The distribution and evolution of fungal symbioses in ancient lineages of land plants

An accurate understanding of the diversity and distribution of fungal symbioses in land plants is essential for mycorrhizal research. Here we update the seminal work of Wang and Qiu (Mycorrhiza 16:299-363, 2006) with a long-overdue focus on early-diverging land plant lineages, which were considerably under-represented in their survey, by examining the published literature to compile data on the status of fungal symbioses in liverworts, hornworts and lycophytes. Our survey combines data from 84 publications, including recent, post-2006, reports of Mucoromycotina associations in these lineages, to produce a list of at least 591 species with known fungal symbiosis status, 180 of which were included in Wang and Qiu (Mycorrhiza 16:299-363, 2006). Using this up-to-date compilation, we estimate that fewer than 30% of liverwort species engage in symbiosis with fungi belonging to all three mycorrhizal phyla, Mucoromycota, Basidiomycota and Ascomycota, with the last being the most widespread (17%). Fungal symbioses in hornworts (78%) and lycophytes (up to 100%) appear to be more common but involve only members of the two Mucoromycota subphyla Mucoromycotina and Glomeromycotina, with Glomeromycotina prevailing in both plant groups. Our fungal symbiosis occurrence estimates are considerably more conservative than those published previously, but they too may represent overestimates due to currently unavoidable assumptions. Electronic supplementary material The online version of this article (10.1007/s00572-020-00938-y) contains supplementary material, which is available to authorized users.


Introduction
Fungi colonize plants and interact with their living tissues in a variety of ways; these interactions can be detrimental (parasitic), neutral (symptomless) or beneficial (mutualistic) to the host plant. More than 85% of vascular plant species are considered to form mutually beneficial symbioses in their roots, termed mycorrhizas, with soil fungi (Brundrett and Tedersoo 2018). This percentage is only an estimate because investigating every plant species is neither practical nor currently possible given that not all species are known and ca. 2000 new vascular plants species are described each year (Pimm and Raven 2017). For the most part, fungal symbiosis occurrence rate estimates are lacking for early-diverging plant lineages as little effort has been directed towards compiling the data required to allow these estimations to be made. This also reflects an overall paucity of data available on these groups, including information on the type of interaction formed, i.e. whether the interaction is mycorrhizal or mycorrhizal-like in plants such as liverworts and hornworts that lack true roots. However, in the last decade, there has been an increased research focus on the diversity and distribution of fungal associations in liverworts, hornworts and lycophytes, largely driven by the discovery of Mucoromycotina fungi in association with these plants Desirò et al. 2013; Rimington et al. 2015) and the demonstration that at least some of these associations are mycorrhizal or mycorrhizal-like-i.e. those between lycophytes and Mucoromycotina ); between liverworts and Glomeromycotina (Field   and Glomeromycotina and Mucoromycotina together (Field et al. 2016b); and between liverworts and Ascomycota ). We address this lacuna by compiling published fungal symbiosis status for these early-diverging plant lineages with the caveat that some of the reported symbioses, e.g. those in hornworts, are considered such on the basis of morphology and/or involvement of fungi known to be mycorrhizal with other plant lineages but are yet to be confirmed experimentally. A comprehensive list of which plant species enter into fungal symbioses and with which fungi not only serves as a useful resource for future studies but also provides insight into the origins and distribution of these relationships and how they evolved across plant lineages (Wang and Qiu 2006). This is particularly pertinent today as recent studies are finally providing much improved resolution on the phylogenetic relationships among the earliest-diverging bryophytes (liverworts, mosses and hornworts) and vascular plants, which have been contested for decades (e.g. Hornworts and some liverworts also form endosymbioses with cyanobacteria (Nostoc sp.) (Adams and Duggan 2008). In hornworts, these associations are ubiquitous (Renzaglia et al. 2007), while in liverworts, they occur only in two Marchantiopsida species that lack fungal symbionts, Blasia pusilla and Cavicularia densa . Associations with cyanobacteria have also been reported in some moss species; however, these are exclusively epiphytic or endophytic in the dead hyaline cells in Sphagnum leaves (Kostka et al. 2016; Warshan et al. 2017).
Recently, it has been shown that lycophytes also form associations with Mucoromycotina and Glomeromycotina fungi , with emerging evidence of carbonfor-nutrient exchanges between these early-diverging vascular plants and their Mucoromycotina symbionts . A better understanding of fungal associations in lycophytes is important when considering the early evolution of land plant-fungus symbiosis. Lycophytes, which comprise ca. 1360 species (Hassler and Schmitt 2018), are the earliest branching lineage of vascular plants (tracheophytes) and represent the transition from non-vascular to seed plants (Kenrick and Crane 1997). They are of particular importance because putative transitional 'pre-vascular' plants, including Rhynie Chert fossils such as Aglaophyton, are all extinct (Remy et al. 1994). As such, extant lycophytes are considered the best modern analogues for the first vascular plants (Kenrick and Crane 1997).
Lists detailing the fungal symbiosis status of plants have been published for many years; for example, the first list of fungal symbiosis in liverworts was produced 70 years ago . Such lists require regular updating as the number of studies increases and so does our knowledge of the diversity of symbioses within and across plant clades. Earlier compilations usually focused on a local scale and only on certain, almost invariably vascular, plant groups . It was not until 2006 that a worldwide literature survey of fungal symbioses across all land plant groups was performed (Wang and Qiu 2006). This landmark publication by Wang and Qiu (2006) captured the status of over 3000 species (143 of which were bryophytes) and, unsurprisingly, has been highly influential ever since. In the 13 years since its publication, this paper has been one of the most cited on mycorrhizas (over 1500 citations as of January 2020) and has provided important insights on the evolution of mycorrhizas; for example, evidence that arbuscular mycorrhizas (AM) are found throughout the land plant phylogeny has been used as a key argument for Glomeromycotina symbiosis being an ancestral trait of land plants  Fungal symbiosis occurrence rate estimates are commonly used to highlight the near-ubiquity of these relationships. For instance, few publications concerning AM fail to mention that at least 80% of plant species form these symbioses, most commonly citing the reference book 'Mycorrhizal Symbiosis' (Smith andRead 1997, 2008). These estimates are useful for emphasizing the importance of mycorrhizas to broad audiences and to highlight the diversity of these relationships between fungi and plants. These estimates are useful starting points for more refined estimates; recently, re-examination has shown that 80% may be an overestimation for AM symbioses, with the true value probably closer to 71% (Brundrett and Tedersoo 2018). Fungal symbiosis occurrence estimations for early-diverging plants have been more sporadic and highly variable including Glomeromycotina symbioses occurring in 60% and 100% of liverwort and hornwort species, respectively (Brundrett 2009) and 25% of bryophytes forming fungal associations, the majority of which involve Glomeromycotina (Brundrett and Tedersoo 2018). The last figure fails to take on board the fact that mosses, the most speciose group of bryophytes with ca. 12,000 species, lack fungal symbionts.
We present a new global compilation of the fungal symbiosis status of liverworts, hornworts and lycophytes. Our compilation more than triples the number of early-diverging plant species listed in Wang and Qiu (2006) and is the first to focus on early-diverging plant lineages on a global scale.

Literature survey
A survey of the published literature on fungal symbioses in liverworts, hornworts and lycophytes was performed. Reexamination of Wang and Qiu's survey (Wang and Qiu 2006) revealed that some key references for these plants were missing and that fungal symbiosis status was often reported only as 'fungal association' without specifying the fungus involved; thus, a full search was performed, including studies prior to 2006. In trying to capture all available references, several keywords were used as search terms in Google Scholar. In each search, one of the following plant terms was used: 'liverwort', 'hornwort', 'lycopod' and 'lycophyte'. Each plant term was combined with one of t h e f o l l o w i n g f u n g a l t e r m s : ' f u n g i ' , ' f u n g u s ' , 'Glomeromycotina', 'Mucoromycotina', 'Glomeromycota', 'Glomus tenue' and 'fine endophyte'. Additionally, for liverworts, which are known to form more diverse fungal symbioses than the other two lineages, the following terms were also used: 'Ascomycota', 'Basidiomycota', 'Rhizoscyphus', 'Pezoloma', 'Sebacina' and 'Tulasnella'. Using these criteria, a total of 34 searches were performed. The titles and abstracts of all references returned by the searches were scrutinized to identify reports of the fungal status of any liverwort, hornwort or lycophyte species. Where the search terms returned more than 500 hits (e.g. 'lycopod fungi' returned 14,600 hits), only the first 500 results were investigated. Fungal symbiosis status was recorded as Glomeromycotina, Mucoromycotina, Ascomycota, Basidiomycota or non-symbiotic. Additionally, the presence of dark septate endophytes (DSE) was recorded for lycophytes. For some liverwort and hornwort species, only the presence of a 'fungal association' was recorded as the fungal lineage could not be assigned. As well as recording the fungal status, the identification method (microscopy and/or DNA sequencing) was noted for all species. The publications found through Google Scholar that were deemed relevant to the investigation were read and any literature found within those publications, but not returned directly by Google Scholar, was also included. This secondary search method returned exclusively microscopy studies published prior to 1990 (and dating back to 1891); thus, we are confident that all relevant molecular studies were found with our main search method. Additionally, information on the fungal symbiosis status of some liverwort species was obtained either from the liverwort flora of Paton (1999) or from our own unpublished microscopy observations (25 species; see Table S2).

Plant nomenclature
Nomenclature for liverworts and hornworts follows the most recent floras (Söderström et al. 2016; Stotler and Crandall-Stotler 2017) and the Tropicos database (www.tropicos.org); taxonomic rankings above genus level follow Söderström et al. (2016). For lycophytes, nomenclature follows the Checklist of Ferns and Lycophytes of the World by Hassler and Schmitt (2018). When currently accepted names differ from those in the original reports, both are given in Table 1, with the latter appearing in parentheses.

Estimating symbiosis occurrence rates
Fungal symbiosis occurrence rates were estimated for each of the three early diverging plant lineages: liverworts, hornworts and lycophytes. The number of species per genus or family and the total number of species per lineage were based on Söderström et al. (2016) for liverworts and hornworts and on Hassler and Schmitt (2018) for lycophytes. When making estimates for hornworts and lycophytes, if a species within a genus was colonized by a fungal lineage, then it was assumed that all members of the genus have the potential to be colonized by that fungal lineage. Underlining this assumption was the finding of fungi by our own observations on fresh specimens of the same genera. The total number of species potentially colonized in a plant lineage was divided by the total number of species in that lineage and multiplied by 100 to produce an estimate for the fungal symbiosis occurrence rate. In instances where the fungal status of a genus was unknown or reported only as 'fungal association', the genus was not included in the calculations and the total number of species was reduced accordingly. The same method was applied to liverworts but using the family level rather than the genus, Table 1 The fungal symbionts of early-diverging plants. Mucoro -Mucoromycotina, Glom -Glomeromycotina, Asco -Ascomycota, Basid -Basidiomycota, FA -Fungal association with unidentified fungi, NS -non-symbiotic, DSE -dark septate endophytes. Species labelled 'Mucoro (FRE)' were reported only as being colonized by fine root endophytes (i.e. Glomus tenue). A question mark after 'Mucoro' signifies it was not reported in the original publication but microscopy images are indicative of Mucoromycotina colonization. Checks indicate whether DNA sequencing and/or microscopy were used for fungal identification. An asterisk specifies our unpublished personal observations. In the column labelled Fungi, a hash indicates a report considered incorrect as a result of further studies. A cross signifies a likely incorrect report that is discussed in the main text. Species in bold had conflicting reports of symbiotic status. Where appropriate, the species names used in original reports are provided in parentheses. Reference numbers are listed below the                    (Table 1) plus our own observations on freshly collected specimens of a range of Riccardia species. 2) Plagiochilaceae-This is the most speciose family in the Jungermanniales with 767 species in ten genera; however, fungal symbiosis has only been reported in the fourspecies genus Pedinophyllum. For calculations, we considered Pedinophyllum to be the only Plagiochilaceae genus (Feldberg et al. 2010) that can be colonized by symbiotic Basidiomycota and the rest were considered nonsymbiotic. Re-enforcing this assumption is that fact that neither Schuster (1980) nor Paton (1999) mention fungi other than in Pedinophyllum, and we have never seen them in fresh specimens of over 50 species in the family. 3) Gymnomitriaceae-This relatively speciose family (97 species) of nine genera contains only one genus (Nardia) for which fungal symbiosis has been reported, and the rest are non-symbiotic; thus, for calculations, we considered Nardia to be the only symbiotic genus in Gymnomitriaceae. As for the Plagiochilaceae, we have never seen fungi in freshly collected specimens other than in Nardia. The Gymnomitriaceae predominantly grow on bare rock, a substrate ill-suited to fungal symbioses. 4) Jungermanniaceae-Fungal symbiosis has only been reported in Eremonotus, a single species genus . All the other members of this family (37 species) that have been investigated (Paton 1999; do not enter into fungal symbiosis, so only Eremonotus was considered to be symbiotic in our calculations.
Numbers of species per genus/family are given in Table S1.

Plant species numbers
The fungal symbiosis status of up to 648 liverwort, hornwort and lycophyte species, belonging to 194 genera, 82 families and 23 orders, was compiled (Table 1) by combining data from 84 publications. The number of species for each of these early-diverging plant groups and the fungal lineages that colonize them are listed in Table 2. The total value, 648 species, includes seven subspecies and 53 samples identified only to the genus level (sp.) that may represent duplicates (except when they are the only entry for that genus, e.g. Lepidogyna sp.). Thus, at least 591 species are included in our survey ( Table 2). This represents a considerable increase on the number of early-diverging plant species, 180, included in Wang and Qiu's survey (Wang and Qiu 2006). The hornworts and lycophytes are well represented; our survey includes members of every hornwort and lycophyte family and of most genera except for one hornwort and four lycophyte genera ( Table 1). The liverworts are less well represented; this is because of their higher diversity, comprising over twenty times the number of genera found in hornworts or lycophytes. While  Table 1. The remaining 210 genera with unknown fungal symbiosis status are members of the Metzgeriidae and Jungermanniidae. This reflects a research bias, as most studies have focused on species from known symbiotic clades (e.g. 24% of Pelliidae species and 17% of Marchantiopsida species have been investigated) while neglecting those from clades considered to be largely asymbiotic. Indeed, only 5% of Jungermanniidae species have been investigated to date, reflecting that the Lejeuneaceae, the most speciose Jungermanniidae family with ca. 2000 species, is asymbiotic ). Since the survey by Wang and Qiu was published in 2006, the use of DNA sequencing to identify plant fungal symbionts has increased dramatically. To date, the fungal status of 259 fully named early-diverging plant species has been analysed by molecular methods versus only six reported in Wang and Qiu (2006).
Our survey unveiled contradictory reports on the fungal symbiotic status (symbiotic vs. non-symbiotic) of 51 species (42 liverworts, one hornwort and eight lycophytes) probably reflecting low fungal colonization levels , habitat type and/or seasonal variation in colonization (personal observations) in these species. Colonization by two fungal lineages has been reported in 51 species (35 liverworts, 11 hornworts and 5 lycophytes). We found no report of more than two fungal lineages colonizing the same plant species. All dual colonisations involve either members of Mucoromycotina and Glomeromycotina (Mucoromycota) or Ascomycota and Basidiomycota (Dikarya), with the former (45 species) being more common than the latter (5 species).

Estimating symbiosis occurrence rates
Our estimates of fungal symbiosis occurrence rates for the different fungal lineages in liverworts, hornworts and lycophytes show that fungal symbiosis appears to be the norm in hornworts and lycophytes, but not in liverworts (Table 3). Occurrence rates were easier to estimate for hornworts and lycophytes than for liverworts, as these two groups contain less species and engage in less diverse symbioses than liverworts. We estimated that 69% of hornwort species can be colonized by Mucoromycotina fungi and 78% by Glomeromycotina (Table 3). In lycophytes, colonization by Glomeromycotina is higher than by Mucoromycotina; 99% of lycophyte species can potentially form AM while only 4% are estimated to be symbiotic with Mucoromycotina Table 2 The numbers of earlydiverging plant species for which fungal symbiosis status has been reported. M -Mucoromycotina, G -Glomeromycotina, B -Basidiomycota, A -Ascomycota, FA -Fungal association. Where reports were contradictory (symbiotic and non-symbiotic), the symbiotic report is included. The number between parentheses represents the maximum number of different species, reflecting that some species were identified as 'sp.' so could represent duplicates of fully identified species  Table 3). The fungal status of each hornwort and lycophyte genus is found in Table S1. Our estimates of fungal symbiosis occurrence rates in liverworts had to be calculated at the family, rather than genus, level (except for four families, as explained previously) because this group contains many more genera (ca. 386) than hornworts and lycophytes (12 and 18 genera, respectively) and the fungal symbiosis status of less than half (169) of these genera is currently known. However, the fungal symbiotic status of most liverwort families has been reported, with that of only 15 out of 87 families remaining unassigned. These 15 families all have low species numbers: less than ten species except for one family. Thus, the fungal symbiosis status of liverworts is well represented at the family level (Table S1).
We estimated that only 4% and 5% of liverwort species are colonized by Mucoromycotina and Glomeromycotina, respectively (Table 3). Symbioses involving Basidiomycota (7%) and Ascomycota (17%) appear to be more common in liverworts but an absence of fungal symbiosis is by far the prevalent state (71%). The sum of these estimates is greater than 100% due to several liverwort species forming dual colonization with both Mucoromycotina and Glomeromycotina. Below we consider the major liverwort groups individually: Haplomitriopsida-Up to 100% of these earliest-diverging liverworts can be colonized by Mucoromycotina fungi. There has been a single molecular report of Glomeromycotina symbiosis in Haplomitrium chilensis ); however this report was published prior to the discovery of Mucoromycotina colonization in liverworts and has since been questioned by several molecular investigations Marchantiopsida-These are the earliest-diverging liverworts to form Glomeromycotina symbioses; however, fungal colonization is relatively low and 22% and 38% of Marchantiopsida liverworts are estimated to be colonized by Mucoromycotina and Glomeromycotina, respectively. These results are skewed by the absence of symbionts from the most speciose Marchantiopsida family, Ricciaceae (Table S1), where both terrestrial and aquatic taxa lack symbionts. When Ricciaceae is excluded from calculations, the colonization estimates increase to 43% for Mucoromycotina and 74% for Glomeromycotina.
Pelliidae-This is the latest-diverging liverwort group to form Mucoromycotina and Glomeromycotina symbioses, and colonization is common at 97% and 99%, respectively.
Metzgeriidae-Basidiomycota colonization is estimated to occur in 44% of Metzgeriidae liverworts. If no assumption of 50% colonization in Riccardia species was applied to our calculations (see exception 1 in 'Methods'), then this estimate would increase to 75%.
Jungermanniidae-Ascomycota and Basidiomycota have only been reported in the Jungermanniales and are not present in the Porellales or Ptilidiales. Our calculations suggest that 5% of Jungermanniidae species can be colonized by Basidiomycota while 20% can be colonized by Ascomycota.
Our occurrence rate estimations for Glomeromycotina colonization in early-diverging land plants disagree with those published previously by Brundrett (2009), except for lycophytes. For the latter, our results agree with 100% colonization (Brundrett 2009) (Table 3). For hornworts, our estimate of 78% is lower than the previous one of 100% (Brundrett 2009), although it confirms that colonization by Glomeromycotina in hornworts in common. The most striking discrepancy is between our finding that only 5% of liverworts likely form arbuscular mycorrhizal-like associations and the  Brundrett (2009). Furthermore, our results indicate that previous estimates for the formation of any type of fungal symbiosis in bryophytes have also been excessive. Wang and Qiu (2006) estimated that 46% of bryophytes enter into symbiosis with fungi, whereas Brundrett and Tedersoo (2018) put this value at 25%, while also stating that in bryophytes the majority of these relationships involve Glomeromycotina fungi. In our study, after accounting for the ca. 13,000 non-symbiotic moss species, we estimate that only 11% of bryophytes enter into a symbiosis with fungi and that the most widespread symbiosis is with Ascomycota (53%) rather than Glomeromycotina (33%). The large number of species in Lepidoziaceae, within Jungermanniidae (751 species), is principally responsible for the Ascomycota occurrence rate estimate being higher than that of the other fungal lineages combined. Even though our fungal symbiosis occurrence rates are considerably lower than previously published ones, they too may represent overestimates since our calculations are based on the assumption that all members of a plant genus (or family for liverworts) can be colonized by a fungal lineage if at least one member of the genus (or family) is colonized by that lineage. While efforts were made to prevent overestimation in four liverwort families where an absence of symbiosis is common (Aneuraceae, Gymnomitriaceae, Jungermanniaceae and Plagiochilaceae), more data are needed to determine which families are fully symbiotic and for which symbiosis is more variable. Another important consideration in these estimations is the symbiotic status of the fungi colonizing plants. All lineages of Mucoromycotina related to the Endogonales and Glomeromycotina are considered to be mycorrhizal-like when in association with early-diverging plants Field et al. 2016a, b). This is however not the case for Ascomycota and Basidiomycota, which are far more diverse than Mucoromycotina and Glomeromycotina and regularly colonize these plants as commensals or parasites . The structures formed by Ascomycota and Basidiomycota while colonizing early-diverging plants are not necessarily diagnostic of mutualisms, and thus, it is difficult to infer mutualistic, commensal or parasitic relationships based on morphology alone ). Therefore, morphological observations of Basidiomycota in Metzgeriidae and Ascomycota and Basidiomycota in Jungermanniidae may not necessarily reflect mycorrhizallike relationships. An additional complication is that, at present, Hyaloscypha (Pezoloma, Rhizoscyphus) ericae is the only Ascomycota species for which mutualistic nutrient exchange with liverworts has been confirmed ); thus, reports of colonization by Ascomycota that have not been identified as H. ericae using DNA sequencing may not represent mutualisms. For Basidiomycota, so far only Tulasnella and Serendipita (Sebacina) have been reported as genera symbiotic with liverworts (Bidartondo and Duckett 2010); however, both associations await physiological tests for exchange between partners. It follows that colonization of liverworts by mycorrhizal-like Ascomycota and Basidiomycota may have been overestimated and efforts are now required to identify molecularly the fungal symbionts of these plants as well as testing for nutrient exchange.
In contrast, Mucoromycotina occurrence rates are likely underestimates, especially for lycophytes. Traditionally, the unique structures of Glomeromycotina, in particular the arbuscules, made them easily and accurately identifiable through microscopy (Smith and Read 2008). However, the recent discovery of endosymbiotic Mucoromycotina, which cannot be distinguished from Glomeromycotina cytologically (Desirò et Table 1 are actually incorrect, although, at present, it is not possible to determine if and how these potential misidentifications might have influenced our occurrence rate estimations.
These caveats aside, our estimates can still be considered the best fungal symbiosis occurrence rates to date for earlydiverging plants. While those for early-diverging liverworts are based on fairly comprehensive information and are unlikely to change with additional data, those for later-diverging groups are likely to improve as more data become available for these plants.

Inferring gains and losses of symbiosis
The gains and losses of fungal symbiosis during the evolutionary history of the early-diverging plant families included in Table 1 have been inferred (Fig. 1). These are discussed below: Liverworts-The liverworts have had a more diverse history of losses and gains of symbiosis than the hornworts and lycophytes. Mucoromycotina likely formed the ancestral symbiosis with liverworts and appear to have been maintained as the sole symbionts in the Haplomitriopsida (Rimington et al. 2019). There have been losses of Mucoromycotina symbiosis in Marchantiopsida liverworts during the divergence of the Blasiales, Sphaerocarpales and Marchantiales. In the Marchantiales, symbiosis has been regained in three families, Monocleaceae, Aytoniaceae and Targioniaceae. In Pelliidae, Mucoromycotina symbiosis has been maintained in all families except four (Hymenophytaceae, Phyllothalliaceae, Petallophyllaceaeand Noterocladaceae). Conversely, Glomeromycotina symbiosis likely had a single origin in liverworts after the divergence of the Haplomitriopsida, followed by several losses in the Marchantiopsida, from Sphaerocarpales and six families of the Marchantiales, but only one loss in the Pelliidae, from the Phyllothalliaceae. After the divergence of the Pelliidae, there was a complete loss of both Mucoromycotina and Glomeromycotina symbioses in liverworts. Basidiomycota and Ascomycota symbioses appear to have been gained and lost multiple times during the evolution of the Metzgeriidae and Jungermanniidae. In the Metzgeriidae, there was a single gain of Basidiomycota symbiosis within the Aneuraceae and a subsequent loss from a large number of the later-diverging Riccardia species (Rabeau et al. 2017). Because the fungal symbiosis status of many Jungermanniidae families remains unresolved, it is not yet possible to accurately estimate gains and losses of Ascomycota and Basidiomycota symbioses in this subclass. Based on the better-studied families (highlighted in grey in Fig. 1), Ascomycota symbiosis appears to have evolved at least six times, with two major losses, while Basidiomycota symbiosis appears to have been gained on at least four occasions, with at least one loss. Alternatively, it is possible that Ascomycota and Basidiomycota symbioses had a single origin in the Jungermanniales followed by a large number of losses. Although this seems less likely, multiple losses of AM and rhizobia have been inferred in angiosperms, so until the fungal symbiosis status of these liverworts is fully resolved for all families, ancestral reconstruction will be of limited value to further our understanding of fungal associations in these plants.
Hornworts-Apart from some individual losses and apparent regains in certain hornwort species ), both Mucoromycotina and Glomeromycotina symbioses have been maintained throughout the Anthocerotopsida. Fungal symbiosis has never been recorded in the single species class Leiosporocerotopsida that contains the earliest-diverging extant hornwort Leiosporoceros dussii. Leiosporoceros dussii is notable not only for its lack of fungal symbiosis but also for its unique cyanobacterial symbiosis (Villarreal and Renzaglia 2006). With the order of divergence of the bryophytes under debate (Puttick et al. 2018), it is unknown whether Mucoromycotina and Glomeromycotina are both ancestral symbionts of all hornworts and were lost from Leiosporocerotopsida or whether these symbioses were gained in the hornworts only after Leiosporocerotopsida branched off. It also remains to be determined whether members of the Phymatocerotaceae are colonized by Mucoromycotina, Glomeromycotina or both fungi since the only record for this family is a report of 'a fungal association' ; however, the regular colonization of the other Anthocerotopsida families by Mucoromycotina and Glomeromycotina suggests this family is also colonized by both fungal lineages.
Lycophytes-Phylogenetic inference (Fig. 1) and fossil evidence (Strullu-Derrien et al. 2014) both support that the ancestor of all vascular plants entered into symbiosis with Mucoromycotina and Glomeromycotina. Within the lycophytes, there have only been losses of symbiosis and no subsequent gains. The loss of Mucoromycotina symbiosis appears to have occurred on a larger scale than that of Glomeromycotina symbiosis, with a major loss after the divergence of the Lycopodiaceae which resulted in Isoëtaceae and Selaginellaceae apparently being colonized only by Glomeromycotina. It should be noted however that no fungal molecular data have been generated from Isoëtaceae and all microscopy reports predate the discovery of Mucoromycotina in lycophytes; therefore a symbiosis with Mucoromycotina cannot be ruled out. Additionally, only three of the 688 Selaginellaceae species have been analysed molecularly ; therefore, this family may also enter into symbiosis with Mucoromycotina as well as Glomeromycotina. There have also been losses of Mucoromycotina symbiosis within the Lycopodiaceae and the subfamily Huperzoideae is only colonized by Glomeromycotina. Within the subfamily Lycopodioideae there appears to have been a complete loss of symbiosis in the Lycopodiastrum-Pseudolycopodium-Austrolycopodium-Dendrolycopodium-Diphasium clade (Field et al. 2016a). The low levels of colonization of lycophytes by symbiotic fungi and the evolution of non-symbiotic species suggest that these plants may have a low dependence on their mycorrhizal partners when mature . On the other hand, the gametophytes of lycophytes are often subterranean and achlorophyllous and therefore fully dependent on their symbiotic fungi for nutrition .

Identifying lycophyte fungal symbionts
The identity of the fungi that enter into symbiosis with lycophytes and the extent of these symbioses remain poorly resolved. While the available evidence indicates that only Mucoromycotina and Glomeromycotina colonize members of this lineage (Pressel et al. 2016), more work is needed to confirm this and to determine which symbionts dominate in n a t u r e ( L e h n e r t e t a l . 2 0 1 7 ) . S y m b i o s i s w i t h Glomeromycotina has been reported more frequently than with Mucoromycotina (53 species vs. 6); however, most of these reports precede the discovery of Mucoromycotina-plant symbiosis and also lack molecular identification. Indeed, a recent molecular survey found a smaller difference in incidence of colonization between the two fungal lineages, albeit with Glomeromycotina also being the dominant type . There has only been one report of colonization by Basidiomycota in lycophytes ). However, because of a lack of electron microscopy evidence and of molecular methods suitable for detecting Mucoromycotina in this report, and as it contradicts all previous and subsequent reports (Table 1)

Conclusions
In concluding their seminal work, Wang and Qiu (2006) highlighted that 'more basal land plants should be investigated, as they occupy an especially important position in our understanding of the origin of mycorrhizal symbiosis'. In the subsequent 13 years considerable effort has gone into addressing some of these gaps in knowledge so that the fungal symbiosis status of more than three times the number of earlydiverging species reported in Wang and Qiu is now known. Nevertheless, further research is still required as to date only 6% of liverwort, 13% of hornwort and 5% of lycophyte species have been examined. Within liverworts, our survey highlights Jungermanniidae as the group in most need of further investigation. Lycophytes also require further investigation; it is likely that estimates of the occurrence of Mucoromycotina symbiosis in this lineage will increase with additional use of molecular methods.
Compiling this survey of fungal symbioses in earlydiverging plants has highlighted the importance of both DNA sequencing and microscopy for determining the identity of plant fungal symbionts. Microscopy alone is not enough to identify fungi unless they display truly diagnostic characteristics; DNA sequencing allows us to determine fungal presence, but not whether this represents a symbiosis. Combining these two complementary methods is essential to fully understand the distribution and diversity of fungal symbiosis in plants, while physiological studies of resource exchange between partners are needed to assess whether the plant-fungus association is functionally mycorrhizal or mycorrhizal-like. Fig. 1 The phylogenetic position and fungal symbiosis status of earlydiverging plant families. Branch lengths have no value and only show how the families are currently considered to be related. Initials in the table denote: M Mucoromycotina, G Glomeromycotina, A Ascomycota, B Basidiomycota. A check indicates presence, a cross absence, and a question mark indicates an unknown identity reported only as 'fungal association'. Checks highlighted in grey are likely accurate reports and were used for occurrence rate estimations, whereas the mutualistic status of un-highlighted checks remains unknown (only relevant for Ascomycota and Basidiomycota symbioses in liverworts). An asterisk indicates a likely incorrect report of symbiosis