Background

Isolation and screening of microorganisms have been applied as a strategy to obtain strains able to produce industrially-relevant enzymes. Considering the increased number of available genomes, new rational approaches, such as genome mining, provide an attractive alternative to labor-intense screenings [1, 2]. This is also an interesting alternative to target the prospection of enzymes in fungi deposited in culture collections. Previous successes of genome mining have been documented for lipases [2], lignocellulosic-degrading enzymes [3, 4] and peptidases, particularly in Aspergillus species [5].

Peptidases (EC 3.4) consist of a large group of hydrolytic enzymes that catalyze the hydrolysis of proteins by cleavage of the peptide bonds between amino acid residues [6]. The use of microbial peptidases provides technological and economic advantages in industries including detergent, textile, leather, dairy and pharmaceutical production. Peptidases are one of the most important groups of industrial enzymes representing and accounting for approximately 65% of the total enzyme production worldwide [7, 8].

In industrial processes enzymes are often subjected to extreme physicochemical conditions, which are suboptimal for mesophilic ones [1]. Enzymes with potentially unusual properties such as those from thermophilic fungi are thus much sought after. Their enzymes usually have higher thermostability when compared to the mesophilic species, although this is not the case for all proteins. Not only are they of immediate industrial interest but they also enable us to investigate their thermostability patterns and use this knowledge in the rational engineering of thermostability into thermolabile enzymes. Heat-tolerant fungi, often found in composting systems, have been reported as producers of thermostable enzymes with industrial applications [9]. Peptidases from thermophilic fungi have been evaluated in relation to their biochemical properties (e.g. thermal stability) and industrial applications, for instance, Thermoascus aurantiacus and its hydrolytic activity on bovine casein [10], Thermomucor indicae-seudatiacea and Rhizomucor miehei in milk clotting activity [11, 12].

Here, we investigated in silico the diversity of peptidases in the genomes of heat-tolerant fungi and their phylogenetically related mesophilic counterparts. In order to predict the determinants of their thermostability, we investigated the peptidase profile, i.e. catalytic type and families, and amino acid composition of these peptidases and predicted the structural patterns of the representatives from the A1 family aspartic peptidases.

Methods

Fungal genomes retrieval and phylogenetic analysis

The annotated genomes of thermophilic, sensu Oliveira et al. [13], thermotolerant and mesophilic species listed in Table 1 were retrieved from public databases, including the National Center for Biotechnology Information (NCBI; http://www.ncbi.nlm.nih.gov/), DOE Joint Genome Institute (JGI, http://genome.jgi.doe.gov/) and Genozyme (http://genome.fungalgenomics.ca/).

Table 1 List of fungal genomes mined for peptidase encoding genes
Full size table

We inferred a phylogenetic tree to evaluate the evolutionary relationships between the selected species. A super alignment of the selected fungal proteomes was constructed with the Hal pipeline [14], allowing for no missing data. Poorly aligned positions and ones with gaps were removed with Gblocks 0.91b [15]. The following stringent parameters were used: the maximum number of contiguous non-conserved positions was limited to six amino acids, and the minimum length of a block to 15 amino acids. This produced a 106,488-bp-long alignment, which was used for the estimation of the phylogeny. We estimated the best protein evolution model with ProtTest 3.2.1 [16]. The species tree was generated in PhyML 3.3 [17]. We calculated the Approximate Bayes (aBayes) branch supports. The analysis was run using the LG model of evolution. The ProtTest estimate of the α-parameter of the γ-distribution of six substitution rate categories (1.019), and the determined proportion of invariable sites (0.067) were used. The phylogeny data, including alignments, are available in the Treebase repository (http://purl.org/phylo/treebase/phylows/study/TB2:S22179).

Thermomucor indicae-seudaticae genome retrieval and annotation

Few annotated fungal genomes of the order Mucorales were present in the databases. Thus, in the present study we annotated the genome for T. indicae-seudeticae (Mucorales: Lichtheimiaceae). The pipeline MAKER was used to annotate the previously unannotated genome of T. indicae-seudeticae (GenBank accession number JSYX01.1). Since the transcriptome of this species was not available, we used the following data as evidence to support the annotation in the pipeline: (i) all proteins contained in the MEROPS protease database (downloaded 15. 7. 2016); (ii) all proteins of the Swissprot database (downloaded 15. 7. 2016); and (iii) the transcriptome of Lichtheimia ramosa (GenBank GCA_000945115.1), a related species that belongs to the same order. We used three gene predictors in the MAKER pipeline: (i) Semi-HMM-based Nucleic Acid Parser (SNAP) [18], bootstrap-trained within MAKER; (ii) unsupervised-trained GeneMark-ET [19] and (iii) Augustus [20] trained for Rhizopus oryzae.

Search for putative peptidases, thermal adaptation and analysis of the enzymatic profiles in fungi

We mined the proteomes of all investigated fungi for putative protease sequences using the BLAST against the peptidase database MEROPS [21] (http://merops.sanger.ac.uk/). The putative peptidases were classified according to their catalytic site and families by the MEROPS server. An analysis of similarity (ANOSIM) was performed to check for differences in the catalytic type composition between mesophilic and thermophilic species and we applied the Percentage of Similarity analysis (SIMPER) to identify which catalytic type contributes the most to the differences in the enzymatic profile. We conducted the same analyses to evaluate the difference in composition of peptidases families.

The percentage ratio of each type of amino acids and the percentage ratio of charged, polar and hydrophobic amino acids was calculated using the PEPSTATS utility included in the EMBOSS suite. We carried out a paired t-test to determine if single amino acid residues or groups contributed to significant differences between thermophilic and mesophilic species in the set of the whole proteins (114,946 and 102,521 proteins, respectively) and the set of peptidases (3340 and 3590 peptidases, respectively); thermotolerant species were not included in the analysis. All analyses were performed in Past v. 2.17c [22]. All results are presented as the changes from mesophilic to thermophilic species.

Selection of functional homologs and representative proteins from the subfamily A1A aspartic peptidase (AP)

We chose the A1A AP family because it is the most well characterized peptidase. The dataset was scrutinized for the presence of typical AP hallmarks defined as D[TS]G, Y, XXG, D[TS]G, and XXG (where X is any of the hydrophobic residues AFILMV). Sequences lacking any of the hallmarks were considered as non-functional homologues and excluded from further analysis. We made the first alignment manually by the catalytic motif D[TS]G site as described in Revuelta et al. [23]. The second alignment was performed by ClustalW [24].

We performed phylogenetic analysis to identify a cluster of functional sequences in MEGA7 [25]. The evolutionary history was inferred by using the Maximum Likelihood method based on the JTT matrix-based model [26]. Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Joining and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model, and then selecting the topology with superior log likelihood values.

From the initial tree a cluster with 12 protein sequences was selected. This cluster was composed of proteins from Aspergillus fumigatus, A. niger, Chaetomium globosum, C. thermophilum, Myceliophthora fergusii, M. sepedonium, M. thermophila, Myriococcum thermophilum, Rasamsonia byssochlamydoides, Thermoascus crustaceus, Thielavia terrestris and T. australiensis.

Construction of three-dimensional models by homology

We selected homologous proteins of the family A1A AP as described above. Their amino acid sequences were used to build models using SWISS-MODEL (https://swissmodel.expasy.org/). The SWISS-MODEL template library (SMTL version 2016–09-07, PDB release 2016–09-02) was screened with Blast [27] and HHBlits [28] for evolutionary related structures matching the query sequence. We evaluated the accuracy of each predicted model and its stereo-chemical properties by PROCHECK [29]. The model was selected on the basis of various factors such as overall G-factor, number of residues in core allowed, generously allowed and disallowed regions in Ramachandran plot. We further analyzed the model by QMEAN [30], Lgscore in ProQ [31] and Z-score in ProSA [32].

The sequences were submitted to a bidimensional eletrophoresis in silico using the JVirGel 2.0 [33] to predict the theoretical pI (isoelectric point) and Mw (molecular weight). The number of α-helix, β-strand and β-sheet structures, number of cavities, superficial area and volume were estimated using Swiss-PdbViewer (http://www.expasy.org/spdbv/). We took into consideration that we observed differences in the number of cavities, so we included additional peptidases in the analysis. A total of 50 peptidases were randomly sampled from mesophilic (25) and thermophilic (25) species to the analysis. We tested all the above data with selected proteins from mesophilic and thermophilic species for significance using ANOVA (for continuous data) and Kruskal-Wallis (for counting data).

Results

Taxonomic insight into thermophilic fungi

We compared the genomes of heat-tolerant fungi and their phylogenetically related mesophilic counterparts. A total of 23 species (eight mesophilic, two thermotolerant and 13 thermophilic) were evaluated (Table 1). We inferred a phylogenetic tree using the fungal proteomes to evaluate the evolutionary relationship between the species (Fig. 1). The tree shows that some taxa are not grouped into a monophyletic group, such as Thermomyces and Chaetomium, indicating that species in these taxa do not belong to the same genus and their phylogenetic position should be reevaluated.

Fig. 1
figure 1

Phylogenetic tree of heat-tolerant fungi and phylogenetically related mesophilic counterparts based on fungal proteomes. Chi2-based branch supports are shown, calculated according to the approximate Likelihood-Ratio Test, as implemented in PhyML 3.3. Thermophilic species are in bold

Full size image

Peptidases found in fungal genomes

The total number of putative genes encoding for peptidases in the fungal genomes investigated in this study ranged from 241 to 820 (Table 2). Only Penicillium roqueforti genome contained all catalytic types (Serine, Aspartic, Metallo, Threonine, Cysteine, Glutamic and Asparagine). P. roqueforti has the largest number of putative peptidases (total of 820) followed by Talaromyces stipitatus (686), Rhizopus microsporus (652), Myceliophthora sepedonium (494), Chaetomium globosum (469), Rhizopus delamar (464), Aspergillus niger (437), Penicillium chrysogenum (397) and Mucor circinelloides (396). These fungi, excepted for R. microsporus, are classified as mesophilic (Table 2). Two species of the thermophilic genus Thermomyces, T. dupontii and T. lanuginosus, contained the smallest number of putative peptidases (241 and 246, respectively).

Table 2 Total number of genes encoding for peptidases from heat-tolerant and mesophilic fungal species
Full size table

Asparagine and Glutamic peptidases are not widely distributed among the genomes of fungi explored in the present study. For example Thermomyces stellatus and P. roqueforti were the only species that presented Asparagine peptidases while Glutamic peptidases were absent in the species belonging to Mucorales. We observed differences between thermophilic and mesophilic species even between closely related ones, e.g. between Myceliophthora fergusii, Myceliophtora thermophila and M. sepedonium (Table 1). The analysis of similarity (ANOSIM) showed that peptidase profiles of mesophilic and thermophilic species differ especially in the number of predicted peptidases (P = 0.0001, R = 0.7516). According to the percentage of similarity analysis (SIMPER) the overall peptidase profile between thermophilic and mesophilic species differed by 26.08% (the contribution of each catalytic type is shown in Additional file 1: Table S1).

The entire list of peptidases families and the number of homologous peptidases are shown in Additional file 1: Table S2. From the 138 families of peptidases found, nine are Aspartic peptidases, 32 are Cysteine, one is Glutamic, 52 are Metallo, two are mixed, two are Asparagine, 34 are Serine and six are Threonine. Considering the enzyme families, 11 families contributed to almost 50% of the total difference between thermophilic and mesophilic species (see Additional file 1: Table S3).

Putative adaptations to thermostability in peptidases

We evaluated the amino acid frequencies in both, the whole proteins in the genome and in the putative proteases. The comparison between the whole proteins in the datasets from mesophilic and thermophilic species showed significant changes in all single and groups of amino acids residues (Fig. 2a and c) in the direction from mesophilic to thermophilic species (P < 0.05). We observed an increase in the amino acids Ala, Glu, Gly, Pro, Arg and Val, while a decrease was observed in the other amino acids. The proteins from thermophilic species also showed an increase in charged and hydrophobic residues and a decrease in polar residues.

Fig. 2
figure 2

Comparison of amino acid composition between proteins (a and c) and peptidases (b and d) from mesophilic (white box plot) and thermophilic (grey box plot) fungi. *P < 0.05; **P < 0.01; ***P < 0.0001

Full size image

Regarding the peptidases, we noted the same pattern observed for all proteins of thermophilic species (P < 0.05), such as an increasing in the amino acids Ala, Glu, Gly, Pro, Arg and Val, an increase in charged and hydrophobic residues, and a decrease in polar residues (Fig. 2b and d). On the other hand, we observed a decrease in the amino acids Cys, His, Ile, Lys, Met, Asn, Gln, Ser, Thr, Trp and no differences were found for Asp, Phe, Leu, and Tyr residues. A detailed table with statistical data is available in Additional file 1: Table S4.

To evaluate characteristics in the three-dimensional structures of orthologous peptidases, we built 3D models of A1A AP peptidases, one of the most characterized family. They were evaluated via Procheck, QMEAN, ProSa (Z-score) and ProQ (Lgscore) and the values support the models shown in Additional file 1: Table S4. The stereo-chemical quality of the model structures showed that the majority of amino acids are in the most favored and additionally allowed favored regions (Additional file 1: Table S5). No significant differences were found in the number of α-helix, β-strand and β-sheet, superficial area, volume, molecular weight and isoeletric point. On the other hand, the number of cavities decreased in peptidases from thermophiles (P = 0.0185, Kruskal-Wallis test) (Table 3). Although the proteins presented differences in amino acid composition, the conformational structure is the same, maintaining the basic structure of the family A1A (Additional file 1: Figure S1).

Table 3 Characterization of the Aspartic Peptidase protein and the three-dimensional structure
Full size table

Due to the observed difference in the number of cavities, we evaluated additional peptidases from the family A1A (a total of 50 peptidases) to confirm this pattern. A significant difference (P = 0.0009441, Kruskal-Wallis test) was observed in the peptidases from mesophilic and thermophilic species (6.52 ± 2.08 and 4.44 ± 1.78, respectively), confirming our previous findings (Fig. 3).

Fig. 3
figure 3

Differences between the number of cavities in peptidases from mesophilic and thermophilic fungi. Different letters represent statistical difference (P = 0.0009441, Kruskal-Wallis test)

Full size image

Discussion

Genome streamlining by genome reduction has been reported to prokaryotic organisms as negatively correlated to growth temperature [34]. Thermophilic fungi have also experienced a genome reduction in response to thermal adaptation and consequently they lost many genes during their evolution [35], among them, the peptidases coding-genes, as shown for the first time in this study. Our results showed the largest reduction in the peptidases families with a higher number of genes, while those with fewer or single copies were less affected. The observed reduction is in contrast with the observations for cellulolytic enzymes, which were expanded in thermophilic fungal genomes but there was no mention regarding the peptidases-coding genes and how they are affected with this reduction [35].

Thermostable peptidases acting at high temperatures (65–85 °C) have already been applied in the baking, brewing, detergent and leather industries [36]. Thermophilic fungi are recognized as an interesting source of hydrolytic enzymes with industrial application, for example amylases, cellulases, hemicellulases, lipases and peptidases [9]. Despite the reduction in the number of copies of peptidases-coding genes, here we report a large catalogue of these enzymes, providing a good basis for further investigation and application.

A promising strategy to improve thermostability in proteins is the site-directed mutagenesis [37]. However, there is no consensus about the relationship of amino acid composition and its role in thermal adaptation. Increases in charged or hydrophobic residues, or both, are often reported, but their contribution to thermostability is still a topic of discussion [38].

Although we confirmed some of the previously observed changes in amino acid composition in our peptidases dataset (e.g. increased hydrophobic and charged residues), some of our observations differ from previous reports on thermal adaptations. These differences include observed increase in Trp [35, 39], lower frequency of Asp in eukaryotic proteins [35], an increase in Tyr and Ile and less Glu and Arg in M4 peptidases in prokaryotes species [40] all of which were not confirmed in our study. These observations suggest that while amino acid substitutions in thermoadaptation follow some general patterns, there are specific adaptations that differ between Archaea, Bacteria and Eukarya. These differences are observed even for different groups of proteins, as detected in the peptidases evaluated in this study when compared to the other proteins in the same genomes. It warrants the need to study thermoadaptation on a case-to-case basis.

The increase in Ala, Glu, Gly, Pro, Arg and Val content of peptidases from thermophilic fungi are in line with reports that some of these amino acids increase thermostability of proteins. They can improve the thermal stability by (i) forming a large number of electrostatic interactions (e.g. hydrogen bond and salt bridges), such as is the case of Glu and Arg [41, 42], (ii) increasing the rigidity of proteins, such as by the cyclic structure in the side chain of Pro [42], (iii) maintaining hydrophobic pockets, e.g. with Ala [43], or (iv) increasing the number of weak interactions, e.g. with Gly [44].

On the other hand, other amino acids that had a content decrease are known to reduce thermal stability, as Met and Asn, by the chemical instability of these residues at high temperatures [39]. Asn and Gln deaminate easily and Cys is susceptible to oxidation at elevated temperature [45]. Unless it is required for activity or formation of disulfide bonds, Cys is often absent from thermophilic proteins [45].

Gly and Pro residues have a major influence on the kinetics of loop formation in proteins. Glycine accelerates loop formation by decreasing the activation energy and it is known to contribute to conformational flexibility of polypeptide chains and to flexibility of some loops associated with enzymatic catalysis [46, 47]. Cis Prolyl shows the fastest kinetics of all sequences despite an increased activation energy [46]. The frequency of Pro in the modeled proteins was increased in the proteases from thermophilic fungi mainly in the loop areas. Although the increase of Pro is often seen in organisms with high GC content, in thermophilic fungi this content does not differ significantly between mesophilic and thermophilic species [35].

Although we observed significant differences between the amino acid composition of peptidases from thermophilic and mesophilic fungi, their predicted structures remained relatively unchanged. The only exception was the significant decrease in the number of cavities in peptidases from thermophiles. In addition to the observed reduction in the number of peptidases, we also observed the possible effect of natural selection. In this sense, there are two possible evolutionary scenarios: i) thermophilic fungi have lost peptidases with large number of cavities and kept only those that are compactly folded or ii) peptidases from thermophilic fungi were optimized to contain fewer cavities.

In general, cavities are considered as packing defects destabilizing the native structure [48]. The peptidases of thermophilic fungi present the same adaptation observed in thermophilic enzymes and it was interpreted as an adaptation for protein thermostability [39]. However, the low number of cavities was only observed for hyperthermophilic enzymes but in this study we report for moderate thermophilic enzymes as well.

Conclusions

Although thermophilic fungi present less genes encoding for peptidases, they have adaptations that could play a role in thermal resistance. These can occur from the genome to the protein structure level. Exploring the patterns that improve thermal stability in specific proteins can accelerate the process of finding species able to produce enzymes with the desired properties. This strategy combined with genome mining can drive the selection of target enzymes with characteristics indicating higher thermal stability. Moreover, this approach can find patterns to improve mesophilic proteins by site-directed mutagenesis for engineering enzymes adapted to high temperatures. Our results are not only of biotechnological interest but they also have an evolutionary appeal. In addition, the results prompt hypotheses on the structural differences related to temperature and stability that can be experimentally tested.