New species and records of Trichoderma isolated as mycoparasites and endophytes from cultivated and wild coffee in Africa

A survey for species of the genus Trichoderma occurring as endophytes of Coffea, and as mycoparasites of coffee rusts (Hemileia), was undertaken in Africa; concentrating on Cameroon and Ethiopia. Ninety-four isolates of Trichoderma were obtained during this study: 76 as endophytes of healthy leaves, stems and berries and, 18 directly from colonized rust pustules. A phylogenetic analysis of all isolates used a combination of three genes: translation elongation factor-1α (tef1), rpb2 and cal for selected isolates. GCPSR criteria were used for the recognition of species; supported by morphological and cultural characters. The results reveal a previously unrecorded diversity of Trichoderma species endophytic in both wild and cultivated Coffea, and mycoparasitic on Hemileia rusts. Sixteen species were delimited, including four novel taxa which are described herein: T. botryosum, T. caeruloviride, T. lentissimum and T. pseudopyramidale. Two of these new species, T. botryosum and T. pseudopyramidale, constituted over 60% of the total isolations, predominantly from wild C. arabica in Ethiopian cloud forest. In sharp contrast, not a single isolate of Trichoderma was obtained using the same isolation protocol during a survey of coffee in four Brazilian states, suggesting the existence of a ‘Trichoderma void’ in the endophyte mycobiota of coffee outside of Africa. The potential use of these African Trichoderma isolates in classical biological control, either as endophytic bodyguards—to protect coffee plants from Hemileia vastatrix, the fungus causing coffee leaf rust (CLR)—or to reduce its impact through mycoparasitism, is discussed, with reference to the on-going CLR crisis in Central America.

Species of the ascomycete genus Trichoderma (Hypocreales: Hypocreaceae) are widely distributed in different environments and have a variety of biological activities 1 . In the last two decades various studies have investigated the diversity and taxonomy of Trichoderma and numerous novel species have emerged using DNA sequence data [2][3][4][5] . However, despite the various surveys aimed at covering the diversity of this genus, such studies have been concentrated mostly in Asia, Europe and the Americas [6][7][8][9][10][11][12][13] . In contrast, until now, Africa has been poorly covered in terms of assessing the diversity of Trichoderma, with the exception of some studies involving specific regions or ecological niches, such as soil in South Africa 14 . In the case of Trichoderma occurring as endophytes in Coffea, there is a single study covering species isolated from the rhizosphere of C. arabica in Ethiopia 15 .
Fungi belonging to the genus Trichoderma have a recognized role as decomposers 16,17 and for a long time they were considered to be soil saprotrophs of little practical relevance 18,19 . Currently, it is widely accepted that such a generalization was erroneous and many species of Trichoderma are now recognized as mycoparasites, as well as endophytes of woody plants 18,[20][21][22][23][24] . The endophytic interaction between Trichoderma and their host-plants is intimate and may be complex, involving many steps at each level from direct contact to internal colonization of www.nature.com/scientificreports/ this suggests strong endemism and isolation of the Trichoderma mycobiota of coffee in Kenya, this should be treated with caution since sampling in Kenya was limited and, thus, this may be an artefact. However, the fact that there was no commonality in species between Kenya and the other countries is intriguing and warrants further investigation. The only Brazilian isolate was identified as T. andinense. The occurrence of T. andinense appeared in an ad hoc isolation during a search for mycoparasites of CLR pustules in Brazil. The coffee survey in Brazil-aimed at obtaining endophytic Trichoderma from plants growing in semi-wild conditions at eight localities in four states (Table 3)-yielded isolates comprising a range of different genera but just a single isolate was identified as belonging to the genus Trichoderma.
When comparing the number of taxa of Trichoderma present in C. canephora and C. arabica, it was found that both harbour five species but these are distributed differently in the host tissues. The only Trichoderma spp. occurring in both C. arabica and C. canephora were T. breve and T. botryosum sp. nov. In C. arabica, the highest diversity of Trichoderma was found in berries (four species), whereas in C. canephora the highest diversity was found in stems (four species): however, since most of the sampling was focused on C. arabica, it would be premature to reach a firm conclusion. Stems yielded the highest number of isolates from the aerial tissues of Coffea, with 40 (42.6% of the total), followed by leaves with 29 (30.8%), and seven isolates (7.4%) from berries. Mycoparasites, with 18 isolates, formed 19.2% of the total (Tables 1 and 2).
Taxonomy. Four additions to the genus Trichoderma emerged from the phylogenetic study of the isolates obtained during this survey of Coffea in Africa. Morphological and cultural information proved useful to confirm their separation from closely-related known species of Trichoderma; providing further evidence for their recognition as novel species. Type cultures were deposited in the internationally-recognized culture collection of the Universidade Federal de Viçosa (COAD). The following species were collected and identified, with authority names and publication details as recommended by Bissett et al. 52 .
Trichoderma aggressivum Samuels  Description and illustration, see 5,56 .  The tree was based on tef1 sequence dataset. Bootstrap values (≥ 70%) of the ML analyses, as well as posterior probability scores (≥ 0.9) from a Bayesian analysis of the same dataset, are indicated at well supported nodes together with thickened branches. The isolates belonging to known species, obtained in this study are in bold. Isolates of new species, described in this study, are in bold red. The tree was rooted with Trichoderma asperellum (TR3). The phylogenetic tree was edited using Inkscape 1.0 (https ://inksc ape.org/pt-br/).  61 . This is the first report of T. aggressivum as a mycoparasite of Hemileia and of rusts, in general; having been recorded previously only in mushroom farms in both North America and Europe 5 . Thus, this appears to be the first record of T. aggressivum from the tropics. It has been shown to produce various antifungal compounds 63 , and falls within the clade defined by a mycoparasitic mode of nutrition in a consensus phylogenetic tree 24 . The tree was based on rpb2 sequence dataset. Bootstrap values (≥ 70%) of the ML analyses, as well as posterior probability scores (≥ 0.9) from a Bayesian analysis of the same dataset, are indicated at well supported nodes together with thickened branches. The isolates belonging to known species, obtained in this study, are in bold. Isolates of new species, described in this study, are in bold red. The tree was rooted with Trichoderma asperellum (TR3). The phylogenetic tree was edited using Inkscape 1.0 (https ://inksc ape.org/pt-br/). Notes: This fungus, in the Longibrachiatum clade, was originally described from its sexual morph collected from a log in the Venezuelan Andes 65 . Similar isolates from soil in Saudi Arabia, Amazonian Peru and Hawaii were reported later 64 but were considered to represent new taxa within the T. andinense sub-clade and this species "remains known only from a single collection" 64 . Thus, this is the first report of T. andinense as a mycoparasite of Figure 7. Bayesian phylogenetic tree of clades Longibrachiatum and Viride. The tree was based on rpb2 sequence dataset. Bootstrap values (≥ 70%) of the ML analyses, as well as posterior probability scores (≥ 0.9) from a Bayesian analysis of the same dataset, are indicated at well supported nodes together with thickened branches. The isolates belonging to known species, obtained in this study are in bold. Isolates of new species, described in thes study, are in bold red. The tree was rooted with Protocrea pallida (CBS 121552). The phylogenetic tree was edited using Inkscape 1.0 (https ://inksc ape.org/pt-br/). Notes: This is a cosmopolitan species, in the Viride clade, but it is more commonly isolated from soil in tropical regions 5 and its sexual morph is rarely formed 3 . Chaverri and Samuels 24 show its habitat preference is soil, but in their phylogenetic tree, T. atroviride groups with species having a mycoparasitic mode of nutrition. In our study, it was isolated from a leaf of C. arabica and this is the first record of T. atroviride as an endophyte of coffee. Trichoderma atroviride is more associated with the plant rhizosphere, rather than aerial tissues, and it has been shown recently that colonization of both maize and tomato roots by this fungus induces foliar herbivory resistance 66 . Previously, it has been known to protect plants against root pathogens (Pythium, Rhizoctonia) through induced resistance and antibiosis 67   Notes: Trichoderma botryosum grouped phylogenetically close to T. afarasin and T. endophyticum 4 in the Harzianum clade. The new species is morphologically similar to its close relatives in: the pyramidal-type conidiophores; size of conidia and ampulliform phialides; and growth rate on PDA at 25 °C. The most distinctive morphological features in this species are the presence of chlamydospores and the grape-like clusters of conidia.  Notes: Trichoderma breve is similar morphologically to the T. harzianum complex, and has previously been isolated from soil in the north of China 12 . Phylogenetic analyses indicate that T. breve is closely related to T. bannaense 12 , and in our study it lies close to T. lentiforme. This is the first report of T. breve in Africa and also as an endophyte of coffee; being recorded from both C. canephora and C. arabica in Cameroon and Ethiopia, respectively, with a single isolate from each country ( Conidiophores pyramidal with verticillate, paired lateral branches; phialides generally formed on terminal branches, in divergent whorls of three to four, ( Notes: Phylogenetic analyses placed T. caeruloviride close to T. amazonicum 8 and T. pleuroticola 68 in the Harzianum clade. The new species can be distinguished from its nearest relatives by: no growth at 35 °C; the presence of a coconut-like odour on PDA; and the blue-green conidia in SNA microculture. Morphologically, T. caeruloviride is distinct from T. amazonicum which has a branching pattern of the pachybasium type, elliptical to subglobose conidia, ampulliform phialides, and chlamydospore-like structures in clusters. Trichoderma caeruloviride shares some morphological characteristics with T. pleuroticola, such as the pyramidal-type branching pattern, the globose conidia and the formation of chlamydospores; but can be separated by the larger, lageniform phialides of T. caeruloviride. Both isolates were from berries of C. arabica in Ethiopia and it appears to be a rare species (Table 2). Notes: Trichoderma guizhouense, in the Harzianum clade, was first isolated from soil in Guizhou Province of China 55 , and has since shown promise as a biocontrol agent of Rhizoctonia root rot 69 . It was also isolated as an endophyte from the woody liana, Ancistrocladus korupensis (Ancistrocladaceae)-extracts of which are active against HIV 70    Notes: Trichoderma hamatum, in the Viride clade, is a cosmopolitan species, originally isolated from soil but it has also been reported as an endophyte in both stems and pods of wild Theobroma gileri from sub-montane forest in western Ecuador 20 . In addition, it was identified as a mycoparasite of frosty pod disease (Moniliophthora roreri) on the same host in this ecosystem 20 . It has also been isolated from the rhizosphere of C. arabica in Ethiopia 15 . In our study, T. hamatum was obtained from stems and berries of C. arabica in both cultivated and wild coffee plants, also in Ethiopia. It is most frequently cited as a colonizer of the rhizosphere, and some soil strains have been shown to promote crop growth, to activate biocontrol mechanisms against root pathogens and to induce systemic resistance to foliar pathogens 71 . Previously, an endophyte strain from the pod of a wild Theobroma species was found to promote the growth and delay drought symptoms in cacao plants 72 . Notes: Trichoderma koningiopsis, in the Viride clade, is a cosmopolitan species, but it is more frequently recorded in tropical rather than temperate regions, and mostly from soil. During a survey of Trichoderma diversity in soil and leaf litter from the Amazonian rainforest of Colombia, T. koningiopsis was amongst the commonest species isolated 73 . In the Atlantic rainforest of Brazil, T. koningiopsis was found to be the dominant Trichoderma species in leaves being carried by Atta leaf-cutting ants, and subsequently rejected by them from the nest: it was posited that the ants recognized the threat posed by this mycoparasite to the fungal garden 74 . It has also been reported as a common stem endophyte in a species of Theobroma in sub-montane forest in western Ecuador 22 . It has also been shown to colonize cacao plants via the leaf trichomes 75 . An isolate of Trichoderma koningiopsis from the stem of a Vinca species in Iran was found to produce a range of anti-microbials, including trichodermin, as well as cytotoxic compounds 76 . Trichoderma koningiopsis has also previously been isolated from the rhizosphere of C. arabica in Ethiopia 15 . Here, it is reported for the first time as an endophyte of Coffea: all seven isolates being recovered from the leaves and stem of both cultivated and wild C. canephora in Cameroon (Table 2), where it appears to be common. Notes: Trichoderma lentissimum is phylogenetically close to T. gamsii 77 and T. lieckfeldtiae 38 in the Viride clade. The new species is also morphologically similar to T. gamsii in: the branching pattern (pyramidal type); the lageniform phialides; and the formation of chlamydospores; but can be separated based on the smaller phialides and conidia of T. lentissimum as compared with those of T. gamsii. The most prominent differences between T. lentissimum and T. lieckfeldtiae, is the pachybasium-type branching pattern which is found only in T. lieckfeldtiae and the absence of chlamydospores in the latter. This species appears to be rare and it was only isolated once during the survey, as a stem endophyte in Kenya. Notes: Trichoderma parareesei, in the Longibrachiatum clade, was originally isolated from soil of a subtropical rainforest near Iguazu Falls, Argentina, and is reported to have a pantropical distribution in both rainforest and agricultural soils 59,64 . During a survey of Trichoderma species in the sapwood and dead branches of cacao trees in south-eastern Brazil, it was exclusively isolated from dead wood-where it was the dominant species-and was never recorded from the sapwood 79 . In our study, T. parareesei was isolated as an endophyte from stems of C. arabica and also as a mycoparasite of Hemileia cf. coffeicola. This species has never been reported before either as an endophyte in coffee or as a mycoparasite, and probably this is the first record as an endophyte, in general (see 79 ). Trichoderma parareesei was described as a sympatric, clonal, agamospecies (reproducing only asexually) closely related to T. reesei and is its likely ancestor 59,78 . The latter is a critically important species in the biotechnology industry as a producer of cellulases and hemicellulases and-because of its ability to express recombinant proteins-it is now being targeted for a role in the production of biofuels 80 . Many of these industrial strains have been shown to be T. parareesei by Druzhinina et al. 78 . They also found that this species is strongly mycoparasitic, compared to T. reesei, showing significant antagonism to a range of aerial plant pathogens in dual-culture tests. It has also been demonstrated that T. parareesei has biocontrol potential against both fungal and oomycete plant pathogens and, moreover, that it enhances root development and promotes growth, in general, of tomato plants 81 . Finally, it was posited that the ecological niche of T. parareesei is not soil but the canopy of tropical forest 78 . Our results confirm their supposition since this species was recorded in stems of wild to semi-wild C. arabica, as well www.nature.com/scientificreports/ as being found as a mycoparasite of Hemileia rust in the canopy of wild coffee trees in the understorey of cloud forest: all isolates from Ethiopia (Tables 1 and 2 Notes: With the exception of the soil isolate DAOM 165782 (North Carolina), T. petersenii was known previously only from ascospore isolations 60 . However, it has since shown to be common on woody hosts in Southern Europe, especially in Spain, and is reported to be ubiquitous on the Canary Islands, being found in the sexual stage on stromata of the Xylariaceae 3 , presumably as a mycoparasite. In our study, T. petersenii was found for the first time in Africa, and also this is the first report of it as a rust mycoparasite, specifically, on a Hemileia species close to H. coffeicola ( Table 1). The latter species, however, is a pathogen of C. canephora in the lowland tropics of West Africa and thus this high-altitude rust of wild Arabica coffee in Ethiopia is considered to be undescribed. Notes: Most of the species found in this study could be identified with high support using the combination of tef1 and rpb2 genes. However, for the isolates assigned to T. pseudopyramidale, it was also necessary to include the calmodulin gene in the analysis in order to separate the novel species from the closely related T. pyramidale. Trichoderma pseudopyramidale grouped phylogenetically close to T. pyramidale 4 in the Harzianum clade. The two species share several characteristics in common, such as pyramidal conidiophores, a similar growth rate on PDA and SNA, and the formation of amorphous pustules with white-yellow borders. Trichoderma pyramidale has larger phialides and conidia, compared to T. pseudopyramidale, whilst the latter species forms chlamydospores on CMD and produces a yellow reverse on PDA; and, unlike T. pyramidale, it is able to grow at 35 °C. Trichoderma pseudopyramidale forms two monophyletic subclades, one containing endophytic isolates and another including isolates obtained directly from CLR pustules (as mycoparasites). Since both subclades come from a phylogenetically well-supported clade by ML, MP and BI, we decided to keep them in a single species and consider them to represent an infra-specific grouping not warranting taxonomic recognition at this stage. No significant differences in morphology or in the growth rates for isolates belonging to these subclades were found. In our study, T. pseudopyramidale represented by far the commonest isolate (35 isolates), the greater majority as an endophyte in both stems and leaves of C. arabica, as well as a mycoparasite of Hemileia cf. coffeicola on wild Arabica coffee in Ethiopia (Tables 1 and 2). There was a single record from Cameroon, as a mycoparasite of Hemileia coffeicola on C. canephora.
Trichoderma spirale Bissett- Notes: Trichoderma spirale, in the Strictipile clade, is a cosmopolitan species. It was isolated for the first time from soil in Thailand, but has since found to be common in sapwood of cacao and other Theobroma spp. 79,82 . It was also the dominant Trichoderma species isolated from roots of Pinus densiflora in South Korea and it was suggested that Trichoderma may be playing a role in stimulating plant colonization by the ectomycorrhizal fungus Tricholoma matsutake 83 . In our study, it was found for the first time as an endophyte of coffee; all three isolates were from stems of wild C. canephora in Cameroon rainforest of the Congo basin. Notes: Trichoderma theobromicola, in the Viride clade, was found for the first time growing as an endophyte in the trunk of wild cacao in Amazonian Peru and, in subsequent greenhouse studies, it demonstrated promise as an antagonist against frosty pod disease (Moniliophthora roreri) after being inoculated into and re-isolated from seedlings of cacao 22 . Similarly, an isolate of T. theobromicola from Cola sp. in the Cameroon 38 revealed biocontrol potential after it was shown to be parasitic on and reduced the disease incidence of Phytophthora in Capsicum annuum 84 . All the isolates in our study were from stems of wild C. canephora in Cameroon rainforest (Table 1). This is the first report of this species as an endophyte of coffee. Notes: Trichoderma virens is a cosmopolitan species, commonly isolated from soil samples, but its sexual morph appears to be rare; having been found only once on dead wood 5 . In our study, it was isolated from the stem of a wild species of Coffea in Cameroonian rainforest on a single occasion. This is the first record of T. virens as an endophyte of coffee and, seemingly, as an endophyte of aerial plant tissues. Previously, it has been shown to colonize sugar-cane roots; forming dense mycelium in the intercellular spaces 85 . It was also reported that T. virens secretes proteins to facilitate colonization of maize roots in which plant-host immune responses are suppressed 86 . Earlier, it was demonstrated that T. virens promotes growth of Arabidopsis by stimulating the www.nature.com/scientificreports/ root system through an auxin-dependent mechanism 87 . The isolate from our study may have additional mechanisms to colonize woody stems and, perhaps, to form a similar beneficial interaction with its wild coffee host.

Discussion
Previous studies have investigated the diversity of endophytic fungi associated with coffee [88][89][90][91][92] , but these were based on surveys restricted to the Americas and Hawaii, where coffee is an exotic introduced species. The endophytic mycobiota found in these studies is dominated by genera such as Colletotrichum, Fusarium, Penicillium, Pestalotia and Xylaria. Such assemblages consist mainly of opportunistic endophytes-seemingly, of little biological significance to their hosts 93 -with Trichoderma appearing only infrequently. Only one study involved sampling of all the coffee tissues (leaf, berry, stem, root system) 89 and, of the 843 isolates obtained, only four were identified as belonging to the genus Trichoderma. Conversely, and in sharp contrast, the Coffea samples from Africa in this study yielded 76 endophytic isolates of Trichoderma from the aerial plant tissues of a relatively small sample size, with a highly diverse taxonomic range, including four new species. At this stage, it is not possible to determine whether the new taxa described herein are geographically restricted to Africa or even to coffee. Nevertheless, we find it significant that a far richer diversity of Trichoderma was found in association with coffee in its African centre of origin compared to that elsewhere, especially in the Neotropics. We also find particularly relevant the complete absence of endophitic Trichoderma species isolates amongst the plethora of fungal isolates obtained from our sampling in semi-wild situations in Brazil. This was entirely unexpected and may indicate the existence of a 'Trichoderma void' in the coffee endophyte mycobiota outside of Africa. The occurrence of Trichoderma in association with C. arabica has been reported previously in Ethiopia 15,94 , but these studies focused on strains isolated from the rhizosphere and root tissues. The isolates included: T. harzianum sensu lato, T. hamatum, T. asperelloides, T. spirale, T. atroviride, T. koningiopsis, T. gamsii and T. longibrachiatum. Only three of these taxa were isolated during our study of stems, leaves and berries: namely, T. hamatum, T. spirale and T. koningiopsis. These are cosmopolitan species that are frequently isolated from tropical habitats, especially from soil 6,7,60,95 . Certain Trichoderma species were isolated from more than one plant tissue type: T. koningiopsis and T. spirale from the leaves and stems of C. canephora; T. hamatum from the stems and berries of C. arabica. T. hamatum, T. koningiopsis and T. spirale have also been reported as endophytes in other tropical woody plants, notably cacao and rubber 5 . Nevertheless, only T. hamatum had previously been reported as endophytic in C. arabica; occurring in the root system 15,94 . Apart from the four novel species described here, other Trichoderma species were found for the first time as endophytes in coffee: T. atroviride, T. guizhouense, T. breve and T. theobromicola. These species were known from other habitats, such as: tropical soils; decaying wood and bark; as mycoparasites; on mushroom compost; in leaf-cutting ant colonies; and as endophytes in Theobroma spp. (Malvaceae) 4,5,20,96,97 . Trichoderma guizhouense has a worldwide distribution, mainly in soil, and had only been reported previously as an endophyte in the endemic woody liana, Ancistrocladus korupensis, and in the stems of Cola trees in primary forest in south-west Cameroon 4 . Previously, T. theobromicola was known only from South America, and reported to be a common endophyte in sapwood of cacao 4,22,79 whilst T. breve, a recently described species isolated from soil, was previously known only from northern China 12 . These two species are new geographical and host records for Africa, but this may simply reflect the poor sampling of Trichoderma in the region, particularly for endophytes. The results of the surveys also suggest that many species of Trichoderma are either cosmopolitan or pantropical.
Mycoparasitism-the ecological relationship where one fungus parasitizes another 98 -has now been reported for a number of species of Trichoderma, notably: T. atroviride, T. hamatum, T. longibrachiatum, T. reesei and T. virens, and it has recently been established that mycoparasitism is an ancestral trait of the genus 19,99 Mycoparasitic Trichoderma spp. have a wide range of hosts, including true fungi, such as Botrytis cinerea, Rhizoctonia solani, Alternaria alternata and Fusarium spp., as well as Oomycetes, such as Pythium ultimum 49,98 . However, the species found as mycoparasites of H. vastatrix reported here-T. aggressivum, T. andinense, T. parareesei, T. petersenii and T. pseudopyramidale-are the first in the genus to be reported attacking the Hemileia rusts associated with coffee. Three of the species of Trichoderma obtained during the surveys are well-known mycoparasites, but were found here only growing as endophytes in coffee, namely: T. atroviride, T. hamatum and T. virens. Trichoderma pseudopyramidale may deserve special attention as a potential biocontrol agent of CLR, since it was the most common mycoparasitic species obtained from both Cameroon and Ethiopia (77.8% of total mycoparasites). In Ethiopia this species was commonly associated with a purported new species of Hemileia (cf. H. coffeicola) on wild C. arabica in cloud forest (ca. 2000 m). It was also frequently isolated as an endophyte from the leaves and stems of both semi-wild and wild C. arabica in Ethiopia (see Table 1). It may encompass dual roles as an endophytic bodyguard of coffee and also as a contact mycoparasite of CLR.
Mycoparasitic fungi associated with coffee rust have been studied n regions of the world where coffee is not a native species, such as in Mexico 100 . It is interesting to note that this Mexican survey identified six purported mycoparasites: Acremonium byssoides, Calcarisporium ovalisporum, C. arbuscula, Fusarium pallidoroseum, Sporothrix guttuliformis and Verticillium (= Akanthomyces lecanii). A more recent publication reporting the results of an investigation in Mexico and Puerto Rico, involving the use of single-molecule DNA sequencing of fungal rRNA gene barcodes to identify putative mycoparasites in pustules of H. vastatrix, yielded 15 fungal taxa associated with CLR, none of which belonged to Trichoderma 101 . Information on the ecology of the new Trichoderma species described here, and their role in nature, is limited because relatively few strains of each species were isolated during the survey; the exceptions being T. botryosum and T. pseudopyramidale, which constituted over 60% of the total isolations and seem to have a close association with their Coffea hosts, in both Cameroon and Ethiopia, in wild, semi-wild and cultivated situations.
The aim of the present study was to collect and catalogue endophytes of Coffea species-as well as the mycoparasites of the associated Hemileia rusts-in their African centres of origin, as part of a project to screen and www.nature.com/scientificreports/ assess these isolates as potential biological control agents of CLR. The target area is Central America where the rust has become a critical constraint to coffee production, as well as causing a socio-economic crisis, over the past decade 53,102,103 . The work presented here covers only the taxonomy with some observations on the ecology of the Trichoderma isolates resulting from the surveys in Africa, but these data will be pivotal for selecting candidate biocontrol agents for the potential management of H. vastatrix in the Americas. The philosophy behind the overall project is based on the concepts of classical biological control and, in the case of CLR, on the Enemy Release Hypothesis which posits that exotic species become invasive and achieve pest status because of increased fitness in the absence of their coevolved natural enemies 104 . One solution to address the problem of invasive alien pests is to source, import and release coevolved natural enemies from the centres or regions of origin of the target species in order to reduce 'pest' fitness: the classical biological control strategy. This approach using fungal natural enemies, such as entomopathogens and plant pathogens, has been employed successfully to control invasive alien arthropod pests and weeds 105,106 , but never against alien plant diseases using mycoparasites. The evidence from our study indicates that there is a guild of Trichoderma species, potentially antagonistic to H. vastatrix in Africa, which could be exploited for biological control of CLR in Central America following the classical approach. There are claims that non-specific, indigenous mycoparasites; notably, Lecanicillium lecanii-now Akanthomyces lecanii 107 -can reduce the impact of CLR in the Americas 108 , but this is not evident based on the continuing rust outbreaks.
Another scenario has been suggested to further explain the invasiveness of alien plant species: the Endophyte-Enemy Release Hypothesis 109 , which posits that alien plants arrive not only without their coevolved natural enemies but also deficient in, or completely lacking, coevolved endophytes, some of which may be acting as symbionts ('bodyguards'); protecting their hosts against adverse abiotic and biotic factors. Thus, in their absence, exotic crops thrive and alien weeds invade, with no natural enemies reducing plant fitness and fecundity and no bodyguards to 'pay' for protection. In crop species, the consequences can be catastrophic when coevolved natural enemies-lacking their own natural enemies, such as mycoparasites (in the case of fungal pathogens)eventually catch up with their endophyte-deficient plant hosts. Such may be the case with H. vastatrix in Central America-and, of course, this may explain the devastating rust epiphytotics that destroyed coffee cultivation in Sri Lanka (Ceylon) in the nineteenth century, as well as in all the global regions where the rust has invaded 110 .
Thus, the ideal classical biological control agent for CLR would combine the best of both worlds in the form of an endophytic mycoparasite, and-as our results indicate-the genus Trichoderma contains such candidates. Potentially, these would not only be used to colonize the coffee leaf and parasitize the external rust pustulesas well as to target the invasive, intercellular mycelium of the rust-but also to bolster host defences through induced resistance 66,84,111 . There is increasing evidence that, in addition to induced resistance to diseases and pests, endophytic Trichoderma species confer a range of other benefits to their plant host, in particular, drought tolerance, resistance to abiotic factors such as salt stress and growth stimulation 5,21,23,27,28,30,32,36,43,71,72,81,[85][86][87]112 .
Preliminary data, using Trichoderma isolates from the survey, are showing positive results in the laboratory with evidence of reduction in rust disease severity 113 (Authors, unpublished). Greenhouse screening of four isolates of Trichoderma (COAD 2418, COAD 2417, COAD 2535 and COAD 2439), belonging to T. hamatum and T. pseudopyramidale sp. nov.., showed their ability to inhibit the germination of H. vastatrix urediniospores above 70% in vitro. Isolate COAD 2396 (T. atroviride) reduced the severity of the disease to less than 50% of the levels observed in the controls when applied before or simultaneously with H. vastatrix on coffee leaf discs. In addition, an isolate of T. parareesei (COAD 2482) promoted the growth and increased the biomass of tomato roots by 33% and 57%, respectively; whilst others are now showing the ability to increase drought tolerance 113 (Authors,unpublished).
The methodology employed during the survey for the isolation of endophytes has proven to be robust. It has been emphasized previously that endophyte isolation is a method-dependent process and this will determine the quality and quantity of fungi obtained 114 . In our experience, isolating in situ-directly in the field from tree stems-or immediately after collection, eliminates or reduces contamination by many of the opportunistic endophytes and favours the slower-growing, potentially obligate endophytes. This has consistently been demonstrated not only during the present coffee survey in Africa, but also from previous surveys of wild species of Theobroma and Hevea in South American rainforests where this approach was pioneered 20,36,39 These surveys resulted not only in the discovery of numerous new Trichoderma taxa-which are still being described 4,41 -but in many other taxonomic novelties, including new endophytic lineages of Tolypocladium and a new class of Pezizomycotina 115,116 . Moreover, they reveal the paucity of endophytes in cultivated exotic plants-in this instance, cacao and rubber-compared to wild populations of Theobroma and Hevea in natural ecosystems 39,41,[117][118][119] . This has been confirmed during the present study, when a survey of coffee endophytes in four states of Brazil, failed to isolate any species of Trichoderma, providing compelling evidence that centres of origins or diversity of plants harbour unique guilds of endophytic Trichoderma species-as well as other genera-that could be exploited not only for classical biological control but also as potential reservoirs of novel metabolites.
In conclusion, our surveys in Africa for endophytes and mycoparasites associated with the genus Coffea and with its Hemileia rusts have revealed a highly diverse range of fungi, with many novel species; Trichoderma being just one component. Because of the relatively few countries (3) and localities (18) visited, and the restricted number of host plants sampled, this can only be viewed as a snapshot of the actual diversity of endophytes, as well as of mycoparasites, associated with Coffea in Africa, especially in forest ecosystems. Potentially, in Madagascar, where the diversity of the genus is richer with 59 confirmed species 120 , this still-untapped diversity could be even higher. Loss of forest habitats in Africa and Madagascar means that many of these fungi will go extinct, along with their host plants, before being described. The potential loss of such key antagonists of the CLR fungus-as well as of Coffea germplasm-should be cause for concern to coffee stakeholders.  (Figs. 13a,b, 14a-c, 15a,b). At each selected site, Coffea plants were examined for rust pustules-with particular attention to collecting rust colonies exhibiting mycoparasitism, or appearing to be abnormal (unusual colour, poor sporulation) (Figs. 14d, 15c,d). Specimens were dried in a plant press for later processing in the laboratory (preliminary identification and isolation). Also, at each site, samples of at least three separate adult plants were collected, consisting of healthy leaves, berries and 3-cm diam or thicker stem sections of each individual, and bagged for examination and processing later the same day. Isolations were made from healthy leaves, stems and berries of C. arabica, C. brevipes, C. canephora and C. eugenioides (Figs. 13c-f, 15e). The isolation protocol followed the procedure described by Evans et al. 20 with modifications, and were performed as described below.
1. Stems in situ were thoroughly rubbed with cotton wool soaked in 70% alcohol and, after the alcohol had evaporated, the bark was removed using a flamed knife or machete blade (Fig. 13a). The exposed panel was then cleaned with a scalpel (Swann Morton 10) and the surface further pared with a smaller blade (Swann Morton 11). Nine, triangular slivers of sapwood (ca. 8 × 5 mm) were excised with a scalpel (Swann Morton 10A) from the panel and transferred individually with fine forceps to three plastic Petri plates (3-or 5-cm diam; 3 samples/plate), containing selective media: potato dextrose agar (PDA), one-fifth strength (20% PDA), supplemented with 10 mg/l penicillin-streptomycin solution. These were sealed immediately with electrical tape and stored in plastic boxes. During these procedures, all instruments were surface sterilized in 90% ethanol and flamed using a portable, alcohol burner. On arrival at the laboratory, the plates were transferred to a 25 °C incubator and examined regularly over an 8-week period. Hyphal tips or spores were excised or picked from colonies as they appeared on or around the wood samples and transferred to 5-cm diam, plastic Petri plates containing 20% PDA or potato carrot agar (PCA) and incubated under black light at 25 °C to promote sporulation. This procedure was firstly described for the isolation of endophytic fungi by Evans et al. 20 but was applied here for the first time for endophytic fungi from coffee. 2. Young mature healthy leaves (third from the branch tip) were thoroughly rubbed with cotton wool soaked in 70% alcohol and, after the alcohol had evaporated, three small (ca. 5 × 5 mm) square fragments were excised from the leaf centre (including the midrib) and were surface sterilized for 3 min by immersion in 10% bleach, followed by immersion in sterile water in stoppered plastic tubes and, following a thorough agitation, were plated as described for stems. Subsequent processing was as described for stems. 3. Whenever available, Coffea berries were also sampled and treated similarly as described above for leaf samples but, after surface cleaning with alcohol, each fruit was skinned and inner parts were divided into three slices which were then surface sterilized before plating. Further steps followed the same procedure as described above.
For the isolation of mycoparasites, conidia from parasitized rust pustules were selected and picked-off with a sterile needle, using a dissecting microscope, and transferred to PDA plates. The dried samples were processed within 2 weeks of collection after transport to the laboratory in the UK or Brazil.
The same endophyte isolation protocol described above was utilized for samples collected at eight localities in four Brazilian states (Espírito Santo, Minas Gerais, São Paulo and Rio de Janeiro). Survey sites closest to those where coffee was sampled in Africa were selected; concentrating on those where coffee plants were growing in semi-wild or forest situations, such as abandoned coffee farms and invasive populations in Atlantic rainforest (Fig. 16).
DNA extraction, PCR amplification and sequencing. Strains were grown in 3-cm diam plates containing 5 mL of potato dextrose broth (PD) at 25 °C in the dark for 4-5 days. DNA was extracted from the mycelium grown on the surface of the broth. DNA was extracted with the Wizard Genomic DNA Purification kit (Promega, Madison, EUA) by following the manufacturer's instructions. The fragments rpb2 (primers fRPB27cR-RPB25F2) 121 and tef1 (primers EF2-EF1728M) were amplified for all isolates and additionally cal (primers CAL228-CAL737) 122 was amplified for a subset of 12 isolates.
The polymerase chain reaction (PCR) amplifications were performed in a total reaction volume of 12.5 μl, including 0. 25  www.nature.com/scientificreports/ Phylogenetic analysis. Consensus sequences were assembled from forward and reverse sequencing chromatograms using SeqAssem 123 tef1, rpb2 and cal contigs of all strains were compared to homologous sequences deposited in NCBI GenBank. Sequences generated in the present study were deposited in the NCBI GenBank database (Table 1) and sequences obtained in other studies were used in our phylogenetic analyses and were retrieved from the NCBI GenBank database (Supplementary Table S1) T. Sequence alignments were performed using MUSCLE implemented in MEGA 10 124 . In total, the dataset comprised 324 partial tef1 (sequences 664 pb); 169 partial rpb2 sequences (951 pb) and 25 partial cal sequences (443 pb). Two concatenated trees with tef1 and rpb2 sequences were created, one with taxa of the clades Harzianum (more numerous), Stricpile and Virens, and the other with the rest of the taxa (Figs. 1 and 2); a third concatenated analysis with partial sequences of three genes, tef, rpb2 and cal, was constructed with a subgroup of sequences to clarify thephylogenetic relationships of www.nature.com/scientificreports/ some species within the clade Harzianum (Fig. 3), such trees containing 168 taxa with 2515 characters, 86 taxa with 2422 characters and 25 taxa with 1927 characters, respectively. The concatenated alignments were generated in Sequence matrix v1.8 125 . Single-gene trees were also generated. Maximum parsimony (MP), Maximum likelihood (ML) and Bayesian Inference (BI) were performed for the concatenated and single-gene trees. Prior to phylogenetic analyses, the most appropriate nucleotide substitution model for each locus was selected using MRMODELTEST v.2 126 . Nucleotide substitution models in the two-gene concatenated trees were HKY + I + G and SYM + I + G (Figs. 1 and 2), for tef1 and rpb2, respectively. For the three-gene concatenated tree, the models were HKY + I, K80 + I and K80 + G (Fig. 3) for tef, rpb2 and cal, respectively. For all trees the BI and ML analysis were estimated in the CIPRES Science Gateway Platform using Mr. Bayes 3.2.6 and RaxML-HPC v.8, respectively 127,128 and MP in MEGA 10. Phylogenetic trees were visualized using FigTree (http://tree.bio.ed.ac. uk/softw are/figtr ee/). Phylogenetic species were recognized based on two main previously accepted criteria 129 Genealogical Concordance (the clade was present in the majority of the single-locus genealogies, as revealed by a majority-rule consensus tree) and Genealogical Non-discordance (the clade was well supported in the least one single-locus genealogy, as judged both by MP and BI and was not contradicted in any other single-locus genealogy at the same level of support).

Morphological characteristics.
The results of the phylogenetic analysis of the assemblage of Trichoderma isolates guided the selection of isolates to be included in the morphological analysis and characterization of novel taxa. One or two isolates of each new taxon were examined. Procedures for morphological observation of Trichoderma followed the protocol established by Samuels and Hebbar 5 . Macroscopic characteristics of colonies-mycelium colour, radial growth, presence/absence of concentric rings, sporulation "pustules" 62 , pigmentation and presence/absence of odour-were evaluated on PDA, CMD (Corn-meal Agar) and SNA (Synthetic Nutrient Deficient Agar) after 7 days at 25 °C under a 12-h daily light regime (light provided by two white and one near-UV lamps placed 35 cm above the plates). Rates of growth were evaluated at 72 and 96 h on the three culture media at 25, 30 and 35 °C in the dark. Observations of fungal structures were made using an Olympus BX 51 microscope and were based on slide cultures prepared with colonies of each isolate growing from PDA and CMD blocks 130 . After 4-5 days of growth at 25 °C under the same light regime described above, the slides were mounted in 3% KOH for observation and illustration. Descriptions included biometric data of phialides, conidia and chlamydospores. Measurements were taken from images generated with a digital camera Olympus Q-Color 3.  Informed consent. Informed consent was obtained to publish the names/information/images of all study participants appearing in the publication.
Ethical approval. All experimental protocols were approved by a named institutional and/or licensing committee/s. All methods were carried out in accordance with relevant guidelines and regulations.