Lem2 mediates repression of non-coding RNAs and meiotic genes

The INM protein Lem2 anchors and silences constitutive heterochromatin38,43. To examine whether Lem2 regulates gene expression through additional mechanisms, we performed transcriptome analysis by RNA sequencing (RNA-seq). A substantial portion of transcripts were upregulated in cells deleted for lem2+ (lem2∆), whereas only a few were decreased (838 versus 35, from a total of 6,642 transcripts; Fig. 1a,b). The S. pombe genome contains roughly 70% protein-coding and 21% non-coding genes44. Non-coding RNAs (ncRNAs) were significantly over-represented (61%) among the upregulated transcripts in lem2∆ cells (Fig. 1a,c). Upregulated ncRNAs include sme2, a key player during meiosis45, and the snoRNA sno20 (Fig. 1d). Consistently, the Analysis of Gene Lists (AnGeLi) tool46 revealed ‘ncRNA’ as the group of genes most significantly altered in lem2∆ cells and also uncovered other genes linked to meiosis and sporulation (Fig. 1e and Supplementary Data 1). Several long terminal repeat (LTRs) transcripts were also increased (Fig. 1a,c). We further examined mutants of Man1 and Ima1, two other integral envelope proteins known to interact with chromatin47. Genome-wide analysis of man1∆ cells revealed no major transcriptional changes (Extended Data Fig. 1a). Reverse transcription followed by quantitative PCR (RT–qPCR) confirmed that selected meiotic genes and ncRNAs were upregulated in lem2∆ cells; in contrast, these were largely unaltered in man1∆ and ima1∆ cells (Fig. 1f and Extended Data Fig. 1d), indicating a specific role for Lem2 in the regulation of these transcripts.

Fig. 1: Lem2 represses ncRNAs and meiotic genes.

a, Volcano plot depicting RNA-seq data from lem2∆ versus WT cells. Genes significantly up- (red) or downregulated (blue) are highlighted (log2(fold change) >1 or < –1 with P adjusted value < 0.01 by the Wald test, as implemented within the DESeq2 framework). Prominent transcripts are in bold. b, MA-Plot of lem2∆ cells relative to WT. x and y axes show the log2(mean expression) and log2(fold change), respectively, in lem2∆ over WT. c, Pie charts showing distributions of ncRNA, LTR, protein-coding, and other transcripts (pseudogene, rRNA, snoRNA, snRNA, tRNA). Left, genome-wide distribution of transcript features in a WT genome. Right, transcript feature distribution of the significantly upregulated transcripts in lem2∆ cells (log2(fold change) > 1 with P adjusted value < 0.01 by the Wald test). d, Coverage plots showing upregulated transcripts in lem2∆ cells from three independent biological replicates. Reads are presented as counts per million (CPM). Genomic coordinates are shown in base pairs (bp). e, Table with selected results from gene list enrichment analysis of lem2∆ mutants. The AnGeLi tool with a two-tailed Fisher’s exact test and a false-discovery rate of 0.05 was used for this analysis46. GO B.P., Gene Ontology biological process; reg. of transc. during mei., regulation of transcription during meiosis; mei., meiotic. f, Top, domain structures of Lem2, Man1, and Ima1 (length, amino acids). Bottom, transcript levels of sme2 and sno20, quantified by RT–qPCR (n = 4 independent biological replicates). g, sme2, sno20, and tlh1 transcript levels, quantified by RT–qPCR (n = 4 independent biological replicates; except clr2∆, ago1∆: n = 3; HC, heterochromatin). h, ChIP–qPCR analysis of Pol II-S5P enrichment at sme2+, sno20+, and tlh1+ genes (n = 3 independent biological replicates). For RT experiments in f and g, data are normalized to transcript levels of act1 or the average of selected euchromatic genes (act1+, tef3+, ade2+), respectively, and shown relative to WT. For h, ChIP data are divided by the input and normalized to the average level of selected euchromatin loci (act1+, tef3+, ade2+). For f–h, the individual replicates are shown in floating bar plots and the line depicts the median.

Source data

Since Lem2 represses transcription by recruiting SHREC to constitutive heterochromatin38, we examined whether heterochromatic factors also regulate meiotic transcripts and ncRNAs. RNA-seq in cells lacking the H3K9 methyltransferase Clr4 revealed increased levels of pericentromeric ncRNAs and subtelomeric mRNAs, as previously shown48,49 (Extended Data Fig. 1b). Although many of these transcripts were also increased in lem2∆ cells, the number of upregulated transcripts was significantly lower in clr4∆ cells than in lem2∆ cells (103 versus 838, Extended Data Fig. 1b,c). We examined selected targets by RT–qPCR in clr4∆ and other mutants deficient in heterochromatin assembly (swi6∆), RNAi (ago1∆), and SHREC (clr1∆, clr2∆, clr3∆). In stark contrast to heterochromatic transcripts, sme2 and sno20 transcript levels were unaltered in these mutants (Fig. 1g and Extended Data Fig. 1e); ssm4 showed modest upregulation in clr4∆ and ago1∆ mutants (Extended Data Fig. 1e), in agreement with its location within a heterochromatin island5. These data suggest that Lem2 regulates the expression of these exosome targets largely independently of heterochromatin formation.

Although heterochromatin is controlled at the transcriptional and post-transcriptional levels, meiotic genes are mainly regulated through RNA degradation by the nuclear exosome2,3,20,50. To determine whether Lem2 acts transcriptionally or post-transcriptionally, we performed chromatin immunoprecipitation followed by qPCR (ChIP–qPCR) with Ser5-phosphorylated RNA polymerase II (Pol II-S5P). As expected, clr4∆ cells showed strong enrichment of Pol II-S5P and increased transcription of tlh1+, a subtelomeric gene within heterochromatin (Fig. 1g,h). We also observed moderate Pol II-S5P enrichment at tlh1+ in lem2∆ cells, consistent with its heterochromatin function38. However, Pol II-S5P abundance was unaltered at the sme2+ and sno20+ loci in lem2∆ cells (Fig. 1g,h). This implies that Lem2 regulates meiotic and non-coding transcripts through a transcription-independent mechanism distinct from its role in heterochromatin silencing.

Lem2 and the nuclear exosome cooperate in RNA surveillance

Since meiotic transcripts and ncRNAs are major nuclear exosome substrates9, we examined whether Lem2 mediates post-transcriptional regulation through RNA degradation. We performed RNA-seq in mutants lacking components of the nuclear exosome pathway, i.e., Rrp6 (nuclear exosome), Red1 (MTREC), Erh1 (EMC complex), Iss10 (MTREC/EMC-bridging factor), Ccr4 (CCR4–NOT complex), or Air1 (TRAMP) (Fig. 2a). Principal component analysis (PCA) revealed mutant-specific groups displaying high reproducibility across independent biological replicates (Extended Data Fig. 2a). Differential expression analysis followed by unsupervised K-means clustering revealed a striking similarity between transcriptome profiles of lem2∆, rrp6∆, and red1∆ mutants. However, lem2∆ showed only limited transcriptome overlap with erh1∆, iss10∆, air1∆, and ccr4∆ (Fig. 2b). Pairwise transcriptome comparisons revealed strong positive correlations for lem2∆ with rrp6∆ and red1∆ (R = 0.79 and 0.65, respectively; Fig. 2c and Extended Data Fig. 2b) and a weaker correlation with air∆ (R = 0.53; Extended Data Fig. 2c). No correlation was seen with erh1∆ or ccr4∆ at the genome-wide level (Extended Data Fig. 2c). We also found a strong accumulation of antisense transcripts in lem2∆ cells, similar to findings in rrp6∆ cells (Extended Data Fig. 2d). Using AnGeLi46, we analyzed the top clusters containing upregulated transcripts in lem2∆, rrp6∆, and red1∆ strains (clusters 1–5). Cluster 1 was enriched for features related to early meiosis (63% frequency) and Red1-mediated degradation (30%). Many of these transcripts were also increased in cells lacking Erh1, consistent with its function in binding to exosome substrates as part of the EMC (Fig. 2b and Extended Data Fig. 2e). In contrast, transcripts present in clusters 2 to 5 were predominantly enriched for ncRNAs (between 53% and 69%), or genes related to splicing, stress regulation, and late meiosis. For cluster 3, we observed an overlap with transcripts specifically increased in air1∆ (Fig. 2b and Supplementary Data 2). Together, these results suggest that Lem2 cooperates with distinct exosome factors to control transcripts through multiple degradation pathways.

Fig. 2: Lem2 collaborates with the nuclear exosome.

a, Scheme highlighting the main players in the nuclear exosome pathway, with different targeting/bridging complexes (MTREC/NURS, EMC, CCR4–NOT, and TRAMP) providing substrate specificity. Mmi1 recognizes DSR motifs in meiotic genes. Heterochromatic domains (HOODs) are partially controlled through CCR4–NOT and TRAMP. b, K-means clustering of RNA-seq data, based on differential expression. Clusters with most upregulated transcripts in lem2∆ (clusters 1–5) were analyzed with AnGeLi with a two-tailed Fisher’s exact test and a false-discovery rate of 0.05 (ref. 46). The red–blue color scale represents log2(fold change expression) relative to the WT level for each given gene and mutant. c, Scatterplot of genome-wide log2(fold change) expression from transcripts in rrp6∆ versus lem2∆, both relative to WT. The linear regression and the Pearson correlation coefficient (R) are shown. d, Expression changes in selected genes regulated by the nuclear exosome analyzed by RT–qPCR (‘Mmi1 regulon’ and others). HC, heterochromatin controls; EC, euchromatin controls. Color scale represents log2(fold change expression) relative to WT for each given gene and mutant. e, Clustering based on Pearson’s correlation coefficient (PCC) of RT–qPCR data, with genes regulated by the exosome in the indicated strains. f, sme2 and ssm4 transcript levels quantified by RT–qPCR. Data are normalized to act1 transcript levels and are shown relative to WT on a log2 scale (n = 6 independent biological replicates). Letters denote groups with significant differences from one-way analysis of variance (ANOVA) followed by Tukey’s post hoc tests at P < 0.05. g, sno20 and snR42 transcript levels quantified by RT–qPCR with primers specific to either the precursor or mature forms. Data are normalized to act1 transcript levels and shown relative to WT (n = 4 independent biological replicates). h, Northern blot analyses of RNAs in the indicated strains. Asterisks denote the precursor species of RNAs. rRNA served as a loading control. Two independent experiments were performed with similar results. For f and g, individual replicates are shown in a floating bar plot, and the line depicts the median.

Source data

To confirm co-regulation of these transcripts by Lem2 and the exosome, we analyzed a subset of transcripts recognized by Mmi1 (the ‘Mmi1 regulon’)51. In accordance with transcriptomics (Fig. 2b), RT–qPCR revealed significant changes in Mmi1-regulated RNAs in lem2∆, rrp6∆, and red1∆ cells and cells lacking the poly(A)-binding protein Pab2 (Fig. 2d,e). Conversely, transcript levels of rrp6 and other exosome genes were largely unaltered in the lem2∆ mutant. Therefore, upregulation of exosome targets is not due to loss of exosome gene expression (Extended Data Fig. 2f).

Red1 associates with chromatin at Red1-dependent heterochromatin (HC) islands5. We noticed that genes upregulated in lem2∆, such as mei4+, are often part of these HC islands, whereas Red1-independent HC islands were mostly unaffected by lem2∆ (with the exception of a ncRNA and an LTR gene; Extended Data Fig. 3a). This prompted us to examine the chromatin environment of mei4+. Deleting iss10+ decreased Red1 binding, as expected. However, we observed no change in chromatin association of Red1 or Mmi1 in lem2∆ cells (Extended Data Fig. 3b,c). Moreover, this locus retained H3 dimethylation at K9 (H3K9me2) in lem2∆ cells, but not in rrp6∆ cells (Extended Data Fig. 3d). We also examined H3K9me at HOODs, which assemble heterochromatin upon loss of Rrp6 (ref. 27), but observed no H3K9me2 increase in lem2∆ cells (Extended Data Fig. 3d). These findings further support the hypothesis that Lem2 acts at the post-transcriptional level to control exosome targets, rather than through chromatin changes.

The close resemblance of transcriptional profiles indicates that Lem2 may act through a pathway that also involves Pab2, Red1, and Rrp6. We examined prominent exosome targets (sme2, ssm4, sno20, and snR42) in single and double mutants lacking these factors in combination with lem2∆ by RT–qPCR. Although pab2∆, red1∆, and rrp6∆ single mutants displayed higher transcript levels of sme2 and ssm4 than did the lem2∆ mutant (Fig. 2f and Extended Data Fig. 3e), additional deletion of lem2+ resulted in a non-additive phenotype, implying an epistatic interaction. In contrast, snoRNA levels showed an additive increase in lem2∆ red1∆ double mutants (Extended Data Fig. 3e). SnoRNAs are highly abundant and derived from Pab2-mediated 3′ processing of RNA precursors, which accumulate in the absence of Rrp6 (ref. 15). Consistent with a role for Lem2 in 3′ processing, we found that the precursors of sno20 and snR42, but not their mature forms, accumulated in lem2∆ cells (Fig. 2g,h). Together, these results indicate that Lem2 plays a broad role in RNA degradation by collaborating with distinct pathways with different substrate specificities.

Lem2 interacts with exosome-targeting factor Red1

Given the epistatic interaction between lem2+ and red1+ (Fig. 2f), we tested for physical interaction between Lem2 and Red1 by expressing Lem2 fused to green fluorescent protein (Lem2-GFP) and Red1 tagged with 6 copies of the hemagglutinin epitope (Red1-6×HA) from their endogenous loci, as described previously28,38. Co-immunoprecipitation (coIP) revealed that Red1 associates with Lem2 in vivo (Fig. 3a). This association was insensitive to RNase and benzonase treatment, similar to MTREC interactions with other factors9,28.

Fig. 3: Lem2 physically interacts with the nuclear exosome through the MSC domain.

a, Co-immunoprecipitation of Red1-6xHA with Lem2-GFP in untreated cells, or cells treated with RNase or benzonase. H3 served as loading control. Shown is a representative example of experiments reproduced at least three times. b, Y2H analysis of spRed1 (S. pombe Red1), hsRed1 (H. sapiens homolog of Red1 corresponding to ZFC3H1 492–1308 aa), or spRrp6 with MSC-spLem2 or MSC-hsLem2, grown for 3 days on medium with increasing auxotrophy selection (SDC, synthetic media with dextrose and complete amino acid mix; -Trp, -Leu, -His, without tryptophan, leucine, histidine, respectively). Fusions with Gal4 activation domain (pGADT7-AD) or Gal4 DNA-binding domain (pGBKT7-BD) are shown. c, Schematic representation of Lem2 truncation constructs. Protein domains and positions (amino acids) are highlighted. All constructs were C-terminally GFP-tagged and expressed from the endogenous locus. d, Immunoblot of Lem2-GFP constructs in c. H3 served as a loading control. e, Transcript levels of sme2, ssm4, mei4, and ade2, quantified by RT–qPCR on the Lem2 truncation mutants shown in c. Data are normalized to act1 transcript levels and shown relative to WT (n = 4 independent biological replicates). The individual replicates are shown in a floating bar plot and the line depicts the median.

Source data

Lem2 has distinct structural domains that mediate different functions. The LEM domain contributes to centromere association, whereas the MSC domain mediates heterochromatin silencing38. In addition, a region adjacent to the first transmembrane domain interacts with the integral membrane protein Bqt4 (refs. 42,52). Although deleting the amino terminus did not impair Red1 association, removing the carboxy-terminal MSC domain abolished Red1 binding (Extended Data Fig. 4a). To test whether the MSC domain is sufficient and essential for Red1 association, we performed yeast two-hybrid (Y2H) assays. Consistent with our coIP results, Lem2-MSC and full-length Red1 interacted in the Y2H assay (Fig. 3b). Lem2-MSC also associated with Iss10, but not with Rrp6, Pab2, or Mtl1 (Extended Data Fig. 4b). In humans, the helicase Mtr4 associates with the zinc finger protein ZFC3H1, forming the PAXT complex, the human homolog of MTREC12. Remarkably, we found a strong interaction between hLEM2-MSC and ZFC3H1 across species, indicating that binding between Lem2 and MTREC is conserved (Fig. 3b).

Next, we analyzed selected targets by RT–qPCR to test whether the MSC domain mediates the repression of exosome substrates using a series of published Lem2 truncation mutants (Fig. 3c)42. We confirmed the expression of these constructs by immunoblots (Fig. 3d). N-terminal truncations of Lem2 did not impact the level of sme2 or other transcripts. However, mutants lacking the C-terminal region largely phenocopied the repression defect of the full deletion (Fig. 3e). To further test whether the perinuclear localization of the MSC domain is important for target repression, we generated a soluble C-terminal Lem2 fragment lacking the two transmembrane domains. This fragment was fused to GFP and the SV40 nuclear localization signal (NLS), and expressed using either the endogenous lem2 promoter or the strong TEF promoter (Extended Data Fig. 4c). Both constructs produced a diffuse nuclear pattern that differs from the rim shape of full-length Lem2 (Extended Data Fig. 4d). Notably, expression of soluble MSC-GFP failed to suppress the accumulation of meiotic transcripts in lem2∆ cells regardless of their protein levels (Extended Data Fig. 4e,f), similar to previous observations of heterochromatin transcripts38. We therefore conclude that Lem2 cooperates with the nuclear exosome through Red1 interaction and that perinuclear localization of the MSC domain is crucial for regulating exosome targets.

Lem2 regulates silencing of exosome targets at the nuclear periphery

Our data (Extended Data Fig. 4f) and previous studies38,43 indicate that perinuclear location of Lem2 is critical for its gene repression function. This implies that exosome targeting or degradation of Lem2-dependent RNA substrates takes place at the nuclear periphery. To test this hypothesis, we first examined the subnuclear localization of the sme2 RNA, which contains 25 DSR motifs and is one of the most upregulated transcripts in the lem2∆ mutant (Fig. 1a,b). Using single-molecule fluorescence in situ hybridization (smFISH), sme2 transcripts were readily detectable in vegetative cells lacking Red1 or in wild-type (WT) cells undergoing meiosis (Extended Data Fig. 5a), during which sme2 forms the Mei2 dot31. However, sme2 smFISH signals were undetectable in WT or lem2∆ cells during mitotic growth, precluding further analysis (Extended Data Fig. 5a). We overcame this technical challenge using an engineered strain23, which expresses a reporter containing 14 DSRs and 4 U1A small nuclear RNA (snRNA) stem loops that can be visualized by co-expressing U1A-yellow fluorescent protein (YFP) (Fig. 4a). We confirmed that this reporter undergoes Iss10-, Red1-, and Lem2-dependent transcript elimination (Fig. 4b and Extended Data Fig. 5b). Using live-cell imaging, we studied the localization of the DSR-containing RNA reporter relative to the nuclear periphery (Cut11-mCherry). In both WT and lem2∆ cells, expression of the 14×DSR reporter resulted mostly in a single dot. We next determined the frequency at which the DSR dot localizes to specific nuclear areas using a zoning assay (Extended Data Fig. 5c)53. Interestingly, the DSR dot preferentially localized to the nuclear periphery in 40% of WT cells (zone I), whereas only 20% of the lem2∆ cells showed this pattern, indicating that Lem2 promotes the perinuclear localization of this exosome RNA substrate (Extended Data Fig. 5c). This result differed from the localization of genomic loci, as shown for the lacO array/GFP-lacI marked sme2+ locus, which did not preferentially appear at the nuclear periphery (Extended Data Fig. 5d). DSR-containing transcripts localize within nuclear Mmi1 foci, which have been proposed to be degradation sites of meiotic transcripts13,18,26. We therefore assessed the subnuclear localization of Mmi1 and the DSR reporter. Expression of CFP-Mmi1 resulted in several foci, some of which overlapped with the DRS reporter dot, as previously reported23. Notably, foci that co-localized with the DSR dot displayed Lem2-dependent perinuclear localization (43% versus 23% for zone I; Fig. 4c), implying this exosome substrate is recognized close to the NE. In contrast, the total pool of CFP-Mmi1 foci showed a general subnuclear distribution in WT and lem2∆ strains (29% versus 26% for zone I) (Fig. 4c). These data indicate that Lem2 regulation of exosome targets occurs at the nuclear periphery, where a subfraction of the nuclear pool of the elimination factor Mmi1 localizes.

Fig. 4: Exosome substrates localize at the nuclear periphery.

a, Schematic representation of the engineered DSR-containing construct, adapted from ref. 23. Luciferase is expressed with 14 copies of DSR and 4 copies of the U1A tag. Padh1, adh1+ promoter; Tnmt1, nmt1+ terminator. b, Transcript levels of luciferase and tef3, quantified by RT–qPCR in strains encoding 0×DSR copies (0×DSR), 14×DSR copies (14×DSR), or 14×DSR and the NE marker in a WT (14×DSR Cut11-mCherry) or lem2∆ background (14×DSR Cut11-mCherry lem2∆). Data are normalized to act1 transcript levels and shown relative to the 14×DSR strain (n = 3 independent biological replicates). The individual replicates are shown in a floating bar plot, and the line depicts the median. Coloured dots underneath the x-axis denote combinations of conditions shown left of the graph. c, Top left, representative images from live-cell microscopy of the DSR-containing strain, with CFP-Mmi1 and Cut11-mCherry in a WT or lem2∆ background. Cut11-mCherry marks the NE. A single z-stack is shown. Top right, schematic representation of the S. pombe nucleus, divided into three equal areas (zones) designated I–III. Bottom, quantification of CFP-Mmi1 location in WT and lem2∆ backgrounds relative to the periphery expressed in percentage of dots. Locations of CFP-Mmi1 foci were determined in cells in which the Mmi1 dot exclusively overlapped with the DSR dot (left) and in all cells (right); n denotes the number of cells counted in two independent experiments. Statistical analysis was performed using χ2 test. **P = 0.0048; n.s., not significant (P = 0.7469). Scale bar, 1 μm.

Source data

Lem2 assists exosome substrate targeting

Given its interaction with Red1 and role in repression and localization of exosome targets (Figs. 2–4), we speculated that Lem2 supports RNA turnover through recognition and handover to MTREC. The MSC domain of the human Lem2 homolog MAN1 (hMAN1) binds to DNA in vitro54 and contains an RNA-recognition motif55. However, S. pombe homologs lack this C-terminal extension, and whether Lem2 can bind RNA is unknown.

We used RNA immunoprecipitation (RIP) followed by RT–qPCR to assess binding of sme2 or ssm4 transcripts, which contain several DSRs (Fig. 5a)3. When immunoprecipitating Lem2-GFP, we were unable to detect those transcripts (Fig. 5b). We therefore tested the alternative hypothesis that Lem2 plays an accessory role in loading RNAs onto exosome-targeting factors. We expressed GFP-Mmi1 and Red1-Myc from their endogenous loci and confirmed that the epitope-tagged versions are functional (Extended Data Fig. 6a). In agreement with previous reports2,13, we found that sme2 and ssm4 transcripts were abundantly enriched with Mmi1 and Red1 (Fig. 5c,d). Strikingly, deletion of lem2+ markedly reduced binding of these transcripts to Mmi1 and Red1 (Fig. 5c,d). Binding of exosome substrates to Mmi1 was less affected in the absence of Red1 (Extended Data Fig. 6b), which is proposed to act downstream of substrate recognition by Mmi1 (ref. 4). Moreover, these substrates can be captured by Mmi1 even when they accumulate to high levels, as seen in red1∆ cells, implying that the substrate-binding capacity of Mmi1 is not limited under these conditions (Extended Data Fig. 6c). Together, these results reveal a critical role for Lem2 in the early step of RNA recognition.

Fig. 5: Lem2 promotes binding of RNA targets with exosome-targeting factors.