The NSL complex is required for piRNA production from telomeric clusters

Abstract

The NSL complex is a transcriptional activator. Germline-specific knockdown of NSL complex subunits NSL1, NSL2, and NSL3 results in reduced piRNA production from a subset of bidirectional piRNA clusters, accompanied by widespread transposon derepression. The piRNAs most transcriptionally affected by NSL2 and NSL1 RNAi map to telomeric piRNA clusters. At the chromatin level, these piRNA clusters also show decreased levels of H3K9me3, HP1a, and Rhino after NSL2 depletion. Using NSL2 ChIP-seq in ovaries, we found that this protein specifically binds promoters of telomeric transposons HeT-A, TAHRE, and TART Germline-specific depletion of NSL2 also led to a reduction in nuclear Piwi in nurse cells. Our findings thereby support a role for the NSL complex in promoting the transcription of piRNA precursors from telomeric piRNA clusters and in regulating Piwi levels in the Drosophila female germline.

Figure 1.. Germline depletion of NSL complex subunits results in transposon up-regulation.

(A) Representative images of ovaries with germline (nanos-GAL4) knockdowns of nsl1 (left), nsl2 (middle), and nsl3 (right). An ovary with knockdown of a control gene, white, is shown in each case. See also: Fig S1A. (B) Heatmap depicting log₂ fold changes of RNA abundance of all transposon families upon NSL1, NSL2 or NSL3 RNAi in ovaries and NSL1 or NSL3 RNAi in S2 cells compared with control knockdowns. NSL1 and NSL3 RNAi in ovaries have been normalized to virgin white RNAi ovaries. NSL2 RNAi has been normalized to normal white RNAi ovaries. Data from S2 cells are derived from Gaub et al (2020). RNA-seq data represent the mean of three biological replicates, that is, ovaries collected from females from three separate crosses. (C) Scatterplots comparing steady-state RNA abundance between white RNAi and NSL1 RNAi (left), NSL2 RNAi (middle), and NSL3 RNAi (right) ovaries relative to controls. Virgin white RNAi ovaries are used as the control reference for NSL1 and NSL3 RNAi, and standard (nonvirgin) white RNAi ovaries are used as the control reference for NSL2 RNAi. Genes are coloured grey and transposons are coloured blue. Data represent the mean of three independent biological replicates. (D) Scatterplots comparing steady-state RNA abundance between white RNAi and NSL2 RNAi ovaries. Group 1 (left), Group 2 (middle), and Group 3 (right) transposons are highlighted in red. Transposons were classified according to Li et al (2009). Data represent the mean of three independent biological replicates. (E) Heatmap showing Spearman correlation of all RNA-seq datasets. (F) PCA plot of all RNA-seq datasets. (G) Overlap of all differentially expressed genes and transposon elements in RNA-seq upon NSL1, NSL2, and NSL3 knockdown in ovaries.

Figure S1.

(A) Stereomicroscope photos of ovaries of flies subjected to nanos-GAL4 (25754)-driven RNAi-mediated depletion of white or NSL proteins. The Bloomington or VDRC stock center identifiers are indicated in brackets for each fly line. The same objective (magnification) was used for all images. (B) Barplots showing RT–qPCR fold changes for nsl1 (left, ovary), nsl2 (middle left, ovary), nsl2 (middle right, unfertilized eggs), and nsl3 (right, ovary) genes upon their respective germline knockdowns compared with controls. Data shown for nsl1 and nsl3 represent a comparison with white RNAi virgin ovaries as control to ensure a comparison between similar tissue morphology. Values were normalized to rp49. Values shown are an average of three biological replicates. Error bars represent SD. NSL1 RNAi (ovary) P = 8.67 × 10⁻⁷; NSL2 RNAi (ovary) P = 0.023; NSL2 RNAi (unfertilized eggs) P = 0.024; NSL3 RNAi (ovary) P = 1.05 × 10⁻⁶ (paired t test). (C) Barplot showing the percentage of hatched eggs laid by females with nos-GAL4-driven white RNAi and NSL2 RNAi for two independent crosses (#1 and #2). 439 and 495 eggs were counted for white RNAi. 339 and 259 eggs were counted for NSL2 RNAi. (D) Pairwise scatterplots showing correlations between RNA-seq replicates for each sample.

Figure S2.

(A) Volcano plots showing the log₂ fold-change and P-value of all genes and transposons from the RNA-seq of NSL2 RNAi compared with white RNAi. Data shown are obtained from the DEseq2 analysis of three biological replicates. Shown in green are all genes and transposons with an absolute log₂FC > 1 and a P-value < 0.05. Shown in red are all Group 1 transposons (from Li et al [2009]) with the same requirements. (B) Barplot showing RT–qPCR fold changes for transposons HeT-A, TAHRE, blood, and burdock in the unfertilized eggs (pure germline RNAs) upon NSL2 RNAi (orange) compared with white RNAi (white). Values were normalized to rp49. Values shown are an average of three biological replicates. Error bars represent SD. piwi P = 0.49; vasa P = 0.33 (paired t test). (C) Immuno-detection of the HeTA-Gag protein in the ovaries, a translation product of the HeT-A transposon, upon white RNAi and NSL2 RNAi. Scale bar, 10 μm. A representative image of n = 6 ovaries is shown. (D) Immuno-detection of γ-H2Av protein in white RNAi and NSL2 RNAi ovaries. Arrowheads indicate nurse cells showing accumulation of γ-H2Av. DAPI, 4′,6-diamidino-2-phenylindole. Scale bar, 10 μm. A representative image of n = 6 ovaries is shown.

Figure 2.. Small RNA sequencing reveals selective loss of piRNAs upon NSL1 or NSL2 depletion.

(A) Left: heatmap showing log₂ fold changes of sense and antisense piRNA abundance mapping to Group 1 transposons between NSL1 RNAi and virgin white RNAi ovaries. Right: heatmap showing log₂ fold changes of sense and antisense piRNA abundance mapping to Group 1 transposons between NSL2 RNAi and white RNAi ovaries. Unaffected transposons are in red or dark orange, down-regulated transposons are in white and pale orange and very strongly down-regulated transposons are in blue. Transposon classification was used from Li et al (2009). Data shown are a representative replicate from two or three replicates with high correlation to each other. (B) Sense (blue) and antisense (orange) piRNA abundance over the consensus regions of TAHRE (top), HeT-A (middle), and HMS-Beagle (bottom) is shown for the white RNAi and NSL2 RNAi small RNA-seq data. (C) Log₂ fold changes comparison between NSL1 RNAi and NSL2 RNAi on all Group 1 transposons. R represents Pearson correlation. (D) Left: heatmap showing log₂ fold changes of sense and antisense piRNA abundance between NSL1 RNAi and virgin white RNAi mapping to 50 piRNA clusters with the largest changes between NSL2 RNAi and virgin white RNAi. Right: heatmap showing log₂ fold changes of sense and antisense piRNA abundance between NSL2 RNAi and white RNAi mapping to 50 piRNA clusters with the largest changes between NSL2 RNAi and white RNAi. Data shown are a representative replicate from two or three replicates with high correlation to each other. piRNA clusters showing an increase in mapping piRNAs are in red, showing no change are in dark orange, showing a decrease in white or pale orange, and showing a very strong decrease are in blue. See Table S1. (E) Sense (orange) and antisense (blue) piRNA abundance over consensus regions of cluster 22 (top, telomeric), cluster 3 (middle, telomeric), and cluster 42AB (bottom, pericentric) is shown for the white RNAi and NSL2 RNAi small RNA-seq data. (F) Log₂ fold changes comparison between NSL1 RNAi and NSL2 RNAi on all piRNA clusters. R represents Pearson correlation.

Figure S3.

(A) Heatmap showing Spearman correlation between small RNA-seq replicates for each sample. (B) Scatterplot showing piRNA abundance of sense (left) and antisense (right) reads for white RNAi and NSL2 RNAi ovaries. Selected transposons are labeled. (C) Log₂ fold-change comparison between RNA-seq and small RNA-seq upon depletion of NSL2. R represents Pearson correlation. (D) Ping-pong analysis for all piRNAs mapping to transposons in white RNAi (left) and NSL2 RNAi ovaries (right). (E) Analysis showing the frequency of each nucleotide at each position of every transposon-mapping piRNA in white RNAi (left) and NSL2 RNAi ovaries (right). (F) Analysis showing the number of sense (blue) and antisense (red) reads of 23–29 nt length for all transposon-mapping piRNAs in white RNAi (left) and NSL2 RNAi ovaries (right).

Figure 3.. Depletion of NSL2 leads to decreased nuclear Piwi levels.

(A) Immuno-detection of Piwi and Vasa in white RNAi and NSL2 RNAi ovaries. Scale bar, 10 μm. A representative image of n = 6 ovaries is shown. (B) Immuno-detection of Aub in white RNAi (left) and NSL2 RNAi (right) ovaries. Scale bar, 10 μm. A representative image of n = 8 ovaries is shown. (C) Immuno-detection of Ago3 in white RNAi (left) and NSL2 RNAi (right) ovaries. Scale bar, 10 μm. A representative image of n = 8 ovaries is shown. (D) Barplot showing the total frequency count of full-length or partial transposon insertions belonging to the listed transposon families contained within the 20 piRNA clusters showing the highest changes in piRNA abundance (in other words, most affected) upon NSL2 RNAi (small RNA-seq log₂FC; see Fig 2D and Table S1). (D, E) Dotplot characterizing the transposon composition of piRNA clusters which contain both an NSL2 MACS2 peak and appear in the list of top 20 piRNA clusters showing the most deregulation upon NSL2 RNAi (see panel (D)). The sizes of the bubbles indicate the number of copies of a given transposon element present in that particular piRNA cluster. The colour of the bubble indicates the log₂ of the number of transposon elements in that particular cluster.

Figure S4.

(A) Barplot showing the RNA-seq normalized read counts for selected piRNA pathway genes that are important for the regulation of the pathway. Grey bars show values for white RNAi and orange bars show values for NSL2 RNAi. Values shown are an average of three biological replicates. Asterisk denotes a P-value < 0.05 (Wald test). Data shown are mean ± S.D. (B) Barplot showing RT–qPCR fold changes for piwi and vasa upon NSL2 RNAi over white RNAi in unfertilized eggs. Values are normalized to housekeeping gene rp49. Values shown are an average of three biological replicates. Error bars represent SD. (C) Western blots of Piwi, Armi, Aub, and Ago3 from white RNAi and NSL2 RNAi ovaries. Expected sizes of proteins are indicated using arrowheads (Piwi ∼100 kD; Armi 140–150 kD; Aub ∼100 kD; Ago3 ∼100 kD). (D) Flies expressing a GFP-FLAG–tagged Zucchini (Pacman BAC clone CH322-41M17 containing the zuc locus tagged with GFP-Precission-V5-3xFLAG; Hayashi et al, 2016) were subjected to either white RNAi or NSL2 RNAi. Top: ovary lysates were probed using anti-FLAG. Expected size of the ZUC::GFP::FLAG protein is indicated using an arrowhead (∼56 kD). Bottom: immuno-detection of ZUC::GFP::FLAG in the ovaries. To amplify the GFP signal (to compensate for loss of signal during fixation), the flies were immunostained with anti-GFP^Alexa488 antibody before imaging. (E) Sashimi plot showing the coverage over the piwi gene of raw alignments recovered from the RNA-seq from ovaries upon white (red) and NSL2 (blue) RNAi.

Figure S5.

(A) Heatmap depicting densities of NSL3 (left, S2 cells) and HA-3xFLAG-NSL2 (right, ovaries) binding over all genes in the fly genome. A profile depicting the average signal is shown above the heatmaps. Data shown are from one representative replicate from two well-correlating biological replicates. NSL3 S2 cell ChIP-sequencing data are derived from Lam et al (2012). (B) GO term analysis performed for the genes containing ovary-specific MACS2 peaks from the HA-3xFlag-NSL2 ChIP-seq. (C) Genome browser snapshot of the telomeric end of chromosome X showing input-normalized ChIP-seq profiles of NSL1 and HA-3xFLAG-NSL2. The zoomed-in region depicts peaks of NSL1 and NSL2, highlighted in light green shaded boxes, over the telomeric piRNA cluster, cluster 22. (D) A plot showing density of HA-3xFLAG-NSL2 over the consensus regions of blood, mdg1, HMS-Beagle, and burdock obtained by ChIP-seq. The input reads are shown in grey and the immunoprecipitated reads are shown in red. (E) Barplot showing the total frequency count of full-length or partial transposon insertions belonging to the listed transposon families contained within the 20 piRNA clusters showing the lowest changes in piRNA abundance (in other words least affected) upon NSL2 RNAi (small RNA-seq log₂FC; see Fig 2D and Table S1).

Figure 4.. The NSL complex binds to the promoters of telomeric transposons.

(A) Genome browser snapshot of the telomeric end of chromosome 4 showing input-normalized ChIP-seq profiles of NSL1 and HA-3xFLAG-NSL2. The zoomed-in region depicts peaks of NSL1 and NSL2, highlighted in green, over the telomeric piRNA cluster, cluster 3. Data show a merged bigwig of two independent replicates, that is, ovaries collected from females from two separate crosses. (B) A plot showing density of HA-3xFLAG-NSL2 over the consensus regions of HeT-A (left), TAHRE (middle), and TART-C (right) obtained by ChIP-seq. The input reads are shown in grey and the immunoprecipitated reads are shown in red. A schematic of the domain structure of each transposon is presented below. The arrow represents the location of the telomeric promoter.

Figure S6.

(A) Genome browser snapshot of the piRNA cluster 97 showing input-normalized ChIP-seq profile of HA-3xFLAG-NSL2. This cluster exhibits a MACS2-called NSL2 peak. Data show a merged bigwig of two independent replicates. (B) Genome browser snapshot of the piRNA cluster 20A and cluster 80F showing input-normalized ChIP-seq profile of HA-3xFLAG-NSL2. Note that these two clusters do not exhibit a MACS-identified NSL2 peak. Data show a merged bigwig of two independent replicates.

Figure S7.. Pairwise scatterplot showing ChIP-seq replicates correlation for each sample.

Figure S8.

(A) Profiles showing average density of H3K9me3 over all transposon insertions in white RNAi and NSL2 RNAi ovaries. Note that the y-axis begins at 2.5. TSS- 5′-end of transposon. TES- 3′-end of transposon. (B) MA plots depicting the DEseq2 results of ChIP-seq performed to assay the density of H3K9me3 in white RNAi and NSL2 RNAi ovaries. All genes and transposons showing a difference in log₂ fold-change with a P-value < 0.05 are coloured in pink. Transposon insertions showing a difference in log₂ fold-change with P-value < 0.05 are coloured in blue. Group 1 (left), Group 2 (middle), and Group 3 (right) transposons are shown. Data represent an average of three biological replicates. Only unique reads are considered for this analysis. (C) MA plots depicting the DEseq2 results of ChIP-seq performed to assay the density of H3K9me3 in white RNAi and NSL2 RNAi ovaries. All genes and transposons showing a difference in log₂ fold-change with P-value < 0.05 are coloured in pink. The 13 piRNA clusters showing a difference in log₂ fold-change with a P-value < 0.05 are coloured in blue. Data represent an average of three biological replicates. Only unique reads are considered for this analysis. (C, D) There is an overlap of six piRNA clusters between the 13 piRNA clusters exhibiting statistically significantly reduced H3K9me3 after NSL2 RNAi (from panel (C)) and the top 20 piRNA clusters showing the highest decrease in piRNA production upon NSL2 RNAi (as analyzed in Fig 3D and E). Dotplot showing the number of transposons belonging to the indicated transposon families contained within each of these six piRNA clusters. The sizes of the bubbles indicate the number of copies of a given transposon element present in that particular piRNA cluster. The colour of the bubble indicates the log₂ of the number of transposon elements in that particular cluster. (D, E) Summary of the characteristics of the six piRNA clusters from panel (D). “HTT insertion” indicates the presence of at least 1 HeT-A, TAHRE or TART transposon insertion within that cluster. NSL2 peak indicates that at least 1 NSL2 ChIP-seq peak is detected within that cluster. “Chromosome tip” indicates that the cluster is located within 70 kb of the chromosome end. (F) Density of H3K9me3, obtained from input-normalized ChIP-seq, in white RNAi (blue) and NSL2 RNAi (orange), plotted over consensus regions of telomeric transposons HeT-A (top), TAHRE (middle), and TART-C (bottom).

Figure 5.. Loss of NSL2 leads to a reduction of H3K9me3, Rhino, HP1a, and H3K4me3 over telomeric piRNA clusters.

(A) Genome browser snapshot of the piRNA cluster cluster 3, showing input-normalized ChIP-seq profiles of H3K9me3 (green), Rhino (red), HP1a (maroon), and H3K4m3 (light blue) upon white and NSL2 RNAi. Blue blocks depict peaks called by MACS2 from the HA-3xFLAG-NSL2 ChIP-seq. Data show a merged bigwig of two independent replicates, that is, ovaries collected from females from two separate crosses. (B) Genome browser snapshot of the piRNA cluster cluster 22, showing input-normalized ChIP-seq profiles of H3K9me3 (green), Rhino (red), HP1a (maroon), and H3K4m3 (light blue) upon white and NSL2 RNAi. Blue blocks depict peaks called by MACS2 from the HA-3xFLAG-NSL2 ChIP-seq. Data show a merged bigwig of two independent replicates, that is, ovaries collected from females from two separate crosses. (C) Genome browser snapshot of the piRNA cluster 42AB, showing input-normalized ChIP-seq profiles of H3K9me3 (green), Rhino (red), HP1a (maroon), and H3K4m3 (light blue) upon white and NSL2 RNAi. Blue blocks depict peaks called by MACS2 from the HA-3xFLAG-NSL2 ChIP-seq. Data show a merged bigwig of two independent replicates, that is, ovaries collected from females from two separate crosses.

Figure S9.

(A) Immuno-detection of HP1a upon white RNAi (left) and NSL2 RNAi (right). Scale bar, 10 μm. A representative image of n = 5 ovaries is shown per condition. (B) Venn diagram showing overlap between transposon elements up-regulated in the NSL2 RNAi, mael^M391/r20 and rhi^2/KG mutant ovaries. mael^M391/r20 and rhi^2/KG mutant RNA-seq data were taken from Chang et al (2019). (C) Venn diagram showing overlap between NSL2 MACS2-called ChIP-seq peaks and transposon elements up-regulated in all three conditions (NSL2 RNAi, mael^M391/r20, and rhi^2/KG mutants). (B) The common regions are not identical to panel (B) because of the fact that more than one transposon element can overlap with one ChIP-seq peak (in this case it is counted as 1).

Figure 6.. Comparison of NSL2 RNAi with mael^M391/r20 and rhi^2/KG RNA-seq data.

(A) Genome browser snapshots of the piRNA clusters cluster 3 and cluster 22, showing RNA-seq profiles of white RNAi and NSL2 RNAi (magenta); RNA-seq profiles in control w¹¹¹⁸ and mael^M391/r20 mutant ovaries (light green); GRO-seq profiles of w¹¹¹⁸ and mael^M391/r20 mutant ovaries (dark green); RNA-seq profiles of w¹¹¹⁸ control and rhi^2/KG mutant ovaries (light blue); Rhino ChIP-seq profiles of white RNAi and NSL2 RNAi (dark blue); and NSL2 ChIP-seq profile in WT (red). Dark blue blocks depict peaks called by MACS2 from the HA-3xFLAG-NSL2 ChIP-seq. White RNAi RNA-seq, NSL2 RNAi RNA-seq, NSL2 ChIP-seq, and Rhino ChIP-seq data show a merged bigwig of two independent biological replicates. RNA-seq datasets in w¹¹¹⁸ (SRR8078485, SRR8078482, SRR8078483), mael^M391/r20 (SRR8078565, SRR8078564, SRR8078563) and rhi^2/KG (SRR8078593, SRR8078594, SRR8078595) mutants and GRO-seq datasets from w¹¹¹⁸ (SRR8078585, SRR8078586, SRR8078583) and mael^M391/r20 mutants (SRR8078587, SRR8078588, SRR8078581) are from Chang et al (2019). Data show a merged bigwig of three independent replicates. (B) RT-qPCR analysis of expression of mael, NSL complex members nsl1, nsl2, nsl3, and three classes of transposon (HeT-A, TAHRE, Burdock) in NSL2 RNAi, combined NSL2 + white RNAi, mael RNAi, and combined NSL2 + mael RNAi ovaries. Each bar represents the mean ± SD of four independent biological replicates. All values were normalized first to rp49 and then to white RNAi (white RNAi level is set at 1 and is indicated by the dotted horizontal lines).

Figure S10.. Loss of NSL2 leads to mitotic defects in early embryos.

Immuno-detection of γ-tubulin upon white RNAi (left) and NSL2 RNAi (right) in fertilized 0–2-h-old embryos. Scale bar, 10 μm. A representative image of n = 10 eggs is shown.