Dried herbarium specimens from the following public herbaria were used: B, FH, G, GLM, GZU, HBG, KR, M, PUR, S and W (acronyms according to Index Herbariorum, Holmgren et al. 1990).

Urediniospores and cross sections of sori (uredinia) from dried Milesina specimens were mounted in a mixture of lactic acid and glycerol (Kirk et al. 2001) and examined with a light microscope (Zeiss Axioskop 2 plus) at a magnification of 400× or 1000×. If a sufficient amount of spore material were available, 30 spores per specimen were arbitrarily selected and measured. The number of examined specimens was between two (M. magnusiana) and 23 (M. kriegeriana). The number of spores examined depended on sample size and varied for each measurement and also between specimens. The length and width of 30 spores (2–4 specimens per species) and sori (only specimens with Woodwardia host), the length of 15 spines, the cell wall thickness of 10 spores and the distance between 20 spines were measured for each specimen (2–16 specimens per species). For spine base diameters, see next chapter.

Germ pore number and their position in the wall of urediniospores were evaluated by an adapted technique originally developed for the genus Tranzschelia (Scholler et al. 2014). Spores were mounted in Hoyer’s medium (Cunningham 1972) on a slide, then cover slips were pressed until the spores were disrupted and released the plasma. Then the slides were placed on a drier at 40 °C. After two to five days, the numbers of germ pores were counted in phase contrast illumination at 400× magnification for 120 spores of each species. Only the specimens with the best observable germ pores were used for the analysis. In addition, the diameter of pores was measured at 400× magnification.

Specimens were photographed with a Jenoptik ProgRes CT3 digital camera attached to a Zeiss Axioskop 2 plus light microscope (Oberkochen), using differential interference contrast (DIC) and phase contrast as illumination techniques. Images were captured with PROGRES CAPTUREPRO version 2.10.0.1 software. The pictures of the uredinia of Milesina sp. were taken with a ProgRes CT3 digital camera (Jena) attached to a Zeiss Stemi 508 (Zeiss, Oberkochen). All values determined in this study were rounded to one decimal place and outliers were not included in the species description.

Uredinia and urediniospores of dried specimens of Milesina spp. were placed on a holder with conductive double-sided tape (Leit-Tabs, Plano GmbH). Scanning electron microscope images were obtained on a Philips XL 30 FEG environmental scanning electron microscope operated at acceleration voltages of 12 kV at a chamber pressure of 133 Pa (1 Torr). In order to achieve a better contrast and less charge effects, the samples were coated first with a mixture of gold (80%) and palladium (20%) (MED 020, BAL-TEC).

SEM studies were carried out to study surface structures which are not visible by light microscopy. Spine base diameters (30 per species) were also measured with SEM and the software IMAGEJ 1.5.

The statistical analyses for germ pore numbers and boxplots were carried out with the programme R 3.4.3 (R Core Team 2017).

Samples were prepared from herbarium specimens by excising single rust pustules including the plant material. They were placed into micro tubes with 8–12 ceramic beads, 1.4 mm diameter (Bio-Budget technologies, Krefeld, Germany), frozen at -20 °C overnight and homogenised on a Bead Ruptor (biolabproducts, Bebensee, Germany) at a speed of 7.45 m/s for 25 s. After freezing the samples again for 10 min at -20 °C, homogenisation was repeated. DNA was extracted with the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s protocol. Selected samples were homogenised with glass mini mortars and pestles (Roth, Karlsruhe, Germany) in 400 µl of the homogenisation buffer included in the extraction kit.

Molecular barcodes were generated for three loci: ITS (Internal Transcribed Spacer of the ribosomal DNA in the nucleus), 28S (coding for the large subunit of the ribosomal RNA gene located on the ribosomal DNA in the nucleus), nad6 (coding for subunit 6 of NADH dehydrogenase, mitochondrial DNA). Primer sequences are listed in Table 1.

PCR was performed with the Accuprime Taq Polymerase System (Life Technologies, Karlsruhe, Germany) using the supplied buffer II and the following final concentrations: 2 mM MgCl₂, 0.2 mM of each dNTP and 500 nM of each primer. The PCR programme was as follows: 3 min denaturation at 94 °C, 40 amplification cycles (94 °C for 30 s, 50 °C for 30 s and 68 °C for 60 s) and 7 min strand completion at 68 °C. PCR products were visualised in 1.6% agarose gel. Deviations from the 50 °C annealing temperature are listed in Table 1.

After purification of the PCR product with QIAquick-PCR Purification Kit (Qiagen, Hilden, Germany), it was sent to GATC Biotech AG (Konstanz, Germany) for sequencing. Sequencing was performed with the same primers used for the PCR. Forward and reverse sequences were edited and assembled with the software package GENEIOUS 10.0 (Biomatters, Auckland, New Zealand).

Several comparison sequences were selected in order to compare the branch length between Milesina species with branch lengths between related genera. Criteria of selection were availability within the GBOL project and membership in the Pucciniales suborder Melampsorineae sensu Aime (2006) and Aime et al. (2018b). The genera included Puccinia (GenBank) Pucciniastrum, Uredinopsis, Cronartium (GenBank) and Melampsoridium as outgroup. GenBank accessions of Cronartium ribicola ITS sequences are DQ445908 (Hietala et al. 2008), GU727730 (Mulvey and Hansen 2011) and KX574673 (Vogler et al. 2017). GenBank accessions for Puccinia graminis are AY874141, AY874143 and AY874146 (Abbasi et al. 2005).

Sequences were aligned with the ClustalW algorithm implemented in the programme BioEdit, version 7.1.3.0 (Hall 1999) using the standard parameters offered by the programme. Alignments were used for phylogenetic reconstruction by three different methods:

i) Neighbour-Joining (NJ) analysis was performed with the programme PAUP* 4.0b10 (Sinauer, Sunderland, MA, USA) using the Kimura-2-parameter substitution model. Node support values for NJ were calculated from 1000 bootstrap replicates.

ii) Maximum-Likelihood (ML) analysis: The original NEXUS alignment was reformatted to the extended PHYLIP format using the programme Mesquite 2.75 (http://mesquiteproject.org/mesquite/mesquite.html). The PHYLIP alignment was analysed under the ML criterion on the web-based RAxML black box (Stamatakis et al. 2008https://www.genome.jp/tools/raxml/. Both formats were accessed on 01.09.2018). The used substitution model was GTR without GAMMA correction for amongst-site heterogeneity. Node support values were calculated from 100 bootstrap replicates.

c) Bayesian Inference (BI) analysis: The DNA-Substitution model GTR+I+G was used for performing Bayesian analysis with the programme MrBayes 3.2 (Ronquist et al. 2012). Two independent MCMC runs were performed, each with four chains over 1 000 000 generations. Every 100^th tree was sampled. Initial burn-in was 25% and summarisations were calculated after the standard deviation of split frequencies reached below 0.01. The resulting tree file contained posterior probability values for node support.

Tree files resulting from the three methods were visualised using the programme TreeGraph 2 (Stöver and Müller 2010). Alignments are provided as NEXUS files in the Online Supplemental Material (Suppl. materials 1–3).

ITS sequences were generated for 72 specimens of 11 Milesina species (Table 2). These include 10 of 11 Milesina species known to be present in Europe. Only for M. magnusiana no material was available. In addition, we sequenced an unknown Milesina species from the Canary Islands which politically belongs to Europe but geographically may belong to Africa. All specimens are from the fungus collection of the State Museum of Natural History Museum Karlsruhe, Germany (Acronym KR) as listed in the Methods section. An additional 43 specimens (data not shown) yielded no sequences. Thus, the success rate for sequencing was 63%. Amongst the unsuccessful specimens, 17 were collected before 2010 and 12 failed specimens were collected in 2017. The oldest successfully sequenced specimens were from 1999 (M. murariae, KR-M-0025191; M. scolopendrii KR-M-0025400). No attempts have been made to sequence M. magnusiana because only two old species from 1933 and 1964 (M-0290299, M-0205474) were available. Nine ITS sequences were generated for the genera Chrysomyxa, Melampsoridium and Uredinopsis (Table 3).

All 72 specimens with ITS sequences were sequenced for the loci nad6 and 28S. Twenty nine specimens yielded barcode sequences at the locus nad6 (sequencing success 40%), while 24 specimens were successfully sequenced at the locus 28S (sequencing success 33%, Table 2). Since no nad6 or 28S sequences could be generated for specimens of the genus Chrysomyxa, specimens of the genus Pucciniastrum were sequenced as replacements in nad6 and 28S phylograms (three sequences for nad6, one sequence for 28S, Table 3).

Phylogenetic analysis of the ITS barcode revealed four clades for clades within Milesina species. The nodes for the first, second and fourth clade have maximum support values of 100/1/100 for the three phylogenetic reconstruction methods ML, BI and NJ (Figure 1). In clade 1, the ITS sequences of the pairs M. whitei/kriegeriana and M. blechni/woodwardiana sp. nov. are almost identical within the pairs but differ between the pairs by one nucleotide (Figure 3). At position 381, the nucleotide T (M. whitei kriegeriana) is replaced by C (M. blechni/M. woodwardiana sp. nov.). Position 1 is the first nucleotide after the signature TCATTA for the 3´end of the 18S rDNA. This difference is also reflected in the ITS phylogram by a node with weak support of 67/0.69/54 (Figure 1).

Figure 1. 

ITS Phylogram of 11 Milesina species (excluding M. magnusiana). The phylogram is based on a 733-bp alignment. A Maximum Likelihood (ML) tree is shown with support values for ML, Bayesian Inference (BI) and Neighbour Joining (NJ), in the order ML/BI/NJ. Support values are presented when they are above 50 (ML, NJ) or 0.5 (BI). The host is indicated in brackets. Milesina specimens without species designation (host Abies alba) are not colour-coded. For comparison, several sequences were included from closely related genera. They were all newly generated within the GBOL project, except the GenBank sequences for Cronartium spp. The drawings on the right side present the typical arrangement of spines and germ pores (grey dots) on the Milesina urediniospores.

In clade 2, M. scolopendrii, M. polypodii and M. murariae cannot be distinguished by ITS sequences. Apart from single nucleotides at unspecific positions, the ITS sequences are identical. The sequence of M. feurichii differs from the other three species by two nucleotides with specific positions (Figure 4). This difference, however, is not reflected by bootstrap or posterior probability support (Figure 1). Clade 3 with M. vogesiaca and M. exigua has only weak support in the BI and NJ analysis (0.73 and 63) while the bootstrap support was below 50 for the ML analysis (Figure 1). In a version of the ITS phylogram with Puccinia graminis as outgroup (Online Suppl. material 1: Figure S1), the support values for clade 3, including M. exigua are 100/1/100. The only member of the fourth clade is M. carpatica with two identical sequences. The ITS sequence is clearly different from all other Milesina species investigated. In summary, amongst the 11 Milesina species, only four species (M. feurichii, M. vogesiaca, M. exigua and M. carpatica) can be unambiguously assigned by their ITS sequences.

Amongst the specimens on the aecial host Abies alba, nine grouped into clade 1 and two into clade 2 (Figure 1). Following the distinction at position 381, six Abies-dwelling specimens of clade 1 belong to the pair M. whitei/kriegeriana and three to the pair M. blechni/woodwardiana sp. nov.

The ITS phylogeny (Figure 1) does not confirm monophyly of Milesina species. High support values are available only for a clade containing all Milesina species and Uredinopsis filicina (node support (100/1/99, Figure 1). This indicates that the genus Milesina may be paraphyletic. The specimens of the three genera Melampsoridium, Cronartium and Chrysomyxa form a clade with a node support of 81/-/0.9, indicating that probably none of them is a sister group of Milesina. When the branch lengths from the clade defining node to the next deeper node are compared, it its apparent that the branch lengths for the Milesina clades are as long or even longer when compared to the branch lengths between the different genera Melampsoridium, Cronartium and Chrysomyxa. This indicates a relatively large genetic distance between the clades within the genus Milesina.

Due to the low sequencing success of these two markers, only seven (nad6) and eight (28S) Milesina species could be included in the analysis. Although no sequences are available for clade 4 (M. carpatica), the general pattern of the clades is the same as for the ITS phylogeny. Clade 1 and clade 2 consist of the same species (Figure 2) as in the ITS analysis and cannot be distinguished. In addition and in contrast to the ITS data, the distinction between the species pairs M. whitei/kriegeriana and M. blechni/woodwardiana sp. nov. is not possible. In confirmation of the ITS data, the support for a clade that contains all Milesina species and Uredinopsis filicina is high (99/1/100 for nad6, 96/0.96/64 for 28S, Figure 2). This again indicates that the genus Milesina is not monophyletic. The branch lengths from the clade defining node to the next deeper node are shorter between the Milesina clades as compared to the branch length between related species. This is in contrast to ITS data.

Figure 2. 

Phylograms of supplementary barcodes. The nad6 phylogram is based on a 550 bp alignment, the 28S phylogram on a 680 bp alignment. The technical description is the same as for Figure 1. All Milesina specimens from Figure 1 were attempted to sequence for the supplementary barcodes. Only the shown specimens resulted in sequences. The non-Milesina species were altered depending on availability. No GenBank sequences were included and the genus Chrysomyxa was replaced by Pucciniastrum.

Figure 3. 

Deviations from the consensus ITS sequence of section Milesina. The first line indicates the nucleotide positions in base pairs, the second line the consensus sequence. The order of specimens is as shown in Figure 1. “Milesina sp” denotes specimens from Abies alba. Deviations for single specimens can be found at 5 positions. All specimens of M. blechni and M. woodwardiana deviate at position 381 from M. whitei and kriegeriana.

The ambiguity in ITS data to determine a clade 3, consisting of both M. vogesiaca and M. exigua, is also found in the nad6 and 28S data. In the nad6 phylogram, M. exigua has an unsupported position next to Uredinopsis filicina. In the 28S phylogram, M. exigua is only in the same clade with M. vogesiaca if Uredinopsis filicina is included. Even then, the support values of 70/0.8/52 are relatively low.

Germ pores

The number and position of the germ pores of all species were visualised. Germ pores provided three important features, namely (i) the number, (ii) the position and, finally, (iii) the size of pores. The four species with the highest number of germ pores per spore all belong to the section Milesina (Figure 5). All other species had similar germ pore numbers.

Figure 4. 

Deviations from the consensus ITS sequence of section Scolopendriorum. Description as for Figure 3. Milesina feurichii deviates from the other three species in positions 288 (A) and 521 (G).

Figure 5. 

Boxplot of germ pore numbers of urediniospores of 12 Milesina spp. and four sections. For each species 120 spores from two (M. magnusiana), three (M. feurichii) or four (all other species) specimens were evaluated. Median, whisker, quantile and outliers (dots) are shown.

The following key to European Milesina sections and species is based on urediniospore (abbreviated Us) features listed in Table 4. It requires light-microscopical equipment and methods described in the Methods section. The lengths of the urediniospores refer to the main values.

In previous studies of the genus Milesina (e.g. Wróblewski 1913; Faull 1932; Kuprevič and Tranzschel 1957; Berndt 2008), size and shape of urediniospores, wall thickness, spine length and density were the main features used to characterise their morphology. In this study, additional morphological features and criteria are provided to distinguish species (Figure 5, Table 4).

The number, position and size of germ pores have not been documented even in more recent studies of Milesina (Berndt 2008). Additionally, in Chrysomyxa, another genus of Pucciniastraceae (Cao et al. 2017), no germ pores were shown. Germ pores in Milesina are documented in Cummins and Hiratsuka (2003). The authors report “bizonate, obscure” germ pores for species of the genus. We found this character only in the two species of the section Vogesiacae. A further observation of germ pores is reported for two North American species Milesina polypodophila (Bell) Faull and Milesina marginalis Faull & Watson (Moss 1926) where germ pores showed a scattered distribution. With our light microscopic method, detection of germ pores was easy and could be realised within short time. In two species, M. carpatica and M. exigua, germ pores were more difficult to visualise and need more time to evaluate. In general, however, this method is suitable to document an important morphological and taxonomically relevant feature. It may also help to characterise other genera in the Pucciniastraceae. Another feature, smooth areas on the surface of urediniospores has not been documented so far. It is a special character of species of section Scolopendriorum. All of these features in combination allow identification of Milesina species in Europe by urediniospore features alone, using a light microscope even without knowledge of the host plant species. Only one pair of species, the common M. blechni and M. kriegeriana, is difficult to distinguish morphologically. We only found differences in spore length measurements.

In general, identification using only the host is unreliable, since the range of telial hosts in Milesina has been only scarcely studied. This holds true even for common species like M. kriegeriana, a species which has obviously a much wider host range with species in different host families (Berndt 2008) than listed in European compilatory literature (e.g. Gäumann 1959; Majewski 1977; Klenke and Scholler 2015). Polystichum aculeatum is known for hosting M. carpatica and M. vogesiaca (e.g. Gäumann 1959; Majewski 1977; Klenke and Scholler 2015). In the present study, M. exigua was also found on P. aculeatum, demonstrating again that the host range may be wider for several Milesina species and that host identification is not sufficient to identify Milesina spp.

We were able to classify four sections by phylogenetic analysis of ITS sequences (Fig. 1) but we were only able to differentiate 4 of 11 species. This low differentiation is in contrast to another genus in the suborder Melampsorineae sensu Aime (2006). In the genus Chrysomyxa, almost all species could be resolved on the basis of ITS barcodes (Cao et al. 2017). Still, amongst the three tested barcodes ITS, nad6 and 28S in the present study, ITS showed the highest species resolution because it could resolve the pairs M. whitei/kriegeriana and M. blechni/woodwardiana.

The alternative barcode nad6 has been tested on different rust species (Vialle et al. 2009, Feau et al. 2011, Vialle et al. 2013). It was chosen for the present study because, in a study on Melampsora spp., it was the only barcode amongst six tested barcodes (ITS, 28S; CO1, nad6, MS277, MS208) that could distinguish the species Melampsora laricis-tremulae and Melampsora pinitorqua (Vialle et al. 2013) on the basis of a Single Nucleotide Polymorphism. As seen from the branch lengths in Figure 1 and Figure 2 (compare also the scale bars), nad6 sequences show the lowest variation between the sections Milesina, Scolopendriorum, and Vogesiacae as compared to the other two markers. Within the sections Milesina and Scolopendriorum, the sequences were completely identical amongst the species. This low variation makes this marker more suitable for studies on infrafamiliar level than for species distinction on the infrageneric level.

The fungal barcode 28S rDNA is the second most widely used, following ITS (Schoch et al. 2012). It it used for phylogenetic analysis of rusts on the infrafamiliar or higher taxonomic level (Maier et al. 2003, Aime et al. 2006, Yun et al. 2011), but also the identification of closely related species (V, Beenken 2014, Maier et al. 2016, Beenken et al. 2017, Demers et al. 2017, Zhao et al. 2017). In the studies that directly compare ITS with 28S data, species resolution of ITS and 28S are comparable (Ali et al 2016, Beenken et al. 2012, Beenken et al. 2017, Feau et al. 2011) or ITS (namely the sub region ITS2) shows a slightly better resolution (Kenaley et al 2018, McTaggart and Aime 2018). However, when species could not be distinguished with ITS data, distinction was also not possible with 28S data. For instance, European specimens of Coleosporium species C. euphrasiae (Schumach.) G. Winter, C. campanulae (Pers.) Lév. and C. senecionis (Pers.) Fr., each occurring on telial hosts in different plant families, could neither be distinguished by ITS nor by 28S sequences. This is also clearly the case in the present study for the Milesina species in the sections Milesina and Scolopendriorum that colonise different fern species, but have almost identical ITS and 28S sequences.

One possible solution to lacking species resolution is to declare all specimens with the same ITS sequence data as one species, which was the original concept of ITS barcoding (Schoch et al. 2012). However, it is not only the telial host that differs between the Milesina species in the section Scolopendriorum, but also the features of the urediniospores (see identification key and Table 4). In the case of contradicting results, it is not advisable to weight the ITS data as more reliable than the morphological/host data. In the ascomycete genus Fusarium, several species complexes could not be resolved by ITS data, but with newly developed barcodes TOPI (Topoisomerase I) and PGK (phosphoglycerate kinase, Al-Hatmi et al. 2016). It is also possible that, for the rust fungi, new barcodes can be developed on the basis of genomic data (Aime et al. 2018b) that finally allow morphologically determined species to be resolved.

Not all specimens studied morphologically were used for sequencing (i.e. old specimens, type specimens, specimens with little spore material) and not all of those specimens, where DNA was extracted, were successfully sequenced. The rate of successful ITS sequencing (63%) is relatively low. In a previous study on Melampsora rust fungi on Salix, the rate of ITS sequencing success (93%) was much higher (Bubner et al. 2014). The Melampsora study comprised exclusively freshly collected specimens not older than half a year, whereas we also analysed herbarium material that was several years old. However, freshly collected specimens (from 2017) also failed, while the two oldest successful specimens are from 1999. Therefore, it is not only the age of the samples that explains the low success of sequencing. It is possible that the small and fragile sori of Milesina species (in comparison, for instance, to Melampsora sori on Salix) are more prone to DNA decay on the herbarium specimens.

Even more surprising is the low success rate for the 28S sequencing. The 28S sequencing was performed only on samples with successful ITS sequencing. Template DNA should be present because both loci belong to the same multicopy rDNA region on the nuclear DNA. Despite this linkage, 28S is reported to have a PCR success rate of only 80% as compared to ITS in a large scale study on Basidiomycota including Pucciniomycotina (Schoch et al. 2012). Nevertheless, in recent phylogenetic studies on rusts, often concatenated alignments of ITS and 28S are used (e.g. Beenken 2017, Demers et al. 2017, McTaggart and Aime 2018, Vialle et al. 2013) which requires that both ITS and 28S can be sequenced. Although phylogenetic studies rarely report a success rate, it can be assumed that routine sequencing of both loci is possible. The low sequencing success in Milesina could be a genus-specific problem. We used 28S primers (ITS4BRF, LR5) which Vialle et al. (2013) successfully used for sequencing Melampsora spp. on poplars for Milesina spp., however, with much less success. Possibly 28S rDNA of Milesina is more difficult to sequence than in other rust genera. Other primer combinations for sequencing 28S rDNA are available and could be tested. The requirements of further testing demonstrate that, in the genus Milesina, species identification by barcode sequencing is still far from being routine.

The support values for section Vogesiacae are smaller than for the other three Milesina sections when M. exigua is included. Interestingly, the support values are 100/1/100 for an ITS tree with Puccinia graminis as outgroup. The decision to place both M. vogesiaca and M. exigua into one section is more strongly supported by morphological than by molecular data. Milesina vogesiaca and M. exigua are the only two species that have urediniospores without ornamentation and a bizonate position of germ pores. Further support for a molecularly and morphologically defined clade is given through Uredinopsis filicina. In the ITS phylogram, it groups behind a node with high support values that includes both M. vogesiaca and M. exigua. The inclusion of U. filicina within a clade, that comprises all Milesina species (also M. carpatica), indicates that the genus Milesina is paraphyletic. The paraphyly is also indicated in the nad6 and 28S phylograms. Furthermore, M. vogesiaca, M. exigua and U. filicina form a group with high support values in 28S phylogram.

By morphology of urediniospores, U. filicina (the type species of Uredinopsis) is similar to the two Milesina species in the section Vogesiacae because it also has smooth urediniospores (Gäumann 1959). Germ pores have not been analysed so far. Germ pores are reported for two North American species, U. osmundae Magnus and U. atkinsonii Magnus (Moss 1926). Urediniospores are smooth and show the same bizonate position of germ pores as documented for M. vogesiaca and M. exigua. A recent study on molecular age estimates in Pucciniales presents 28S data of Melampsorineae that also comprises a limited selection of Uredinopsis and Milesina specimens (Figure 3 in Aime et al. 2018a). The species U. osmundae, U. filicina and M. vogesiaca group together, while M. scolopendrii and an undetermined Milesina species form a sister clade. This confirms the topology of the trees in the present study. The molecular and the morphological data indicate that at least U. filicina is actually a Milesina species in the clade Vogesiaceae. To place U. filicina (and possible other species of Uredinopis) in the genus Milesina, however, requires a more comprehensive sampling of Uredinopsis species and sequencing of both ITS and 28S rDNA (study in preparation).

Amongst the 11 sequences of specimens found on the aecial host Abies alba, nine could be assigned to the section Milesina. Although this section contains four species, the question which species is able to form aeciospores on Abies alba can be narrowed down to three species. Only telial hosts of M. blechni, M. kriegeriana and M. whitei grow in the distribution area of Abies alba in Europe. Therefore, M. woodwardiana can be excluded, because the host Woodwardia radicans is restricted to Macaronesia and the Mediterranean (Hohenester and Welss 1993, Li et al. 2016) from where no Milesina sequences from the aecial host (Abies spp.) are available. In addition, M. woodwardiana obviously does not form telia and, consequently, no basidia and basidiospores. Basidiospores, however, are necessary to infect Abies.

The answer to the question which of the two species M. whitei / M. kriegeriana has an alternate host needs further field observation and experimental studies (inoculation experiments). Our specimens most probably belong to M. kriegeriana, because they were all collected in the Black Forest area (SW Germany) where we found M. kriegeriana many times on the telial host but not M. carpatica. Inoculation experiments should not only include the hosts of M. whitei (Polystichum spp.) and M. kriegeriana (Dryopteris spp.), but also Struthiopteris spicant, the host of M. blechni. This would further help to answer the question whether M. blechni and M. kriegeriana are distinct species or not. Despite the SNP at position 381, both species are very similar in urediniospore morphology. Inoculation experiments would provide further arguments to clarify the status of the two species. Another approach to analyse both host alternation and species distinction in the section Milesina would be to measure gene flow between the aecial (Abies) and the different telial hosts (ferns) by population genetics. Gene flow measurements in rust fungi have been applied to Melampsora larici-populina (Barres et al. 2012; Elefsen et al. 2014).