Identification of Oxygen-Independent Pathways for Pyridine Nucleotide and Coenzyme A Synthesis in Anaerobic Fungi by Expression of Candidate Genes in Yeast

ABSTRACT Neocallimastigomycetes are unique examples of strictly anaerobic eukaryotes. This study investigates how these anaerobic fungi bypass reactions involved in synthesis of pyridine nucleotide cofactors and coenzyme A that, in canonical fungal pathways, require molecular oxygen. Analysis of Neocallimastigomycetes proteomes identified a candidate l-aspartate-decarboxylase (AdcA) and l-aspartate oxidase (NadB) and quinolinate synthase (NadA), constituting putative oxygen-independent bypasses for coenzyme A synthesis and pyridine nucleotide cofactor synthesis. The corresponding gene sequences indicated acquisition by ancient horizontal gene transfer (HGT) events involving bacterial donors. To test whether these enzymes suffice to bypass corresponding oxygen-requiring reactions, they were introduced into fms1Δ and bna2Δ Saccharomyces cerevisiae strains. Expression of nadA and nadB from Piromyces finnis and adcA from Neocallimastix californiae conferred cofactor prototrophy under aerobic and anaerobic conditions. This study simulates how HGT can drive eukaryotic adaptation to anaerobiosis and provides a basis for elimination of auxotrophic requirements in anaerobic industrial applications of yeasts and fungi.

requiring enzymes of the kynurenine pathway. Instead, their genomes were reported to harbor genes encoding an L-aspartate oxidase (NadB) and quinolinate synthase (NadA), two enzymes active in the bacterial pathway for NAD 1 synthesis (6) (Fig. 1, right). Since bacterial and plant aspartate oxidases can, in addition to oxygen, also use fumarate as electron acceptor (19,20), it is conceivable that NadA and NadB may allow for oxygen-independent NAD 1 synthesis in anaerobic fungi. No hypothesis has yet been forwarded on how these fungi may bypass the oxygen requirement for the canonical fungal CoA biosynthesis route.
The goals of this study were to identify the pathway responsible for oxygen-independent synthesis of CoA in Neocallimastigomycetes and to investigate a possible role of NadA and NadB in oxygen-independent synthesis of pyridine nucleotide cofactors. A candidate L-aspartate decarboxylase (Adc)-encoding gene was identified by genome analysis of Neocallimastigomycetes, and its phylogeny was investigated. Candidate Neocallimastigomycetes genes for L-aspartate oxidase and quinolinate synthase, previously reported to have been acquired by HGT (6), as well as the candidate Adc gene, were then functionally analyzed by expression in S. cerevisiae strains devoid of essential steps in the native cofactor synthesis pathways. As controls, previously characterized genes involved in oxygen-independent NAD 1 biosynthesis by Arabidopsis thaliana (21) and a previously characterized Adc-encoding gene from the red flour beetle Tribolium castaneum (TcPAND) (22) were also expressed in the same S. cerevisiae strains. The results demonstrate how heterologous expression studies in yeast can provide insight into evolutionary adaptations to anaerobic growth and selective advantages conferred by proposed HGT events in Neocallimastigomycetes. In addition, they identify metabolic engineering strategies for eliminating oxygen requirements for cofactor biosynthesis in anaerobic industrial applications of S. cerevisiae.

RESULTS
Identification of a candidate oxygen-independent L-aspartate decarboxylase involved in CoA synthesis in anaerobic fungi. Decarboxylation of L-aspartate to b-alanine by L-aspartate decarboxylase (Adc), an enzyme that occurs in many species across all domains of life (23), enables an oxygen-independent alternative for the canonical fungal pathway for CoA synthesis (Fig. 1). To explore its occurrence in anaerobic fungi, a set of 51 amino acid sequences of Adc homologs listed by Tomita et al. (23) were used as queries against all proteins from 5 Neocallimastigomycetes species deposited in the TrEMBL section of the UniProt database. This search yielded 16 Neocallimastigomycetes hits (E value , 10 26 ) (see Table S1 in the supplemental material), six of which originated from Neocallimastix californiae. Only one of these hits, A0A1Y1ZL74, did not reveal annotation errors upon transcriptome sequencing (RNAseq) read mapping, showed the highest read coverage (see Fig. S1), and was selected as the best Neocallimastigomycetes Adc candidate.
The amino acid sequence A0A1Y1ZL74 (here referred to as NcAdcA) was used for a second round of homology search to obtain a broad set of Adc-like sequences, with a similar sequence representation of taxa across the three domains of life (104 sequences from Bacteria, 101 from Eukarya, and 120 from Archaea) (see Data Set S1). The complete set of NcAdcA homologs (together with the set defined by Tomita et al. [23] and their Neocallimastigomycetes homologs) (see Data Set S2) was subjected to multiple-sequence alignment. A subsequent phylogenetic tree ( Fig. 2; Data Set S3) showed that NcAdc sequences are closely related to those of chytrid fungi (e.g., A0A1S8W5A4 from Batrachochytrium salamandrivorans) and anaerobic bacteria (e.g., B8I983 from Clostridium cellulolyticum, currently known as Ruminiclostridium cellulolyticum [24]; we used the former name for consistency with UniProt identifiers). These Neocallimastigomycetes, chytrid, and bacterial Adc homologs were more closely related to each other than to characterized eukaryotic Adc and bacterial PanD sequences. Furthermore, HMMER E values obtained from using NcAdcA as the query against the bacterial database were more significant than when using the eukaryotic or archaeal databases (see Fig. S2; Data Set S1). These results suggest that a bacterial ancestor donated an Adc-encoding sequence to a common ancestor of chytrids and Neocallimastigomycetes.
To further investigate the potential bacterium-to-chytrid HGT event, a refined ortholog search and phylogenetic analysis were performed. Full proteomes of all species showing an NcAdcA homolog, in addition to predicted proteomes from six chytrids used in a previous phylogenomic analysis (8), were retrieved and used to obtain all possible co-ortholog groups. From a total number of 103 NcAdcA orthologs obtained, 85 were bacterial, 5 were archaeal, and 13 were eukaryotic (Table 1; Data Set S2). Eukaryotic NcAdcA orthologs were only found in fungi, and 12 of 13 were found in species from the Chytridiomycota phylum. The latter included five of the six chytrids analyzed in the phylogenomic study by Wang et al. (8) and all Neocallimastigomycetes considered in this study. Further phylogenetic analysis of the 103 NcAdcA orthologs indicated a common origin for bacterial and chytrid NcAdcA ( Fig. 3; Data Set S4). The closest bacterial relatives to NcAdcA were found in the facultative anaerobe and waterborne bacterium Aeromonas hydrophila subsp. hydrophila ATCC 7966 T (25) and the ruminal anaerobe C. cellulolyticum strain H10 (24,26). Additional close bacterial relatives were also strict anaerobes, such as the syntrophic bacterium Syntrophus aciditrophicus (27) and members of the Desulfobacteraceae family (28).
The Adc bacterium-to-chytrid HGT event was further confirmed by using Abaccus, an automated phylogeny-aware and topology-based algorithm (29). Abaccus uses the FIG 2 Unrooted maximum likelihood phylogenetic tree of aspartate decarboxylase and glutamate decarboxylase homologs. Sequences of proteins with demonstrated enzyme activity are marked with white triangles (L-aspartate decarboxylases) or black triangles (glutamate decarboxylases). Interactive visualizations with all sequence identifiers, branch distances, and bootstrap values can be accessed at https://itol.embl.de/tree/838448017961605604402 and https://itol.embl.de/tree/8384480476641615985323. topology of a given tree to determine taxonomic level "jumps" (J) and "losses" (L) between a seed sequence (NcAdcA) and every other node in the tree. The tree of NcAdcA orthologs resulted in a J of 4 and L of 3, meaning that the node comprising NcAdcA "jumps" 4 taxonomic levels, which could only be explained by complete losses in 3 of these taxonomic levels. These J and L values obtained for the tree of NcAdcA orthologs are higher than Abaccus' default HGT cutoff values (J $ 2 and L $ 3) and are independent of the evolutionary model used to infer the tree (PROTGTR [30], JTT [31], and LG [32]).
Comparison of bacterial PanDs (Q0TLK2 from Escherichia coli and P9WIL2 from Mycobacterium tuberculosis) against Adcs from other bacteria (B8I983 from C. cellulolyticum) and eukaryotes (including A7U8C7 from Tribolium castaneum) showed only little sequence homology between NcAdcs, known bacterial PanDs, and eukaryotic Adcs (Data Set S5). The only conserved region encompassed the full length of PanDs (126 to 139 amino acids), which represents less than 60% of the full length of other Adc sequences (e.g., NcAdcA is 625 amino acids long). These sequence comparisons, together with the intron-exon structures verified with RNA-seq data (Fig. S1), show that NcadcA has acquired eukaryotic features while retaining homology to its bacterial ancestor, as is typical for prokaryotic genes acquired by fungal genomes (33).
Neocallimastigomycetes PfnadB, PfnadA, and NcadcA genes support aerobic pyridine nucleotide and CoA synthesis in yeast. Neocallimastigomycetes were previously reported to have acquired an L-aspartate oxidase (nadB) and a quinolinate synthase (nadA) gene by HGT (6). Hence, UniProt entries A0A1Y1V2P1 and A0A1Y1VAT1 from Piromyces finnis were functionally reassigned as NadA and NadB candidates, and the corresponding genes were tentatively named PfnadB and PfnadA. These sequences, together with NcadcA, were codon optimized and tested to bypass the corresponding oxygen-requiring reactions in S. cerevisiae.
The BNA2 and FMS1 genes of S. cerevisiae were deleted by Cas9-mediated genome editing. The inability of strain IMK877 (bna2D) to synthesize quinolinic acid and of strain IMX2292 (fms1D) to synthesize b-alanine was evident from their inability to grow on glucose synthetic medium lacking nicotinic acid (SMDDnic) and pantothenate (SMDDpan), respectively (Table 2). Strain IMK877 was used for heterologous complementation studies with codon-optimized expression cassettes for PfnadB and PfnadA, while an expression cassette for N. californiae NcadcA (A0A1Y1ZL74) was introduced into strain IMX2292. Congenic strains expressing previously characterized NADB and NADA genes from Arabidopsis thaliana (AtNadB and AtNadA; Q94AY1 and Q9FGS4, respectively) (21), and a previously characterized gene from Tribolium castaneum encoding an aspartate decarboxylase (TcPanD; A7U8C7) (22) were tested in parallel.
Aerobic growth of the engineered S. cerevisiae strains was characterized in shakeflask cultures on SMD or on either SMDDnic or SMDDpan (Table 2). In contrast to the Fungal NAD 1 and Coenzyme A Oxygen-Independent Pathways ® reference strain IMK877 (bna2D), S. cerevisiae IMX2301 (bna2D PfnadB PfnadA) grew in SMDDnic, indicating complementation of the bna2D-induced nicotinate auxotrophy by PfnadB and PfnadA. However, the specific growth rate of the engineered strain in these aerobic cultures was approximately 3-fold lower than that of the reference strain IMX585 (BNA2) ( Table 2). Strain IMX2302 (bna2D AtNADB AtNADA) did not grow in SMDDnic, suggesting that the plant NadB and/or NadA proteins were either not functionally expressed or not able to complement the nicotinate auxotrophy in these aerobic yeast cultures. Strain IMX2300 (fms1D NcadcA) grew in SMDDpan, indicating complementation of the pantothenate auxotrophy. However, this strain reproducibly showed a lag phase of approximately 48 h upon its first transfer from SMD to SMDDpan and grew exponentially thereafter at a rate of 0.34 6 0.01 h 21 . To explore whether the lag phase of strain IMX2300 reflected selection of a spontaneous mutant, it was subjected to three sequential transfers in SMDDpan. A single-colony isolate, IMX2300-1, from the adapted population showed a specific growth rate of 0.346 0.01 h 21 in both SMD and SMDDpan ( Table 2). Whole-genome sequencing of IMX2300-1 did not reveal any mutations in coding DNA sequences that were considered physiologically relevant in this context compared to the nonadapted strain IMX2300 (BioProject accession number PRJNA634013). This observation indicated that the lag phase of strain IMX2300 most likely reflected a physiological adaptation or culture heterogeneity rather than a mutational event (34).
The specific growth rate of S. cerevisiae IMX2305 (fms1D TcPAND) on SMDDpan did not significantly differ from that of the reference strain IMX585 on SMD, and it was almost 4-fold higher than the specific growth rate of the reference strain on SMDDpan. These results are consistent with a previous study on functional expression of TcPAND in S. cerevisiae (35).
Expression of Neocallimastigomycetes PfnadB, PfnadA, and NcadcA suffices to enable anaerobic pyridine nucleotide and CoA synthesis in yeast. To investigate whether expression of heterologous PfnadB and PfnadA and that of NcadcA were  Table 1. An interactive visualization with all sequence identifiers, branch support, distances, and bootstrap values can be accessed at https://itol.embl.de/tree/8384480267191615280152. sufficient to enable anaerobic growth in the absence of nicotinate and pantothenate, respectively, growth of the engineered S. cerevisiae strains on SMD, SMDDnic, and/or SMDDpan was monitored in an anaerobic chamber (Fig. 4).
Growth experiments on SMDDnic or SMDDpan were preceded by a cultivation cycle on the same medium, supplemented with 50 g liter 21 instead of 20 g liter 21 of glucose to ensure complete depletion of any surplus cellular contents of pyridine nucleotides, CoA, or relevant intermediates. Indeed, upon a subsequent transfer to SMDDnic or SMDDpan, the reference strain IMX585 (BNA2 FMS1), expressing the native oxygen-dependent pathways for nicotinate and b-alanine synthesis, showed no growth (Fig. 4A, B, and C).
Characterization of engineered yeast strains in anaerobic batch bioreactors. The anaerobic chamber experiments did not allow quantitative analysis of growth and product formation. Therefore, growth of the S. cerevisiae strains expressing the Neocallimastigomycetes genes, IMX2301 (bna2D PfnadB PfnadA) and IMX2300-1 (fms1D NcadcA), was studied in anaerobic bioreactor batch cultures on SMDDnic or SMDDpan and compared to growth of S. cerevisiae IMX585 (BNA2 FMS1) on the same media.
The reference strain IMX585, which typically grows fast and exponentially in anaerobic bioreactors when using complete SMD (36), exhibited extremely slow, linear growth on SMDDnic and SMDDpan (Fig. 5). Similar growth kinetics in "anaerobic" bioreactor cultures of S. cerevisiae on synthetic medium lacking the anaerobic growth factors Tween 80 and ergosterol were previously attributed to slow leakage of oxygen into laboratory bioreactors (37)(38)(39).
In contrast to the reference strain IMX585, the engineered strains IMX2301 and IMX2300-1 exhibited exponential anaerobic growth on SMDDnic and SMDDpan, respectively ( Fig. 5; Table 3). The specific growth rate of strain IMX2301 (bna2D PfnadB PfnadA) on SMDDnic was not significantly different from that of the reference strain on complete SMD (36), indicating full complementation of the anaerobic nicotinate auxotrophy of S. cerevisiae. The specific growth rate of strain IMX2300-1 (fms1D NcadcA) on SMDDpan was only 20% lower than this benchmark (Table 3). Biomass and ethanol yields of strain IMX2301 grown in anaerobic batch cultures on SMDDnic and strain IMX2300-1 grown on SMDDpan were not significantly different from those of the reference strain IMX585 grown on complete SMD (P value . 0.05) ( Table 3).

DISCUSSION
This study shows how oxygen-independent pantothenate and nicotinate prototrophy can be conferred to the facultatively anaerobic yeast S. cerevisiae by heterologous expression of NcadcA, PfnadB, and PfnadA genes from Neocallimastigomycetes as well as corresponding orthologs from other species (TcPAND, AtNADB, and AtNADA). These Fungal NAD 1 and Coenzyme A Oxygen-Independent Pathways ® results also provide insights into how acquisition of these genes by HGT conferred selective advantage to Neocallimastigomycetes' ancestors under anaerobic conditions. Results from phylogenetic analysis of Adc sequences ( Fig. 2) were consistent with an earlier report on multiple evolutionary origins and variable evolutionary rates of pyridoxal-59-phosphate-dependent enzymes, including Adc and glutamate decarboxylases (40,41). A separate clade of Neocallimastigomycetes sequences shows homology with characterized glutamate decarboxylases (e.g., Q04792 from S. cerevisiae and K4HXK6 from Lactobacillus brevis) (Fig. 2). These results further support acquisition of an Adc-encoding DNA sequence by HGT rather than by neofunctionalization of a glutamate decarboxylase gene.
The characterized NcAdcA (A0A1Y1ZL74) yielded the highest homology with orthologous sequences from chytrid fungi and anaerobic bacteria. This observation is in agreement with previous research showing that HGT events played a major role in shaping the genomes of Neocallimastigomycetes (4, 6, 7), with Firmicutes and Proteobacteria as prominent sequence donors (6). Specifically, closer bacterial orthologs to NcAdcA were found in genome sequences of A. hydrophila (Proteobacteria) and C. cellulolyticum (Firmicutes). These bacterial species are anaerobic, and considering their ecological niches (waterborne and decayed grass/ruminal fluid, respectively [24,26]), the results agree with current hypotheses of these types of bacteria donating genes to anaerobic gut fungi and subsequently driving a mammalian transition to herbivory (6,8). Since NcAdcA orthologs were found in 5 of the 6 chytrids analyzed, the Adc HGT transfer event appears to have preceded the 66 (610) million years ago (MYA) estimate for divergence of Neocallimastigomycetes from other chytrids (8), although this estimate may be contended by more recent phylogenomic analyses for the whole fungal kingdom (42).
NcAdcA orthology and phylogenetic analyses revealed Phialocephala subalpina as the only other nonchytrid non-Neocallimastigomycetes eukaryote to have a separate Adc-like protein. This fungus is a root endophyte and was previously proposed have obtained multiple genes by HGT from bacterial donors (43). However, A0A1L7WVG4 , and IMX2302 (bna2D AtNADB AtNADA) (C) transferred to medium with 2% glucose with (~) or without (n) nicotinate after a carry-over phase in SMDDnic containing 4% glucose (l in gray box). Strains IMX585 (D), IMX2300-1 (fms1D NcadcA) (E), and IMX2305 (fms1D TcPAND) (F) transferred to medium with (4) or without (h) pantothenate after a carry-over phase in SMDDpan containing 4% glucose (* in gray box). Anaerobic conditions in the chamber were maintained using a palladium catalyst and a 5% hydrogen concentration. Error bars represent the mean deviations from independent cultures (n = 2).
(PAC_06602), here identified as an Adc ortholog, was not among the 21 genes of P. subalpina listed as likely acquired by HGT from nonfungal species. Since the phylogenetic placement of the putative P. subalpina Adc was close to bacterial as well as archaeal sequences, further studies are needed to reveal its evolutionary history.
Whereas an alternative to the kynurenine pathway for NAD 1 synthesis was previously inferred from genome sequence analysis, the pathway by which Neocallimastigomycetes synthesize coenzyme A had not previously been explored. Six pathways for synthesis of the essential CoA precursor b-alanine are known: (A) decarboxylation of L-aspartate (44), (B) transamination of malonate semialdehyde with L-glutamate as amino donor (45) or L-alanine (46), (C) reduction of uracil followed by hydrolysis of the resulting dihydrouracil (47), (D) oxidative cleavage of spermine to 3-aminopropanal followed by oxidation of the aldehyde group (16), (E) 2,3-aminomutase of alanine (48), and (F) addition of ammonia to acryloyl-CoA, followed by hydrolysis of the resulting CoA thioester (48). Of these pathways, all but option D, in principle, can occur in the absence of oxygen. Yeasts and other filamentous fungi typically form b-alanine from spermine (pathway D), but in some species, the use of pathway C was also reported (49).

Strain
Medium Fungal NAD 1 and Coenzyme A Oxygen-Independent Pathways ® While the aspartate decarboxylation route (A) has not previously been demonstrated in wild-type fungi, functional expression of bacterial and T. castaneum TcPanD was used in metabolic engineering of S. cerevisiae to boost supply of b-alanine as a precursor for 3-hydroxypropionate production (22,35). Wild-type S. cerevisiae strains cannot grow in anaerobic environments unless supplemented with pantothenate. Expression of either NcadcA or TcPAND in an fms1D S. cerevisiae strain, which lacks the native oxygen-dependent pantothenate biosynthesis pathway, enabled growth in pantothenate-free medium under aerobic and anaerobic conditions. Although the different specific growth rates of S. cerevisiae strains expressing NcadcA or TcPAND indicate that changing expression levels and/or origin of ADC-encoding genes may be required to achieve optimal growth, these results provide a proof of principle for a simple metabolic engineering strategy to eliminate oxygen requirements for pantothenate synthesis.
Genomic analyses previously suggested that genomes of Neocallimastigomycetes encode a putative L-aspartate oxidase (NadB) and quinolinate synthase (NadA) as alternatives to the canonical kynurenine pathway found in other fungi (6). Additionally, Neocallimastigomycetes appear to have acquired both nadB and nadA through HGT (6). Until now, functionality of these Neocallimastigomycetes proteins in an oxygen-independent pathway for synthesis of quinolinate from L-aspartate had not been demonstrated.
Our results demonstrate that expression of nadB and nadA homologs, either from the Neocallimastigomycetes P. finnis or from the plant A. thaliana (21), suffice to allow anaerobic synthesis of NAD 1 of S. cerevisiae. Due to the involvement of the Bna2 and Bna4 oxygenases in NAD 1 synthesis by S. cerevisiae, nicotinate is an essential growth factor for this yeast under anaerobic conditions (14,50,51). A similar strategy was recently successfully applied to enable oxygen-independent synthesis of pyridine nucleotides in the bacterium Pseudomonas putida (52). The present study represents the first demonstration of a metabolic engineering strategy to eliminate oxygen requirements for NAD 1 synthesis in a yeast.
Functional expression of heterologous NadA quinolinate synthases in S. cerevisiae was observed despite the fact that these enzymes are [4Fe-4S] iron-sulfur cluster proteins (53,54), which are notoriously difficult to functionally express in the yeast cytosol (55)(56)(57)(58). However, earlier studies on functional expression of the [4Fe-4S] activating protein of bacterial pyruvate-formate lyase (59,60) demonstrated that low-levels of expression can occur without modification of the yeast machinery for cytosolic assembly of [Fe-S] clusters. The inability of AtNadB and AtNadA to support NAD 1 synthesis in aerobic cultures may be due to oxygen sensitivity of the [4Fe-4S] cluster in the AtNadA quinolinate synthase domain (61). In contrast to PfNadA, AtNadA carries an N-terminal SufE domain which, in other organisms, has been demonstrated to allow this oxygensensitive enzyme to remain active under aerobic conditions by reconstituting its [Fe-S] cluster (61).
This work contributes to the understanding of how Neocallimastigomycetes adapted to their anaerobic lifestyle by acquiring genes that enable oxygen-independent synthesis of central metabolic cofactors. Experiments with engineered S. cerevisiae strains showed that contribution of the heterologous genes to in vivo oxygen-independent cofactor synthesis did not require additional mutations in the host genome. These results indicate how acquisition of functional genes by HGT, even if their expression was initially suboptimal, could have conferred an immediate advantage to ancestors of anaerobic fungi living in cofactor-limited anoxic environments. A similar approach was recently applied to study the physiological impact on S. cerevisiae of expressing a heterologous gene encoding squalene-tetrahymanol cyclase, which in Neocallimastigomycetes, produces the sterol surrogate tetrahymanol (13). Functional analysis by heterologous expression in S. cerevisiae circumvents the current lack of tools for genetic modification of Neocallimastigomycetes (2) and can complement biochemical studies (3)(4)(5) and genome sequence analyses (6, 7).
Pantothenate and nicotinate, together with the other compounds belonging to the B-group of water-soluble vitamins, are standard ingredients of chemically defined media for aerobic and anaerobic cultivation of yeasts (62). S. cerevisiae strains have been shown to contain the genetic information required for de novo synthesis of these vitamins and can even be experimentally evolved for complete prototrophy for individual vitamins by prolonged cultivation in single-vitamin-depleted media (63,64). In large-scale processes, addition of nutritional supplements increases costs, reduces shelf-life of media, and increases the risk of contamination during their storage (62). Therefore, metabolic engineering strategies for enabling oxygen-independent synthesis of NAD 1 and pantothenate are of particular interest for the development robust yeast strains with minimal nutritional requirements that can be applied in anaerobic biofuel production (62). Further studies of the unique evolutionary adaptations of Neocallimastigomycetes may well provide additional inspiration for engineering robust fungal cell factories that operate under anaerobic conditions.

MATERIALS AND METHODS
Homology and phylogenetic analyses. A set of 51 amino acid sequences previously used to discriminate between L-aspartate decarboxylases (Adc) and glutamate decarboxylases (23) was reused to identify candidate Neocallimastigomycetes Adc sequences. These sequences were used as queries against a database containing all 58,109 Neocallimastigomycetes proteins deposited in UniProt trembl (Release 2019_02), which represented 5 species (Neocallimastix californiae, Anaeromyces robustus, Piromyces sp. E2, Piromyces finnis, and Pecoramyces ruminantium), and extracted according to the NCBI taxonomic identifier (taxid) 451455. Sequence homology was analyzed using BLASTP 2.6.01 (65) with 10 26 as the E value cutoff, resulting in 16 Neocallimastigomycetes sequences as shared hits from all 51 queries (see Table S1 in the supplemental material). Four of these sequences showing homology to experimentally characterized Adc proteins originated from N. californiae and were checked for RNA-seq read coverage and splicing junction support, revealing A0A1Y1ZL74 as the best candidate (Fig. S1). For this purpose, Illumina libraries were obtained from the Sequence Read Archive using accession SRR7140690 (66) which were then mapped using STAR 2.6.1a_08-27 (67) against genome assembly GCA _002104975. Alignments were processed using SAMtools 1.3.1 (68) and visualized using Artemis (69).
A0A1Y1ZL74, also referred to as NcAdcA, was used for a second round of homology search using HMMER 3.2 (70) against 3 different databases built from UniProt release 2019_02 to include all RefSeq sequences from Bacteria (taxid 2), Eukarya (taxid 2759), and Archaea (taxid 2157; TrEMBL and Swiss-Prot categories were also included in this case). Selection for hits with more than 60% alignment length over the query sequence and an E value of ,10 26 resulted in a total of 325 sequences (103 from Bacteria, 101 from Eukaryota, and 121 from Archaea) (Data Set S1).
The set of 325 A0A1Y1ZL74 homologous sequences, together with those from Tomita et al. (23), and the 16 Neocallimastigomycetes sequences from that described above were used for further phylogenetic analyses. A total of 387 sequences (Data Set S2) were aligned with MAFFT v7.402 (71) in "einsi" mode, and alignments were trimmed with trimAl v1.2 (72) in "gappyout" mode and then used to build a maximum likelihood phylogenetic tree with RAxML-NG 0.8.1 (72) using default parameters, with the exception of the use of the PROTGTR1FO model and 100 bootstrap replicates. The resulting phylogenetic tree drawn with iTOL (73) is shown in Fig. 2, and corresponding sequences and the unannotated tree are provided in Data Sets S2 and S3.
Proteomes from species showing an Adc homolog were extracted into individual fasta files and used for (co)orthology search with ProteinOrtho6 (74). A0A1Y1ZL74 ortholog groups were then extracted and subjected to alignment, trimming, and phylogenetic analysis as described above. The resulting phylogenetic tree is shown in Fig. 3, and corresponding sequences and the unannotated tree are provided in Data Sets S2 and S4.
Abaccus v1.1 (29) (https://github.com/Gabaldonlab/Abaccus) was used to search the tree presented in Fig. 3 (Data Set S4) for evidence of HGT. For this purpose, the taxonomy table provided as default was supplemented with definitions for the additional chytrids considered in this study.
Multiple-sequence alignment was also performed with Clustal omega 1.2.4 (75) to compare selected amino acid sequences showing candidate and experimentally characterized Adcs against bacterial PanDs. These sequences and alignments are shown in Data Set S5.
Strains, media, and maintenance. S. cerevisiae strains used and constructed in this study (Table 4) were derived from the CEN.PK lineage (76). Yeast cultures were routinely propagated in YP (10 g liter 21 Bacto yeast extract [Becton, Dickinson and Co., Sparks, MD], 20 g liter 21 Bacto peptone [Becton, Dickinson and Co.]) or synthetic medium (SM) (77). YP and SM were autoclaved at 121°C for 20 min. SM was then supplemented with 1 ml liter 21 of filter-sterilized vitamin solution (0.05 g liter 21 D-(1)-biotin, 1.0 g liter 21 D-calcium pantothenate, 1.0 g liter 21 nicotinic acid, 25 g liter 21 myo-inositol, 1.0 g liter 21 thiamine hydrochloride, 1.0 g liter 21 pyridoxol hydrochloride, 0.20 g liter 21 4-aminobenzoic acid). Where indicated, nicotinic acid or pantothenic acid was omitted from the vitamin solution, yielding SM without nicotinic acid (SMDnic) and SM without pantothenic acid (SMDpan), respectively. A concentrated glucose solution was autoclaved separately for 15 min at 110°C and added to SM and YP to a concentration of 20 g liter 21 or 50 g liter 21 , yielding SMD and YPD, respectively. SMD with urea or acetamide instead of ammonium sulfate (SMD-urea and SMD-Ac, respectively) was prepared as described previously (78,79). For anaerobic growth experiments, sterile media were supplemented with Tween 80 (polyethylene Fungal NAD 1 and Coenzyme A Oxygen-Independent Pathways ® glycol sorbate monooleate; Merck, Darmstadt, Germany) and ergosterol ($95% pure; Sigma-Aldrich, St. Louis, MO) as described previously (39). Yeast strains were grown in 500-ml shake flasks containing 100 ml medium or in 100-ml shake flasks containing 20 ml medium. Shake-flask cultures were incubated at 30°C and shaken at 200 rpm in an Innova Incubator (Brunswick Scientific, Edison, NJ). Solid media were prepared by adding 15 g liter 21 Bacto agar (Becton, Dickinson and Co.) and, when indicated, 200 mg liter 21 G418 (Thermo Scientific, Waltham, MA). After genotyping, engineered strains were restreaked twice to select single clones. Removal of the guide RNA (gRNA)-carrying plasmid was conducted as previously described (80). Stock cultures were prepared by adding glycerol to a final concentration of 33% (vol/vol), frozen, and stored at 280°C.
Molecular biology techniques. DNA was PCR amplified with Phusion Hot Start II high-fidelity polymerase (Thermo Scientific) and desalted or PAGE-purified oligonucleotide primers (Sigma-Aldrich) according to the manufacturers' instructions. DreamTaq polymerase (Thermo Scientific) was used for diagnostic PCR. Oligonucleotide primers used in this study are listed in Table 5. PCR products were separated by gel electrophoresis using 1% (wt/vol) agarose gel (Thermo Scientific) in Tris-acetate-EDTA (TAE) buffer (Thermo Scientific) at 100 V for 25 min and purified with either GenElute PCR clean-up kit (Sigma-Aldrich) or with Zymoclean gel DNA recovery kit (Zymo Research, Irvine, CA). Plasmids were purified from E. coli using a Sigma GenElute plasmid kit (Sigma-Aldrich). Yeast genomic DNA was isolated with the SDS-lithium acetate (LiAc) protocol (81). Yeast strains were transformed with the lithium acetate method (82). Four to eight single colonies were restreaked three consecutive times on selective media, and diagnostic PCRs were performed to verify their genotypes. Escherichia coli XL1-Blue was used for chemical transformation (83). Plasmids were then isolated and verified by either restriction analysis or by diagnostic PCR. Lysogeny broth (LB; 10 g liter 21 Bacto tryptone, 5 g liter 21 , Bacto yeast extract with 5 g liter 21 NaCl) was used to propagate E. coli XL1-Blue. LB medium was supplemented with 100 mg liter 21 ampicillin for selection of transformants. The overnight-grown bacterial cultures were stocked by adding sterile glycerol at a final concentration of 33% (vol/vol), after which samples were frozen and stored at 280°C.
Plasmid construction. Plasmids used and cloned in this study are shown in Table 6. Plasmids carrying two copies of the same gRNA were cloned by Gibson assembly (80,84). In brief, an oligonucleotide carrying the gene-specific 20-bp target sequence and a homology flank to the plasmid backbone was used to amplify the fragment carrying the 2-mm origin of replication sequence by using pROS13 as the template. The backbone linear fragment was amplified using primer 6005 and pROS11 as the template (85). The two fragments were then gel purified and assembled in vitro using the NEBuilder HiFi DNA assembly master mix (New England BioLabs, Ipswich, MA) according to the manufacturer's instructions. Transformants were selected on LB plates supplemented with 100 mg liter 21 ampicillin or 50 mg liter 21 kanamycin. Primer 11861 was used to amplify the 2-mm fragment containing two identical gRNA sequences for targeting BNA2. The PCR product was then cloned in a pROS11 backbone yielding plasmid pUDR315.
The coding sequences for AtNADA, AtNADB, PfnadA, PfnadB, and NcadcA were codon optimized for expression in S. cerevisiae and ordered as synthetic DNA through GeneArt (Thermo Fisher Scientific). The plasmids carrying the expression cassettes for TcPAND, AtNADA, AtNADB, PfnadA, and PfnadB were cloned by Golden Gate assembly using the Yeast Toolkit (YTK) DNA parts (86). These plasmids were cloned using the pYTK096 integrative backbone that carries long homology arms to the URA3 locus and a URA3 expression cassette allowing for selection on SM lacking uracil. The TcPAND coding sequence was amplified using the primer pair 11877/11878 and pCfB-361 as the template. Then, the linear TcPAND gene and plasmids pUD1096, pUD1097, pUD652, and pUD653 carrying the coding sequences for AtNADA, AtNADB, PfnadA, and PfnadB, respectively, were combined together with YTK-compatible part plasmids in BsaI (New England BioLabs) Golden Gate reactions to yield plasmids pUDI168, pUDI245, pUDE931, pUDI243, and pUDI244, respectively.
For deletion of the FMS1 gene, IMX581 (can1D::Spycas9-natNT2 ura3-52) was transformed with 500 ng of the FMS1-targeting gRNA plasmid pUDR652 together with 500 ng of the annealed primer pair 13527/13528 as the repair dsDNA oligonucleotide, resulting in IMX2293. Then, 500 ng each of plasmids pUDI168 and pUDI242 carrying the expression cassettes for TcPAND and NcadcA, respectively, were digested with NotI (Thermo Fisher) and separately transformed in IMX2293, yielding IMX2305 and IMX2300, respectively. Selection of IMX2305 and IMX2300 was performed on an SMD agar plate, since the integration of each Adc-encoding cassette also restored the URA3 phenotype. In contrast, selection of IMK877 was conducted on SMD-Ac agar plates, while selection of IMX2302, IMX2301, and IMX2293 was conducted on YPD-G418 agar plates. Strains IMK877, IMX2300, IMX2302, and IMX2301 were stocked in SMD, while IMX2305 and IMX2293 were stocked in SMDDpan and YPD, respectively.
Aerobic growth studies in shake flasks. For the determination of the specific growth rate of the engineered strains under aerobic conditions, a frozen aliquot was thawed and used to inoculate a 20-ml wake-up culture that was then used to inoculate a preculture in a 100-ml flask. The exponentially growing preculture was then used to inoculate a third flask to an initial optical density at 600 nm (OD 660 ) of 0.2. The flasks were then incubated, and growth was monitored using a 7200 Jenway Spectrometer (Jenway, Stone, United Kingdom). Specific growth rates were calculated from at least five time points in the exponential growth phase of each culture. Wake-up and precultures of IMX2301 and IMX2302 were grown in SMDDnic. Wake-up and precultures of IMX2300 and IMX2305 were grown in SMDDpan, while wake-up and precultures of IMK877 and IMX2292 were grown in SMD. Anaerobic growth studies in shake flasks. Anaerobic shake-flask based experiments were performed in a Lab Bactron 300 anaerobic workstation (Sheldon Manufacturing Inc., Cornelius, OR) containing an atmosphere of 85% N 2 , 10% CO 2 , and 5% H 2 . Flat-bottom shake flasks of 50 ml were filled with 40 ml SMD-urea medium containing 50 g liter 21 glucose as the carbon source to ensure depletion of the vitamin/growth factor of interest and 20 g liter 21 glucose for the first transfer. Media were supplemented with vitamins, with and without pantothenic acid or nicotinic acid as indicated, and in all cases, supplemented with Tween 80 and ergosterol. Sterile medium was placed inside the anaerobic chamber 24 h prior to inoculation for removal of oxygen. Traces of oxygen were continuously removed with a regularly regenerated Pd catalyst for H 2 -dependent oxygen removal placed inside the anaerobic chamber. Aerobic overnight shake-flask cultures on SMD-urea were used to inoculate the anaerobic shake flask without pantothenic acid or without nicotinic acid at an initial OD 600 of 0.2. Cultures were cultivated at 30°C with continuous stirring at 240 rpm on an IKA KS 260 Basic orbital shaker platform (Dijkstra Verenigde BV, Lelystad, the Netherlands). Periodic optical density measurements at a wavelength of 600 nm using an Ultrospec 10 cell density meter (Biochrom, Cambridge, United Kingdom) inside the anaerobic environment were used to follow the growth over time. After growth had ceased and the OD 600 no longer increased, the cultures were transferred to SMD-urea with 20 g liter 21 glucose at an OD 600 of 0.2 (39).
Anaerobic bioreactor cultivation. Anaerobic bioreactor batch cultivation was performed in 2-liter laboratory bioreactors (Applikon, Schiedam, the Netherlands) with a working volume of 1.2 liters. Bioreactors were tested for gas leakage by applying 30 kPa overpressure while completely submerging them in water before autoclaving. Anaerobic conditions were maintained by continuous sparging of the bioreactor cultures with 500 ml N 2 min 21 (#0.5 ppm O 2 , HiQ nitrogen 6.0; Linde Gas Benelux, Schiedam, the Netherlands). Oxygen diffusion was minimized by using Fluran tubing (14 Barrer O 2 , F-5500-A; Saint-Gobain, Courbevoie, France) and Viton O-rings (Eriks, Alkmaar, the Netherlands). Bioreactor cultures were grown on either SMDDpan or SMDDnic with ammonium sulfate as the nitrogen source. pH was controlled at 5 using 2 M KOH. The autoclaved mineral salts solution was supplemented with 0.2 g liter 21 sterile antifoam emulsion C (Sigma-Aldrich). Bioreactors were continuously stirred at 800 rpm, and temperature was controlled at 30°C. Evaporation of water and volatile metabolites was minimized by cooling the outlet gas of bioreactors to 4°C in a condenser. The outlet gas was then dried with a PermaPure PD-50T-12MPP dryer (Permapure, Lakewood, NJ) prior to analysis. CO 2 concentrations in the Frozen glycerol stock cultures were used to inoculate aerobic 100-ml shake-flask cultures on either SMDDpan or SMDDnic. Once the cultures reached and OD 660 of .5, a second 100-ml aerobic shake-flask preculture on the same medium was inoculated. When this second preculture reached the exponential growth phase, biomass was harvested by centrifugation at 3,000 Â g for 5 min and washed with sterile demineralized water. The resulting cell suspension was used to inoculate anaerobic bioreactors at an OD 660 of 0.2.
Analytical methods. Biomass dry weight measurements of the bioreactor batch experiments were performed using preweighed nitrocellulose filters (0.45 mm; Gelman Laboratory, Ann Arbor, MI). Tenmilliliter culture samples were filtrated, and then the filters were washed with demineralized water prior to drying in a microwave oven (20 min at 360 W) and weight measurement. Metabolite concentrations in culture supernatants were analyzed by high-performance liquid chromatography (HPLC). In brief, culture supernatants were loaded on an Agilent 1260 HPLC system (Agilent Technologies, Santa Clara, CA) fitted with a Bio-Rad HPX 87 H column (Bio-Rad, Hercules, CA). The flow rate was set at 0.6 ml min 21 , and 0.5 g liter 21 H 2 SO 4 was used as the eluent. An Agilent refractive-index detector and an Agilent 1260 variable wavelength detector (VWD) were used to detect culture metabolites (88). An evaporation constant of 0.008 divided by the volume in liters was used to correct HPLC measurements of ethanol in the culture supernatants, taking into account changes in volume caused by sampling (89). Statistical analysis on product yields was performed by means of an unpaired two-tailed Welch's t test.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. DATA SET S1, XLSX file, 0.