Immune gene expression is activated in mRNA m⁶A writer complex mutants

To understand the roles of mRNA m⁶A in Arabidopsis, we asked what groups of genes were most affected by the loss of function of the m⁶A writer complex protein, VIRILIZER. We previously characterised gene expression changes in vir-1 mutants using a combination of Illumina RNA sequencing (RNA-seq) and Oxford Nanopore Technologies direct RNA sequencing (ONT DRS) [12]. We analysed three genotypes with different VIR activity: a wild-type Col-0 control, a hypomorphic Arabidopsis vir-1 mutant defective in VIR function, and a complementation line expressing VIR fused to Green Fluorescent Protein (GFP) (VIR complemented; VIRc) that partly restores VIR activity in the vir-1 mutant background [8,12]. For each genotype grown in sterile conditions at 22°C, we sequenced RNA purified from seedlings of at least six biological replicates with Illumina RNA-seq and four with ONT DRS (S1 File).

Using the Illumina RNA-seq data, we identified differentially expressed genes between vir-1 and WT Col-0 by fitting a quasi-likelihood model in edgeR [47] (threshold: adj.p < 0.001, log2FC > 2.0). We found 806 genes significantly upregulated in vir-1 compared to Col-0 and 349 genes significantly downregulated (S2 File). We examined GO (gene ontology) term distribution among the differentially expressed genes using gProfiler [48]. The most significantly enriched GO terms were related to response to external stimuli and defence. For example, 92 of the 597 upregulated genes with GO term annotation were annotated with the biological process ‘defense response’ (GO:0006952), a significant enrichment (adjusted p = 1.32x10^-8) compared to the background of all genes (Fig 1A and S2 File).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Immune gene expression is activated in m⁶A writer complex mutants.

A) Top 10 gene ontology terms enriched in the set of 597 genes significantly upregulated (FDR < 0.001, log2FC > 2.0) in vir-1 mutants compared to Col-0 WT at 20°C. B) Heatmap showing the TMM-FPKM normalised log2 centred fold expression between vir-1, Col-0 and VIRc for all genes with TAIR annotations including the term ‘defense/defence’. PR1 (AT2G14610) is highlighted with an arrow. C) Normalised log2 counts per million of PR1 (AT2G14610) in Col-0 (n = 7), vir-1 (n = 6) and VIRc (n = 6) in Illumina RNA-seq. Boxes represent the interquartile range of the logged values. D) Normalised log2 counts per million of PR1 (AT2G14610) in Col-0, fip37-4 and fio1-1 at 20°C in ONT DRS, showing the upregulation of PR1 in fip37-4.

https://doi.org/10.1371/journal.pgen.1011925.g001

To examine the global expression trends for defence-related genes, in a different way, we identified 1033 genes that included ‘defence’ or ‘defense’ in their TAIR annotation description [49]. Of these genes, 86 were differentially expressed in at least one condition. Plotting the zero-centred fold change of these genes shows the extent of expression recovery in the VIRc complementation line (Fig 1B). For example, expression of the defence marker gene PR1 (AT2G14610) was 311-fold (8.28 log2FC) higher in vir-1 than Col-0 but restored to similar levels as Col-0 in the VIRc complementation line (Fig 1C). The upregulation of PR1 in vir-1 is also detected in the orthogonal ONT DRS data (S1A and S1B Fig). We next asked if other m⁶A writer mutants had elevated PR1 expression. We analysed ONT DRS data of fip37–4 mutants that disrupt the Arabidopsis m⁶A writer complex ortholog of WTAP [21]. Genes which are significantly upregulated in fip37–4 mutants are significantly enriched for GO terms related to defence (S3 File). Like vir-1 mutants, fip37–4 mutants have elevated PR1 expression (Fig 1F). In contrast, PR1 is not upregulated in fiona1 mutants that disrupt the Arabidopsis N6 methyladenosine methyltransferase METTL16 ortholog that adds m⁶A to U6 snRNA [21] (Fig 1D). We conclude that a phenotype of Arabidopsis mRNA m⁶A writer complex mutants is the upregulation of genes involved in defence signalling.

Loss of the mRNA m⁶A writer complex triggers temperature-sensitive autoimmunity

The upregulation of defence response genes in vir-1 raised the question of whether autoimmunity might explain the gene expression changes and pleiotropic developmental phenotypes of mRNA m⁶A writer complex mutants. Arabidopsis autoimmunity is frequently temperature-sensitive, with autoimmune phenotypes suppressed at elevated ambient temperatures [38]. Therefore, to address whether the gene expression changes we detected in vir-1 mutants reflected an autoimmune response, we compared the gene expression profiles of vir-1 and WT Col-0 seedlings grown in sterile conditions at 17°C and 27°C. We used a combination of Illumina RNA-seq and ONT DRS to analyse gene expression. We performed four biological replicates with each RNA sequencing technology, genotype and temperature treatment. The resultant sequencing statistics are detailed in S1 File.

We analysed the Illumina RNA-seq data using a quasi-likelihood model (glmQLFit) in edgeR [47] and identified genes differentially expressed between vir-1 and WT Col-0 at each temperature. This revealed 1215 genes which were significantly upregulated (adj.p < 0.001, log2FC > 2.0) in vir-1 at 17°C (S4 File). Remarkably, 91% of these genes (1103) were not significantly upregulated in vir-1 at 27°C (Fig 2A), revealing that most gene expression changes in vir-1 are temperature-sensitive. Principal component analysis (PCA) separates the biological replicates by genotype and temperature. The first component, which explains 40% of the variance, captures gene expression changes specific to vir-1 at 17°C. In contrast, vir-1 and Col-0 are indistinguishable at 27°C in this component (Fig 2B). Likewise, correlation matrix analysis reveals that the gene expression features of vir-1 mutants grown at 17°C are the most distinct among all the datasets (Fig 2C). We found elevated expression of PR1 in vir-1 grown at 17°C, just as we had previously seen at 22°C, but PR1 expression was at similar levels to WT Col-0 in vir-1 grown at 27°C (Fig 2D). We also detected this differential PR1 expression pattern with orthogonal ONT DRS data (S2A Fig) and RT-qPCR (S2B Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Loss of m⁶A triggers a temperature-sensitive autoimmune response.

A) Upset plot showing the overlap in significantly upregulated and downregulated genes (log2FC + /- 2.0 FDR < 0.001) in vir-1 at 17°C compared to Col-0 at 17°C and vir-1 at 27ºC compared to Col-0 at 27ºC. B) Principal component analysis showing the clustering of samples by experimental condition (including genotype) and temperature. C) Correlation matrix and hierarchical clustering of expression profiles for each condition. The clustering shows that gene expression patterns are distinct for all conditions. In addition, biological replicates within conditions cluster together. However, the gene expression patterns detected in vir-1 separate as the most different of all possible comparisons. D) Significant upregulation of PR1 (AT2G14610) in vir-1 at 17°C, shown by a boxplot of normalised expression (log 2 counts per million) in Illumina RNA-seq data (n = 4 samples per genotype). E) Most enriched GO terms among genes significantly upregulated (FDR < 0.001, log2FC<2.0) in vir-1 at 17°C compared to the average of: Col-0 17°C, 27°C and vir-1 at 27°C. Source data available in S6 File. F) Heatmap showing the TMM-FPMK normalised log2 centred fold change for all differentially regulated genes in vir-1 at 17°C (log2FC + /- 2.0 FDR < 0.001) with TAIR annotations including ‘defense/defence’ for conditions vir-1 at 17°C, 27°C, Col-0 at 17°C and 27°C. PR1 (AT2G14610) is highlighted with an arrow.

https://doi.org/10.1371/journal.pgen.1011925.g002

We used a generalised linear model (GLM) to model all conditions simultaneously and identify differential gene expression specific to vir-1 at 17°C. The GLM design contrasts vir-1 at 17°C minus the average of Col-0 at 17°C and 27°C and vir-1 at 27°C. This model identified 931 genes with significantly increased expression (adj.p < 0.001, log2FC>2.0) in vir-1 at 17°C compared to the other conditions (S2C Fig and S5 File). GO-term analysis revealed that the biological processes most significantly enriched in genes upregulated in vir-1 at 17°C were related to defence responses (Fig 2E and S6 File). The defence annotation GO terms were similar in describing the gene expression changes previously detected in vir-1 grown at 22°C (S2D Fig). As an orthogonal approach, we analysed protein domain enrichment using DAVID [50] (S6 File). The most significantly enriched protein domains in genes upregulated in vir-1 at 17°C included RLP23-like, which is found in receptor-like kinases that function as PRR proteins (15-fold enrichment), Defensin_plant, a domain found in highly expressed marker proteins of defence (10-fold enrichment), and WRKY_plant (6-fold enrichment), which is found in WRKY transcription factors that frequently control defence response transcription networks [51]. We used AME (Analysis of Motif Enrichment) from the MEME Suite to identify enriched motifs in promoter regions of genes upregulated in vir-1 at 17 °C compared to Col-0 at 17 °C. There was significant enrichment (p < 0.01) of 15 transcription factor binding motifs from the DAP-seq database [52], including NAC, MYB-related, and W-Box (WRKY binding) motifs, consistent with the enrichment of WRKY protein domains in the genes upregulated in vir-1 at 17 °C (S7 File). The genes upregulated in vir-1 at 17 °C versus all other conditions were enriched for ATAF1, KAN2, AT5G56840, and AT2G20400 binding motifs (S7 File). There was no enrichment of any motif in the promoters of genes upregulated in vir-1 at 27 °C compared to Col-0 at 27 °C. Together, these motif analyses are consistent with an immune state in vir-1 at 17 °C, but also indicate that other transcription factor classes may contribute to a more complex stress-related transcriptional program.

Next, we used a different approach to ask how defence gene expression was affected by temperature using the GLM analysis. We plotted the zero-centred log2FC normalised gene expression of 136 genes with defence/defense included in their TAIR annotation in vir-1 at 17°C compared to other conditions. This analysis reveals that the expression level of most genes with a TAIR defence/defense annotation in vir-1 at 17ºC is at WT Col-0 levels when vir-1 mutants are grown at 27°C (Fig 2F).

To determine whether the major defence gene expression differences observed in vir-1 at 17ºC compared to all other conditions might be explained by overlooked pathogen contamination of our experimental material, we examined our RNA-seq data for non-Arabidopsis sequences. Although the poly(A) mRNA purification step used in Illumina RNA-seq may limit the sensitivity of comprehensive bacterial RNA detection, this control should indicate if plant pathogens (including fungal and oomycete species) were a major concern. We used all vir-1 Illumina RNA-seq data to produce a de novo transcriptome assembly and searched the resulting contigs against the GenBank NR (non-redundant) database using BLASTP [53]. No significant enrichment of plant pathogen sequences was found in vir-1 17ºC samples compared to Col-0 (S8 File).

In summary, by exploiting the established frequent temperature sensitivity of Arabidopsis immunity, we discovered that the major annotation terms associated with the upregulated genes in vir-1 mutants at 17°C are related to defence. Therefore, at the gene expression level and strikingly dependent on temperature, we conclude that autoimmunity is a major phenotype of the Arabidopsis mRNA m⁶A writer complex mutant, vir-1.

Genes that function in diverse aspects of immunity are upregulated in vir-1 mutants

Given the global trend of defence gene activation in vir-1 mutants, we next analysed individual gene expression changes to understand what type of defence genes were affected. Expression of mRNA encoding the master defence transcription factor SARD1 [54,55] was upregulated (AT1G73805: log2FC 4.0) in vir-1 at 17°C but not 27°C, consistent with the established temperature sensitivity of SARD1 transcription [56] (Figs 3A and S3A). The expression of mRNA encoding the FLS2 receptor, which detects the bacterial flagellin flg22 peptide [58] and is one of the best-characterised Arabidopsis PRR proteins, was upregulated in vir-1 at 17°C (AT5G46330: log2FC 3.0) but not at 27°C (vir_27 vs Col-0_27: log2FC 1.3) (Figs 3B and S3B). We detected the upregulation of 15 annotated NLRs - 13 TIR-NLRs, one CC-NLR (LOV1) and one RPW8-NLR (HR4) at 17°C but not 27°C (S9 File). For example, the TIR-only TX0 was upregulated (AT1G57630: log2FC 4.1) (Figs 3C and S3C). TX0 can hydrolyse nucleic acids, particularly RNA, and synthesise 2’,3’,-cAMP/cGMP molecules that ultimately signal cell death in the hypersensitive response [59].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Defence genes that function in diverse aspects of immunity are upregulated in vir-1 mutants A) Significant upregulation of SARD1 (AT1G73805) in vir-1 at 17°C, shown by a boxplot of normalised expression (log2 counts per million) in Illumina RNA-seq data (n = 4 samples per condition).

B) Significant upregulation of FLS2 (AT5G46330) in vir-1 at 17°C, shown by a boxplot of normalised expression (log2 counts per million) in Illumina RNA-seq data (n = 4 samples per condition). C) Significant upregulation of TX0 (AT1G57630) in vir-1 at 17°C, shown by a boxplot of normalised expression (log2 counts per million) in Illumina RNA-seq data (n = 4 samples per condition). D) Significant upregulation of ACD6 (AT4G14400) in vir-1 at 17°C, shown by a boxplot of normalised expression (log2 counts per million) in Illumina RNA-seq data (n = 4 samples per condition). E) Upset plot showing modest overlap in differentially upregulated genes between vir-1 at 17°C and previously published acd6-1 RNA-seq data [57].

https://doi.org/10.1371/journal.pgen.1011925.g003

We next asked whether genes previously associated with Arabidopsis autoimmunity were misregulated in vir-1. The TIR-NLR RPS6 (AT5G46470) is recurrently associated with autoimmunity. For example, the extreme phenotypes of Arabidopsis nonsense-mediated RNA decay (NMD) and mitogen-activated kinase mutants have been attributed to RPS6 [60,61], although the mechanisms involved are not understood [60,61,62]. RPS6 expression is not significantly altered in vir-1 RNA-seq data, but ONT DRS analysis indicates that the TIR-only gene located downstream of the RPS6 locus is upregulated [62] (S3D Fig). The TIR-NLR SNC1 has been used as a model to understand autoimmune signalling [63]; SNC1 was not significantly upregulated in vir-1 at 17°C (log2FC 1.22), but SIDEKICK3, a TIR-NLR required for SNC1-mediated autoimmunity [64], is one of the most upregulated genes (log2FC 8.32) in vir-1 at 17°C (S9 File). Finally, we examined ACD6, which encodes a multipass transmembrane protein with intracellular ankyrin repeats, that mediates a trade-off between growth and defence [65]. First identified in lab-based mutant screens [66,67], high-activity ACD6 alleles are frequently found in natural Arabidopsis accessions [68,69,44]. ACD6 is upregulated in vir-1 at 17°C (log2FC 6.27), but the expression level is similar to WT Col-0 in vir-1 grown at 27°C (Figs 3D and S3E). We asked whether the changes in gene expression between vir-1 and acd6 mutants were related by re-analysing a recently published acd6–1 Illumina RNA-seq dataset [57]. We found a subset of differentially expressed genes overlap between these two mutants; 205 upregulated genes and 21 down-regulated genes in common (Fig 3E). However, 931 genes are uniquely upregulated in acd6–1 and 701 genes are uniquely upregulated in vir-1 at these thresholds (adj.p < 0.001, log2FC>2.0), indicating that the misregulation of ACD6 alone does not simply explain vir-1 autoimmunity gene expression phenotypes.

A group of flowering time genes paralleled the expression of immune response genes. The floral pathway integrator, FT, [70] is upregulated in vir-1 at 17°C but not 27°C (S9 File). In addition, the expression of a group of genes known to function downstream of FT in floral development, including FUL, SOC1, SEP3, SPL4, SPL5, AGL19 and AGL24 phenocopied defence gene expression patterns (S9 File and S3F-M Fig).

In summary, diverse genes attributed to the different ETI and PTI layers of Arabidopsis innate immunity were upregulated in vir-1 mutants at 17°C but not 27°C, together with mRNA encoding the SARD1 master defence transcription factor that controls SA-dependent and SA-independent defence responses.

vir-1 mutants exhibit temperature-sensitive pathogen resistance and localised cell death

A characteristic of autoimmunity is that plants show enhanced resistance to pathogens because defence gene expression is already upregulated prior to infection. We, therefore, asked whether the global changes in gene expression detected in vir-1 resulted in a functional impact on pathogen infection. We examined the susceptibility of WT Col-0, vir-1 and fls2 (flagellin sensing 2) mutants to the biotrophic pathogenic bacterium Pseudomonas syringae pv. tomato (Pto) DC3000. The fls2c mutant allele [71] lacks the receptor for bacterial flagellin and is susceptible to infection, so is a positive control for infection in these experiments. We flooded seedlings grown at 17°C, 21°C and 27°C with P. syringae pv. tomato (Pst) DC3000. We found that fls2c mutants were more susceptible than Col-0 to infection at all temperatures, consistent with previous reports [71] (Figs 4A and S4 and S10 File). In contrast, vir-1 mutants were more resistant to infection than Col-0 when grown at 17°C and 21°C, but there was no significant difference in infection levels between Col-0 and vir-1 plants grown at 27°C. Therefore, the temperature-sensitive patterns of defence gene expression detected in vir-1 convert to a corresponding change in immunity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. vir-1 mutants exhibit temperature-sensitive pathogen resistance and localised cell death.

A) Four-week-old Col-0 WT, vir-1 and fls2c seedlings flood-inoculated with a bacterial suspension of Pst DC3000 (5x10⁶ CFU/ml) and 0.025% v/v Silwet L-77. Bacterial populations were quantified at 3 days post inoculation (dpi) (n = 3 per condition). One way ANOVA tests on each genotype revealed a significant effect of temperature in the vir-1 genotype (F = 37.23, p = 0.0358) which was not present in Col-0 WT. Source data available in S9 File. This experimental analysis was replicated independently in S4 Fig. B) Trypan blue staining of Col-0 WT and vir-1 mutant leaves imaged with a Zeiss histology microscope at 10x magnification. C) Estimation of trypan blue staining patterns using ImageJ (n = 10 per condition). Two-way ANOVA revealed significant effects of temperature (F = 22.34, p = 3.45 × 10 ⁻ ⁵), genotype (F = 22.18, p = 3.63 × 10 ⁻ ⁵), and their interaction (F = 22.16, p = 3.66 × 10 ⁻ ⁵). Post hoc comparisons using Tukey’s HSD test indicated that vir-1 at 17°C significantly differed from all other conditions (p < 0.0001). Source data available in S10 File.

https://doi.org/10.1371/journal.pgen.1011925.g004

Localised premature cell death is a phenotype of plant immune responses to infection, and leaf lesions are a feature of some autoimmune genotypes [39,40]. To investigate whether vir-1 mutants exhibited elevated levels of cell death compared to WT Col-0 in the absence of pathogen infection, we stained seedlings grown in sterile conditions with the vitality marker trypan blue (TB). We recorded microscopy images and quantified the levels of detectable TB staining for 10 individual leaves of each genotype grown in each condition using ImageJ [72]. We detected the highest levels of cell death in vir-1 grown at 17°C (Fig 4B and 4C and S11 File). However, at 27°C, cell death patterns in vir-1 were at negligible levels, comparable to those detected in WT Col-0 at 17°C and 27°C. We, therefore, conclude that vir-1 mutants show elevated levels of premature cell death at 17°C when immune response pathways are autoactivated.

Overall, vir-1 mutants’ temperature-sensitive response to two orthogonal analyses of autoimmunity - enhanced pathogen resistance and increased premature cell death - is consistent with the global patterns of gene expression changes we detect at the RNA level. These findings suggest that the mRNA m⁶A writer complex is required to dampen defence pathway signalling to prevent autoimmunity in the absence of pathogens but not for the defence responses that suppress P. syringae pv. tomato (Pto) DC3000 infection. We, therefore, conclude that vir-1 mutants have a temperature-sensitive autoimmune phenotype.

mRNA m⁶A levels in vir-1 mutants are not temperature-sensitive

An ethyl methanesulfonate-induced 5’ splice site mutation in intron 5 (G + 1 to A) causes the vir-1 allele, resulting in cryptic splicing events within exon 5 that disrupt the VIR open reading frame [8]. Since mutations that disrupt splice sites can be temperature-sensitive [73], we asked if the expression of VIR mRNA was restored at 27°C. However, we found no evidence from our RNA-seq data to support this idea (S5A and S5B Fig). We next asked whether mRNA m⁶A levels in vir-1 mutants were restored to wild-type levels at 27°C. We first used liquid chromatography-tandem mass spectrometry (LC-MS/MS) to analyse the m⁶A/A (adenosine) ratio in poly(A)+ RNA purified from Col-0 and vir-1 grown at 17°C and 27°C. The level of poly(A)+ RNA m⁶A modification in the hypomorphic vir-1 allele was reduced to approximately 10% of that detected in Col-0 at both 17°C and 27°C (Fig 5A and S12 File), consistent with previous reports of reduced m⁶A levels in vir-1 mutants [8,12]. In addition to LC-MS/MS, we used the ONT DRS data to examine mRNA m⁶A levels. We have previously mapped m⁶A in ONT DRS data using the Differr and Yanocomp programs, which depend upon comparing WT and mutant genotypes [12,74]. We supplemented Yanocomp analysis with m6Anet, a neural-network-based method that can call read-level m⁶A stoichiometry without genotype comparison [75]. We found no restoration of m⁶A levels identified by m6Anet (Fig 5B) or Yanocomp analysis (S5C Fig) in vir-1 at 27°C compared to 17°C. We conclude that the suppression of immune gene expression detected in vir-1 at 27°C is not explained by an accompanying change in mRNA m⁶A modification.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. m⁶A levels in vir-1 mutants are not temperature-sensitive.

A) LC-MS/MS analysis showing the significant effect of genotype on the m⁶A/A ratio and the lack of significant interaction between genotype and temperature on m⁶A levels (two-way ANOVA; p < 0.001) (n = 3 per condition). Source data available in S11 File. B) Density distribution of the ratio of m⁶A modification per site for all sites with probability modification > 0.9 predicted by m6Anet. Individual replicates are plotted as solid lines, with the combined density of a condition (genotype and temperature) plotted as a dashed line. 3545 sites were predicted to have an m⁶A modified site in at least one Col-0 sample, compared to only 327 in vir-1.

https://doi.org/10.1371/journal.pgen.1011925.g005

The visible developmental defects of vir-1 mutants are separable from autoimmune gene expression

Arabidopsis autoimmune mutants often show developmental defects that can be rescued by growth at elevated temperatures [38]. We asked whether the impact of autoimmunity might explain the developmental phenotypes of vir-1. We examined the development of Col-0 and vir-1 mutant plants from germination to flowering and seed-set at 17°C and 27°C. We included acd6–1 as a positive control for an autoimmune mutant compromised in development at 17°C, which appears more like WT Col-0 when grown at 27°C. We found that the short stature and lack of apical dominance characterising the visible developmental phenotypes of vir-1 mutants were not rescued by growth at 27°C (Fig 6A). However, we were able to replicate the previously reported developmental rescue of acd6–1 mutants at 27°C, compared to 17°C (S6A Fig). We, therefore, conclude that the impact of the loss of the mRNA m⁶A writer complex on development and defence gene expression programmes is separable.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. The primary visible developmental defects of vir-1 mutants are not rescued by growth at 27°C.

A) Developmental phenotype of Arabidopsis WT Col-0 and mutant vir-1 grown at 17°C and 27°C. Plants are 28 days old and were grown at the indicated temperatures throughout their development following a 2-day stratification treatment.

https://doi.org/10.1371/journal.pgen.1011925.g006

Next, we asked if we could detect gene expression changes that might underpin developmental change in vir-1 mutants. We identified significant gene expression differences consistently affected by the vir-1 mutation across all our datasets, irrespective of temperature (adj.p < 0.001, log2FC>2.0) (S13 File). The flowering time control gene, FLC, was the only gene significantly downregulated across all our vir-1 experimental conditions. Only 58 genes were consistently upregulated. However, the most enriched GO term biological processes for these genes were “defence response”, “systemic acquired resistance”, and “defence response to other organism” (S6B Fig and S14 File), indicating that defence gene upregulation remains an important vir-1 gene expression phenotype. For example, RNA encoding the flavin-dependent monooxygenase, FMO1, was upregulated in vir-1 relative to WT Col-0 in all tested conditions (S6C Fig). FMO1 is a critical regulator of systemic acquired resistance to pathogen infection [76]. We also identified consistent upregulation of the AtNUDT24 nudix hydrolase. AtNUDT24 is uncharacterised, but other members of this protein family function to modify signalling nucleotides in plant defence [59,77] or in RNA decapping and hydrolysis, among other roles [78,79].

Overall, we conclude that although defence gene activation in vir-1 mutants is suppressed at 27°C, visible developmental defects are not, and autoimmune gene expression programmes remain a component of the vir-1 mutant phenotype.

Altered poly(A) tail length distributions are a temperature-sensitive phenotype of vir-1 mutants

We asked if immune response gene mRNAs were more likely to be m⁶A-modified than other mRNAs by analysing our ONT DRS datasets. It is not possible to determine m⁶A stoichiometry for every immune response gene because most are either not expressed in wild-type Col-0 in the absence of pathogen infection or are expressed at levels below the read count required by m6Anet to predict m⁶A sites. However, for the transcripts with sufficient coverage (greater than 20 mean reads in Col-0 across 17ºC and 27ºC treatments), we observe 41% fewer m⁶A sites among genes with TAIR annotations including the terms ‘defense’ or ‘defence’ compared to the Col-0 transcriptomes as a whole (Fisher’s exact test: p = 6.24x10^-6, odds ratio = 0.59).

To understand how the loss of mRNA m⁶A might trigger an autoimmune response, we next asked if RNA processing was affected in vir-1 mutants grown at different temperatures. Consistent with our previous reports [12,21], we found that genetic disruption of the mRNA m⁶A writer complex resulted in a global shift to proximal poly(A) site usage in diverse mRNAs (Fig 7A). We identified 981 genes with significant differences in poly(A) site usage between Col-0 and vir-1 at 17 °C and 820 genes with significant differences in poly(A) site usage between Col-0 and vir-1 at 27 °C. In contrast, only a small number of genes showed significant temperature-sensitive differences in poly(A) site usage, with 48 affected genes in Col-0 and 30 in vir-1. Genes with significantly altered poly(A) site usage were not enriched among those genes significantly misexpressed (FDR < 0.001) in vir-1 at 17ºC compared to the other conditions (Fisher’s exact test, p = 0.38), and there was no correlation between gene expression and the changes in poly(A) site usage observed between Col-0 and vir-1 at either temperature (Pearson’s correlation, r = 0.014, p = 0.23).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Poly(A) tail length distributions of vir-1 are disrupted in a temperature sensitive manner.

A) Shifts towards upstream (promoter-proximal) poly(A) site usage are detected in vir-1 mutants at 17°C and 27°C compared to Col-0 WT. Source data available in S14 File. B) Normalised expression (log 2 counts per million) of PR1 (AT2G14610) in ONT DRS analysis of cpsf30-yth mutants at 20°C (n = 4 per condition). C) End points of reads mapping to the 3’ end of ISTL2 (AT1G25420), showing the shift from the predominate use of the poly(A) from a single downstream site in Col-0 samples to two upstream sites in vir-1 mutants. D) Normalised expression (log 2 counts per million) of PR1 (AT2G14610) in ONT DRS analysis of te234 triple mutants at 20°C (n = 4 per condition). E) Density distribution of estimated poly(A) tail lengths in Col-0 and vir-1 at 17°C and 27°C. The distribution of each replicate is plotted individually. F) Histograms depicting the distribution of Wasserstein distance metric for significant changes in mean estimated poly(A) tail length per gene between Col-0 and vir-1 at 17°C and 27°C, and between Col-0 at 27°C and 17°C. At 17°C, 13 genes have significantly shorter poly(A) tails in vir-1, while 7,894 genes have significantly longer tails. At 27°C, 6,669 genes displayed significantly shorter poly(A) tails in vir-1, whereas 7 genes had significantly longer tails. G) In Col-0 there are 1,425 genes with significantly shorter mean estimated poly(A) tails at 17ºC compared to 27ºC. These findings are derived from data pooled across all replicates.

https://doi.org/10.1371/journal.pgen.1011925.g007

We could identify vir-1-dependent poly(A) site shifts at individual genes. For example, at ISTL2 (AT1G25420), a distal poly(A) site is preferentially used in Col-0 (Fig 7B), but two promoter-proximal poly(A) sites are preferentially used in vir-1, and the distal site, which was preferentially used in Col-0, is almost unused in vir-1. Therefore, consistent with the aggregated data, poly(A) site usage shifts are predominantly evident between Col-0 and vir-1, rather than between temperature treatments, at individual genes.

Altered 3’ end formation may trigger autoimmunity, but the subsequent signalling might be suppressed at 27°C, so we addressed this question differently. If alternative poly(A) site usage triggered the autoimmune response, it might be mediated by the CPSF30-YTH m⁶A reader. We examined gene expression changes in cpsf30-yth mutants using ONT DRS. The cpsf30-yth mutants express a truncated protein comprising the N-terminal Zinc-Finger domains that bind the AAUAAA poly(A) signal, but lack the C-terminal YTH domain. We found PR1 is not upregulated in cpsf30-yth, although it is upregulated in fip37–4 mutants analysed alongside here as a positive control (Fig 7C). Therefore, this combination of data does not provide evidence that altered poly(A) site usage triggers the autoimmunity phenotypes of Arabidopsis mRNA m⁶A writer complex mutants.

The cytoplasmic YTH reader domain proteins likely mediate specific impacts of m⁶A on RNA fate [80]. The most abundant of these are ECT2, 3 and 4 [81]. We analysed a triple mutant in each gene, te234 [82] using ONT DRS. We found no evidence that PR1 was upregulated in te234. In contrast, PR1 upregulation was again detected in fip37–4 mutants included here as a positive control (Fig 7D).

Next, we asked if poly(A) tail length was altered in vir-1 mutants. We have previously reported a change in poly(A) tail length profiles at specific genes in vir-1 [12], and we asked if this was a more widespread phenotype. We used the ONT software Dorado to estimate transcript poly(A) tail length in our ONT DRS data. This analysis reveals a characteristic periodicity in the estimated lengths of read poly(A) tails, which likely reflects the footprint of binding of multiple poly(A) binding proteins (PABPs) [83,84]. We found that the distribution of estimated poly(A) tail lengths was markedly different in vir-1 compared to WT Col-0 - at 17°C; relatively fewer transcripts with short poly(A) tails and more with longer poly(A) tails are detected in vir-1 compared to WT Col-0 (Fig 7E). At 27°C, vir-1 mutant mRNA poly(A) tail length profiles are different again, with a distribution more enriched in short poly(A) tails compared to WT Col-0 (Fig 7D). In contrast, the 30 nt poly(A) tail of Saccharomyces cerevisiae ENOLASE II RNA, used here as a spike-in calibration standard during ONT DRS library preparation, is consistent across all samples, with an estimated median tail length of 33 nt (S7A Fig). We asked if we could detect changes in estimated poly(A) tail lengths at individual genes. We assessed poly(A) tail length distributions for GAPC2 (AT1G13440) and UBQ10 (AT4G05320) transcripts, because they are highly expressed and stably detected across the experimental conditions analysed here. These genes display tail length distribution patterns that closely mirror the global read poly(A) tail length estimates: fewer short poly(A) tails and more long poly(A) tails detected in vir-1 compared to Col-0 at 17ºC and more short poly(A) tails and fewer long poly(A) tails detected in vir-1 compared to Col-0 at 27ºC (S7B and S7C Fig). In a previous study [12], we reported changes in poly(A) tail length distribution of CAB1 (AT1G29930) transcripts between wild-type Col-0 and vir-1 mutants. In the current dataset, CAB1 also exhibits temperature- and genotype-dependent shifts in tail length, with fewer short poly(A) tails and more long poly(A) tails detected in vir-1 compared to Col-0 at 17 °C (S7D Fig). However, at 27°C, more of the shortest CAB1 poly(A) tail length distributions are detected in Col-0 than in vir-1, suggesting a degree of gene-specific variation.

Changes in the global distributions of poly(A) tail length may be partly due to differences in the sets of genes expressed at detectable levels in the different conditions. To examine the differences in tail length per gene and exclude genes that were only expressed in single conditions, we used the Wasserstein distance metric to quantify shifts in mean per gene tail length distributions between conditions [62]. This analysis identified a shift towards longer mean poly(A) tails in vir-1 at 17°C compared to Col-0 and a shift towards shorter mean poly(A) tails in vir-1 at 27°C, consistent with the different distributions of estimated poly(A) tails lengths (Fig 7F and S15 File). Temperature-sensitive changes in mean poly(A) tail length in Col-0 were modest by comparison (Fig 7G).

We next asked if there was any relationship between poly(A) tail length and those mRNAs that are misexpressed in the vir-1 mutant. This analysis is limited by the fact that many of the genes upregulated in vir-1 at 17 °C are either expressed at low or undetectable levels in other conditions. For those genes for which we could make this comparison, we found the same pattern of poly(A) tail length distributions for genes expressed at significantly higher levels in vir-1 at 17 °C compared to other conditions (S7E Fig). Therefore, global changes in poly(A) tail length are not explained by the immune response genes upregulated in vir-1 at 17 °C.

Next, we asked whether the shift in mean poly(A) tail length was directly associated with the loss of m⁶A modification. We found that while genes with m6Anet-predicted m⁶A sites had shorter poly(A) tails, the temperature-dependent differences in poly(A) tail length distribution were seen in genes predicted to be either m⁶A-modified or non-modified (S7F Fig).

In summary, our findings reveal global changes in poly(A) tail length distributions as the primary temperature-sensitive mRNA processing phenotype of vir-1 mutants. This global phenotype is not restricted to genes that function in defence responses or to those transcripts predicted to have lost m⁶A modification. There is no evidence that immune response genes are more likely to be m⁶A-modified than other mRNAs in the absence of pathogen infection.

Plant material

Wild-type Arabidopsis thaliana accession Col-0 and te234 [109] were obtained from the Nottingham Arabidopsis Stock Centre. The vir-1 and VIR-complemented (VIR::GFP-VIR) lines [8] were from K. Růžička, Brno, Czechia; acd6–1 was from J. Greenberg, Chicago, USA; fip37–4 was from R. Fray, Nottingham, UK; fls2c (SAIL_691C4) [71] was from P. Hemsley, Dundee, UK; cpsf30-yth (GK477H04) was from Ł. Szewc, Poznań, Poland.

Plant growth conditions

Seeds were sown on MS10 medium plates, stratified at 4°C for 2 days, germinated in a controlled environment at 22°C under 16 hr light/8 hr dark conditions and harvested for RNA purification 14 days after transfer to 22°C. For temperature assays, plant growth chambers were set to either 17°C or 27 °C, with all other conditions the same as above. Seedlings were harvested 14 days after germination during the first two hours of the light period following an 8-hour dark phase. Four-week-old plants were used for phenotyping the adult plants at 17°C or 27 °C.

Trypan blue staining

Trypan blue staining was performed on leaves from 4-week-old plants of WT Col-0 and vir-1 grown at 17°C and 27°C. Leaves were stained in a solution of Tris-EDTA equilibrated phenol (pH 8) (25%), glycerol (25% v/v), lactic acid (25% v/v) with trypan blue (10mg/ml). Leaves were treated with staining solution for 10 minutes at 95°C then incubated overnight at room temperature. Leaves were destained in chloral hydrate solution twice, for 4 h and overnight. Stained leaves were imaged under a Zeiss histology microscope at 10x magnification. Images were imported into ImageJ [72], and the total stained area was measured in pixels, with the stained area expressed as a percentage of the total leaf area. Data collected from 10 leaves per condition was plotted, and a two-way ANOVA test with post hoc Turkey’s HSD tests was used to assess the effects of genotype and temperature and their interaction on the percentage of leaf stained.

Pathogenesis assays

Arabidopsis Col-0, vir-1 and fls2c seedlings were treated with Pseudomonas syringae pv tomato (Pto) DC3000 using flood inoculation. Bacteria were cultured on MG agar supplemented with rifampicin at 28°C for 24–48 h. Four-week-old seedlings flood-inoculated with a bacterial suspension of P. syringae DC3000 (5x10⁶ Colony-Forming Units (CFU)/ml) containing 0.025% Silwet L-77. Sterilised seedlings were grown on half-strength MS agar for 2–3 weeks at 17°C, 21°C and 27°C before inoculation. The bacterial suspension was applied to the Arabidopsis seedlings, and plates were incubated at room temperature for 2–3 minutes. Excess liquid was drained, and seedlings were maintained at growing temperatures for three days. Bacterial growth was calculated using serial dilution of material from three seedlings per plate and recorded as CFU/mg. The significance of changes in bacterial growth in the differing conditions was tested using ANOVA. The experiment was repeated to confirm results.

RNA isolation

Total RNA was isolated using the RNeasy Plant Mini kit (Qiagen) and treated with RNAse-free DNase (Promega-M6101). RNA concentration and integrity were measured using a NanoDrop one spectrophotometer and Agilent 4150 Tapestation.

Gene expression analysis by RT-qPCR

Total RNA was extracted from 14-day-old seedlings. Total RNAs were treated with RNAse-free DNase (Promega-M6101). First-strand cDNAs were synthesised using SuperScript™ III Reverse Transcriptase (Thermo Fisher Scientific-12574026). qPCR was carried out on a LightCycler^® 96 Instrument using Brilliant III Ultra-Fast SYBR Green qRT-PCR Master Mix (Agilent-600886). Three biological replicates (independently harvested samples) with three technical replicates for each were analysed. Relative expression levels were determined using the 2^−ΔΔCT method. Arabidopsis UBQ10 (AT4G05320) was used as internal control. qPCR primer pairs are listed in S16 File.

Preparation of libraries for Illumina RNA-sequencing

Illumina RNA-seq libraries were prepared by Genewiz (Azenta LifeScience) using NEB Next Ultra Directional Library Prep Kit according to the manufacturer’s instructions. Paired-end sequencing with a read length of 150 bp was carried out on the Illumina NovaSeq X following the manufacturer’s instructions. Raw sequence data was converted to fastq and de-multiplexed using Illumina bcl2fastq version 2.20.

Processing of Illumina RNA-seq data and differential gene expression analysis

Quality assessment of RNA-seq reads was performed using FastQC [110]. For digital expression, the Salmon index was built using Arabidopsis Araport11 transcript annotations [111]. Transcript and gene-level counts were estimated using Salmon (with gtf option and Araport 11 annotation) (version 1.9.0) [112]. Differential expression analysis was performed in edgeR (version 4.2.2) using a quasi-likelihood generalised linear model (glmQLFit). Annotation of genes of interest, categorising them as defence, flowering or other and returning GO annotation with further annotation was performed using custom scripts: https://github.com/bartongroup/PT_Arabidopsis_names_to_annot. To visualise dimensional reduction in the context of RNA-seq quality control, PCA plots were created using ggplot2 (version 3.5.1). Correlation and heatmap plots were generated with the ptr script from the Trinity RNAseq package (version 2.15.2) [113]. GO enrichment heatmap were made using msbio (metascape)(version 3.5.20240901) [114].

Functional enrichment analysis was performed using a combination of Goseq (version 1.42.0) [115] and g:Profiler (version e111_58_p18_f463989d) with the g:SCS multiple testing correction method and a significance threshold < 0.05 [48]. Domain enrichment analysis was performed in DAVID [50] using a FDR significance threshold of < 0.05.

Tests for motif enrichment in promoter regions were carried out using Analysis of Motif Enrichment (AME) (version 5.5.8) in MEME (version 5.5.7). Promoters were defined as the 1.5kb region upstream of the transcript start site. A set of 5000 promoter sequences from randomly selected genes were used as the background.

To determine whether pathogen contamination was present in the vir-1 RNA-seq samples, reads were mapped to the TAIR10 genome, and from the resulting bam file, unmapped reads were returned using STAR (version 2.7.11b) [116]. BBnorm (October 19, 2017) was then used to normalise the “unmapped” reads with the following setting “target=75 min=3” [117]. The normalised reads were assembled using Trinity [113] (version 2.15.2) with –trimmomatic –no_normalise. The transcriptome assembly was processed using cd-hit-est (version 4.8.1) (-c 0.90 -n 8 -T 24 -M 0) [118] to reduce redundancy at 90%. The resulting final transcriptome assembly was then searched against Genbank NR with Diamond-BLASTP using Diamond (version v2.0.5.143) [119]. The diamond BLASTP output was post-taxonomically annotated using https://github.com/peterthorpe5/public_scripts/tree/master/Diamond_BLAST_add_taxonomic_info. The final taxonomically assigned BLAST output was then interrogated for the presence of plant pathogens, as defined here (https://phytopathdb.org/pathogens_eg/). Digital expression per condition to this assembly was estimated using Salmon [120], and differential expression analysis was performed as described above.

ONT DRS library preparation

Total RNA was isolated, as detailed above. Poly(A)+ RNA was purified from approximately 100µg of total RNA using the Dynabeads mRNA purification kit (Thermo Fisher Scientific) following the manufacturer’s instructions. The quality and quantity of mRNA were assessed using the Nanodrop one spectrophotometer (Thermo Fisher Scientific) and Tape station 4150 (Agilent Technologies). ONT DRS libraries were prepared from 100ng poly(A)+ RNA for the WT Col-0 - vir-1 comparison at 17°C and 27°C. All other ONT DRS libraries were prepared from 500ng poly(A)+ RNA. Libraries were made using the Direct RNA sequencing kit (SQK-RNA002; Oxford Nanopore Technologies) according to the manufacturer’s instructions. The poly(T) adapter was ligated to the mRNA using T4 DNA ligase (New England Biolabs) in the Quick Ligase reaction buffer (New England Biolabs) for 15 min at room temperature. First-strand cDNA was synthesised by SuperScript III Reverse Transcriptase (Thermo Fisher Scientific) using the oligo(dT) adapter. The RNA–cDNA hybrid was then purified using Agencourt RNAClean XP magnetic beads (Beckman Coulter). The sequencing adapter was ligated to the mRNA using T4 DNA ligase (New England Biolabs) in the Quick Ligase reaction buffer (New England Biolabs) for 15 min at room temperature followed by a second purification step using Agencourt beads (as described above). Libraries were loaded onto R9 version SpotON Flow Cells (Oxford Nanopore Technologies) and sequenced using a GridION device at the Tayside Centre for Genomic Analysis, School of Medicine, University of Dundee, for a 48-hour runtime. Four biological replicates were performed for each genotype.

ONT DRS mapping

Reads were basecalled using Dorado version 0.5.3 (Oxford Nanopore Technologies) using the rna002_70bps_hac@v3 high accuracy model. Reads were aligned to the Araport11 transcriptome [111] and the TAIR10 Arabidopsis genome [121] using minimap2 version 2.17 [122] cond for spliced alignment. The following parameters were used for both alignments: --end-seed-pen = 15 for end seed penalties, -A1, -B1, -O2,32, -E1,0 and -C9 to tune alignment scoring. For genomic alignment, splice junction information was incorporated using the --junc-bed parameter, which referenced the annotated introns BED file. A junction bonus of 10 (--junc-bonus = 10) was applied to prioritise alignments that utilised known splice junctions, increasing alignment accuracy for spliced reads. The spliced alignment was optimised with parameters -k14, -uf, -w5, --splice, and -g2000, along with a maximum intron size of 200,000 (-G200000), --splice-flank = yes for spliced alignment flanking detection, and -z200 for seeding thresholds. Alignments were converted to BAM files and indexed using samtools version 1.18.

Prediction of m⁶A in ONT DRS data using m6Anet

Event information was extracted from raw signal data and transcriptome alignments using the f5c implementation of eventalign with event scaling [123]. Aligned event files were processed using m6anet [75]. Data preparation and inference was performed using the pretrained Arabidopsis model with the default read probability threshold of <0.0033. Predicted sites of modification were filtered using the recommended probability-modified threshold of >0.9 [75]. The distribution of predicted modification ratios for all sites passing this threshold was plotted for each condition.

Analysis of poly(A) site usage in ONT DRS data

Differential 3’ analysis was performed on bam files using the d3pendr tool as described previously [62]. The statistical significance of the 3’ shift was assessed by permuting read alignments between the control and treatment distributions to determine the maximum distance achieved by random sampling.

Estimation of poly(A) tail length in ONT DRS data

The length of poly(A) tails per read was estimated using “--no-trim --estimate-poly-a” with the following model: rna002_70bps_hac@v3 in Dorado (version 0.5.3). Reads were mapped to the Araport 11 transcriptome using minimap2 (see above), and the resulting bam file was used to generate a read-to transcript table for further interrogation. Differences in mean poly(A) tail length per gene between conditions were calculated as previously described [12]. In brief, poly(A) tail lengths were aggregated by gene ID, and where genes were present with at least ten reads in both conditions, the distributions of poly(A) tail lengths were compared using the Wasserstein distance. Significance was assessed using a permutation test with 999 bootstraps. Genes were classed as m⁶A-modified if they had at least one site above the probability-modified threshold of >0.9 in at least one Col-0 sample.

Gene tracks

Gene track figures were generated using Matplotlib (version 3.9.2) from normalised bigwig files of Illumina RNA-Seq coverage and pooled bam files of reads per condition. For tracks with >100 ONT DRS reads per condition, a random subsample of 100 reads per track was plotted.

m⁶A liquid chromatography with tandem mass spectrometry

m⁶A analysis using tandem liquid chromatography-mass spectroscopy (LC/MS-MS) was performed as previously described [12,62]. LC/MS-MS was carried out by the FingerPrints Proteomics facility at the University of Dundee. A two-way ANOVA test was used to assess the effects of genotype and temperature and their interaction on the ratio of m⁶A to A.