A-to-I mRNA editing controls spore death induced by a fungal meiotic drive gene in homologous and heterologous expression systems

Abstract

Spore killers are meiotic drive elements that can block the development of sexual spores in fungi. In the maize ear rot and mycotoxin-producing fungus Fusarium verticillioides, a spore killer called Sk^K has been mapped to a 102-kb interval of chromosome V. Here, we show that a gene within this interval, SKC1, is required for Sk^K-mediated spore killing and meiotic drive. We also demonstrate that SKC1 is associated with at least 4 transcripts, 2 sense (sense-SKC1a and sense-SKC1b) and 2 antisense (antisense-SKC1a and antisense-SKC1b). Both antisense SKC1 transcripts lack obvious protein-coding sequences and thus appear to be noncoding RNAs. In contrast, sense-SKC1a is a protein-coding transcript that undergoes A-to-I editing to sense-SKC1b in sexual tissue. Translation of sense-SKC1a produces a 70-amino-acid protein (Skc1a), whereas the translation of sense-SKC1b produces an 84-amino-acid protein (Skc1b). Heterologous expression analysis of SKC1 transcripts shows that sense-SKC1a also undergoes A-to-I editing to sense-SKC1b during the Neurospora crassa sexual cycle. Site-directed mutagenesis studies indicate that Skc1b is responsible for spore killing in Fusarium verticillioides and that it induces most meiotic cells to die in Neurospora crassa. Finally, we report that SKC1 homologs are present in over 20 Fusarium species. Overall, our results demonstrate that fungal meiotic drive elements like SKC1 can influence the outcome of meiosis by hijacking a cell’s A-to-I editing machinery and that the involvement of A-to-I editing in a fungal meiotic drive system does not preclude its horizontal transfer to a distantly related species.

Introduction

In the late 1860s, the father of modern genetics, Gregor Mendel, was the first to recognize that the inheritance of an organism’s traits was predictable. Today, a key tenant of Mendel’s Laws of Inheritance is that a gene (allele) present in only 1 parent of a sexual cross will be inherited by 50% of the resulting progeny. Despite this tenant, some genes are inherited at a much higher frequency, often close to 100%. These “selfish” genes are said to drive through meiosis and are referred to as meiotic drive elements (Zimmering et al. 1970).

Several meiotic drive elements have been identified in the Ascomycota, a fungal phylum in which sexual spores, ascospores, are produced in sack-like structures known as asci (Raju 1994; Bravo Nunez et al. 2018). In the Ascomycota, known meiotic drive elements are called spore killers because they disrupt ascospore production. While most asci have either 4 or 8 ascospores, a spore killer can reduce the number of viable ascospores by half, with most of the survivors inheriting the spore killer genotype.

Among the Ascomycota, spore killers in species of Podospora and Neurospora have been examined most extensively. Podospora anserina has at least 2 groups of spore killers. One group currently consists of a single member (het-s) while the other contains many members (the Spok family). The het-s allele encodes a prion that interacts with an alternate, nonprion forming allele to cause spore death (Dalstra et al. 2003; Saupe 2011). In contrast, Spok genes encode a single protein that can both kill spores and provide resistance to killing (Grognet et al. 2014; Vogan et al. 2019). Neurospora spp. has 3 spore killers: Spore killer-1 (Sk-1), Sk-2, and Sk-3 (Turner and Perkins 1979). The Sk-1 locus consists of a single gene responsible for both killing and resistance (Svedberg et al. 2021). The Sk-2 and Sk-3 loci contain multiple genes and killing and resistances are conferred by different genes. In the Sk-2 locus, gene rfk-1 confers Sk-2-mediated spore killing, while gene rsk confers resistance to killing (Campbell and Turner 1987; Hammond et al. 2012; Harvey et al. 2014; Rhoades et al. 2019). In the Sk-3 locus, gene rsk confers resistance, but the gene that confers killing has not been identified (Hammond et al. 2012; Svedberg et al. 2018).

In contrast to Podospora and Neurospora, understanding of spore killers in other genera of the Ascomycota, such as Fusarium, is limited. While meiotic analyses of 3 Fusarium species, Fusarium verticillioides, Fusarium subglutinans, and Fusarium proliferatum, have identified the presence of spore killing (Kathariou and Spieth 1982; Sidhu 1984; Raju 1994), little is known about the genetic basis of spore killing in these fungi. All 3 species are of agricultural concern because they cause maize ear rot, which is an economically significant disease under some environmental conditions (White 1999). The fungi are also a food and feed safety concern because they produce mycotoxins. Fumonisin (FB) mycotoxin production by F. verticillioides is of particular concern because the toxins frequently contaminate maize kernels and have been associated with cancer, neural tube defects, and stunted growth in human populations for which maize is a dietary staple (Marasas 2001; Munkvold et al. 2021). FB toxicity is enhanced by coexposure with aflatoxin, a mycotoxin produced by Aspergillus species that also infect maize (Carlson et al. 2001; Xue et al. 2019). Crops contaminated with mycotoxins cause billion-dollar losses world-wide each year. Although efforts to minimize the impact these toxins have on society have been significant, none have resulted in a control strategy that is effective under all conditions.

Analysis of the spore killer system in F. verticillioides indicated isolates have either a spore killer (Sk^K) or spore killer sensitive (Sk^S) genotype. An Sk^K x Sk^S cross yielded asci with 4 ascospores, whereas Sk^K x Sk^K and Sk^S x Sk^S crosses yield asci with 8 ascospores. After initial analyses in the early 1980s, Sk^K was mapped to a 102-kb segment of chromosome V (Xu and Leslie 1996; Pyle et al. 2016). Comparison of this segment in Sk^K and Sk^S strains of F. verticillioides revealed multiple differences, including the presence of a unique, albeit hypothetical, gene in Sk^K strains and its absence in Sk^S strains. This gene was named Spore Killer Candidate-1 (SKC1) (Pyle et al. 2016). The objective of the current study was to further characterize SKC1 and determine whether it is responsible for Sk^K-mediated spore killing in F. verticillioides.

Materials and methods

Fungal strains and general propagation methods

Fusarium verticillioides strains (Table 1) were routinely grown on V8 juice agar (V8A) (Tuite 1969) and Carrot agar (CA) at 22°C in an incubator with continuous fluorescent light (Philips F34T12). CA was prepared as described by Klittich and Leslie (1988) except that we hand-peeled and diced large organically grown carrots for use in the medium. Neurospora strains (Table 2) were routinely grown on Vogel’s minimal medium (VMM, Vogel 1956) at 32°C in an incubator or at room temperature on a benchtop.

Table 1.

Fusarium verticillioides strains.

Name (Aliases)a	SKC1 b	SKC1 RIPc	Markerd	Mating type	Genetic lineage
Fv149 (FGSC 7600, M-3125)	Absent	No	None	MAT1-1	SkS
Fv999 (FGSC 7603, M-3120)	Native	No	None	MAT1-2	SkK
ΔSKC1::HYGB (tJML33.1)	Replaced with HYGB	No	ΔSKC1::HYGB	MAT1-2	SkK
ΔSKC1::GENR (tJML34.46)	Replaced with GENR	No	ΔSKC1::GENR	MAT1-2	SkK
Addback-A (tJML69.37)	Native + ectopic	No	GENRE	MAT1-2	SkK
Addback-B (tJML69.143)	Native + ectopic	No	GENRE	MAT1-2	SkK
Addback-C (tJML69.253)	Native + ectopic	No	GENRE	MAT1-2	SkK
Offspring-A3 (tJML69.37 P3)	Native	No	None	MAT1-1	SkS × SkK
Offspring-A5 (tJML69.37 P5)	Native	Yes	Nonee	MAT1-1	SkS × SkK
Offspring-A6 (tJML69.37 P6)	Native	No	None	MAT1-1	SkS × SkK
Offspring-B25 (tJML69.37 P25)	Native	Yes	None	MAT1-1	SkS × SkK
Fv-SKC1TAG>TAA (tJML56.208)	Native SKC1TAG>TAA + ectopic	No	None	MAT1-2	SkK
Offspring-C11 (tJML56.208 P11)	Native SKC1TAG>TAA	No	None	MAT1-2	SkS × SkK
Offspring-C37 (tJML56.208 P37)	Native SKC1TAG>TAA	No	None	MAT1-2	SkS × SkK

^aStrain sources: Fv149 and Fv999 (Leslie et al. 1992). All other Fusarium strains in this study were derived from Fv999 and/or Fv149.

^bThe status of SKC1 in each strain is indicated. Addback strains carry multiple copies of SKC1, 1 at the native locations and 1 or more at ectopic locations. Offspring-A3 and -A6 carry a functional version of SKC1 at the native location. Offspring-A5 and -B25 carry nonfunctional versions of SKC1 at the native location and at ectopic locations(s).

^cThe presence of multiple copies of SKC1 in Addback strains correlates with RIP-like mutations in the copy/copies of SKC1 passed to some descendants (see Table 3).

^dThe status of selectable markers in each strain is indicated. Each addback strain carries at least 1 ectopic copy of GENR^E.

^eOffspring-A5 has multiple, nonfunctional copies of GENR, due likely to RIP.

Table 2.

Neurospora crassa strains.

Name	Short name	Genotype
F2-26 (RTH1005.2)	wt	rid; fl a
ISU-3037 (RTH1623.2)	sad-2 Δ	rid; fl; sad-2 Δ ::hph+ a
ISU-4915 (CNR310.1.2)	SKC1	rid his-3+::SKC1::nat+; mus-52 Δ ::bar+ A
ISU-4987 (CNR315.1.1)	SKC1TAG>TGG	rid his-3+::SKC1TAG>TGG::nat+; mus-52 Δ ::bar+ A
ISU-4988 (CNR324.3.2)	SKC1TAG>TGA	rid his-3+::SKC1TAG>TGA::nat+; mus-52 Δ ::bar+ A
ISU-4989 (CNR325.2.2)	SKC1TAG>TAA	rid his-3+::SKC1TAG>TAA::nat+; mus-52 Δ ::bar+ A
P8-43	Host	rid his-3; mus-52 Δ ::bar+ A

The Neurospora strains used in this study are in the Oak Ridge genetic background (Perkins 2004). The rid genotype suppresses RIP (Freitag et al. 2002). The his-3 genotype results in histidine auxotrophy, while the fl genotype suppresses macroconidiation (Perkins et al. 2000). The mus-52^Δ allele increases transformation efficiency (Ninomiya et al. 2004). The sad-2^Δ genotype suppresses MSUD (Shiu et al. 2006). Mating types are A and a. The his-3⁺::SKC1 transgene in strains ISU-4915, ISU-4987, ISU-4988, and ISU-4989 were amplified from genomic DNA with primer set I297/I298 and confirmed to be free of unintended mutations by Sanger sequencing.

Crossing techniques

Crosses of F. verticillioides were performed using previously published protocols with some modifications (Klittich and Leslie 1988; Leslie and Summerell 2006). Crosses were incubated at 22°C with continuous white light provided by 2 fluorescent light bulbs and 12-h intervals of longwave ultraviolet light provided by 1 blacklight bulb (General Electric F40T12BL). Conidia from the female parent of each cross were collected from a V8A culture and placed at the center of a 100-mm Petri plate containing 30.0 ml of CA. After incubation for 1 week, the female parent was fertilized with conidia of the male parent by flooding a V8A culture (100 mm Petri plate) of the male parent with 6.0 ml of 2.5% Tween 60. A 2.0-ml aliquot of the resulting conidial suspension was then added to the 1-week-old CA culture of the female parent and spread over the surface with an L-shaped spreader. The female aerial hyphae and conidia were not removed prior to fertilization, rather they were mixed with the male conidia while spreading. The resulting crossing plate was returned to the incubator and washed at 2-day intervals as reported by Cavinder et al. (2012). That is, an 8.0 ml aliquot of 2.5% Tween 60 was spread over the surface of the plate and the resulting mixture of liquid and disrupted tissue was then poured off. Washes continued until perithecia were observed without magnification on the agar surface.

Unidirectional crosses were performed with Neurospora crassa strains on synthetic crossing medium as previously described (Hammond et al. 2011; Samarajeewa et al. 2014). Crosses were performed at room temperature under ambient lighting.

Perithecial dissections

Fusarium verticillioides perithecia were harvested 3 weeks postfertilization under a dissecting microscope, rolled on 5% water agar to remove excess vegetative tissue, and squashed under a coverslip in 25% glycerol. Perithecial contents (asci) were examined by standard light microscopy. N. crassa perithecia were harvested 12–16 days postfertilization (dpf) under a dissecting microscope and placed in a drop of 50% glycerol on a microscope slide. Syringe needles were used to tease perithecial contents into the surrounding liquid and remove vegetative debris and perithecial wall fragments. The perithecial contents were overlayed with a coverslip prior to examination by standard light microscopy.

General molecular techniques

FungiDB was used to download genome information for F. verticillioides and N. crassa (Basenko et al. 2018). Fungal DNA was routinely isolated with the Quick-DNA Fungal/Bacterial Miniprep Kit (Zymo Research, D6005) or with the Mini Genomic DNA Kit for Plants and Fungi (IBI Scientific, IB47231). PCR was typically performed with Primestar Polymerase (Clontech) or Q5 DNA polymerase (New England BioLabs). The sequences of all primers used in this study are provided in Supplementary Table 1. Gel extraction and DNA/PCR-product purification were performed with the PureLink Quick Gel Extraction Kit (Invitrogen, K210012) or the GEL/PCR DNA Fragment Extraction Kit (IBI Scientific, IB47030). Plasmid DNA was typically isolated from Escherichia coli using a standard alkaline lysis miniprep protocol. Gel electrophoresis was performed with standard 0.8–2% agarose-TAE (Tris-Acetate-EDTA) gels. Ethidium bromide staining and UV light sources were used to visualize DNA in agarose gels. Routine DNA sequence analysis was performed with Sequencher (Version 5.4.6, Gene Codes Corp.), DNAMAN (Version 7, LynnonBiosoft), MEGA X (Version 10.2.4, Kumar et al. 2018), and/or BioEdit (Version 7.2.5, Hall 1999). CLC Genomics Workbench (Version 20.0, Qiagen) and Tablet (Version 1.21.02.08, Milne et al. 2010) were used to visualize alignments of reads from genome and RNA sequencing datasets. Sanger sequencing was performed at NCAUR or the University of Illinois Urbana-Champaign CORE sequencing center.

Transformation vector and plasmid construction

Plasmid pJML38.1 contains SKC1-flanking sequences on either side of a hygromycin resistance cassette (HYGB). The upstream and downstream SKC1-flanking sequence were amplified from Fv999 genomic DNA with primer sets 2759/2760 and 2761/2762, respectively. HYGB was amplified from plasmid pJML31.1 (Kim et al. 2020) with primer set 2763/2764. The SKC1-flanking sequences were fused to HYGB by PCR with primer set 2767/2768 according to the PCR fusion protocol of Szewczyk et al. (2006). The amplified fusion product was inserted into the commercial vector pCR-XL-2-TOPO with the TOPO XL-2 Complete PCR Cloning Kit (Invitrogen, K805010) to create pJML38.1.

Plasmid pJML39.1 contains SKC1-flanking sequences on either side of a geneticin resistance cassette (GENR). The pJML39.1 construction method was nearly identical to the pJML38.1 construction method except that a fragment of DNA containing GENR was used in the fusion reaction instead of HYGB. GENR was amplified from plasmid pGenNotI (Desjardins et al. 2004) with primer set 2765/2766.

Plasmid pJML51.2 contains the SKC1 complementation (i.e. “addback”) fragment. pJML51.2 was constructed by amplifying a 2.5-kb product from Fv999 genomic DNA with primer set 2767/2768 and inserting the product into pCR-XL-2-TOPO.

Plasmid pJML53.3 is nearly identical to pJML51.2 except for a single site-directed mutation changing the sense-SKC1a stop codon from TAG to TAA. To construct this plasmid, primer set I1262/I1263 was first used to amplify and insert SKC1 into pJET1.2 with the CloneJET PCR Cloning Kit (ThermoFisher, K1231) to create plasmid pNR210.1. The G-to-A mutation was generated using the Q5 site-directed mutagenesis kit (New England BioLabs, E0554S) on pNR210.1 with primer set P2080/P2081 to create plasmid pNR229. A 1.3-kb fragment containing SKC1^TAG>TAA was then amplified from pNR229 with primer set 2915/2916 and fused by PCR to a 997-bp SKC1 upstream fragment (created with primer set 2759/2914) and a 795-bp SKC1 downstream fragment (created with primer set 2762/2917) using primer set 2767/2768 to generate a 2.5-kb fragment that was subcloned into pCR-XL-2-TOPO to create plasmid pJML53.3 (SKC1^TAG>TAA).

Plasmid pTH1325.7 contains an SKC1-spanning interval of DNA between the NotI and EcoRI sites of plasmid pNR28.12 (GenBank MH553564.1). The SKC1 fragment in pTH1325.7 was amplified from Sk^K (Fv999) genomic DNA with primer set I1262/I1263. Plasmids pNR214.4, pNR231.8, and pNR232.1 are nearly identical to pTH1325.7 except for a single site-directed SKC1 mutation in each plasmid. The Q5 Site-Directed Mutagenesis Kit was used with pNR210.1 (see above) and 3 different primer sets to create the following mutant constructs: SKC1^TAG>TGG (primer set I2039/I2040), SKC1^TAG>TAA (primer set I2080/I2081), and SKC1^TAG>TGA (primer set I2080/I2082). The mutant sequences were then inserted between the NotI and EcoRI sites of pNR28.12 to create plasmids pNR214.4 (SKC1^TAG>TGG), pNR231.8 (SKC1^TAG>TAA), and pNR232.1 (SKC1^TAG>TGA). All site-directed mutations were verified by Sanger sequencing.

Fusarium and Neurospora transformations

The F. verticillioides transformation procedure used in this study was modified from previously published protocols (Turgeon et al. 1987; Salch and Beremand 1993; Proctor et al. 1999) as described below. Conidia were collected from the recipient strain after 7 days of growth on 20 ml of V8A in a 100-mm Petri plate for approximately 1 week into 15 ml of water. The spore suspension was then filtered through 3 layers of sterile Miracloth (MilliporeSigma, 475855), added to 100 ml of GYP (2% glucose, 0.3% yeast extract, 1% peptone), and incubated at 28°C and 200 rpm until germ tubes were approximately 3–10 times the length of an ungerminated conidium (approximately 8 h).

Germlings were collected by centrifugation at 1,408g for 10 min, washed with 10 ml of 0.7 M NaCl, and resuspended in 20 ml of 0.7 M NaCl containing 15 mg/ml Lysing Enzymes (MilliporeSigma, L1412), 40 mg/ml Driselase (MilliporeSigma, D9515), and 0.1 mg/ml chitinase (MilliporeSigma, C6137). The digestion reaction was incubated at 30°C and 100 rpm until most germ tubes had been digested (1–3 h). Protoplasts were collected by centrifugation for 10 min at 1,408g, washed once with 10 ml of 0.7 M NaCl, and resuspended in STC buffer (1.2 M sorbitol, 10 mM CaCl₂, and 10 mM Tris-HC1, pH 8.0) at a concentration of 5–10 × 10⁸ protoplasts/ml.

To replace SKC1 with HYGB, DNA was amplified from pJML38.1 with primer set 2767/2768. To replace SKC1 with GENR, DNA was amplified from pJML39.1 with primer set 2767/2768. To complement ΔSKC1::HYGB with SKC1, 2 DNA molecules were amplified; 1 DNA molecule was amplified from pJML51.2 with primer set 2767/2768, and the other was amplified from pGenNotI with primer set 739/740. For transformation of protoplasts with a single DNA molecule, 15–20 µg of PCR product was added directly to 300 µl of protoplasts. A solution of STC-PCT was freshly prepared by mixing equal volumes double strength (2×) STC with PCT (30% polyethylene glycol 8000, 50 mM CaCl₂, and 10 mM Tris-HCl, pH 8.0). A 1.2 ml aliquot of STC-PCT was then added to the transformation reaction (i.e. the DNA + protoplast suspension). After 1-h incubation at room temperature without agitation, 4 ml of STC was added to the transformation reaction, which was followed by the addition of 2 ml of PCT. The transformation reaction was then incubated for 1 h at room temperature without agitation before plating. For cotransformation of protoplasts with 2 DNA molecules, 50–60 µg of each PCR product was added to 600 µl of protoplasts. All other reagent volumes were doubled relative to transformation with one molecule except for the second STC and PCT treatments in which the STC (4 ml) and PCT (2 ml) volumes remained the same.

To plate transformation reactions, the entire reactions were combined with 100 ml of molten (50–55°C) regeneration medium [RGM, 34.2% sucrose, 0.1% yeast extract, 0.1% casein hydrosylate (enzymatic), and 2% agar], poured (in 10-ml aliquots) onto 20 ml of solidified RGM in 100 mm culture plates, and incubated overnight. The next day, the plates were overlaid with 10 ml of molten (50–55°C) 1% water agar containing 600 µg/ml hygromycin B (Thermo Fisher, 10687010) or 1 mg/ml G418 Sulfate (Thermo Fisher, 10131027). After several days, colonies that grew through the water agar layer were transferred to separate 60-mm culture plates containing 10 ml of V8A without antibiotic. After producing conidia, putative transformants were screened by conidial PCR. PCR-positive transformants were single spore purified before additional screening by genome sequencing and/or PCR-based genotyping assays.

Neurospora crassa transformations were performed as described by Rhoades et al. (2020). All plasmid-based transformation vectors were linearized by restriction digestion with SspI and purified with the GEL/PCR DNA Fragment Extraction Kit. Homokaryotic transformants were obtained by plating dilute concentrations of conidia on selective BDS medium (Brockman and De Serres 1963; Rhoades et al. 2020) and screening single-conidium-derived colonies by PCR. Strain CNR310.1.2 was constructed by transforming strain P8-43 with pTH1325.7. CNR315.1.1 was constructed by transforming P8-43 with pNR214.4. CNR324.3.2 was constructed by transforming P8-43 with pNR232.1. CNR325.3.2 was constructed by transforming P8-43 with pNR231.8. The homokaryotic nature of CNR310.1.2, CNR315.1.1, CNR324.3.2, and CNR325.3.2 were confirmed with primer set I297/I298.

Preliminary screening of transformants by conidial PCR

Fusarium verticillioides transformants with the desired genetic modification were screened by conidial PCR essentially as described by Henderson et al. (2005). For each transformant, an inoculation loop of conidia (see Henderson et al. 2005 for definition) was suspended in 300 µl of TE (10 mM Tris-HCl, 1 mM EDTA, pH 8.0) in a 0.5-ml microcentrifuge tube and heated at 100°C for 10 min in a thermal cycler. After cooling to room temperature, cellular debris was pelleted by centrifuging for 5 min at 9,391g. PCR was performed on 2.0 µl of the supernatant with the Phire Plant Direct PCR Master Mix (Thermo Fisher, F170S; 10 µl final reaction volume). The following primer sets were used to screen transformants for (1) ΔSKC1::HYGB: 2757/2759, 2758/2762, and 2356/2357; (2) ΔSKC1::GENR: 2695/2759, 2696/2762, and 2356/2357; and (3) SKC1 and SKC1^TAG>TAA addback strains: 2705/2759, 2706/2762, and 2356/2357.

Genome sequencing and analysis

Genomic DNA for sequencing was isolated with the Quick-DNA Fungal/Bacterial Miniprep Kit. DNA libraries were prepared with the Nextera XT DNA Library Preparation Kit for Illumina Systems (FC-131-1096) and 1 ng of genomic DNA. DNA libraries were sequenced with MiSeq Reagent Kit Version 3 and a MiSeq sequencer (Illumina). Raw reads were processed by removing adapters and low-quality reads were trimmed and filtered with CLC Genomics Workbench as previously reported (Brown et al. 2020).

RNA isolation

Fusarium verticillioides total RNA was isolated from Sk^K-vegetative tissue, Sk^S-vegetative tissue, and Sk^S × Sk^K perithecia. Vegetative tissues were collected from liquid shaking cultures of Sk^K (Fv999) or Sk^S (Fv149) (100 ml GYP, 200 rpm, 28°C, and ambient light in 500 ml flask), while perithecia were collected from an Sk^S (Fv149) × Sk^K (Fv999) cross, where Sk^S served as female, 20 dpf. The vegetative and perithecial tissue were lyophilized and stored at −80°C or on dry ice until RNA isolation, which was performed as described by Rhoades et al. (2019) with the exception that dried rather than fresh perithecia were ground in TRIzol reagent (ThermoFisher, 15596026). The procedure of Rhoades et al. (2019) was also used to isolate total RNA from fresh 10 dpf perithecia of an N. crassa sad-2^Δ (ISU-3037) × SKC1 (ISU-4915) cross, where sad-2^Δ served as the female parent.

RNA sequencing and analysis

Two strand-specific RNA sequencing datasets were produced in this study. The mycelial dataset was made with a 4-day vegetative culture of Fv999 (described above), while the perithecial dataset was made with 20 dpf perithecia from an Sk^S (Fv149) × Sk^K (Fv999) cross (described above). RNA sequencing was performed on mRNA-enriched total RNA by Novogene (Sacramento, CA, USA). The mRNA enrichment was performed with an oligo(dT)-based method, and sequencing was performed with a PE150-sequencing strategy and Illumina’s NovaSeq 6000 sequencing system. Raw reads were filtered by Novogene to remove reads containing adapters, reads containing more than 10% undetermined base sequences, and reads with more than 50% low quality (Phred) base calls. Filtered reads were trimmed in CLC Genomics Workbench using the NGS Core Tools Trim sequences tool set at default values. The Adapter trim list consisted of the 2 Illumina stranded RNA adapter sequences. The Quality trim parameter limit was 0.05 and the Ambiguous trim parameter was set at 2 residues.

To examine SKC1 transcription, “Read 2” from each pair of reads was mapped to positions 95,939–100,297 of GenBank KU963213.1, which contains SKC1, the intergenic sequences flanking SKC1, and the predicted coding sequences of SKC1’s flanking genes. A custom python script was used to calculate coverage levels and Microsoft Excel was used to produce coverage level charts. Reads that mapped to the reference with more than 2 mismatches were ignored. Coverage levels in the mycelial and perithecial datasets were used to predict transcript boundaries. Specifically, positions with a coverage level greater than 0.5% that of the highest level along the predicted sense or antisense transcripts were included in the transcript models.

RT-PCR and cDNA analysis

For RT-PCR analysis of sense SKC1 transcripts, first-strand cDNA synthesis was performed on total RNA with the ProtoScript II First Strand cDNA Synthesis Kit (New England BioLabs, E6560S) with the included Randomized Primer Mix. The reverse transcription products were used as templates in PCR reactions with primer set I2070/I2071. RT-PCR products were analyzed by gel electrophoresis, purified by gel extraction, and analyzed by Sanger sequencing with primer I2074. A similar process was used for RT-PCR analysis of antisense SKC1 transcription except that the reverse transcription step was primed with an antisense SKC1 specific primer (I2221 or I2222) instead of random primers. The reverse transcription products were used as templates in PCR reactions with primer set I2225/I2226. The resulting PCR products were gel purified, cloned into pJET1.2 (thereby creating plasmids pNR283.1 and pNR283.6), and sequenced by Sanger sequencing.

SCK1 phylogenetic analysis

SKC1 homologs from Fusarium species other than F. verticillioides were detected by BLASTN analysis of a previously described genome sequence database housed locally in CLC Genomics Workbench (Kim et al. 2020; Geiser et al. 2021). The sequences were generated locally or downloaded from GenBank/NCBI as previously described (Kim et al. 2020; Geiser et al. 2021). Query sequence for BLASTN analysis (default parameters of word size of 11, match 2, mismatch -3 with gap costs of existence 5 and extension 2) was the SKC1 homolog from F. verticillioides Fv999. The sequences of the BLASTN hits were retrieved and subjected to phylogenetic analyses as implemented in MEGA X (Kumar et al. 2018). First, the homologous SKC1 ORFs were predicted based on comparison to SKC1 from F. verticillioides. Next, the coding sequences were converted to amino acid sequence, aligned using the MUSCLE method (Edgar 2004), converted back to the nucleotide sequence, and subjected to maximum parsimony (MP) and maximum likelihood (ML) analyses with 500 bootstrap replicates.

Results

Deleting SKC1 abolishes Sk^K-mediated spore killing

Crosses between Sk^S and Sk^K strains of F. verticillioides produce asci with 4 ascospores instead of 8 because of spore killing (Kathariou and Spieth 1982; Raju 1994). To determine if SKC1 is required for spore killing, we deleted the gene in an Sk^K strain by replacing it with either of the selectable marker genes HYGB or GENR (Fig. 1a and Supplementary Figs. 1 and 2). When the SKC1 deletion mutants (skc1 mutants) were crossed with an Sk^S strain, the resulting asci had 8 ascospores (Fig. 1, c and d). In contrast, asci from control, Sk^S × Sk^K, crosses had 4 ascospores (Fig. 1b). These results indicate that SKC1 is required for Sk^K-mediated spore killing.

Fig. 1.

Deletion of SKC1 eliminates spore killing. a) A diagram of the SKC1 deletion strategy is shown. Coding regions are represented by white rectangles. Promotor, terminators, untranslated sequences, and intergenic sequences are shown as black bars. Intron locations are not indicated. Fv3165 and Fv3164 are SKC1-flanking genes. SKC1 coding sequences span positions 97,921 through 98,223. The deleted region (orange bar) includes the SKC1 coding sequences up to the predicted stop codon in the genomic sequence (97,921–98,133). The deleted region was replaced with HYGB to create ΔSKC1::HYGB and GENR to create ΔSKC1::GENR. Positions are according to GenBank KU963213.1. The diagram is drawn to scale and the orange bar represents 213 base pairs (bp). b) Sk^S × Sk^K produces 4-spored asci. c) Sk^S × ΔSKC1::HYGB produces 8-spored asci. d) Sk^S × ΔSKC1::GENR produces 8-spored asci. Unidirectional crosses were performed with Sk^S as the female parent. Red and blue arrows mark 4-spored and 8-spored asci, respectively.

Complementation with SKC1 restores Sk^K-mediated spore killing

To confirm that SKC1 is required for spore killing, we co-transformed an skc1 mutant (ΔSKC1::HYGB) with 2 DNA molecules: GENR and a 3-kb DNA fragment from the wild-type Sk^K locus that included SKC1. Our goal was to restore SKC1 to the Sk^K locus of the skc1 mutant via homologous recombination by replacing HYGB with a wild-type copy of SKC1 and at the same time introduce GENR at an ectopic location (Fig. 2a). Following the cotransformation protocol, we recovered 3 skc1 mutant-derived transformants with a wild-type copy of SKC1 reintegrated into the Sk^K locus replacing HYGB. The 3 transformants were designated as SKC1 addback strains Addback-A, Addback-B, and Addback-C (Supplementary Fig. 3). In crosses with an Sk^S mating partner, Addback-A and Addback-B produced some asci with 4 ascospores and some asci with 8 ascospores (Fig. 2, b and c). In contrast, nearly all asci produced from crosses of Sk^S and Addback-C had 4 ascospores (Fig. 2d). Taken together, these results demonstrate that loss of Sk^K-mediated spore killing, hereafter referred to as spore killing, in the skc1 mutant was restored by complementation with SKC1.

Fig. 2.

SKC1 restores spore killing to ΔSKC1::HYGB. a) A diagram of the cotransformation-based SKC1 complementation strategy is shown. The procedure was designed to allow HYGB (marked by the orange bar) in ΔSKC1::HYGB to be replaced with the missing segment of SKC1 (positions 97,921–98,133) while GENR integrates at an ectopic location in the genome. A portion of pGenNotI derived plasmid sequences was amplified along with the GENR cassette (dashed black line) for use in the transformation procedure. The diagram was drawn to scale. For reference, the orange bar represents 1,451 bp. Other style conventions are as in Fig. 1a. b) Sk^S × Addback-A produces a mix of 4-spored and 8-spored asci. c) Sk^S × Addback-B produces a mix of 4-spored and 8-spored asci. d) Sk^S × Addback-C produces 4-spored asci (8-spored asci were not observed). Unidirectional crosses were performed with Sk^S as the female. Red and blue arrows mark 4- and 8-spored asci, respectively.

Multiple copies of SKC1 trigger RIP in F. verticillioides

To investigate the cause of the inconsistent spore killing complementation phenotypes of Addback-A and Addback-B, we generated genome sequence data for the 3 addback strains. This analysis revealed that all 3 strains lacked HYGB (Supplementary Fig. 4), which is consistent with successful replacement of ΔSKC1::HYGB deletion construct with SKC1 in each strain. Furthermore, all 3 strains had GENR (Supplementary Fig. 5), which is consistent with our use of GENR in the addback transformation procedure. In addition, all 3 strains possess SKC1, which was absent in their progenitor strain, the skc1 mutant (Supplementary Fig. 6). However, the number of SKC1-specifc reads present in the genome sequence datasets of the addback strains suggests that each genome contained more than 1 copy of SKC1 (Supplementary Fig. 6). Because our PCR-based genotyping assays indicated that the Sk^K locus of each addback strain had only 1 copy of SKC1 (Supplementary Fig. 3), the additional copy(ies) of SKC1 must have been located at an ectopic location(s).

Repeat-induced point (RIP) mutation occurs in some ascomycetous fungi during a cross when 1 parent contains 2 or more copies of a DNA sequence and results in the presence of multiple C-to-T transitions in the repeated sequence in progeny (Leslie and Dickman 1991; Aramayo and Selker 2013; Gladyshev 2017). If the repeated sequence is a gene, RIP can render all copies nonfunctional. Given the multiple copies of SKC1 in the addback strains, RIP provides a possible explanation for the incomplete complementation phenotype of Addback-A and Addback-B. Therefore, we analyzed the genome sequences of 3 progeny from crosses of an Sk^S strain and Addback-A (progeny A3, A5, and A6) and 1 progeny from Addback-B (progeny B25). This analysis revealed that SKC1 sequences in A5 and B25 had many C-to-T transitions, whereas SKC1 sequences in A3 and A6 lacked such mutations (Table 3). The C-to-T transitions in SKC1 sequences in A5 and B25 resulted in changes in the amino acid sequences as well as premature stop codons. Thus, RIP of SKC1 sequences of some addback progeny and no RIP in other progeny is consistent with the incomplete complementation phenotype of Addback-A and Addback-B. That is, C-to-T transitions that affect SKC1 function are expected to block spore killing and result in asci with 8 ascospores, whereas the absence of C-to-T transitions are expected to allow spore killing to occur and result in asci with 4 ascospores.

Table 3.

An analysis of SKC1 mutations in offspring from parents with multiple copies of SKC1.

	SKC1 and flanking DNA				SKC1 coding sequence only
	Pos: 97,592–98,261				Pos: 97,921–98,223
	C > T	G > A	other	Total	C > T	G > A	Other	Total
Addback-A (parent)	0	0	0	0	0	0	0	0
Offspring-A3	0	0	0	0	0	0	0	0
Offspring-A5	32	18	0	50	15	13	0	28
Offspring-A6	0	0	0	0	0	0	0	0
Addback-B (parent)	0	0	0	0	0	0	0	0
Offspring-B25	0	15	0	15	0	10	0	10

Mutations are classified as C to T, G to A, and all others (e.g. A to C, A to G). Genome sequencing reads from each strain were mapped to positions 95,939–100,297 of GenBank KU963213.1. This region is present in Sk^K (Fv999) but not Sk^S (Fv149). Mapped reads were examined for mutations. Only mutations present in at least 20% of reads covering each position were counted.

SKC1 transcripts undergo A-to-I editing during sexual development

Previously, we examined transcript levels of 42 SKC1-linked genes during 6 stages of sexual development, from 2 h postfertilization to the early spore ejection stage (Pyle et al. 2016). We accomplished this by analyzing RNAseq datasets that had been generated from a cross between Sk^K (Fv999) and Sk^S (Fv149), where Sk^K served as female (Sikhakolli et al. 2012). A significant finding from this analysis was the discovery that SKC1 transcripts are present at relatively high levels through all analyzed stages of sexual development. For example, of the 42 genes analyzed, SKC1 transcripts were the most numerous at the first time point and the second most numerous at the last time point. The finding that SKC1 transcript levels are relatively high even at early stages of sexual development raises the possibility that SKC1 transcription is not tightly linked to sexual development.

To further examine SKC1 expression, we performed strand-specific RNA sequencing on RNA isolated from 2 developmental stages of F. verticillioides: (1) vegetative tissue of an Sk^K after 4 days of growth on CA (mycelial dataset) and (2) perithecia from an Sk^S × Sk^K cross 20 dpf (perithecial dataset). Mapping RNA sequence reads from vegetative and perithecial tissue to the SKC1 locus revealed that SKC1 is bidirectionally transcribed (Fig. 3, a and b). Here, we define reads that matched the reference SKC1 mRNA (GenBank KU963213.1) as “sense” sequences (Fig. 3, blue bars) and those that were complementary to the reference SKC1 mRNA as “antisense” sequences (Fig. 3, red bars).

Fig. 3.

SKC1 is bidirectionally transcribed. a) RNA sequences from the mycelial dataset were mapped to an SKC1-containing reference sequence using custom python scripts (Supplementary Fig. 15). The chart depicts RNA coverage levels for positions 97,586 through 98,819 of the previously defined 102-kb SKC1 locus (GenBank KU963213.1). Blue and red vertical bars indicate coverage levels for the reference (“sense”) strand and nonreference (“anti-sense”) strand, respectively. Antisense coverage levels were used to predict the location of a + 1 site, 2 exons, an intron, and a transcriptional cleavage site (TCS) for a transcript called antisense-SKC1a (see Materials and methods). A variant transcript called antisense-SKC1b results when the intron is retained in the antisense transcript. b) RNA sequences from the perithecial dataset were mapped to an SKC1-containing reference sequence. Sense coverage levels were used to predict the location of a +1 site, 2 exons, an intron, a TCS, a start codon, and 2 stop codons for transcripts called sense-SKC1a and sense-SKC1b. The transcripts are nearly identical, except that codon 71 is a UAG codon in sense-SKC1a and a UIG codon in sense-SKC1b.

The ratio of antisense to sense SKC1 sequences was higher in mycelia than perithecia (Fig. 3, a and b). Thus, we used the mycelial dataset to derive a model of antisense SKC1 transcription. Specifically, we predicted that 2 antisense SKC1 transcripts, antisense-SKC1a and antisense-SKC1b (with intron), are transcribed from SKC1. Our model posits that the transcripts possess the same +1 site and transcriptional cleavage site, but they differ with respect to intron sequence that is removed from antisense-SKC1a and retained in antisense-SKC1b (Fig. 3a and Supplementary Fig. 7a). Results of an RT-PCR analysis of RNA isolated from F. verticillioides mycelia were consistent with the existence of both antisense-SKC1a and antisense-SKC1b (Supplementary Fig. 7, b–d). While the role these transcripts play in spore killing is unclear, both lack obvious coding sequences, which suggests that they are noncoding RNAs.

The ratio of sense to antisense SKC1 sequences was higher in perithecia than mycelia (Fig. 3, a and b). Thus, we used the perithecial dataset to derive a model of sense SKC1 transcription. By examining RNA sequence coverage levels across SKC1, we identified a transcript called sense-SKC1 that included +1 site, an intron, a transcriptional cleavage site, a translational start codon, and a UAG stop codon (Fig. 3b). However, manual inspection of aligned sequences revealed many reads that spanned the stop codon of sense-SKC1 but some reads had UGG, which is a codon sequence for tryptophan, rather than the stop codon sequence UAG. The change in sequence from UAG to UGG is evidence of A-to-I RNA editing (Bian et al. 2019). Using both the mycelial and perithecial datasets, we then examined all sense RNA sequence reads and calculated the percentage of sequences showing evidence of A-to-I editing (Table 4). There was evidence for A-to-I editing in 84% of reads from perithecia but only 0% of reads from mycelia (Table 4). These results suggest that the A-to-I edit in sense-SKC1 is specific to sexual development. We confirmed the read mapping results by sequencing RT-PCR products that spanned the UAG codon in sense-SKC1 (Fig. 4, b and c). As expected, transcripts with the A-to-I edit were absent in mycelia and present in perithecia. In contrast to the A-to-I edit, analysis of the RT-PCR amplicon sequence revealed that the SKC1 intron in sense SKC1 transcripts was absent from both mycelia and perithecia (Fig. 4c). Overall, these data suggest that SKC1 is transcribed to produce at least 2 sense transcripts: transcript sense-SKC1a, which has the originally predicted UAG stop codon, and transcript sense-SKC1b, which has the UGG, tryptophan coding codon, formed by A-to-I editing and a UGA stop codon formed after excision of the intron (Figs. 3b and 4a).

Table 4.

RNA editing during sexual development.

RNAseq dataset	Totala	UAGb	UGGc	A → Id
Vegetative (Mycelia)	215	215	0	0.0%
Perithecial (Perithecia)	997	161	835	83.8%

^aThe number of sense RNA sequencing reads that map to the SKC1 A-to-I edit site in the mycelial or perithecial datasets.

^bNumber of UAG reads.

^cNumber of UGG reads.

^dPercentage of A-to-I editing (UGG/[UAG + UGG]). One read was UCG in the perithecial dataset (161 + 835 + 1 = 997). Three reads had multiple base changes across UAG and thus were ignored.

Fig. 4.

Fungal A-to-I mRNA editing of SKC1’s UAG stop codon to UIG occurs in perithecia. a) The sequence of sense-SKC1 is shown. Two sequence motifs of unknown function are present in the 5′ UTR: ACACAA (double underline) and ACACAAUUCCAUCAAUCCACU (single underline). The locations of a putative start codon (green), UAG stop codon (blue), and intron (gray) are highlighted. The sense-SKC1b sequence contains a single nucleotide difference relative to the sense-SKC1a sequence: the UAG stop codon has been edited to a UIG tryptophan codon. The stop codon for sense-SKC1b is the UGA codon (red highlight) formed by intron removal. b) SKC1 RT-PCR products were analyzed by gel electrophoresis. Lanes 1 and 8 contain 0.5 µg of GeneRuler 1 kb Plus DNA Ladder (ThermoFisher, 10787018). RNA: K, S, and P indicate that total RNA came from Sk^K-vegetative tissue, Sk^S-vegetative tissue, or Sk^S × Sk^K perithecia (see RNA isolation). RT: “+” indicates that reverse transcriptase was included in the first strand synthesis reaction, while “–” indicates that the enzyme was omitted from the first strand synthesis reaction. c) Sequence analysis of the RT-PCR products in lanes 2 (top) and 6 (bottom) of (b). The chromatogram segments span either the A-to-I edit site (left) or the exon–exon junction (right). The green arrow in the bottom left panel highlights UAG to UIG editing detected only in perithecial tissue. Only the “A” peak is present in vegetative tissue. The black arrows highlight the exon–exon junction as expected in vegetative and perithecial tissue.

While designing PCR primers for analysis of sense SKC1 transcripts by RT-PCR, we discovered repeated elements within the putative 5ʹ untranslated regions (UTR) of sense-SKC1a and sense-SKC1b (Fig. 4a). Element ACACAA is repeated 5 times within the first 116 nucleotides of the transcripts. The other repeat is 21 nucleotides long, contains the first repeat, and is found 3 times within the first 104 nucleotides of each transcript (Fig. 4a). As these repeats occur in the 5′ UTR of SKC1, they do not impact Skc1 protein sequence.

SKC1 disrupts ascospore development in N. crassa

To determine if SKC1 can cause spore killing in other fungi, we inserted F. verticillioides SKC1 into the genome of N. crassa (Fig. 5a). During meiosis in N. crassa, genes present in unpaired DNA are repressed by a process called meiotic silencing by unpaired DNA (MSUD) (Aramayo and Selker 2013; Hammond 2017). Thus, we expected SKC1 to be silenced when wild-type N. crassa (Nc-wt) was crossed with an N. crassa strain with SKC1 (Nc-SKC1) because SKC1 would be unpaired in such crosses. Indeed, asci from Nc-wt × Nc-SCK1 crosses had the same number of ascospores as asci from crosses of 2 Nc-wt strains. To overcome MSUD, we crossed Nc-SKC1 with an N. crassa strain (Nc-sad-2^Δ) in which MSUD was suppressed (Shiu et al. 2006). In Nc-sad-2^Δ × Nc-SCK1 crosses, ascus development stalled and/or aborted in most perithecia and only a few viable ascospores formed (Fig. 5, d and e and Supplementary Fig. 8a). In Nc-sad-2^Δ × Nc-wt crosses, by contrast, the vast majority of asci and ascospores developed to maturity and produced viable ascospores. To determine if SKC1 is transmitted in a biased manner to the few viable ascospores produced by the Nc-sad-2^Δ × Nc-SCK1 cross, we examined 32 offspring for the presence of SKC1. Nineteen of the offspring carried the SKC1 allele, which, according to a simple chi-square test, is consistent with a 50% transmission rate. Together, these results indicate that SKC1 restricts both ascospore and ascus development in N. crassa but also that, among the small number of viable ascospores that are formed, SKC1 is not inherited in a biased manner.

Fig. 5.

SKC1 restricts ascus development in N. crassa. a) A DNA fragment containing SKC1, SKC1’s intergenic flanks, and short segments of SKC1’s flanking genes’ coding sequences were ligated to a nourseothricin resistance cassette (nat-1) and used to replace a 919-bp segment of intergenic DNA (red bar) between the his-3 and rat-1 genes on chromosome I in N. crassa. b) The SKC1 gene is unpaired during meiosis in meiotic cells of a Nc-wt × Nc-SKC1 cross. This phenomenon is symbolized by the “unpaired loop” in the diagram of an alignment between the his-3 locus of a wt parent and the his-3 locus of an SKC1 parent. The SKC1 gene is silenced during meiosis by MSUD because it is found within this unpaired loop. c) Nc-wt × Nc-host crosses and Nc-wt × Nc-SKC1 crosses produce 8-spored asci. “Host” is the transformation host used for transformation with various SKC1-containing transgenes. d) Nc-sad-2^Δ× Nc-host crosses produce 8-spored asci while Nc-sad-2^Δ× Nc-SKC1 crosses produce underdeveloped and/or aborted asci. e) Culture plates and lids for 4 types of crosses are shown. In N. crassa, mature spores are shot upwards by the perithecia and stick to the lids of culture plates. The 3 “black” circular shapes in each culture dish are masses of perithecia. The 3 diffuse black circular areas on the culture lids are spores that have collected on the lids. The lids have been removed from the culture plates to allow a qualitative comparison of perithecia and spore levels. The images show that, despite producing perithecia at levels that are roughly equivalent to the 3 control crosses (i.e. Nc-wt × Nc-host, Nc-wt × Nc-SKC1, and Nc-sad-2^Δ × Nc-host), Nc-sad-2^Δ × Nc-SKC1 crosses produced few spores.

SKC1 transcripts undergo A-to-I editing in a heterologous system

We also examined SKC1 transcripts in N. crassa for evidence of A-to-I editing of the UAG stop codon. To do this, RNA recovered from perithecia from an Nc-sad-2^Δ × Nc-SKC1 cross was used as a template to amplify and then sequence SKC1 transcripts. The analysis indicated that some SKC1 transcripts had the original stop codon sequence (UAG), while others had the edited sequence (UIG) (Fig. 6a). The analysis also revealed that the intron sequence was absent in SKC1 transcripts in N. crassa regardless of whether the stop codon sequence was edited (Fig. 6b). These results provide evidence that SKC1 transcripts undergo A-to-I editing and intron processing in N. crassa in the same manner as in F. verticillioides.

Fig. 6.

SKC1 function in N. crassa requires A-to-I mRNA editing. a) Sequence of an SKC1 RT-PCR product amplified from total RNA of Nc-sad-2^Δ × Nc-SKC1 perithecia (10 dpf). The chromatogram shows a major “G” peak and a minor “A” peak at the edit site, indicating that A-to-I editing of SKC1’s UAG stop codon occurs in N. crassa as it does in F. verticillioides. b) Sequence of the same amplicon described in (a) spaning the SKC1 exon–exon junction. The chromatogram shows that the exon–exon junction forms properly in N. crassa. c, top row) Nc-wt × Nc-host, Nc-wt × Nc-SKC1, and Nc-wt × Nc-SKC1^TAG>TGA crosses produce 8-spored asci. c, bottom row) Nc-sad-2^Δ× Nc-host and Nc-sad-2^Δ× Nc-SKC1^TAG>TGA crosses produce 8-spored asci, while Nc-sad-2^Δ× Nc-SKC1 crosses produce underdeveloped/aborted asci. d) Crosses of Nc-wt × Nc-SKC1^TAG>TGG produce 8-spored asci, while Nc-sad-2^Δ× Nc-SKC1^TAG>TGG crosses produce underdeveloped/aborted asci.

Codon change required for spore killing

The A-to-I editing of the SKC1 transcript observed in both F. verticillioides and N. crassa raises the question: is the change in the codon sequence from UAG to UIG required for spore killing? To address this in F. verticillioides, we complemented the skc1 mutant with a variant of SKC1 in which the stop codon sequence was changed from TAG to TAA (SKC1^TAG>TAA) (tJML56.208) (Fig. 7a). We rationalized that the SKC1^TAG>TAA transcript could undergo A-to-I editing but that the edit from UAA to UGA would result in a stop codon that would prohibit production of the 84 amino acid protein (Skc1b).

Fig. 7.

Sense-SKC1b is required for spore killing. a) Control Sk^S (Fv149) × Sk^K (Fv999) cross produced 4-spored asci. b) Sk^S × Fv-SKC1^TAG>TAA C-11 and Sk^S × Fv-SKC1^TAG>TAA C-37 crosses produced 8-spored asci. Unidirectional crosses were performed with Sk^S (Fv149) as the female strain. Red and blue arrows mark 4- and 8-spored asci, respectively.

PCR and genome sequence analysis of the skc1 mutant complemented with SKC1^TAG>TAA (transformant Fv-SKC1^TAG>TAA) revealed that a single copy of SKC1^TAG>TAA had replaced HYGB and that additional copies of SKC1^TAG>TAA had integrated elsewhere in the genome (Supplementary Figs. 9 and 10). We removed the ectopic copies of SKC1^TAG>TAA through meiotic reassortment by crossing transformant Fv-SKC1^TAG>TAA with Sk^S strain Fv149. Genome sequence analysis of 2 of the resulting progeny, Offspring-C11 (tJML56.208 P11) and Offspring-C37 (tJML56.208 P37) revealed that both had only a single copy of the SKC1^TAG>TAA gene that was positioned at the native SKC1 locus and did not harbor any mutations indicative of RIP (Supplementary Fig. 11). Crosses of Offspring-C11 or Offspring-C37 with Sk^S strain Fv149 yielded asci with 8 ascospores, which indicated an absence of spore killing (Fig. 7c). These results provide evidence that the A-to-I editing that changes the stop codon (UAG) of the SKC1 transcript to a tryptophan-specifying codon (UIG) is required for spore killing in F. verticillioides. This in turn indicates that transcript sense-SKC1b is required for spore killing but sense-SKC1a is not required.

To address the role of A-to-I editing of the SKC1 stop codon in N. crassa, we introduced 2 variants of the SKC1 gene into the fungus. In the first variant, we changed the stop codon sequence to TAA (SKC1^TAG>TAA), as described above for F. verticillioides. In the second variant, the stop codon sequence was changed from TAG to TGA (SKC1^TAG>TGA). For SKC1^TAG>TGA, we rationalized that transcripts could not undergo A-to-I editing because the target adenosine had been changed to guanine. In Nc-sad-2^Δ × Nc-SKC1^TAG>TGA and Nc-sad-2^Δ × Nc-SKC1^TAG>TAA crosses, ascus and ascospore development was not restricted, indicating that both SKC1^TAG>TGA and SKC1^TAG>TAA did not kill ascospores or interfere with their development (Fig. 6c and Supplementary Fig. 12).

To further refine the sequence that is responsible for spore killing, we generated an N. crassa strain with a third SKC1 gene variant (SKC1^TAG>TGG), in which the stop codon sequence was changed from TAG to the tryptophan codon sequence TGG. In Nc-sad-2^Δ × Nc-SKC1^TAG>TGG crosses, ascus and ascospore development were restricted in the same manner as in Nc-sad-2^Δ × Nc-SKC1 crosses (Fig. 6, c and d). Together the results of the SKC1 variant expression in F. verticillioides and N. crassa indicate that the absence of the stop codon at position 383 of the SKC1 transcript is required for spore killing, and that a mixture of edited and unedited transcripts is not required for spore killing.

Distribution, phylogenetic relationships, and location of SKC1 in Fusarium

To investigate the distribution of SKC1 in Fusarium, we used BLASTn analysis of an inhouse database with genome sequences of 186 species representing all 23 multispecies lineages (species complexes) of the genus (Kim et al. 2020; Geiser et al. 2021). With SKC1 as query, we detected 22 homologs in 19 species from 6 closely related species complexes: the Fusarium burgessii (Burgessii), Fusarium newnesense (Newnesense), Fusarium nisikadoi (Nisikadoi), Fusarium oxysporum (Oxysporum), Fusarium redolens (Redolens), and Fusarium fujikuroi (Fujikuroi) species complexes (Fig. 8). The 6 species complexes constitute a monophyletic lineage within Fusarium, however, within most of these complexes the presence of SKC1 was discontinuous. For example, our Fusarium genome sequence database includes over 40 species within Fujikuroi, but SKC1 was detected in only 12 of the species, including F. verticillioides. In the database used for this analysis, most species were represented by a single genome sequence. Therefore, a lack of detection of SKC1 in a genome sequence could be because the species lacks SKC1 or because the sequence was from an Sk^S strain of a species that has SKC1. With 1 exception, all species in which SKC1 was detected had only 1 copy of the gene per genome sequence. The exception was Fusarium foetens; the 1 genome sequence examined had 3 SKC1 paralogs. Two strains of F. oxysporum had SKC1, but the SKC1 sequences from the 2 strains exhibited a relatively high level of divergence (Fig. 8).

Fig. 8.

Genealogy of Fusarium SKC1 homologs inferred by MP and ML analysis of nucleotide coding sequences. Statistical support for branches within the phylogenetic tree were generated by bootstrap analysis with 500 pseudoreplicates and are indicted by numbers near the internodes: # = MP, ML. A single number indicates that the bootstrap values were the same. Major clades are demarked by horizontal lines. F. foetens A–C indicate 3 SKC1 paralogs in one strain. F. oxysporum A and B refer to 2 different SKC1 homologs from 2 different strains, Fo4287 (NRRL 34936) and Foii5 (NRRL 54006). Location number refers to different genomic locations of the F. verticillioides homologs of the genes flanking the SKC1 homologs. The Fusarium species complex of each strain is indicated on the right and is highlighted by a unique background color in the tree (i.e. Fujikuroi is an abbreviation for F. fujikuroi specie complex). cd = could not determine because SKC1 was present on a small contig that did not include any flanking genes.

MP and ML analysis resolved the 23 SKC1 homologs into 4 clades (Fig. 8). The SKC1 MP and ML trees mirrored recently reported Fusarium species trees (Kim et al. 2020; Geiser et al. 2021) in some respects but had some noteworthy differences. First, unlike species in the species trees, some SKC1 sequences from the same species complex were more distantly related to one another than sequences from different complexes. For example, in the SKC1 tree, clade 1 included sequences from both the Fujikuroi and Newnesense complexes, while clade 2 included sequences from the Fujikuroi, Nisikadoi, and Oxysporum complexes. Likewise, the relationships of SKC1 sequences from species of the Fujikuroi complex did not always correlate with relationships of the species within the complex. This was apparent when the SKC1 trees were considered in the context of the 3 major clades (African, American, and Asian) into which members of the Fujikuroi complex have been resolved in previous phylogenetic analyses (O'Donnell et al. 1998; Kim et al. 2020). For example, in the SKC1 tree, sequences from the African clade species F. thapsinum grouped in Clade 2 and were relatively distantly related to sequences from other African clade species (i.e. Fusarium musae, Fusarium secorum, Fusarium udum, and F. verticillioides), which grouped in Clade 1 (Fig. 8). Furthermore, the 3 SKC1 homologs from F. foetens were relatively distantly related to one another; each was resolved in a different SKC1 clade (Fig. 8).

The relative genomic location of the SKC1 homologs in each Fusarium strain was assessed by examining their flanking genes (Supplementary Table 2). Here, we determined that SKC1 homologs occurred at 11 different locations (Fig. 8). The locations were not necessarily correlated with the 4 major SKC1 clades shown in Fig. 8 or with species complexes. For example, 4 of the 6 SKC1 homologs in Clade 2 occurred at location 7, whereas the 11 SKC1 homologs in Clade 1 occurred at any 1 of 6 locations. Further, SKC1 homologs from members of the Fujikuroi complex occurred at any 1 of 7 locations, and each of the 3 F. foetens paralogs occurred at a different location. SKC1B of F. foetens, a member of the Oxysporum complex, is most closely related to and occurs in the same location (9) as the 2 F. oxysporum SKC1 homologs, whereas F. foetens SKC1A and SKC1C were more distantly related to and at different locations (8 and 11, respectively) than the F. oxysporum SKC1 homologs.

It is noteworthy that of the 20 Fusarium species with SKC1, only the F. musae homolog was identical in sequence to F. verticillioides SKC1 (Supplementary Figs. 13 and 14). Even though SKC1 homologs from other species differed from the F. verticillioides homolog at numerous positions of both the nucleic acid and predicted amino acid sequences, all homologs were conserved with respect to the TAG stop codon (Supplementary Fig. 13). This raises the possibility that the SKC1 homologs in the other species undergo A-to-I RNA editing in the same manner as in F. verticillioides.

Discussion

Spore-killing meiotic drive elements, or gene drives, in species of the Sordariomycetes have been a topic of study for 4 decades. However, genes responsible for spore killing as well as resistance to killing, have been identified and characterized only within the last decade. Gene drives have the potential to revolutionize efforts to limit mycotoxin contamination and plant disease by providing farmers with a highly efficient, targeted approach to either kill or neutralize problematic organisms (Wedell et al. 2019; Gardiner et al. 2020). Like any new technology, effective use of gene drives to control agricultural problems will require a thorough understanding of the process. Some important questions about gene drives that require answers are, what genes confer gene drive/killing in fungi; how do natural gene drives function; do genes linked to gene drives impact competition among isolates in the environment or selection by host plants? In the current study, we identified SKC1 as a gene within the Sk locus that confers spore killing in F. verticillioides, a fungus that contaminates maize with the carcinogenic mycotoxins fumonisins worldwide. We also used multiple approaches to elucidate how SKC1 confers spore killing.

Our results demonstrated that deletion of SKC1 eliminated the Sk-mediated spore killing, allowing production of 8 ascospores per ascus during sexual development, while complementation of the deletion mutant restored spore killing. Transcriptional profiling of vegetative and sexual tissue indicated differential regulation of SKC1 and formation of 4 distinct SKC1 transcripts: 2 sense transcripts, of which one occurred exclusively in sexual tissue; and 2 antisense transcripts that occurred almost exclusively in vegetative tissue. The single nucleotide difference between the 2 sense transcripts occurred within the second position of predicted SKC1 stop codon (TAG) based on the genomic sequence of the gene. Transcript sense-SKC1a has the stop codon (UAG) sequences, whereas transcript sense-SKC1b has the codon UGG instead of the stop codon. As a result, sense-SKC1a is predicted to encode a 70-aa protein (Skc1a) and sense-SKC1b is predicted to encode an 84-aa protein (Skc1b), which consists of Skc1a with a 14-aa extension at the carboxy terminal end.

The single-nucleotide difference between sense-SKC1a and sense-SKC1b is consistent with A-to-I mRNA editing, a posttranscriptional modification process that involves deamination of adenosine (A) to form inosine (I) (Bian et al. 2019). Ribosomes recognize I as a guanine (G), which can lead to a change in codon specificity and, therefore, a change in amino acid sequence. In Fusarium graminearum and F. verticillioides, thousands of A-to-I editing events have been documented specifically during sexual development, and in F. graminearum some of the editing events change protein function (Liu et al. 2016; Bian et al. 2019; Liang et al. 2021). In N. crassa, transcripts for the gene rfk-1, which is required for the meiotic drive element Sk-2, also undergo A-to-I mRNA editing exclusively in sexual tissue (Rhoades et al. 2019). In the current study, identification of the putative A-to-I editing of SKC1 transcript raised the question, is spore killing induced by the edited SKC1 (sense-SKC1b) transcript or the larger Skc1b protein? Protein structure predictions by Rosetta, as implemented on the Robetta server (Kim et al. 2004), indicate that the C-terminus of Skc1b models includes an alpha-helix extending 4–5 turns that is not present in Skc1a models. Because neither Skc1a nor Skc1b share sequence similarity to proteins of known function, it is not clear how the carboxyterminal alpha-helix of Skc1b could cause spore killing.

In the initial transcriptomics experiment, vegetative tissue was meant to serve as a negative control for SKC1 expression, because we expected SKC1 expression to occur only during sexual development. Consistent with this expectation, we did not observe any phenotypic differences between in the skc1 mutant and the wild type during growth on/in solid or liquid media. Instead, we detected significant SKC1 expression during vegetative growth, which, included 2 antisense transcripts. One possible role for the antisense transcripts could be to protect vegetative tissue by interfering with or repressing sense SKC1 transcript function (Pelechano and Steinmetz 2013).

To shed light on the potential roles that the alternative forms of SKC1 transcripts and/or proteins play in spore killing, we expressed a modified version of SKC1 in F. verticillioides and 4 versions of SKC1 in the model fungus N. crassa. These experiments revealed in both fungi that SKC1 transcripts undergo A-to-I editing to form sense-SKC1b and that the sense-SKC1b transcript is required. In contrast to F. verticillioides, in N. crassa SKC1 does not act as a meiotic driver in that only 50% of surviving progeny inherit SKC1 and ascus developmental defects do not require a mixture of unedited and edited transcripts. In F. verticillioides, the presence of the Sk^K loci in almost 100% of surviving ascospores suggest that the loci also provide resistance to killing, similar to Sk-1 in N. crassa (Svedberg et al. 2021). The failure of SKC1 to drive in N. crassa suggests that resistance afforded by Sk^K in F. verticillioides is separate from SKC1 or is dependent on F. verticillioides-specific spatial or temporal cues.

Our survey of genome sequences detected SKC1 in species of 6 closely related Fusarium species complexes, indicating that Sk^K-mediated spore killing could be more widespread in the genus than reflected in the literature (Sidhu 1984; Raju 1994; Pyle et al. 2016). The presence of SKC1 was discontinuous within most species complexes in which it was detected. However, given that Sk^S strains of F. verticillioides lack SKC1 and that we examined only a single genome sequence of most species, the absence of SKC1 in some genome sequences could reflect the presence of 2 genotypes (Sk^K and Sk^S) within a species rather than absence of SCK1 in the species as a whole. Nevertheless, if SKC1 was more widely distributed in Fusarium than indicated by the results of our survey, we expected that it would have been detected in at least some genomes representative of other species complexes, particularly if the frequency of SKC1 in the species was similar to the frequencies observed in this study for strains of F. verticillioides (∼50%) and F. proliferatum (∼90%). Indeed, the GenBank database includes genome sequences for 19 strains of the species F. fujikuroi, but SKC1 was not detected in any of them.

As far as we are aware, the spore killing phenotype has been reported in 3 Fusarium species: F. verticillioides (Kathariou and Spieth 1982), F. subglutinans (Sidhu 1984), and F. proliferatum (Raju 1994). In our survey, we did not detect SKC1 in 8 genome sequences of F. subglutinans. It is possible that these 8 strains were Sk^S, which suggests that the frequency of Sk^K strains was lower than the 50% that we observed for F. verticillioides. Another possibility is that the spore killing system in F. subglutinans is distinct from the SKC1-mediated system in F. verticillioides. Yet another possibility is that the spore killing described by Sidhu (1984) occurred in F. temperatum, a species that was considered synonymous with F. subglutinans prior to 2011 (Scauflaire et al. 2011). However, we did not detect SKC1 in any of the 5 F. temperatum genome sequences included in our survey. In contrast to F. subglutinans/F. temperatum, we did detect SKC1 in F. proliferatum, which indicates that the previously reported spore killing in this fungus (Raju 1994) could be mediated by SKC1. It is note-worthy that the species F. siculi and F. hostae have SKC1 homologs that are likely to be nonfunctional because each contains premature stop codons (Supplementary Table 2). Translation of the shorter ORFs would result in proteins that are 22 and 29 amino acids shorter at the carboxy-terminus than Skc1a from F. verticillioides. Because all 3 premature stop codons are UAA, A-to-I editing would not be expected to lead to a longer ORF in transcripts because the codon formed from the editing would be UGA, which is also a stop codon.

The different genomic locations and relatively distant phylogenetic relationships of the 3 SKC1 paralogs in F. foetens indicate that they did not arise from a recent gene duplication event. The presence of at least 2 paralogs could be the result of horizontal gene transfer (HGT). While F. foetens is a member of the Oxysporum complex, SKC1A is most closely related to homologs from members of the Fujikuroi and Nisikadoi complexes. If SKC1A was horizontally transferred to F. foetens, a member of one of these 2 complexes would be a likely donor. Likewise, branch conflicts between the SKC1 tree (Fig. 8) and recently reported Fusarium species trees (Kim et al. 2020; Geiser et al. 2021) are consistent with HGT of SKC1 within and among species complexes. For example, in the SKC1 tree, sequences from 2 species of the Newnesense complex are nested within a clade of sequences from the Fujikuroi complex, which is consistent with HGT of SKC1 from a member of Fujikuroi to an ancestor to the 2 Newnesense species. In support of this, the SKC1 homologs in the 2 Newnesense species are located at the same relative genomic location (4) as the SKC1 homolog in F. secorum, a member of Fujikuroi. Alternative explanations for conflicts also include incomplete lineage sorting of ancestral SKC1 alleles or paralogs resulting from ancient duplication events (Proctor et al. 2013; Steenwyk et al. 2019). It is also possible that the multiple conflicts between the SKC1 and species trees have resulted from a combination of these evolutionary processes. Together, the results of our survey and phylogenetic analyses of SKC1 indicate a complex evolutionary history of the gene during divergence of at least 6 Fusarium species complexes. It remains to be determined whether such complexities have affected the ability of SKC1 to confer spore killing. Ongoing experiments to determine whether SKC1 homologs from other species can complement an F. verticillioides skc1 mutant could provide further insight into the evolutionary history and functionality of the gene.

Data availability

Strains and plasmids are available upon request. The RNA sequencing datasets and the mutant genome sequences reported in this study have been deposited in NCBI’s Sequence Read Archive (Leinonen et al. 2011) under BioProject accession number PRJNA768118 and the following: SRR16169787 (ΔSKC1::HYGB), SRR16169786 (ΔSKC1::GENR), SRR16169784 (Addback-A), SRR16169783 (Addback-B), SRR16169782 (Addback-C), SRR16169781 (Offspring-A3), SRR16169780 (Offspring-A5), SRR16169779 (Offspring-A6), SRR16169778 (Offspring-B25), SRR16169777 (Dataset: Mycelial), and SRR16169785 (Dataset: Perithecial). Strains Fv149 and Fv999 can be obtained from the Fungal Genetics Stock Center (McCluskey et al. 2010). Fusarium genome sequence data are available at GenBank and the accession numbers are listed in Supplementary Table 2.

Supplemental material is available at GENETICS online.