1  
  

    1  
  

  
  
  

 
  
 

Influence of polyploidy on morphology, genetic 
differentiation and reproductive strategy amongst varieties 

of Rhodohypoxis baurii (Hypoxidaceae). 

 
by 

Bianca Tasha Ferreira 
 
 
 

Submitted in fulfilment of the requirements for the 
degree Master of Science (MSc) 

in 
Animal, Plant and Environmental Sciences 

in the Faculty of Science, University of the Witwatersrand, Johannesburg,  
South Africa 

 
 

Supervisor: Dr. Kelsey L. Glennon 
Co- supervisor: Prof. Glynis V. Goodman-Cron 

 
 

03 May 2023 
 
  



 
  

      
  

2 

  
  

Declaration  
 

I declare that this dissertation is my own, unaided work. It has been submitted in 
fulfilment of the requirements of a Master of Science at the University of the 
Witwatersrand. It has not been submitted before for any degree or examination to 
another university or similar institution.   

  

 
  

  

Bianca Tasha Ferreira   

  

3rd May 2023 

  
  
  
  
  
  
  
  
  
  
  
  
  
  

Supervisors:  

Dr. Kelsey Glennon  

Prof. Glynis Goodman-Cron    
  



 
  

      
  

3 

TABLE OF CONTENTS  

Declaration ............................................................................................................................. 2 

TABLE OF CONTENTS ..................................................................................................... 3 

LIST OF TABLES AND FIGURES .................................................................................... 4 

ABSTRACT ........................................................................................................................... 6 

ACKNOWLEDGMENTS .................................................................................................... 8 

CHAPTER 1: INTRODUCTION ........................................................................................ 9 

1.1 Rationale ................................................................................................................. 9 
1.2 Literature Review ........................................................................................................ 11 

Polyploidy overview .................................................................................................. 11 
Evolutionary importance of polyploidy .................................................................... 14 
Reproductive isolation and shifts in breeding systems ............................................. 16 
Species, subspecies and varieties .............................................................................. 20 
Rhodohypoxis baurii complex ................................................................................... 21 

1.3 Aim and objectives ...................................................................................................... 26 
CHAPTER 2: METHODS ................................................................................................. 27 

Morphometrics Analysis ........................................................................................... 27 
Cross-pollination viability ................................................................................................ 39 
Agamospermy/apomixis ................................................................................................... 40 
Pollinator identification .................................................................................................... 40 
Pollen viability .................................................................................................................. 41 

CHAPTER 3: RESULTS ................................................................................................... 42 

Morphometric analysis of R. baurii varieties and ploidy levels ............................... 42 
Microsatellite analyses - Genetic differentiation and gene flow .............................. 53 
Reproductive strategies ............................................................................................. 58 
Pollinator observations across R. baurii varieties ................................................... 65 

CHAPTER 4: DISCUSSION ............................................................................................. 71 

How does polyploidy influence morphology in R. baurii varieties? ......................... 71 
Reproductive strategies and escaping MCE in R. baurii .......................................... 74 
Genetic diversity and gene flow across R. baurii varieties ...................................... 77 
Latitudinal polyploid gradient .................................................................................. 80 
Taxonomic implications ............................................................................................ 82 

CHAPTER 5: CONCLUSION ........................................................................................... 87 

REFERENCE LIST ............................................................................................................ 88 

APPENDIX ........................................................................................................................ 103 

 
  
 
 
 



 
  

      
  

4 

LIST OF TABLES AND FIGURES  
 

Figure 1: Genetic differentiation, morphological change and changes in polyploidy, 
when exhibited together, could lead to polyploid speciation events. ………….……..11 
   
Figure 2: The four main regions of the Drakensberg (Northern Drakensberg,     
Central Drakensberg, Southern Drakensberg and Eastern Cape Drakensberg). ….…..22   
 
Figure 3: Images of the three Rhodohypoxis baurii varieties………………………....24 
  
Table 1: Character list for multivariate analysis of Rhodohypoxis baurii 
complex……………………………………………………………………..………....28  
  
Figure 4: Trichome branching patterns and sheaths across R. baurii varieties…….....29  
  
Figure 5: Reproductive morphology measurements in R. baurii varieties………...….30  
  
Figure 6: Geographic distribution of Rhodohypoxis baurii herbarium specimens  used 
for morphometric analyses. ……………………………………………………….…..32  
   
Table 2: Sampled localities and GPS co-ordinates of 12 populations of Rhodohypoxis 
baurii varieties used in molecular microsatellite analyses and seven populations  used 
in morphometric analyses. ………………………………………………………..….33 
  
Figure 7: Distribution map of Rhodohypoxis baurii varieties known range, and the  
11 sampled populations. ……………………………………………………………..34 
  
Figure 8: Inter-varietal and intra-varietal crosses conducted between the three 
Rhodohypoxis baurii varieties with all varieties acting as both pollen donors             
(paternal) and pollen receivers (maternal).  ……………………………………...…..40 
    
Figure 9: Rhodohypoxis baurii pollen grains stained with Alexander’s stain showing   
viable pollen (red) and inviable pollen (green/blue).  …………..………………..…..42 

Table 3: List of contributions (loadings) of each morphological character measured 

for the first four Principal Component Axes, based on the PCA analysis……..…..43 

Figure 10: Non-metric multidimensional scaling (NMDS) analysis of 116 herbarium 
specimens of R. baurii varieties. ……………………………………………………..44          
  
Figure 11: Comparison of vegetative, and floral morphological traits of 116 herbarium 
specimens represented by Rhodohypoxis baurii varieties. …………………………..46 
  
Figure 12: Comparison of floral morphological traits of 31 collected greenhouse specimens 
represented by Rhodohypoxis baurii varieties. ………………………………………48          
  
Figure 13: Comparison of vegetative morphological traits of 43 herbarium specimens 
represented by Rhodohypoxis baurii varieties. …………………………………….…49    
        



 
  

      
  

5 

  
Figure 14: Image showing the size difference of reproductive organs such as anther 
and ovules between Rhodohypoxis baurii varieties. ………………………………..49   
  
Figure 15: Variation in flower colour of R. baurii  individuals housed in the       
greenhouse. …………………………………………………………………….…...50   
  
Figure 16: Comparison of floral morphological traits across 31 collected greenhouse 
specimens of Rhodohypoxis baurii varieties grouped by ploidy level ……………..52           
  
Figure 17: Comparison of vegetative morphological traits of 31 collected greenhouse 
specimens of Rhodohypoxis baurii varieties grouped by ploidy level …………......53          
  
Table 4: Indices of genetic diversity based on 12 microsatellite markers across 11sampled 
population, the three Rhodohypoxis baurii varieties and three ploidy levels. ……...55          
  
Figure 18: Population genetic structure, allelic diversity and ploidy of the R. baurii        
populations based on auto-polyploidized data from 12 microsatellite markers.   ……….…57     
  
Figure 19: DAPC based on Bruvo genetic distance matrix grouped by A) taxonomic 
varieties, B) grouped by ploidy levels, C) grouped by populations. ……………….58          
  
Table 5: Differences in seed set between inter-varietal and intra-varietal crosses as well 
as between like ploidy and unlike ploidy crosses. …………………………………60       
  
Table 6: Differences in germination success between seeds from intra-varietal and inter-
varietal crosses	as well as between like ploidy and unlike ploidy crosses. ………..62    
  
Figure 20: Comparison of pollen viability between (A) Rhodohypoxis baurii varieties 
and (B) different ploidy levels. ………………………..…………………………...63         
  
Figure 21: Desktop Scanning Electron Microscope (SEM) images of R. baurii pollen 
grains. ………………………………………………………………………………64            
  
Table 7: Differences in clonal reproduction between Rhodohypoxis baurii varieties and 
ploidy levels. ………………………………………………………………………65       
  
Table 8: Summary of insect activity of species observed on inflorescences of 
Rhodohypoxis baurii at seven sites throughout the Drakensberg during        
the November 2021– January 2022 flowering season. ………………..……..67–68  
   
Figure 22: Insect visitors of Rhodohypoxis baurii flowers and evidence of herbivory 
on the tepals and anthers of Rhodohypoxis baurii varieties. ……………………....69  
        
Figure 23: Frequent insect visitors caught whilst visiting Rhodohypoxis baurii 
flowers at two Northern Drakensberg populations.  . ………………………….…..70  
  
 
 

 



 
  

      
  

6 

ABSTRACT   
 

Evolutionary mechanisms, such polyploidy (increase in chromosome sets), alters plant 

morphology, gene flow and reproductive strategies, which can facilitate the generation 

or loss of species. Rhodohypoxis L. (Hypoxidaceae) is a small near-endemic 

Drakensberg genus comprising six species, one of which is Rhodohypoxis baurii. 

Rhodohypoxis baurii contains three morphologically distinct varieties, with varying 

ploidy-levels: R. baurii var. baurii (2×, 4×), R. baurii var. platypetala (2× 3× , 4×), 

and R. baurii var. confecta (2×). Therefore, R. baurii is an ideal system to evaluate 

whether polyploidy leads to lineage divergence or homogenization and contributes to 

biodiversity in this lineage. The aim of the study was to assess the influence of ploidy 

on morphology, genetic differentiation and reproductive strategy among varieties of 

Rhodohypoxis baurii (Hypoxidaceae), as well as to better recognise the three varieties. 

Thirty vegetative, floral, and reproductive traits were measured across 124 herbarium 

specimens and 43 individuals housed in the greenhouse. A matrix containing 20 

quantitative and 12 qualitative characters was constructed and a Principal Coordinates 

Analysis (PCoA) and Non-Metric Multidimensional Scaling analysis (NMDS) were 

conducted. Important distinguishing morphological features that had a high 

eigenvalue (shown via a PCA) were selected for direct comparison using box and 

whisker box plots to compare means ± standard errors (SE).  Certain morphological 

traits such as anther length, peduncle length and tepal sizes differed significantly 

among the varieties and ploidy levels, with polyploid individuals exhibiting the gigas 

effect. This was especially evident in R. baurii var. platypetala, which contained 

many polyploid individuals and exhibited larger flowers (longer and wider tepals) and 

larger anthers compared to other varieties.   

 

Rhodohypoxis baurii var. confecta and R. baurii var. platypetala are genetically, 

geographically and morphologically similar, differing only in flower colour, flower 

size and peduncle length. However, most of these differences can be attributed to 

differences in ploidy and /or altitude with R. baurii var. confecta occurring at higher 

altitudes and R. baurii var. platypetala containing multiple ploidy-levels. It is therefore 

evident that ecological differences and polyploidy have shaped the morphological 

differences in these two taxa. In addition, R. baurii var. baurii populations in the 



 
  

      
  

7 

Eastern Cape Drakensberg were morphologically, geographically and genetically 

distinct from all other populations and varieties, and may be a new/ undescribed taxon; 

however, this warrants further investigation. Out of 231 experimental crosses, 113 

intra-varietal and inter-varietal crosses produced seeds. Rhodohypoxis baurii polyploid 

individuals show a shift away from sexual reproduction to asexual reproduction as 

they all showed higher rates of clonal reproduction than the diploid individuals. 

Moreover, crosses between polyploids yielded lower seed sets and lower germination 

rates than diploid-diploid crosses. Genetic differentiation and gene flow were 

quantified for 280 individuals among the varieties and ploidy-levels across 11 

populations using 12 microsatellite markers labelled with the FAM NED dyes. Leaf 

material was collected from 237 individuals of Rhodohypoxis baurii (R. baurii var. 

confecta n = 88, R. baurii var. baurii n = 87, R. baurii var. platypetala n = 62) and 

flow cytometry conducted to estimate ploidy. A latitudinal ploidy gradient was evident 

across sampled populations that corresponds with shifts in reproductive strategy, and 

changes in the extent of gene flow. Population genetic structure coincided primarily 

with geographic localities, with diploid Northern Drakensberg populations having 

similar allelic diversity to one another. The Central and Southern Drakensberg mixed 

ploidy populations also showed similar allelic diversity but differed from the tetraploid 

Eastern Cape Drakensberg populations. Furthermore, gene flow was higher between 

geographically close populations irrespective of ploidy-level, with geographically 

isolated regions (such as the Eastern Cape Drakensberg) and outlying populations (i.e. 

Karkloof) showing unique genotypes, indicating little gene flow and allele sharing. 

Consequently, shifts in reproductive strategy and geographic isolation are likely 

changing gene flow patterns among varieties and ploidy levels which appears to be 

facilitating both lineage diversification and homogenization in this species.  

  
  
  
     



 
  

      
  

8 

ACKNOWLEDGMENTS  
 

This research was funded by the National Research Foundation (Reference Number: 

118526) and the University of the Witwatersrand Friedel Sellschop Award to KLG. I 

thank the University of the Witwatersrand (Postgraduate Merit Award) and the South 

African Association of Botanists (SAAB) for funding and bursary support. Thank you to 

the C.E. Moss Herbarium (J), Pretoria National Herbarium (PRE) and the University of 

KwaZulu-Natal (NU) for the loan and use of specimens. Thank you to Dr Kenneth 

Oberlander and Sinethemba Ntshangase from the University of Pretoria for the use of, 

and training in, flow cytometry. Thank you to Shane McEvey from the Australian 

Museum Research Institute for assistance in fly identification.   

 

I would like to thank my supervisors, Dr. Kelsey Glennon and Prof. Glynis Goodman-

Cron, for all the guidance, assistance and encouragement throughout the years. You 

both inspire me every day and have taught me more than I could have asked for! A 

special thanks to Dr. Kelsey Glennon for always going above and beyond the role of a 

supervisor. Thank you for investing so much time and energy into my growth as a 

scientist and my future as a researcher. I cannot thank you enough for all you have 

taught me and done for me.   

Thank you to my wonderful lab mates, Jessica Minnaar, Verona Govender, Paulo 

Ribeiro, Saness Moodley and Timothy Hall. All of the laughs together made the long 

lab days so much fun and made these past two years fly by. Saness Moodley, thank 

you for your kindness and support, your cheery deposition made all the difficult days 

easier. Thank you for always being willing to lend a helping hand and answer my 

endless number of questions. Paulo Ribeiro, thank you for always welcoming me in 

your lab and space, and for being the best company during the long writing days. 

Timothy Hall, there are no words to explain how grateful I am for you. Thank you for 

encouraging me and for always being willing to help me in the lab, greenhouse and in 

the field. You never hesitated to aid me or support me throughout the past two years.   

    
 

 



 
  

      
  

9 

CHAPTER 1: INTRODUCTION  
 

1.1 Rationale  

Understanding the evolutionary path of mountain endemic plant species helps to 

quantify, understand the history of, and predict the future of the biodiversity in 

hotspots such as the Drakensberg Mountain Centre, under the threat of global change. 

Evolutionary mechanisms, such as whole genome duplication (polyploidy), have been 

noted to provide advantages in extreme environmental conditions and are often cited 

for their potential role in ameliorating the harm of global change to biodiversity and 

ecosystem functionality (Ahrens et al., 2020; Innes et al., 2021; Van de Peer et al., 

2017). Polyploidy often results in morphological and reproductive changes which 

may, subsequently, lead to diversification of plant lineages (Soltis et al., 2014), 

resulting in an increase in biodiversity. Two opposing views exist on the effects of 

polyploidization and genome changes within a long-term evolutionary setting. One 

view is that polyploidization can lead to long term plant diversification and, therefore, 

is a driver of speciation and biodiversity (De Storme & Mason, 2014; Levin, 1983; 

Wood et al., 2009). On the other hand, others hypothesize that polyploidization is 

‘evolutionary noise’ or an ‘evolutionary dead-end’ due to lower diversification rates of 

polyploids and lower longterm survival, and therefore insignificant to the main process 

of evolution (Mayrose et al., 2011; Stebbins, 1950; Wagner, 1970) .   

Rhodohypoxis baurii (Baker) Nel is an ideal system to evaluate whether polyploidy 

can lead to speciation and assist in maintaining and creating biodiversity in this group. 

Rhodohypoxis baurii contains three varieties, comprising a range of different ploidy 

levels: R. baurii var. baurii (Baker) Nel  (2×, 4×), R. baurii var. platypetala (Baker) 

Nel  (2×, 3×, 4×, and R. baurii var. confecta Hilliard & B.L. Burtt (2×) (Saito 1975, 

Glennon unpublished data). Gene flow between varieties and closely related species 

may be facilitated by polyploidy and reflected in morphology, suggesting that few 

reproductive barriers exist between Rhodohypoxis baurii varieties. Consequently, 

many shared morphological traits exist between the R. baurii varieties, and 

distinguishing features need to be identified to enable better recognition of these three 

varieties.   

  



 
  

      
  

10 

Chromosome duplication has been noted to alter plant morphological traits, for 

example, resulting in features such as increased flower number per inflorescence and 

larger reproductive organs (Balao et al., 2011; Knight & Beaulieu, 2008; Robertson et 

al., 2011). Observed increases in reproductive organ size are indicative of reduced 

self-incompatibility (Robertson et al., 2011), often resulting in increased occurrences 

of selfing in polyploid individuals (Barringer, 2008; Robertson et al., 2011). However, 

the reproductive consequences of polyploidy are not fully understood (Vamosi et al., 

2007). Therefore, corroborating ploidy-levels with morphological trends in R. baurii 

will further our understanding of the effects that polyploidy may have on plant 

morphology, reproduction, and its role in lineage diversification. Recent greenhouse 

observations suggest that R. baurii may reproduce clonally or be self-compatible (B. 

Ferreira, pers. obs.). In this study, individuals of different ploidy and different taxa will 

be assessed to describe their           reproductive biology. Quantifying gene flow and 

reproductive strategies, together with examining ploidy-level variation in R. baurii, 

will facilitate our understanding of the role that polyploidy may have played in this 

group’s evolutionary past and potentially have in its future trajectory.  

  

The overall aim of the study is to assess the influence of ploidy on differences in 

morphology and reproductive strategy among varieties of Rhodohypoxis baurii 

(Hypoxidaceae). Using microsatellites, genetic differentiation and gene flow were 

quantified among the varieties and individuals with different ploidies. Reproductive 

strategies of each variety were examined in the greenhouse and pollen and seed viability 

of intra-variety and inter-variety crosses were examined as well as seed viability of 

selfed individuals. Field-based pollinator observations were conducted to understand 

what, if anything, is pollinating R. baurii. Quantifying reproductive strategies, together 

with examining ploidy-level and morphological variation, of R. baurii varieties, 

advance our understanding of the species evolutionary past and assist in predicting its 

trajectory. This information will contribute to our understanding of the role that 

polyploidy may play in the generation and perpetuation of species (Figure 1).  

 

 

 

 

  



 
  

      
  

11 

 
  

Figure 1. Genetic differentiation, morphological change and changes in polyploidy, when 
exhibited together, could lead to polyploid speciation events through an increase in 
variability (both genetic and morphological), phenotypic changes (such as the gigas effect) 
and lineage divergence.  

  

1.2 Literature Review  

Polyploidy overview  

Understanding the evolutionary path of polyploid species and the processes involved 

in speciation events enables one to understand the history and more accurately predict 

the future of biodiversity. Most speciation theories are based around natural selection, 

with emphasis on driving selective pressures that cause species divergence and 

speciation (Mayr, 1963; Schluter & Nagel, 1995; Lynch & Force, 2000; Werth & 

Windham, 1991). On the other hand, others view selection as any cause of genetic 

divergence between allopatric populations, resulting in reproductive isolation and 

subsequent divergence (Sobel et al., 2010). How polyploidy fits into speciation is still 

an open question (Schluter & Nagel, 1995).  

  

Genetic  
differentiation   

Polyploidy   

phenotypic  
changes   Variability   

Morphology   

Polyploid   
speciation   

Lineage  
divergence   



 
  

      
  

12 

Consequently polyploid events are a likely mechanism of plant diversification and speciation 

(Sobel et al., 2010; Soltis & Soltis, 2009; Levin & Soltis, 2018; Mayrose et al., 2014); these 

events result in genetic and phenotypic variation on which natural selection may act.   

  

Polyploid events result in organisms that can be classified into one of two main groups: 

autopolyploids or allopolyploids. Autopolyploidy is the formation of polyploids within 

a single species (Clausen et al., 1945; Stebbins, 1947) where multiple sets of 

chromosomes come from the same taxon (intra-specific hybridisation), whereas 

allopolyploids result from interspecific hybridisation (Lewis, 1980) and chromosome 

sets are combined from two different taxa. However, additional types of polyploids 

were later identified, such as aneuploidy, where chromosome numbers differ from the 

exact multiple of the base chromosome number (Ramsey & Schemske, 2002) and 

neopolyploids which are newly formed auto- or allopolyploids (Ramsey & Schemske, 

2002). Within each of these broader classifications, polyploid individuals are further 

identified based on the number of chromosome sets they possess. For example, 

triploids (3×) contain three full sets of chromosomes and tetraploids (4×) contain four 

chromosome sets (Ramsey & Schemske, 1998). These polyploids can arise from 

multiple different, recurrent and independent origins (Soltis et al., 1989; Werth et al., 

1985; Wyatt et al., 1988).  

Types of polyploid formation  

Polyploids can have multiple origins with different mechanisms involved in their 

formation. Recurrent formation of polyploids from genetically different parents is 

common and relatively frequent for both allopolyploids (Bayer & Crawford, 1986; 

Ogihara & Tsunewaki, 1988; Ranker et al., 1989; Werth et al., 1985; Wyatt et al., 

1988) and autopolyploids (Shore, 1991; Soltis & Soltis, 1993; Soltis et al., 1989; Wolf 

et al., 1990). Numerous ancient polyploidization events have occurred across 

angiosperms (Barker et al., 2009; Blanc & Wolfe, 2004; Bowers et al., 2003; Cui et 

al., 2006; Jaillon et al., 2007; Paterson et al., 2004; Shi et al., 2010). For example, 

Soltis, Soltis and Ness (1989) suggested that alumroot, Heuchera micrantha Douglas 

ex Lindl. (Saxifragaceae), had at least three separate origins of autopolyploid 

formation in the species’ evolutionary history based on ancestral chloroplast genomes 

shared by diploid and tetraploid populations. Due to the multiple origins of 



 
  

      
  

13 

polyploids, polyploid species can assimilate genetic diversity from many sources 

/populations of diploids (Symonds et al., 2010). This can maintain high levels of 

‘segregating genetic variation’ (Soltis et al., 2014; Soltis & Soltis, 1999; Soltis & 

Soltis, 2000). From somatic doubling to the generation and importance of unreduced 

gametes, triploid and tetraploid individuals can be formed via multiple routes 

(D’Amato, 1952; Khoshoo, 1959; Navashin, 1925;  Thompson & Lumaret, 1992).  

The nonreduction of gametes is one of the main processes responsible for polyploid 

formation (Brownfield & Köhler, 2011; Ramsey & Schemske, 1998). Unreduced 

gametes (diploid or 2n gametes) occur via meiotic defects (e.g., nondisjunction) 

where ploidy-levels are not reduced during meiosis (Bretagnolle & Thompson, 1995; 

Harlan & deWet, 1975). Similarly, somatic doubling occurs during an error in mitosis 

where cells are produced with twice the number of normal chromosomes (Bretagnolle 

& Thompson, 1995; Brownfield & Köhler, 2011). Both the nonreduction of gametes 

(also known as “meiotic nuclear restitution”) and somatic doubling results in 2n 

gametes which contain full somatic chromosome numbers (Ramsey & Schemske, 

1998). The combination of a reduced and unreduced gamete leads to triploid 

formation and is more common in diploid populations than polyploid populations 

(Bretagnolle & Lumaret, 1995; Khoshoo, 1959). Likewise, the union of two 

unreduced gametes is likely to result in tetraploid formation (Ramsey & Schemske, 

1998). Selfing via triploids and tetraploids will commonly result in triploid and 

tetraploid offspring, respectively (Ramsey & Schemske, 1998). Autotriploid gametes 

are often not functional due to aneuploidy but can mediate and lead to tetraploid 

production via self-fertilisation or backcrossing with diploids, where triploids act as 

triploid bridges (Dermen, 1931; Husband, 2000; Johnsson, 1940; Levin, 1975; 

Ramsey & Schemske, 1998; Yamauchi et al., 2004). These resulting tetraploids 

generally have functional gametes and can reproduce further (Yamauchi et al., 2004). 

Self-fertilisation and backcrossing of hybrid triploids with diploids often leads to 

allotetraploid formation (Ramsey & Schemske, 1998). Polyploid formation can be 

either maladaptive or advantageous to individual plants and populations. 

Chromosome doubling may provide a basis for evolution within a population level 

framework and can either result in individuals that may be adapted to changing 

environments or may be deleterious to individuals in environments where diploids are 

well adapted (Levin, 1983).   

 



 
  

      
  

14 

Evolutionary importance of polyploidy  

Due to the dynamic nature of polyploid genomes, there is extensive potential to 

generate novel genetic variation (Doyle et al., 2008; Flagel & Wendel, 2009; Gaeta et 

al., 2007; Leitch & Leitch, 2008; Otto & Whitton, 2000), which in turn, may lead to 

plant speciation or diversification (De Storme & Mason, 2014; Levin, 1983). 

However, polyploidy is not always viewed as a contributing mechanism of 

diversification. For instance, Mayrose et al., (2011) described polyploids as 

“evolutionary dead-ends,” arguing that polyploidy does not contribute significantly to 

evolution. Initially, Stebbins, (1950) and Wagner, (1970) argued that polyploids are 

more likely to go extinct than diploids and, therefore, are of less evolutionary 

importance. Even though it is widely understood that polyploids are likely to go 

extinct expeditiously at the population level (Levin, 1975; Ramsey & Schemske, 1998; 

Soltis et al., 2010), they are still of evolutionary importance. Polyploids have the 

potential for rapid adaptation due to their unique genomic background (Otto & 

Whitton, 2000; Schoenfelder & Fox, 2015; te Beest et al., 2012). Polyploidy has short-

term adaptive abilities and the unique characteristic of duplicated genes, and the 

retention of multiple gene sets, may explain long- term evolutionary transformation 

(Otto & Whitton, 2000; Van de Peer et al., 2017). Short- term success of polyploids 

may be due to genetic changes and subsequent increases in genetic variation (Doyle et 

al., 2008; Parisod et al., 2010; Schoenfelder & Fox, 2015;  Soltis et al., 2014 ;Otto & 

Whitton, 2000; Van de Peer et al., 2017).   

 

The increase in genetic variation due to polyploidy per se affects the morphology and 

ecology of neopolyploids and may alter interspecific interactions and gene flow 

(McCarthy et al., 2016; Soltis et al., 2014; te Beest et al., 2012). Furthermore, 

duplicated genes can have new or different and diverse functions leading to adaptive 

advantages in polyploids (Adams & Wendel, 2005; Lynch & Force, 2000; Madlung, 

2013; Soltis & Soltis, 1993). Polyploids have also played a historically advantageous 

role in angiosperm diversification (WeissSchneeweiss et al., 2013). Because 

polyploids often arise from multiple origins, there can be an increase in genomic 

diversity within polyploid individuals and populations (Soltis & Soltis, 1999; Weiss‐

Schneeweiss et al., 2012). It has also been documented that there is a positive 

correlation between species richness and frequency of polyploidy as genera with 50–



 
  

      
  

15 

75% polyploid species also show greater species richness (Otto & Whitton, 2000; 

Petit & Thompson, 1999; Vamosi & Dickinson, 2006). Likewise, it has been noted 

that species richness increases drastically in many angiosperm clades after ancient 

whole genome duplication events, which supports the suggestion of higher 

diversification rates in polyploids (Soltis et al., 2009). Furthermore, Crow & Wagner, 

(2006) found reduced risk of extinction and increased rates of evolution in polyploids. 

Therefore, despite arguments to the contrary there is compelling evidence that 

polyploidy may lead to plant speciation and diversification. However, for polyploidy 

to contribute to evolutionary processes, polyploid individuals first need to form and, 

more importantly, establish for the effects of polyploidy (be it advantageous or 

detrimental) to have a long-term impact.   

  

Two stages are crucial to polyploid evolution and success: formation and 

establishment. These two early stages of polyploid evolution entail that the new 

cytotype must be viable and fertile in order for the organism and lineage to have a 

higher likelihood of persisting (Ramsey & Schemske, 2002). The likelihood of new 

polyploid establishment and survival depends on the extent of reproductive isolation 

(RI) within and between ploidy levels (Ramsey & Schemske, 1998). Genetic 

incompatibility and reproductive isolation (RI) are often a consequence of 

polyploidization. Polyploidization can have detrimental effects on plant fertility due to 

the unstable nature of polyploid genomes and the meiotic aberrations that exist (Van 

de Peer et al., 2017). Yet, the main hindrance to the success of newly formed 

polyploid individuals is minority cytotype exclusion (MCE) (Levin, 1975; Van de 

Peer et al., 2017). MCE is where one cytotype is dominant (often diploids) and more 

common than other or new cytotypes (such as triploids or tetraploids). The less 

common cytotypes are less likely to mate and reproduce due to a lack of like ploidy 

partners (Husband, 2000). The reproductive success of a new cytotype is therefore 

frequency dependent as the minority cytotypes are at a mating disadvantage (Levin, 

1975). If minority cytotypes do not overcome MCE, they will be rapidly excluded and 

eventually go extinct in the population (Levin, 1975). In order to overcome MCE, 

polyploid individuals need to have like-polyploid mates in a diploid dominated 

environment in order to reproduce (Van de Peer et al., 2017) or must be able to self-

pollinate. Polyploidization often occurs with a shift in breeding systems from cross-



 
  

      
  

16 

pollination or sexual reproduction to self-pollination or asexual reproduction 

respectively (Comai, 2005).  

Reproductive isolation and shifts in breeding systems  

Selfing in polyploid individuals is aided by a breakdown in self-incompatibility (Levin, 

1983). A decrease in self-incompatibility allows self-pollination to occur, leading to further 

polyploid formation (Ramsey & Schemske, 1998) and an escape from MCE. This change in 

reproductive systems due to chromosome doubling is relatively common in many taxa, such 

as Solanum L., Trifolium repens L., T. hybridium L. and Lycopersicon peruvianum L. 

(Brewbaker, 1954, 1958; de Nettancourt et al., 1974; Lewis, 1943; Magoon et al., 1958). 

Assortative mating by cytotypes may aid in overcoming MCE (Mable et al., 2011), but 

selfpollination and asexual reproduction tend to be more common than sexual reproduction, 

particularly in neopolyploids (Alix et al., 2017; Ramsey & Schemske, 2002). New 

allopolyploids formed via hybridization often show agamospermic/apomictic reproductive 

strategies (Alix et al., 2017). For example, Sorbus L. species are known for producing 

microspecies via hybridisation that exhibit apomictic reproduction (Robertson et al., 2011). 

There are clear similarities and shared characteristics between polyploid formation and 

gametophytic apomixis: both processes rely on the formation of unreduced gametes (Majeský 

et al., 2012; Ramsey & Schemske, 1998; Whitton et al., 2008) and agamospermy is almost 

exclusively restricted to polyploids in Erigeron annuus (L.) (Noyes & Rieseberg, 2000).   

Mating systems can influence and change the long-term evolutionary trajectory and success of 

polyploids ( Barrett et al., 2003; Barrett et al., 1996; Barringer, 2007; Lande & Schemske, 

1985; Stebbins, 1950). Many questions exist about the evolution of self-fertilization or 

asexual reproduction and the trade-offs associated with this shift in breeding system 

(Barringer, 2007). Self-fertilization can be either advantageous as a “reproductive assurance 

mechanism” (Morgan & Wilson, 2005; Pannell & Barrett, 1998; Stebbins, 1950) or 

detrimental as it can result in inbreeding depression (Charlesworth & Charlesworth, 1987; 

Lande & Schemske, 1985). Even though the pattern of increased self-pollination with changes 

in ploidy has been noted and accepted widely, the relationship and mechanism between self-

fertilization and polyploidy is poorly understood (Barringer, 2007; Mable, 2004; Stebbins, 

1950). However, several hypotheses exist around why polyploidy may be associated with 

increased levels of selfing (Barringer, 2007).  

  



 
  

      
  

17 

Polyploidy usually results in a breakdown of self-incompatibility, resulting in 

increased frequency of selfing events (Barringer, 2007; Mable, 2004). On the other 

hand, the ability to self-fertilize may facilitate polyploid evolution (Barringer, 2007). 

Neopolyploids are likely to be in sympatry with their diploid progenitors when they 

first form, and because crosses between different cytotypes often lead to reduced 

fitness (Jackson, 1976; Levin, 1975; Ramsey & Schemske, 1998), minority cytotypes 

experience MCE and minority cytotype disadvantages (Levin, 1975). Therefore, the 

ability to self-pollinate allows polyploid individuals to escape MCE, survive, and 

reproduce (Levin, 1975; Ramsey & Schemske, 1998; Stebbins, 1971). Lastly, 

polyploids may exhibit higher levels of self-fertilization as they experience less 

inbreeding depression than diploids due to the multiple copies of genes they possess 

(Lande & Schemske, 1985). Increased selfing without the consequences of inbreeding 

depression would result in the selection of increased selfing instances in polyploid 

individuals and populations (Barringer, 2007).   

  

Asexual reproduction is another common breeding system shift in many polyploid 

individuals (Levin, 1983). A shift to asexual reproduction with polyploidization is 

clearly seen in Themeda australis Stapf. as diploids reproduce sexually, yet 

tetraploids reproduce asexually (Woodland, 1964). In addition to selfing, reproductive 

isolation (RI) of polyploids and MCE can be facilitated by changes in pollinator 

interactions between diploid and polyploid individuals due to changes in polyploid 

morphology (McCarthy et al., 2016; Van de Peer et al., 2017). Increased genetic 

variation among polyploids may lead to differences and divergence in morphological 

traits of polyploid individuals (McCarthy et al., 2016; Soltis et al., 2014). Changes in 

traits that attract pollinators (such as inflorescence colour, size and number of 

inflorescences per individual) will lead to changes in pollinator interactions (Gross & 

Schiestl, 2015; McCarthy et al., 2015; Segraves & Thompson, 1999). Consequently, 

pollinators may avoid or favour polyploids due to their different flower morphology 

and/or chemical scent/compounds, leading to assortative mating and resulting in RI of 

polyploids (Porturas et al., 2019; Segraves & Anneberg, 2016).   

  

Changes in floral morphology (and the subsequent change in pollinator relationships) 

are relatively common in polyploid taxa (Taylor & Smith, 1980). Higher ploidy-levels 

often exhibit increases or decreases in the number of inflorescences/flowers per 



 
  

      
  

18 

individual plant compared to their diploid progenitors (Levin, 1983). For example, in 

White clover, Trifolium repens L. (Fabaceae) tetraploid individuals showed a 20% 

decrease in number of flowers per head, whereas in Strawberry clover, T. fragiferium 

L., and Egyptian clover, T. alexandrinum L., tetraploids showed a 20% increase in the 

number of flowers per head (Mackiewicz, 1965). Similarly, Honey clover, Melilotus 

albus Medik. (Fabaceae) tetraploids also showed an increase in flowers per head by 

30% (Jaranowski & Kalasa, 1971). Interestingly, RI due to differentiation in pollinator 

communities (as a result of morphological changes) is not necessarily negative but may 

promote speciation events (Van de Peer et al., 2017). Differentiation in pollinator 

communities in a mixed population of diploid and polyploids will often cause RI, which 

may facilitate polyploid establishment (Segraves & Anneberg, 2016). If the “new” 

polyploid traits are selected for by pollinators, MCE will be overcome, and 

establishment success will increase, subsequently, a speciation event is more likely to 

occur (Van de Peer et al., 2017).   

  

Polyploidization not only leads to changes in floral morphology, but can cause a 

multitude of morphological changes induced by chromosome duplication (Balao et al., 

2011; Stebbins, 1947). After whole genome duplication, polyploids experience 

processes that can shrink or enlarge their genomes (Leitch & Leitch, 2008). This 

genome modification can induce phenotypic variation (Bennet & Leitch, 2005; Chen, 

2007). Cell enlargement and an increase in trait sizes is a well understood and an 

accepted effect of whole genome duplication (Knight & Beaulieu, 2008; Levin, 2002; 

Stebbins, 1971). Known as the ‘gigas effect’, the size of organs generally increases 

with chromosome number (Müntzing, 1936; Otto & Whitton, 2000; Stebbins, 1971). 

The ‘gigas effect’ is more common in new polyploids (Ramsey & Schemske, 2002). 

The underlying process behind this relationship is known as the “nucleotype effect” 

where organs are larger merely because the cells comprising them are larger (Bennett, 

1971; Doyle & Coate, 2019).  However, the ‘gigas effect’ is not consistent for all 

species (Otto & Whitton, 2000; Vamosi et al., 2007) as many polyploid taxa exhibit 

smaller or equal sized phenotypic traits compared to their diploid progenitors (Ning et 

al., 2009; Porturas et al., 2019; Segraves & Thompson, 1999; Trojak-Goluch & 

Skomra, 2013). For example, in Common hop, Humulus lupulus L. (Cannabaceae), 

tetraploid individuals have larger cones (infructescenses) compared to diploid 

individuals but similar sized flowers and significantly smaller leaves than diploids 



 
  

      
  

19 

(Trojak-Goluch & Skomra, 2013). Phenotypic changes in reproductive traits are 

commonly seen in polyploids, such as larger pollen grains (Porturas et al., 2019). In 

some Solanaceae species, for example, polyploids had larger reproductive organs than 

their diploid progenitors (Robertson et al., 2011). Other common morphological 

changes in polyploid individuals include increases in stomatal diameters (Mtileni et al., 

2021), changes in petal length (Bennett, 1971), changes in plant stature (Balao et al., 

2011), thicker foliage and thicker stems (Ramsey & Schemske, 2002). Changes in 

reproductive morphology due to polyploidization not only lead to changes in 

reproductive strategies but can also result in changes in phenology. These shifts often 

affect pollinator interactions and may affect gene flow and genetic diversity within and 

between populations and species (Ramsey & Schemske, 2002). Due to the larger 

morphological structures found in polyploids, and the larger cells, developmental 

processes tend to take longer (Cavalier-Smith, 1978; Ramsey & Ramsey, 2014) as more 

time is needed for cells to divide (Bennett, 1987; Francis et al., 2008). This can 

potentially result in later flowering times and a shift in flowering phenology in 

polyploid individuals (Jersáková et al., 2010; Oswald & Nuismer, 2011; Roccaforte et 

al., 2015; Segraves & Anneberg, 2016). Delays in, and prolonged flowering times are 

common in neopolyploids where growth and development are slowed (Ramsey & 

Schemske, 2002). For example, induced autotetraploids of Medicago L. and Melilotus 

Mill.. (Fabaceae) species exhibited flowering times that were several weeks longer than 

their diploid counter parts (Jaranowski & Kalasa, 1971).  

  

Changes in morphology, reproductive strategies, flowering phenology and the 

subsequent changes in pollinator interaction, together, may alter gene flow and allele 

sharing patterns. The breakdown in self incompatibility combined with changes in 

pollinator interactions leading to changes in gene flow can facilitate hybridisation. 

Lower reproductive incompatibility and fewer interspecific reproductive barriers often 

lead to higher levels of gene flow and allele sharing among distinct taxa (Alix et al., 

2017; Schluter & Nagel, 1995), which subsequently can result in a higher numbers of 

hybrid individuals or can facilitate hybrid swarms (Alix et al., 2017; Wood et al., 

2009). Hybridization often results in introgression, where recurrent backcrossing 

creates new allele combinations, generating genetic diversity and possibly driving 

diversification and speciation (Rieseberg, 1995). Therefore, hybridization may cause 

high genetic variability by producing new or novel genotypes, potentially leading to 



 
  

      
  

20 

speciation events (Alix et al., 2017). Hybridization together with polyploidization may 

act as drivers of evolutionary diversification by increasing gene flow, resulting in 

speciation (Alix et al., 2017; Levin, 2002; Ramsey & Schemske, 1998; Rieseberg, 

1995). High levels of gene flow can lead to the homogenization of distinct taxa or can 

produce new and novel lineages (Ramsey & Schemske, 1998). The resulting 

homogenization or diversification often blurs species boundaries, resulting in the 

division or combination of species into subspecies or varieties, depending on the 

degree of reproductive and geographic isolation.   

Species, subspecies and varieties   

Historically, species were recognised according to morphological similarities (Mayr, 

1969), but are now defined in a multitude of ways based on various species concepts. 

These species concepts are based on phenotypic distinguishability, reproductive 

isolation, ecological niche occupation, and lineage monophyly through time 

(phylogenies) (Coyne & Orr, 2004; de Queiroz & Donoghue, 1988). A species is the 

primary taxonomic unit and the basic evolutionary unit (Radford, 1986). There are 

many different and varied definitions of a species with several different concepts and 

criteria for recognizing species. The unified species concept recognises species as 

separately evolving metapopulation lineages as the only criterion necessary for 

describing distinct species. (De Queiroz, 2007). However, this concept does not assist 

in identifying differences between isolated populations or varieties that may or may 

not be evolving separately. Therefore, for the purposes of this study, a species is 

defined as the smallest group of populations which is permanently distinct and 

distinguishable from another with a reasonable degree of reproductive isolation from 

other species under natural conditions (following Cronquist, 1968). However, 

infraspecific classifications (e.g., subspecies, variety) are required when species lines 

are blurred and there is recognisable variation within a taxon that has a geographic or 

ecological link (Du Rietz, 1930; Stace, 1989).   

Traditionally, subspecies are a taxonomic rank lower than species, and many authors 

use subspecies and varieties interchangeably as synonyms (Hamilton & Reichard, 

1992). Nonetheless, Du Rietz (1930) defined both subspecies and varieties and these 

definitions are widely accepted and are currently used in botanical classification. 

Subspecies are defined as a population of several biotypes forming a relatively distinct 



 
  

      
  

21 

regional group (Du Rietz, 1930) and, therefore, are geographically isolated ecotypes or 

“topodemes” (Stace, 1989). Varieties are defined as one or several  populations of one 

or several biotypes forming a somewhat distinct local group (Du Rietz, 1930) and, 

therefore, are local or ecological ecotypes or “topodemes” (Stace, 1989).  

 

Species may also be defined by the “taxonomic circle” (Damm et al., 2010). In this 

framework, organisms or populations can only be defined as a species when at least 

three types of data, with DNA data as the foundation, support the species’ recognition. 

These additional data sources include reproductive isolation, ecological data such as 

habitat information, geographical data such as the species distribution, and 

morphological data (Damm et al., 2010; DeSalle et al., 2005; Fitze et al., 2011). 

However, this approach cannot be applied at subspecies or varietal levels due to the 

lack of reproductive isolation and the morphological and ecological differentiation 

often seen between infraspecific taxa, especially varieties (Clausen, 1941). Populations 

of subspecies are usually geographically isolated from one another and, therefore 

cannot interbreed in situ, however, they may not be reproductively isolated and when 

in sympatry such subspecies could potentially share alleles. Similarly, populations of 

varieties that are close enough to each other (not in sympatry, but not completely 

isolated) will likely interbreed.  

  
Rhodohypoxis baurii complex  

Rhodohypoxis is a small genus within the Hypoxidaceae that comprises six species 

(Hilliard & Burtt, 1978). Rhodohypoxis is a near endemic montane to alpine genus 

distributed in eastern South Africa, centred in the Drakensberg Mountain Centre and 

Lesotho plateau with  additional populations found in the KwaZulu-Natal Midlands, 

Free State and Mpumalanga (Hilliard & Burtt, 1978) (Figure 2). Rhodohypoxis is 

distinguished from Hypoxis (its sister taxon) by its white and pink flowers (versus 

yellow), the presence of a short perigone tube and subsessile anthers arising from the 

perigone at two levels (Hilliard & Burtt, 1978). Other distinguishing features of the 

genus are trichomes (hirsute/pilose) on the leaves, pedicels, and ovary, and a tufted 

indumentum on the back of the outer tepals. The leaves are basal and either spreading 

on the soil surface or erect, elliptic to filiform. The anther thecae have acute, usually 



 
  

      
  

22 

free tips. Stigma lobes consist of three fleshy papillose flanges attached to a very short 

style. Seeds are either ellipsoid or spherical (Hilliard & Burtt, 1978).   

 

 

 
Figure 2. The four main regions of the Drakensberg (Northern Drakensberg, Central 
Drakensberg, Southern Drakensberg and Eastern Cape Drakensberg) with black points 
showing the 11 sampled populations, including two populations with localities 
described as “away from the main Drakensberg” by Hilliard & Burtt (1978).  

  

Rhodohypoxis baurii is the type for the genus and is characterised by Hilliard and Burtt 

(1978) as having 4–10 leaves that are linear-lanceolate to lanceolate and hairy on the upper 

surface and margins, anthers ranging from 1.5–2 mm, ovaries 2–4 mm in diameter, and seeds 

1.5 mm in diameter on average. Rhodohypoxis baurii, commonly known as the ‘red star,’ is a 

morphologically variable species containing three varieties: Rhodohypoxis baurii var. baurii , 

R. baurii var. platypetala, and R. baurii var. confecta (Hilliard & Burtt, 1978) (Fig. 3). 

Rhodohypoxis baurii var. baurii has been noted to have two distinct morphological forms 

associated with geographic distribution (Hilliard & Burtt, 1978). The first form occurs “away 

from the main Drakensberg”, is found at lower altitude (1140 – 1740 m a.s.l.) and has larger 

  

Northern Drakensberg   
Central Drakensberg   
Southern Drakensberg   
Eastern Cape Drakensberg   
Away from the main Drakensberg   



 
  

      
  

23 

leaves, whilst the other form occurs “on the main Drakensberg” or “facing the main berg at 

heights” and occurs at higher altitudes (2070 – 2640 m) and has smaller leaves (Hilliard & 

Burtt, 1978, p. 57). Several taxonomic changes have been made within the species’ history 

((Hilliard & Burtt, 1978). Most notably, R. milloides (Baker) Nel. was originally considered a 

variety of Rhodohypoxis baurii by Jahrbücher in 1914 (R. baurii var. milloides (Hilliard & 

Burtt, 1978). When the genus was revised by Hilliard & Burtt in 1975 it was given species 

status as Rhodohypoxis milloides (Hilliard & Burtt, 1978). The most obvious and variable 

characters among the three varieties of R. baurii are the flower number and colour as flowers 

are either solitary or paired and can be pink, white, red or a mix within a single population.   

  

R. baurii var. baurii and R. baurii var. confecta both have erect leaves while R. baurii 

var. platypetala may have either spreading or erect leaves (Hilliard & Burtt, 1978). 

Regarding flower colour, R. baurii var. baurii flowers tend to be red and pink or very 

rarely white. In contrast, R. baurii var. platypetala generally has white flowers that are 

rarely pigmented, and R. baurii var. confecta flowers tend to be white, red or pink 

(Hilliard & Burtt, 1978). These three varieties differ mainly in habitat and show slight 

but definite ecological distinctions (Hilliard & Burtt, 1978). R. baurii var. baurii is 

found in moist habitats on damp and partly shaded cliff faces or in other damp habitats 

such as in small marshes and streambanks. Rhodohypoxis baurii var. confecta occurs in 

moist habitats on damp grass slopes, often among rocky outcrops or between rock 

sheets or on summits of plateaus in short damp turf among rock sheets and often occurs 

at higher altitudes (Hilliard & Burtt, 1978). In contrast,  R. baurii var. platypetala is 

found in dry habitats, often on shallow stony soils on rock sheets at lower altitudes 

(Hilliard & Burtt, 1978). Even in mixed populations that contain multiple varieties, R. 

baurii var. platypetala is found on drier ground than the other varieties. R. baurii var. 

confecta occurs at high altitudes (i.e., above 2600 meters above sea level (m a.s.l.) and 

forms sheets of mixed white, pink and red flowers (Hilliard & Burtt, 1978).   

 

 

  
 
 



 
  

      
  

24 

  
  

Figure 3. Rhodohypoxis baurii varieties A = Rhodohypoxis baurii var. platypetala, B = 
Rhodohypoxis baurii var. confecta, C = Rhodohypoxis baurii var. baurii. Varieties shown in 
the field (A1 = Karkloof, B1 = Sentinel, C1 = Tenahead Lodge) and individuals housed in the 
greenhouse (A2, B2, C2).   

  
  

A1   A 2   

B2   B1   

C1   C2   



 
  

      
  

25 

Rhodohypoxis populations are morphologically variable and isolated hybrid individuals and 

populations as well as introgressed populations occur (Hilliard & Burtt, 1978). This suggests 

that there may be few genetic barriers between varieties. Where R. baurii var. baurii and R. 

baurii var. platypetala occur sympatrically, hybrid lineages form, but are restricted to 

intermediate habitats and do not lead to merging of parent populations (Hilliard & Burtt, 

1978; Glennon pers. obs.). The degree of inter- and intra-specific allele sharing and gene flow 

that may occur between the species and varieties in Rhodohypoxis is unknown. Little is 

known about the reproductive strategies of R. baurii, yet the frequency of hybridization and 

allele sharing that have been observed suggest that few reproductive barriers exist between R. 

baurii varieties and other Rhodohypoxis species. Preliminary cytological analyses indicate 

varying levels of ploidy in and among the three currently recognised varieties: R. baurii var. 

baurii (2×), R. baurii var. platypetala (2×, 3× 4×), and R. baurii var. confecta (2×, 3×). Yet, 

the impacts of these differing ploidy-levels on reproduction and interbreeding are not fully 

understood. Due to the likely high levels of gene flow between the varieties, distinguishing 

morphological features are largely lacking. Gene flow between varieties and closely related 

species may be facilitated by polyploidy and reflected in morphology. Quantifying 

morphology, potential reproductive isolation, and reproductive strategies, together with 

assessing ploidy-level variation, in R. baurii will be useful to describe the role that polyploidy 

may have played or potentially will play in this group’s evolutionary past and its future 

trajectory. Therefore, the overall aim of this study is to assess the influence of ploidy on 

divergence in morphology, genetic differentiation and reproductive strategy amongst varieties 

of Rhodohypoxis baurii (Hypoxidaceae) as well as to better distinguish the three 

Rhodohypoxis baurii varieties.   

  

  

  

  

  

  

  

  

  

  

    



 
  

      
  

26 

1.3 Aim and objectives   

The overall aim of this study is to assess the relationship between ploidy and variation 

in morphology, genetic differentiation and reproductive strategy amongst varieties of 

Rhodohypoxis baurii (Hypoxidaceae) as well as to better distinguish the three 

Rhodohypoxis baurii varieties.  

  

Four main objectives are associated with the aim of this study:   

1. Use morphometric tools to assess how the three Rhodohypoxis baurii varieties may be 

morphologically distinguished from one another, and identify if morphological 

differences coincide with ploidy rather than taxonomy.  

2. Use molecular data to estimate the gene flow, genetic differentiation and genetic 

diversity between and among the three R. baurii varieties.  

3. Evaluate if different varieties or ploidies in R. baurii differ in reproductive morphology 

and/or adopt different reproductive strategies.  

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

    



 
  

      
  

27 

CHAPTER 2: METHODS  

Morphometrics Analysis  

A total of 126 herbarium specimens of Rhodohypoxis baurii were examined from Pretoria 

National Herbarium (PRE), H.G.W.J. Schweickerdt Herbarium (PRU), CE Moss Herbarium 

(J) and Bews Herbarium (NU) for morphometric analysis (Appendix, Table 1), including the 

lectotype of R. baurii var. confecta and specimens from the type localities of R. baurii var. 

baurii and R. baurii var. platypetala that were  good matches of the type specimens (as 

compared with scans received from K and TCD respectively). The specimens were 

representative of the morphological variation across the geographical distribution of the 

species and its three varieties (Fig. 6). The specimens were initially sorted into 

varieties/groups based on the collector’s identification, then further sorted based on the 

distinguishing characteristics noted in the original descriptions by Hilliard & Burtt, (1978). 

For each specimen, 21–32 floral, vegetative and reproductive characters were measured 

depending on the quality and completeness of the specimens used (Table 1; Fig. 4; Fig. 5). 

The locality of each specimen as well as altitudinal data were recorded. Mature leaves and 

mature flowers were used for all leaf and flower related measurements, respectively. A 

minimum of three measurements were taken for each feature and the average was log 

transformed and used for analysis. Qualitative features were scored as either binary or ordered 

multi-state characters and used to construct a numerical matrix for analysis using PCoA and 

NMDS. The analyses were conducted on both the quantitative and qualitative traits separately 

as well as on the combine matrix.  

 

 

 

 

 

 

 

 

 

 



 
  

      
  

28 

Table 1. Character list for multivariate analysis of Rhodohypoxis baurii complex 
including all floral, vegetative, and reproductive traits. For all trichome branching 
pattern types (traits 18–21) and sheath types (trait 23), refer to Figure 4 below, and for 
reproductive traits (traits 26–32) refer to Figure 5 below. 
 

Characters   Averaging / Coding  
Peduncle length (mm)     Single measurement  
Leaf lamina length (mm)  Average of 3 measurements   
Leaf lamina width (at widest 
point)(mm)  

Average of 3 measurements   

Receptacle diameter (mm)    Single measurement 
Peduncle trichome length (mm)  Average of 3 measurements   
Receptacle trichome length (mm)  Average of 3 measurements   
Leaf trichome length (mm)  Average of 3 measurements   
Tepal trichome length (mm)  Average of 3 measurements   
Flowers/ peduncle (mm)    ?? 
Outer tepal length (mm)  Average of 3 measurements   
Outer tepal width (mm)  Average of 3 measurements   
Inner tepal length (mm)  Average of 3 measurements  
Inner tepal width (mm)  Average of 3 measurements  
Trichome density on peduncle  Absent (0)  Sparse (1)  Dense (2)  
Trichome density on leaf lamina  Absent (0)  Sparse (1)  Dense (2)  
Trichome density on receptacle  Absent (0)  Sparse (1)  Dense (2)  
Trichome density on tepal  Absent (0)  Sparse (1)  Dense (2)  
Trichome branching pattern peduncle  Simple (0)  Bifid (1)  Stellate even (2)  Stellate uneven (3)  
Trichome branching pattern leaf lamina   Simple (0)  Bifid (1)  Stellate even (2)  Stellate uneven (3)  
Trichome branching pattern leaf margin  Simple (0)  Bifid (1)  Stellate even (2)  Stellate uneven (3)  
Trichome branching pattern tepal  Simple (0)  Bifid (1)  Stellate even (2)  Stellate uneven (3)  
Flower colour   White (0)  Light pink (1)  Dark pink (2)  
Spiky fibrous sheaths   Present (0)  Absent (1)  
Tufted trichomes on tepal tips  Present (0)  Absent (1)  
Trichome density pattern on leaves   Absent (0)  Sparse (1)  Dense (2)  
Anther length  Average of 3 measurements  
Anther width  Average of 3 measurements  
Anther lobe width  Average of 3 measurements  
Stigma length   Single measurement 
Stigma lobe length    Single measurement 
Stigma lobe width    Single measurement 
Ovule diameter across centre  Average of 3 measurements  



 
  

      
  

29 

 
Figure 4: A–D) Trichome branching patterns on leaves across R. baurii varieties showing (A) 
single trichome on R. baurii var. baurii from Hebron Farm, (B) bifid trichome on R. baurii 
var. baurii from Impendhle region in Natal, (C) uneven stellate trichome with a single long 
trichome surrounded by a stellate cluster at the base seen on R. baurii var. confecta, and (D) 
typical even stellate trichome on R. baurii var. baurii from Vrederus. E–F). Differences in 
sheaths surrounding the base on the peduncle and leaves just above the bulb with (E) showing 
‘spiky’ fibrous sheath hairs on R. baurii var. platypetala from Karkloof and (F) showing a 
smooth sheath on R. baurii var. confecta from Sentinel population. Scale bar = 2.8 mm.  

  

  

A   

D   C   

B   

F   E   



 
  

      
  

30 

 

 
Figure 5. Reproductive morphology measurements in R. baurii varieties: (AL) Anther 
Length, (AW) Anther Width, (ALW) Anther Lobe Width, (SLL) Stigma Lobe Length, (SLW) 
Stigma Lobe Width, (OD) Ovule Diameter, (SL) Stigma Length.  

  
Morphometric analysis of the 32 morphological features of 126 specimens was used to 

confirm the identity of the three putative Rhodohypoxis baurii varieties. A matrix containing 

20 quantitative and 12 qualitative characters was constructed using the measurements of the 

characteristics for all the specimens. A PCA was conducted on the combined data to generate 

eigenvalues. A PCA was conducted on both logged transformed measurements and unlogged 

measurements. Both logged and unlogged data resulted in the similar values and trends were 

the same between logged and unlogged data. Traits that exhibited very low eigenvalues 

(<0.0001), indicating a lack of variability, were removed from the data set (Appendix, Table 

3). Traits that were not significant and did not help distinguish the three varieties were 

removed. A total of 8 traits were removed (tufted trichomes on tepal tips, peduncle trichome 

length, tepal trichome length, trichome density on peduncle, trichome branching pattern 

peduncle, trichome branching pattern on leaf margins, trichome branching pattern on tepals 

and ‘spiky’ fibrous sheath hairs presence), with 16 quantitative and 8 qualitative traits 

remaining (Table 3). All trichome traits removed had eigenvalues that were near 0 (i.e. 

0.0000002) and were causing morphologically different individual to group together 

AL   

SL   

OD   

SLW   
SLL   

AW   
ALW   



 
  

      
  

31 

incorrectly. Therefore, these traits were removed.  A Principal Coordinates Analysis (PCoA) 

and Non-Metric Multidimensional Scaling analysis (NMDS) were conducted in R v3.4.3 

(RStudio Team, 2015) using the package ‘vegan’ (Oksanen et al., 2022). The Gower general 

similarity coefficient was used for the PCoA as it is best suited for examining similarities and 

differences between individuals/varieties for morphometric data input (Gower, 1971). PCoA 

was used rather than a Principal Component Analysis (PCA) as the data set contains both 

qualitative and quantitative characters and a PCoA can be used with limited missing data 

(Borcard et al., 2018). In addition, a PCA was conducted on the remaining quantitative data 

to generate eigenvalues and to assess the relative roles of quantitative vs. qualitative 

characters. Important distinguishing morphological features that had a high eigenvalue 

(shown via a PCA) were selected for direct comparison using box and whisker box plots to 

compare means ± standard errors (SE). An additional 35 individuals were collected from 

seven populations (Table 2) in South Africa and near the Lesotho border and were housed in 

the greenhouse at Wits University. Measurements of these individuals were compared 

separately to the herbarium specimens using box and whisker plots. The greenhouse and 

herbarium specimens were analysed separately as the morphology of the collected specimens 

changed slightly in greenhouse conditions. For example, many of the plants in the greenhouse 

had longer peduncles and leaves compared to the natural populations (and herbarium 

specimens), likely due to the lack of wind in the greenhouse. High eigenvalues indicate 

features that had a considerable influence in the directionality and positioning of the 

individuals on the principal axes. An ANOVA was conducted in R v3.4.3 (RStudio Team, 

2015) to evaluate whether any significant differences exist in selected morphological features 

among, and between, the Rhodohypoxis baurii varieties. Two one-way ANOVAs were run; 

the first with variety as the fixed effect (testing if trends in morphological trait are due to 

variety) and the second with ploidy as the fixed variable (testing if trends in morphological 

traits are due to variety). Additionally a two-way multifactorial ANOVA was run with both 

ploidy and variety entered as explanatory variables. The results did not differ between the 

two types of ANOVA’s. Tukey post-hoc analyses were performed on both ANOVA types to 

test for paired differences post ANOVA at a 95% confidence interval (a = 0.05).  

 

 



 
  

      
  

32 

 

Figure 6. Geographic distribution of Rhodohypoxis baurii herbarium specimens used 
for morphometric analyses in this study. Dark pink: Rhodohypoxis baurii var. 
baurii, light pink: Rhodohypoxis baurii var. confecta, white: Rhodohypoxis baurii 
var. platypetala and pale blue: mixed individuals.  
  

Flow cytometry was used to estimate the ploidy levels of 86 R. baurii var. baurii 

individuals, 30 R. baurii  var. confecta individuals and 31 R. baurii var. platypetala 

individuals that were housed in the greenhouse. All individuals were housed in the 

same greenhouse under the same growth conditions.  Greenhouse conditions 

attempted to mimic natural environmental conditions with the average summer 

temperatures ranging from 15°C to 30°C.. Flow cytometry provides an estimate of 

DNA content relative to a set standard from which ploidy level can be inferred. Flow 

cytometry was performed following Doležel et al., (2007) at the H.G.W.J. 

Schweickerdt Herbarium (PRU), University of Pretoria. Oxalis articulata Savigny 

genome size is known (pg = 1.96) and was used as the standard for flow cytometry 

on Rhodohypoxis individuals. Approximately 1 cm2 of O. articulata and 1 cm2 of R. 

baurii leaf tissue were finely chopped together using razor blades in 500 µl of Otto I 

lysis buffer. The solution was filtered through a 30 µm mesh filter to and incubated 

for 15 minutes at room temperature. One milliliter of Otto II buffer containing the 

fluorescent dye DAPI (4’, 6-diamidino-2-phenylindole) was added to the solution to 

stain the cells and left to incubate at room temperature for a further 15 – 40 min. The 

samples were analysed using a CyFlowR Space flow cytometer. FloMax® software 

(Partec GmbH, 2001) was used to calculate the mean size of each peak (sample and 

  
60 k 
m   



 
  

      
  

33 

standard) and calculate the coefficient of variation. Relative DNA content was 

calculated. Ploidy levels were inferred by correlating the estimated DNA content to 

the base number of somatic chromosomes presented by Saito (1975).   

 

Microsatellites Data Collection  

Leaf material was collected from 237 individuals of Rhodohypoxis baurii (R. baurii 

var. confecta n = 88, R. baurii var. baurii n = 87, R. baurii var. platypetala n = 62) 

across 11 populations (Figure 7; Table 2).  

Table 2. Sampled localities, GPS co-ordinates and ploidy-levels present of 11 
populations of Rhodohypoxis baurii varieties used in molecular microsatellite analyses 
and eight populations used in morphometric analyses (bolded/highlighted). No fresh 
leaf tissue was available for Sehlabethebe and Sani Chalet populations for flow 
cytometry, therefore ploidy levels were estimated based on microsatellite data (inferred 
from dosage). It is important to note that pure R. baurii var. confecta individuals are 
diploids only, but hybridization of this variety with other Rhodohypoxis species may 
lead to polyploid hybrids as observed in the Platberg population of R. baurii var. 
confecta.  

Variety  Locality  Altitude 
(m a.s.l.)  

GPS co-ordinates  Ploidy levels 
present  

R. baurii var. baurii  Naude’s Nek  2400  -30.731515, 28.135147  Tetraploid (4×)  
R. baurii var. baurii  Vrederus  1950  -30.790036, 28.262599  Tetraploid (4×)  
R. baurii var. baurii  Tenahead lodge  2500  -30.709617, 28.136370  Tetraploid (4×)  
R. baurii var. confecta  Sentinel   2600  -28.751953, 28.890260  Diploid 2×  
R. baurii var. confecta 
(Hybrid population)  

Platberg  2300  -28.245808, 29.161346  Diploid, triploid 
and tetraploid  

R. baurii var. platypetala  Golden Gate  2059  -28.517355, 28.616570  Diploid 2× 
R. baurii var. platypetala 
&  R. baurii var. baurii  

Hebron Farm  1680  -33.441026, 150.868042  Diploid, Triploid 
and tetraploid  

R. baurii var. platypetala  Karkloof  1770  -29.298840, 30.269599  Diploid and 
tetraploid  

R. baurii var. platypetala  Sunset Farm   1934  -29.742937, 30.550833  Diploid, triploid 
and tetraploid  

R. baurii var. baurii  Sehlabethebe  2800  -29.824725, 29.200379  Diploids and 
polyploids  

R. baurii var. confecta  Sani Chalet  2900  -29.585149, 29.288145  Diploids and 
polyploids  

  



 
  

      
  

34 

 
Figure 7. Distribution map of Rhodohypoxis baurii varieties with pink areas showing 
the known range of the varieties, and maroon points showing the 11 sampled 
populations.  

  
  

The collected leaf material was dried in silica gel immediately to preserve the DNA 

for extraction. DNA was extracted from the sampled leaf material using a modified 

CTAB method following  Doyle & Doyle, (1987) and Cullings, (1992). The CTAB 

extraction method was optimized for Rhodohypoxis with incubation times specified as 

follows. After addition of the CTAB buffer, the samples were incubated at 55°C for 

90 minutes and after the addition of isopropanol the samples were incubated at −80°C 

for 5 minutes and then moved to −4 °C for an additional 15 minutes. These 

incubations times were optimal for large amounts of high-quality DNA with minimal 

contaminants that could inhibit PCR amplification. In the case of a very thin aqueous 

phase being present after the initial chloroform:isoamyl alcohol step, the 

chloroform:isoamyl alcohol step was repeated and samples re-centrifuged before 

pipetting off the aqueous phase. Ammonium acetate was excluded from this modified 

protocol. All DNA samples were quantified using a microvolume UV-Vis 

spectrophotometer (Nanodrop, Thermo Fisher Scientific) and good quality genomic 

DNA (A260/280 1.6 > 2.2) was diluted to a final concentration of 25–50 ng/µL before 



 
  

      
  

35 

amplification. Samples with low quality genomic DNA were re-extracted permitting 

enough leaf tissue remaining.   

Sixty microsatellite markers were developed for Rhodohypoxis by AllGenetics 

(AllGenetics & Biology SL, A Coruña, Spain). The 60 primers developed were trialled 

and 28 amplified successfully; the 12 most variable markers were selected, optimized, 

and used in this study (Appendix, Table 2). The oligonucleotide tails used were the 

universal sequences CAG (CAG TCG GGC GTC ATC), and T3 (AAT TAA CCC TCA 

CTA AAG GG). The two oligonucleotides were labelled with the FAM dye, and the 

NED dye, respectively. Amplifications were performed using a single-plex fluorescent 

microsatellite PCR reaction. Amplifications were carried out using a BioRad T100TM 

thermal cycler using the following cycling profile: initial denaturation at 95 °C for 1 

minute, followed by 33-35 cycles of 95 ºC for 20 s, 57 ºC for 60 s, 68°C for 45 s; and a 

final extension step at 68 ºC for 5 min. Success of the amplification was assessed via gel 

electrophoresis by mixing 3 µl 6x loading gel to 5 µl amplification reaction and loading 

samples onto a 2% agarose gel stained with SYBRSafe dye. The successfully amplified 

products were sent for fragment analysis at the Central Analytical Facility (CAF), 

Stellenbosch. The resulting chromatographs were scored manually. Peaks with a 

fluorescent intensity greater than 100 were scored and peaks with a fluorescent intensity 

less than 100 were removed. Distance between scored peaks was 4 base units or greater 

in order to avoid double scoring peaks that had stutters due to replication slippage. 

 

A matrix was generated based on the scored peaks using the microsatellite plug in for 

Geneious Prime Version 11.0.6+10 (Biomatters Development Team, 2005).  As the 

relative fluorescence of the peaks varied based on the success of the PCR reaction, 

dosage could not be inferred from the height (fluorescent intensity) of the peaks. 

Therefore, in all subsequent analyses using Genodive version 3.05 (Meirmans, 2020) 

the option to “correct for unknown dosage of alleles” was chosen. The method 

implemented by Genodive version 3.05 (Meirmans, 2020) to correct for unknown 

dosage assumes random mating within populations. The dosage correction uses a 

maximum likelihood statistical method based on random mating within populations. 

For every incomplete marker phenotype (e.g. AB) it is possible to calculate the 

likelihood of observing that phenotype given some set of allele frequencies, by 

combining the likelihoods of the underlying genotypes (AAAB, AABB, or ABBB) 



 
  

      
  

36 

(Meirmans, 2020). The algorithm starts by calculating the likelihood of all observed 

phenotypes for a population, given the uncorrected (biased) allele frequencies 

(Meirmans, 2020). The allele frequencies are then slightly changed and the likelihood 

is recalculated. If the fit improves, the changed allele frequencies are accepted and the 

old ones discarded (Meirmans, 2020). This process is continued until convergence is 

reached.  

 

Microsatellites Data Analyses   

Microsatellite data analyses were conducted to evaluate the extent of genetic 

differences and diversity between and within varieties, ploidy levels, and geographic 

populations. To compare genetic diversity both expected heterozygosity (HE) and 

observed heterozygosity (Ho) were calculated using Genodive version 3.05 

(Meirmans, 2020). Genetic distances, percentage polymorphic loci, Nei’s gene 

distance index and inbreeding coefficient (F), were calculated and evaluated using 

Genodive (Meirmans, 2020). Any departures from HWE were assessed for each locus 

to ensure that the locus is not under selection and is suitable for further analyses. Nei’s 

genetic distance was used to evaluate how genetically distinct the groups are and to 

what extent the taxa or geographic populations share alleles. The dataset was also 

evaluated for the presence of clones (see objective 3: Reproductive strategies) and 

clonal individuals were removed before performing population structure analyses.   

  

The dataset contains mixed ploidy levels (diploids and polyploids). Therefore, the data 

were divided/ transformed to create five different data sets before performing 

population structure analyses. Firstly, the data set was divided into subsets based on 

ploidy levels, with diploids and tetraploids analysed separately, as well as together in a 

combined data set. The combined ploidy data set was further transformed with all 

diploids transformed into auto-polyploids by replicating the diploids alleles to create a 

polyploidized data set based on the maximum likelihood of different genotypes given 

the population allele frequencies (with the assumption of random mating within a 

population) (i.e. 170, 182, 0, 0 diploid genotypes were changed to 170, 182, 170, 182). 

Lastly, all polyploid individuals in the combined data set were made into diploids by 

removing alleles based on the most likely allele combinations using maximum 



 
  

      
  

37 

likelihood of different genotypes given the population allele frequencies with the 

assumption of random mating within a population. Removed alleles were made into a 

subsequent second individual, resulting in a diploidized data set. All data 

transformations were done in Genodive (Meirmans, 2020). Auto-polyploidizing the 

data set reinforces trends that already exist, but may reinforce trends that are not 

necessarily represented in the initial dataset. Therefore, it is important to perform 

analyses on the original data set as well as the transformed data sets to ensure trends 

stay the same and to ensure congruency and validity of the results.   

 

Admixture (gene flow) was estimated and population structure was assessed using 

STRUCTURE 2.3.1 (Falush et al., 2007). STRUCTURE 2.3.1 uses a clustering method to 

assign individuals to clusters where Hardy-Weinberg Equilibriums (HWE) is achieved. To 

estimate K clusters, an admixture model with correlated allele frequencies was used, with an 

initial burn-in of 25 000, followed by 250 000 MCMC iterations. The K (possible number of 

genetic clusters) was set from one to 13 and with eight replicates per K value. K was set from 

one to 13 to accommodate the possible outcome of one genetic cluster (as they all belong to a 

single species) and the possible outcome of up to 13 genetic clusters due to the 11 populations 

present. The best fit value of K for the data was identified using STRUCTURE 

HARVESTOR (Earl & VonHoldt, 2011, https://taylor0.biology.ucla.edu/structureHarvester/) 

in order to identify the number of genetic clusters present and and whether these genetic 

cluster coincide with the recognised varieties, geographic populations, or ploidy. The 

proportion and pattern of shared alleles did not differ between the diploidized and 

polyploidized data sets, therefore the polyploidized data set was used for ordination and 

phylogenetic network analyses.  

 

In the initial round of ordination analyses, two non-parametric methods were used to analyse 

the microsatellite data: Principal Component Analysis (PCA) and Discriminant Analysis of 

Principal Components (DAPC). While both PCAs and DAPCs provide a multivariate 

summary of genetic data, a DAPC also assesses the fit of data to a varying number of 

population clusters (Everson et al., 2021). Therefore, a DAPC yielded more accurate results 

for our data over a PCA. A Bruvo genetic distance matrix was generated using Genodive 

(Meirmans, 2020) and used to conduct a DAPC in R v3.4.3 (RStudio Team, 2015) using the 

package ‘adegenet’(Jombart, 2008). Lastly, the Bruvo genetic distance matrix was imported 



 
  

      
  

38 

into SplitsTree (Huson & Bryant, 2021) to generate a NeighborNet phylogenetic network and 

Neighbour Joining consensus tree to understand the relationships and reticulation events 

between and within the three varieties.  

 

Reproductive strategies  

To identify the breeding systems of Rhodohypoxis baurii, and to evaluate viability of crosses 

between varieties, greenhouse plants were divided into five experimental groups to test for the 

presence of the following: clonality, autonomous self-pollination, hand-manipulated self-

pollination, cross-pollination by hand and agamospermy/apomixis.  

 

Clonal reproduction  

To check for and infer clonal reproduction, root connections were examined by 

uprooting plants that were originally planted as single individuals, but where the pots 

subsequently contained three or more individual stems within the same pot. Ten pots 

per variety (R. baurii var. baurii, R. baurii var. confecta and R. baurii var. platypetala) 

containing multiple plants were uprooted and the number of root connections recorded. 

From each of these pots, leaf tissue for three individuals per pot was collected for a 

total of 30 tissue samples per variety (90 tissue samples in total). DNA extractions and 

microsatellite amplification were performed (see 2.2 Microsatellites Data Collection) 

to confirm genetic similarity of these individuals. Genetic similarities and extent of 

clonality were identified using Genodive (Meirmans, 2020). If genetically identical, 

these individuals represent clonal reproduction. Uprooted plants were repotted with a 

single plant/bulb per pot. The pots were observed after one year and any new sprouts 

or root connections that formed were recorded.  

 

Autonomous-self-pollination and Hand-self pollination  

Five to fifteen greenhouse plants per variety were selected and bagged to test for 

autonomous self-pollination. All already open flowers were removed, and flower buds 

were enclosed in mesh organza bags. Hand self-pollinations were performed by 

removing all dehiscent anthers of a single flower and the pollen was applied to the 

stigma on the same flower by picking a dehiscent anther with forceps and gently 

dabbing the pollen onto the stigma, with care taken not to damage the stigma or tepals. 



 
  

      
  

39 

All flowers were then bagged using mesh organza bags and left to fruit. Seed set was 

recorded, and percentage of the seeds germinated was used to assess seed viability. 

Seed germination trials were carried out in the greenhouse. Seeds were placed in 

between pre-packaged sterile cotton wool that was dampened with autoclaved water, 

then placed in sealed plastic pre-packaged sterile containers and left in the greenhouse 

with average daytime temperatures between 25°C and 29°C under ambient light 

conditions. Seeds were checked weekly and all germinated seeds recorded. The 

germination trails were conducted for three consecutive months. Seeds that had not 

germinated after three months were marked as inviable. No tetrazolium salt staining 

was conducted on the seeds as the seeds were too small to yield accurate viability 

results from such staining techniques.    

Cross-pollination viability   

Ten to twenty plants of each variety were selected for cross-pollination by hand. 

Crosspollinated flowers were emasculated prior to the anthers maturing, with care 

taken not to transfer any pollen onto the stigma of the same plant. Pollen from a 

different variety was transferred onto the emasculated maternal plant’s stigma for 

inter-varietal crosses, and pollen from the same variety was transferred onto the 

emasculated maternal plant’s stigma for intravarietal crosses by picking a dehiscent 

anther with forceps and gently dabbing the open anther onto the stigma to transfer 

pollen, taking care not to damage the stigma or tepals. Intervarietal and intra-varietal 

crosses were performed between all three varieties, with at least 10 crosses 

representing each cross type; (cross types = b x c, b x p, c x p, c x c, b x b, p x p; Fig. 

8). All flowers were bagged and left to set seed. Seed set, seed number, and seed 

viability (via germination) were recorded for each cross.  

  



 
  

      
  

40 

 
Figure 8. Both inter-varietal and intra-varietal crosses were conducted between the three 
Rhodohypoxis baurii varieties (b = R. baurii var. baurii, c = R. baurii  var. confecta, p = 
R. baurii var. platypetala) with all varieties acting as both pollen donors (paternal) or 
pollen receivers (maternal). Arrows indicate direction of pollen transfer.  

Agamospermy/apomixis  

Ten to fifteen flowers of each variety were emasculated to test for agamospermy. 

After emasculation, all flowers were bagged using mesh organza bags and left to set 

seed. For all the above treatments, the number of seeds produced (if any) per 

plant/treatment were recorded and all seeds produced were germinated to test for 

viability.  

Pollinator identification   

In the field, pollinator observations were conducted at seven sites (Golden Gate National 

Park, Sentinel, Langalibalele Pass, Karkloof Nature Reserve, Minerva Reserve, Highmoor  

Nature Reserve and Vrederus Trout Farm). Pollinator observations were conducted between 

November and January 2021/2022 at the peak of the Rhodohypoxis baurii flowering season 

for one to two days per site. Observations were conducted for two continuous hours every six 

hours within a 24-hour period (i.e., for a total of eight hours within a 24-hour period), starting 

at approximately 6 am, 12 pm and  7 pm for the various observation sessions. Observations 

were conducted during both day and night hours. The frequency of insect visitation per plant 

and the type of visitation/ interaction was recorded and visiting insect(s) were caught when 

possible and placed in separate vials to avoid cross-contamination of pollen loads. The insects 

were identified to genus or species level using ‘Field guide to insects of South Africa’ by 

  
  



 
  

      
  

41 

Picker et al. (2004), and confirmed via consultation with entomologists. Pollen was washed 

off caught insects using ethanol and stained using Alexander’s stain following Peterson et al.'s 

(2010) protocol. After staining, three to four drops of distilled water were used to rinse the 

slides and excess fluid absorbed with filter papers. Pollen slides were sealed and fixed using 

Hoyer’s mounting medium (Anderson, 1954). Any pollen grains present were compared to 

Rhodohypoxis baurii pollen from greenhouse plants to ascertain if visiting insects were 

possibly acting as pollinators.    

 

Pollen viability   

In order to assess pollen viability, all anthers were removed from a single flower and 

were immediately placed into 1.5 ml microcentrifuge tubes containing Carnoy’s 

Fixative to fix the pollen. The top three anthers (top whorl) and bottom three anthers 

(bottom whorl) were collected into separate microcentrifuge tubes as it was 

hypothesized that the whorls may mature at different times. The anthers and pollen 

were left in the Carnoy’s Fixative for a minimum of 48 hours and a maximum of four 

months. Thereafter, both the top and bottom anthers were removed, and pollen was 

stained using Alexander’s stain following Peterson et al.'s (2010) protocol. After 

staining, three to four drops of distilled water were used to rinse the slides and excess 

fluid absorbed with filter paper. Pollen slides were sealed and fixed using Hoyer’s 

mounting medium. The numbers of viable (red) and inviable (green) pollen grains 

(Fig. 9) were counted per slide using a Zeiss compound light stereo-microscope at 

100× magnification.   

 

 



 
  

      
  

42 

 
Figure 9.  Rhodohypoxis baurii pollen grains stained with Alexander’s stain showing 
viable pollen (red) and inviable pollen (green/blue).   

    
CHAPTER 3: RESULTS  

 

Morphometric analysis of R. baurii varieties and ploidy levels   

Of the 126 herbarium specimens selected for morphometric analyses, 10 specimens 

were removed from the data set as they were identified as hybrid individuals (as they 

fell far outside the morphometric space of the other individuals/ known varieties) and 

were from localities where high levels of introgression and hybridization between R. 

baurii and other Rhodohypoxis or Hypoxis species were noted by Hilliard & Burtt 

(1978). A total of 8 traits were removed (tufted trichomes on tepal tips, peduncle 

trichome length, tepal trichome length, trichome density on peduncle, trichome 

branching pattern peduncle, trichome branching pattern on leaf margins, trichome 

branching pattern on tepals and ‘spiky’ fibrous sheath hairs presence), with 16 

quantitative and 8 qualitative traits remaining (Table 3). Clear overlap of the 

specimens in the ordination analyses meant that little morphological distinction was 

found between the three varieties (Fig. 10), when qualitative data were excluded, no 

separation of the varieties was found with all three varieties overlapping completely. 

Rhodohypoxis baurii var. platypetala shows the largest morphological variation and 

overlaps mainly with R. baurii var. confecta to some extent, and R. baurii var. baurii 

and R. baurii var. confecta overlap considerably, indicating that they are less 

morphologically distinct from each another. Three outlier individuals from the 

Northern and Central Drakensberg were identified in the R. baurii var. confecta group 

Inviable pollen grains   

Inviable  
pollen  
grains   

Viable  pollen  
grains   

Viable  
pollen  
grains   

50 um   50 um   



 
  

      
  

43 

(Hc113, Hcpv122 and Hcm114; Fig. 10) that were collected from Cathkin Peak, Mont 

Aux Sources and Cathedral Peak, respectively. These three localities have multiple 

Rhodohypoxis and Hypoxis species present with introgression and hybridisation 

possibly occurring.  

 

Table 3: List of contributions (loadings) of each morphological character measured 
for the first four Principal Component Axes, based on the PCA analysis. 
 PC 1 PC 2 PC 3 PC 4 
Peduncle Length (mm) 0.934 -0.357 -0.007 -0.002 
Leaf length 0.897 0.441 -0.001 0.002 
Leaf width 0.220 -0.298 0.262 0.713 
Peduncle trichome length 0.390 0.002 0.214 0.417 
Ovary trichome length 0.328 0.088 0.191 0.369 
Leaf trichome length 0.331 -0.046 0.045 0.520 
Tepal Trichome length 0.338 0.131 0.181 0.313 
# Flowers/ peduncle 0.231 -0.081 -0.016 -0.127 
Outer tepal length 0.569 -0.040 0.787 -0.146 
Outer tepal width 0.358 -0.161 0.763 0.038 
Outer tepal ratio -0.103 0.232 -0.080 0.177 
Trichome density peduncle -0.155 0.132 0.148 0.188 
Trichome density leaf -0.357 -0.182 0.054 0.014 
Trichome density ovary 0.251 -0.065 -0.015 0.281 
Trichome density tepal -0.0689 -0.155 0.145 0.157 
Trichome branching pattern peduncle -0.202 -0.065 0.113 -0.086 
Trchome branching pattern leaf centre 0.013 0.167 -0.135 -0.683 
Trichome branching pattern leaf margin and 
midrib  0.129 -0.013 -0.030 -0.011 
Trichome branching pattern tepal 0.072 -0.136 0.005 -0.163 
Flower colour (as noted by collector) -0.127 0.341 -0.316 0.192 
Spiky sheaths  -0.141 -0.099 0.051 -0.033 
Tufted trichomes on tepal tips 0.025 -0.028 -0.024 0.153 
Trichome density pattern on leaves  -0.175 -0.197 0.278 -0.242 
 

 

 

 



 
  

      
  

44 

 
 

 
Figure 10. Non-metric multidimensional scaling (NMDS) analysis of 116 herbarium 
specimens of R. baurii varieties based on 23 morphological traits using the Gower 
similarity coefficient. A) PC axes 1 and 2 explain 65% of the variation, B) PC axes 2 
and 3 explain 54% of the variation. Three outliers of R. baurii var. confecta (arrowed) 
showed traits commonly seen in Hypoxis species: Hcm104 from Cathedral Peak; 
Hc113 from Monks Cowl Reserve and Hcpv122 from Mont Aux Sources.  

  
 
  

PC 1 

PC
 2

 

A 

  PC 2   

PC
 3

   
 

  

R. baurii  var.  baurii   
R. baurii  var.  platypetala   
R. baurii  var.  confecta   
  

B 

R. baurii  var.  baurii 
R. baurii  var.  platypetala 
R. baurii  var.  confecta 



 
  

      
  

45 

Distinguishing morphological features and those with high eigenvalues were selected 

for direct comparison using box and whisker box plots (Fig.11–15). Due to the 

polyploid nature of the sampled R. baurii individuals, it is important to note that 

population ploidy levels may have changed over time, and changes in ploidy may lead 

to morphological changes and shifts. Therefore, herbarium specimens and greenhouse 

individuals were analysed separately. Trichome traits such as the length of trichomes 

present on the peduncle of herbarium specimens did not differ significantly between 

varieties (df = 108, F = 1.59, P = 0.21) and leaf lamina length did not differ 

significantly between varieties (df = 108, F = 2.18 P = 0.12) (Fig. 11). Leaf widths of 

R. baurii var. baurii herbarium specimens was significantly smaller than those of R. 

baurii var. confecta (df = 105, F = 3.41 P = 0.032) and R. baurii var. platypetala (df = 

105, F = 3.41, P = 0.012) (Fig. 11). Certain traits of Rhodohypoxis baurii var. 

platypetala specimens were significantly larger than in R. baurii var. baurii and R. 

baurii var. confecta, such as peduncle length (df = 108, F = 8.03, P = 0.00056), outer 

tepal width (df = 100, F = 8.37, P = 0.00043) and outer tepal length (df = 100, F = 

10.48, P < 0.001) (Fig. 11).   

  



 
  

      
  

46 

 
Figure 11. Comparison of vegetative, and floral morphological traits of 116 herbarium 
specimens represented by Rhodohypoxis baurii varieties:  baurii = Rhodohypoxis baurii 
var. baurii, confecta = Rhodohypoxis baurii var. confecta and platypetala = 
Rhodohypoxis baurii var. platypetala. All measurements were log transformed and 
significant differences are indicated above each box to a 95% confidence interval (α = 
0.05). The compact letter display is based on Tukey Posthoc tests derived from two-way 
ANOVAs.   

  
Rhodohypoxis baurii individuals housed in the greenhouse were collected throughout the 

Drakensberg region between 2017 and 2020 and all measurements were conducted in 2021. A 

total of seven vegetative, peduncle and trichome traits were measured across 43 greenhouse 

individuals, and 12 floral traits measured across 31 greenhouse individuals. Rhodohypoxis 

baurii var. platypetala has significantly larger floral features than R. baurii var. baurii and R. 



 
  

      
  

47 

baurii var. confecta, such as outer tepal width (df = 26, F = 27.69, P < 0.001), inner tepal 

length (df = 26, F = 8.09, P = 0.0019) and inner tepal width (df = 26, F = 6.39, P = 0.0055), 

but has significantly shorter leaves than R. baurii var. confecta (df = 21, F = 3.66, P = 0.043) 

(Figs.12, 13). Rhodohypoxis baurii var. platypetala and R. baurii var. baurii have, on average, 

similar leaf widths (df = 21, F = 3.66, P = 0.24), whereas R. baurii var. platypetala has 

significantly wider leaves than R. baurii var. confecta (df = 21, F = 3.96, P = 0.028) (Fig. 13).  

Rhodohypoxis baurii var. baurii and R. baurii var. confecta do not differ in most floral size 

measurements, including outer tepal length (df = 26, F = 3.81, P = 0.90), inner tepal length 

(df = 26, F = 8.09, P = 0.75) and inner tepal width (df = 26, F = 6.39, P = 0.50) (Fig. 12). 

Female reproductive traits do not differ between the three varieties, such as pistil length (df = 

26, F = 6.39, P = 0.309) and ovule width (df = 29, F = 0.76, P = 0.48) (Fig. 12). On the other 

hand, male reproductive traits namely anther length and anther width do differ, where R. 

baurii var. platypetala has significantly longer anthers than R. baurii var. baurii (df = 29, F = 

2.35, P = 0.096) (Fig. 12; Fig. 14). Flower colour is one of the main distinguishing features 

used to identify R. baurii varieties, with R. baurii var. baurii exhibiting dark pink tepals, R. 

baurii var. confecta flowers showing various shades of white and pinks, and R. baurii var. 

platypetala showing exclusively white flowers with the occasional pink flushes on tepal tips 

(Fig. 14; Fig. 15).  

  

  



 
  

      
  

48 

  
  

Figure 12. Comparison of floral morphological traits of 31 collected greenhouse 
specimens represented by Rhodohypoxis baurii varieties: baurii = Rhodohypoxis baurii 
var. baurii, confecta = Rhodohypoxis baurii var. confecta and platy = Rhodohypoxis 
baurii var. platypetala. All measurements are log transformed and significance levels are 
indicated above each box to a 95% confidence interval (α = 0.05). The compact letter 
display is based on Tukey Posthoc tests derived from two-way ANOVAs.   

.   
  
  



 
  

      
  

49 

  
  

Figure 13. Comparison of vegetative morphological traits of 43 herbarium specimens 
represented by Rhodohypoxis baurii varieties with baurii = Rhodohypoxis baurii var. 
baurii, confecta = Rhodohypoxis baurii var. confecta and platy = Rhodohypoxis baurii 
var. platypetala. All measurements were log transformed and significance levels are 
indicated above each box to a 95% confidence interval (α = 0.05). The compact letter 
display is based on Tukey Posthoc tests derived from two-way ANOVAs.   

.   
  
  

 
Figure 14. Image showing the size difference of reproductive organs such as anthers 
and ovules between (A) Rhodohypoxis baurii var. platypetala, (B) Rhodohypoxis baurii 
var. baurii, (C) Rhodohypoxis baurii var. confecta.   

  

A   C   B   

5 mm   



 
  

      
  

50 

  
Figure 15. Variation in flower colour of R. baurii individuals housed in the greenhouse: A) 
R. baurii var. baurii from the Vrederus population, Eastern Cape Drakensberg;  B) R. baurii 
var. baurii from the Hebron Farm population, Southern Drakensberg, exhibiting dark pink 
tepals; C and D) R. baurii var. confecta flowers from the Sentinel population, Northern 
Drakensberg, showing various shades of white and pinks; E) R. baurii var. platypetala from 
Sunset Farm population (away from the main Drakensberg) showing white flowers with the 
occasional pink flushes on tepal tips;  F) exclusively white flowered R. baurii var. 
platypetala from Karkloof, which region (away from the main Drakensberg). Images taken 
in 2021.  

A   

F   E   

D   C   

B   



 
  

      
  

51 

Rhodohypoxis baurii varieties exhibit multiple ploidy levels (Table 2; Appendix, Table 4), 

and changes in ploidy may lead to morphological shifts, therefore, morphological traits were 

grouped by ploidy level to test if morphological variation or differentiation is due to ploidy 

level. Only three triploids were present in the dataset and were excluded due to the low 

sample size. Leaf length did not differ between diploids and tetraploids (df = 22, F = 0.72, P = 

0.41) nor did leaf width (df = 22, F = 0.82, P = 0.38) (Fig.17). However, there are two distinct 

groups present in the tetraploid leaf width measurements. The first group comprises R. baurii 

var. baurii individuals from the Eastern Cape Drakensberg, noted by Hilliard & Burtt (1978, 

p.57) as the “on the face of the main Berg at heights”. In this dataset, these individuals 

consisted exclusively of tetraploids (Table 2; Appendix, Table 4). The second group consists 

of R. baurii var. baurii “away from the main Drakensberg” individuals and R. baurii var. 

platypetala polyploid individuals. The “on the face of the main Berg” R. baurii var. baurii 

tetraploid individuals' leaves are 57.95% narrower (2.52 mm ± 0.24) than “away from the 

main Drakensberg” R. baurii var. baurii diploid individuals (5.99 mm ± 0.416). There is no 

significant difference in leaf widths between diploids a