We used the self-compatible, para-aminobenzoic acid auxotrophic AmutBmut1 pab1-1 #326 strain of C. cinerea for genetic modification and as a control in all experiments. The fungus was maintained on solid YMG (yeast-malt-glucose) medium. All media and buffers used for culturing and transformation (YMG, minimal medium, regeneration medium, top agar, PEG/CaCl₂ solution, MM and MMC buffers) were prepared as described by Dörnte and Kües (2012). Oidiation was induced by culturing at 37°C under constant white light. After transformation, we selected for the complementation of PABA auxotrophy using synthetic minimal medium.

Homologs of ort2 were retrieved by blastp (e < 10^-5) and an InterPro search (search terms: IPR001313, IPR033133) from a proteome dataset comprising 13 agaricomycete species, two additional basidiomycetes (Cryptococcus neoformans, Ustilago maydis) (Nagy et al., 2023) complemented by two ascomycete yeasts, Saccharomyces cerevisiae and Schizosaccharomyces pombe. The proteome of Phanerochaete chrysosporium is included in two versions. For our model species, C. cinerea version 2 was used. We refer to C. cinerea proteins based on the annotation of Hegedüs et al. (2025) (https://mushroomdb.brc.hu/); the IDs we supply can be found in the JGI annotation as gene (model) names (https://mycocosm.jgi.doe.gov/Copcin2/Copcin2.home.html). Homologs fitting the HMM of IPR040000 were excluded, as these are orthologs of the nucleolar protein Nop9 of S. cerevisiae (C. cinerea protein: 438228). Retained homologs were aligned by PRANK v.170427 (Löytynoja, 2014) by using the -F parameter and a maximum likelihood guide tree inferred in preliminary analyses by RAxML version 8.2.12 (Stamatakis, 2014), from a multiple sequence alignment generated by Mafft-linsi v7.453 (Katoh and Standley, 2013). The PRANK alignment was used for tree inference and bootstrapping by RAxML version 8.2.12, under the WAG model with gamma-distributed rate heterogeneity. Branch support was assessed by 500 non-parametric rapid bootstrap replicates. Trees were visualized in FigTree v1.4.4 (http://tree.bio.ed.ac.uk/software/figtree).

We intended to perform gene deletion through Cas9 cleavage followed by homologous repair, and overexpression via ectopic integration. We designed two crRNAs and a plasmid (p.Δort2) as a repair template, and another plasmid (p.ort2OE) was constructed to generate overexpressing strains ( Supplementary Figure S2 ). The plasmids, primers and crRNAs were designed using SnapGene version 4.3.11 (https://www.snapgene.com/). The sequences of primers and crRNAs are listed in Supplementary Table S1 . Both plasmids possess a backbone derived from pUC19 and carry the positive selection marker gene pab1-1 with its promoter and terminator regions (CcPAB1) amplified from the pMA412 vector (Stanley et al., 2014). p.Δort2 carries the marker gene between homologous arms (1 kb sequences upstream and downstream of the target sequence). p.ort2OE carries the ort2 gene driven by the promoter of a heat shock protein gene, Cc.hsp3 (ID: 423239), given that the promoter of its ortholog has been shown to be useful for overexpression in Schizophyllum commune (Ohm et al., 2011).

Fragments were amplified by Phusion Plus DNA polymerase (Thermo Fisher Scientific) using overlapping Gibson primers. Plasmids were assembled using the Gibson Assembly Cloning Kit (New England Biolabs, USA), following the manufacturer’s instructions. The ligation mixture was introduced into competent NEB5α cells following the protocol of Rabe and Cepko (2020), and the plasmids were purified from the cultures of colonies for which the presence and proper composition of the plasmids had been confirmed, using the NucleoBond Xtra Midi kit (Macherey-Nagel, Germany).

Oidia were harvested from 5-day-old cultures grown under constant light, by washing with distilled water, filtering through a 40 µm cell strainer, and pelleting by centrifugation (1000 g, 10 min). Pellets were washed with 5 ml MM buffer, pelleted again, and resuspended in 1 ml sterile-filtered MM buffer containing ~20% w/v VinoTaste Pro (Novozymes, Denmark). Protoplasting was carried out by shaking oidia at 70–90 rpm in an incubator at 36°C for 3–4 hours. Once the majority of oidia turned into spherical protoplasts, the process was terminated by adding 5 ml MMC buffer, and the protoplasts were collected (1000 g, 15 min).

The pelleted protoplasts were resuspended in 100 µl MMC buffer. We added 10 µg plasmid DNA; in the case of the knockout transformation, the two RNP complexes [assembled as described by Pareek et al. (2022)]; and for both transformations, 30 µl PEG-CaCl₂ solution. The cells were incubated on ice for 30 min, then 0.5 ml PEG-CaCl₂ solution was added, and the mixture was incubated at room temperature for 10–20 minutes. After adding 20 ml MMC buffer, the solution was gently mixed, and the cells were pelleted (1000 g, 15 minutes). The cells were resuspended in 1 ml MMC buffer, mixed with top agar, and poured onto regeneration medium. Colonies were collected onto selective minimal medium plates. Transformants were checked by PCR, amplicon sequencing, and RT-qPCR. To obtain DNA template for the PCR reactions, vegetative mycelium was lysed thermally as described by Dörnte and Kües (2012). Amplicon sequencing was conducted commercially (Microsynth, Switzerland) after purification using the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel, Germany). To purify mixed-genotype cultures, we prepared single-spore cultures from oidia.

Cultures grown on cellophane were homogenized in a mortar with liquid nitrogen. RNA extraction was performed using the Quick-RNA MiniPrep kit (Zymo Research, USA) following the manufacturer’s instructions. We screened the overexpressing mutant candidates by RT-qPCR. The colonies of the transformants and the wild-type controls were exposed to 42°C heat shock for 2 hrs before sample collection. Reverse transcription was carried out with the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific), and qPCR reactions were performed using SsoAdvanced Universal SYBR Green Supermix (BioRad), following the manufacturers’ instructions. RNA extracted from the wild-type strain was used as a control, and the β-tubulin gene (protein ID: 393528) was chosen as the reference gene. Expression changes were calculated using the formula FC = 2^−ΔΔCt (Livak and Schmittgen, 2001). qPCR was conducted in six technical replicates: two reactions per experiment, repeated three times starting from the same RNA extracts.

Since light- and dark-grown colonies of C. cinerea produce fruiting bodies of different morphology, we investigated fruiting under both conditions. Strains were inoculated onto ‘half-sugar’ YMG (containing 2 g/L glucose) (Muraguchi et al., 2015), incubated in darkness at 28°C for six days, then half of the plates were transferred to alternating light conditions (12 hrs light, 12 hrs dark; Panasonic FL40SS-ENW/37 light bulb), while the rest were kept in darkness. Both groups were further incubated at 28°C for two weeks. After several smaller-scale phenotyping experiments, fifty plates of the deletion mutants and one hundred of the selected overexpression strains were phenotyped and compared to a wild-type control group. For the overexpression mutants, after the additional two-week incubation, we quantified fruiting, along with making morphological observations. We counted etiolated fruiting bodies (dark stipes) longer than 1 cm per Petri dish formed by dark-grown colonies, and structures that had at least started to elongate (P4 and older) in the case of plates exposed to alternating light. Phenotyping experiments were also conducted in jars containing 150 ml ‘half-sugar’ YMG with 6 g/L agar, with the same protocol.

Following the method described by Venegas et al. (2011) and Galeota-Sprung et al. (2022), we performed qPCR on a nuclear (β-tubulin gene, JGI protein ID: 393528) and a mitochondrially encoded gene (NADH dehydrogenase subunit 1, protein ID: 384263), using DNA templates extracted from an ort2OE, four Δort2 mutants, and the wild-type strain, by using the Blood & Cell Culture DNA Maxi Kit (Qiagen, Germany). We used five extracts from all the strains; and for each sample, two technical replicate qPCR reactions were performed. qPCR was carried out in the same way as described above.

Research Group Julien Gagneur (Munich Centre for Machine Learning) provided us with a list of putative Ort2 recognition motifs and mRNA targets, predicted by their species-aware DNA language model (Karollus et al., 2024). We utilized the hits belonging to the most significant motif. The list contains 772 Okayama 7 genes; 655 remain when converted to AmutBmut identifiers. A Gene Ontology (GO) term enrichment analysis was performed on the AmutBmut gene set. A FIMO (Grant et al., 2011) search was also conducted for UTRs of C. cinerea (Hegedüs et al., 2025), using the S. cerevisiae Puf3p recognition motif PWM as described by Gerber et al. (2004). We aimed to select the top 100 most significant hits; however, due to the identical p-values among several entries, a total of 116 were included. We also generated three mitochondrial gene sets. Okayama 7 genes with a mitochondrial GO annotation were downloaded from QuickGO (https://www.ebi.ac.uk/QuickGO) and converted to AmutBmut identifiers. Mitochondrial gene sets were also obtained based on predicted subcellular localization. We employed TargetP 2.0 (Almagro Armenteros et al., 2019) and DeepLoc 2.0 (Thumuluri et al., 2022) to identify proteins potentially localizing to mitochondria. Hits with 0.7 or higher probability scores were considered significant. An overlap analysis was conducted on the five gene sets. An Euler diagram was drawn using the eulerr R package. For predicting the subcellular localization of Ort2, we used DeepLoc 2.0 (Thumuluri et al., 2022) as implemented at https://services.healthtech.dtu.dk/ and WolfPSort (Horton et al., 2007) at https://wolfpsort.hgc.jp. Both software were run with default parameters and the setting ‘slow’ for DeepLoc 2.0 and ‘fungi’ for WolfPsort.

Through BLASTp and InterPro searches, we collected the Pumilio protein sequences from the proteomes of 13 agaricomycete species, two additional basidiomycete species (Cryptococcus neoformans, Ustilago maydis) (Nagy et al., 2023), and two ascomycete yeasts, Saccharomyces cerevisiae and Schizosaccharomyces pombe. We identified 4 to 7 Pumilio proteins in the examined species of Basidiomycota, which is similar to copy numbers reported for Ascomycota (Matsushima et al., 2007). In Coprinopsis cinerea, six proteins that possess Pumilio-family RNA-binding repeat domains (IPR001313) were found. C. cinerea 438228 was excluded from the analyses, since its sequence is highly divergent, contains another domain (IPR040000), and, as an ortholog of the nucleolar Nop9 of Saccharomyces cerevisiae, it might function differently. Nop9 orthologs also formed a distant subfamily in previous studies (Najdrová et al., 2020).

The Pumilio family formed 5 clearly discernible and strongly supported (maximum likelihood bootstrap > 70%) subfamilies, each of which contained S. cerevisiae and Sch. pombe sequences, indicating conservation of major Pumilio subfamilies retained across Asco- and Basidiomycota ( Figure 1 ; Supplementary Figure S1 ). Each subfamily harbors a single protein of each agaricomycete species, with few exceptions (e.g., two Pleurotus ostreatus paralogs in clade puf4-puf5). Subfamily topologies mostly reflect the species tree (Naranjo‐Ortiz and Gabaldón, 2019): agaricomycete proteins group together, while more basal branches represent the two ascomycetes and the two basidiomycete species outside Agaricomycetes. This indicates highly conserved evolutionary trajectories and low overall divergence within subfamilies. Subfamilies differ from each other considerably, as indicated by the long branches separating them. C. cinerea 354109 (CopciAB_ort2) grouped with Puf3 of Sch. pombe, Puf3p of S. cerevisiae, and Pum1 of Cryptococcus neoformans ( Figure 1 ), whose function has been investigated (Webster et al., 2019; Liu et al., 2018), especially that of S. cerevisiae Puf3p (García-Rodríguez et al., 2007).

We inspected the expression dynamics of the five C. cinerea genes encoding Pumilio proteins in Krizsán et al.’s RNA-Seq dataset (2019) ( Figure 1 ). The expression of C. cinerea 354109 (CopciAB_ort2) peaks in the hyphal knot (HK), the earliest stage of fruiting body formation (Kües and Navarro-González, 2015), suggesting it has functions in the early developmental events. We note that Xie et al.’s data (2020) indicate a maximum expression in vegetative mycelium; however, both datasets agree that its expression decreases as the developmental program progresses. For the phylogenetic and expression-related reasons discussed above, we selected C. cinerea 354109, later named ort2, for experimental characterization.

Using the RNP-mediated CRISPR-Cas9 genome editing technique (Pareek et al., 2022), we generated four Δort2 disruptant strains (see the plasmid in Supplementary Figure S2a ). The lack of ort2 expression was demonstrated by RT-qPCR ( Supplementary Table S2 ). Though we used two crRNAs, instead of complete deletion, we could achieve only gene disruption. The disruptant genomes were not screened for off-target integrations, given that the phenotypic similarity of the independent disruptant strains implies the observed characters likely result from the intended disruption of ort2 rather than off-target effects.

Phenotyping of the four Δort2 disruptants revealed no differences in vegetative growth patterns and oidiation relative to the parent strain (not shown), whereas fruiting body and dark stipe development seem to be impaired to a degree ( Figure 2 ), the latter process considerably. Grown under alternating light conditions in jars, stage 1 primordia appeared at least 3 days later on disruptant cultures than on the wild-type ones, and no Δort2 fruiting bodies reached maturity (i.e., failed to produce basidiospores and deliquesce). We counted structures whose stipes had already started to elongate, including P4 and P5 stage primordia (Kües and Navarro-González, 2015), as well as young and mature fruiting bodies, together on four wild-type cultures and 16 Δort2 jar cultures ( Figure 2a ). On average, on wild-type cultures 10.5; while on Δort2 cultures 1.875 such structures formed, which is a significant difference (p = 0.0009, two-sample t-test); although we acknowledge the limited power of the analysis due to the low number of data points and the presence of outliers. We conducted a larger-scale experiment with 20 wild-type and 20 mutant jar cultures. The wild-type cultures formed 16.05 ± 9.56 fruiting bodies (P4 to MFB), while none of the disruptants produced any fruiting bodies or initials, indicating the decreased readiness of the mutants for fruiting compared to the parent strain. No consistent difference was observed in the tissue structure of primordia; however, many of them exhibited diverse abnormalities ( Figure 2b ). When cultured under constant darkness on plates, wild-type cultures produced dark stipes, whereas the Δort2 cultures did not. Across four wild-type jar cultures, a total of 34 dark stipes were counted, approximately 20 of which exceeded 1 cm in length/height. In contrast, on 16 Δort2 jar cultures, only three dark stipe-like structures were observed altogether (all measuring less than 1 cm), along with some rudimentary structures that failed to elongate. This indicates that Δort2 strains are capable of producing dark stipes, though to a considerably lower extent.

Utilizing an ectopically integrating plasmid ( Supplementary Figure S2b ), additional copies of ort2 were introduced into the genome of the AmutBmut strain, under the control of the hsp3 heat shock promoter. RT-qPCR was performed to measure ort2 expression in mycelia grown at 28°C ( Supplementary Table S3 ). We selected the strain with the highest expression change (ort2OE(10); FC = 4.14 ± 2.79), along with another overexpression strain (ort2OE(7); FC = 1.28 ± 0.65) for preliminary phenotyping. After confirming the phenotypes were consistent across the independent mutants in preliminary experiments, ort2OE(10) was characterized further in larger-scale experiments.

According to our observations, the overexpression of ort2 did not affect vegetative growth characteristics and oidiation, however, we observed an effect on fruiting. Under alternating light, the ort2OE strains are able to complete the fruiting program; from hyphal knots to mature fruiting bodies, all developmental stages were observed, as well as sclerotia and dark stipes. For plates grown under alternating light conditions, we counted structures whose stipes had already started to elongate (P4 to MFB), together. On 50-50 Petri dish cultures, we found that the number of fruiting bodies reaching stage 4 primordium or older was similar between the wild type and overexpression strains (p = 0.2354, two-sample t-test) ( Figure 3a ). Mature, deliquescent fruiting bodies were not observed in the wild-type strain during the phenotyping, whereas, on 50 ort2OE Petri-dish cultures, altogether 8, and on 4 cultures grown in jars, 6 fruiting bodies were able to complete development by the end of the experiment. In the case of dark-grown colonies, we counted etiolated fruiting bodies taller than 1 cm, as a proxy for the number of dark stipes. The overexpression mutant produced significantly more dark stipes than the wild-type strain (p = 7.23 × 10^-9, two-sample t-test) ( Figures 3b, c ). The average number of these structures per Petri dish was approximately 7.5 times higher in the case of the ort2OE mutant than in the wild-type strain. This, combined with the disruptant phenotypes, suggests that ort2 influences the development of dark stipes.

In addition, the ort2 overexpression strains exhibit diverse developmental abnormalities, which occur at low frequencies (and therefore cannot be subjected to statistical tests) ( Supplementary Figure S3 ). For dark-grown colonies, these phenotypes include the intertwining of etiolated fruiting bodies, forming amorphous structures ( Supplementary Figure S3d ), and the formation of robust prostrate dark stipes. Some of these reached lengths of up to 20 cm, nearly encircling the dish ( Supplementary Figures S3c-d ), a phenomenon that is uncommon in the wild-type strain. Branched fruiting body morphologies, occurring on both dark-grown colonies and those exposed to alternating light, constitute another low-frequency developmental abnormality observed upon ort2 overexpression ( Figure 3d ; Supplementary Figure S3a ). We documented fruiting bodies with branched stipes, as well as ones bearing protrusions on their caps (or tips). The branched fruiting bodies and snake-like prostrate dark stipes led us to name C. cinerea 354109 ort2 after Orthrus, the two-headed, snake-tailed dog from classical mythology.

We attempted obtaining mechanistic insights into the role of Ort2 in C. cinerea, by (1) functional enrichment analyses of putative targets inferred by a large language model (Karollus et al., 2024) and (2) by specifically addressing whether it is involved in mitochondrial biogenesis, a function that was postulated to be a Saccharomycotina-specific innovation (Jiang et al., 2010).

Using a language model (Karollus et al., 2024), reconstruction probability arrays were generated for the C. cinerea Okayama 7 genome (Stajich et al., 2010) (300 bps downstream of STOP codons, since Pumilio proteins primarily bind the 3’-UTRs of transcripts (Nishanth and Simon, 2020). Three motifs were identified that significantly matched the S. cerevisiae Puf3p motif (using TF-MoDISco and TOMTOM). The most significant motif ( Figure 4a ) was found in 772 C. cinerea Okayama 7 transcripts, 655 of which were found in the AmutBmut strainThe information content of the reconstruction probabilities of motif sites was significantly higher than that of sites containing the shuffled motif (p = 4.02×10^-166, Mann-Whitney U test), suggesting the DNA language model recognizes the specific nucleotide arrangement defining the motif, not just general sequence composition. This supports the motif’s robustness and increased confidence in future use. It is worth noting that the central position in the predicted C. cinerea motif – where it differs from the S. cerevisiae Puf3p motif (A) – can similarly be occupied by a cytosine in the experimentally determined binding site of the Neurospora crassa ortholog (Wilinski et al., 2017) ( Figure 4a ), supporting our prediction. We analyzed GO enrichment within the 655 C. cinerea AmutBmut target genes. This analysis did not reveal a clear signal, with only five significantly enriched terms (adjusted p-value < 0.05, Supplementary Table S4 ). The cause of the lack of signal in this dataset may be manifold, though we note a clear functional enrichment was not found for N. crassa either, perhaps meaning that these proteins bind functionally diverse transcripts. Nevertheless, our GO results did not comprise terms related to the mitochondrion, in agreement with previous results on filamentous fungi (Jiang et al., 2010).

To verify this finding, we first compared mtDNA abundance in the Δort2, ort2OE, and wild-type strains, using quantitative PCR targeting a nuclear (β-tubulin) and a mitochondrially encoded gene (NADH dehydrogenase subunit 1, Figure 4b ; Supplementary Table S5 ). If Ort2 regulated mitochondrial biogenesis, one would expect overexpression and deletion to have opposite effects on mtDNA abundance. However, no such differences were observed. The measured mtDNA levels in the disruptants were not significantly different from those in the wild type. Although the selected overexpression strain displayed a more pronounced difference, this change occurred in the opposite direction than one would expect based on the role of S. cerevisiae Puf3p as a negative regulator of mitochondrial biogenesis (García-Rodríguez et al., 2007).

As an alternative approach, we compared candidate Ort2 targets inferred by in silico methods to mitochondrial proteins of C. cinerea (1) Gene Ontology (GO) annotations, (2) subcellular protein localization predicted by DeepLoc 2.0, and (3) by TargetP 2.0. The three mitochondrial gene sets considerably overlap ( Figure 4c ), validating these predictions. Mitochondrial protein lists are provided in Supplementary Table S6 . However, when these gene sets were compared with 655 putative Ort2 targets predicted by the language model, and with 116 C. cinerea mRNA targets predicted with FIMO (using the S. cerevisiae Puf3p recognition motif as query), we obtained small overlaps ( Figure 4c ; Supplementary Table S6 ). Only one comparison, between the set of genes with mitochondrial GO annotation and the list of Ort2 targets predicted by the language model, revealed a significant overlap: 14.77% of the LM hits are in the GO set (p < 10^-5, Fisher’s, exact test).

Prediction of the subcellular localization of ort2 using in silico tools suggested that ort2 is cytoplasmic and nuclear. DeepLoc 2.0 yielded scores of 0.7543 and 0.5299 for cytoplasmic and nuclear localization, respectively, with much lower (>0.09) scores for other localizations. WolfPsort ranked a nuclear localization first (score: 13), followed by cytoplasmic (6), mitochondrial (4), and peroxisomal (3). We hypothesize that ort2 may shuttle between the cytoplasm and the nucleus, possibly depending on posttranslational modifications. This is consistent with the behavior of other Pumilio family proteins (Gerber et al., 2004). Overall, we interpret these results as an absence of evidence for a mitochondrial function of Ort2 is in line with the data of Jiang et al. (2010) on filamentous Ascomycota. It remains to be tested whether the predicted mRNA targets are biologically significant and how these contribute to fruiting and dark stipe development remains unclear.