (Bowers 1930;Sax 1930), with the most extensive and genuswide chromosome counts reported in the 1950s (Ammal et al. 1950;Darlington et al. 1955). In Rhododendron species, mitotic chromosomes in the root tips are notably small and difficult to distinguish under the microscope (Jones et al. 2007;Zaytseva et al. 2018). This difficulty may increase for counting the multiple sets of chromosomes in polyploids (Comai 2005; Van de Peer et al. 2017). Meiotic chromosome number can be more easily determined by counting haploid chromosomes in pollen mother cells (PMCs) (Windham et al. 2020), but little information is available on optimal bud harvest time for meiotic observation in Rhododendron. In contrast, flow cytometry (FCM) saves time by enabling rapid determination of nuclear DNA content (genome size) for a large number of samples (Doležel et al. 1998;De et al. 2010;Zaytseva et al. 2018). Apart from the measurement of genome size (Bou Dagher-Kharrat et al. 2013;Khan et al. 2021;Choi et al. 2022), FCM has been applied to estimate Rhododendron ploidy levels in several studies (Doležel et al. 1998;De Schepper et al. 2001;Jones et al. 2007;Zhou et al. 2008;Zaytseva et al. 2018;Khan et al. 2021). Cytological studies and FCM generally require access to living material. However, this can be hindered due to difficulties in accessing remote Rhododendron habitats or living accessions on sites of ex situ collections internationally. For FCM at least, the use of dehydrated leaf tissues has proven to be reliable for ploidy estimation in other species (Suda and Trávníček 2006;Tomaszewska et al. 2021), but this approach has not yet been tested on Rhododendron.
Rhododendron is taxonomically complex, divided into nine subgenera (if considering Vireya as a subgenus) with further sections and subsections of varying sizes (Chamberlain et al. 1996;Frodin 2004;Argent 2015). Davidian (1982) systematically described the morphology of Rhododendron. Chamberlain et al. (1996) 'lumped' a number of previously recognised species as synonyms, which is considered the most comprehensive reference for the taxon checklist to date. In this study, we initially consider the whole genus, and then focus on subsection (ss.) Maddenia (subgenus Rhododendron, section Rhododendron) as a case study. Due to the complex taxonomy and continuous morphological variation within ss. Maddenia, many questions remain about species boundaries, which is identified as a general problem in the genus (Chamberlain et al. 1996;Cullen 2005;Gibbs et al. 2011;Donald 2012;MacKay 2018;Jamieson 2021). Ss. Maddenia is the largest among all the subsections in subgenus Rhododendron (if the vireyas are treated as subgenus Vireya rather than the broad section Schistanthe under subgenus Rhododendron), and several new species (Mao and Bhaumik 2015;Mao et al. 2017;Chang et al. 2021;Rushforth et al. 2022) have been published since Chamberlain et al. (1996). With 'lumping' species as synonyms and/or changes in placement of species over time, the number of accepted species in ss. Maddenia varies in treatments by different authors. In The Rhododendron Species, Davidian (1982) used the concept of 'series' rather than 'subsection' and listed a total of 56 species in the two series Ciliatum and Maddenii. In The Genus Rhododendron: Its classification & synonymy that we are following as the major reference for taxonomic classification, Chamberlain et al. (1996) listed 52 species (57 taxa) in ss. Maddenia. Cox and Cox (1997) included 31 species of ss. Maddenia in The Encyclopedia of Rhododendron Species, while Khan et al. (2021) defined 56 species in the phylogenetic study. In ss. Maddenia, the R. maddenii complex is especially problematic, with 12 taxa placed in synonymy under the two subspecies of R. maddenii (R. maddenii ssp. maddenii and R. maddenii ssp. crassum) (Chamberlain et al. 1996). Previous studies also found exceptional occurrence of polyploids (tetraploid, hexaploid, octoploid and dodecaploid) in R. maddenii, which raises questions on the uniqueness of this species complex and patterns of species diversification in ss. Maddenia (Ammal et al. 1950;Cubey 2003;Khan et al. 2021).
Previous phylogenetic studies indicate that ss. Maddenia may not be monophyletic (Donald 2012;Shrestha et al. 2018;Khan et al. 2021). However, two of these studies encompassed the whole genus, with few species included from ss. Maddenia (Shrestha et al. 2018;Khan et al. 2021), while Donald (2012) only considered the yellow-flowered species that include part of ss. Maddenia species. The species coverage as well as number of molecular markers applied in previous studies were limited, which does not provide adequate evidence on the relationships between ss. Maddenia and other possibly related species or subsections.
Subsection Maddenia includes rhododendrons that are all lepidote (scaly), originating from the Himalayan region through to southern China and northern Vietnam. The taxa not only present great morphological diversity ( Fig. 1) but also possess unique horticultural value because of their scented flowers and high resistance to thrips (Cullen 1980;MacKay et al. 2018;Jamieson 2021). According to conservation assessments for 51 ss. Maddenia taxa, 33 were placed in either a threatened category or Data Deficient (Gibbs et al. 2011;MacKay et al. 2018;Chang et al. 2021). Due to the variable taxonomy and species definitions derived from traditional morphology, conservation assessments and subsequent conservation action are subject to debate (Cubey 2003;Gibbs et al. 2011;Donald 2012;Jamieson 2021).
In this study, we collected leaf samples from living accessions to examine the ploidy variation in ss. Maddenia. Our methodology involved (i) reviewing the literature to establish a comprehensive dataset of ploidy levels across Rhododendron taxa; (ii) investigating the basic cytogenetics of ss. Maddenia by estimating ploidy, especially of the previously reported polyploid taxa in the R. maddenii complex and of taxa not previously studied. Flow cytometry was used for ploidy estimation, and when possible, confirmed with meiotic chromosome counts from living material. The resulting data will not only inform our future molecular phylogenetic studies but also assist in developing an understanding of the speciation and ecological features of ss. Maddenia for conservation management.

Literature review of ploidy in Rhododendron
Prior knowledge on ploidy levels of Rhododendron taxa was compiled in a spreadsheet, with data from chromosome counting (2n) and flow cytometry (x) both included but listed separately [see Supporting Information- Table S1]. Chromosome data were compiled from the Chromosome Counts Database (CCDB, Rice et al. 2015), Index to Plant Chromosome Numbers (IPCN, Goldblatt and Johnson 1979) and literature not covered by these two databases (Sax 1930;Ammal et al. 1950;Li 1957;Cubey 2003;Contreras et al. 2007). FCM ploidy data were collected from previous reports (De Schepper et al. 2001;Jones et al. 2007;De et al. 2010;Khan et al. 2021;Choi et al. 2022). Taxonomic classification (subgenus, section, subsection) followed Chamberlain et al. (1996), except that the vireya species were treated as subgenus Vireya following Argent (2015) (subgenus Vireya may be treated as the broad section Schistanthe under subgenus Rhododendron as in recent studies). Taxa together with their synonyms were indexed according to Chamberlain et al. (1996) except for those in subgenus Vireya (Argent 2015). We integrated the classification for the convenience of indexing taxon names and summarizing ploidy levels, as these two publications are the most comprehensive and latest references for the respective groups. Irrespective of the positioning of the Indomalesian species as subgenus Vireya or the broad section Schistanthe (Goetsch et al. 2011;Shrestha et al. 2018;Khan et al. 2021), the group of species involved is still largely as described by Argent (2015). Taxa were recorded as polyploids if polyploidy was reported from either chromosome counting or flow cytometry. Reported ploidy, chromosome number and proportion of polyploid taxa were summarized for each taxonomic group.
All taxa of ss. Maddenia that have been previously reported for ploidy, except for R. vanderbiltianum (Atkinson et al. 2000) and R. yungchangense (Cubey 2003), were included in the present study. Further to the taxa listed by Chamberlain et al. (1996), four taxa were separated from species complexes and analysed as distinct entities: R. iteophyllum , R. sinonuttallii (Gibbs et al. 2011), R. taronense (Gibbs et al. 2011) and R. valentinioides (ined.) (Gibbs et al. 2011;Donald 2012). Three new species published post-1996 were examined: R. pseudomaddenii (Mao and Bhaumik 2015), R. leptocladon (Rushforth and Nguyen 2019) and R. kuomeianum (Chang et al. 2021). R. vanderbiltianum was also included because of its suggested placement in this subsection (Argent et al. 2008;Donald 2012;MacKay et al. 2018). Individual accessions from New Zealand collections were identified following Davidian (1982), based on herbarium specimens and photographs taken in the field.
While fresh leaves were collected locally in New Zealand, overseas samples were silica-gel dried and imported from RBGE and RSBG to New Zealand. In both cases, fully expanded young leaves were routinely sampled for flow cytometry, although sometimes leaves from the previous season's growth were used when young leaves were unavailable. Fresh samples from the New Zealand sites were chilled and shipped overnight to the laboratory for FCM ploidy estimation. For some local accessions, dehydrated leaf samples from herbarium specimens were used, as fresh or silica gel-dried leaf materials were unavailable at the time of the FCM experiment [see Supporting Information- Table S2]. A subset of samples was replicated to verify the consistency of ploidy results for the same accession using fresh leaf vs. silica gel-dried leaf vs. dried leaf from an herbarium specimen.
Leaf tissue collected from R. fortunei was routinely used as the internal diploid standard for flow cytometry. When fresh leaves of R. fortunei were unexpectedly unavailable, R. parryae was used as the internal standard. Both species were previously reported as diploids (Ammal et al. 1950;Jones et al. 2007) and available as living plants for our sampling. Analysis of a subset of samples was repeated to verify the consistency of ploidy results for the same accession using either of the diploid standards. Genome sizes of these two Rhododendron species were measured [see Supporting Information- Table S2], using Pisum sativum L. (2C = 8.8 pg) and Zea mays L. (2C = 5.33 pg) as standards.

Flow cytometry preparation and analysis
Cell nuclei suspensions from leaf tissue were prepared for flow cytometry following Doležel et al. (2007) with minor modifications. Both fresh and dehydrated leaf samples were processed using the same protocol at Manaaki Whenua-Landcare Research (Lincoln, New Zealand). Approximately 1 cm 2 of leaf tissue of each sample was co-chopped with the diploid standard using a sharp razor blade in approximately 1 mL ice-cold Otto I buffer in a Petri dish, then left to incubate for 1-2 min at room temperature. The homogenate was filtered into a sample tube through a 20-µM nylon mesh to remove large particles. Next, DAPI (4ʹ6-diamidino-2-phenylindole) stain, prepared with Otto II solution, was added to the sample tube at 4 µg/mL. The samples were run on a Partec PAII flow cytometer, using a 375-nm UV laser and FloMax software, until the particle count was at least 5000. Histogram peaks were manually gated for all samples. The relative fluorescence values of the peak positions of DAPI-stained nuclei (Mean-x) and the coefficient of variation (CV-x %) of the RN1/RN2 peak were evaluated. Where there was any uncertainty to gate a peak, the sample was run alone, then co-chopped with the diploid standard to confirm the peak of the standard. Data were transferred to an Excel spreadsheet from which the FCM ploidy (x) was calculated [see Supporting Information- Table S2]. Genome size measurements of the two diploid standards of Rhododendron were performed using the same protocol and reagents, except that 10 µg/mL propidium iodide was used as the stain, and samples were run on a Partec CyFlow Space with a blue laser at 488 nm. Each plant sampled was measured sequentially in triplicate using leaves harvested on the same day, from which the average 2C value (pg) was calculated for the genome size.

Validation of FCM ploidy with meiotic chromosome counts
Chromosome counts were performed on a sub-sample of six accessions to validate the estimated ploidy levels from flow cytometry. Developing flower buds were harvested on sunny mornings in a local garden in New Zealand, mostly between 0730 and 0930 hours. Prior to the harvest, a series of observations were made to determine the correct stage of the meiotically dividing PMCs. Serial sampling of flower buds was performed during the growth of anthers. The outer layers of bud scales were removed before the buds were immersed in fixative (1 part of glacial acetic acid to 3 parts of absolute ethanol) for at least 24 h. Under a stereo microscope (Leica MZ9.5), young anthers were removed from the buds, mashed and stained with 1 % acetocarmine. After removing the debris, slides with a coverslip were heated to ~50 °C for ~30 s and set aside for 1.5-2 h for deeper staining, followed by the final 'squash' onto the slide. The meiotic chromosomes of PMCs were observed under a compound microscope (Leica DM500) and images were captured with 100× objective under oil immersion.

Taxonomic distribution of polyploidy in Rhododendron
Our summary of ploidy levels from existing databases and the literature demonstrated that polyploidy occurs in five of the nine subgenera of Rhododendron (considering Vireya as Table 1. Taxonomic distribution of polyploidy in Rhododendron. Values in parentheses indicate the number of (polyploid taxa)/(tested taxa)/(all taxa) in the corresponding groups. Ploidy data were combined from flow cytometry and chromosome counts. Taxa were counted as polyploids when there was an occurrence, even if diploids were also found. E.g., of the 38 taxa in subgenus Azaleastrum, 11 have been examined for ploidy, in which one taxon was reported with occurrence of polyploids. Taxonomic classification of genus Rhododendron is according to Chamberlain et al. (1996) and Argent (2015). Non-vireya taxon names were indexed following Chamberlain et al. (1996) (2003) are not included due to inaccessibility of data. Reported ploidy data are in Supporting Information- Table S1 Subgenus Section Subsection
Ploidy level is highly varied in Rhododendron taxa, including 2x (387 taxa), 3x (9), 4x (76), 5x (1), 6x (25), 8x (4) and 12x (1) [see Supporting Information- Table  S1]. Only diploids have been reported in four of the nine subgenera and several sections or subsections of the other five subgenera (Table 1; Fig. 2). Intraspecific ploidy variation has been observed in a total of 55 taxa in subgenera Hymenanthes (6), Pentanthera (7), Rhododendron (35) and Vireya (7). The highest number of intraspecific ploidy levels was revealed in the R. maddenii complex, with four levels in both R. maddenii ssp. maddenii and R. maddenii ssp. crassum. Over the genus, a discrepancy occurs that polyploidy was reported only from either chromosome counting or flow cytometry when both methods have been used to study the same taxa, which was observed in 22 taxa among the total 424 reported [see Supporting Information- Table  S1]. For example, polyploids were identified from only flow cytometry in subgenus Vireya and ss. Rhodorastra in subgenus Rhododendron. In such cases, these taxa were recorded as polyploids for our examination of frequency of polyploidy in Rhododendron groups.  (Table 2; [see Supporting Information- Table S2]). All taxa outside the R. maddenii complex were diploids, with 43 taxa reported using flow cytometry for the first time, among which 12 taxa had no previously reported ploidy data. Polyploids were only identified within the two subspecies of R. maddenii, and there  Table  1, number of 'diploid' taxa = number of ('tested taxa' -'polyploid taxa'), while number of 'untested' taxa = number of ('all taxa' -'tested taxa'). Ploidy data were compiled from literature. Details are shown in Supporting Information- Table S1. See online for colour figure. Table 2. Ploidy of subsection Maddenia in the present study in comparison to published ploidy levels compiled from the literature. Diploids are 2n = 2x = 26. 1 Identified taxa in alphabetical order. Taxon list follows Chamberlain et al. (1996). 2 'Yes' indicates the first report of flow cytometry ploidy estimation from the present study, with chromosome counts reported previously. '*' indicates the first ploidy report from the present study. '/' indicates the accession was not considered as a distinct species, as it is an affinity or synonym and shown in the parentheses. Ploidy data in Supporting Information- Table S2. 3 Identification of accessions undetermined for subspecific taxonomy in the complex, based on morphology, thus listed as the species. 4 Considered as a distinct species according to literature (Fang et al. 2005;Gibbs et al. 2011;Donald 2012;Cox 2013;Mao and Bhaumik 2015;Mao et al. 2017;Chang et al. 2021). # Reported as a polyploid from a single accession (Cubey 2003;Khan et al. 2021) Ss were seven diploids as well as 55 polyploid accessions. At least one wild accession was tested for each of 43 taxa (a total of 135 wild accessions examined). When 20 accessions that had been first evaluated from fresh leaves were replicated using dehydrated samples (silica gel-dried or herbarium specimen), ploidy of the seven diploid taxa was consistent, regardless of how the sample was dried. For silica gel-dried samples, FCM ploidy was generally the same as from fresh leaves (Table 3 and see Supporting  Information-Table S3). Among the 13 tested polyploids from the R. maddenii complex, nine samples showed reproducible ploidy levels, while ploidy for four accessions was not certainly determined. However, when herbarium specimens were used for the same accessions, the ploidy level was the same as using fresh leaves for only three of the polyploid accessions. Polyploidy was not determined for three accessions, or one level lower than ploidy estimated from fresh leaves for five accessions (Table 3).

Ploidy levels of taxa in subsection
As two species were used as the diploid standard, we tested replicate samples of the R. maddenii complex to validate FCM ploidy using both standards. Interestingly, all 10 polyploid accessions showed discrepant ploidy results between the two runs (Table 4 and see Supporting Information-Table  S3). Ploidy estimation using R. parraye was one or two levels lower than that using R. fortunei, and there was a higher incidence of odd-numbered results with the R. parryae standard. Further examination demonstrated that the genome size of R. parryae (2C = 1.70-1.75 pg) was larger than R. fortunei (2C = 1.52-1.57 pg) [see Supporting Information- Table  S4]. This explains the lower FCM ploidy level calculated for the same accession, when using the former as the diploid standard.

Discussion
High ploidy variation in genus Rhododendron, with the most frequent polyploidy in subgenera Pentanthera and Rhododendron Our summary dataset (Fig. 2) Table S1]. This difference may result from accessions that were rare aneuploids (Jones and Brighton 1972), or likely miscounts due to technological limitations at the time of early cytological studies of Rhododendron (Bowers 1930). Natural ploidy series, with both diploids and polyploids, have been reported in several taxa of Rhododendron. Ploidy variation has been reported among species in complexes such as R. maddenii, R. telmateium and R. yunnanense, as well as polyploid series within species R. flavidum and R. occidentale [see Supporting Information- Table S1]. Jones et al. (2007) proposed that the polyploid series within R. occidentale was from species diversification while that within R. flavidum may result from an unresolved species boundary with R. calendulaceaum. Ploidy variation could arise from intraspecific variability (i.e. cytotype diversity) (Husband et al. 2013;Farhat et al. 2019) and reflects the high diversification rate of flora in the Himalayan region (Schwery et al. 2015;Yan et al. 2015;Shrestha et al. 2018). However, it could also be related to the unresolved taxonomy. That is, in cases where multiple taxa have been 'lumped' into one, it is possible that one or some of the previously recognised taxa had a particular ploidy level, which presents as a series within a set of accessions that encompass the synonymous taxa. A third issue might be incorrect field labelling leading to incorrect grouping of samples, although this could be greatly improved with morphological identification by the Table 3. Flow cytometry estimates of ploidy level for samples from fresh and dehydrated leaves (silica gel-dried or air-dried herbarium sample). 1 All samples were analysed using R. fortunei as the diploid standard. FCM histograms of accessions with inconsistent ploidy in different runs are shown in Supporting Information- researcheR. This identification issue was also addressed for taxonomic correction in studies of other plant genera such as Deutzia Thunb. (Hydrangeaceae) (Hembree et al. 2020). Resolving the taxonomy of polyploid complexes will significantly inform decision making on conservation of biodiversity at several levels (e.g. phenotypic, phylogenetic or species diversity) (Ennos et al. 2005;Laport and Ng 2017).
Notably, a discrepancy among ploidy levels was found between chromosome counts and flow cytometry reported in previous studies of Rhododendron. Polyploids were identified only from either flow cytometry (e.g. subgenus Vireya, subsection Rhodorastra) or chromosome counting (e.g. R. baileyi) when both methods have been applied [see Supporting Information- Table S1]. This occasional discrepancy is observed more likely from the two largest ploidy studies of Rhododendron: Ammal et al. (1950) using chromosome counting and Khan et al. (2021) using flow cytometry. Limited coverage of samples might be the primary reason for this discrepancy, such that intraspecific ploidy variation was not captured. Another factor that may influence the ploidy discovered is more recent sampling from garden cultivation rather than from the wild, as in cultivation there may be more polyploids due to their favourable horticultural features (Jones et al. 2008;Rodionov et al. 2019). Given that flow cytometry is less time-consuming than chromosome counting, additional sampling for flow cytometry from wild populations should be considered to further understand the ploidy variation in Rhododendron taxa.
Subsection Maddenia consisting of diploids except for the R. maddenii complex Our ploidy estimations using flow cytometry present the most comprehensive ploidy analysis of ss. Maddenia reported to date. We have made the first ploidy reports for 12 taxa, increasing the known number of diploid taxa in this subsection to 47 of the total 51 taxa studied (Table 2 and [see  Supporting Information-Table S2]. In agreement with previous studies, polyploids are present in ss. Maddenia, but only in the two subspecies of R. maddenii. The only two polyploid exceptions reported outside the R. maddenii complex were one octoploid (2n = c.104) R. taronense (Cubey 2003) and Table 4. Flow cytometry ploidy estimates of samples repeated with two diploid standards. 1 All results based on fresh leaves. FCM histograms are shown in Supporting Information- Table S3 Ss. Maddenia taxon Acc. ID FCM ploidy 1 Diploid standard one tetraploid (4x) R. carneum (Khan et al. 2021), but both were from one accession only. Our FCM results from multiple accessions together with other previous chromosome counts identified only diploids in both species, supporting the conclusion that these two species are diploids (Table 2 and see  Supporting Information-Table S1). The R. maddenii complex reveals a ploidy series consisting of diploids and polyploids varying from pentaploid to octoploid (5-8x) ( Table 2). This is consistent with the previous reports from flow cytometry (Jones et al. 2007;De et al. 2010;Khan et al. 2021), where hexaploids (6x) and octoploids (8x) were reported. Tetraploids (2n = 52) were identified in previous chromosome counts but not in the present study (Ammal et al. 1950;Darlington et al. 1955). However, it is not known whether the polyploid complex within R. maddenii results from intraspecific variability or a combination of ploidy levels due to the 'lumped' taxa. Also, we lack data as to whether the R. maddenii polyploids are allopolyploids or autopolyploids. Cubey (2003) proposed an autopolyploid origin of the polyploids in R. maddenii, when there were no distinguishable morphological characters among plants of different ploidy levels, and diploids were rarely found in this species. Although R. maddenii was assessed as a species of Least Concern in the Red List (Gibbs et al. 2011), we advocate studies of the two putative subspecies to determine whether the synonymous taxa require species recognition, which might result in a revision of their conservation status.
In the whole genus, natural anisoploidy is not common, although triploids (3x or 2n = 39) have been reported in nine species and a pentaploid (5x) in one species [see Supporting Information- Table S1]. Our study of ss. Maddenia revealed some anisoploids (5x, 7x) in the R. maddenii complex (Table  2). More Rhododendron anisoploids, usually triploids, are bred for horticulture, but little is known about their reproductive biology (Li 1957;Jones and Ranney 2009). The pentaploid (5x) and heptaploid (7x) plants in our study do produce flowers, but their fertility and the mechanism as to how these anisoploids arose await further investigation.
In general, R. maddenii ssp. crassum tends to present a higher ploidy level with more octoploids (8x) than R. maddenii ssp. maddenii which includes more hexaploids (6x) ( Table 2 and  see Supporting Information-Table S1). This may be an infraspecific distinction and related to the geographic distribution of the two subspecies. Cullen (1980) compared the geographic distribution of the two subspecies of R. maddenii. While he suggested that the morphological variation in R. maddenii was unrelated to geography, our results show a possible trend in the ploidy levels between the two subspecies. However, the relationship between ploidy and geographic pattern in this species complex is still under investigation and requires further carefully structured field surveys.
The R. maddenii complex shows a similar pattern of ploidy variation and geographic distribution as the genus Buddleja L. (Scrophulariaceae) which is also from the Sino-Himalayan region (Cullen 1980;Chen et al. 2007). One hypothesis for the ploidy differences between the two subspecies of R. maddenii may be the ongoing in situ speciation in this area (Hughes 2017). As one of the world's youngest mountain ranges, with a high frequency of polyploidy in plants, the Sino-Himalayan region has been identified as a centre of species diversification that may be attributed to polyploidization (Irving and Hebda 1993;Schwery et al. 2015;Xing and Ree 2017;Shrestha et al. 2018;Rice et al. 2019;Xia et al. 2021). However, the Sino-Himalayan origin does not explain why polyploids should have continued to occur in R. maddenii, while other species in ss. Maddenia which are predominantly from the same region do not exhibit polyploids. Further research on a wider range of wild samples, from across the geographic range, and particularly from the R. maddenii complex, would inform this question.
Flow cytometry as a useful tool for estimating ploidy level of Rhododendron, even with dehydrated leaves Despite few reports in previous studies (De Schepper et al. 2001;Jones et al. 2007;Khan et al. 2021), the consistency of ploidy identified by flow cytometry and meiotic chromosome counts in our study supports the reliability of flow cytometry for ploidy estimation of Rhododendron (Fig. 3). Our trial with replicated samples verified higher reliability of silica gel-dried over herbarium leaves of Rhododendron for FCM ploidy assessment (Table 3). Although not preferred, dehydrated, particularly silica gel-dried leaf tissue, has been successful in other plant groups for ploidy estimation with the standard DAPI protocol using flow cytometry, thereby eliminating the inconvenience of collecting and preserving fresh samples (Šmarda 2006;Šmarda and Stančík 2006;Suda and Trávníček 2006;Farhat et al. 2019;Tomaszewska et al. 2021). The efficiency of flow cytometry with dehydrated leaves from herbarium specimens might be limited by several factors, including insufficient amount of tissue, sampling of mature leaves rather than newly expanding ones, incorrect drying, storage and preservation of samples and the limited efficacy of nuclei isolation due to degradation (Tomaszewska et al. 2021).
Although flow cytometry can be convenient for ploidy determination of a large number of samples, especially for identifying the frequency of polyploids (Jones et al. 2007;Kron et al. 2007;Hembree et al. 2020;Tomaszewska et al. 2021), this technique may not always yield definitive results . In our results for all samples that were interpreted as polyploids, histogram peaks of the higherploidy samples tended to be significantly lower than that of the diploid standard. Such small peaks may be missed, or the decision between diploid or polyploid may be difficult to interpret in data analysis [see Supporting Information- Table S3]. In addition, in some cases the ratio of two peaks on the histogram was near the midpoint between euploids, and the interpreted ploidy was therefore approximate (e.g., PK17 R. maddenii ssp. maddenii: 7-8x; Table 4 and [see Supporting Information- Table S3]). In such cases, we repeated samples without co-chopped standard tissue, to verify the results. In addition, our use of two diploid standards (R. fortunei and R. parryae) due to constraints on availability of fresh material in some gardens, brought further challenge to interpreting the ploidy of some accessions. Differing values of DNA contents from different standards slightly change the calculated ploidy ratio, particularly for polyploids. Our results for R. fortunei and R. parryae fall within the range of genome sizes of reported Rhododendron diploids (Bou Dagher-Kharrat et al. 2013;Khan et al. 2021;Choi et al. 2022). However, the larger genome size of R. parryae likely caused the interpreted ploidy for the same accessions being one level lower than using R. fortunei as the standard (Table 4 and [see Supporting Information- Table S4]). Consistent use of a single standard in a study is therefore recommended, and we anticipate further use of these two Rhododendron species as diploid standards in future research for the genus. Nevertheless, due to the degree of unavoidable uncertainty of ploidy determination by flow cytometry, we suggest that chromosome counting is still the 'gold standard' approach to determining ploidy level.
In previous studies, most Rhododendron chromosome counts were made from root tips grown from seed (Jones et al. 2007;De et al. 2010;Zaytseva et al. 2018). However, this may introduce errors in the ploidy determination, as seedlings from species in cultivation are likely to be from openpollinated seeds and hence may be hybrids. For this reason, we used developing stamens in flower buds as the source of tissue for chromosome counts, which requires careful observation of the timing of flower bud development. Rhododendron flower buds move into dormancy shortly after flower differentiation. The onset and duration of rest depend upon the presence of the flower bud scales (Schneider 1968). Mirgorodskaya et al. (2015) reported that the microspores of the evergreen species R. catawbiense in Russia underwent meiosis at the end of the summer (i.e. in August) and overwintered at the vacuolization stage. Mitosis with the formation of bicellular pollen grains occurred shortly before flowering at the beginning of summer in the following year (i.e. in June). In light of these observations, we suggest harvesting flower buds with dividing microspore mother cells after blooming and close to winter dormancy, at the late stage of flower differentiation. However, it may take the entire season to observe development and identify the correct stage for sampling. Meiotic chromosomes can only be observed once the buds are dissected under the microscope after fixation and staining, which increases the difficulty of harvesting inflorescence buds at the desired development stage and requires constant sampling in the field. Rhododendron chromosomes are small and difficult to view under the microscope (Jones et al. 2007), which makes it more difficult to assess polyploids with multiple sets of chromosomes (Windham et al. 2020). In some cases, even a physical chromosome count cannot confirm the number of chromosomes, due to overlapping chromosomes, or abnormal chromosome behaviours (e.g. lagging chromosomes) in polyploids (Li 1957;Contreras et al. 2007). Other possible approaches to ploidy estimation such as targeted capture sequencing may allow discovery of polyploid characteristics (Viruel et al. 2019;Tahir et al. 2020).

Conclusion
Phylogenetic analysis, commonly used to provide knowledge on species relationships and in turn for conservation planning, can be confounded by the presence of polyploids in a set of samples. In the 'big genus' (Frodin 2004) Rhododendron that presents both complex taxonomy and considerable conservation problems, prior studies reported polyploids in several species. We investigated the taxonomic distribution of polyploidy in the whole genus, and particularly ploidy levels of taxa in ss. Maddenia. Polyploidy occurs across the genus with 22 % polyploids among the reported 424 species, with the highest frequency in subgenera Pentanthera and Rhododendron. However, the genus remains largely underexamined for ploidy, with no report for 69 % of Rhododendron taxa.
Flow cytometry is a suitable tool for ploidy estimation in Rhododendron. When fresh tissue is unavailable, silica gel-dried leaves are more reliable than leaves from herbarium samples. We used flow cytometry to estimate ploidy for 47 taxa in ss. Maddenia, including 12 taxa that had never been investigated in previous studies. In this subsection, polyploids have been definitively identified in only the R. maddenii complex, where its two subspecies exhibit ploidy series consisting of diploids and various polyploidy levels. The ploidy variation in the R. maddenii complex may be a factor of the unresolved taxonomy or of the diversification of the species across a broad geographic range. Broader sampling from wild populations should be considered in future research to resolve the relationship between taxon geography and ploidy levels.
Current botanic garden accessions of ss. Maddenia, especially those from the wild, can be analysed with next-generation sequences mapped against published Rhododendron genomes (Zhang et al. 2017;Soza et al. 2019;Yang et al. 2020;Ma et al. 2021;Zhou et al. 2022) to understand character evolution, especially for those characters used to resolve species taxonomy. More immediately, the present ploidy estimations from our samples will prompt our phylogenetic study of ss. Maddenia. Eventually, knowledge of resolved taxonomic debates will underpin the priorities in ss. Maddenia for conservation actions.

Supporting Information
The following additional information is available in the online version of this article- Table S1. Reported ploidy of taxa in Rhododendron L. (Ericaceae). Table S2. Ploidy estimation of taxa in subsection Maddenia using flow cytometry in the present study. Table S3. Flow cytometry histograms of subsection Maddenia accessions with inconsistent ploidy in different runs. Table S4. Genome size measurements of R. fortunei Lindl. and R. parryae Hutch.

Conflicts of Interest Statement
None.

Declarations
Plant specimens from Royal Botanic Garden Edinburgh were collected under Convention on Biological Diversity (permit ID: 266). Plant material from Royal Botanic Garden Edinburgh and Rhododendron Species Botanical