An RNAi-based platform for spatiotemporal control of histone gene expression during animal development

Abstract

Background

Mutational analysis of histones provides an important means of studying the function of histone post-translational modifications (PTMs) in epigenetic gene regulation. However, several technical challenges have impeded direct tests of histone residue function in metazoans, including the massive abundance of histone gene products, multiple copies of histone genes in the genome, and the necessity of histones for cell viability.

Results

Here, we describe a new experimental approach in Drosophila for selective depletion of the replication-dependent histone H3.2. Using short hairpin RNA (shRNA) transgenes, we demonstrate effective depletion of endogenous H3.2 gene expression, which causes defects in cell proliferation and organ development. We further show that a histone replacement transgene engineered to be insensitive to RNA interference (RNAi) fully rescues shRNA-mediated developmental defects. Last, we demonstrate that this selective depletion platform recapitulates phenotypes caused by histone gene mutation.

Conclusions

We conclude that shRNA-mediated depletion of endogenous histone H3.2, coupled with histone replacement transgenes engineered to be insensitive to RNAi, is an effective experimental approach for studying the role of histone PTMs in animal development.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13072-025-00657-8.

Background

During development, a single fertilized egg generates a diversity of cell types with specialized functions to form an adult organism. This cellular diversity arises not from changes to the underlying DNA sequence, but instead through differential gene expression and execution of lineage-specific developmental programs. One essential layer of differential gene regulation comes in the form of epigenetic modifications to chromatin, including post-translational modification of histone proteins.

Histones are the protein cores of nucleosomes, the fundamental repeating unit of chromatin. Nucleosomes are composed of four core histones: H2A, H2B, H3, and H4 [1]. These proteins are encoded by two categories of histone genes. The replication-dependent histone genes are expressed at high levels only during S phase of the cell cycle when large amounts of histone protein are required to package the newly synthesized genome into chromatin [2]. By contrast, the replication-independent histone genes are expressed throughout all cell cycle stages and help to maintain chromatin structure at sites of nucleosome turnover, in addition to their contribution to genome packaging in S phase [3].

Histone post-translational modifications (PTMs) have been directly implicated in regulation of gene expression [4]. In particular, the N-terminal “tails” of histones are subject to a wide array of chemical modifications. Certain modifications, such as trimethylation of histone H3 at lysine residue 27 (H3K27me3), contribute to formation of a repressive chromatin state that silences target gene expression [5]. Other PTMs, such as methylation of histone H3 at lysine residue 4, are correlated with active chromatin states [6]. The co-occurrence of specific histone PTMs in chromatin serves as the basis of the “histone code” hypothesis, which proposes that specific combinations of histone PTMs regulate DNA-templated processes such as transcription, replication, and repair [7]. However, despite strong correlations with gene activity states, the extent to which histone PTMs play a direct role in gene regulation remains unclear [8]. One approach employed to study histone PTM function has been to mutate histone modifying enzymes, such as methyltransferases or acetyltransferases. However, these enzymes often have non-histone substrates and catalytic-independent functions, thus any phenotypes observed may not be due to histone PTM loss alone [8–10]. An alternative approach is to directly interrogate the phenotypic consequences of histone PTM loss by mutating the histone residues themselves to alternative residues that cannot be post-translationally modified.

Histone gene manipulation is challenging in most animals because histone genes often exist in many copies that reside at different genomic locations. For instance, the human histone genes are found at three different genomic loci [11]. By contrast, the replication-dependent histone genes of Drosophila melanogaster exist at a single genomic locus, referred to as HisC, which contains approximately 100 tandemly arrayed copies of a 5 kb histone gene repeat unit [12]. Each 5 kb repeat contains the four core histone genes plus the linker histone H1 gene. This organization allows for the complete removal of the replication-dependent histone genes with a single genetic deletion, referred to as ΔHisC [13]. Homozygous HisC deletion results in early embryonic lethality. However, ΔHisC lethality can be rescued with transgenes bearing tandem arrays of the wild-type histone gene repeat unit [13, 14]. These “histone replacement transgenes” can be mutated through site-directed mutagenesis to express mutant histones that cannot be post-translationally modified at specific amino acid residues. Aside from methionine substitutions, which have a dominant phenotype when expressed [15–18], most mutant histone replacement transgenes can be maintained in ΔHisC heterozygotes without detectable developmental phenotypes because the transgenic histones are expressed at low levels relative to the 100 copies of the remaining endogenous histone genes [14, 19]. Homozygous HisC deletion then results in exclusively mutant histone expression from the replacement transgene, permitting study of the roles of histone post-translational modifications in animals [14, 20–28].

Although histone gene replacement methods have proven valuable for studying the role of histone PTMs in animal development, limitations still exist [29]. Due to the central role of histones in regulating genome function, many histone mutations are lethal. As a result, histone mutant phenotypes are often investigated through genetic mosaic analysis of homozygous HisC deletion clones generated by mitotic recombination. While powerful, recombination-based approaches are limited in spatiotemporal control and by the number of mutant cells that can be analyzed. We sought to determine whether these limitations could be circumvented by employing the GAL4/UAS binary expression system to deplete histone transcripts via short hairpin RNA (shRNA)-induced RNA interference (RNAi). Thousands of GAL4 drivers exist for D. melanogaster allowing tight spatiotemporal control of UAS-driven shRNAs targeting histones. We reasoned that shRNA-mediated depletion of endogenous histones could be coupled with transgenic histones engineered to be insensitive to RNAi, thereby allowing for selective expression of transgenic histones in specific tissues and developmental stages. We focus here on the sole replication-dependent histone H3 gene, H3.2.

Materials and methods

Drosophila stocks and genetics

All crosses were maintained on standard corn media (Archon Scientific) at 25⁰C. The following fly stocks were used: w; tub-GAL4/CyO act-GFP (gift of Duronio lab), w; UAS-GFP (gift of Matera lab), w; ey-GAL4/CyO (BDSC #5535), w; GMR-GAL4 (BDSC #1104), prd-GAL4 (BDSC #1947), yw; act > STOP > GAL4 UAS-GFP (BDSC #4411), nub-GAL4 (BDSC #25754), CoinFLP-GAL4 (BDSC #58754), UAS-E(z) RNAi (BDSC #33659), ΔHisC^Cadillac [29], H3.2^WT−mel [30]. Transgenic lines reported here for the first time were created at BestGene by phiC31 integration into the 86FB attP site (UAS-H3.2-shRNAs) and GenetiVision by phiC31 integration into the VK33 attP site (H3.2^WT−sim, H3.2^K27R−sim). For adult viability measurements, target genotypes were plotted as a proportion of total progeny. For developmental viability assays, at least 150 embryos were scored for hatching 48 h after egg laying. Survival of those larvae was subsequently scored. Conventional dissecting scope images were acquired with a Canon EOS Rebel T3i digital camera. To induce mitotic clones, vials were heat-shocked at 37⁰C for 10 min at 48–53 h after egg laying (h AEL). For generating FLP-out clones, vials were heat-shocked at 48–53 h AEL (early expression) or 67–72 h AEL (late expression), or 55–60 h AEL for CoinFLP experiments. Larvae were dissected 24-h after heat-shock for the late expression condition or at wandering third instar stage for all other imaging of wing and eye-antennal imaginal discs. Pupae were dissected at the pharate adult stage and scored according to the following classes. Class I: Head completely missing to remnants of head cuticle. Class II: Head present with highly reduced to completely missing eyes. Class III: Wildtype head with slightly reduced eyes. Class IV: Wildtype head and eyes. Additional details on crosses are available upon request.

shRNA design and cloning

The DSIR tool [31] was utilized to generate hairpin sequences against the D. melanogaster H3.2 3’-UTR sequence. The following four sequences were selected:

shRNA #1: 5′–UUAUAGAGUACGCUAGCGCUU–3′.

shRNA #2: 5′–UUUAUCUGCAAGUUAAUGCCG–3′.

shRNA #3: 5′–UAUCUGCAAGUUAAUGCCGUG–3′.

shRNA #4: 5′–UAGCGCUUUAUCUGCAAGUUA–3′.

Oligos were cloned into the VALIUM20 vector [32] following established protocols [33].

shRNA#1 Top Strand:

5′–CTA GCA GTA AGC GCT AGC GTA CTC TAT AAT AGT TAT ATT CAA GCA TAT TAT AGA GTA CGC TAG CGC TTG CG–3′

shRNA#1 Bottom Strand:

5′–AAT TCG CAA GCG CTA GCG TAC TCT ATA ATA TGC TTG AAT ATA ACT ATT ATA GAG TAC GCT AGC GCT TAC TG–3′

shRNA#2 Top Strand:

5′–CTA GCA GTC GGC ATT AAC TTG CAG ATA AAT AGT TAT ATT CAA GCA TAT TTA TCT GCA AGT TAA TGC CGG CG–3′

shRNA#2 Bottom Strand:

5′–AAT TCG CCG GCA TTA ACT TGC AGA TAA ATA TGC TTG AAT ATA ACT ATT TAT CTG CAA GTT AAT GCC GAC TG–3′

shRNA#3 Top Strand:

5′–CTA GCA GTC ACG GCA TTA ACT TGC AGA TAT AGT TAT ATT CAA GCA TAT ATC TGC AAG TTA ATG CCG TGG CG–3′

shRNA#3 Bottom Strand:

5′–AAT TCG CCA CGG CAT TAA CTT GCA GAT ATA TGC TTG AAT ATA ACT ATA TCT GCA AGT TAA TGC CGT GAC TG–3′

shRNA#4 Top Strand:

5′–CTA GCA GTT AAC TTG CAG ATA AAG CGC TAT AGT TAT ATT CAA GCA TAT AGC GCT TTA TCT GCA AGT TAG CG–3′

shRNA#4 Bottom Strand:

5′–AAT TCG CTA ACT TGC AGA TAA AGC GCT ATA TGC TTG AAT ATA ACT ATA GCG CTT TAT CTG CAA GTT AAC TG–3′

Drosophila simulans H3.2 3’-UTR and tandem array cloning

The D. simulans H3.2 3’-UTR sequence was identified via BLAST [34] and cloned into the 5 kb histone gene unit using Q5 mutagenesis (NEB). Tandem arrays containing twelve copies of the histone gene unit were generated as previously described [35].

Sequence comparisons of H3.2 paralogs

The H3.2 gene copy number was determined via BLAST using the D. melanogaster H3.2 reference sequence and the complete assembly of the histone locus [36]. Sequence identity was analyzed using MATLAB (The MathWorks, Inc.) Bioinformatic Toolbox.

Immunofluorescence

Standard immunostaining protocols were performed for embryos, wing imaginal discs, and eye-antennal imaginal discs. Primary antibodies and concentrations are as follows: chicken anti-GFP (Abcam, cat# ab13970) 1:250, mouse anti-H3 (Active Motif, cat# 39064) 1:3000, mouse anti-Ubx (Developmental Studies Hybridoma Bank, cat# FP3.38) 1:30, mouse anti-Abd-B (Developmental Studies Hybridoma Bank, cat# 1A2E9) 1:30, rabbit anti-H3K27me3 (Cell Signaling Technology, cat# 9733) 1:500. Secondary antibodies used at 1:1000: Alexa Fluor 555 (Thermo Fisher Scientific, cat # A21424), Alexa Fluor 647 (Thermo Fisher Scientific, cat# A21245), Alexa Fluor 647 (Thermo Fisher Scientific, cat# A21449), Alexa Fluor 647 (Thermo Fisher Scientific, cat# A32728). Confocal images were acquired using a Leica SP8 and processed in ImageJ. Uniform adjustments were made to brightness and/or contrast for figures. Image quantification of prd-GAL4 embryos was performed on raw image files, and signal intensity in the region of interest was determined using ImageJ and analyzed using MATLAB. Thresholding of GFP signal was used to determine GAL4-expressing regions. After background subtraction, the mean signal intensity for each region was calculated for each embryo and plotted.

Scanning electron microscopy

Scanning electron microscopy (SEM) images of newly eclosed (≤ 48 h) and dissected pharate adults were acquired with a Hitachi TM4000Plus tabletop SEM microscope. Dissected pharate adults were imaged at 10 kV and 100 × magnification while GMR-GAL4 flies were imaged at 15 kV and 200 × magnification.

Results

An RNAi based platform for manipulating histone gene expression

For an RNAi-based approach to be effective at depleting endogenous histone gene expression (Fig. 1A), the shRNA must successfully target transcripts produced by each of the endogenous H3.2 genes present in the genome. Any variation in shRNA target sequence across the tandem array would reduce efficacy of H3.2 depletion. To investigate sequence composition across histone H3.2 gene repeats, we examined a recent long-read sequence assembly of the entire HisC locus [36]. Bioinformatic analysis identified 108 copies of the endogenous histone H3.2 gene. All but one copy is predicted to encode identical full-length H3.2 proteins. One gene copy is truncated and predicted to be non-functional; this gene was excluded from further analysis. Alignment of sequences corresponding to the H3.2 3'-UTR revealed 100% sequence identity for 103/107 gene copies, whereas 4/107 gene copies share the same single nucleotide polymorphism (Fig. 1B, Supplemental Fig. 1). Sequence identity across the 3'-UTRs is even greater than that observed for the H3.2 coding sequence, for which 92/107 gene copies are identical and 15/107 have synonymous polymorphisms (Supplemental Fig. 1). The high degree of sequence identity across all H3.2 gene copies makes the 3'-UTR a favorable target for shRNA-mediated depletion. Targeting the 3'-UTR also provides a straightforward means of generating H3.2 transcripts insensitive to shRNA-mediated depletion by swapping the D. melanogaster sequence with the 3'-UTR from the closely related species, D. simulans. Importantly, the sequence of the stem loop, a cis-element located within the 3'-UTR required for histone mRNA processing, is conserved between the two species (Fig. 1B), suggesting that the D. simulans H3.2 3'-UTR may functionally replace the D. melanogaster 3'-UTR. Furthermore, sequence divergence upstream of the stem loop allows for design of shRNA constructs whose products would hybridize perfectly with D. melanogaster histone H3.2 3'-UTRs, leading to their degradation, whereas mismatches with D. simulans 3'-UTRs would render them insensitive to shRNA-mediated degradation, thereby enabling selective depletion of endogenous H3.2 (Fig. 1B).

Fig. 1.

A system for selective depletion of endogenous histone H3.2 transcripts. A Schematics of H3.2 shRNA transgene (top), endogenous histone genes (middle), and transgenic histone genes with D. simulans H3.2 3'-UTR indicated in purple (bottom). B DNA sequence corresponding to the H3.2 3'-UTR from D. melanogaster and D. simulans. Asterisks indicate nucleotide identity across all D. melanogaster H3.2 paralogs. Shading indicates nucleotide identity between D. melanogaster and D. simulans H3.2 3'-UTRs. Horizontal lines indicate target sequences for each H3.2 shRNA. C Stacked bar charts depicting the proportion of larvae ubiquitously expressing the H3.2 shRNA via tubulin-GAL4 vs control siblings that do not express the H3.2 shRNA. The expected proportion is indicated with a dashed line. *p < 0.05, ***p < 0.0001 chi-square goodness of fit. D Bar charts depicting the survival of larvae from panel C to pupal stages (pupariation) and adulthood (eclosion)

We designed four GAL4/UAS-driven shRNA constructs targeting overlapping regions in the D. melanogaster H3.2 3'-UTR (Fig. 1B) and generated independent transgenic lines via site-specific integration. We also generated a histone replacement transgene with 12 tandem repeats encoding wild-type histone proteins in which the sequence corresponding to the D. melanogaster H3.2 3'-UTR was replaced by the D. simulans sequence (hereafter, H3.2^WT−sim). Below, we test whether these elements allow for spatiotemporal control of histone gene expression via selective depletion. We first determine whether these shRNAs can deplete endogenous histone H3.2 expression and generate developmental phenotypes consistent with histone loss-of-function. Next, we examine whether shRNA-insensitive replacement transgenes encoding wild-type histone H3.2 rescue endogenous histone gene deletion and shRNA-induced phenotypes. Last, we test whether shRNA-insensitive replacement transgenes encoding mutant H3.2 proteins result in histone mutant phenotypes upon shRNA expression.

Histone H3.2-shRNA expression causes cell proliferation defects due to reduced H3 protein

Diploid D. melanogaster cells contain approximately 200 histone H3.2 gene copies, and their mRNA and protein gene products are among the most abundant in the cell [37, 38]. Due to this abundance, we first sought to determine whether shRNAs targeting endogenous H3.2 would be capable of generating developmental phenotypes associated with histone depletion. Homozygous deletion of the HisC locus results in embryonic lethality due to cell cycle arrest before the onset of mitosis during the 15th nuclear cycle [13, 39]. Prior to this stage, maternally supplied histone gene products are used to package the newly replicated genome. However, maternal histone mRNAs are degraded at the end of cycle 14 [40], and the absence of zygotically encoded gene products in ΔHisC mutants leads to a reduction in the pool of histone proteins and cell cycle arrest [41]. In addition to H3.2, ΔHisC mutants lack all other replication-dependent histone genes (H4, H2A, H2B, H1). Whereas viability of animals lacking only histone H3.2 has not been reported, clones of cells lacking H3.2 were reported to exhibit severe growth defects [42], presumably due to incomplete nucleosome assembly during DNA replication. To determine the consequences of histone H3.2 shRNA expression during development, we first used a ubiquitous tubulin-GAL4 driver that turns on early in embryogenesis. We found that expression of each of the four shRNAs results in lethality. Few to no embryos ubiquitously expressing shRNA #1 or shRNA #4 hatched into larvae, whereas approximately 30% of embryos expressing shRNA #2 or shRNA #3 hatched (Fig. 1C). However, nearly all shRNA-expressing larvae died, with only a single larva progressing to the pupal stage (Fig. 1D). Importantly, we recovered zero shRNA-expressing adults (Fig. 1D). Thus, expression of each H3.2 shRNA is deleterious for development. To determine the impact of shRNA expression on histone H3 protein levels, we performed immunofluorescence in early embryos using the prd-GAL4 driver, which is active in segmentally repeated stripes in the epidermis (Fig. 2A). Similar to our findings with the tubulin-GAL4 driver, expression of H3.2 shRNA under control of prd-GAL4 causes embryonic lethality (Supplemental Fig. 2A). Total histone H3 levels were evaluated approximately 1.5-h after the onset of prd-GAL4 expression, a time in development when epidermal cells are actively cycling [41, 43]. In contrast to the control genotype, we observed a significant reduction in histone H3 protein in shRNA-expressing cells relative to non-expressing cells for each of the four shRNA lines (Fig. 2B, Supplemental Fig. 2). Image quantification indicated an approximate 22% decrease in histone H3 signal intensity in shRNA-expressing regions relative to control regions for shRNA lines #2, #3, and #4 and a 33% decrease in H3 signal intensity for shRNA #1 (Fig. 2B). To put this finding into perspective, the maximum expected reduction in total H3 protein levels is 50% because only a single S phase occurs after the onset of prd-GAL4 activation [41], and shRNA-mediated depletion is not expected to affect the levels of H3 protein already present in chromatin prior to prd-GAL4 activation. Hence, the observed decrease in total H3 protein levels are 50–66% of the maximum expected reduction. The antibody we employed for immunofluorescence is also expected to recognize the replication-independent H3.3 protein, which may contribute to the total H3 protein levels observed. We also compared DAPI levels in H3.2 shRNA-expressing to non-expressing cells and observed a moderate but significant decrease for shRNA #1 and #3, as well as a noticeable decrease in several individual embryos expressing shRNA #2 and #4 (Fig. 2B). Prior studies of ΔHisC embryos determined that S-phase duration is extended in the absence of all replication-dependent histone genes. Thus, it is possible that the reduced DAPI levels we observe here is a similar consequence of a slower rate of DNA replication in H3.2 shRNA-expressing cells. We conclude that shRNA expression quickly and effectively reduces histone H3.2 protein levels despite the high copy number of H3.2 genes.

Fig. 2.

Expression of H3.2 shRNA causes depletion of H3 protein. A Confocal images of embryos stained for DAPI (left), GFP (middle), and total H3 (right) for no shRNA control (top) and H3.2 shRNA #3-expressing embryos (bottom). Scale bars–100 µm. B Representation of prd-GAL4 expression pattern. Signal was quantified for the regions within the dashed box (a, b, c). For each embryo, the ratio of the signal in GAL4-expressing cells (region b) to GAL4 non-expressing cells (mean of regions a and c) is plotted. The median is indicated by the horizontal line, and individual data points are shown with purple dots. *p < 0.05, ***p < 0.0001 Wilcoxon rank sum test

To examine the consequences of shRNA expression on organ development, we turned to the Drosophila eye, which has been widely used as a model to understand developmental control of cell proliferation [44, 45]. The adult compound eye is derived from a larval tissue known as the eye-antennal imaginal disc, which along with the eye and antenna, produces most of the external structures of the adult head [46]. Development of the eye-antennal imaginal disc is characterized by a proliferative phase in which cells asynchronously progress through the cell cycle, followed by a post-proliferative phase when most of the cells exit the cell cycle to establish a terminal fate. Within the eye, these two phases can be interrogated separately using the ey-GAL4 driver [47], which is active early in eye development during the proliferative phase, and the GMR-GAL4 driver [48], which is active later during the last round of cell division in the eye and in the post-proliferative phase (Fig. 3A).

Fig. 3.

H3.2 shRNA expression causes proliferation defects. A Cartoon of eye-antennal disc development depicting spatiotemporal patterns of ey-GAL4 and GMR-GAL4 expression. Temporal expression patterns are indicated by blue and magenta arrows. Spatial expression patterns are indicated by green shading. B Cartoon of D. melanogaster head (BioRender) alongside scanning electron micrographs of pharate adult heads. Representative images of four phenotypic classes of decreasing severity are presented. The percentage of control and H3.2 shRNA #3-expressing adults belonging to each class is indicated. Scale bars–500 µm. C Confocal images of wandering 3rd instar eye-antennal discs (left). Control eye-antennal disc depicting ey-GAL4 driven GFP expression (top). ey-GAL4 H3.2 shRNA #3- expressing eye-antennal disc (bottom). Stereomicrograph of pharate adult heads from the corresponding genotypes (right). Scale bars–100 µm. D Confocal images of wandering 3rd instar eye-antennal discs (left). Control eye-antennal disc depicting GMR-GAL4 driven GFP expression (top). GMR-GAL4 H3.2 shRNA #3-expressing eye-antennal disc (bottom). Stereomicrograph of pharate adult heads from the corresponding genotypes (middle). Scale bars–100 µm. Scanning electron micrographs of adult eyes from the corresponding genotypes (right). Scale bars–200 µm. E Confocal images of 3rd instar wing imaginal disc clones expressing GFP (left), or GFP and H3.2 shRNA #3 (right). Clones were induced 72–77 h before dissection (“Early expression”) or 24–29 h before dissection (“Late expression”). Scale bars–100 µm. (G) Box plots of clone size for Late expression clones. The median is indicated by the horizontal line, and individual data points are shown with black dots. ***p < 0.0001 Wilcoxon rank sum test

H3.2 shRNA expression under control of ey-GAL4 resulted in a severe reduction in viability. H3.2 shRNA-expressing animals represented no more than 10% of the observed progeny, which is a significant reduction relative to the expected mendelian ratio of 50% (Supplemental Fig. 3A). As described below, lethality of these H3.2 shRNA-expressing animals is likely due to disruption of head development, although we observed that this ey-GAL4 driver is also active in the salivary gland, suggesting that activity outside of the eye field may also contribute to the observed reduction in viability. All surviving adults exhibited defects in eye morphology, indicating that phenotypes caused by H3.2 shRNA expression are fully penetrant (Fig. 3B, C, Supplemental Fig. 3). Closer inspection of pharate adults that failed to eclose from their pupal cases using scanning electron microscopy (SEM) revealed that shRNA expression under control of ey-GAL4 resulted in extensive loss of eye tissue and other head structures, including animals that completely lacked heads (Fig. 3B). Consistent with these observations, shRNA-expressing 3rd instar eye-antennal imaginal discs exhibited highly reduced to absent eye tissue in the most extreme cases (Fig. 3C). Interestingly, we observed increased severity of phenotypes in shRNA-expressing males relative to females. For instance, none of the surviving adults were male (Supplemental Fig. 3A), and pharate adult males exhibited more severe defects in eye and head morphology relative to females (Supplemental Fig. 3B). The cause for the increased phenotypic severity in males is unclear; however, the vector we used for shRNA expression has been reported to have stronger effects in males than in females [49], consistent with our observations.

Loss of eye and head tissue upon shRNA expression during the proliferative stage of eye development suggests that H3.2 protein was sufficiently depleted to cause a proliferation defect. An alternative explanation is that tissue loss was due to cytotoxicity or off-target effects of shRNA expression. To investigate this possibility, we drove shRNA expression in eye imaginal discs using the later-acting GMR-GAL4. Replication-dependent histones are only expressed during S phase [2]. Therefore, we reasoned that if the observed phenotypes were independent of histone H3.2 depletion, then shRNA expression during the post-proliferative phase would also result in tissue loss. Instead, we observed that GMR-GAL4 driven H3.2 shRNA expression resulted in no detectable changes to eye imaginal disc morphology (Fig. 3D). Moreover, adult eyes examined five days after GMR-GAL4 activation appeared normal under conventional dissecting microscopes (Fig. 3D). However, closer examination of shRNA-expressing adults using SEM revealed subtle changes to eye morphology relative to control GMR-GAL4 > UAS-GFP flies, including disruptions to the hexagonal array of ommatidia and missing interommatidial bristles (Fig. 3D, inset). Similar rough eye phenotypes and loss of interommatidial bristles have previously been reported when the GMR promoter was used to drive expression of the cell cycle inhibitor, p21 [50, 51]. Gene expression driven by GMR can disrupt the second mitotic wave, reducing the number of precursor cells available for terminal differentiation in the eye, leading to rough eye phenotypes [50]. Likewise, interommatidial bristles are produced by two cell divisions that take place during pupal stages of eye development, after GMR-GAL4 becomes active [51]. Thus, subtle rough eye phenotypes and interommatidial bristle loss is expected if histone H3.2 shRNA expression disrupts cell proliferation. Together, these findings indicate that loss of tissue upon histone H3.2 shRNA expression is not due to cytotoxicity or off-target effects. Instead, these findings suggest that H3.2 shRNA expression results in a cell proliferation defect due to depletion of histone H3.2 protein.

In a final set of experiments, we examined the consequences of H3.2 shRNA expression with increased temporal and cellular resolution by performing mosaic analysis in developing wing imaginal discs. Clones of cells expressing GFP alone or GFP plus histone H3.2 shRNA were generated via FLP/FRT mediated excision of a transcriptional stop cassette located between the constitutively active Actin 5C promoter and GAL4 coding sequence (i.e. “FLP-out” clones) [52]. Induction of FLP 67–72 h prior to dissection yielded large GFP-positive clones located throughout the wing imaginal disc epithelium in the control genotype (n = 10 wings) (Fig. 3E, “Early expression”), indicating these cells were able to proliferate normally next to their neighboring wild-type cells. By contrast, GFP-positive clones were completely absent from the H3.2 shRNA-expressing genotype (n = 21 wings) (Fig. 3E, “Early expression”), consistent with elimination of shRNA-expressing cells from the wing imaginal disc epithelium caused by a growth disadvantage relative to neighboring wild-type cells [53]. Shortening the duration of shRNA expression by decreasing the length of time between FLP induction and dissection led to the recovery of H3.2 shRNA-expressing clones (Fig. 3E, “Late expression”). Quantification of clone size revealed that shRNA-expressing clones were approximately twofold smaller than control clones (median clone size of 4 cells in control clones versus 2 cells in shRNA-expressing clones) (Fig. 3F), consistent with a proliferation defect caused by depletion of histone H3.2 protein.

We next attempted to compare the growth of control clones and H3.2 shRNA-expressing clones in the same tissue by employing the CoinFLP system, which allows for clonal expression of either GAL4 (shRNA) or LexGAD (control) via mutually exclusive FLP-induced recombination [54]. Distinct fluorescent reporters distinguish GAL4-expressing clones (CD8::RFP) from LexGAD-expressing clones (CD8::GFP). However, comparison of clone sizes between control and shRNA-expressing cells was not possible due to the recovery of very few control LexGAD clones for unknown reasons (Supplemental Fig. 4). Nevertheless, comparison of GAL4-expressing clones from shRNA-containing and shRNA-lacking wing imaginal discs yielded similar results as the FLP-out experiments described above in that shRNA-expressing clones were smaller than shRNA-lacking clones (Supplemental Fig. 4). Closer inspection revealed that shRNA-expressing clones appeared fragmented relative to shRNA-lacking clones (Supplemental Fig. 4, inset), and fragments of fluorescently tagged membrane (CD8::RFP) were often detected at the basal surface of the wing imaginal disc. These observations are consistent with elimination of shRNA-expressing cells from the epithelium due to a growth defect [53]. Altogether, these findings demonstrate that H3.2 shRNA expression leads to a proliferation defect, likely due to reduced H3.2 protein.

The 3'-UTR of D. simulans H3.2 rescues histone locus deletion in D. melanogaster

Replication-dependent histone transcripts are the only non-polyadenylated mRNAs in metazoans [55]. Instead of a poly-A tail, they terminate in a 3' stem loop that is responsible for proper transcript processing, translation, and degradation [56]. As noted above, the histone H3.2 stem loop sequence is identical between D. melanogaster and D. simulans (Fig. 1B). However, sequence composition varies upstream of the stem loop, raising the question whether the D. simulans 3'-UTR would function in D. melanogaster. To test the functionality of the D. simulans H3.2 3'-UTR, we asked whether H3.2^WT−sim could support development in the absence of the endogenous replication-dependent histone genes by generating a transgenic line bearing twelve tandem histone gene repeats integrated at the same genomic locus as the existing H3.2^WT−mel transgene [30]. We found that one copy of the H3.2^WT−sim transgene rescued homozygous HisC deletion. More importantly, H3.2^WT−sim adults eclosed at the expected mendelian ratio, similar to the H3.2^WT−mel control genotype (Fig. 4A). We conclude that the 3'-UTR of D. simulans H3.2 supports normal histone H3.2 expression in D. melanogaster.

Fig. 4.

H3.2^WT−sim replacement transgene rescues H3.2 depletion phenotypes. A Table of viability for H3.2^WT−mel and H3.2^WT−sim replacement transgenes in D. melanogaster. The p values for the chi-square test are shown. B Bar chart depicting the proportion of total adults with ubiquitous H3.2 shRNA expression (tubulin-GAL4) that are female for H3.2^WT−mel and H3.2^WT−sim replacement transgenes. The expected proportion is indicated with a dashed line. C Scanning electron micrographs of pharate adult heads. Scale bars–500 µm. D Stacked bar chart depicting the proportion of pharate adults in four phenotypic classes of head defects. H3.2 shRNA #3 expression driven by ey-GAL4 in panels C and D

The 3'-UTR of D. simulans is insensitive to shRNA-mediated H3.2 depletion in D. melanogaster

We next sought to determine whether mismatches in the D. simulans 3'-UTR sequence render it insensitive to shRNA-mediated H3.2 depletion. First, we crossed H3.2^WT−sim flies bearing a ubiquitous GAL4 driver to each H3.2 shRNA and scored adult viability (Fig. 4B). As before, these experiments were conducted in the presence of the endogenous replication-dependent histone genes. To account for the extra twelve H3.2 gene copies contributed by the histone replacement transgene, we performed a control cross using H3.2^WT−mel. Whereas H3.2^WT−mel crosses produced zero adult progeny for all four shRNAs, crosses with the H3.2^WT−sim replacement transgene produced adult progeny for three out of the four shRNAs (Fig. 4B). No adults were obtained for shRNA #1. It is possible that shRNA #1 has off-target effects or it may deplete H3.2^WT−sim expression in addition to endogenous H3.2. Interestingly, we observed differences in the numbers of shRNA-expressing adult males and females rescued by H3.2^WT−sim. Whereas the number of shRNA-expressing H3.2^WT−sim females matched the number of non-shRNA expressing control H3.2^WT−sim females, fewer than expected shRNA-expressing H3.2^WT−sim males were recovered for each of the four hairpins (Supplemental Fig. 5A). This finding is consistent with the increased severity of H3.2 shRNA-mediated defects observed in ey-GAL4 males lacking a histone replacement transgene, as described above (Supplemental Fig. 3A). Although the basis for this sex difference is unclear, the penetrance of male lethality roughly correlates with the levels of H3 protein depletion observed across the four shRNA constructs (Fig. 2B), suggesting that the severity of the defects may be due to the strength of H3.2 knockdown.

We next tested the ability of the H3.2^WT−sim replacement transgene to rescue shRNA-mediated defects in the developing eye. Similar to our findings with ubiquitous shRNA expression, the H3.2^WT−sim replacement transgene rescued adult viability when H3.2 shRNAs were expressed under control of ey-GAL4, whereas the H3.2^WT−mel replacement transgene did not rescue adult viability (Supplemental Fig. 5B). Moreover, the H3.2^WT−sim replacement transgene fully rescued shRNA-mediated defects in head and eye development (Fig. 4C, D), whereas H3.2^WT−mel replacement flies exhibited the same distribution of head and eye defects caused by ey-GAL4 driven shRNA expression in the absence of a histone replacement transgene (Fig. 3B). Surprisingly, we observed no differences in male and female viability in H3.2^WT−sim flies expressing shRNAs under control of ey-GAL4 (Supplemental Fig. 5C), unlike our findings when ey-GAL4 was used to drive shRNA expression without a histone replacement transgene (Supplemental Fig. 3A). We also observed that the H3.2^WT−sim replacement transgene rescues viability of shRNA #1 driven by ey-GAL4 (Supplemental Fig. 5B), unlike our findings with ubiquitous expression of shRNA #1 (Fig. 4B), indicating that some of the phenotypes caused by this shRNA are cell-type specific.

We next examined the ability of the H3.2^WT−sim replacement transgene to rescue growth of shRNA-expressing clones in developing wing imaginal discs. Similar to results obtained in the absence of a histone replacement transgene, we failed to obtain shRNA-expressing clones in the presence of the H3.2^WT−mel replacement transgene (Fig. 5A, “Early expression”). However, we observed large shRNA-expressing clones in the presence of the insensitive H3.2^WT−sim replacement transgene (Fig. 5A, “Early expression”). As before, we next shortened the duration of time during which the shRNA was expressed to recover clones before they were eliminated from the wing imaginal disc epithelium (Fig. 5A, “Late expression”). Quantification of clone size indicated that shRNA-expressing clones were approximately two-fold larger in the presence of the insensitive H3.2^WT−sim replacement transgene relative to the H3.2^WT−mel replacement transgene (median clone size of 4 cells in H3.2^WT-sim clones versus 2.5 cells in H3.2^WT-mel clones) (Fig. 5B).

Fig. 5.

H3.2^WT−sim replacement transgene rescues cell proliferation and H3.2 protein levels in H3.2 shRNA expressing cells. A Confocal images of 3rd instar wing imaginal discs from H3.2^WT−mel (left) and H3.2^WT−sim (right) replacement transgenes. Clones expressing H3.2 shRNA #3 are marked with GFP and were induced 72–77 h before dissection (“Early expression”) or 24–29 h before dissection (“Late expression”). Scale bars–100 µm. B Box plots of clone size for Late expression clones. The median is indicated by the horizontal line, and individual data points are shown with black dots. ***p < 0.0001 Wilcoxon rank sum test. C Confocal images of embryos stained for DAPI (left), GFP (middle), and total H3 (right) for no shRNA control (top) and H3.2^WT−sim H3.2 shRNA #3-expressing embryos (bottom). Scale bars–100 µm. D Box plots depicting the relative signal in GAL4-expressing cells to non-expressing cells, as depicted in Fig. 2B, for DAPI and H3. The median is indicated by the horizontal line, and individual data points are shown with purple dots. Not significant by Wilcoxon rank sum test

Last, we examined if H3 protein levels were rescued by the H3.2^WT−sim replacement transgene in embryos expressing the shRNA under control of prd-GAL4. In contrast to the depletion of H3 observed in the absence of an insensitive replacement transgene (Fig. 2), H3 protein levels were restored by the H3.2^WT−sim replacement transgene (Fig. 5C, D). Notably, DAPI levels were also rescued, suggesting that the decrease observed in the absence of the H3.2^WT−sim replacement transgene was due to a proliferation defect as opposed to off-target effects of the H3.2 shRNA. In addition to restoring H3 protein levels, the H3.2^WT−sim replacement transgene also rescued the embryonic lethality phenotype caused by H3.2 shRNA expression driven by prd-GAL4 (Supplemental Fig. 2A). Together, these findings indicate that the D. simulans H3.2 3'-UTR is insensitive to shRNA-mediated knockdown, and it supports normal viability and cell proliferation by restoring H3 protein levels despite depletion of endogenous H3.2 expression.

Selective depletion of endogenous H3.2 with a mutant histone replacement transgene results in Polycomb target gene derepression

Histone gene replacement approaches allow for investigation of epigenetic mechanisms in animals by combining genetic deletion of the endogenous replication-dependent histone genes with replacement transgenes encoding mutant histones [29]. To determine whether shRNA-mediated selective depletion of endogenous histone gene expression can be used to study the consequences of histone residue mutation, we generated an shRNA-insensitive histone replacement transgene that encodes a mutant H3.2 protein. We focused on histone H3.2 lysine 27 (H3.2K27), whose methylation is required for silencing of Polycomb target genes. Mutation of H3.2K27 to a non-modifiable arginine residue (H3.2K27R) results in derepression of Polycomb target genes, such as the Hox gene Ultrabithorax (Ubx) in developing wings (Supplemental Fig. 6) [14, 20, 57]. We performed a series of clonal analyses to determine the impact of shRNA expression on Ubx expression in the main epithelium of 3rd instar wing imaginal discs. Consistent with expectations, clones expressing shRNAs targeting Enhancer of zeste (E(z)), the sole methyltransferase responsible for H3K27 methylation in D. melanogaster, exhibited derepression of Ubx, and a decrease in H3K27me3 levels (Fig. 6A). Next, we made clones of cells expressing histone H3.2 shRNAs in the presence of control H3.2^WT−sim replacement transgenes and observed no derepression of Ubx or change in H3K27me3 levels (Fig. 6B). By contrast, H3.2 shRNA-expressing clones in which the insensitive histone replacement transgene encodes mutant H3.2K27 (H3.2^K27R−sim) exhibited strong Ubx derepression and decreased H3K27me3 levels (Fig. 6C). Similarly, the Hox gene Abd-B was also derepressed in shRNA-expressing clones with the H3.2^K27R−sim replacement transgene (Supplemental Fig. 6B). Thus, shRNA-mediated knockdown of endogenous histone H3.2 in the presence shRNA-insensitive H3.2K27R replacement histones reduces the abundance of wild-type histone below the minimum threshold necessary for maintaining repression of Polycomb target genes.

Fig. 6.

Selective depletion of endogenous H3.2 with the H3.2^K27R−sim replacement transgene causes Ubx derepression and adult wing defects. A–C Confocal images of 3rd instar wing imaginal discs stained for H3K27me3 (blue) and Ubx (magenta). Clones expressing shRNA are marked with GFP (green) for: A E(z) shRNA, B H3.2 shRNA #3 with H3.2^WT−sim replacement transgene, and C H3.2 shRNA #3 with H3.2^K27R−sim replacement transgene. D Confocal images of 3rd instar wing imaginal discs with H3.2 shRNA #3 driven by nub-GAL4 stained for DAPI, GFP, Ubx, and H3K27me3. Control H3.2^WT−sim replacement transgene (top), H3.2^K27R−sim replacement transgene (bottom). Arrow marks tracheal tissue that endogenously expresses Ubx. Scale bars–100 µm. E Representative images of adult wings for control animals and genotypes from panel D. Arrow marks tissue corresponding to adult wing

In a further test of this platform’s ability to selectively express mutant histones with high spatiotemporal precision, we expressed histone H3.2 shRNA under control of nubbin-GAL4, which is active in the developing wing imaginal disc pouch [58]. Whereas the control H3.2^WT−sim replacement transgene supported normal H3K27me3 levels and Ubx repression in wing imaginal discs (Fig. 6D, n = 4) and normal adult wing morphology (Fig. 6E, n = 20), the H3.2^K27R−sim replacement transgene resulted in decreased H3K27me3 levels and robust Ubx derepression in 100% of wing imaginal discs (Fig. 6D, n = 9) and severely malformed adult wings (Fig. 6E, n = 10). Ubx derepression was noticeably weaker in the peripheral regions of the wing imaginal disc pouch. Nub gene expression is thought to be activated in these more proximal cells later in wing disc development relative to more distal cells located in the center of the wing pouch [59]. As a result, these peripheral cells likely express H3.2 shRNA under control of nub-GAL4 for a shorter duration, limiting the time and number of cell divisions during which endogenous H3.2 expression can be depleted and replaced with mutant histones. Consistent with this interpretation, H3K27me3 levels are higher in these peripheral, shRNA-expressing, non-Ubx-derepressing cells (Fig. 6D). Altogether, we conclude that sufficient mutant histone is incorporated into chromatin to trigger Polycomb target gene derepression due to shRNA-mediated endogenous histone H3.2 depletion. More generally, these findings demonstrate that selective histone depletion can be used to generate mutant tissues, expanding the set of experimental approaches that can be used to examine histone mutant phenotypes.

Discussion

Technical challenges associated with mutating histone genes in vivo has impeded the study of histone function in animal models. As a result, our understanding of metazoan histone PTM function has been largely inferred from indirect evidence obtained via mutation of the genes encoding chromatin binding and modifying enzymes. We describe here a new experimental approach that relies on selective depletion of endogenous histone gene expression coupled with histone replacement transgenes engineered to be insensitive to shRNA targeting. Expression of shRNAs effectively depletes endogenous histone H3.2. Replacement transgenes encoding insensitive wild-type histone H3.2 fully rescue phenotypes caused by shRNA-mediated depletion, and replacement transgenes encoding insensitive mutant histone H3.2K27 result in phenotypes associated with Polycomb loss of function upon H3.2 shRNA expression. Thus, this selective depletion system can be used to directly test histone residue function in vivo.

Current methods to generate histone mutants in Drosophila rely on homozygous deletion of the histone locus coupled with expression of tailor-made transgenic histones [29]. However, many histone mutations are recessive lethal which limits the developmental contexts in which they can be studied. For instance, H3.2K27 mutants die during embryogenesis [14, 20]. These embryos exhibit mild homeotic transformations and Polycomb target gene derepression because most embryonic cells divide a limited number of times after zygotic transcription of H3.2K27 mutant histones initiates, and maternally deposited wild-type histones suppress mutant phenotypes prior to the onset of zygotic transcription. In such circumstances, genetic mosaic analysis via mitotic recombination has been utilized to overcome these limitations. However, only a relatively small number of homozygous mutant cells that are mixed with wild-type cells can be obtained through mitotic recombination, limiting the types of downstream analyses such as biochemical assays or genomic profiling that can be performed. Selective depletion of endogenous H3.2 through RNAi overcomes this shortcoming due to the availability of GAL4 drivers with a wide range of spatiotemporal expression patterns, thus providing the opportunity to study histone mutant phenotypes at the level of whole tissues and organs.

One major question regarding the efficacy of an RNAi-based system is whether it would suffice to deplete the massive quantities of histone gene products present in cells. Each S phase, millions of H3 proteins are needed to package the newly replicated genome, raising the possibility that shRNA-mediated knockdown would fail to deplete endogenous histone gene products to sufficiently low levels for a phenotypic consequence. A related potential drawback is whether there would be a delay in knockdown while the shRNA accumulated to functional levels. However, we observed significantly reduced histone H3.2 protein levels within one cell cycle of shRNA expression in embryos (Fig. 2). We also observed rapid loss of shRNA-expressing cells from the wing imaginal disc (Fig. 3E), consistent with many prior studies demonstrating elimination of cells that have a growth disadvantage from a proliferating tissue [53]. Collectively, our findings demonstrate that shRNA-mediated knockdown in proliferating tissues can rapidly overcome the high expression levels of histone H3.2, one of the most abundant gene products in cells. We note, however, that this RNAi-based system would not be appropriate for use in post-replicative cells. We also note that the replication-independent histone H3.3 genes are still expressed despite targeted depletion of histone H3.2 in our system. H3.3 and H3.2 can perform overlapping functions [60, 61], meaning that H3.2 mutant phenotypes may become more severe upon mutation of H3.3.

Success of our approach depends on the presence of shared sequence identity such that all histone gene copies can be targeted by the shRNA. Recent assembly of the D. melanogaster histone gene locus enabled us to determine that sequences within the 3'-UTR were identical across all histone H3.2 gene copies. The applicability of this experimental approach to other histone genes would depend on finding similar stretches of high sequence identity across all paralogs. Alternatively, pools of shRNA constructs could be employed to simultaneously knockdown endogenous histone gene expression. Although the feasibility of CRISPR-based editing of all twenty-eight histone H3 alleles was recently achieved in mouse embryonic stem cells [62], an shRNA-mediated approach may also be useful in mammals or other animal models by providing an inducible means of histone gene depletion. If so, selective depletion of endogenous histone gene expression would provide a broadly powerful experimental approach for interrogating histone mutant phenotypes.

Conclusions

We describe here an experimental approach for examining the function of histone post-translational modifications during animal development, building upon the capabilities of existing histone gene replacement methods. We show that transgenic shRNAs can deplete endogenous replication-dependent histone H3.2 despite its exceptional abundance. We demonstrate that this loss of endogenous H3.2 causes defects that are similar to those observed upon mutation of H3.2. Moreover, transgenic wild-type histones engineered to be insensitive to RNAi can rescue the developmental defects and H3 protein depletion caused by shRNA expression. Last, we demonstrate that this experimental platform causes derepression of Polycomb target genes when coupled with an shRNA-insensitive H3.2^K27R transgene. Thus, selective depletion of endogenous histones permits study of the consequences of histone gene mutation during animal development with high spatiotemporal resolution.

Author contributions

OMA and DJM conceived the study and designed the experiments. OMA performed all experiments and data analysis. OMA and MPLJ cloned the constructs used to generate transgenic lines. OMA and DJM wrote the manuscript.

Funding

This work was supported by National Institute of General Medical Sciences grants R35GM128851 to DJM, and R25GM055336 and T32GM135128 to OMA. This work was also supported by a grant to DJM and OMA from the Howard Hughes Medical Institute through the Gilliam Fellows Program.

Data availability

Data and materials will be made available upon request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.