Fig. 6: Loss of NDX function induces increased sRNA expression and R-loop formation.

a Left: sRNA expression levels are significantly increased in the ndx1–4 mutant compared to Col-0. The size distribution of sRNA molecules peaks at 21-nt and 24nt, respectively. Sample size n = 3 biologically independent experiments. Statistical sifnificance is indicated (prop.test, two sided, Bonferroni correction). Right: Distribution of sRNA levels in Col-0 and ndx1–4 plants over the nine functional sRNA clusters identified by⁵¹. Box plots show the medians and 95% confidence intervals. Clusters 1–3 are associated with protein coding ORFs and miRNA loci and sRNA levels show no difference between ndx1–4 and Col-0. Clusters 4–9 are associated with Pol IV, and in part, Pol V and sRNA levels are significantly increased in ndx1–4 (p < 2.2e-16, Kruskal–Wallis test with Mann–Whitney test (for multiple pairwise comparison) and Benjamini–Hochberg correction). Sample size n = 3 biologically independent experiments. b Heatmap showing upregulated (orange) and downregulated (blue) sRNA loci in ndx1–4. sRNA reads were aligned to sRNA loci from⁵¹. Three independent biological replicates are shown in the diagram. P < 0.0001, Mann–Whitney rank sum test. c, d Representative examples of upregulated sRNA loci and their validation by stem loop rt-qPCR. The bar chart shows positive strand expression levels normalized to PP2AA3 RNA expression. Sample size n = 3 biologically independent experiments. Error bar: SEM. e Increase of siRNA-levels in the absence of NDX function. The ratio of aligned siRNA reads (ndx1–4/Col-0) is significantly increased in pericentromeric TEs, Knob TEs and chromosome arm TEs relative to “negative control” regions (tRNA and rRNA genes). Statistics: Kruskal–Wallis test with Mann–Whitney test (for multiple pairwise comparison) and Benjamini–Hochberg correction. Sample size (n) is indicated. f The number of aligned sRNA reads show a significant increase in ndx1–4 over the functional categories of TEs (RdDM TEs vs. CTM2 TEs). sRNA read counts were normalized to TE length. The siRNA levels from RdDM loci are also significantly higher than those from CMT2-only loci (both in Col-0 and ndx1–4; Statistics: Kruskal–Wallis test with Mann–Whitney test (for multiple comparison) and Benjamini–Hochberg correction. Sample size (n) is indicated. g The same as f but sRNAs were sized as 21, 22, 23, 24 nt siRNAs. In the 24nt class, there is a statistically significant difference between the expression status of ndx1–4 and Col-0 samples. Bounds of boxes in e–g describe the interquartile range with the median; whiskers indicate minimum and maximum values; outliers are not shown. h Stem-loop rt-qPCR validation of increased sRNA production from Copia28 and MULE1 transposons (normalized to PP2AA3 expression). Error bar: SD. Sample size n = 4 biologically independent replicates.