Diversity of xerotolerant and xerophilic fungi in honey

Fungi can colonize most of the substrata on Earth. Honey, a sugary food produced by bees (and other insects) has been studied little in terms of its fungal diversity. We have surveyed and evaluated the presence of xerotolerant and xerophilic fungi in a set of honey bee samples collected from across Spain. From 84 samples, a total of 104 fungal strains were isolated, and morphologically and phylogenetically characterized. We identified 32 species distributed across 16 genera, most of them belonging to the ascomycetous genera Aspergillus, Bettsia, Candida, Eremascus, Monascus, Oidiodendron, Penicillium, Skoua, Talaromyces and Zygosaccharomyces. As a result of this survey, eight new taxa are proposed: i.e. the new family Helicoarthrosporaceae, two new genera, Helicoarthrosporum and Strongyloarthrosporum in Onygenales; three new species of Eurotiales, Talaromyces affinitatimellis, T. basipetosporus, and T. brunneosporus; and two new species of Myxotrichaceae, Oidiodendron mellicola, and Skoua asexualis.


INTRODUCTION
Honey is a natural sweetener produced by honey bees (insects of the genus Apis of the order Hymenoptera) from nectar (blossom honey or nectar honey) or from carbohydrate-rich secretions of living green parts of plants or excretions of plant-sucking phytophagous aphids (insects of the family Aphidida, order Hemiptera) (honeydew honey) after combination with the bee's specific substances, placement, dehydration, and storage in the honey comb to ripen and mature. Honey is mostly composed of monosaccharides (dextrose and fructose), at a concentration of not lower than 60% and a much lesser amount of oligosaccharides, organic acids, enzymes (amylases and α-glucosidase) and solid particles. Due to its particular physicochemical nature and biological origin, honey should be an ideal substratum for the development of xerotolerant and xerophilic fungi. However, little information has been gathered about these fungi and their relationships with honey and honey products. Nonetheless, most of the fungal species from honey had been reported as new for science.

Fungal isolation
A total of 83 samples of honeydew and blossom (nectar) honey from different locations in Spain (Fig. 1), and one from Argentina (San Martín, Buenos Aires province), have been processed. All samples were of the harvest in 2014, stored in settling tanks, and after a variable period of time clarified by filtration (with one exception, which was by centrifugation). Seventy-two of the Spanish samples corresponded to honeydew honeys, 45 from trading companies and 27 collected and processed by beekeepers. A few of the samples provided by commercial companies were categorized (according to the nature of the honeydew) as oak, holm oak and forest honey. The 11 samples of blossom honey were provided by beekeepers, and these were classified as multifloral. All samples provided by commercial companies were subjected to a thermal treatment, subjecting the honey at 45-55°C for a few hours up to 2 days, or pasteurized (2 min at 80°C). The samples provided by beekeepers have not undergone any heat treatment. For each sample, 10 g of honey was dissolved into 90 mL of sterile water in a sterile disposable plastic container, and 1 mL of such dilution (1:10) was aseptically plated onto two 90 mm diam. plastic Petri dishes and mixed with 15 mL of molten (at 50-55°C) 18% glycerol agar (G18; DG18 [Hocking & Pitt 1980] without dichloran: 5 g peptone, 10 g dextrose, 1 g KH 2 PO 4 , 0.5 g MgSO 4 ·7H 2 O, 15 g agar-agar, 110 g glycerol, 1 L tap water, and supplemented with 250 mg/L of L-chloramphenicol). Once the medium had solidified, one of the Petri dishes was incubated in darkness at 15°C and the other at 25°C for up to 2 months. The colonies developed were examined under a stereomicroscope. Fungal structures from selected (representative of all morphological variety) colonies were transferred to 50 mm diam. Petri dishes containing G18 by using a sterile insulin-type needle and incubated in the same conditions to obtain pure cultures.

Phylogenetic analysis
A preliminary molecular identification of the isolates was carried out with LSU sequences using Basic Local Alignment Search Tool (BLAST; https://blast.ncbi. nlm.nih.gov/Blast.cgi) and only the type sequences or reliable reference strains from GenBank were considered for identification, and a maximum level of identity (MLI) of ≥98% was used for identification at the rank of species and < 98% at the rank of genus. BenA for to the genera Aspergillus, Penicillium, and Talaromyces, and ITS for the genera Monascus, Oidiodendron and Skoua were used for identification at the rank of species. An LSU tree was built to determine the phylogenetic relationships of all our isolates. Phylogenetic trees of ITS and a combination of ITS-BenA-CaM-rpb2 were also built to distinguish the members of Myxotrichaceae and the genus Talaro

Growth at different water activities (a w )
To test the capacity of growth in different water activities, media containing malt extract (1% w/w), yeast extract (0.25% w/w) and agar-agar (1% w/w) at pH 5.3 were adjusted at six different a w (0.97, 0.95, 0.93, 0.92, 0.88 and 0.82) by adding equal weights of fructose and glucose (corresponding to 22, 30, 40, 44, 48, and 55% w/ w of sugars, respectively) (Pitt & Hocking 1977). Water activity was measured in duplicate by a water activity meter (Aqualab, Decagon Devices CX3 02734) with an accuracy of ±0.002 at 25°C. Triplicate plates were inoculated at their centre with 5 μL of spore suspension of selected fungi, and incubated at 25°C in darkness, with the exception of FMR 15880, FMR 15883 and FMR 16031, which were at 15°C (because of their poor growth at 25°C). The colony diam. was measured after 21 days.

Fungal diversity
All honey samples produced fungal colonies on G18 at 15°C as well as at 25°C. Table 1 summarizes the fungal strains identified phenotypically and molecularly. With

Molecular phylogeny
Our first phylogenetic study included 206 LSU sequences with a total of 606 characters, including gaps, 352 of them being parsimony informative. The ML analysis was congruent with that obtained in the BI analysis, both displaying trees with similar topologies. The isolates were distributed across two main clades ( Fig. 2a-c), the first (100% BS / 1 PP) corresponding to the Ascomycota and including 99 isolates, and the second (100% BS / 1 PP) involving the rest of the isolates and pertaining to the Mucoromycota. The first main clade was divided into six subclades: A (82% BS / 1 PP), which represents Onygenales; B (75% BS / 0.96 PP), Eurotiales; C (100% BS / 1 PP); Pleosporales, D (unssuported) as incertae sedis; E (100% BS / 1 PP), Schizosaccharomycetales, and F (94% BS / -PP), Saccharomycetales. Subclade A contains seven well-supported groups, six of which represent the known families of Onygenales, i.e. Gymnoascaceae (A1), Arthrodermataceae (A3), Nannizziopsiaceae (A4), Eremascaceae (A7), Ascosphaeriaceae (A8), and Spiromastigaceae (A9), and a seventh group (A5) composed of five of our strains probably representing a new family. The groups representing Ajellomycetaceae (A6) and Onygenaceae (A2) were unsupported. Strains in subclade A were distributed as follows: the five mentioned above into A5, FMR 16121 into a separate branch of the Ajellomycetaceae (A6), four strains conspecific with Eremascus albus (A7), and one (FMR 16318) identified as Ascosphaera atra (A8). Thirty-nine strains were placed in Eurotiales (Subclade B). One (FMR 16566) was placed together with Talaromyces flavus and T. kabodanensis in an unsupported branch, and 16 strains near to T. minioluteus into a well-supported sister clade (B1). Into B2 (unsupported), which includes species of Aspergillus, eight of the strains were placed in a branch (99% BS / 1 PP) together with A. glaucus, A. montevidensis and A. pseudoglaucus (sect. Aspergillus). For the final identification of these eight strains, we used BenA sequence comparison, which were found to be A. montevidensis (one strain) and A. pseudoglaucus (seven strains). FMR 16310 was placed in a branch together with the ex-type sequence of A. asperescens (sect. Nidulantes). show the trees resulting from the phylogenetic analyses of Myxotrichaceae and Talaromyces, respectively. The phylogenetic tree based on the analysis of the ITS (Fig. 3), included 67 sequences belonging to Myxotrichaceae and Pseudeurotiaceae, whose alignments encompassed a total of 547 characters, including gaps, 204 of which were parsimony informative. The ML and BI analyses showed a similar tree topology. It comprised a main clade of Myxotrichaceae, where 20 strains were located, 17 of Skoua (14 identified as S. fertilis), and the remaining three in a separate branch that might represent a new species of the genus. Finally, three strains phylogenetically distant from the others appeared in a separate branch close to Myxotrichum setosum and Oidiodendron truncatum.
The tree based on four concatenated loci (BenA, CaM, rpb2 and ITS; Table 2; Fig. 4) was built to resolve the phylogenetic relationships of the Talaromyces strains. The dataset contained 123 sequences with a total of 2265 characters, including gaps, (520 of them for ITS, 377 for BenA, 516 for CaM and 852 for rpb2), of which    Page 13 of 30 BenA, 308 for CaM and 349 for rpb2). The sequence datasets did not show conflict in the tree topologies for the 70% reciprocal bootstrap trees, which allowed the multi-locus analysis. The ML analysis showed similar tree topology and was congruent with the Bayesian analysis. In this tree (Fig. 4), the five Talaromyces strains we obtained were located in two different clades: one corresponding to the section Trachyspermi (100% BS / -PP), with four strains phylogenetically distant from T. atroroseus, one of them (FMR 9720) in a separate branch; and the second corresponding to the section Purpurei (74% BS / -PP), where the fifth strain (FMR 16566) was located in a distant branch.
Diagnosis: Distinguished from other genera of Onygenales by the production of thick-walled globose arthroconidia, and because this fungus is an obligate xerophile. Diagnosis: Strongyloarthrosporum catenulatum is phylogenetically close to the Ajellomycetaceae, a family of non-xerophilic fungi characterized by their thermally dimorphic nature and, consequently, pathogenic for animals. By contrast, S. catenulatum is an obligate xerophilic fungus with globose conidia sometimes disposed in chains. Description: Colonies on G18 reaching 20-21 mm diam after 3 wk. at 25°C, elevated, velvety, sulcate, sporulation sparse, exudate absent, yellowish white (4A2) at the centre and white (3A1) at the edge; reverse orange-grey (5B2), diffusible pigment absent. Mycelium composed of hyaline, septate, smooth, thin-to thick-walled, anastomosing hyphae, 1.5-4 μm wide. Conidiophores reduced mostly to single fertile side branches and to the terminal part of the vegetative hyphae, 5-60 μm long, hyaline, disarticulating in conidia. Conidia hyaline, mostly onecelled, occasionally two-celled, holo-and enteroarthric, solitary, disposed terminally, intercalary or sessile on the fertile hyphae, or produced in basipetal chains of up to ten conidia, smooth-walled, thicker than the hyphae, thickener at the ends, mostly globose, 3-6 μm diam, flattened or not at one or both ends, disarticulating by rhexolytic secession from the conidiogenous hyphae. Chlamydospores and racquet hyphae absent.

Subclade B: Eurotiales
Due to both LSU-based ( Fig. 2; sister clade B1) and ITS-BenA-CaM-rpb2-based (Fig. 4) phylogenetetic trees, four of our Talaromyces strains were placed in section Trachyspermi in a well-supported subclade divided in two branches, and one more strain was placed into the section Purpurei in a basal position (Fig. 4) Etymology: After the morphological similarity to the asexual morph of Basipetospora (formerly applied to the asexual morph of Monascus).
Diagnosis: Differs from other species in sect. Trachyspermi in that the conidiogenesis is very similar to that of Monascus (syn. Basipetospora), characterized by retrogressively produced conidia, which have not been previously described in Talaromyces  Description: Colonies on MEA reaching 10-11 mm diam after 3 wk. at 25°C, slightly elevated, velvety to floccose, Fig. 6 Strongyloarthrosporum catenulatum CBS 143841 T . a Colonies on G18, G25 N and MY70FG at 25°C (from left to right), surface and reverse (from top to bottom). b-e Conidiophores and conidia. Scale bar = 10 μm margins entire, yellowish grey (4B2) at the centre and white (4A1) at the edge, exudate absent, sporulation sparse; reverse brownish red (8C8) at the centre and greyish orange (5B6) at the edge, diffusible pigments absent. Mycelium abundant, composed of subhyaline to pale brown, smooth to echinulate, thin-walled, septate, anastomosing hyphae, of 2-3 μm wide. Conidiophores mostly reduced to a single conidiogenous cell, sometimes slender and with an additional conidiogenous locus near the base, arising alternately or oppositely at both sides of the vegetative hyphae, mostly separate from the vegetative hyphae by a basal septum. Conidiogenous cells smooth-walled to echinulate, mostly cylindrical and occasionally slightly slender towards the apex, sometimes broadening below the apex, but also flask-or barrel-shaped, very variable in length, 3-20(− 45) × 1-2.5 μm, conidiogenesis retrogressive. Conidia one-celled, hyaline and echinulate when young, becoming brown to dark brown and nearly smooth-walled with the age, formed basipetally, in false chains of up to ten conidia, mostly globose, 3.0-5.0 μm diam. Sexual morph not observed.
Etymology: From Latin brunneus-, brown, and -sporarum, spore, in reference to the colour of the conidia.
Diagnosis: Distinguished from other species in sect. Purpurei, with the exception of T. purpurei (the type species of the section), by the production of solitary phialides and monoverticillate conidiophores (biverticillate conidiophores in the other species of the section). However, T. brunneosporus can be differentiated from T. purpureus because lack of a sexual morph (present in the latter species), Description: Colonies on MEA reaching 13-14 mm diam after 3 wk. at 25°C, slightly elevated, velvety to floccose, margins irregular, yellowish white (4A3), exudate absent, sporulation sparse, reverse light brown (6D8) at the centre and yellowish brown (5D6) at the edge, diffusible yellowish brown (5E6) pigment present. Mycelium abundant, composed of subhyaline, smooth-and thin-walled, septate, anastomosing hyphae 2-3 μm wide. Conidiophores mostly stalked, monoverticillate, smooth-and thin-walled, bearing one to four conidiogenous cells at the top, frequently arising oppositely at both sides of the vegetative hyphae, sometimes reduced to a single conidiogenous cell, sessile or integrated to the vegetative hyphae (= adelophialides). Conidiogenous cells phialidic, smooth-walled, mostly slender towards the apex, flaskshaped, 8-12 × 2.5-3.5 μm, with a darkened apical area when the conidiogenous cells have produced several conidia, conidiogenesis enteroblastic. Conidia one-celled, globose, hyaline and smooth-walled when young, becoming brownish-green to dark brown and verrucose with the age, 3-4 μm diam, in long false chains of up to 25 conidia. Sexual morph not observed.
Minimum, optimal and maximum temperature of growth on G18 are 15, 25, and 30°C, respectively; no growth on CYA at 37°C nor on CREA at 25°C.
Notes: Talaromyces brunneosporus and T. purpureus grow more slowly on CYA and MEA than other species of the section. However, T. brunneosporus produces dark brown colonies with a dark brown diffusible pigment on CYA, while the colonies of T. purpureus are pale beige and without diffusible pigments. Also, the colonies on OA and MEA are purplish in T. purpureus and pale coloured and dark brown in T. brunneosporus. Talaromyces affinitatimellis Rodr.-Andr., Stchigel & Cano, sp. nov. Fig. 9. MycoBank MB 823591.
Etymology: From Latin affinitatis-, affinity, and -mellis, honey, after the substrate from which the fungus was isolated.
Diagnosis: Differing from all other species in sect. Trachyspermi (with the exception of T. basipetosporus) by the production of conidia by retrogressive conidiogenesis. Talaromyces affinitatimellis differs from T. basipetosporus by the production cylindrical, smooth-walled to echinulate conidiogenous cells ending in a greenish brown, broad collarette-like structure (conidiogenous cells irregularly-shaped, smooth-walled, and without such apical structure in T. basipetosporus).
Description: Colonies on MEA reaching 29-30 mm diam. After 3 wk. at 25°C, flat, floccose, not sulcate, margins entire, olive (3D3) at the centre and white (4A1) at edge, exudate absent, sporulation sparse; reverse pale orange (5A3) at centre and pale yellow (4A3) at edge, diffusible pigment absent. Mycelium abundant, composed of subhyaline to pale brown, smooth-and thinwalled, septate, anastomosing hyphae, of 2-4 μm wide. Conidiophores hyaline to pale brown, reduced to a single conidiogenous cell, occasionally with an additional conidiogenous locus near the base or lateraly disposed, or short-stalked and bearing two conidiogenous cells, sometimes with an additional lateral conidiogenous cell arising alternately at both sides of the vegetative hyphae, separate from them by a basal septum. Conidiogenous cells hyaline to pale brown, smooth-walled, mostly cylindrical and occasionally slightly slender towards the apex, sometimes ending in a greenish-brown, broad collarette-like structure, 3-20 × 1.5-3 μm, conidiogenesis retrogressive but enteroblastic. Conidia one-celled, hyaline and echinulate, becoming brown to dark brown and nearly smooth-walled with the age, produced basipetally in false chains of up to ten in number, mostly globose, 3.0-5.0 μm diam. Sexual morph not observed.

Subclade D: Incertae sedis
Based on both LSU-based ( Fig. 2; sister clade D1) and ITS-based (Fig. 3) phylogenetic trees, ten of our strains were located in a well-supported and separated branch related to species of the genera Oidiodendron and Myxotrichum, and phylogenetically distant from the most similar taxa included in the study, M. setosum and O. truncatum (Fig. 3). Recognition of all of these distinct strains was also supported by unique phenotypic characteristics; therefore, we propose the recognition of the new species Oidiodendron mellicola. Furthermore, because three of our strains were placed near Skoua fertilis in both LSU-based ( Fig. 2; sister clade D2) and ITSbased (Fig. 3) phylogenies and because they showed different phenotypic features and enough phylogenetic distance relative to S. fertilis, we also propose the introduction of a further new species, Skoua asexualis.
Etymology: From Latin mellis-, honey, and -cola dwelling on, referring to the habitat.
Diagnosis: Forming a terminal clade together with O. truncatum and M. setosum at a significant phylogenetic distance (5.3% from the other two species), and differing morphologically from other known species of Oidiodendron and the asexual morphs of Myxotrichum in the absence of well-differentiated conidiophores, and the slow growth. (seen after 6 wk. of incubation), exudate absent, reverse orange-white (6A2) at the centre and orange-grey (6B2) at the edge, diffusible pigment absent. Mycelium composed of hyaline, septate, smooth-and thin-walled hyphae, 1-3 μm wide. Conidiophores reduced to fertile side branches and the terminal part of a vegetative hyphae, mostly simple or once branched near or at the base, 10-40 μm long, pale olive, disarticulating in conidia. Conidia one-celled, mostly holoarthric, sometimes enteroarthric, mostly in chains of up to ten, occasionally solitary and sessile, mostly barrel-shaped, sometimes cylindrical, conical or "Y"-shaped, 5-14 × 2.5-5 μm, pale olive, disarticulating by schizolytic or rhexolytic secession from the hyphae. Chlamydospores absent. Sexual morph absent. Description: Colonies on PDA reaching 6-7 mm diam after 3 wk. at 25°C, elevated, velvety, sporulation abundant, exudates absent, diffusible pigment absent, colonies brown (7E6) at the centre and whitish at the edge, reverse brownish orange (6C5) at the centre and greyish orange (5B3) at the edge. Mycelium composed of hyaline, repeatedly septate, smooth-and thin-walled hyphae, 2-6 μm wide. Conidiophores absent. Conidia mostly one-celled, occasionally two-to three-celled, hyaline, solitary or in short chains, smooth-and thick-walled, mostly globose, occasionally broadly ellipsoidal, pyriform, or irregular-shaped, truncate at one or both ends, 3-7 μm diam, conidiogenesis holoblastic when sessile or terminal, and holothallic when intercalary, disarticulating by rhexolytic secession; the holoblastic and holothallic conidia produce a succession of secondary holoblastic conidia, forming a big, radiating mass of cells of up to 50 μm diam, which eventually detach as complex asexual propagules from the fertile hyphae. Chlamydospores similar to the conidia but thicker, mostly non-or occasionally one-septate, intercalary or terminal. Sexual morph unknown.

DISCUSSION
This is the most comprehensive assessment of the diversity of the xerotolerant and xerophilic fungi of honey intended for human consumption to date. We have isolated selectively and identified, by a polyphasic approach, six species of ascomycetous yeasts and 27 of filamentous ascomycetes, some representing new taxa, from honey samples. The yeasts, Candida magnoliae, C. sorbosivorans, Schizosaccharomyces octosporus, Zygosaccharomyces barkeri, Z. mellis, and Z. gambellarensis, had been reported from honey before, and C. magnoliae has also been associated with living honeybees (Gilliam et al. 1974b). All these yeasts have been described as osmophilic and able to grow at a w of 0.80 or lower (Tilbury 1967;van Eck et al. 1993;Ganthala et al. 1994;Erickson & Mc-Kenna 1999;Torriani et al. 2011). We found C. magnoliae and C. sorbosivorans were phylogenetically closely related (see Fig. 2), and it was reported that both differ only in a few physiological characteristics (James et al. 2001). To our knowledge, none of the species of Aspergillus that we isolated (A. asperescens, A. montevidensis, and A. pseudoglaucus) have previously been reported from honey. Aspergillus asperescens was originally isolated from soil and bat dung (Stolk 1954), but also from rotten wood and soybean seeds; however, most of the isolates were from cave soil (probably linked to bat dung). Aspergillus montevidensis and A. pseudoglaucus have been reported as the most important food-spoilage species of the genus (Pitt & Hocking 1977;Kozakiewicz 1989), but are known from extreme environments such as salterns (Butinar et al. 2005). Aspergillus montevidensis has been reported from various environmental samples (air, soil, etc.), and even on honeybees and bee larvae (http://gcm.wfcc.info/; Talice & Mackinnon 1931;Gilliam et al. 1974a); A. pseudoglaucus has been reported in air, paper and soil (http://gcm.wfcc.info/; Blochwitz 1929). Aspergillus montevidensis and A. pseudoglaucus are able to grow at a w values of 0.80 (Snow 1949;Armolik & Dickson 1956;Guynot et al. 2003). Monascus is a well-known genus with species (especially M. purpureus and M. ruber) of economic importance due to their use in production of foodstuffs, bioactive compounds, pigments and enzymes. Currently, Monascus is placed in Aspergillaceae (syn. Trichocomaceae) based on phylogenetic studies, and closely related to Leiothecium ellipsoideum and Xeromyces bisporus (Houbraken & Samson 2011;. Recently, three new species   Lin 1975;Hawksworth & Pitt 1983). Monascus ruber has also been found in soil and human clinical specimens (Hawksworth & Pitt 1983). Species of Monascus have been previously reported in honey by Snowdon & Cliver (1996) and by Barbosa et al. (2017). Monascus pilosus, M. purpureus, and M. ruber were reported previously (Hawksworth & Pitt 1983) as able to grow well on G25 N (a w = 0.93). The species of Penicillium we found in honey included P. camemberti, P. citrinum, P. corylophilum, and P. cravenianum. The most common source of isolation of P. camemberti is blue cheeses, but it can also be found on a wide variety of substrata (Thom 1906; http://gcm.wfcc.info/). Penicillium citrinum was originally reported in milk and bread in the USA (Thom 1910), but it is found globally and easy to recover from spoiled foods and diverse environmental sources (www. cabri.org/collections.html) including honey, pollen and bee nests (Barbosa et al. 2018). Penicillium corylophilum (Dierckx 1901) mostly occurs in damp buildings in North America and Western Europe, but also in foods and mosquitoes (Da Costa & De Oliveira 1998;McMullin et al. 2014), and honey (Sinacori et al. 2014). The minimum a w reported for the growth of P. camemberti, P. citrinum and P. corylophilum was around 0.80 (Abellana et al. 2001;Fontana 2008;Kalai et al. 2017). Penicillium cravenianum, a species moderately xerotolerant (grows on G25 N), has only been reported in soil (Visagie et al. 2016). Notably, all the isolates of Talaromyces that we found in honey belonged to three unrecognized species. Talaromyces basipetosporus was recovered from a honey sample in Buenos Aires province, Argentina, and is characterized by simple conidiophores that mimic those of the asexual morph of Monascus (syn. Basipetospora), which develops conidia by a retrogressive mode of conidiogenesis, a feature not previously reported in Talaromyces. Talaromyces affinitatimellis displays a similar conidiogenesis to T. basipetosporus and both species are phylogenetically closely related but phenotypically differentiated as T. affinitatimellis grows faster and produces more complex conidiophores. Talaromyces brunneosporus differs from the other species of sect. Purpurei, apart from T. purpureus, in having monophialidic and monoverticillate conidiophores (they are biverticillate in the other species). However, both species are distinguishable because T. brunneosporus produces penicillate conidiophores (not aspergillate as in T. purpureus), longer phialides, and verrucose conidia with a flattened base (T. purpureus conidia are ornamented by spiral ridges). Talaromyces basipetosporus has a high xerotolerance, with similar growth rates on MEA with sugars up to a w 0.82. Despite the decreasing growth rates of T. brunneosporus and T. affinitatimellis when sugar concentration increases, both fungi are able to grow at a w 0.82 (Fig. 12). Xerochrysium xerophilum (Pitt et al. 2013;syn. Chrysosporium xerophilum, Pitt 1966), is an extreme xerophile with a minimum a w for growth of 0.66 (Gock et al. 2003;Leong et al. 2011). This fungus, previously reported from chocolate, coconut, dried prunes, and stored corn (Pitt & Hocking 2009;Pitt et al. 2013), has not been found in honey until now. This species is phylogenetically close to Monascus (Pitt et al. 2013). Among the species of Onygenales, Ascosphaera atra and Eremascus albus were recovered once and four times, respectively. Ascosphaera atra (Skou & Hackett 1979) was originally reported from dead larvae of the alfalfa leafcutter bee covered in cysts of Ascosphaera aggregata (Skou 1975), and from pollen in the gut of healthy leafcutter larvae. This fungus was subsequently reported from grass silage (Skou 1986). Ascosphaera atra is homothallic and saprobic, probably being a common contaminant of pollen (Skou & Hackett 1979), which would explain its presence in honey samples. Eremascus albus is a well-known xerophilic fungus, with spores that can germinate at a w as low as 0.70 (Pitt 1965). This fungus has been reported to spoil malt extract (Eidam 1883), chocolate cake, dried fruits, and mustard powder (Harrold 1950), but never previously from honey. We identified several isolates belonging to the newly described family Helicoarthrosporaceae, which only includes the new monotypic genus Helicoarthrosporum, and a single strain belonging to the new monotypic genus Strongyloarthrosporum (Ajellomycetaceae). The morphology of Helicoarthrosporum mellicola resembles species of Scytalidium (S. cuboideum, S. ganodermophthorum, and S. sphaerosporum) because of the production of cuboid arthroconidia in long chains. However, Helicoarthrosporum is phylogenetically distant from Scytalidium, as the latter is related to Myxotrichaceae. Strongyloarthrosporum catenulatum was found to be phylogenetically close to Ajellomycetaceae, whose members are thermally dimorphic and pathogenic to animals (including the humans), and has never been reported as xerotolerant. However, having features not seen in that family, S. catenulatum is unequivocally a xerophilic fungus, only growing on G18, G25 N and MY70FG, and producing globose arthroconidia, either singly or in chains. The sole xerophilic fungus phylogenetically close to S. catenulatum is Eremascus albus (Eremascaceae), but it only develops a sexual morph. Regarding the family Myxotrichaceae, Skoua fertilis, which was detected in all honey samples, resembles Eremascus albus (Eidam 1883) in having naked asci arising directly out of the mycelium and formed by the fusion of two equal cells borne on short entwined hyphae. Both taxa can be only morphologically differentiated by the shape of the ascospores and by sexual reproductive details. While S. fertilis (syn. E. fertilis) belongs to Leotiomycetes, closely related to Myxotrichaceae (Wynns 2015), E. albus is located in Eurotiomycetes, closely related to Onygenales (Cai et al. 1996;Berbee 2001;Wynns 2015). Skoua was introduced for E. fertilis (i.e. Skoua fertilis) and has been reported on bee bread, honeycomb, dried prunes and spoiled moist prunes, green compost, and shortcake (www.cabri.org/collections.html; http:// gcm.wfcc.info/; Harrold 1950), but not so far on honey.
The minimum a w for growth and sporulation reported for S. fertilis was 0.77 (Pitt 1965;Wynns 2015), a similar value observed in all our strains (0.82). We isolated three strains of Skoua phylogenetically different from S. fertilis, and named them as Skoua asexualis because they form asexual spores instead of the sexual spores as observed in the type species of the genus. Bettsia alvei (Skou Fig. 12 Relatedness between the growth of the new fungal taxa and the decreasing water activity (a w ) of the culture medium.  1972,1975), the other fungus identified in all honey samples, belongs to Pseudeurotiaceae and is characterized by dark, closed ascomata (usually called "spore cysts") and hyaline globose ascospores, forming a sticky mass. Bettsia alvei has been isolated from hives in Europe as well as the USA (Burnside 1929), and from bakery products, spoiled chocolate, desiccated coconut, honeycomb, concentrated jelly, dried and spoiled prunes, pollen, table jelly, bee wax, and wine starters (www.cabri. org/collections.html; http://gcm.wfcc.info/). It was also isolated from chocolate in Austria (a w less than 0.3), but thus far had not been recorded from honey. The lowest a w tested for growth of this species was 0.88 (Beuchat & Pitt 1990) and 0.89 (Udagawa & Toyazaki 2000), similar values to those we found. All our isolates of B. alvei developed the chrysosporium-like asexual morph but failed in the production of the sexual morph. Among the most frequent species we isolated was an undescribed species of Oidiodendron, O. mellicola. Species of this genus are mostly recovered from soil and other substrata rich in cellulose, and are found worldwide (Domsch et al. 1980;Calduch et al. 2004;Rice & Currah 2005). Oidiodendron mellicola is phylogenetically related to O. truncatum and M. setosum, the former characterized by welldifferentiated dark conidiophores and barrel-shaped conidia with a dark scar at one or both ends (typical features of Oidiodendron), and the latter by hyaline conidiophores and conidia, and by dark brown to black, spinose, gymnothecial ascomata (typical of the genus Myxotrichum). Interestingly, M. setosum is reported as a common hive fungus in Europe (Burnside 1929). Oidiodendron mellicola is the only species of the genus reported from honey, and it can be distinguished morphologically from other species of the genus by its absence of stipitate conidiophores, and the production of long chains of conidia, which are pale, smooth, ellipsoidal to cylindrical, truncated (but not darkened, as in O. truncatum) at one or both ends, and by the slow growing colonies. Like most of the species of the genus, O. mellicola grows better at 15°C than 25°C. Other fungi rarely found in our study were Alternaria multiformis, previously only reported from soil (Simmons 1998), and the mucoralean Cunninghamella bertholletiae, Mucor plumbeus, and Rhizopus oryzae, all found worldwide. These probably represent environmental contaminants. Although all the new taxa that we propose displayed a high xerotolerance, only Strongyloarthrosporum catenulatum can be considered an obligate xerophile, because it was able to grow faster at the lowest a w tested (Fig. 12).

CONCLUSION
The application of G18 as a selective culture medium for isolation of xerotolerant/xerophilic fungi from honey samples enabled the recovery and identification of 13 genera and 29 species of Ascomycota, and three genera (one species for each) of Mucoromycota. Many of these fungi have never reported from honey before. Among them, we proposed a new family (Helicoarthrosporaceae), two new genera (Strongyloarthrosporum and Helicoarthrosporum) and seven new species (Strongyloarthrosporum catenulatum, Helicoarthrosporum mellicola, Oidiodendron mellicola, Skoua asexualis Talaromyces basipetosporus, T. brunneosporus, and T. affinitatimellis). All fungal taxa that we isolated from honey were able to grow at low water activity (up to 0.82), but only Ascosphaera atra, Bettsia alvei (two fungi strongly associated to honeybees and their life-style), Eremascus albus, Strongyloarthrosporum catenulatum (one of the new taxa we described) and Xerochrysium xerophylum can be considered obligate xerophiles. Also, because several of the honey samples were thermally treated, these fungi can be considered as hot-resistant. Honey is evidently a reservoir of xerotolerant and xerophilic fungi, which survives to the thermal treatment used to make honey non-crystallisable. Some of these fungi are related to the honeybee life-style; however, as is in the case of the new taxa described here, the origin in nature remains unknown. In the latter case, flowers and aphids could play an important role as a source of such fungi. During the course of the study, the most important pathogenic fungi for honeybees, Aspergillus flavus and Ascosphaera apis, were not found. Several of the fungi found in honey samples (Aspergillus and Pencillium spp.) are potential producers of mycotoxins, but this does not mean that the honey may represent a risk to the health of the consumer, because (in general) the production of mycotoxins or the fungal growth are suppressed at water activities lower than 0.70 (Mannaa & Kim 2017), as is the case of honey (a w of 0.60 or less). Honey should be considered as a "living food" and, consequently, its "normal" mycobiota merits more extensive study. It is expected that such "normal" mycobiota may vary qualitatively and quantitatively, depending on the geographic origin, the botanical type and water activity of the honey, among other physicochemical and biological parameters. Honey is clearly one of the relatively unexplored habitats for the missing fungal diversity, especially as the new taxa we found came from samples from just two countries. Malt extract yeast extract 70% fructose-glucose; nrRNA: Nuclear ribosomal ribonucleic acid; OA: Oatmeal agar; PDA: Potato dextrose agar; PYE: Phytone yeast extract agar; rpb2: fragment of the RNA polymerase II subunit 2 gene; SEM: Scanning electron microscopy; TOTM: Test opacity tween medium; TreeBASE: a repository of user-submitted phylogenetic trees and data used to build them; YES: Yeast extract sucrose agar