The CIRM-CF fungal collection: a treasure trove of taxonomic and geographic diversity
As of March, 2021, the CIRM-CF offers to the scientific community a collection of 2,824 fungal strains, from 259 genera and 557 species (Fig. 1a). This huge diversity results from the efforts deployed by the CIRM-CF since 2006 to acquire and preserve the fungal diversity living in French territories (Fig. 1b), through field collection (in natural habitats) and strain deposits by mycologists. To do so, the CIRM-CF has established strong connections with a consortium of expert mycologists (from Universities or learned societies) to carry out macro- and micro- morphological identification, as well as strain isolation. As shown in Fig. 1b, the CIRM-CF took part to 27 field collecting expeditions, in tropical rainforests (mainly French Guiana, Martinique, Guadeloupe) as well as temperate forests. As a result, more than 3,300 isolated strains and their associated information (taxon, geographic location, substrate of isolation, …) have been acquired by the CIRM-CF, 44% of which issued from fungal specimens living in tropical or sub-tropical climate. Before entering the CIRM-CF collection, and to ensure that high quality BRC standards are met, the purity and the viability of candidate strains are checked by 3 successive sub-cultures. Moreover, strain identity deduced from morphological identification keys are checked by molecular authentication, such as the genotyping of one barcoding gene (usually ITS1-5.8S-ITS2). Consequently, about one third of isolated strains are rejected due to (i) impurity, (ii) poor viability or (iii) when the molecular information was not in agreement with specimen morphological identification. This elevated exclusion rate highlights the technical hurdles and challenges pertaining to proper cultivation and/or isolation of pure fungal strains from natural habitats. After authentication, the fungal strains are maintained with three different storage modes (cryopreservation under liquid nitrogen, water and oil-submerged cultures at 4°C) in different locations and the viability of the cryopreserved cultures are checked after six months.
From the complete collection of the CIRM-CF, a total of 1,031 fungal strains were selected for this functional study. This set represents a wide taxonomic diversity of 400 species, 99% from the Ascomycota and Basidiomycota phyla, and spanning 26 orders, 78 families (43 Basidiomycota, 31 Ascomycota and 4 Mucoromycota families) and 231 genera (135 Basidiomycota, 91 Ascomycota, 5 Mucoromycota genera (Supplementary Tables 1 and 2). 68% of these strains were collected in French territories, 40% of which coming from overseas territories (mainly French Guiana and Martinique). Of note, fungal orders for which at least 10 strains were tested (shown in bold characters in Supplementary Table 2), encompass 365 different species. Furthermore, although the strain sampling of some orders could appear as downsized (e.g., 20 vs 392 strains in the Gloeophyllales and Polyporales orders, respectively), the diversity sampling is still ensured since these seemingly downsized orders usually contain less families (e.g., 1 vs 18 families, respectively)12,13. We underscore that the CIRM-CF collection is dedicated to filamentous fungi able to decay plant biomass transformation, the represented fungi are thus almost exclusively saprobic (growing on wood particularly).
Set-up of the large-scale multi-phenotyping
To assess the degrading potential of the fungal strains towards molecules engineered by mankind (Fig. 2), we selected five compounds known to be highly recalcitrant to degradation14− 17: two different synthetic azo dyes, Reactive Black 5 (RB5) and Basic Blue 41 (BB41), which are amongst the most commonly used dyes in the textile industry, and thus major environmental toxic pollutants; soluble lignosulfonate (LGS) as a pollutant coproduct from the pulp and paper industry; Impranil® DLN-SD (IMP) as a soluble polyurethane used in plastic and textile industries; and microcrystalline cellulose (Avicel®PH-101; AVI), a recalcitrant refined wood pulp, notably widely used in the food industry as texturizer (Supplementary Fig. 1). In the screening assay, LGS, IMP and AVI were the sole carbon source, whereas azo dyes were supplemented with malt to allow fungal growth. For each strain, a six-well plate was used to assess simultaneously the different growth conditions (Fig. 2a), including a positive control culture on malt medium. Of note, a negative control culture on agar medium devoid of additional carbon source was systematically assessed in parallel. After an incubation period of 13 days at 25°C, different phenotypes were observed: the decolorization of RB5 and BB41, phenol oxidation of LGS, clearing halo formation of IMP, and growth on AVI. The biosorption (i.e. absorption on the mycelium) of the dyes was not considered as decolorization. To obtain semi-quantitative information on the efficiency of degradation, we attributed scores (from 0 to 4) as follows (Fig. 2b): growth was estimated by comparison of the growth diameter and density of mycelia with cultures on agar plate (negative control, score 0) and with cultures on malt agar plate (positive control, score 4); scores for azo dyes (RB5 and BB41) decolorization and LGS phenol oxidation were 0 (negative), 1 (light), 2 (medium), 3 (strong) or 4 (maximum), whereas scores for clearing halo formation on IMP were 0 (negative) or 4 (positive) (raw data are available in Supplementary Data 1).
Most fungal strains were able to grow on RB5 (99%) and BB41 (91%) supplemented with malt extract, whereas only 34% and 20% of them significantly achieved decolorization, respectively (Supplementary Fig. 2). Also, whereas all strains grew properly in the presence of RB5, about 20% of the strains showed limited growth (score ≤ 2) in the presence of BB41. Thus, BB41 seems less favorable than RB5 to fungal growth. Regarding LGS, most strains (92%) were able to grow, albeit the growth extent remained limited as only 13% reached the scores 3 and 4, most probably due to the difficulty to assimilate LGS that was used as sole carbon source in the culture medium. Yet, 59% of the strains managed to oxidize phenols from LGS (observed by browning). As to crystalline cellulose, a substrate that could be expected to be readily degraded by saprotrophic fungi, only 15% of strains succeeded to significantly grow. Regarding IMP, our large-scale phenotyping experiment allowed to evidence a few set of strains (2%) showing a significant growth (score ≥ 3) (and 43% of the strains showed a limited growth), indicating that IMP was the most recalcitrant compound tested in the present study. Nonetheless, 23% of the strains succeeded to form a clearing halo, underlying that limited growth could be sufficient to produce degrading enzymes.
Functional phenotypes are not correlated with growth capacities.
As mentioned above, we did not observe any correlation between growth capacities and (i) decolorization of azo dyes, (ii) oxidation of LGS or (iii) clearing of IMP (Supplementary Fig. 3a). This result suggests that extensive fungal biomass development is not required for degrading the targeted compounds. In an attempt to identify potential correlations between the phenotypes, we focused our attention on the best-performing strains (with the highest score of 4) for each phenotype (Supplementary Data 1), and generated an Upset plot (Fig. 3a). Remarkably, most of the best-performing strains were shown to have a single, preferred target compound, with little overlap observed between functional phenotypes (Fig. 3a). Strikingly, this observation was also true for RB5 and BB41, two azo dyes characterized by different chemical structures and number of azo bonds (Supplementary Fig. 1). This lack of correlation between phenotypes suggests that the best-performing strains make use of specialized enzymatic activities not universally shared. Regarding the phylogenetic distribution of the observed phenotypes (Fig. 3b), for orders in which > 10 strains could be characterized (shown in bold characters/pink histograms in Fig. 3b), the following general trend stood out: Basidiomycota appear to degrade or modify a broader range of compounds than Ascomycota. RB5 was predominantly decolorized by Basidiomycota, whereas BB41 was decolorized by both Basidiomycota and Ascomycota (mainly Gloeophyllales, Hypocreales and Xylariales orders). The oxidation of LGS was mainly observed for Basidiomycota, although some Ascomycota showed relatively high mean scores (Pleosporales, Hypocreales, Xylariales). The clearing activity on IMP was observed in a large array of fungal orders with Ascomycota fungi (Pleosporales, Eurotiales, Hypocreales) displaying the highest scores. Finally, the Polyporales and Russulales orders, belonging to Basidiomycota phylum, showed the best mean scores for growth on microcrystalline cellulose.
Fungi display a great functional diversity at each taxonomic rank.
The global analysis presented hereinabove highlighted that most of the best-performing strains display a main phenotype although minor “activity” towards other targets could be detected. To probe the order/species-dependent functional diversity, we categorized each target compound/strain couple as either non-active (“N”, i.e. when score was 0) or as active (“A”, i.e. for scores from 1 to 4). Given that 5 compounds were tested, this 2-level categorization yields in theory 25 possible phenotype combinations (Fig. 4a). Interestingly, the 32 theoretically possible functional profiles were all observed at least once, albeit profiles with single- and dual phenotypes were preponderant. We observed that some profiles were more frequent than others such as the functional profile #3 (LGS oxidation) and #4 (LGS oxidation and growth on microcrystalline cellulose). Taking the Russulales order as an example, while profile #4 gathers one-third of the tested strains, the remaining two-thirds are distributed across eleven different profiles, indicating the presence of a significant functional diversity within a single order.
To delineate the “strain effect” on the observed functional diversity, we then analyzed to which extent the number of tested strains impacted the number of detected functional profiles (Fig. 4b-e). Unsurprisingly, at the order level, the higher the number of tested strains, the more functional profiles were observed (Fig. 4b). We also observed a similar correlation at the family (Fig. 4c) and genus (Fig. 4d) levels. Astonishingly, a high functional diversity was also noticed between strains belonging to a same species (Fig. 4e). Depending on the species, 3 to 12 distinct profiles were observed when more than 10 strains of the same species were screened. For instance, 12 distinct profiles were observed with the phenotyping of 19 strains of Pycnoporus cinnabarinus or the phenotyping of 49 strains of Agaricus bisporus. Using the non-linear regressions shown in Fig. 4b-e, and provided that good enough taxonomic coverage/species diversity is ensured, we estimated the minimal set of strains to be screened to cover a functional diversity as large as possible to be 150, 75, 40 and 20 strains, at the order, family, genus and species level, respectively.
A roadmap for the selection of fungal families for dedicated applications.
Amongst the wide diversity offered by the fungal kingdom, identifying the most suitable fungal family(ies) for a dedicated application is a true challenge. Here, we set off to provide knowledge-based selection guidelines regarding the degradation of the 5 selected industrial compounds, and applications thereof. To this end, we analyzed the statistical distribution of the functional phenotype scores within fungal families with satisfying diversity coverage (i.e., > 10 tested strains), representing a total of 900 fungal strains (Fig. 5, Supplementary Data 2). Horizontal reading of Fig. 5 shows that some compounds are preferentially targeted by specific fungal families (e.g., RB5 by Basidiomycota families) while vertical analysis informs that some fungal families display a marked preference for specific industrial compounds (e.g., Phanerochaetaceae with AVI). Strikingly, almost all families revealed the presence of high-score outliers, highlighting that even fungal families with seemingly overall poor activity on a particular compound can contain a few efficient strains.
In details, regarding microcrystalline cellulose, out of the 154 strains that showed growth, 142 belong to the Basidiomycota phylum (Fig. 5a), with a significantly higher prevalence of strains from the Phanerochaetaceae, Polyporaceae and Meruliaceae families. We underscore that strains from the phlebioid clade, within the order Polyporales, (Phanerochaetaceae, Irpicaceae, Meruliaceae)12,21, show the highest mean scores of growth on microcrystalline cellulose. For Ascomycota, the Hypocreaceae, Aspergillaceae and Xylariaceae families stood out with numerous strains able to slightly grow on microcrystalline cellulose (median score ≤ 2).
Concerning LGS oxidation, families from lignicolous Basidiomycota showed relatively good mean scores, with highest prevalence being observed for strains from the Ganodermataceae, Pleurotaceae, Agaricaceae, Polyporaceae and Hymenochaetaceae families (Fig. 5b). In contrast, the mean scores obtained by brown-rot families (Gloeophyllaceae, Fomitopsidaceae) were very low. This observation is consistent with the acknowledged limited abilities of brown-rot fungi to degrade or modify lignin in nature (see Discussion).
Regarding the decolorization of RB5 (Fig. 5c), stark differences between the fungal phyla were observed, Basidiomycota being in general much more efficient than Ascomycota. In particular, the Basidiomycota families Pleurotaceae (69% of tested Pleurotaceae), Meruliaceae (66%), Hymenochaetaceae (63%), and Polyporaceae (56%) showed the strongest prevalence for decolorizing RB5. The fact that families of wood decaying Basidiomycota were overall the most efficient for RB5 decolorizing is in line with previous observations made on synthetic dyes22. For Ascomycota, a few strains were outstanding, in particular amongst the Fusarium species (Nectriaceae). Strikingly, the taxonomic distribution of phenotypes observed on the second tested dye, BB41, was totally different from that of RB5 (Fig. 5d), Gloeophyllaceae, Nectriaceae, Xylariaceae and Fomitopsidaceae families displaying the best mean scores.
Finally, concerning the commercial polyester polyurethane IMP, we were surprised to observe that all families except two had at least one strain able to efficiently clear IMP (Fig. 5e). Indeed, Fig. 5e shows outliers in all families, except the Pleurotaceae and Phanerochaetaceae families. In general, Ascomycota were more efficient (higher mean and median scores) than Basidiomycota at degrading such artificial polymer. In particular, the Pleosporaceae family showed the best mean-score, with 14 out of 15 tested strains clearing IMP with success, followed by the Aspergillaceae family that displayed more than 50% of positive strains. It is noteworthy that genera containing pathogenic species (Botrytis, Sclerotinia, Alternaria, Phoma, Colletotrichum, Mycosphaerella, Verticillium, Trichothecium, Fusarium…) showed a propensity to be good IMP degraders. Yet, some Basidiomycota families (in particular Hymenochaetaceae, Gloeophyllaceae, Strophariaceae, Agaricaceae) also emerged (because of few strains from Phellinus (sensu stricto), Tropicoporus, Gloeophyllum, Heterobasidion, Amylosporus, Trichaptum, Laetisaria, Agrocybe and Agaricus genera) as potential good candidates to uncover novel polyesterases (see Discussion).