Class-II dihydroorotate dehydrogenases from three phylogenetically distant fungi support anaerobic pyrimidine biosynthesis

In most fungi, quinone-dependent Class-II dihydroorotate dehydrogenases (DHODs) are essential for pyrimidine biosynthesis. Coupling of these Class-II DHODHs to mitochondrial respiration makes their in vivo activity dependent on oxygen availability. Saccharomyces cerevisiae and closely related yeast species harbor a cytosolic Class-I DHOD (Ura1) that uses fumarate as electron acceptor and thereby enables anaerobic pyrimidine synthesis. Here, we investigate DHODs from three fungi (the Neocallimastigomycete Anaeromyces robustus and the yeasts Schizosaccharomyces japonicus and Dekkera bruxellensis) that can grow anaerobically but, based on genome analysis, only harbor a Class-II DHOD. Heterologous expression of putative Class-II DHOD-encoding genes from fungi capable of anaerobic, pyrimidine-prototrophic growth (Arura9, SjURA9, DbURA9) in an S. cerevisiae ura1Δ strain supported aerobic as well as anaerobic pyrimidine prototrophy. A strain expressing DbURA9 showed delayed anaerobic growth without pyrimidine supplementation. Adapted faster growing DbURA9-expressing strains showed mutations in FUM1, which encodes fumarase. GFP-tagged SjUra9 and DbUra9 were localized to S. cerevisiae mitochondria, while ArUra9, whose sequence lacked a mitochondrial targeting sequence, was localized to the yeast cytosol. Experiments with cell extracts showed that ArUra9 used free FAD and FMN as electron acceptors. Expression of SjURA9 in S. cerevisiae reproducibly led to loss of respiratory competence and mitochondrial DNA. A cysteine residue (C265 in SjUra9) in the active sites of all three anaerobically active Ura9 orthologs was shown to be essential for anaerobic activity of SjUra9 but not of ArUra9. Activity of fungal Class-II DHODs was long thought to be dependent on an active respiratory chain, which in most fungi requires the presence of oxygen. By heterologous expression experiments in S. cerevisiae, this study shows that phylogenetically distant fungi independently evolved Class-II dihydroorotate dehydrogenases that enable anaerobic pyrimidine biosynthesis. Further structure–function studies are required to understand the mechanistic basis for the anaerobic activity of Class-II DHODs and an observed loss of respiratory competence in S. cerevisiae strains expressing an anaerobically active DHOD from Sch. japonicus.

Class I-A and Class-II DHODs contain a single FMN cofactor per subunit. In heterotetrameric Class I-B enzymes, two subunits also each contain an FMN cofactor, while the other two both contain an FAD cofactor and an [2Fe-2S] cluster [4]. Class-I DHODs contain an active-site cysteine residue involved in deprotonation of C5 of dihydroorotate, while Class-II enzymes have a serine in the same position acting as the catalytic base [9][10][11][12][13].
Most eukaryotes harbor a monomeric Class-II DHOD that donates electrons to the quinone pool of the mitochondrial respiratory chain [8,[14][15][16]. Bacterial Class-II DHODs have an N-terminal sequence that localizes them to the inside of the cytoplasmic membrane, whereas eukaryotic enzymes are targeted to the outside of the mitochondrial inner membrane [5]. This mitochondrial targeting also applies to fungal Class-II DHODs, which include the Ura9 orthologs of yeasts such as Lachancea kluyveri and Schizosaccharomyces pombe [17]. Since respiration in yeasts requires oxygen as electron acceptor, reliance on these respiration-coupled enzymes precludes pyrimidine-prototrophic anaerobic growth [8].
Class-I DHODs predominantly occur in gram-positive bacteria and Archaea [18] but are also found in a small number of yeasts, including Saccharomyces cerevisiae and closely related species [19,20]. S. cerevisiae is among the few yeast species that are able to grow under strictly anaerobic conditions [21]. ScUra1, a Class-IA, fumarate-coupled DHOD, enables S. cerevisiae to synthesize pyrimidines in the absence of oxygen [19,20]. A small number of other Saccharomycetes, including Kluyveromyces lactis and L. kluyveri, harbor Ura1 as well as Ura9 orthologs [19,22,23]. Based on sequence similarity of yeast ScURA1 orthologs with Lactococcus genes, they are assumed to have been acquired by horizontal gene transfer [15,19].
In line with a proposed essentiality of ScURA1 orthologs for anaerobic pyrimidine synthesis by yeasts [19,24], replacement of ScURA1 by a Class-II DHOD gene from L. kluyveri (LkURA9) or Sch. pombe (SpURA3) yielded strains that were only pyrimidine prototrophic under aerobic conditions [8,19]. Conversely, replacement of ScURA1 by LkURA1 supported aerobic as well as anaerobic pyrimidine prototrophy [19]. Introduction of ScURA1 in URA9dependent yeasts was proposed as a metabolic engineering strategy for enabling anaerobic, pyrimidine-prototrophic growth of yeasts lacking a native ScURA1 ortholog [25].
The long-held assumption that expression of a Class-I DHOD is required for anaerobic pyrimidine biosynthesis in eukaryotes was first challenged when Dekkera bruxellensis, which only harbors a putative URA9 ortholog, was shown to grow anaerobically in pyrimidine-free media [26,27]. A hypothesis that DbUra9 is able to use a non-quinone electron acceptor [17,28,29] was, however, not experimentally tested.
We recently observed that D. bruxellensis may not be the only eukaryote in which anaerobic pyrimidine synthesis involves a Class-II DHOD. Inspection of the genome of the fission yeast Sch. japonicus, which shows fast anaerobic growth in synthetic media without uracil [30,31], suggested that it only contains a URA9 ortholog. Moreover, genomes of Neocallimastigomycetes, a group of deepbranching, obligately anaerobic fungi that lack mitochondria and instead harbor hydrogenosomes [32,33], also appeared to lack orthologs of soluble Class-I DHOD.
The goals of the present study were to investigate whether URA9 orthologs in eukaryotes capable of anaerobic growth indeed support anaerobic pyrimidine biosynthesis, and to gain more insight into underlying mechanisms and tradeoffs. To this end, we expressed putative Class-II DHOD genes from the obligately anaerobic Neocallimastigomycete Anaeromyces robustus (Arura9), the facultative anaerobes Sch. japonicus (SjURA9) and D. bruxellensis (DbURA9), as well as from the oxygen-requiring yeasts Ogataea parapolymorpha (OpURA9) and Kluyveromyces marxianus (KmURA9), in an S. cerevisiae ura1Δ background. After studying aerobic and anaerobic growth of the resulting strains in uracil-supplemented and uracil-free synthetic media, we analyzed subcellular localization of Ura9-eGFP fusion proteins in S. cerevisiae and assessed the impact of a conserved amino-acid substitution in anaerobically functional Ura9 orthologs. To identify possible natural electron acceptors, we performed enzyme assays in cell extracts of an S. cerevisiae strain expressing ArUra9 and resequenced the genomes of laboratory-evolved S. cerevisiae strains whose anaerobic growth depended on expression of DbURA9. We found that instead of quinone, ArUra9 uses free flavins as electron acceptors and that expression of SjURA9 in S. cerevisiae results in loss of respiration.

Obligately anaerobic Neocallimastigomycetes and facultatively anaerobic yeasts harbor putative Class-II DHODs
A preliminary exploration of a set of selected fungal proteomes for Class-I and Class-II DHODs was based on a sequence similarity search with Class-I and Class-II enzymes of L. kluyveri (LkUra1 and LkUra9, respectively; [19]; Table 1) as queries. Consistent with earlier studies, the S. cerevisiae proteome showed a single sequence with strong similarity to LkUra1 (ScUra1; [10,19,23]), while proteomes of the facultatively fermentative yeasts O. parapolymorpha and K. marxianus, which both require oxygen for biosynthesis-related reactions [34,35], yielded previously described Class-II DHOD sequences with strong homology to LkUra9 (OpUra9 and KmUra9, respectively; [23,36]). K. marxianus additionally showed a sequence with high sequence similarity to LkUra1 (KmUra1; [22]). The search also confirmed earlier reports that the facultatively anaerobic yeast D. bruxellensis only harbors a putative DHOD sequence with high similarity to LkUra9 (DbUra9; [17,23,28]).
Sequence comparison with LkUra1 provided no evidence for presence of a Class-I DHOD in the Neocallimastigomycetes Piromyces finnis, Neocallimastix californiae and A. robustus. Instead, these obligately anaerobic fungi yielded predicted protein sequences with high similarity to the Class-II DHOD LkUra9 (Table 1), which we tentatively called PfUra9, NcUra9 and ArUra9, respectively. A similar result was obtained for the facultatively anaerobic fission yeast Sch. japonicus [30,31], whose putative Class-II DHOD was tentatively named SjUra9 ( Table 1). Dependence of anaerobic, pyrimidineprototrophic growth of phylogenetically distant fungi on Class-II ('Ura9') DHODs would be remarkable in view of the well-documented coupling of canonical eukaryotic Class-II DHODs to aerobic mitochondrial respiration [5,8,19,37,38].
To study the phylogeny of fungal Ura9 orthologs, the amino-acid sequence of LkUra9 was used as query for a sequence analysis with a hidden Markov model method (HMMER; [39]) against all fungal proteomes in Uniprot [40], from which the best hits were used for orthology prediction (see "Methods"). A similar strategy was performed to obtain bacterial LkUra9 orthologs. A phylogenetic tree constructed based on the resulting 331 fungal and 73 bacterial Ura9 orthologs showed a clear separation of bacterial and fungal Ura9 orthologs ( Fig. 1; Additional file 1, Additional file 2, Additional file 3). The topology of the tree followed fungal phylogeny, as exemplified by the LkUra9 ortholog of Sch. japonicus, which clustered with that of closely related species Sch. pombe whose Class-II DHOD was previously shown not to support anaerobic uracil prototrophy [8]. Similarly, the D. bruxellensis Ura9 ortholog clustered with those of other Pichiacaea yeasts, including O. parapolymorpha whose Ura9 ortholog does not support anaerobic uracil prototrophy (see below). These results indicated that, if Ura9 orthologs of Neocallimastigomycetes, D. bruxellensis and Sch. japonicus share properties that enable anaerobic pyrimidine synthesis, these are likely to have evolved independently, without involvement of horizontal gene transfer (HGT). Evolution of oxygen-independent pyrimidine synthesis in these fungi would then differ from the proposed HGT-mediated acquisition of a respirationindependent Class-I DHOD by an ancestor of S. cerevisiae [19,20] and from proposed HGT-driven adaptations of Sch. japonicus [31,41] and Neocallimastigomycetes [42,43] to bypass other oxygen requirements for biosynthetic processes.

Heterologous URA9 genes complement aerobic pyrimidine auxotrophy of ura1Δ S. cerevisiae
To assess functionality of the putative Class-II DHOD genes of the Neocallimastigomycete A. robustus and  while the congenic ura1Δ strain IMK824 only grew on SMUD + ura (0.30 h −1 , Fig. 2, Additional file 5: Table S1). The lower specific growth rate of the uracilauxotrophic strain IMK824 on SMUD + ura probably reflected a growth-limiting rate of uracil uptake [44]. S. cerevisiae ura1Δ strains carrying expression cassettes for Arura9, DbURA9, KmURA9 or OpURA9 all grew aerobically on SMUD + ura and SMUD, at specific growth rates of 0.34 h −1 to 0.35 h −1 . A ura1Δ strain expressing SjURA9 grew almost two-fold slower (0.17-0.19 h −1 ) on SMUD than the other uracil-prototrophic strains (Fig. 2, Additional file 5: Table S1). The growth rate of the SjURA9-expressing strain on SMUD + ura also was approximately one-third lower than that of the ura1Δ strain IMK824, indicating that expression of SjURA9 negatively affected aerobic growth of S. cerevisiae on glucose. These results showed that ArUra9, SjUra9 and DbUra9 are functional DHODs that complement a ura1Δ mutation in S. cerevisiae under aerobic conditions. A similar complementation of S. cerevisiae ura1 null mutants was previously demonstrated for Class-II DHODs of the oxygen-requiring yeasts L. kluyveri (LkUra9 [19]) and Sch. pombe (SpUra3 [8]).

Class II-DHODs of A. robustus, D. bruxellensis and Sch. japonicus support anaerobic growth of an S. cerevisiae ura1Δ strain
A. robustus, D. bruxellensis and Sch. japonicus were previously reported to grow anaerobically in synthetic media without pyrimidine supplementation [26,27,30,45]. While, based on genome-sequence analysis, previous studies [17,23,28,29] inferred that anaerobic pyrimidine prototrophy of D. bruxellensis was supported by a Class-II DHOD (DbUra9), its activity was not experimentally confirmed. To assess anaerobic functionality of ArUra9, DbUra9 and SjUra9, we investigated whether these Class-II DHODs could support anaerobic, pyrimidine-prototrophic growth of ura1Δ S. cerevisiae strains.
In contrast to expression of URA9 orthologs from oxygen-requiring yeasts, expression of Arura9 in S. cerevisiae supported fast anaerobic growth (0.25 h −1 ) on SMUD (Fig. 2, Additional file 5: Table S1). An S. cerevisiae strain in which ScURA1 was replaced by a SjURA9 expression cassette also grew anaerobically on SMUD as well as SMUD + ura, at an approximately two-fold lower growth rate than that of the strain expressing Arura9 (Fig. 2, Additional file 5: Table S1).
Strain IMI439, in which DbURA9 replaced ScURA1, did not show anaerobic growth on SMUD during the first 30 h of incubation (Additional file 5: Figure S1). When, after 68 h, OD 600 had increased to 2.9, the strain was transferred to fresh SMUD, resulting in immediate anaerobic growth. In a parallel experiment with strain IMI447, which expressed KmURA9, no growth was observed upon transfer to fresh SMUD.
These results indicated that Class-II DHODs from the Neocallimastigomycete A. robustus and from the facultatively anaerobic yeast Sch. japonicus support pyrimidine synthesis under anaerobic conditions. The delayed anaerobic growth of a ura1Δ strain expressing DbURA9 suggested that anaerobic functionality of DbURA9 in S. cerevisiae requires physiological or genetic adaptations.

Fast anaerobic, uracil-prototrophic growth of DbUra9-dependent S. cerevisiae strains correlates with mutations in FUM1
To further explore anaerobic functionality of DbUra9, we investigated whether adaptation of S. cerevisiae IMI439 (ura1Δ::DbURA9) to anaerobic, pyrimidineprototrophic growth (Additional file 5; Figure S1) was associated with mutations in the DbURA9 expression cassette or in native S. cerevisiae genes. Two independent anaerobic cultures of this strain on SMUD were incubated until growth occurred and then transferred to fresh anaerobic SMUD, in which they instantaneously grew (Additional file 5: Figure S1). Upon reaching stationary phase, two single colonies were isolated from each culture. To check whether these isolates (IMS1167, IMS1168, IMS1169 and IMS1170) had acquired stable mutations that stimulated anaerobic, uracil-prototrophic growth, they were first grown under non-selective conditions (aerobic growth on SMUD + ura) and then transferred to anaerobic medium with a reduced uracil content (SMUD + ura0.1). Inoculation of anaerobic cultures on SMUD with the resulting anaerobic, pyrimidine-limited cultures resulted in instantaneous anaerobic growth of all four isolates (Additional file 5: Table S1). Comparison of whole-genome sequences of the four isolates with that of their parental strain IMI439 (ura1Δ::DbURA9) revealed no mutations in the DbURA9 expression cassette. Strains IMS1167 and IMS1168, which originated from the same anaerobically adapted culture, both contained point mutations in VPS1 (VPS1 I410L ), which encodes a dynamin-like GTPase involved in vacuolar sorting and in FUM1 (FUM1 M432I ), encoding fumarase ( Table 2). Strains IMS1169 and IMS1170, which were isolated from the second anaerobically adapted culture, each harbored a different mutation in FUM1 (FUM1 A294V and FUM1 T218M , respectively; Table 2). These results indicated that anaerobic activity of DbUra9 in S. cerevisiae does not require changes in its protein sequence but, based on the presence of mutations in FUM1 in all four isolates, may depend on the intracellular fumarate concentration.

A cysteine residue in the active site of SjUra9 is required for activity under anaerobic conditions
To identify potentially relevant differences in the aminoacid sequences of Ura9 orthologs from oxygen-requiring yeast strains and those of anaerobically functioning Ura9 enzymes, sequences of ArUra9, SjUra9, DbUra9, KmUra9 and OpUra9 were subjected to a multiple sequence alignment, along with those of the characterized Class II DHODs of L. kluyveri (LkUra9l), Sch. pombe (SpUra3), and DHOD sequences of the Neocallimastigomycetes N. californiae (NcUra9) and P. finnis (PfUra9) (Additional file 5: Figure S2). In comparison with the yeast Ura9 sequences, those of the three Neocallimastigomycetes showed a 76-81 amino-acid truncation N-terminal truncation. In canonical fungal Ura9 enzymes, the N-terminus contains a mitochondrial targeting sequence [5,38] and is proposed to be involved in quinone binding [46,47].

Table 2
Mutations in DbURA9 expressing strains, evolved for anaerobic pyrimidine prototrophy Two cultures of IMI439 (ura1Δ::DbURA9) were independently evolved under anaerobic conditions on SMUD (Additional file 5: Figure S1). Two single colony isolates from each flask were subjected to whole-genome resequencing and predicted amino-acid substitutions were only found in Vps1 and Fum1 The Neocallimastigomycete Ura9 sequences as well as those of the two facultatively anaerobic yeasts (DbUra9 and SjUra9) contained a cysteine residue instead of the conserved serine residue that acts as catalytic base in canonical Class-II DHODs [11,12,48]. Of 331 fungal Ura9 orthologs (Additional file 1) only three additional proteins (from Coemansia reversa, Smittium culisis and Gonapodya prolifera) harbored a cysteine at this position, but did not show an N-terminal truncation (Additional file 1 and Additional file 5: Figure S2).

Evolution line Strain Mutations
Since soluble fumarate-and NAD + -dependent Class-I DHODs also use a cysteine residue as catalytic base [12,13], we investigated the relevance this residue for in vivo activity of ArUra9 and SjUra9 in S. cerevisiae. Because of the slow adaptation to anaerobic, uracil-prototrophic growth of IMI439 (ura1Δ::DbURA9), we did not include DbUra9 in these experiments. Point mutations were introduced in Arura9 (C168S) and SjURA9 (C265S) to change the cysteine codon for a serine, yielding strains IMG007 (ura1Δ::Arura9 C168S ) and IMG008 (ura1Δ::SjURA9 C265S ). In addition, a point mutation in the corresponding serine codon of KmURA9 was introduced to change it to a cysteine, yielding strain IMG005 (ura1Δ::KmURA9 S263C ).
Changing the active-site serine residue (S263) of KmUra9 to a cysteine did not affect aerobic growth of S. cerevisiae, as shown by identical specific growth rates of strains IMG005 (ura1Δ::KmURA9 S263C ) and IMI446 (ura1Δ::KmURA9) on SMUD (Fig. 3, Additional file 5: Table S1). This result indicated that, at least in KmUra9, the serine catalytic base that is strongly conserved in canonical Class-II DHODs is not essential for activity under aerobic conditions. However, strain IMG005 did not grow anaerobically without uracil supplementation (Fig. 3, Additional file 5: Table S1), indicating that replacement of the catalytic-base serine residue of KmUra9 by a cysteine is not sufficient to enable anaerobic functionality in S. cerevisiae.
Replacing the active-site cysteine residue in ArUra9 by a serine residue did not cause a different aerobic or anaerobic growth rate on SMUD of strain IMG007 (ura1Δ::Arura9 C168S ) relative to its parental strain (Fig. 3, Additional file 5: Table S1). Apparently, despite our observation that an active-site cysteine residue occurred in different anaerobically active Ura9 orthologs, it is not required for anaerobic functionality of ArUra9 in S. cerevisiae. A strikingly different result was obtained upon changing the cysteine residue in the active site of SjUra9 to a serine. Under aerobic conditions, exponential growth of strain IMG008 (ura1Δ::SjURA9 C265S ) on SMUD was nearly two-fold slower than that of its parental strain IMI452 (ura1Δ::SjURA9). In contrast, strain IMG008 (ura1Δ::SjURA9 C265S ) failed to grow on SMUD under anaerobic conditions. These results indicated that the active-site cysteine residue in SjUra9, but not in ArUra9, is required for DHOD activity under anaerobic conditions.

Subcellular localization of heterologous Ura9 orthologs expressed in S. cerevisiae
To investigate subcellular localization of Ura9 orthologs, eGFP fusions of anaerobically active ArUra9, DbUra9 and SjUra9, as well as of OpUra9, were expressed from multicopy (mc) plasmids in an S. cerevisiae ura1Δ strain, followed by fluorescence-microscopy analysis of the resulting strains (Fig. 4). A co-expressed mRuby2 fluorescent protein fused to the preCOX4 mitochondrial targeting sequence [49] was used as marker for mitochondrial localization.
In strain IME602 (ura1Δ mcSjURA9-eGFP), mRuby2 fluorescence did not reveal the punctuate mitochondrial structures seen in the other strains. Instead, eGFP fluorescence was associated with tubular structures, that partially overlapped with a less defined mRuby2 fluorescence (Fig. 4D). Although elongated mitochondrial morphologies occur in fermenting S. cerevisiae cells [50], the diffuse mRuby2 fluorescence in strain IME602 did not allow for clear localization of SjUra9-eGFP. This strain was therefore also stained with the dye MitoTracker Deep Red. This approach only yielded vague tube-like structures or no fluorescence at all (Fig. 4E). Since staining by both preCOX4-MTS-mRuby2 and MitoTracker Deep Red depend on mitochondrial membrane potential [49], we hypothesized that expression of SjURA9 reduces or abolishes mitochondrial membrane potential, possibly as a consequence of a loss of respiratory capacity [51].

Expression of SjURA9 in S. cerevisiae causes loss of respiratory capacity and mitochondrial DNA
To investigate whether expression of SjURA9 causes loss of respiratory capacity, SjURA9-expressing S. cerevisiae strains were tested for their ability to grow on non-fermentable carbon sources. In contrast to the reference strain S. cerevisiae CEN.PK113-7D (Fig. 5A), three independently constructed strains expressing SjURA9 (IMI452, IMI462 and IME571; Additional file 5: Table S2) failed to grow on synthetic medium supplemented with ethanol and glycerol (SMEG, Fig. 5D, E, F, G, respectively). The inability of SjURA9-expressing strains to grow on SMEG resembled that of the respiratory-deficient strain IMK242 [52]. In contrast, strain IME603, which expressed ScURA1 from a multicopy plasmid, as well as ura1Δ strains expressing other URA9-expressing did grow on SMEG, (Additional file 5: Figure S3). Removal of the SjURA9 expression plasmid from strain IME571, yielding strain IMS1206, did not restore growth on these non-fermentable carbon sources (Fig. 5H). To explore whether the loss of respiratory competence in SjURA9-expressing strains was caused by loss of mitochondrial DNA, its presence was evaluated by staining with the DNA-specific dye 4′,6-diamidino-2-phenylindole (DAPI, [53]) and by whole-genome sequencing. The reference S. cerevisiae strain CEN.PK113-7D (Fig. 5A) and a strain expressing ScURA1 from a multicopy plasmid (IME603; Fig. 5B) showed punctuated extranuclear DAPI staining. The SjURA9-expressing S. cerevisiae strains IMI462, IMG008 and IME571 only showed fluorescent staining of nuclear DNA, while strain IMI452 (ura1Δ::SjURA9) showed vague DAPI-stained tubular structures. Whole-genome sequencing showed absence of mitochondrial DNA in all the four strains SjURA9expressing S. cerevisiae strains (Fig. 5). These observations indicated that expression of SjURA9 in aerobic S. cerevisiae cultures causes loss of mitochondrial DNA and respiratory capacity.

ArUra9 uses free FAD and FMN as electron acceptors
The absence of mitochondria in Neocallimastigomycetes and the cytosolic localization of ArUra9-eGFP in S. cerevisiae implied that its activity was unlikely to involve components of the mitochondrial respiratory chain. To identify possible natural electron acceptors of ArUra9, enzyme-activity assays were performed in cell extracts of strain IME569 (ura1Δ mcArura9). Cell extracts of strains IME603 (mcScURA1) and IMK824 (ura1Δ) were included as references.

Discussion
Fumarate-dependent Class-IA DHODs in S. cerevisiae (ScUra1) and closely related yeasts circumvent the oxygen requirement of respiration-dependent pyrimidine synthesis via the mitochondrial Class-II DHODs that occur in most other fungi [19,20,22]. The assumption that presence of a Class-IA DHOD is required for anaerobic pyrimidine prototrophy in fungi was first called into question when analysis of the genome of D. bruxellensis indicated absence of a Class-IA DHOD gene and, instead, revealed a sequence (DbURA9) with similarity to Class-II DHOD genes [17,23,28]. We identified similar situations in obligately anaerobic Neocallimastigomycetes and in the facultatively anaerobic fission yeast Sch. japonicus (Table 1, Fig. 1). Heterologous expression studies in ura1Δ S. cerevisiae showed that DbUra9, as well as orthologs from the Neocallimastigomycete A. robustus (ArUra9) and Sch. japonicus (SjUra9), supported anaerobic pyrimidine prototrophy.   This phenotype was not observed for ura1Δ S. cerevisiae strains expressing URA9 genes from the aerobic yeasts K. marxianus and O. parapolymorpha (Fig. 2) nor in similar experiments involving URA9 genes from other aerobic yeasts [8,19,25].
A cysteine as catalytic base is considered a hallmark of Class-I DHODs [57] while, with few reported exceptions, Class-II enzymes have a serine in this position [16]. Instead, the three anaerobically active Ura9 orthologs investigated in this study had a Cys in the catalyticbase position and shared this feature with only 3 of 331 other predicted fungal Ura9 orthologs. The latter three sequences originated from C. reversa, Sm. culicis and G. prolifera (Fig. 1), which were all originally isolated from microaerobic or anoxic environments (dung [62], insect guts [63] and submerged fruits [64], respectively). A Cysto-Ser change in the active site of Sch. japonicus SjUra9 specifically abolished its ability to support anaerobic pyrimidine prototrophy of S. cerevisiae (Fig. 3). While this result suggested that a Cys as catalytic base can be relevant for anaerobic functionality, changes at the same position of A. robustus ArUra9 (Cys to Ser) and K. marxianus KmUra9 (Ser to Cys) showed it is neither absolutely required nor sufficient for anaerobic activity of Class-II DHODs (Fig. 3).
Enzyme assays in cell extracts showed that ArAro9 can use free FAD and FMN as electron acceptors (Table 3), which was not previously observed for DHODs. Biochemical standard redox potentials of the non-enzyme-bound FADH 2 /FAD and FMNH 2 / FMN redox couples (E 0 ' = − 219 mV; [65]) and those of DHOD-bound FMNH 2 /FMN (− 242 to − 330 mV; [58][59][60]) are compatible with either of these flavin cofactors acting as physiological electron acceptor of ArUra9. In anaerobic chemostat cultures of S. cerevisiae, intracellular FAD and FMN contents of 0.17 and 0.09 μmol (g biomass) −1 were reported [66]. While use of free flavins in cellular redox reactions is relatively rare, the S. cerevisiae fumarate-reductase Osm1 can re-oxidize free FADH 2 [67]. Since combined deletion of OSM1 and its paralog FRD1 abolishes anaerobic growth of S. cerevisiae [68], we could not experimentally verify their involvement in the in vivo anaerobic activity of ArUra9. Sequences with strong homology to S. cerevisiae Frd1 and/or Osm1 were found in Neocallimastigomycetes, Sch japonicus and D. bruxellensis (Additional file 5: Table S3). In silico prediction of subcellular localization indicated that these putative fumarate reductases were mitochondrial in Sch. japonicus and D. bruxellensis. Consistent with the inferred cytosolic localization of ArUra9, they were predicted to occur in the cytosol of Neocallimastigomycetes (Additional file 5: Table S3).
Previous studies implicated the N-terminal domains of Class II DHODs in quinone binding [46,69]. Based on analysis of 1500 Class II DHOD sequences and structural alignments, Sousa et al. [16] proposed conserved residues involved in quinone binding, stabilizing and pocket entry (Additional file 5: Figure S2). Ura9 sequences from Neocallimastigomycetes lacked several of these conserved residues as a consequence of their N-terminal truncation. In addition, a conserved positively-charged residue implicated in quinone binding by other Class-II DHODs was replaced by a hydrophobic residue in ArUra9 and orthologs from other Neocallimastigomycetes (Additional file 5: Figure S2). These observations, together with the absence of quinone-dependent DHOD activity in cell extracts containing ArUra9 (Table 3) support the notion that, even though quinones have been detected in these anaerobes [70], they are not involved in the activity of ArUra9 and orthologs in other Neocallimastigomycetes,.
Consistent with the N-terminal truncation of Ura9 orthologs from Neocallimastigomycetes (Additional file 5: Figure S2), an ArUra9-eGFP fusion expressed in S. cerevisiae was localized to the cytosol (Fig. 4). Sch. japonicus and D. bruxellensis Ura9 orthologs retained an N-terminal mitochondrial targeting sequence, as well as conserved residues proposed to be involved quinone binding (Additional file 5: Figure S2). The only difference in quinone-associated residues was a Tyr-to-Phe change in SjUra9 (Y137), which, in view of the similarity of these amino acids, may not have affected functionality (Additional file 5: Figure S2). We were unable to measure activities of DbUra9 and ScUra9 in cell extracts with the artificial electron acceptor PMS or other potential electron acceptors.
DbUra9-expressing ura1Δ isolates of S. cerevisiae evolved for fast anaerobic pyrimidine-prototrophy revealed three different mutations in the FUM1 fumarase gene. Mutations in the human fumarase gene (FH) have been implicated with different types of cancer due to increased fumarate concentrations [71]. Thr218 and Ala294 of DbUra9 correspond to Val197 and Ala274, respectively, in FH and are both located in highly conserved regions [72,73], while Ala274 resides in the active site [74]. Mutation of Ala274 to a valine in human fumarate hydratase (FH) was implicated in ovarian mucinous cystadenoma [75], and resulted in a 50% decreased activity of the enzyme [73]. By analogy, it seems probable that the Fum1 A294V change in strain IMS1169 also led to a reduced fumarase activity. Corresponding amino acids in human FH of Thr218 and Met432 are located in the core helix, and accumulated fumarate resulting from mutations in this helix have been associated with different types of cancer [74]. If further research confirms higher intracellular fumarate concentrations in strains carrying FUM1 mutations, these might favor in vivo use of fumarate as electron acceptor, as in Class I-A enzymes [57] or, alternatively, stimulate reoxidation of FADH 2 via fumarate reductases. Anaerobic pyrimidine synthesis, combined with acquisition by horizontal gene transfer of genes enabling sterol-independent anaerobic growth (squalene hopene cyclase; [31]) and anaerobic deoxynucleotide synthesis (Class-I ribonucleotide reductase; [41]), indicates that Sch. japonicus is remarkably well adapted to anaerobic growth. The observation that independently constructed SjUra9-expressing S. cerevisiae strains consistently showed loss of respiratory capacity and mitochondrial DNA (Fig. 5) indicated a trade-off between anaerobic pyrimidine synthesis and respiratory competence. Quinone-dependent DHODs are known to react with oxygen, leading to formation of hydrogen peroxide and superoxide [55]. In S. cerevisiae, these reactive oxygen species have been implicated in loss of respiratory capacity [76]. Despite presence of respiratory proteins, including low levels of all cytochromes [76], Sch. japonicus strains are naturally respiratory deficient [77][78][79]. It is tempting to speculate whether this phenotype may have provided a driving force for evolution of a respirationindependent DHOD or, alternatively, be a consequence of evolution of SjUra9 for anaerobic functionality. With respect to the latter option, it should be considered that the heterologous context in which loss of respiratory competence was observed in this study does not necessarily reflect conditions in Sch. japonicus or its ancestors.

Conclusions
Our results show that, in addition to the well-established acquisition of a Class-I DHOD by S. cerevisiae and closely related yeasts, at least three separate events in fungal evolution enabled anaerobic pyrimidine synthesis by variants of Class-II DHODs that do not depend on aerobic respiration. These anaerobically active variants were shown to have a cysteine instead of a conserved serine residue in their catalytic sites. Their in vivo activities were not dependent on aerobic respiration, and, in Neocallimastigomycetes, they were not membrane associated and could use free FAD or FMN as electron acceptor. These remarkable differences with canonical Class-II DHODs underline the plasticity of fungal genomes and genes under selective pressure and extend our knowledge on eukaryotic adaptation to anoxic environments.

Molecular biology techniques
Phusion High Fidelity DNA Polymerase (Thermo Fisher Scientific) and PAGE-purified oligonucleotide primers (Additional file 5: Table S5, Sigma Aldrich) were used in polymerase chain reactions (PCRs) for cloning and sequencing. Diagnostic PCRs were performed with desalted oligonucleotides (Additional file 5: Table S5, Sigma Aldrich) and DreamTag Mastermix 2X (Thermo Fisher Scientific). Genomic DNA as template for PCRs was isolated using a YeaStar Genomic DNA kit (Zymo Research, Irvine, CA). PCR products were purified with a GeneElute PCR Clean-Up kit (Sigma Aldrich) or from 1% agarose gels using a Zymoclean Gel DNA Recovery Kit (Zymo Research). Gibson Assembly of purified DNA fragments with 20 bp sequence overlaps was performed with the NEBuilder HiFi DNA Assembly Mastermix (New England Biolabs, Ipswich, MA). Golden-Gate Page 13 of 18 Bouwknegt et al. Fungal Biol Biotechnol (2021) 8:10 assembly was performed according to Lee et al. [87]. E. coli XL1-Blue (Agilent Technologies, Santa Clara, CA) was chemically transformed following manufacturer's instructions and plated on selective media. Correct plasmid assembly was verified by diagnostic PCRs on E. coli transformants [86]. Cas9-mediated genome editing in S. cerevisiae was performed according to Mans et al. [86]. The LiAc/SS-DNA/PEG method [88] was used for yeast transformation with plasmids (at least 1 μg DNA per transformation) or linear DNA fragments (0.5-1 μg per transformation). Single-colony isolates of randomly picked transformants were obtained by three re-streaks on selective media. Integrations and deletions were verified by diagnostic PCRs with genomic DNA as a template. Construction of plasmids and yeast strains is described in detail in Additional file 6.

Cultivation of yeast strains
Aerobic shake-flask cultures were grown in an Innova Incubator (New Brunswick Scientific, Edison, NJ) at 30 °C and 200 rpm. Pre-cultures in 100-mL shake flasks with a working volume of 40 mL were inoculated with frozen stock cultures. Primary pre-cultures of yeast strains expressing plasmid-borne DHOD genes were grown on SMUD and those of other yeast strains on SMUD + ura, and were used to inoculate a secondary pre-culture on SMUD (for plasmid expressing strains) or SMUD + ura0.1. Upon reaching late exponential phase, cultures were centrifuged (5 min at 3000 × g) and washed twice with demineralized water. Washed cell suspensions were used to inoculate 500-mL shake flasks containing 100 mL of SMUD + ura or SMUD, at an initial optical density at 660 nm (OD 660 ) of 0.2.
Anaerobic cultures were grown in 100-mL shake flasks with a working volume of 80 mL. Pre-cultures were grown aerobically on SMUD + ura as described above, until stationary phase, washed twice with sterile demineralized water and transferred to an anaerobic preculture on SMUD + ura0.1, supplemented with Tween 80 and ergosterol. Flasks were incubated on an IKA KS 260 orbital shaker (240 rpm; Dijkstra Verenigde BV, Lelystad, The Netherlands) placed in a Shel Lab Bactron 300 anaerobic workstation (Sheldon Manufacturing Inc, Cornelius, OR) at 30 °C. The gas mixture supplied to the anaerobic workstation contained 10% CO 2 , 5% H 2 and 85% N 2 . Measures to minimize inadvertent oxygen entry were implemented as described by [102]. When anaerobic precultures reached stationary phase, they were used to inoculate cultures on SMUD and SMUD + ura supplemented with Tween 80 and ergosterol.
For spot-plate experiments, yeast strains were pregrown on 20 mL SMUD in 100-mL shake flasks, centrifuged (5 min, 3000g) and washed twice with demineralized water. Washed cultures were used for cell counts with a Z2 Coulter particle count and size analyzer (Beckman Coulter, Brea, CA) set at particle size 2.5-7.5 μm. Cells were diluted to a final concentration of 2.5•10 5 cells mL −1 , and subsequently diluted to 2.5•10 4 cells mL −1 , 2.5•10 3 cells/mL and 2.5•10 2 cells mL −1 . From these four dilutions, 4 μL of each strain and dilution was transferred to SMD, SMD + ura, SMEG and SMEG + ura plates. All strains were pre-grown and plated in duplicate.

Analytical methods
Extracellular metabolite concentrations were measured by high performance liquid chromatography as described by Verhoeven et al. [103] Optical density at 660 nm of aerobic cultures was measured using an Jenway 7200 spectrophotometer (Bibby Scientific, Staffordshire, UK) after accurate dilution to an OD 660 between 0.1 and 0.3. Anaerobic cultures were first diluted to an optical density at 600 nm (OD 600 ) between 0.15 and 0.35, followed by optical densities measurements at 600 nm on an Ultrospec 10 cell density meter (Biochrom, Harvard Biosience, Holliston, MA) that was placed in the anaerobic workstation [102].