The genus Ravenelia (Pucciniales) in South Africa

The genus Ravenelia represents the third largest genus of rust fungi and parasitizes a great number of leguminous shrubs and trees, mainly in the subtropics and tropics. Molecular phylogenetic analyses of this genus using nc 28S rDNA and CO3 sequences are presented with a special focus on South African representatives of Ravenelia. Many of the specimens had been collected by us in recent years, mainly from acacia species of the genera Vachellia and Senegalia. Morphological characters were extensively studied using light microscopy and scanning electron microscopy. The analyses resolved several well-supported phylogenetic groups. By linking these groups to their morphology and life cycle characteristics, it was possible to interpret the outcomes in terms of their evolutionary ecology and biogeography. Several characters previously used to define subgeneric groups within Ravenelia were found to be misleading because of assumed convergent evolution. However, host associations, the ability to induce aecial galls as well as the development of two-layered probasidial cells emerged as useful criteria for inferring monophyletic groups. Six novel Ravenelia species were discovered and described. Furthermore, five species represent new reports for South Africa, species descriptions were emended for two taxa, and a new host report emerged for R. inornata.


Introduction
In 1853, Berkeley introduced the genus Ravenelia within the rust fungi (Pucciniales).The genus initially comprised two species: R. glandulosa Berk.& M.A. Curtis, which was transferred from Sphaeria epiphylla Schwein.collected on Tephrosia virginiana (L.) Pers. in South Carolina and the newly described R. indica Berk.found on pods of an unidentified acacia.Later then Dietel correctly recombined the type species R. glandulosa to R. epiphylla (Schwein.)Dietel (Dietel 1894).In subsequent years, many additional Ravenelia species were found throughout the tropics and subtropics and today, some 200 species are described (Hernández andHennen 2002, Cummins andHiratsuka 2003).Ravenelia thus became the third most species-rich rust fungal genus after Puccinia and Uromyces.
All known species of Ravenelia are confined to a great diversity of Fabaceae residing in all three traditionally recognized subfamilies (Mimosoideae, Faboideae, Caesalpinioideae) (Hennen et al. 2005).The most prominent morphological features that are shared by all species of Ravenelia are the multicellular teliospores, which are borne on compound pedicels composed of two to several hyphae.These spores have an ellipsoidal, reniform, or almost hemispherical shape in side view and bear a variable number of pendent hygroscopic cysts.Other characters include spermogonia of type 5 and 7 (Cummins and Hiratsuka 2003) but these are commonly absent.
All species of Ravenelia are autoecious and their life cycles range from macro-and demi to hemi-, and more rarely to microcyclic (Cummins and Hiratsuka 2003).The aecial stage of several macrocyclic Ravenelia spp.can easily be recognized in the field by their ability to induce galls and witches' brooms in host tissues.Another special feature is the production of uredinoid aecia in numerous species of Ravenelia (Cummins 1959;Cummins and Hiratsuka 2003).
The morphological diversity and the variability of Ravenelia life cycles prompted mycologists early in the twentieth century to establish sections (Long 1903;Dietel 1906) or even to split this genus into several distinct genera (Sydow and Sydow 1915;Sydow 1921).The most sophisticated taxonomic system for Ravenelia was proposed by Sydow (1921) distinguishing eight genera based on teliospore traits in combination with observed life cycles.Details of the competing taxonomic systems are summarized in Table S1.However, a broad genus concept of Ravenelia comprising all these suggested genera or sections (within one genus) remains most widely accepted (compare Cummins and Hiratsuka 2003).
More than 500 rust species are known from South Africa (Crous et al. 2006), making this country relatively well explored for these fungi in comparison with other countries in Africa.Most contributions to the collection and description of rust fungi in South Africa are attributed to the investigations of Ethel M. Doidge during the first half of the twentieth century.In her last comprehensive species list of southern African rust fungi, she mentioned 24 Ravenelia species eight of which she described (Doidge 1927(Doidge , 1939(Doidge , 1950)).The most recent species list was published by van Reenen (1995) but nearly exclusively relied on literature data provided by Doidge.Due to changes in political borders, two species each now only occur in Mozambique (R. deformans and R. le-testui) and Zimbabwe (R. indigoferae and R. bottomleyae) respectively, while R. baumiana was recorded only from Angola.Two rusts, R. atrides and R. bottomleyae, were transferred to the genera Uredopeltis (Wood 2007) and Spumula (Thirumalachar 1946), respectively.Wood (2006) recorded R. ornata for the first time in South Africa and Ebinghaus et al. (2018) described R. xanthophloeae on the Vachellia xanthophloea.Thus, 19 Ravenelia species are currently known for South Africa.
During the course of recent surveys, aiming at recollecting the majority of Ravenelia species from South Africa and especially at investigating all known Acacia s.l. for potential rust infections, we have collected numerous specimens from acacias and fabaceous plants.The overarching aims were to re-evaluate the species diversity and systematics of Ravenelia rusts in South Africa by using microscopic investigations and molecular phylogenetic techniques.For a better understanding of the phylogeny of the genus as a whole also species mainly from the Neotropics were investigated.The emerging phylogenetic clades were interpreted using aspects of biogeographical distributions, life cycle traits, and host associations as well as morphological data.Furthermore, in order to illustrate conflicts when applying the taxonomic system for Ravenelia proposed by Sydow (1921), we mapped his suggested nomenclatural system to the phylogenetic reconstructions and discussed these outcomes.

Specimens examined
The specimens used for the molecular phylogenetic and morphological analyses were collected during several field surveys from 2004 to 2015 in South Africa.In addition, we considered 13 herbarium specimens from BPI originating from North and South America as well as DNA sequences downloaded from GenBank (see Table 1).Freshly collected material was immediately dried between paper sheets in a plant press and deposited after determination at the National Collections of Fungi (PREM) in Roodeplaat, South Africa, and the herbarium of the Natural History Museum in Karlsruhe (KR), Germany.In total, 91 specimens representing 44 Ravenelia species and three outgroup species were included in the molecular phylogenetic analyses and all of them were examined microscopically.For comparative purposes, additional 32 specimens comprising 15 type specimens deposited at PREM were examined only microscopically.All specimens investigated in this study are listed in Table 1.

Light-and electron microscopic investigations
The spores from plant material were scraped from the infected tissues using sterile insect needles and mounted in lactophenol solution on microscope slides.Light microscopic examinations were made using a Zeiss Axioplan light microscope (Carl Zeiss Microscopy, Jena, Germany) with a Color View pretoriensis (PREM1376, PREM60134), R. stictica PREM28255), R. tephrosiae (PREM1419, PREM10700), and R. transvaalensis (PREM27832) were examined at the facilities of the ARC-Plant Protection Institute (ARC-PPRI), Roodeplaat, South Africa, using a Leica Dialux 22 EB microscope and a ColorView III CCD color camera, and measurements for these specimens were made using analySIS LS software (LS Research Software GmbH, Germany).Scanning electron microscopy (SEM) was done using a ZEISS Sigma VP scanning electron microscope.For this purpose, infected leaflets from the herbarium specimens were mounted on double-sided adhesive carbon tape on metal stubs and coated with gold in a sputter coater BAL-TEC SCD OSO (Capovani Brothers Inc., USA).

DNA extraction and PCR
The isolation of spores and DNA extraction procedures were carried out using the INNUPrep Plant DNA Kit (Analytik Jena, Jena, Germany) as described by Ebinghaus et al. (2018).
For PCR of the nc 28S rDNA (LSU), the Taq-DNA-Polymerase Mix (PeqLab, Erlangen, Germany) and the GoTaq G2 HotStart DNA Polymerase Kit (Promega, Mannheim, Germany) were used, whereas only the GoTaq G2 HotStart DNA Polymerase Kit was used for PCR of CO3.To obtain sequences of the LSU, the primer pairs LR0R (Moncalvo et al. 1995) and LR6 (Vilgalys and Hester 1990) and 5.8SrustF/D1D2RustR (Ebinghaus et al. 2018) were used with the following conditions: 3 min at 96 °C followed by

Phylogenetic analyses
Following successful sequencing, the sequences were screened against the NCBI GenBank using the BLASTn algorithm (Altschul et al. 1990) to check for erroneously amplified contaminations and to exclude them from further processing.Forward and reverse strands were then individually assembled and manually edited using Sequencher 5.0 software (Gene Codes Corp., Ann Arbor, MI, USA).A total of 91 DNA sequences were used to construct the alignments of the LSU and 49 sequences for the CO3 sequence data using MAFFT v7.154b (Katoh and Standley 2014) (Lanave et al. 1984) with gamma distributed substitution rates (GTR+G).The analyses were run with a rapid bootstrap analysis using 1000 bootstrap replicates.The ML analyses were first conducted for each dataset separately and topological congruence was checked visually.Because no conflict of supported phylogenetic groupings was observed, a concatenated alignment was constructed for the LSU and CO3 sequence alignments and the subsequent phylogenetic analyses were inferred by applying the same methodology as for individual datasets.
Bayesian inference (BI) was performed with siMBa v.1.0implemented in MrBayes 3.2.5 (Larget and Simon 1999;Ronquist et al. 2012;Mishra and Thines 2014) applying the GTR+G substitution model.The Markov chain Monte Carlo search was run for five million generations with trees sampled every 500 generations.The burnin was set to 0.3.A Bayesian consensus tree was automatically calculated in siMBa and with posterior probabilities plotted on the tree.The phylogenetic trees of all different analyses were viewed and edited in FigTREE v1.4.0 (Rambaut 2009).
The taxonomic system proposed by Sydow (1921) for Ravenelia was applied.The respective generic names in addition to morphological and life cycle characteristics provided by literature were thus plotted on the phylogenetic reconstruction based on the LSU data.

Molecular phylogeny
The LSU sequence data resulted in an alignment comprising 91 sequences of 1016 characters in total length with 436 variable positions of which 372 were parsimony informative, whereas the CO3 alignment comprised 49 sequences with a total length of 605 characters of which 183 were variable and 144 parsimony informative.All alignments are deposited at TreeBase (TB2:S24974, TB2:S24975, TB2:S24976).
The phylogenetic reconstructions of the LSU and CO3 sequence datasets resolved similar tree topologies.Slight differences can be observed in the topologies of both data sets, but only in the placement of weakly or unsupported groupings, e.g., clades II and IV (Fig. 1, 2 and 3, Fig. S1).No significantly different tree topologies were observed in ML and BI approaches for either dataset.We recognized seven clades (i.e., I-VII) that included at least one South African Ravenelia species (Fig. 1).

Taxonomy
The results of the present study, which includes molecular phylogenetic analyses and morphological investigations, led us to propose six taxonomic novelties described in the following section.In addition, four Ravenelia species are newly reported from South Africa, the species descriptions are emended for three rusts and a novel host report is included for one species.
Ravenelia moloto W. Maier,M. Ebinghaus, MycoBank MB831070 Etymology: Name refers to Moloto, which is the common name of the host tree Senegalia erubescens in the local Setswana language.
Spermogonia and aecia not seen.Uredinia amphigenous but predominantly on the adaxial surface of the leaflets, sometimes on pods, sori on leaflets scattered or in small groups, shape ranging from circular to elongated, (60)120-250(460) μm in diameter, up to 6 mm in diameter when occurring on pods where sori form concentric eventually confluenting rings, subepidermal, erumpent; paraphyses peripherically arranged within uredinia, cylindrical or sometimes clavate, often septate, (27)40-55(77) × 6-13 μm, cell wall 1-1.7 μm, transparent to light brown; urediniospores ovoidal to ellipsoidal, 12-16 × 21-28 μm, spore wall evenly 1.4-2.2μm thick, echinulate, aculei approximately 1 μm in height, germ pores 5-6, equatorially arranged.Telia replacing uredinia, chestnut brown to dark brown; teliospores cinnamon brown to chestnut brown, circular to subcircular from above, (63)75-95(103) μm in diameter, upper side of the teliospores slightly convex to flattened, 5-7 probasidial cell across, probasidial cells (20)24-28(34) × (12)15-20(26) μm, cell wall thickened at the top side of the spore and here seemingly bilaminate with an inconspicuously thin, or sometimes distinctly marked hyaline to pale brown outer layer and a chestnut brown inner layer, the inner layer (1)3-5(7) μm thick, each cell with 7-13 verrucose ornamentations, 1-3 μm in height; cysts pendent, globose, hyaline and smooth, in the same number as the probasidial cells, swelling in water but only slightly in lactophenol solution; pedicel multihyphal.Notes: Ravenelia moloto was found to be closely related to R. modjadji on S. polyacantha subsp.campylacantha and to R. doidgeae found on S. polyacantha subsp.polyacantha but were clearly supported as distinct phylogenetic groups based on LSU and CO3 sequence data.The close relationship between these species is reflected by their morphology that makes it difficult to distinguish between them.However, the teliospores of R. moloto tend to be smaller in diameter compared with those of R. doidgeae and its ornamentation often appears more pronounced.These two species show additional minor differences in urediniospore morphology as they often tend to be more globose in R. doidgeae.This rust has been found only on S. erubescens, a tree occurring in the dry savannah in the northeastern part of South Africa.
Additional specimens examined: South Africa, Gauteng, Ditholo Nature Reserve, on leaves of S. mellifera subsp.Notes: Ravenelia spinifera is one of three species including R. transvaalensis and R. acaciae-melliferae occurring on Senegalia mellifera.While R. transvaalensis is also known from South Africa, Ravenelia acaciae-melliferae has been reported only from Eritrea and Ethiopia (Farr and Rossman 2017).Ravenelia spinifera can easily be distinguished from R. transvaalensis by its teliospores that have well-developed spines while those of R. transvaalensis are smooth-walled.Remarkably, we found host individuals with infections caused by both R. transvaalensis and R. spinifera even on a single leaflet.The original description of R. acaciae-melliferae is very limited and it does not provide details of teliospore ornamentation (Baccarini 1917).We thus consider the teliospores of that species as smooth and it remains uncertain whether R. acacia-melliferae species should be reduced to synonymy with R. transvaalensis.Ravenelia spinifera further resembles R. acaciae-nigrescentis on S. nigrescens in overall morphology.Nonetheless, both species were resolved in two wellsupported monophyla in phylogenetic reconstructions based on LSU and CO3 sequence data (Fig. 1 and 3).Additionally, R. spinifera appears to be restricted to S. mellifera subsp.detinens and the two species can thus also be distinguished by their host association.Spermogonia and aecia not seen.Uredinia amphigenous on leaflets, in small groups, subcircular to elongated, sometimes forming concentric rings, light brown, often surrounded by chlorotic areas, 0.1-0.6 mm in diameter, subepidermal, erumpent, peripheral paraphyses cylindrical and elongated to sometimes clavate, with a basal septum, 32-53 × 9-12 μm, cell wall thin and transparent 0.8-1.1(1.7) μm; urediniospores ovoidal to ellipsoidal, sometimes elongated, light brown (23)25-29(33) ×  12-15 μm, spore wall laterally (1)1.3-1.6(2)μm thick, basally slightly thickened and apically often more pronounced thickened, echinulate, aculei approximately 1 μm in height, germ pores 4-5, equatorially arranged.Telia replacing the uredinia, orange brown to chestnut brown; teliospores orange brown to cinnamon brown, circular to subcircular from above, upper side of teliospores convex, (55)85-100(117) μm in diameter with 3-7 probasidial cells across, single probasidial cells (27)31-33(36) × (12)14-18(25) μm, cell wall thickened at the top side of the spore and here distinctly bilaminate with a hyaline brown outer layer and a cinnamon brown inner layer, 3.5-6.5 μm thick, the peripheral cells each with 5-9 small verrucae, 1-1.5(2.5)μm, but central cells often smooth; cysts pendent, globose, hyaline and smooth, in the same number as the probasidial cells, swelling in water but only slightly in lactophenol solution; pedicel multihyphal.
Notes: This rust was found only once on a single tree in Nelspruit, Mpumalanga, in a private garden.The tree was most probably planted as an ornamental as this region is outside its natural distribution range that lies in the eastern part of the North-West Province, in western Limpopo and the northern parts of Gauteng and Mpumalanga (Coates Palgrave 2005;Smit 2008).The teliospores of R. molopa can easily be confused with those of R. pienaarii that infects the widely distributed S. caffra.However, the two species can be distinguished by the peripherally arranged paraphyses in the uredinia of R. molopa (Fig. 6d), while R. pienaarii is aparaphysate.These two rusts are also clearly separated by a significant genetic distance in molecular phylogenetic analyses of LSU and CO3 gene regions; however, the exact phylogenetic position of R. molopa within clade I could not be fully resolved (Figs. 1 and 3).
Notes: Ravenelia doidgeae was found only once and the teliospores are morphologically difficult to discriminate from those of R. modjadji that can frequently be found on S. senegalia subsp.campylacantha.But the urediniospores of this rust tend to be more distinctly ovoidal than those of R. modjadji, which are more ellipsoid.Ravenelia doidgeae further resembles R. moloto in overall morphology but that rust has been found only on S. erubescens.Despite their morphological resemblance, the phylogenetic analyses of the LSU region resolved R. doidgeae as a distinct lineage in a close sister relationship to R. modjadji and R. moloto (Fig. 1).
Ravenelia modjadji M. Ebinghaus, W. Maier, MycoBank MB831074 Etymology: Name refers to the Rain Queen Modjadji of the Balobedu people that live in the region where the holotype specimen was collected.
Notes: Ravenelia modjadji is the only species within a monophyletic lineage of seemingly hemicyclic Senegalia rusts (clade I), which produces primary and secondary uredinia, and the primary uredinia causing malformations in its host.Similar to the gall-forming R. evansii and R. macowaniana, this spore stage is spatially separate from the (secondary) uredinia and telia.Old malformed branches often become detached from the trees during heavy rainfall or strong wind and can then be found in large abundance below the trees.R. modjadji can thus be easily distinguished from its close relatives, R. doidgeae and R. moloto if the aecia are present.It can also be differentiated based on the morphology of the urediniospores, which are ellipsoid in this Ravenelia while the latter two species have more ovoidal urediniospores.Ravenelia modjadji however shares a similar teliospore morphology with its phylogenetic close relatives R. doidgeae and R. moloto.
Ravenelia dumeti M. Ebinghaus, W. Maier, & Begerow sp.nov.(Fig. 9a-h) MycoBank MB831075 Etymology: The name is derived from the Latin word dumetum that describes a plant thicket and indicates the occurrence of the rust fungus in those habitats: the host Senegalia brevispica forms dense and often impenetrable thickets at forest margins and along rivers in South Africa.
Notes: In 1946, this species was first described for India on Acacia arabicae Willd.(= Vachellia nilotica P.J.H. Hurter & Mabb.) by Mundkur and Thirumalachar, and was recently more precisely reported on V. nilotica subsp.indica (Shivas et al. 2013).We report this rust for the first time on V. nilotica subsp.kraussiana that is widespread in southern Africa (Coates Palgrave 2005).Ravenelia acacia-arabicae is similar to R. tandoni that was described on Senegalia catechu (Sydow et al. 1937) but both species can be distinguished by predominantly hypophyllous uredinia in R. acacia-acaciae in contrast to the epiphyllous uredinia in R. tandonii.Furthermore, with 1-1.5 μm, the urediniospore wall of this rust is approximately half as thick as those of R. acaciae-arabicae, which is (2)2.5-3μm thick.The urediniospores of the two rusts also differ in the number and arrangement of germ pores: 8-12 scattered germ pores in R. acaciae-arabicae compared with two rows of four germ pores in R. tandonii.The teliospores of Ravenelia acaciae-arabicae can be distinguished by having 6-9 blunt aculei per probasidial cell in contrast to 3-6 verrucose papillae in R. tandonii.
Phylogenetically, R. acaciae-arabicae is most closely related to R. evansii with which it shares major teliospore character traits such as size and its spinescent ornamentation.However, both species appear not to have a shared host range and can be thus easily distinguished based on their Vachellia hosts.Furthermore, R. evansii frequently causes aecial galls and malformations in infected host tissues while all collections of R. acaciae-arabicae in South Africa lack the aecial stage.Spermogonia and aecia not seen.Uredinia amphigenous on leaflets, circular to elongated, in small groups, brown, single sori minute, (120)180-300(400) μm, subepidermal, erumpent; urediniospores ovoidal, often tapered towards the b a s i s , o c h r a c e o u s b r o w n , e c h i n u l a t e , (24)28-33(38) × (13)15-19(21) μm, spore wall thin but slightly thickened towards the apex and at the basis, 1-2(2.5)μm, germ pores 5-7, equatorially arranged; paraphyses few, peripheral, capitate, about 45 μm in length, capitulum approximately 14 μm wide, light ochraceous brown; telia replacing the uredinia and therefore of same size and shape as the uredinia, dark brown to almost black; teliospores circular to subcircular in diameter, teliospores convex from above but with a concave bottomside, chestnut brown to light brown, (57)84-98(115) μm in diameter with 4-8 probasidial cells across, single probasidial cells (15)27-32(35) × (13)16-19(28) μm, cell wall thickened at the top side of the spore and here bilaminate with a hyaline outer layer and a chestnut brown inner layer, the inner layer (2)3-6(8) μm, very rarely bearing single verrucose ornamentations of 1-2 μm in height; cysts pendent, globose, hyaline, smooth, swelling in water but only slightly in lactophenol solution, number of cysts equal to the number of the probasidial cells; pedicel multihyphal.
Notes: Ravenelia acaciicola was described by Sanwal (1951) in India on Acacia senegal Willd.(now Senegalia senegal (L.) Britton) without providing information regarding which of the four known subspecies of the tree was infected.Two varieties of S. senegal (var.leiorhachis and var.rostrata) occur in South Africa (Coates Palgrave 2005), both of which were found to be infected by R. acaciicola.Interestingly, we found some intraspecific genetic variability between specimens originating from the two different host varieties.Because these differences consisted of only one substitution each in the studied LSU and CO3 genes, we refrain from further splitting of R. acaciicola at this stage.However, it will be interesting to study additional specimens and gene regions for deeper insight.
The specimens examined in this study match the type descriptions given by Sanwal (1951) with respect to the size of the urediniospores as well as in the morphology of the teliospores.However, they differ in the number of germ pores in the urediniospores.Sanwal (1951) described four germ pores in an equatorial position, while 5-7 equatorial germ pores were observed in our collections.We also observed light brown and clavate paraphyses in a single specimen (PREM61847) of S. senegal var.leiorhachis while Sanwal did not describe these structures.Considering the occurrence of this Ravenelia species on different subspecies of S. senegal, the observed morphological differences from type species could reflect phenotypic variability.Alternatively, cryptic species occur on the subspecies of the host tree.The mostly smooth teliospores of Ravenelia acaciicola resemble the closely related R. transvaalensis that infects S. mellifera in South Africa.But both rusts appeared to be host specific to S. senegal and S. mellifera subsp.detinens, respectively and were furthermore distinguished by phylogenetic analyses of the LSU and CO3 gene regions.
Notes: The type specimen was collected in Namibia on Acacia nigrescens Oliv.(Ritschel et al. 2007) but its area of distribution is likely larger because the host tree is common throughout western and southern Africa (Coates Palgrave 2005).The original description lacks comments on the presence of paraphyses in R. acacia-nigrescentis.However, in the present study, a small number of peripherally arranged paraphyses were observed in light and scanning electron micrographs of the uredinia (Fig. 12g).This Ravenelia species resembles R. spinifera morphologically.The two species formed a highly supported monophyletic group representing two closely related but distinct sister species in our phylogenetic analyses (Figs. 1 and 3).Ravenelia acaciae-nigrescentis appeared host specific on S. nigrescens while R. spinifera was only found on S. mellifera.Spermogonia and aecia not seen.Uredinia amphigenous on leaflets, roundish or of irregular shape, loosely in groups or singly, sometimes forming confluently concentric rings and then leading to chlorotic spots, very variable in size, (100)200-400(500) μm, subcuticular, often surrounded by the torn remnants of the cuticle, light ochraceous brown; paraphyses numerous, intrasoral, size of paraphyses increasing from the center to the sorus margin, the central paraphyses very slender and cylindrical, about 25-35 μm in length and 3 μm wide, transparent; paraphyses towards the sorus margin clavate and up to 40 μm in length and 5-10 μm in width; urediniospores globose to rarely ovoidal, light ochraceous brown, echinulate, 14-18 μm, spore wall evenly 1.5-2.5 μm, germpores 9-12, scattered; telia replacing the uredinia and therefore of same size and shape as the uredinia but also on rhachis, dark brown to blackish, teliospores circular in diameter and hemispherical in lateral view, chestnut brown to dark brown, (66)70-85(102) μm in diameter with 5-7 probasidial cells across, single probasidial cells (24)28-34(41) × (12)15-19(25) μm, cell wall thickened at the top side of the spore and here with an inconspicuously thin outer hyaline to pale brown layer, (4)5-7(9) μm, cell wall laterally (1.5)2-3(3.5)μm thick, probasidial cells rarely bear a single short and hyaline papillum up to 3.5 μm in length; cysts pendent, globose, hyaline, smooth, easily detached, swelling in water but only slightly in lactophenol solution, number of cysts equal to number of probasidial cells.Pedicel sometimes light brown, up to 180 μm in length, multihyphal.

Emended species descriptions
In 1948, Doidge added a description of the aecial and uredinial stage of R. modesta based on a specimen (PREM34572) collected from V. luederitzii (Doidge 1948) and which is also another common host of R. evansii (Ebinghaus et al. 2018).The urediniospores of this additional specimen of R. modesta were identical to the urediniospores of the specimens of R. modesta examined in our present study and these were clearly distinct from those of R. evansii.Based on our findings, we consequently consider the uredinial stage of this rust (PREM34572) as representing R. modesta Doidge.
However, it remains doubtful as to whether the aecial stage described by Doidge (1948) for this specimen is conspecific with the uredinia of R. modesta or whether the aecia actually represent R. evansii.We were unable to isolate DNA of sufficient quality from this specimen (PREM34572) and are thus unable to unequivocally confirm the assumed conspecificity of both spore stages with R. modesta.Nonetheless, because both rusts are able to infect the same host species, we propose (i) that the original description Doidge provided in 1939 for the telial stage of R. modesta is not correct but describes the teliospores of R. evansii and that V. hebeclada does not represent a host for R. modesta.(ii) The "connection" of the aecial stage with R. modesta given in Doidge (1948)   stage and with PREM60795 on V. luederitzii var.retinens for the uredinial and telial stages.On the basis of our morphological as well as phylogenetic analyses we furthermore report Vachellia gerrardii and V. rehmanniana as new hosts for R. modesta and define the respective specimens as paratypes.
Notes: Sydow and Sydow (1912) described the uredinia as subepidermal but our microscopic examinations did not support this finding and we rather observed subcuticular sori.We further observed 10-12 scattered or bizonate germ pores in the urediniospores, while Sydow and Sydow (1912) noted the presence of 6-8 scattered germ pores.Arguably, the most important emendation concerns the surface structure of these spores that show a distinct "hub and spoke pattern" (Fig. 14b  and e; see for definition also Gardner and Hodges 1985).
Ravenelia pretoriensis is morphologically similar to R. modesta and both species appeared to be closely related in our phylogenetic analyses based on fragments of the LSU and CO3 gene regions (Figs. 1 and 3).However, R. pretoriensis appeared to be confined to Vachellia karroo and to the closely related V. natalitia, while R. modesta was never observed on these host trees.Vachellia natalitia is here reported as a new host species for this rust fungus.
Additional specimens examined: South Africa, Mpumalanga, South of Barberton, on leaves of V. Notes: Based on collections deposited at the National Mycology Collections (Roodeplaat, South Africa) and our observations, this species appears to have a distribution range in South Africa that is restricted to the southern and southeastern coastal regions and up to the low lying regions of southern Mpumalanga Province.Interestingly, R. inornata appeared to be absent in the central plateau region of South Africa despite the common occurrence of its host trees (Coates Palgrave 2005).This region is climatically distinct in having lower levels of precipitation and a pronounced dry season.The absence of R. inornata in this region might be linked to these environmental factors.
This species shares the hosts V. natalitia and V. karroo with R. pretoriensis and R. macowaniana.Mixed infections by these two species were thus occasionally encountered even on a single leaflet.Co-occurrence of R. inornata with R. pretoriensis may lead to confusion because the teliospores of both rusts bear verrucose ornamentations.However, the teliospores differ distinctly in size with teliospores of R. pretoriensis mostly measuring 70-95 μm, while the teliospores of R. inornata range between 104 and 162 μm (Table S2).They can furthermore be distinguished by their life cycles: Ravenelia pretoriensis might be hemicyclic as only the uredinial and the telial stages are known while R. inornata appears demicyclic.All three species are only distantly related in molecular phylogenetic analyses (Figs. 1 and 3).

Discussion
In the present study, we analyzed numerous Ravenelia specimens with a strong focus on South African species, but included also species from the Neotropics as well as published gene sequences.The South African representatives clustered in seven phylogenetic lineages, one of which was represented by a single species only (clade III).Two of these phylogenetic groups consist only of Paleotropic species while five of these groups consist of Neotropic and Paleotropic species.Two lineages were found that are of exclusively Neotropical origin and were therefore not highlighted in the phylogenetic trees.
In the following sections, the findings of the molecular phylogenetic analyses will be discussed with emphasis on morphological and life cycle traits as well as a consideration of the evolutionary ecological implications relating to host specialization.Finally, the diversity and taxonomy of Ravenelia in South Africa is re-evaluated.

Morphology and life cycle traits linked to phylogenetic lineages
Clade I The species in this clade comprised only South African representatives and are exlusively parasitizing species of the genus Senegalia (Mimosoideae).They were often difficult to discriminate from each other based on urediniospore or teliospore morphology.Nevertheless, the species could be distinguished by molecular phylogenetic analyses and, in addition, showed distinct host preferences.Rusts that shared teliospore traits like a specific ornamentation type tended to represent distinct lineages that mirror their close phylogenetic relationships, e.g., R. acacia-nigrescentis and R. spinifera (spinescent teliospores), R. moloto, R. doidgeae and R. modjadji (verrucose teliospores) or R. transvaalensis and R. acaciicola (smooth teliospores).For species in this clade, only uredinia and telia are known and therefore, they may be hemicyclic.The only exception is here R. modjadji.This species produces gall-inducing primary uredinia as well as secondary uredinia.
Clade II This clade was comprised of both Paleotropical and Neotropical species.The South African R. modesta and R. pretoriensis share a unique urediniospore surface structure with the Neotropical species in this clade that was described as "hub and spoke pattern" (Gardner andHodges 1985, Hernández andHennen 2002, see Fig. 14b and e) and appears to be synapomorphic for this lineage.This urediniospore ornamentation had previously only been reported for a few Ravenelia species from the Neotropics.The African representatives in this clade were only found on Vachellia species, while the South American species had been collected from Vachellia and Mimosa.Further rust species with this character have been reported from other host genera within the Mimosoideae but mostly on Albizia including R. albiziaezygiae, R. albiziicola, R. clemensae among others.Field observations suggested that the uredinial stage of R. pretoriensis and R. modesta is consistently the predominant spore stage, with the teliospores observed more rarely.This finding was irrespective of the time point of species sampling and may represent a lineage-specific life cycle characteristic.
Clade III This clade consisted of a single species, R. dumeti.
Only the uredinial stage is known of this species and specific traits are thus limited to characterize this rust.The numerous intrasoral and distinctly capitate paraphyses distinguish it from all other Ravenelia species investigated in the present study.The extended branch length in the phylogenetic trees that separates this lineage from its sister group of clade II mirrors the distinctiveness of these uredinial traits.Ravenelia dumeti is the only Ravenelia species of southern African origin that parasitizes a species of a lineage of Senegalia with African-American distribution, while all Senegalia rusts of clade I are parasitic on a hosts lineage of an African-Asian distribution (Bouchenak-Khelladi et al. 2010;Kyalangalilwa et al. 2013).In this respect, R. dumeti is more similar to the Ravenelia species studied from South America (Fig. 2), which also parasitize on hosts belonging to this African-American lineage of Senegalia.
Clade IV The members of this clade are heterogeneous in terms of morphology and associated hosts as they parasitize representatives of the two traditionally recognized subfamilies Caesalpinioideae and Mimosoideae.Here, R. halsei (on Senegalia, Mimosoideae) is the only species with aparaphysate uredinia, while R. elephantorhizae (on Elephantorrhiza, Mimosoideae), R. macrocarpa, and R. mesilliana (both on Senna, Caesalpinioideae) have intrasoral paraphyses.In the latter two species, the paraphyses are variable in size and shape (Baxter 1965, Hernández andHennen 2002;Fig. 11).Senna spp.are hosts to approximately 14 species of Ravenelia and are thus an important host group for these rusts (www.indexfungorum.org).Future studies should clarify whether size variation of the paraphyses is a common morphological character that is synapomorphic for Ravenelia species parasitizing Senna.
Clade V Ravenelia holwayi and R. dichrostachydis share the host preference for Mimosoideae but show several striking differences and a significant genetic distance reflecting their distant geographic origin from North America and South Africa, respectively.Most prominently, the species differ in the ability of R. holwayi to induce uredinial galls while R. dichrostachydis does not share this trait.To date, only R. dichrostachydis is known from Dichrostachys while six Ravenelia species have been described on Prosopis, e.g., R. arizonica, R. chacoensis, R. prosopidis, or R. spicigerae (www.indexfungorum.org).The incorporation of these species in future analyses could help to characterize members of this clade in more detail.
Clade VI This is the only lineage comprising Ravenelia species infecting members of the Faboideae suggesting that this host association evolved only once.This is in contrast to the Ravenelia spp.parasitizing on Mimosoideae, which are present in several lineages of the phylogenetic tree and thus of polyphyletic origin.Within this clade, only uredinial and telial stages are known for R. stictica, R. glabra, R. aff.indigoferae, and R. tephrosiae suggesting a hemicyclic life cycle.They are furthermore characterized by two-layered probasidial cells.Since the type species of Ravenelia, R. epiphylla, was collected on a related host of the Faboideae and also shares these characters, we assume that it would reside within this phylogenetic group.Unfortunately, fresh material of R. epiphylla was not available for inclusion in the molecular phylogenetic studies and its relationship to other species could thus not be resolved.
In our analyses, Ravenelia ornata was the sister species to R. platensis.Both species develop uredinioid aecia (Hernández and Hennen 2002;Wood 2006), a trait that is shared with several distinct lineages of Ravenelia rusts, i.e., the neotropical R. echinata var.ectypa, R. havanensis and R. hermosa but also R. holwayi in clade V (Fig. 2).However, R. ornata apparently lacks the ability to induce galls.The two accessions of this rust collected in South Africa showed considerable sequence variation of 3.8% (6.6% when gaps are included) within a fragment of the LSU gene region, even though they are morphologically indistinct based on teliospore morphology (data not shown).Interestingly, collections indicate a wide distribution within the tropics with this rust being reported from the Caribbean, Asia, and Africa (Farr and Rossman 2017).These findings suggest that a broader sampling including specimens from its full geographical range could reveal a species complex including cryptic species.
Clade VII The rusts in this phylogenetic group are confined to Mimosoideae, mainly Vachellia species with the exception of two neotropically distributed rusts, i.e., R. mainsiana and R. cebil infecting Mimosa and Anadenanthera, respectively.Several species in this clade possess two-layered probasidial cells, a feature shared with Ravenelia spp. on Faboideae (clade VI), and most species are able to induce galls (Fig. 2).No spermogonia were reported for those species within the group (R. acacia-arabicae, R. mainsiana, R. cebil, R. inornata) lacking the ability of gall induction.This is in line with observations of Larous and Lösel (1993) that invasion of the vascular system and resulting hypertrophies are only induced by the monokaryotic hyphae of the rust fungi.Remarkable differences in life cycle traits were seen between Ravenelia hieronymi and the closely related R. evansii, R. macowaniana, and R. xanthophloeae by producing telia subsequently on aecial galls in the case of R. hieronymi (Hernández andHennen 2003, Hennen et al. 2005).Hennen et al. (2005) suggested that R. hieronymi could represent a demicyclic rust as pedicellate urediniospores were never observed.This is in contrast to the macrocyclic R. evansii, R. macowaniana, and R. xanthophloeae that induce aecial galls but in which the uredial and telial stages are always spatially separated from the aecia (Doidge 1939;Ebinghaus et al. 2018).While the aeciospores of these rusts obviously re-infect the host, the aeciospores of R. hieronymi were suggested to be non-functional (Hernández and Hennen 2003;Hennen et al. 2005).Similar to R. hieronymi, only aecia and telia are known for R. inornata and they develop intermingled with the aecia, but galls are unknown in this species.However, the two species did not appear to be closely related in our phylogenetic analyses.In the acacia rust Atelocauda koae simultaneously occurring aecio-and teliospores are known to be non-functional and an ongoing transition from demicyclic to microcyclic life cycle has been suggested (Hodges Jr and Gardner 1984, Chen et al. 1996).It remains an open question as to whether this also holds true for R. inornata.Dietel (1906) did not observe germpores in this rust and this could support the view of non-functional aeciospores.
Linking morphological traits to phylogenetic lineages within Ravenelia showed that taxonomic systems as proposed by several authors including Long (1903), Dietel (1906), Sydow andSydow (1915), andSydow (1921) were often not congruent with the phylogenetic reconstructions based on LSU and CO3 gene regions (Fig. 2, Table S1).Rust fungal structures including the morphology of uredinio-and teliospores have been suggested to be highly adaptive and are consequently prone to convergent evolution (Savile 1971(Savile , 1976(Savile , 1978)).Additionally, our finding of close phylogenetic relationship between Ravenelia species that show different life cycle traits highlights the variability of those character states in the rust fungi as has been shown many times before (compare Maier et al. 2007).Nevertheless, several traits such as the host association but also the ability to induce galls in combination with the development of two-layered probasidial cells were found to represent useful criteria to draw conclusions regarding monophyly.In this respect, we found evidence that the macrocyclic and often gall-inducing rusts comprising those on Vachellia but also other members of the Mimosoideae most likely represent a more ancestral lineage in Ravenelia (Shattock and Preece 2000).Likewise, the "hub and spoke" urediniospore ornamentation pattern appeared synapomorphic for a distinct lineage of Ravenelia species.

Evolutionary ecological implications
Species of Senegalia and Vachellia represent the majority of Ravenelia hosts in South Africa.
Two major lineages of Senegalia rusts had either an exclusively Paleotropic (clade I) or Neotropic (i.e., R. cohniana, R. piepenbringiae, and R. hernandezii; compare Ebinghaus and Begerow 2018) distribution.Molecular dating analyses suggest a split of Senegalia species into an Asian-African lineage from American ancestors in the late Miocene, some 18-9 Mya (Bouchenak-Khelladi et al. 2010).The Paleotropic lineage is believed to have subsequently diversified in response to oscillating climatic changes during the Pliocene and species likely spread from Africa into Asia (Bouchenak-Khelladi et al. 2010).The Paleotropically distributed rusts in clade I correspond to this host lineage and reasonably explains the Asian-African distribution of R. acaciicola (Fig. 2).The ancestor of this rust lineage could have jumped from a non-Senegalia host onto an early representative of the diverging African host lineage and diversified simultaneously with the diverging hosts by cospeciation or subsequently via host tracking coevolution (Ehrlich and Raven 1964;Janz and Nylin 1998).In the latter case, the parasite would have overcome host defense of an ancestral host and, following its radiation, colonized and diversified on these related species of the same clade.These two possible scenarios could be assumed for R. modjadji, R. moloto and R. doidgeae that infect the sister species S. polyacantha subsp.campylacantha, S. polyacantha subsp.polyacantha, and S. erubesens, respectively (Bouchenak-Khelladi et al. 2010;Kyalangalilwa et al. 2013).Other members of this rust lineage might rather have evolved by host shifts (Roy 2001), e.g., R. transvaalensis or R. spinifera that both infect S. mellifera subsp.detinens.
The results of the present study further suggest that the Ravenelia species occurring on Senegalia represented in clade I are highly host specific, with most species restricted to a single host.The hosts of the rusts in this clade predominantly occur in the semi-arid savannas of South Africa where they occupy specific niches and form dominant floral elements (Coates Palgrave 2005, Smit 2008).This native semi-arid habitat could also have contributed to a predominance of the more heat-and drought-resistant telial stage compared with the uredinial stage (Savile 1976).
The limited numbers of wind-dispersed urediniospores as well as the obvious lack of spermogonia and the aecial stage in these Ravenelia spp.may imply that rust fungal propagules are restricted from easy movement to a broader range of potential hosts due to dispersal limits.This may have hampered them from acquiring multiple hosts.
Several rust fungi that lack spermogonia were found to reproduce sexually through self-fertility (Anikster et al. 1980;Anikster and Wahl 1985;Ono 2002).If this is also true for the Senegalia rusts, the dispersal limits and the resulting restrictions in gene flow between populations may reasonably explain the observed host ranges but could also have led to an increased speciation rate.On the other hand, apomictic or asexual reproduction was proven to occur in various microcyclic species (Ono 2002) and which potentially can lead to similar speciation patterns.However, examinations of nuclear behavior as well as population genetic approaches are needed to prove either mechanism.
Species in the genus Vachellia (formerly Acacia subsp.Acacia) represent the other major host group for the South African Ravenelia species.Our study clearly showed that the rusts infecting this host genus evolved at least twice independently.The gall-inducing macrocyclic rusts R. macowaniana and R. evansii are among the most abundant rust fungi in southern Africa.This is also consistent with their wide host range and common occurrence of their hosts (Smit 2008).In contrast to the observations in the south African gall rusts, populations of macrocyclic rust fungi were reported to often lose their aecial stage in more arid environments.Savile interpreted this as an adaptation to low water availability and high UV radiation, because aeciospores often possess little pigmentation and thin spore walls (Savile 1971, but see Zwetko and Pfeifhofer 1991).However, we can only speculate under which conditions these rusts have evolved the capacities to induce aecial galls and whether or not ontogenic constraints may restrict life cycle reductions.But considering the scattered distributions of their hosts in open savannas and in the Karoo semi-desert, the massive production and release of wind-borne aeciospores may be still advantageous by effectively bridging the distance between potential host individuals.The production of spermatia and large numbers of aeciospores could also contribute to the wide host range observed in R. macowaniana and R. evansii, with greater numbers of potential hosts being exposed to these spores.The South African and gall-inducing Vachellia rusts could thus have evolved a different strategy compared with Ravenelia species on Senegalia to persist and disperse in a similar environment.
The Vachellia rusts exhibiting urediniospores with "hub and spoke" ornamentation (clade II) most likely originated in the Neotropics.This emerges from the fact that several Ravenelia species displaying this morphological characteristic infect a suite of different mimosoid genera with a Neotropical distribution.We thus assume that an ancestor of R. modesta and R. pretoriensis was introduced to Africa by long-distance dispersal.Spore dispersal over long distances may readily occur and was shown for other rust fungi like Melampsora species (e.g., Barrès et al. 2008).The origin of the African lineage could have coincided with the trans-Atlantic dispersal events that were argued to have occurred in the Vachellia species in the Miocene (16-11 mya;Bouchenak-Khelladi et al. 2010).Ravenelia modesta and R. pretoriensis are difficult to distinguish from each other based on spore morphology or the genetic markers utilized in this study and speciation most likely occurred more recently.
These findings and the fact that Ravenelia species parasitizing the major host genera Vachellia and Senegalia appear in several phylogenetic lineages suggest that diversification of Ravenelia in the current circumscription might have been driven by a combination of host shifts and co-evolutionary host tracking or cospeciation in addition to more rare jumps to co-occurring but more distantly related hosts.This situation might be comparable with patterns observed within Puccinia and Uromyces where a similar situation was observed for the species parasitizing Cyperaceae and Poaceae in relation to all other parasitized groups (Maier et al. 2007;van der Merwe et al. 2007van der Merwe et al. , 2008)).

Diversity of Ravenelia in South Africa
A targeted sampling effort in combination with morphological investigations and molecular phylogenetic analyses revealed that the diversity of Ravenelia species in South Africa greatly exceeds the current knowledge regarding the diversity of this genus in the region.Six Ravenelia species were revealed as new to science and five Ravenelia species were new reports for South Africa.In another recent study, one new Ravenelia species and seven new acacia hosts were reported from the same region (Ebinghaus et al. 2018).Thus, to date, 32 species of Ravenelia are known for South Africa.With a total of 25 species, the majority of Ravenelia spp.are confined to the traditionally recognized host subfamily Mimosoideae (in the classic sense).Of these, 13 species occur on 13 Senegalia hosts, 9 on 15 Vachellia hosts, and 3 on host trees of various other mimosoid genera.Five species occur on Faboideae hosts and a single rust species (R. mesilliana) is known on Senna (Caesalpinioideae) (Table 2).A single species (R. woodii) was collected from an undetermined leguminous host plant (Doidge 1939).Unfortunately, this monotypic rust is not represented in PREM and could not be investigated.
A special effort was made to investigate all of the approximately 42 species of acacias (18 taxa of Senegalia and 24 taxa of Vachellia) in South Africa (Coates Palgrave 2005;Smit 2008) and scrutinize them for infections.With 22 described Ravenelia spp. on at least 28 acacias the genera Senegalia (13 host taxa; 72% of total species) and Vachellia (15 host taxa; 62.5% of total species) represent the major hosts of this genus in South Africa (Table 2).Based on the number of herbarium specimens present at PREM and our own observations, some of these rusts, e.g., R. macowaniana and R. evansii likely represent the most abundant and ecologically significant rust fungal species on acacias in this region.
A total of 161 Vachellia species and 203 species of Senegalia have been described globally (Maslin et al. 2003).Of these, only approximately 26 and 37 species, respectively, were reported as hosts of Ravenelia spp.(www.indexfungorum.org).Considering the high percentage of host taxa we found within the genera Senegalia and Vachellia in South Africa, it is likely that future studies on Ravenelia will significantly exceed species numbers currently known for this genus.

Conclusion
In this study, we have revised and illustrated the diversity of Ravenelia spp. in South Africa.The first molecular phylogenetic analysis is also presented for this genus.Based on phylogenetic reconstructions, it was possible to consider aspects of the presumed evolutionary strategies in Ravenelia spp. that reflect lineage-specific host association patterns and the biogeography of their hosts.It appears worthwhile for future research in these rust fungi to focus on a global species sampling including rusts especially collected from the Caesalpinioideae and Faboideae as well as to incorporate additional members of the family Raveneliaceae.These are likely to shed light on evolutionary pathways aiding to a more comprehensive understanding of the evolution of character states in the rust fungi.

Fig. 1
Fig. 1 Molecular phylogenetic reconstruction of the genus Ravenelia inferred from LSU sequences using BI.Posterior probabilities above 0.90 and MLbootstrap support above 75 are shown.Highlighted in bold are those species that were described as novel taxa in this study

Fig. 2
Fig. 2 Cladogram based on a phylogenetic reconstruction using BI showing character states linked to species.Terminal branches were collapsed

Fig. 3
Fig. 3 Phylogenetic reconstruction based on a combined dataset of CO3 and LSU sequence data.ML bootstrap values above 75 and p values above 0.95 are shown.Species, described as novel taxa in this study, are highlighted in bold

Table 1
List of specimens used in this study