The obligate alkalophilic soda- lake fungus Sodiomyces alkalinus has shifted to a protein diet

Sodiomyces alkalinus is one of the very few alkalophilic fungi, adapted to grow opti-mally at high pH. It is widely distributed at the plant- deprived edges of extremely alkaline lakes and locally abundant. We sequenced the genome of S. alkalinus and reconstructed evolution of catabolic enzymes, using a phylogenomic comparison. We found that the genome of S. alkalinus is larger, but its predicted proteome is smaller and heavily depleted of both plant- degrading enzymes and proteinases, when compared to its closest plant- pathogenic relatives. Interestingly, despite overall losses, S. alkalinus has retained many proteinases families and acquired bacterial cell wall- degrading enzymes, some of them via horizontal gene transfer from bacteria. This fungus has very potent proteolytic activity at high pH values, but slowly induced low activity of cellulases and hemicellulases. Our experimental and in silico data suggest that plant biomass, a common food source for most fungi, is not a preferred substrate for S. alkalinus in its natural environment. We conclude that the fungus has abandoned the ancestral plant- based diet and has become specialized in a more


| INTRODUC TI ON
Soda (or alkaline) soils and lakes are the most alkaline natural habitats on Earth. With pH values typically ranging from 9 to 11, these environments are also often very salty-with high Na + concentrations. Soda soils are usually restricted to arid or semi-arid savanna inland areas interspersed with alkaline water basins called soda lakes, which fluctuate in size throughout the season due to evaporation and rain (Jones & Grant, 2000;Jones, Grant, Duckworth, & Owenson, 1998). A prominent reservoir of soda lakes and soils is located in the Western Siberia, Altai, Trans-Baikal areas (Russia), Mongolia and Africa, where Lakes Magadi and Natron have pH values approaching 12.
Extreme conditions of soda lakes and soils are not rendering them inhabitable as thought before, as an extensive diversity of microorganisms is adapted to thrive there (Grant & Sorokin, 2011;Zavarzin, 1993;Zavarzin, Zhilina, & Kevbrin, 1999). Fungi generally prefer acidic or neutral pH for optimal performance and were therefore not expected in saline soda lakes. However, soda lakes in Russia, Mongolia and Africa unexpectedly were found to harbour fungal species of different taxonomic lineages (Grum-Grzhimaylo, Debets, van Diepeningen, Georgieva, & Bilanenko, 2013;Grum-Grzhimaylo, Georgieva, Bondarenko, Debets, & Bilanenko, 2016;. Some of the recovered fungi not only tolerate high ambient pH, but even prefer that condition for optimal growth, a physiological category called alkalophiles (Horikoshi, 2011). The species that was identified in nearly every soda soil sample was Sodiomyces alkalinus-an obligate alkalophilic ascomycetous fungus, which is a member of the Plectosphaerellaceae family (Grum-Grzhimaylo, .
The high abundance of this fungus in such extreme environment is striking. First, the high pH makes growth of most fungi impossible.
Second, the edges of soda lakes are deprived of plant biomass, which is the dominant food source of most fungi. We hypothesized that S. alkalinus has specific adaptations to deal with those challenges and has changed the preference towards nonplant diet in this plantdeprived environment.
To test this hypothesis, the genome sequence of S. alkalinus was determined and examined for footprints of adaptation to the extreme conditions of soda lakes. The set of encoded carbohydrateactive and proteolytic enzymes was analysed in combination with direct bioassays under various conditions, to shed light on the preferred growth substrates of S. alkalinus. Our results support the shift in preference of the fungus from plants to alternative proteinrich substrates, such as small crustaceans and prokaryotes. Results obtained here not only improve our understanding of how fungi evolved to thrive under extreme natural conditions of soda lakes, but also provide opportunities to exploit S. alkalinus for commercial purposes, as a source of alkaline-active compounds for industrial use.

| Strains, media and growth
The Sodiomyces alkalinus ex-type strain F11 (CBS 110278) was used throughout the study and routinely propagated on the alkaline agar (AA) medium at 28°C (Grum-Grzhimaylo, . For pH-dependent growth experiments, we used appropriate inorganic buffers at final concentration 0.1 M to generate desired pH. We used Aspergillus fumigatus wild type (CBS 140053) and Aspergillus oryzae RIB40 as neutrophilic references. We tested various carbon sources and vitamin utilization on the minimal medium designed for S. alkalinus. For the enzyme assays, S. alkalinus was pregrown in liquid alkaline medium of the same content as AA, omitting agar. The collected pregrown mycelium was washed, filtered and inoculated in the liquid media, based on AA but with different pH values (6, 8 or 10) and substituting malt/yeast extracts for 1% carbon sources. For extended methods, see Supporting Information Appendix S1.

| SDS-PAGE and silver staining
We applied 20 μl of media extracts on 12% (w/v) SDS-PAGE topped with a 5% stacking gel using a MiniProtean II system (Bio-Rad), as described in (Laemmli, 1970). The gels were stained with silver nitrate to visualize the total proteins content (Blum, Beier, & Gross, 1987).

| Enzymatic assays
Culture filtrates after 48 hr of incubation were analysed for the selected enzyme activities involved in plant biomass degradation.
All enzymatic activities were measured in biological duplicates.
Each biological duplicate was measured in technical triplicates in 96-well plates at three pH values using 0.1 M buffers with pH 6, 8 and 10, and at 30°C. Activities were measured using either appropriate p-nitrophenyl (pNP) substrates (Sigma) or by the amount of reducing sugars released, using dinitrosalicylic acid (DNS) reagent (de Vries, Visser, & Graaff, 1999). One unit of enzymatic activity (U) was defined as the amount of enzymes that liberated 1 mmol of the corresponding product per minute of reaction, under assay conditions. The activities were expressed per volume of enzymatic protein-rich food, abundantly available in soda lakes in the form of prokaryotes and small crustaceans.

K E Y W O R D S
alkalophilic fungus, brine shrimps, enzymes, HGT, prokaryotes, Sodiomyces alkalinus solution (U/ml). Total proteolytic activity was measured with Pierce Fluorescent Protease Assay Kit (Thermo Fisher Scientific, USA) using a fluorescein isothiocyanate (FITC)-labelled casein assay according to the manufacturer's instructions. We tested the assay with the standard kit TBS buffer of pH 7.2, but also ensured the assay works properly with other buffers the yield pH 6, 8 and 10. One unit (U) of protease activity was defined as the amount of protein that has an equivalent activity of 1 μg bovine pancreas trypsin (Thermo Fisher Scientific) at pH 8. The activities were expressed per volume of enzymatic solution (U/ml). For extended methods, see Supporting Information Appendix S1.

| DNA and RNA extraction
Genomic DNA was extracted from mycelium of S. alkalinus grown on top of a cellophane membrane put onto AA medium. For genome sequencing, we obtained genomic DNA using the CTAB-based protocol from (Rogers & Bendich, 1988), with minor modifications. The quality and quantity of the DNA was verified on 0.6% agarose gels stained with EtBr, but also using a Nanodrop 2000 (Thermo Fisher Scientific)

| Genome sequencing, assembly and annotation
The S. alkalinus genome and transcriptome were sequenced using Illumina, the former in combination with fragment and long mate pair (LMP) libraries. All libraries were quantified using KAPA Biosystem's nextgeneration sequencing library qPCR kit and run on a Roche LightCycler 480 real-time PCR instrument. The quantified libraries were then prepared for sequencing on the Illumina HiSeq sequencing platform utilizing a truseq paired-end cluster kit, version 3, and Illumina's cBot instrument to generate a clustered flowcell for sequencing. Sequencing of the flowcell was performed on the Illumina HiSeq2000 sequencer using a truseq sbs sequencing kit, version 3, following a 2 × 100 and 2 × 150 indexed run recipe for LMP and fragments/transcriptome, respectively.
Genomic reads from two libraries were filtered and assembled with AllPathsLG (Gnerre et al., 2011). RNA-seq data were de novo assembled into consensus sequences using Rnnotator (Martin et al., 2010). Fungal genome was annotated using the JGI Annotation pipeline and made available via JGI fungal genome portal MycoCosm (http://genome.jgi.doe.gov/Sodal1; Grigoriev et al., 2014), and DDBJ/ENA/GenBank database. For extended methods, see Supporting Information Appendix S1.

| Proteinases
Putative proteinases were found by two successive searches against merops databases (Rawlings, Waller, Barrett, & Bateman, 2014). At first, predicted proteomes of 32 fungi were used as queries for blastp against full merops database (version 12) to remove false positives (Rawlings & Morton, 2008). Then, the output was used as queries for blastp against a smaller merops Scan database of type domains. In both searches, e-value was set to 0.0001. Secretory peptidases were predicted with TargetP 1.1b under default settings (Emanuelsson, Brunak, von Heijne, & Nielsen, 2007).

| qRT-PCR
To track the expression profiles of the lysozyme GH25 (JGI ID 341929), DD-peptidase (JGI ID 350999) and racemase (JGI ID 322460) across various pH, we used RNA isolated from the mycelium grown in liquid (2 days old) at various pHs in biological triplicates. 1 μg of genomic RNA was converted into cDNA with iScript™ cDNA Synthesis Kit (Bio-Rad). The qPCR final mix volume was 8 μl and contained 4 μl 2x iQ SYBR Green SuperMix (Bio-Rad), 0.16 μl 10 μM of each primer (for the primer list, see Supporting Information Table S3), 0.68 μl MQ water and 3 μl of cDNA. Each reaction was run in technical triplicates, in a Bio-Rad CFX96 thermocycler. act1 gene was used as a reference housekeeping gene, and annealing temperature for the qRT-PCRs was set to 61°C (primer efficiencies were verified in pilot runs). The technical triplicates were averaged and used for calculating C t difference relative to the lowest pH point. The ΔΔC t method was employed to calculate relative gene expression. qRT-PCR results were analysed by the bio-r ad cf x manager version 2.0 software. Baseline threshold line was set arbitrarily at exponential phase of PCR (500 relative fluorescence units), and results were normalized to the act1 gene signal.

| DD-peptidase deletion
The strategy for the knockout was based on a double recombination event with the targeted incorporation of a hygromycin resistance cassette disrupting the gene of interest, as described by (Kars, McCalman, Wagemakers, & van Kan, 2005). In brief, two DNA fragments, flanking the ORF of the DD-peptidase, were amplified and fused by overlap PCR with the hph cassette, carrying a gene conferring resistance to hygromycin B (hygB, Formedium). The cassette was introduced into the protoplasts of S. alkalinus by PEG-mediated transformation as in (Kars et al., 2005) and (ten Have, Mulder, Visser, & van Kan, 1998) with a few adjustments. HygB-resistant colonies were subjected to a single-spore bottleneck to ensure a homokaryotic transformant. The putative knockouts were verified with PCR experiments, ensuring the deletion of the DD-peptidase. For extended methods, see Supporting Information Appendix S1 and Supporting Information Table S3.

| Horizontal gene transfer detection
The method described in Marcet-Houben and Gabaldón (2010) was used for the detection of putative horizontal gene transfer (HGT) cases. Shortly, if protein sequences are present in a few fungi as well as in a high number of bacterial genomes, this gene is assumed to have undergone HGT. To identify these cases, we blasted (blastp, e-values 10 −5 , 10 −10 , 10 −15 , 10 −20 , 10 −30 , 10 −50 ) the predicted proteome of S. alkalinus against proteomes (248 fungal and 6,321 bacterial proteomes) obtained from the nonredundant database in NCBI as of April 2017. We tagged a given protein to be horizontally transferred if it fell in the quadrant limited by first 35-quantile for fungal occurrences and 100*SE from loess fit in bacterial distribution. Substantial intersection of the proteins that we obtained after analyses ran at various e-values strengthens our results. We considered an analysis ran at the e-value of 10 −20 in our results. For details, see Supporting Information Appendix S1.

| Growth in situ and in vitro
Sodiomyces alkalinus is an obligate alkalophilic fungus that was iso-

| Genome statistics
The

| Carbohydrate-active enzymes in silico and in vitro
Fungi decompose complex organic carbon compounds externally and use it as an energy source.  in GH18 and CE4 members was observed in both alkalophiles (Supporting Information Table S2). Interestingly, we detected acquisition of lysozymes, which are enzymes involved in degradation of peptidoglycan, the major polysaccharide of bacterial cell walls.
To substantiate the in silico conclusions, a synthetic minimal medium (MM, pH 10) was developed for S. alkalinus and used to test its growth on different carbon sources (Figure 3). In the process of developing MM, deficiencies for two vitamins were detected in S. alkalinus-biotin and thiamine (Supporting Information Figure S2). Since vitamins are exclusively of biogenic origin, an inability to produce

| Sodiomyces alkalinus produces weak but alkaline-active cellulases and hemicellulases
To obtain experimental evidence for the plant polysaccharides hydrolytic capabilities of S. alkalinus, enzyme assays were performed on the crude media extracts after 2 days of growth and compared to neutrophilic distantly related fungus A. oryzae. Extracts were obtained from three types of media at pH values of 8 and 10 (pH 6 for A. oryzae) that differ nutritionally-wheat bran (rich in arabinoxylans, cellulose,

| Proteinases in the genome of Sodiomyces alkalinus
Along with CAZymes, proteinases can play a major role in nutrition of fungi (Yike, 2011). To address the evolution of proteolytic enzymes in S. alkalinus, we searched for the genes that encode putative proteinases and compared them with other fungi. Similarly to CAZymes, we found that the genome of S. alkalinus (and sister A. alcalophilum) has lost about 40%-55% of its total and secreted proteinases suits when compared with its closest relatives, plant pathogenic Verticillium and Plectosphaerella species (Figure 2). Most dramatic losses were observed for the serine-type peptidases (S-clan, Rawlings et al., 2014) such as subtilisins (S08), S09 family peptidases and prolyl aminopeptidases of the family S33 (Supporting Information Table S6).
Despite overall losses, many peptidases from the C-, M-and T-clans

| Sodiomyces alkalinus produces strong alkaline proteases
To directly assess proteolytic capabilities of S. alkalinus, we measured activity of proteases in crude media extracts obtained after growing S. alkalinus on various substrates at different pH. The crude media samples of S. alkalinus displayed very potent protease activities. The highest induction was observed on wheat bran at pH 8 and 10 ( Figure 4).
Less protease activity was detected on sugar beet pulp at pH 10, whereas at pH 8 protease activities were not induced on this substrate.
Interestingly, the chitin medium also induced proteolytic activity. Total proteolytic activity had its optimum at pH 8 and retained about 80% of its activity at pH 10. A. oryzae also produced proteases that were active at high pH, but the activity was 17 times lower than in S. alkalinus. The very active and quickly induced peptidases coupled with global losses of peptidase-encoding genes and retention of certain families of peptidases are consistent with the hypothesis of specialization of S. alkalinus towards narrow diversity of protein substrates found at soda lakes.
Our data support the view that protein-rich small crustaceans (or their eggs), and/or prokaryotes, present in bulk at soda lakes (Figure 1a-c,e), can be the primary food source for S. alkalinus. This hypothesis was corroborated by good growth of the fungus on an array of proteinaceous substrates as sole carbon source in the minimal medium, including crude brine shrimp eggs (Supporting Information Figures S6 and   S7). The bacterial cell extracts did not yield abundant growth probably due to low substrate concentration, but the difference with no carbon source control could still be seen, indicating S. alkalinus can extract energy from bacterial substrates as well.

| Horizontally transferred prokaryotic genes into Sodiomyces alkalinus
Horizontal gene transfer (HGT; also known as lateral gene transfer) is defined as a relocation and stable integration of genetic material between the species' genomes (Doolittle, 1999). It has been shown that fungi are capable of acquiring genes from both prokaryotic and eukaryotic donors (Ros & Hurst, 2009;Slot & Rokas, 2011). Given the extremophilic lifestyle of S. alkalinus and the abundance of prokaryotes from where this fungus was isolated, we searched for evidence of HGT into the genome of S. alkalinus. The distribution of protein occurrences at various e-values among the sampled bacterial (n = 6,321) and fungal (n = 248) predicted proteomes is displayed in Supporting Information Figure S8 with the nine potential HGT cases (e < 10 −20 ). All detected HGTs involve structural genes, encoding enzymes that probably have a narrow catalytic mode. None of the acquired genes are unique to S. alkalinus, as they were found in other fungi as well, often displaying a mosaic distribution, which suggests multiple gains and losses. The acquired genes are involved in several processes: protein and amino acid metabolism, bacterial cell metabolism and some others. The list is enriched for genes (4 out of 9 genes) that encode proteins involved in amino acid and protein metabolism providing an extra line of evidence that these substrates are relevant to S. alkalinus.

| Bacterial DD-peptidase, racemase and lysozyme in Sodiomyces alkalinus
The putative bacterial cell wall-degrading enzymes, such as DDpeptidase and amino acid racemase, are acquired by HGT and characteristic to both alkalophilic fungal species, S. alkalinus and A. alcalophilum, and are not present in the sister group. To address whether these enzymes may have functional significance in the natural habitat of S. alkalinus, the expression of these genes on rich medium at different pHs was quantified. Additionally, we included peptidoglycanase (lysozyme) of the family GH25 into this analysis.
All three genes showed strong upregulation (fivefold to 15-fold) at pH 10, compared to acidic pH and the lysozyme GH25 showed an extra peak of upregulation at neutral pH (Supporting Information Figure S9). A DD-peptidase knockout mutant, constructed by homologous recombination-mediated replacement by an antibiotic resistance cassette, did not produce obvious phenotypic alterations and showed no defects on MM at pH 10 when grown on glucose.
On the proteinaceous carbon sources, the mutant showed healthier looking edges of the colonies, a phenomenon which we cannot explain at the moment (Supporting Information Figures S6 and S7).
Nonetheless, the results suggest that the DD-peptidase is not essential for the performance at high pH on synthetic media supplemented with these substrates. However, when grown with bacterial cell extract as a sole carbon source, the DD-peptidase knockout Although alkaline proteases were detected in A. oryzae, their relevance remains questionable since the fungus cannot sustain growth at high pH. Conversely, the production of alkaline proteases in S. alkalinus seems to be in line with its natural environment. Retention of some aminopeptidase families in the genome of S. alkalinus may also be linked to the ecology of the fungus, as these enzymes tend to have their optimum at neutral or alkaline pH (Pel et al., 2007). We did not detect expansions of chitinases in the genome of S. alkalinus, but the present arsenal may be sufficient to degrade the exoskeleton of brine shrimps and reach their proteinaceous interior. Gains of lysozymes suggest an enhanced capability for the degradation of bacteria. As bacterial cells on average contain about 50% of proteins by dry weight (Neidhardt, 1963), S. alkalinus may have adapted to bacteria as a source of food. The constant physical proximity to brine shrimps with eggs and vitamin-producing prokaryotes at soda soils may also explain vitamin deficiencies that we detected in S. alkalinus, since complementation via diet is possible. Very potent proteases of S. alkalinus can also indicate a strong need for nitrogen in natural environments of soda lakes, as this element becomes limiting at high pH condition due to losses in the form of ammonia gas (NH 3 ).
Evolution of very active proteases in S. alkalinus can be additionally governed by strong competition with prokaryotes for the easily accessible nitrogen-rich protein substrates.
To follow the notion that bacteria could have influenced the evolution of S. alkalinus in soda soils, we searched for genes that may have been horizontally acquired from bacteria. The documented cases of HGT into fungi suggest that some ecologically specialized organisms are rich in horizontally transferred genes (e.g., Schönknecht et al., 2013). The exact mechanisms of HGT are poorly understood, but it is clear that physically intimate and continuous association of microorganisms greatly enhance the chances for gene transfer (Keeling & Palmer, 2008). However, the magnitude of HGT into S. alkalinus from bacteria was similar to that estimated for other fungi (Marcet-Houben & Gabaldón, 2010), indicating no elevated rate of foreign gene acquisition, as one might expect given the extreme environmental niche of this fungus. Despite average HGT rate, some of the transferred genes seem significant as they encode enzymes that can be beneficial under soda-lake conditions. For instance, the acquired DD-peptidase and amino acid racemase most probably act on bacterial cell wall compounds facilitating their decomposition. In the current study, we were able to confirm the bacteria-degrading function at least for the DD-peptidase. Interestingly, this enzyme was also found in the genome of a social amoeba Dictyostelium discoideum, also hypothesized to have entered by HGT, and proposed to facilitate degradation of bacterial cell walls inside the amoeba cells (Eichinger et al., 2005). This is a curious example of how unrelated organisms, a fungus and a slime mould, acquired similar enzymes through HGT to meet similar needs.
The genome of S. alkalinus harbours a gene cluster that encodes the core enzymes required for the biosynthesis of betalactam antibiotics (e.g., penicillins, cephalosporins; van den Berg et al., 2008). This three-gene cluster (pcbC, pcbAB, penDE) is also present in the sister alkalophilic species Acremonium alcalophilum, but not in other closely related fungi. It remains to be demonstrated whether S. alkalinus is capable of producing antibiotics in vivo and if such production would provide any benefit to S. alkalinus in its natural habitat, as beta-lactams rapidly degrade at high pH (Deshpande, Baheti, & Chatterjee, 2004). We are not excluding the possibility that S. alkalinus can produce structurally different antibiotics, which retain stability and activity at high pH, as well as other alkaline metabolites that can be of commercial interest.