In Neurospora, a gene not paired with a homolog in prophase I of meiosis generates a signal that transiently silences all sequences homologous to it by a process called meiotic silencing by unpaired DNA (MSUD). Thus a deletion mutation in a heterozygous cross is formally “ascus-dominant” because its unpaired wild-type partner silences itself. We describe in detail the isolation of a mutation, Sad-1UV, that suppresses the dominance of various ascus-dominant mutations. Additional dominant, semidominant, and recessive Sad-1 alleles have been generated by RIP; the DNA of the dominant RIP alleles becomes methylated, but dim-2-dependent methylation is not necessary for silencing. The barrenness of homozygous Sad-1 crosses is not due to the failure to silence unpaired mating-type mat A-2 mat A-3 genes. Transcripts of sad-1+ can be detected during the sexual phase in a homozygous wild-type cross, indicating that the gene is expressed even if all DNA can pair normally. Meiotic silencing is confined to the ascus in which DNA is unpaired, and silencing does not spread to neighboring asci in a fruiting body of mixed genetic constitution.
NEUROSPORA crassa is usually thought of as a “haploid,” heterothallic, ascomycetous fungus because its two mating types, A and a, remain haploid and propagate themselves separately during the mycelial, macroconidial, or microconidial phases of growth. An optional but important sexual cycle becomes possible when an inoculum of either mating type experiences nitrogen limitation. A precursor of the female tissue of the fruiting body, the protoperithecium, begins to form, and this extends receptive, specialized hyphae called trichogynes into the surrounding medium. If a trichogyne encounters a vegetative cell of the opposite mating type, cytoplasmic fusion occurs, and the nucleus from the “male” element enters the trichogyne, but fusion with resident female-derived nuclei is delayed for several days. Instead, the haploid nuclei in the fertilized fruiting body, now called a perithecium, propagate within specialized structures, the ascogenous hyphae. Nuclei of opposite mating type pair up in a developing tubular structure, the ascus, and fuse (karyogamy) to give a transient diploid cell, the zygote. This zygote quickly embarks on the first of two typical meiotic divisions, resulting in four haploid nuclei. Each of these divides once mitotically, and the resulting eight nuclei, at first still in a common cytoplasm, soon become individualized as separate spores.
In both the ascogenous hyphae of the sexual phase and the mycelia of the vegetative phase of the life cycle, mechanisms exist that, in effect, scan the genome for dual or multiple copies of DNA. Any sequence of ∼1kb or larger that is present in more than a single copy is likely to be a retrotransposon capable of damaging the integrity of the genome. Hence it is not surprising that in Neurospora, as in other eukaryotic organisms, mechanisms have been selected that can destructively mutate these DNA sequences or that can destroy the mRNA transcribed from them. In ascogenous hyphae, greater-than-haploid sequences are subjected to a premeiotic process called repeat-induced point mutation (RIP), which converts many C:G base pairs to A:T pairs. This generally destroys the coding potential of both copies of the duplicated sequences (Selker 1990). The enzymatic steps in this process are only beginning to be understood. A very different silencing process called “quelling” deals with the expression of hyperhaploid DNA sequences during vegetative growth (Cogoni and Macino 2000). Analogous processes of post-transcriptional gene silencing (PTGS) are called “RNA interference” in animals and “cosuppression” in higher plants. Single-stranded RNA, encoded by hyperhaploid sequences (e.g., repeats of a transgene) in Neurospora and transcribed by an unidentified polymerase, is converted to double-stranded RNA (dsRNA) by an RNA-directed RNA polymerase (RdRP) encoded by the qde-1 locus. The dsRNA, in turn, is enzymatically cleaved into “small interference RNA” molecules (siRNA), and, as judged from parallel systems in other organisms, these must act as guide molecules for the destruction of any mRNA transcribed from the original hyperhaploid gene and/ or prime the synthesis of more dsRNA molecules (Zamoreet al. 2000; Lipardiet al. 2001). One of the remarkable features of PTGS in several systems is its ability to propagate itself, not only within the confines of a cell compartment, but also between cells. Palauqui et al. (1997) and Voinnet and Baulcombe (1997) showed that a signal to silence gene expression, presumably an RNA intermediate, could pass between stock and scion in grafted plants, and Fire et al. (1998) found that dsRNA could cause systemic silencing in Caenorhabditis elegans without having been injected intracellularly.
Recently we reported that Neurospora has a third silencing system, one that works during the first meiotic prophase of the brief diploid phase of the life cycle (Shiuet al. 2001). It works to silence the expression of all copies of any gene not paired with its homolog during meiotic homolog pairing, even those copies that are themselves properly paired. We refer to this as meiotic silencing by unpaired DNA (MSUD). A corollary of this is that any gene present in one or three or any uneven number of total copies in the aggregate genomes of the two parents, whether by deletion or duplication, will be silenced; the same holds for any gene present in an even number of copies, but present at different positions in their genomes so that the pairing of their neighbors prohibits their own pairing. If the DNA in question encodes a protein that is required for the completion of meiosis or ascospore development, MSUD will prevent the completion of the sexual cycle, which will be arrested at a stage characteristic of the particular gene that cannot pair. Formally, then, these deletion or insertion mutants are “ascus-dominant,” that is, none of the nuclei of the tetrad will give rise to viable ascospores, including those that, themselves, have a wild-type genome. The deletion mutants, especially, challenge our intuition by being dominant through being absent (Aramayo and Metzenberg 1996; Shiuet al. 2001). The contradiction disappears when we consider that it is really the wild-type allele that is dominant, but is exerting its dominance by making an abnormal product as a result of its failure to find a pairing partner.
In this article, we describe the isolation and characterization of a suppressor of MSUD, called Sad-1 (suppressor of ascus dominance). Sad-1 is itself semidominant, and the basis for this behavior is explored. The wild-type copy of this gene, sad-1+, encodes a putative RdRP, a paralog of the qde-1+ gene involved in quelling, and a homolog of many other genes identified with somatic silencing in animals and plants. We also discuss the complete barrenness of homozygous Sad-1 × Sad-1 crosses and the possibility that the only known heterologous genes in the two mating types of Neurospora, those at the mating-type locus, produce toxic products if they fail to be silenced. We also report the effect of Sad-1 on the ascus dominance of the “Near-Round spore” mutation. Finally, we examine the possibility that MSUD from a silenced locus, the wild-type allele of Round spore, can trigger systemic silencing in a perithecium analogous to that seen in the silencing in a scion by its stock and the propagation of the RNAi signal in nematodes.
MATERIALS AND METHODS
Construction and manipulation of Neurospora strains: The Neurospora strains used in this study are listed in Table 1. For production of mycelia or conidia, Neurospora was grown on Vogel's minimal medium N, with 1.5% sucrose as the carbon source, and routine genetic techniques were used (Davis and DeSerres 1970). Amino acids, purines, and pyrimidine nucleosides, when appropriate, were used at 1 mm; inositol was used at 50 μg/ml, and Ca pantothenate at 2 μg/ml for vegetative growth and 20 μg/ml in crosses; all other vitamins were used at 2 μg/ml. Crosses were done in petri dishes with the salts mixture of Westergaard and Mitchell (1947) with the pH unadjusted (i.e., ∼4.8) and 1.5% sucrose, solidified with 1.5% agar. Usually crosses were made by simultaneous inoculation of the two parents at opposite sides of the plate. Scoring for mating type was done by spotting conidial suspensions onto fluffy tester lawns of the appropriate genotype. Where the spore prints of progeny were to be scored for ejection of round or spindle-shaped spores, the spotted lawns were covered with a thin Plexiglas sheet separated from the lawns by 1-mm shims cut from plastic toothpicks. Transformation by plasmids of ∼10 kbp or less, including plasmids targeted to the his-3 locus, was by electroporation of washed conidia essentially as described by Margolin et al. (1997) but at 1500 V in a 1-mm gap cell (15,000 V/cm). The his-3 mutant allele was 1-234-723 in all cases. Transformation by cosmids employed spheroplasts prepared from germinated conidia essentially as described by Vollmer and Yanofsky (1986).
Mutagenesis and selection of Sad-1UV: Lawns of strain 91-33 (with two copies of asm-1, one wild type and one frameshifted) were prepared on inositol-crossing medium and allowed to grow at room temperature for 6 days, at which time they had formed abundant protoperithecia. An aqueous conidial suspension of the Asm-1Δ (deletion mutant, strain 80-23) was UV-irradiated to near the borderline of detectable killing (86% survival) and ∼5 × 106 conidia were plated to each of 20 lawns. The crop of spores from each lid, almost all of them white and inviable, was collected, heat-shocked, and plated to 20 plates of Vogel minimal plus inositol with the usual sugar mixture (Brockman and Deserres 1963). A total of 166 colonies were obtained from the mature spores from the original 20 lawns. The cultures from these were patched onto ordinary fluffy lawns (81-01 and 81-02) to score for mating type and identify potentially fertile segregants and were also patched to Round spore-carrying cultures (92-36 and 92-37—see below). Of the 166 isolates, one of them formed abundant spindle-shaped spores on the lawn of 92-36 as well as round ones and was the source of the Sad-1UV allele. This isolate was outcrossed and segregants free of Asm-1Δ and inl mutations were then used to construct various strains described in Table 1.
Lawns of Round spore for scoring Sad-1 mutant(s): The lawns carried the fluffy mutation, which eliminates the formation of conidia and is more female fertile than wild type. Round spore is female sterile, but this sterility is recessive. Therefore, to allow Round spore to be used as a female, we prepared heterokaryons in which the inability of Round spore to make protoperithecia was complemented by a categorically sterile helper lacking the mat locus, (round spore+); fluffy; nic-3; matΔ, which nevertheless can produce protoperithecia. To prepare this helper strain, we crossed 89-27, which has the mat A idiomorph deleted from its canonical locus and relocated to the am locus, to strain 61-18. We tested am+ fluffy (fl) nic-3 segregants for mating type. As expected, all of them were either mat a or mat null. One of the latter was chosen and kept as 92-40. The female-sterile fluffy Round spore strains 92-38 and 92-39 were made by conventional crosses. The lawns of nutritionally forced, female-fertile heterokaryons, 92-36 and 92-37, were prepared by coinoculating 92-40 onto minimal medium with either 92-38 or 92-39.
Nucleic acid methods, DNA hybridization, and reverse transcriptase PCR: Standard procedures of molecular biology were used throughout (Sambrooket al. 1989). All DNA sequencing was performed by Biotech Core (Mountain View, CA). Gel electrophoresis and DNA transfer to nylon membrane (Schleicher & Schuell, Keene, NH) were performed according to the manufacturer's instructions. Chemiluminescent detection of immobilized DNA was performed using the DIG High Prime DNA labeling and detection kit (Roche, Indianapolis). For reverse transcriptase (RT)-PCR, poly(A)+-enriched RNA was prepared with the Oligotex mRNA kit (QIAGEN, Chatsworth, CA) and was later used for cDNA synthesis using the First-strand cDNA synthesis kit (Amersham Pharmacia, Piscataway, NJ). PCR of the sad-1+ cDNA was performed using primers that span the intron: SAD4489F (5′ ATGCTCATCTGGCGA CAGCAGATG 3′; positions 4489–4512) and SAD4796R (5′ TTCTCTTTCATGTCGAATGTTTCCC 3′; positions 4796–4772). The PCR product of the cDNA is 250 bp in length, as compared to the 308 bp from that of genomic DNA. The sad-1+ cDNA PCR product was cloned into the PCRII vector using the TOPO-TA cloning kit (Invitrogen, Carlsbad, CA) and subjected to DNA sequencing.
Selection/screening of the suppressor of ascus dominance, Sad-1UV: The low viability of ascospores from crosses between one parent carrying two copies of asm-1 and a second parent with zero or one copy (Shiuet al. 2001) suggested a way of enriching for dominant suppressor mutations that would alleviate the silencing of asm-1+. The viability of the progeny was further reduced if the ectopic copy in the two-copy parent was a frameshift mutant, asm-1 fs. However, even this residual ascospore viability was still large enough to yield many false suppressor candidates. Aramayo and Metzenberg (1996) proposed that several long-known ascus-dominant morphological mutants, Round spore, Banana, and the dominant allele of peak, owed their ascus dominance to the same process that operates with the asm-1 system. Of these mutants, Round spore is particularly tractable, in that it gives almost 100% spherical or near-spherical spores in heterozygous or homozygous crosses, while homozygous wild-type crosses give 100% spores resembling an American football (“spindle shaped”). If one assumes that asm-1 and Round spore are manifestations of the same mechanism, it would make sense to select for production of viable spores in a cross of an Asm-1 deletion strain to a UV-mutagenized strain carrying two copies of asm-1, one of them a wild-type copy and the other having a frameshift mutation; the collection of surviving candidates is then screened by crossing them individually to Round spore. Authentic suppressors were predicted to give spindle-shaped progeny spores. One such suppressed mutant, Sad-1UV, has been studied and is described below. On the order of one-third or more of the ascospores in any fruiting body from a typical heterozygous cross are spindle shaped. In each ascus that does contain any spindle-shaped spores, generally all eight of them will be spindle shaped, including of course the four that are genetically Round spore. That is, within an ascus, all the spores are concordant in phenotype. The fraction of spindle-shaped spores is not markedly different whether Sad-1UV is the female or male parent. It must be emphasized that Sad-1UV is in no sense a suppressor of the vegetative-phase or sexual-phase phenotype of Round spore itself, or of any other mutant allele known to us. A cross between a Round spore strain (84-39) and a Round spore; Sad-1UV strain (92-07) gives virtually 100% round spores. What Sad-1UV does is to make information from the wild-type allele of Round spore in a heterozygous Round spore cross available to all the progeny in an ascus.
Scoring and mapping of Sad-1UV: Initially, we scored conidial suspensions for Sad-1UV by the ability of the mutant allele to give ∼25–75% spindle-shaped spores in crosses to female Round spore lawns, the proportion of these spores varying considerably between early and late-ejected ascospores. When the wild-type allele sad-1+ was used, spindle-shaped ascospores were absent or extremely rare. Later, when Sad-1UV, which is linked to mating type, became available in both mating types, we found that Sad-1UV × Sad-1UV crosses were completely barren; abundant perithecia of outwardly normal morphology were formed, but they produced no ascospores. Once it became clear that this was a property of the mutation itself, it was usually more convenient to use Sad-1UV lawns for scoring (strains 93-13 and 94-02). The mutation was mapped by conventional genetic crosses to markers on the left arm of linkage group I, between un-5 and leu-3 (data not given; see Perkinset al. 2001).
Cloning and characterization of sad-1+: The complete barrenness of homozygous Sad-1UV crosses (Shiuet al. 2001) suggested a strategy for cloning the wild-type allele: restoration of ascospore production by complementation with the appropriate clone from a cosmid library. Accordingly, Sad-1UV a (93-12) conidia were subjected to transformation with pooled cosmids from each of five 96-well pools of a linkage group I-specific library (Fungal Genetics Stock Center, plates M11–M16), selecting for the hygromycin resistance cassette present in the original cosmid vector. The transformants from each pool were allowed to conidiate en masse, and the five pools of conidia were used to fertilize duplicate lawns of Sad-1UV fl A (93-13). Two of these pools gave rise to plates with a small crop of ascospores. The clones in each of the 96-well microtiter plates responsible for the fertile transformants were identified by sib selection as M11:D10 and M13:G8. These correspond to G4:C6 and X4:G3 in the library of Orbach and Sachs (1991). DNA sequence analysis and restriction fragment analysis demonstrated extensive overlap of these two cosmids. Accordingly, only one of them, M11:D10, was studied further. We compared the end sequences of this cosmid with sequences in the Neurospora DNA database (http://www-genome.wi.mit.edu/annotation/fungi/neurospora/). Transformation with restriction fragments of the cosmid gave transformants only with one specifying a presumptive RdRP (Shiuet al. 2001), namely the 6847-bp XbaI-ClaI fragment. Sequencing of this region including the reading frame of the gene of interest and its immediate upstream and downstream regions gave results completely in accord with those in the Whitehead Institute database of the entire Neurospora genome. The sequence showed an open reading frame (ORF) of 4975 bp interrupted by an intron of 58 bp (Figures 1 and 2). The 5′ and 3′ and internal sites of the sad-1+ intron fit the intron splicing consensus sequences (Bruchezet al. 1993). DNA sequences surrounding the sad-1+ ATG start codon agree with the consensus for N. crassa translational initiation (Edelman and Staben 1994). A sequence matching the transcription initiation consensus for N. crassa was found at base pairs 1204–1211 (TCATCANC; Bruchezet al. 1993). A CAAT motif (Bucher 1990) and a TATA box were identified at base pairs 997–1001 and 1119–1125, respectively (Bruchezet al. 1993).
Translation of the sad-1+ ORF predicted a 1638-amino-acid (aa) polypeptide with an Mr of 184,221 (Figure 1). SAD-1 has a predicted isoelectric point of 6.25 and is made up of 43% nonpolar and 28% charged residues. The amino-terminal portion of SAD-1 (aa 19–82) is basic (isoelectric point = 12.68) and is composed of 73% charged or polar residues and 23% proline. The carboxyl-terminal portion of SAD-1 is glycine rich (aa 1569–1690, 77% glycine). Glycine-rich domains are found in a large number of proteins such as extracellular or cell envelope proteins, RNA-binding proteins, and keratin (Steinertet al. 1990). The carboxyl half of the gene has been deleted in the Sad-1UV allele.
An alignment of SAD-1 with RNA-dependent RNA polymerases implicated in various silencing systems such as SDE1 of Arabidopsis thaliana, EGO-1 of C. elegans, and QDE-1 of N. crassa is shown in Figure 3. sad-1+ is also homologous to a putative RdRP (rdp-1) identified in Diaporthe ambigua, host to a mycovirus that confers hypovirulence (O. Preisig and M. J. Wingfield, personal communication; Preisiget al. 2000); it seems possible that this apparent RdRP plays a role in silencing viral genes. A third RdRP of unknown function is also present in the genome of Neurospora (http://www-genome.wi.mit.edu/annotation/fungi/neurospora/).
The sad-1+ sequence seems to be conserved in Neurospora species other than the laboratory crassa strain. A 5.8-kb KpnI-KpnI fragment containing the sad-1+ ORF was used as a probe in Southern analysis, using genomic DNA from various Neurospora species. Under high stringency of hybridization (68°), sequences homologous to sad-1+ DNA can be detected in heterothallic species (N. crassa, N. intermedia, N. discreta, N. sitophila), a pseudohomothallic species (N. tetrasperma), and a homothallic species (N. terricola; data not shown).
Evidence that DNA of a cosmid that complements Sad-1UV maps to the Sad-1 locus: While it seemed likely that the complementing DNA included the wild-type gene corresponding to the mutated one, it was conceivable that the introduced sequences encoded an unlinked suppressor of the mutant gene. To test this possibility, Sad-1UV a (93-12), which had been isolated on the Oak Ridge genetic background, was crossed to Mauriceville-1c A (21-11), a wild-collected strain known to have many polymorphic differences from laboratory strains (Metzenberget al. 1985). Cultures from random ascospores were classified as Sad-1UV or sad-1+, and DNA was prepared from 10 of each. Samples of these and the two parental strains were digested with BamHI, displayed by electrophoresis, blotted to a membrane filter, and probed with the cosmid M13:G8. Differences between the parental strains were obvious in the mobility of at least four fragments, and these segregated in the expected way with the Sad-1UV or sad-1+ phenotype in all 20 progeny (data not shown), ruling out the possibility that an unlinked suppressor of the homozygous barrenness of Sad-1UV had been isolated.
Functional analysis of sad-1 constructs: To determine what portions of SAD-1 are important for function, we obtained and tested several deletion constructs. All or nearly all of the open reading frame of the sad-1+ sequence is essential in conferring fertility in crosses to the Sad-1UV mutant. As shown in Figure 2, a number of 5′-deletion constructs, 3′ deletions, and an in-frame interstitial deletion rendered sad-1 completely inactive.
Detection of sad-1+ mRNA transcript by RT-PCR: It seemed possible that the expression of sad-1+ would be dependent upon the presence of unpaired DNA. To address this question, we attempted to detect sad-1+ mRNA transcripts by RT-PCR. Primers spanning the intron were used to amplify Sad-1 sequences from wild-type vegetative mRNA and from perithecial mRNA isolated from two crosses: wild type (81-02) × wild type (40-27) and wild type (81-02) ×::Bml [β-tubulin duplication strain (95-24)]. The 250-bp cDNA product of sad-1+ mRNA (the identity of which was confirmed by sequencing) could be detected in perithecia of both crosses (data not shown), indicating that the presence of unpaired DNA is not a prerequisite for the expression of sad-1+. No sad-1+ mRNA was detected in extracts of vegetative cells, suggesting that its expression may be specific to the sexual phase.
Isolation of dominant, semidominant, and recessive sad-1 mutants by RIP: We wanted to isolate a representative group of mutants that had suffered RIP in the endogenous sad-1 locus. At the same time, we wanted to avoid the complications of having the ectopic sad-1 copy that induced the RIP process continuing to be present in the final strains. Accordingly, we inserted a 6.85-kb sad-1 XbaI-ClaI fragment into the pBM61 vector and transformed mep his-3; pan-2 A (strain 77-29) to histidine independence. The resulting strain, 96-13, which had the gene order sad-1+ leu-3+ mat A his-3::his-3+ sad-1+ on linkage group I, was then crossed to sad-1+ leu-3 mat a his-3 (strain 95-14) by simultaneous inoculation. Because the proportion of progeny mutated by RIP is higher in ascospores discharged late than in those ejected earlier (Singeret al. 1995), we replaced the lid of the crossing plate on the twenty-third day and used only spores recovered thereafter. We isolated 28 progeny with crossovers between leu-3 and his-3, i.e., sad-1RIP CANDIDATE leu-3+ his-3. All of these were fertile in crosses to ordinary tester lawns of the opposite mating type, but 26 of the 28 were barren in crosses to Sad-1UV tester lawns. These 26 strains were candidate sad-1RIP mutants. The 26 RIP candidate strains were also crossed to lawns for testing suppression of Round spore strains 92-36 and 92-37, and the spore prints were examined for spindle-shaped (silencing-suppressed) spores. Such spores ranged from 0% spindle shaped (five strains) to 100% (two strains), with the rest giving intermediate levels (Table 2). It is well known that DNA sequences that have been altered by RIP very often become methylated on many of their cytosine residues (Selker 1990; Miaoet al. 1994) and it was of interest to examine the degree of methylation of our RIP alleles. We digested aliquots of DNA from each allele with the restriction enzyme HpaII (sensitive to methylation of its recognition tetranucleotide), displayed the digests by electrophoresis, and prepared a membrane for probing (Figure 4). Many of the sad-1-RIP mutants showed a variety of bands of reduced mobility due to the failure to cut at sites in which cytosine had become methylated. Furthermore, there was a strong rank-order correlation between the degree of methylation, as judged from the Southern blot, and the degree of dominance of each allele, as measured by the percentage of spindle-shaped spores recorded in Table 2.
Cytosine methylation and MSUD: Two models suggested themselves as explanations for the correlation between methylation and dominance: (1) The dominance of the particular sad-1RIP allele (i.e., the silencing of the sad-1+ allele), and hence the failure to silence the wild-type alleles of other unpaired genes, requires that the sad-1RIP allele be heavily methylated because methylation signals the machinery that a gene is unpaired and therefore should be silenced, and (2) the dominance of the Sad-1RIP allele and its degree of methylation are not causally related, but both depend on the number of RIP mutations in that allele.
A mutant, dim-2, that lacks all vegetative cytosine methylation and possibly all methylation in any stage of the life cycle (Selkeret al. 1993; Kouzminova and Selker 2001) allowed us to test these alternatives. We prepared and confirmed the phenotype of the double mutant dim-2; Round spore and crossed it to dim-2 of the opposite mating type. If model 1 is correct, the absence of methylation should result in the absence of any silencing, and all ascospores should be spindle shaped. If model 2 is correct, all spores should be round. All the spores from such a cross were, in fact, round. It is clear that cytosine methylation, at least that governed by dim-2, is not essential for the operation of MSUD. In harmony with this model, a dominant Sad-1 mutation (Sad-1RIP141 dim-2) remains dominant even if it is non-methylated when it enters the cross.
Is the barrenness of homozygous Sad-1 crosses due to their failure to silence mat A-2 and/or mat A-3? One hypothesis for the barrenness of the homozygous Sad-1 crosses is that some gene that is present in the genome of one parent but absent from the other (and therefore unpaired in the prophase of meiosis 1) inhibits progress beyond the stage of meiotic pairing of homologs. It would therefore need to be silenced for meiosis to continue, and this would require at least some function of a sad-1+ allele. The only essential difference that has been detected between the two mating types is at the mat locus itself, where the DNA sequences are completely dissimilar and incapable of any pairing that requires Watson-Crick base pair recognition. Indeed, a mat A strain can be converted to a fully fertile mat a by evicting the mat A idiomorph with mat a (Chang and Staben 1994). Thus one or both mat regions could contain a gene or genes that must be silenced to allow the completion of meiosis. These regions contain four genes, mat a-1, mat A-1, mat A-2, and mat A-3 (Ferreiraet al. 1996). The first two of these are indispensable for recognition and mating, and disruption of either of them abolishes fertilization. However, mat A-2 and mat A-3 can be disrupted individually by RIP without major detriment to either fertilization or eventual ascospore production. Disruption of both of them drastically reduces the production of ascospores, but does not totally eliminate it (Ferreiraet al. 1998). Accordingly, we crossed Sad-1Δ mat a to the RIP mutants mat A-2m1, to mat A-3m1, and to mat A-2m3 A-3m3 and isolated strains of genotypes Sad-1Δ mat A-2m, Sad-1Δ mat A-3m1, and Sad-1Δ mat A-2m3 A-3m3 (28-16, 28-17, and 28-18, respectively). Each of these was then crossed to Sad-1Δ mat a. Like homozygous Sad-1 crosses involving a normal mat A, they produced abundant perithecia but no ascospores. In the case of the cross involving Sad-1Δ mat A-2m3 A-3m3, we were concerned that the very low fertility of crosses of sad-1+ mat A-2m3 A-3m3 to Sad-1Δ mat a or even to wild-type (sad-1+) mat a could obscure a real ability of mat A-2m3 A-3m3 to suppress the barrenness of homozygous Sad-1Δ crosses. Therefore, quadruplicate lawns of Sad-1Δ mat a and quadruplicate control lawns of sad-1+ mat a were fertilized with Sad-1Δ mat A-2m3 A-3m3. All four of the control lawns gave rise to at least a small crop of ascospores, whereas none of the Sad-1Δ mat a lawns did so. Thus a failure to silence mat A-2, mat A-3, or both of them does not explain the barrenness of homozygous Sad-1 crosses.
Can Sad-1 mutations suppress the ascus dominance of Near-Round spore? Strains resembling the canonical Round spore mutant can arise in several different ways. Jacobson (1992) generated strains carrying duplications of a segment of linkage group I mapping very close to Round spore but not including it, Dp(I II)MD2. These strains were barren because they were→duplications and also grew very poorly in vegetative culture because they were heterozygous for an incompatibility gene, het-5. Such diploids eventually “escape” vegetative inhibition, often by losing part or all of the duplicated segment and becoming haploid/hemizygous for het-5; if they lose the duplicated segment in its entirety, they may also become fertile. Many such “escapees” produce mostly round spores in crosses to wild type (Jacobson 1992), though never as close to 100% as in the case of the canonical Round spore mutant. Presumably that gene is structurally intact, but its expression is altered. We refer to the escapees of interest as Near-Round spore. The original isolates had been lost, so we generated a number of independent Near-Round spore alleles by the method of Jacobson, subsequently isolating them in both mating types. Four of these are examined in Table 3. In crosses to wild type, they gave predominantly round spores (range: 85.2–98.4%), whereas in crosses to Sad-1UV, they gave predominantly normal spores, a lower proportion being round (4.6–42.0%). Our working model suggests that a homozygous cross of Near-Round spore strains should give normal spores. Jacobson (1992) found that homozygous crosses were, unfortunately, sterile. This was also the case with all but one of our alleles, but that one, allele 31, produced and ejected a modest number of spores. Only 18.8% of the spores were round, as compared with 85.2 and 95.2% of the spores in heterozygous crosses of the two parents to wild type. By comparison, heterozygous and homozygous crosses of the canonical Round spore showed no difference (99.8 and 99.2% round spores, respectively, consistent with the notion that the canonical Round spore, unlike Near-Round spore allele 31, does not contain the genetic information to produce spindle-shaped ascospores and produces none even when it is totally paired (Table 3).
Can MSUD be propagated from one ascus to another in a perithecium? There is abundant evidence in plants (Palauquiet al. 1997; Voinnet and Baulcombe 1997) and in C. elegans (Fireet al. 1998) that the agent(s) of silencing in one tissue can cross plasma membranes and spread to other tissues in the same organism, apparently via an RNA intermediate. We wished to test whether an analogous process occurs with MSUD in Neurospora. To do this, we took advantage of the fact that individual protoperithecia (female components) formed by a genetically marked heterokaryon of a single mating type, when fertilized by homokaryotic conidia of the opposite mating type (male components), can give rise to two genetically different kinds of asci. Johnson (1977) found such mosaic perithecia only on rare occasions, but in the experiment reported below, almost one-third of the perithecia had two female parents. Our experiment is a conceptual repeat of his experiment, but with much larger numbers of perithecia and of spores to allow detection of uncommon events, if they occur. We employed an auxotrophic marker, inl (inositol requiring), that allowed all germinated spores to be classified by inspection, including random spores as well as intact asci. The protoperithecial parent was a forced fluffy (non-conidiating) heterokaryon between two Oak Ridge-compatible strains, fluffy; Round spore; inl a (92-39) and fluffy; his-3; pyr-1 met-2 a (32-20), prepared by coinoculating them onto minimal crossing medium. After protoperithecia had formed, they were fertilized with conidia of inl A (05-34). Ascospores were heat shocked to induce germination and plated to medium lacking inositol. Inositol-requiring germlings push out only a tiny bud, whereas inositol-independent germlings grow out abundantly in ∼16 hr. Since these can arise only from a pairing of the male inl nucleus with a female of the genotype fluffy; his-3; pyr-1 met-2, all of them will have the wild-type allele at the round spore locus. However, if the cross of the inl nucleus to fluffy; Round spore; inl, in which the wild-type allele round spore+ is silenced, can transmit its silencing signal to other inhabitants of the same perithecium, we should observe inl+ segregants that are phenotypically round. Among 56 perithecia that were individually examined, 19 gave only round spores, 18 contained only spindle-shaped spores, and 19 were the desired type containing very roughly equal numbers of ascospores of both shapes. The spores from each of these perithecia were germinated on plates lacking inositol and the inl+ germlings were examined. Of 380 such germlings, all were spindle shaped. In addition, random ascospores ejected to the lid of the plate were germinated on the same medium and 1000 inl+ germlings were examined; they all were spindle shaped. These must have been ejected from many hundreds of perithecia, about one-third of them from mosaic perithecia. Thus ∼300 or more informative germlings from random spores derived from hundreds of mosaic perithecia were sampled. Clearly the silencing due to the presence of Round spore in the asci of mosaic perithecia does not spread to asci not containing this allele. Thus the Round spore allele is dominant within an ascus, but autonomous within a perithecium, as Johnson (1977) concluded from a different experimental design and smaller data set. The same seems likely to be true of other cases of MSUD.
We have described the isolation, cloning, and characterization of the sad-1+ gene and shown here (and previously; Shiuet al. 2001) that alleles extensively altered by RIP (e.g., Sad-1RIP141), or by partial deletion, RIP, and rearrangement of the gene (Sad-1UV), or by complete deletion of the gene (Sad-1Δ) are all formally dominant over the wild-type allele. How can something that is absent dominate something that is present? A normally serviceable word like “dominant” can lead us astray. It is really the unpaired wild-type allele that dominates the ascus phenotype; it does so by making an abnormal product that silences itself and its homologs, if there are any. Probably all deletion and insertion mutants in Neurospora are dominant in the sense that the unpaired wild-type allele prevents expression of the information it contains. They are usually not seen to be dominant because the product of the normal gene is not needed in meiosis or because it can be furnished by surrounding tissue.
The suppression by Sad-1 alleles of Near-Round spore mutations gives additional information that cannot be deduced from the classical mutants such as Round spore itself. Near-Round spore mutations themselves are explainable as deletions that ablate the nearby het-5 heterokaryon-incompatibility allele (Jacobson 1992); however, they cannot be explained easily by deletion of the canonical Round spore gene itself, because the penetrance of the canonical mutation is higher than that of the putative deletion mutants and because at least one of the Near-Round spore mutants, allele 31, gives a higher proportion of spindle-shaped spores in a homozygous cross than in a heterozygous one. A possible explanation is that the canonical Round spore mutation may contain a deletion of the entire ORF of the Round spore gene, whereas the Near-Round spore mutation may contain a deletion that is at or near the promoter region (or 3′ untranslated region) of the Round spore gene. Cloning of the Round spore gene and the characterization of the mutants will provide insights on the effect of unpairing on a nearby ORF.
It seems clear that MSUD requires synthesis of a double-stranded RNA by RdRP, but the enzymatic process by which the first strand is synthesized is completely unknown. It could require transcription by conventional RNA polymerase II, by a modified form of it, or by a presently unrecognized polymerase. At first glance it might seem that one could begin to answer this by making nested deletions that variously spared or destroyed the Pol II promoter of some gene essential for completion of meiosis. One could then ask whether silencing requires that the promoter be intact.
It is not clear why RdRP is necessary for completion of meiotic prophase in otherwise wild-type crosses. In C. elegans, mutations in ego-1 (a sad-1 homolog) affect both germ-line development and RNAi; ego-1 mutants show complex alterations in the switch from spermatogenesis to oogenesis and in early meiotic prophase; in addition, the oocytes contain some unpaired homologous chromosomes (Smardonet al. 2000). Quite possibly RdRP has a required function in meiosis that is independent of silencing. On the other hand, it is possible that certain sexual development genes need to be downregulated by the meiotic silencing system.
A priori, it seems likely that meiotic silencing shares some genetic elements with quelling and other RNAi processes, though the mechanism by which the cell detects abnormal DNA sequences is almost certainly different. Isolation of nonsilencing diplophase mutants in other genes would be informative. Initially, we expected that this would require isolating null mutants, with all the difficulties attendant upon finding recessive mutants that act only in diplophase. Because silencing is reflexive, in that silencing a silencer causes loss of silencing, it may, however, be feasible to isolate semidominant mutants, either as partial duplications or deletions analogous to Sad-1UV or by random insertional mutagenesis with a plasmid large enough to disrupt local pairing.
We have shown that dim-2-mediated cytosine methylation is not required for MSUD. Of course, we cannot rule out the possible existence of an unidentified methyltransferase that is active only during the sexual phase. In plants, data have suggested that methylation of the transgene coding region can be mediated by an aberrant RNA species and that methylation can maintain and amplify the silencing signal (for review, see Bender 2001). It was proposed that plants use DNA methylation as a defense against viral invasion of the genome. Invading sequences marked by methylation would be continuously combatted by the silencing machinery even in the absence of the virus. Moreover, a preexisting silencing signal would protect the plant against invasion by a similar virus. Our data suggest that such a maintenance effect of RNA-directed methylation is absent in Neurospora, as it also is in the nematode (which lacks methylation) and in mammals (which lack efficient propagation of asymmetric, non-CG methylation, the kind associated with transgene silencing).
Although systemic, somatic propagation of a silencing signal is widely observed in plants and animals, it appears to be absent in MSUD. Although a diffusible signal is likely to be present and is freely distributed within an individual ascus, it does not seem to be able to spread to the other asci in the same perithecium (this study) or to the surrounding vegetative cells (our unpublished data). Although there is no clear indication of what the diffusible factor in plants and animals might be, it has been suggested that it might be an aberrant RNA species that has the ability to maintain sequence specificity. The diffusible factor is unlikely to be siRNA, since elimination of the ability to make this category of molecules does not eliminate the systemic spread of silencing (Malloryet al. 2001).
In a wild-type cross, sad-1+ must normally be expressed at or very soon after the onset of meiotic homolog pairing, because in a heterozygous Sad-1Δ cross, the failure of sad-1+ to pair, and therefore the failure to be fully expressed, is sufficient to keep other unpaired genes from being silenced. Therefore sad-1+ must be put to the test at least as early as other genes, perhaps a bit earlier. Clearly, a full understanding of the timing of expression and of silencing of these genes will not come solely from genetic analysis and will await the development of sophisticated cytochemical tools.
We are grateful to David Perkins, Charley Yanofsky, and Virginia Walbot for their hospitality and steady encouragement; to Louise Glass for strains carrying mutations of the mat genes; to David Jacobson for helpful discussions; and to the U.S. Public Health Service (USPHS) for its exemplary flexibility about R.L.M.'s grant. P.H.K.S. was supported by U.S. Public Health Service (USPHS) grant GM08995 to R.L.M.
Communicating editor: M. Zolan
Note added in proof : We have found that crosses homozygous for rid (which encodes a cytosine methyltransferase-like protein active in the sexual cycle), like those homozygous for dim-2, do not suppress the round phenotype of crosses heterozygous for Round spore. We thank Michael Freitag and Eric Selker for the rid deletion strains and for permission to cite their results in advance of publication.
- Received March 11, 2002.
- Accepted May 13, 2002.
- Copyright © 2002 by the Genetics Society of America