Utilization of selenocysteine in early-branching fungal phyla

Selenocysteine (Sec), the 21st amino acid, is co-translationally inserted via an unusual recoding mechanism, wherein UGA, normally a stop codon, is translated as Sec¹. Sec insertion occurs specifically in selenoprotein genes, due to cis-acting RNA structures known as SECIS elements⁹. Sec machinery genes (tRNA^Sec, EFsec, PSTK, SBP2, SecS, SPS) are trans-factors necessary and sufficient for eukaryotic Sec synthesis and insertion^1,2,10. Sec is believed to confer catalytic advantage over Cys, its sulphur-containing analogue^11,12, for specific oxidoreductase functions. Nevertheless, selenoproteins are not found in all organisms. Sec usage is scattered across bacteria^3,4,13 and archaea¹⁴. Within eukaryotes, selenoproteins are present in most metazoans (including all vertebrates¹⁵), some protists and certain algae^4,5,16. They are absent in many insects⁶, few nematodes⁷, plants⁵, and various protists⁴. Notably, fungi were considered the only kingdom of life entirely devoid of Sec^4,5,8. However, here we provide conclusive genomic evidence for Sec utilization by nine fungal species belonging to three early-branching phyla.

We downloaded all available fungal genomes from NCBI (1201 species, Supplementary Table 1), and searched them for the presence of eukaryotic Sec machinery genes (Methods) using Selenoprofiles¹⁷ and Secmarker¹⁸. These automatically generated predictions (Supplementary Figure 1) were analysed for two potential confounders: the occurrence of protein families with similarity to those of interest, and contaminant sequences in fungal genome assemblies. For this, we reconstructed gene trees of candidate proteins together with their most similar annotated sequences (Methods) and inspected them to distinguish protein families (Supplementary Figures 2–5). This procedure led to the dismissal of several candidates. After filtering, Sec machinery proteins (Supplementary Data 1) localized only in a handful of genomes, and co-occurred with tRNA^Sec (Fig. 1). After extensive analysis, we filtered out three species with Sec machinery which we reckoned as resulting from genome contamination from Sec-utilizing bacteria (Supplementary Note 1). On the other hand, we concluded that Bifiguratus adelaidae (Mucoromycota), Gonapodya prolifera (Chytridiomycota), Capniomyces stellatus, Zancudomyces culisetae, Smittium culicis, Smittium simulii, Smittium megazygosporum, Smittium angustum, and Furculomyces boomerangus (Zoopagomycota) were Sec-utilizing fungi (Fig. 2). These species formed distinct clades in three early-branching fungal phyla. The order of Harpellales was particularly well represented: 7 of the 8 species analysed had Sec.

We identified selenoproteins in all Sec-utilizing fungi (Supplementary Data 1), which belonged to seven known selenoprotein families (gene trees provided in Supplementary Figures 6–10). Two of them were found in all Sec-utilizing fungi: SelenoH, a nuclear oxidoreductase possibly involved in redox homeostasis¹⁹, and SPS (Fig. 3), Sec machinery component and also selenoprotein itself in many prokaryotes and eukaryotes⁴. Other fungal selenoproteins included SelenoU, an uncharacterized oxidoreductase²⁰; AhpC (alkyl hydroperoxide reductase C), found as selenoprotein in certain bacteria, protists and porifera²¹; MsrA (methionine sulfoxide reductase A)²², identified as selenoprotein in algae, protists, and various non-vertebrate metazoa; DI-like, homologous to vertebrate iodothyronine deiodinases²³ and present as selenoprotein in various invertebrates, protists and bacteria¹⁶; and finally TXNRD (thioredoxin reductase²⁴), a selenoprotein present in most Sec-utilizing eukaryotes. Notably, this constitutes the first case of animal-like TXNRD described in fungi, since this kingdom uses a shorter and Sec-independent form of TXNRD²⁴. G. prolifera was the species with most selenoproteins, covering all selenoprotein families discussed above. Analysis of a publicly available transcriptome for this species confirmed the expression of all selenoproteins and Sec machinery factors, except for tRNA^Sec (Methods). Selenoprotein transcripts appear to occur at high levels in G. prolifera (Supplementary Figure 11). SPS mRNA was particularly abundant, ranking in the top 1-6% (depending on the background distribution used) transcripts.

We searched fungal selenoprotein genes for occurrence of eukaryotic SECIS elements. Surprisingly, we found canonical SECISes only in 5 of 29 selenoproteins (Fig. 2). We then broadened our searches (Methods) and detected additional SECIS-like structures in 8 genes. These contained the SECIS core motif and typically additional extended paired regions (Supplementary Figure 12). All candidate structures (except one) were located downstream of coding sequences, as expected of eukaryotic SECISes (Supplementary Figure 13). Strikingly, no SECIS candidates were predicted in any Harpellales selenoproteins (with one possible exception). Despite this, we are confident that they represent bona-fide selenoprotein genes: they show homology to known selenoproteins and exhibit UGA aligned precisely with the known Sec position. This observation suggests utilization of non-standard Sec recoding signals. We further performed de novo searches for RNA structures. First, we applied RNAz²⁵ (Methods) to identify evolutionarily conserved structures in SelenoH and SPS genes of Harpellales. This approach identified two structured regions, one per gene, both within coding sequences. The SelenoH CDS structure consisted of two stems, including the Sec UGA and downstream region (Supplementary Figure 14). While all Harpellales sequences support this structure, the majority of pairings break down in the other two Sec-utilizing fungi, suggesting that it may be order-specific. The SPS CDS structure (Supplementary Figure 15) appears conserved across all species. It is located immediately upstream of the Sec UGA and consists of a ~30 pairs stem. Due to the lack of closely related sequences of non-Harpellales Sec-utilizing species, we could not apply RNAz to the rest of fungal selenoproteins. We thus ran RNALfold²⁶ to predict all locally stable RNA structures. All these structures (Supplementary Figure 13) constitute Sec recoding signal candidates.

The reconstructed gene trees for Sec machinery and selenoproteins roughly recapitulate the species tree. These results are illustrated for SPS in Fig. 3, with the rest of protein trees provided in Supplementary Figures 2–10. SPS sequences from Sec-utilizing fungi form a monophyletic cluster, as expected from species phylogeny. This supports the scenario of common descent, with the Sec-containing SPS gene inherited from the ancestor of fungi to these extant species. The same clustering pattern was observed for SBP2. In the case of EFsec and SecS, fungal sequences also form a monophyletic cluster, with the exception of G. prolifera. Noteworthily, G. prolifera possesses two SecS paralogs, both containing similar C-terminal extensions. In these regions, we identified RNA recognition motifs (PFAM family RRM_1). In the PSTK tree, Harpellales form a unique cluster, while the other two Sec-utilizing fungi branch out of a cluster including protists and algae. The position of the fungal clusters relative to the homologs in other lineages is also informative: they are located in the expected position as an outgroup of metazoa, sometimes within protist lineages. Beyond fungal species, Sec machinery trees are in general accordance with the known species phylogeny, with some exceptions. In particular, nematodes and insects are typically placed in trees in more basal positions than expected. This may be caused by diversification of divergence rates across eukaryotic groups, causing long-branch attraction of genes of fast evolving lineages. In a previous study⁴, we concluded that the Sec encoding trait has been likely directly inherited from the root of eukaryotes to extant Sec-utilizing species, while the various non-Sec-utilizing groups underwent a number of parallel Sec losses. The phylogenetic trees presented here are consistent with this scenario, and thus support direct and continuous inheritance of Sec from the root of eukaryotes to Sec-utilizing fungi.

Given the remarkable diversity spanned by fungal genomes, we hypothesized that phylogenetic profiling could be profitably taken to discover fungal genes related to the Sec trait. This technique²⁷ exploits gene co-occurrence across genomes to link genes to pathways. Aiming to uncover potential undiscovered selenoproteins and proteins involved in the Sec pathway, we developed a custom phylogenetic profiling procedure (Methods) to identify genes present in Sec-utilizing fungal genomes but absent in selenoproteinless species. Our procedure resulted in a list of 57 candidate proteins clustered in 27 homologous groups (Supplementary Table 2). PSTK, EFsec, SecS, SPS and SelenoH emerged as top scoring candidates, supporting robustness of the procedure. The rest of candidates had diverse annotated functions, and included oxidoreductases, transporters, uncharacterized proteins and others. We further examined candidates as potential selenoproteins, seeking conserved Sec-UGA codons, but found none (Methods). These genes may be involved in Sec transport, metabolism, or regulation, or represent false positives. Future experiments will be necessary to validate candidates and clarify their role in Sec biology.

In summary, we report here that Sec is genetically encoded by nine species of fungi belonging to three early-branching phyla (Fig. 2). Each phyla includes several other species with no trace of Sec usage in their genomes. We estimated the completeness of the genome assemblies (Methods) and dismissed it as explanation of the scattered pattern of Sec presence (Supplementary Figure 16). We found complete Sec machinery in each Sec-utilizing genome, with two exceptions: SecS in Z. culisetae and SBP2 in G. prolifera. The absence of the former is likely due to incomplete assembly, since we detected SecS in all other Harpellales. In contrast, SBP2 may be truly absent in G. prolifera (or diverged beyond recognition). In fact, while there are no genomes available from closely related species, there is a public G. prolifera transcriptome, and although we found all other machinery proteins transcribed at detectable levels, we still could not identify SBP2. Since this species possesses two SecS paralogs with unique RNA-binding domains, it is tempting (yet far-fetched, from current data) to speculate that one of these may have acquired the SBP2 function.

The scattered distribution of Sec across fungi, together with phylogenetic signal in machinery genes, supports a scenario of vertical inheritance followed by multiple independent Sec losses in this kingdom (Supplementary Note 2). Even in the most parsimonious scenario (Fig. 2), we infer at least 10 different complete losses of Sec encoding capacity in fungi. One of these losses was mapped prior to the split of Dikarya (Ascomycota and Basidiomycota, diverged around 700 mya²⁸), resulting in Sec absence in Saccharomyces cerevisiae and related fungi. The most recent loss occurred in S. mucronatum, as the rest of Harpellales all utilize Sec. We analysed stop codon usage across fungi, and saw no obvious correlation with Sec usage (Supplementary Figure 17). The high incidence of parallel losses may seem surprising; however, this was previously well documented for the Sec trait at various evolutionary scales. Parallel Sec losses were frequently observed in prokaryotes ^3,4,13,14, in various protist groups^4,16 and even within animals, with at least 5 independent Sec loss events meticulously traced within insects⁶. Selenoprotein losses can occur either through whole gene loss or through conversion of Sec codons to Cys, the latter partially retaining enzymatic functions. With few exceptions, Sec-devoid fungi do not contain Cys orthologs of fungal selenoproteins (Supplementary Note 3), suggesting gene losses as primary mechanism of selenoprotein depletion in fungi.

Based on sequence homology, Sec location and phylogenetic signal, we are confident that all identified fungal candidates constitute bona-fide selenoproteins. However, canonical SECIS elements were identified only in a few cases. We further detected non-canonical, SECIS-like structures (with extra paired regions) downstream of various selenoproteins genes (Fig. 2, Supplementary Figures 12–13). If these truly function as Sec insertion signals, then significant alterations occurred in fungi in the SECIS consensus. In any case, none of the selenoproteins of Harpellales (except one) exhibit SECIS/SECIS-like candidates. This indicates non-canonical mechanisms of Sec recoding, at the very least in this fungal order. By analysing sequences of this lineage, we identified two evolutionary conserved structures in the coding regions of SPS and SelenoH (Supplementary Figures 14–15). The SelenoH structure is localized to the Sec TGA and immediately downstream region. In this sense, it resembles recoding-enhancing structures previously characterized in other selenoprotein genes, known as SREs²⁹. The SPS structure locates instead just upstream of Sec TGA. In metazoans, SPS contains a SRE downstream of Sec⁴. It is unclear how the fungal CDS structure relates to the metazoan SRE of SPS, both in phylogeny and in function. Future experiments will be essential to pinpoint the roles of the various RNA elements identified here, as they may as well be unrelated to Sec insertion.

The different Sec-utilizing eukaryotic lineages have typically both common and private selenoprotein families. However, we identified in fungi only selenoproteins with clear homology to known metazoan families. This may be due to purely technical reasons: identification of previously undescribed selenoproteins is a difficult task and typically relies on identification of SECIS elements. Since Sec signals appear altered in fungi, it is not surprising that our SECIS-dependent searches (Methods) did not result in any such candidate. It is conceivable that additional selenoproteins are encoded in Sec-utilizing fungal genomes, possibly including families exclusive of this kingdom. The characterization of fungal Sec recoding signals will facilitate their identification.

Our discoveries may open the way for the bio-engineered production of selenoproteins in fungi. The synthesis of exogenous selenoproteins has previously been attempted in S. cerevisiae: all human Sec machinery genes were transferred and successfully expressed, but Sec insertion was never observed³⁰. This was speculatively attributed to incompatibilities between vertebrates and fungi, due to differences in ribosome structure and translation mechanism. Possibly, the use of Sec machinery native to the fungal kingdom may lead to better results. We also propose G. prolifera as attractive model for eukaryotic selenoprotein expression. The naturally high transcript levels of selenoproteins (Supplementary Figure 11) suggest a tolerance to selenoprotein concentrations higher than most organisms.

In conclusion, we showed here genomic and transcriptomic evidences for Sec usage in fungi. While Sec synthesis machinery matches that of other eukaryotes, fungi may employ unusual mechanisms of Sec insertion involving private Sec signals, requiring further ad-hoc experiments to elucidate. In this work, we provide RNA structure candidates identified in selenoproteins, as well as a list of genes showing genomic co-occurrence with Sec machinery, to facilitate targeted experiments. Lastly, the observation of so many Sec losses within fungi opens the way to investigate the selective forces driving Sec maintenance and loss, and any possible commonalities between fungi and other lineages undergoing Sec extinctions. We anticipate that characterizing the functions of fungal selenoproteins will be key to clarify why some organisms have clung to them for so long, and, by extension, why many other species could instead lose Sec with no apparent consequence.