Actin-related protein 6 facilitates proneural protein-induced gene activation for rapid neural differentiation Free

The basic helix-loop-helix (bHLH) proneural proteins are neural-inducing factors that confer neural fate to neuroepithelium or ectodermal cells. They promote neural fate determination, neural type specification and neural cell differentiation both in vivo and in vitro by transcriptional activation of a series of downstream genes (Bertrand et al., 2002; Wapinski et al., 2013). Nucleosome modification by histone acetyltransferase and demethylase and nucleosome relocation by the SWI/SNF remodeling complex are essential for regulation of proneural proteins to activate transcription in vertebrate cells (Koyano-Nakagawa et al., 1999; Lee et al., 2009; Lin et al., 2017; Seo et al., 2005).

Extensive studies in recent years have revealed that histone variant exchange alters the properties of the nucleosome core, thereby influencing chromatin-associated events, such as transcription, genome stability and DNA repair (reviewed by Martire and Banaszynski, 2020). The canonical H2A has two variants, H2A.Z and H2A.X, involved in nucleosome function and the DNA-damage response, respectively. H2A.Z typically constitutes 5-10% of the total H2A (Bonnet et al., 2019; Redon et al., 2002) and is incorporated into a subset of nucleosomes at nonrandom locations throughout the genome. H2A.Z is enriched within a few nucleosomes surrounding the transcriptional start site (TSS) (Barski et al., 2007; Guillemette et al., 2005; Mavrich et al., 2008; Whittle et al., 2008; Zilberman et al., 2008), although the extent of incorporation varies between different genes.

One of the significant functions of H2A.Z is transcriptional activation. H2A.Z incorporation levels largely correlate with gene expression in the metazoan (Mavrich et al., 2008; Whittle et al., 2008). It is also enriched in genes poised for activation in human cells (Hardy et al., 2009; Ku et al., 2012). Mechanistically, H2A.Z is required to recruit RNA polymerase II (Pol II) to the promoters (Adam et al., 2001). The height of the +1 nucleosome barrier to Pol II inversely correlates with enrichment of H2A.Z (Chen et al., 2019; Weber et al., 2014). Besides transcriptional activation, H2A.Z is implicated in transcriptional repression, gene silencing, and heterochromatin formation and spreading (reviewed by Giaimo et al., 2019).

H2A.Z is incorporated into nucleosomes by the highly conserved SWR1 complex. It partially unwraps DNA from the canonical histone core and catalyzes the replacement of an H2A-H2B dimer with an H2A.Z–H2B dimer (Hong et al., 2014; Willhoft et al., 2018). Actin-related protein 6 (Arp6) is a highly conserved component of the SWR1 complex. It physically interacts with the core ATPase of SWR1 and is essential for complex integrity and histone exchange activity (Scacchetti et al., 2020; Wu et al., 2005). Arp6 is a member of the actin superfamily (Muller et al., 2005; Poch and Winsor, 1997), and both Arp6 and nuclear actin are conserved members of the SWR1 complex in yeast and mammals (Mizuguchi et al., 2004; Ruhl et al., 2006).

Drosophila external sensory (ES) organ development is a classical model in which to study neurogenesis initiated by bHLH proneural proteins. Achaete (Ac) and Scute (Sc) were the first members of the highly conserved proneural protein family to be discovered (Villares and Cabrera, 1987). Both proteins are first expressed in proneural stripes to endow neural potential to the cells. Subsequently, high-level Ac and Sc are restricted to single sensory organ precursors (SOPs) to promote precursor determination, selection and differentiation by transcriptional activation of an array of downstream genes. Here, we report the crucial role of Arp6-mediated H2A.Z replacement for rapid SOP differentiation and division downstream of the proneural protein patterning event. We found that Arp6 is required for efficient transcription of Ac- and Sc-target genes in SOPs. Genetic analyses suggest that Arp6 functions downstream of or in parallel with proneural proteins to activate the SOP differentiation program. Transcriptomic analyses showed that both Arp6 and H2A.Z are preferentially required for proneural protein-induced transcription. Loss of H2A.Z did not reduce proneural protein binding to high-affinity E-box sites, suggesting that E-box binding is independent of H2A.Z enhancer incorporation. phyl is a proneural protein target gene that requires Arp6 and H2A.Z for its maximal expression. H2A.Z was incorporated into the nucleosome around the TSS of phyl in the absence of proneural proteins, and recruitment of proneural proteins to the E-box enhances H2A.Z incorporation through Arp6. Database analysis of the H2A.Z replacement at the +1 nucleosome preceding neurogenesis revealed a strong correlation between H2A.Z enrichment and efficient gene expression by proneural proteins. We propose that, in the absence of H2A.Z replacement at the +1 nucleosome, proneural proteins still bind to the cognate target sites with high affinity but initiate transcription at a slower rate, thus retarding precursor differentiation.

We have previously reported that nuclear actin physically interacts with Ac and Sc and promotes Ac-dependent gene expression (Hsiao et al., 2014). Nuclear actin is a component of several chromatin remodeling complexes, including the SWR1 complex (Cao et al., 2016; Klages-Mundt et al., 2018). Interestingly, the SWR1 complex component Arp6 was also listed as the candidate Ac-interacting protein in the same mass spectrometry analysis. To confirm the interaction between Arp6 and proneural proteins, we performed co-immunoprecipitation using the nuclear extract isolated from S2 cells. We showed that Arp6 was associated with Ac (lane 4 in Fig. 1A). The interaction between Arp6 and Ac was relatively specific as Ac either did not associate or barely associated with the other nuclear actin proteins Arp4, Arp5 and Arp8 (Fig. 1A) (reviewed by Klages-Mundt et al., 2018). We further tested the interaction between Arp6 and other proneural proteins. We found that Arp6 also interacted with Sc and Atonal (Ato) but failed to interact with the proneural protein heterodimeric partner Daughterless (Da) (Fig. 1B). Our results show that Arp6 physically interacts with proneural proteins Ac, Sc and Ato.

To study the role of Arp6 in ac- and sc-dependent ES organ development, we generated a loss-of-function insertion allele (Arp6¹), a deletion allele (Arp6^D) and a HA knock-in allele (Arp6^HA) in which the 3XHA epitope was inserted at the end of the open reading frame by the CRISPR/Cas9 technique (see Materials and Methods and Fig. 1C and Fig. S2A). Compared with controls, Arp6 transcripts were greatly reduced in Arp6¹ hemizygous larvae and not detected in Arp6^D hemizygotes (Fig. 1D). Arp6 mRNA level was moderately reduced by 20-40% in Arp6^HA hemizygotes compared with the controls (Fig. S2B). Arp6¹ and Arp6^D hemizygotes were lethal and most of them were arrested by early pupal stages. Arp6^HA hemizygotes were semi-lethal and a few adults eclosed. The lethality of Arp6¹ could be rescued by the ubiquitous expression of Arp6 from a UAS-Arp6 transgene driven by Act5C-Gal4 (Fig. 1E), confirming that Arp6¹ is an Arp6 mutant allele.

To examine ES organ phenotypes, we generated Arp6¹ and Arp6^D mutant clones on the notum. In both Arp6¹ and Arp6^D clones, macrochaetae (yellow arrowheads) and microchaetae (yellow arrows) were reduced in size, exhibiting shorter and thinner shafts and smaller sockets, compared with non-clonal macrochaetae (red arrowheads) and microchaetae (red arrows) (Fig. 1F,G, and shaft length quantification in Fig. 1I). As a control, we also generated clones for yellow (y), which is used to mark the clone. The ES organs in the y clones showed no difference in shaft length compared with the y/+ heterozygous shaft (Fig. 1I). The reduction in ES organ size in Arp6¹ clones was completely rescued by ubiquitous expression of Arp6 driven by tubP-Gal4 (Fig. 1H), indicating that Arp6 is involved in normal ES organ development. The smaller ES organ size did not result from a general defect in cell growth because the clone size of Arp6¹ homozygous mutant was comparable to that of the homozygous wild-type twin spot (Fig. S1A). We found that the spacing between microchaetae was relatively unaffected in most clones, but was noticeably reduced in the large Arp6 mutant clones (Fig. S1B,C). The size of ES organs appeared normal in Arp6^HA hemizygotes (Fig. S2C).

Although the sizes of shafts and sockets were smaller in Arp6¹ and Arp6^D clones, most of them were present, indicating that the fate determination of outer support cells was largely normal. We also examined the inner daughter cells of the ES organs by dissecting notal tissue 24 h after puparium formation (APF). We found that the expression of Prospero (Pros) and Elav, which label sheath cells and neurons, respectively, were present in the Arp6¹ mutant clones (Fig. 1J,K). Careful examination of the numbers of Pros-positive and Elav-positive cells in Arp6¹ clones showed that loss of Arp6 caused little change in the number of sheath cells (3% reduction compared with heterozygous, n=30 in six clones) and a moderate reduction in the number of neurons (30% reduction compared with heterozygous, n=30 in seven clones). Together, these analyses indicate that the fate determination of ES organ daughter cells is mostly unaffected in Arp6 mutants.

We analyzed further how ES organ development is affected in Arp6 mutant clones by examining notal tissue dissected in a series of time points in pupal stages. We first examined the expression of the proneural protein Ac. We found that the pattern and level of Ac in proneural stripes were unaffected by the lack of Arp6 at 9 h APF (red arrows in Fig. 2A), suggesting that the initiation of Ac expression was normal. Whereas Ac-positive SOPs were isolated in non-clonal Arp6¹/+ cells at 10 h APF, Ac expression remained in clusters of cells in Arp6¹ clones (Fig. 2B). Single Ac-positive cells in Arp6¹ clones were observed at 12 h APF when the Ac level in Arp6¹/+ cells was diminished (Fig. 2C). Ac was undetected in both Arp6¹ clonal and Arp6¹/+ cells at 14 h APF (Fig. 2D). These results suggest a delay in the enrichment of Ac in SOP cells. We also examined Sc expression in a Sc-GFP knock-in line in which the C terminus of Sc is fused with the GFP (Chen et al., 2018). We found a similar delay in Sc expression in SOPs (Fig. 2E,F).

The high-level proneural proteins in SOPs induce the expression of senseless (sens), which is a direct target of proneural proteins (Jafar-Nejad et al., 2003). Delays in Sens expression were also observed in the time series of notal tissues co-stained for Ac and Sens (Fig. 2A′-D′). Quantification of Sens levels in SOPs at 14 h APF confirmed that it was significantly decreased by 32% in Arp6¹ clones compared with the neighboring Arp6¹/+ cells (Fig. 2D′; Fig. S1D), indicating that Arp6 is required for maximal Sens expression. SOPs further divide into two daughter cells that could be labeled by Hindsight (Hnt; also known as Peb) and Cut proteins (Blochlinger et al., 1990; Pickup et al., 2002). Although Hnt and Cut expressions were detected in two adjacent daughter cells after SOP division in Arp6¹/+ cells, they were mostly localized in single, isolated cells in Arp6¹ clones (Fig. 2G,H). A delay in later cell division into three- to four-cell clusters was also observed (Fig. 2I). Taken together, the phenotypic analyses of Arp6 mutants demonstrate defects in the timely differentiation and division of SOP cells.

In mutants lacking ac and sc activity, ES organs on the notum are completely absent (Campuzano et al., 1986; Dambly-Chaudiére and Ghysen, 1987). It is unclear whether partially lacking ac and sc would also lead to a delay in ES organ development. We tested this idea by examining ES organ phenotypes for the hypomorphic allele ac^sbm, a P-element insertion mutant affecting both ac and sc expression (Marcellini et al., 2005). In the ac^sbm clones at the adult nota, ES organs were smaller with a noticeable reduction in the shaft length compared with those in the neighboring heterozygous tissue (Fig. 3A,B). Like Arp6¹ mutant clones, upregulation of Sens was retarded in ac^sbm clones (Fig. 3C). Delays in cell division into two cells and further into three- to four-cell clusters or even into four to five-cell clusters were observed (Fig. 3D-F). Although retarded, sensory organ cell differentiation appeared normal because shaft and socket were found in every adult bristle in ac^sbm clones (Fig. 3A). Elav-positive cells were observed accompanying each Su(H)-positive socket cell at 28 h APF (Fig. 3G), indicating correct cell fate determination of neurons. These analyses together demonstrate that reduced proneural protein level retards SOP differentiation and division, but cell fate determination of daughter cells remain largely unaffected, as observed in Arp6 mutants.

Next, we examined the functional interaction between proneural proteins and Arp6. Given that Ac and Sc levels were not reduced in Arp6 mutant clones, we tested whether Arp6 acts downstream or in parallel with proneural proteins in SOPs by rescue experiment. Overexpression of Ac by sca-Gal4 in a stripe pattern failed to restore the normal levels of Sens in SOPs of Arp6¹ clones (Fig. 3K,K′), indicating that the absence of Arp6 could not be replaced by enhanced Ac expression. We next examined the genetic interaction between Arp6 and ac sc by generating ac^sbm Arp6¹ double mutant clones. Whereas Sens was partially reduced in either ac^sbm or Arp6¹ mutants at 18 h APF (Fig. 3D,H), Sens expression was eliminated in the double-mutant clones (Fig. 3I) and the ES organs were almost absent in the double-mutant clone on adult notum (Fig. 3J). Together, these results support the suggestion that Arp6 functions downstream or in parallel with proneural proteins in SOPs to promote ES organ development.

The retarded SOP differentiation and division in Arp6 mutant clones and the synergistic interaction with proneural genes prompted us to examine whether Arp6 regulates the proneural protein-activated transcriptional program. We first performed transcriptomic analysis to identify genes upregulated by proneural Ac-Da heterodimers. For comparison, we used the DNA-binding defective Ac^R29G mutant, in which arginine is replaced by glycine at the critical residue of the basic domain (Ma et al., 1994). We compared gene expression profiles of Ac-Da- and Ac^R29G-Da-transfected S2 cells. Among the 111 genes that were upregulated by at least 1.5-fold (log₂ fold change>0.58) in two independent experiments, scabrous (sca) and phyllopod (phyl) were the top two upregulated genes (Table S1). Both sca and phyl are direct target genes of proneural proteins in SOPs (Mlodzik et al., 1990; Pi et al., 2004). We then performed quantitative RT-PCR to measure the levels of upregulation and showed that phyl expression was increased by more than 200-fold by Ac-Da heterodimers relative to Ac^R29G-Da, and sca was upregulated by 40-fold (Fig. 4A), confirming that phyl and sca are target genes of proneural proteins.

Therefore, we examined whether the expression of phyl and sca could be modulated by Arp6 in SOPs. It has been shown that the 4.1 kb upstream regulatory region of phyl is sufficient to drive its expression in SOPs and rescues phyl mutant phenotypes when fused with phyl cDNA (Pi et al., 2004). The transcriptional reporter phyl^4.1-nGFP was examined in Arp6¹ clones. At 12 h APF, phyl^4.1-nGFP was expressed in strips of cells in Arp6¹/+ cells, and its expression was diminished in Arp6¹ clones (Fig. 4B). Reduced expression of phyl^4.1-nGFP was also detected in single SOPs of Arp6¹ clones at 14 h APF (Fig. 4C). These results indicate that the Ac- and Sc-target gene phyl is regulated by Arp6 at the transcriptional level. We also examined whether sca transcription is regulated by Arp6 in SOPs. The enhancer trap sca-lacZ showed a markedly reduced expression in SOPs in the Arp6¹ clone compared with control Arp6¹/+ cells (Fig. 4D). Taken together, we conclude that Arp6 regulates the SOP genes phyl and sca at the transcriptional level.

Arp6 is a core component of the H2A.Z-exchange SWR1 complex, which is functionally conserved in yeast, fly and human (Ruhl et al., 2006; Scacchetti et al., 2020; Wu et al., 2005). In Arp6¹ clones, H2A.Z was eliminated from the developing notal cells (Fig. S3A). This could result from the degradation of free H2A.Z (Takahashi et al., 2017). Thus, we investigated whether loss of H2A.Z also impairs ES organ development. H2A.Z is encoded by a single copy of His2Av in the Drosophila genome. In His2Av⁸¹⁰ null mutant clones, deformed bristles with tiny shafts and sockets were observed (Fig. 5A), demonstrating that H2A.Z is required for normal ES organ development. Although Ac expression in proneural stripes was largely unaffected (Fig. 5B), the progression of Ac proneural stripe to single SOP expression and diminish of Ac in SOPs were retarded in His2Av⁸¹⁰ clones (Fig. 5C,D). Upregulation of Sens in SOPs, and SOP division to two daughter cells marked by Sens expression were also slowed down in His2Av⁸¹⁰ clones (Fig. 5D′,E). In summary, His2Av⁸¹⁰ clones, although presenting more severe phenotypes, recapitulate what was observed in Arp6 mutant clones. To evaluate further the functional interaction between Arp6 and H2A.Z in ES organ development, we examined the ES organ size by quantification of the shaft length in mutants partially defective for Arp6 and H2A.Z. H2A.Z was knocked down in the central region of the notum by RNA interference (RNAi) via pnr-Gal4. Compared with controls expressing RFP RNAi, knockdown of H2A.Z reduced the shaft length by 14% (Fig. S2D,E). In the heterozygous mutant carrying one copy of Arp6^HA, knockdown of H2A.Z reduced the shaft length by a further 25% compared with the knockdown of RFP (Fig. S2D,E). The synergistic effect between Arp6 and H2A.Z to regulate shaft length suggests functional interaction in ES organ development.

We next examined in more detail whether the Arp6-included SWR1 complex is regulated during SOP differentiation. In the Drosophila SWR1 complex, the catalytic core ATPase is encoded by the domino (dom)-B isoform (Scacchetti et al., 2020). We assessed Dom-B expression in a Dom-B-GFP protein trap line (Börner and Becker, 2016). The GFP protein was maintained at constant levels in proneural stripes and early Ac-positive SOPs compared with the epithelial cells (Fig. 5F-G′). Interestingly, the GFP level was elevated in the large, Hnt-positive SOPs (Fig. 5H,H′), suggesting that SWR1 activity might be upregulated in mature SOPs. Together, our data support that SWR1-dependent H2A.Z incorporation is required for timely SOP development downstream of the proneural patterning event.

Our analysis suggests that Arp6-mediated H2A.Z replacement might be required for the maximal expression of proneural protein target genes. We next assessed the requirement of Arp6 and H2A.Z for the Ac-Da heterodimer-dependent transcription in S2 cells by RNAi. In cells expressing Arp6 RNAi to reduce Arp6 expression by 93% (Fig. S3C), H2A.Z levels were reduced by 37% (Fig. S3B), indicating that H2A.Z incorporation is partially compromised by Arp6 knockdown in S2 cells. As shown in Fig. 4A, the Ac-Da heterodimer markedly induced phyl transcription in S2 cells. phyl expression was decreased by 24% (log₂ fold change=−0.39) compared with control luciferase (Luc) knockdown (Fig. 6A). We also measured expression of the other target gene, sca, and found that the transcription induction was similarly compromised by Arp6 knockdown (Fig. 6A). Thus, the Ac-Da-induced maximal expression of phyl and sca requires the presence of Arp6.

In cells knocked down for H2A.Z by approximately 90% (Fig. S3B), phyl expression was also decreased by 37% (Fig. S3D). Unexpectedly, examination of exogenously expressed Ac and Da levels revealed that they were much higher in H2A.Z-knockdown S2 cells compared with Luc-knockdown cells (Fig. S3E) but remained comparable between Arp6 knockdown and Luc knockdown cells (Fig. S3E). We then adjusted the amounts of the plasmid in H2A.Z knockdown cells such that they expressed comparable levels of Ac and Da as in the Luc knockdown cells (Fig. S2E). Expression of phyl and sca by comparable levels of Ac-Da was markedly decreased in H2A.Z knockdown cells (Fig. 6A).

We next performed transcriptomic analysis to assess the regulation of Ac-Da-induced genes by Arp6 and H2A.Z. We profiled the expression levels for genes induced by Ac-Da heterodimers. In the group of 111 Ac-Da-upregulated genes, most of their expression levels were compromised in both Arp6 knockdown and H2A.Z knockdown cells (Table S1; Fig. 6B). For comparison, we selected a collection of 117 genes expression levels of which were insensitive to Ac-Da heterodimer (expression changes less than 20%) (see Materials and Methods). Arp6 and H2A.Z knockdown caused a significantly greater reduction in the Ac-Da-upregulated genes than the non-Ac-Da-induced genes (Fig. 6C).

The preferential requirement of Arp6 and H2A.Z for Ac-Da-induced gene expression prompted us to examine whether Arp6 and H2A.Z promote proneural protein binding to their specific DNA-binding sites. The Ac-Da heterodimers bind to E-box sequences CAG(C/G)TG with high affinity (Singson et al., 1994). Chromatin immunoprecipitation (ChIP) analysis was performed for the Ac binding to two CAGCTG in the phyl upstream region (phyl E1 and E2) (Pi et al., 2004). The assay confirmed that Ac-Da preferentially bound the two E-box sites compared with the DNA-binding domain mutant Ac^R29G-Da (Fig. 6D). By contrast, no preferential binding was detected on the low-affinity binding site CACGTG (phyl E4) (Pi et al., 2004; Singson et al., 1994) (Fig. 6D), indicating the specificity of our assay. The sca proximal region contains a CAGGTG element (sca E1) bound by proneural proteins in the gel mobility shift assay (Singson et al., 1994). It was associated weakly with Flag-Ac in the ChIP analysis (Fig. 6D). The binding affinity to these sites was not significantly compromised by Arp6 knockdown (Fig. 6E) and was not reduced by H2A.Z knockdown (Fig. 6F), suggesting that the Ac-Da heterodimer binding to the E-box sites is likely independent of H2A.Z enhancer incorporation.

H2A.Z is enriched in the nucleosomes near the TSS, particularly the nucleosome immediately downstream of the TSS. The H2A.Z enrichment at the +1 nucleosome positively correlates with H2A.Z-dependent activation and facilitates Pol II-mediated transcription in Drosophila (Ibarra-Morales et al., 2021; Weber et al., 2014). To assay the H2A.Z incorporation to the proneural protein target gene, ChIP analysis for Myc-H2A.Z was performed in Myc-H2A.Z-transfected S2 cells and H2A.Z incorporation at the +1 nucleosome of phyl was assayed. The H2A.Z ChIP signal was specifically detected in cells transfected with Myc-H2A.Z, but not in mock-treated cells by anti-Myc antibodies from two different sources (Fig. 6G; Fig. S2C), indicating that phyl is intrinsically permissive for H2A.Z replacement. Because proneural proteins associated with Arp6, we next examined whether Ac influences H2A.Z incorporation of phyl. As shown in Fig. 6H, Ac-Da heterodimers elevated H2A.Z incorporation at the +1 nucleosome compared with Ac^R29G-Da, suggesting that recruitment of Ac-Da heterodimers to the E-box promotes robust H2A.Z replacement. The enhanced H2A.Z incorporation by Ac-Da was significantly diminished in Arp6 knockdown cells (Fig. 6H). Knockdown of Arp6 also decreased the H2A.Z occupancy in cells expressing the inactive Ac^R29G-Da (Fig. 6H), indicating that Arp6 is required for both Ac-Da-induced and Ac-Da-independent, basal H2A.Z incorporation of phyl at the +1 nucleosome.

Our analysis of SOP development suggests that Arp6 and H2A.Z are required for phyl transcription, but dispensable for Ac expression in proneural stripes (Figs 2A and 5B). Consistent with the in vivo data, Arp6 and H2A.Z knockdown did not reduce ac expression in S2 cells (Fig. 6A). Two high-affinity E-boxes in the ac proximal promoter were preferentially bound by Ac-Da relative to Ac^R29G-Da in the ChIP assay and the binding affinity was not significantly compromised in the Arp6 knockdown and H2A.Z knockdown (Fig. 6D-G). Compared with phyl, however, H2A.Z occupancy around the +1 nucleosome location of ac was undetected or significantly lower (Fig. 6F; Fig. S4), suggesting that the +1 nucleosome of ac is intrinsically inaccessible to H2A.Z incorporation.

To investigate further whether intrinsically differential H2A.Z replacement influences proneural protein target gene expression, we analyzed the correlation between H2A.Z replacement at the +1 nucleosome and Arp6- and H2A.Z-dependent gene expression. H2A.Z incorporation is initiated around zygotic genome activation preceding Ac- and Sc-dependent neurogenesis (Ibarra-Morales et al., 2021). By searching the database of H2A.Z replacement during zygotic genome activation, 41 Ac-Da-induced genes exhibited H2A.Z replacement at the +1 nucleosome and 62 genes did not. We analyzed the extent of downregulation in H2A.Z-positive genes and H2A.Z-negative genes by Arp6 knockdown. As shown in Fig. 6I, there was no significant difference in the downregulation of expression between H2A.Z-positive and -negative genes in both Ac-Da-induced and non-Ac-Da-induced gene groups. Upon elimination of H2A.Z by H2A.Z knockdown, however, we found that the Ac-Da-induced gene expression was much more decreased in the H2A.Z-positive genes than in the H2A.Z-negative genes (log₂ fold change: −0.86 versus −0.34; P<0.01) (Fig. 6J). We conclude that the marked reduction was specific to Ac-Da-induced expression because the effects on the non-Ac-Da-induced, H2A.Z-positive genes were small (log₂ fold change=−0.15) and there was no significant difference between H2A.Z-positive and -negative genes in non-Ac-Da induced gene expression (Fig. 6J). Taken together with the genetic interaction between Arp6 and ac sc, our data support the model that H2A.Z acts through its replacement at the +1 nucleosome to facilitate efficient onset and maximal activation of transcription upon proneural protein binding to E-box sites, thus promoting rapid differentiation of neural precursors (Fig. 6K).

bHLH proneural proteins are robust inducers of neurogenesis in vivo and in vitro. They act as pioneer factors to recognize cognate E-box sequence and activate target gene expression for neural determination and differentiation. Our data demonstrate that the histone variant H2A.Z is a crucial chromatin factor for proneural protein-induced maximal gene expression: (1) loss of Arp6 and H2A.Z led to prolonged SOP development and smaller ES organs, the phenotypes also observed in ac sc hypomorphic mutants; (2) SOPs and ES organs were almost eliminated in Arp6 and ac sc hypomorphic double mutants; (3) transcription of proneural protein target genes phyl and sca in SOPs were markedly reduced in Arp6 mutants; and (4) transcriptomic analysis revealed that loss of Arp6 and H2A.Z preferentially diminished gene expression induced by Ac-Da compared with the constitutively activated genes. Unlike the histone acetyltransferase CBP and the chromatin remodeler Brg1 in Xenopus and mouse, which are required for the generation of neural progenitors and neurons by proneural proteins (Koyano-Nakagawa et al., 1999; Lee et al., 2009; Seo et al., 2005), we found that Arp6-mediated H2A.Z incorporation specifically influences precursor differentiation rate without affecting fate determination and formation of SOPs. Therefore, histone variant H2A.Z exchange plays a distinct role in the regulation of proneural protein-induced neurogenesis compared with histone modification and nucleosome remodeling.

Our analyses suggest that Arp6-mediated H2A.Z replacement at the +1 nucleosome facilitates proneural protein target gene expression by both proneural protein-dependent and -independent mechanisms. From the S2 cell studies, we found that proneural proteins enhanced H2A.Z incorporation on its target gene phyl and loss of Arp6 diminished the elevated H2A.Z replacement. These results, together with the physical interaction between Arp6 and proneural proteins, suggest that proneural proteins act upstream and can recruit Arp6 to enhance H2A.Z incorporation. We also observed that H2A.Z enrichment at the +1 nucleosome before neurogenesis was highly correlated with the activation of proneural protein target genes by H2A.Z and Arp6 is required for this basal H2A.Z incorporation, at least to phyl. In addition, Arp6 and H2A.Z do not significantly influence E-box binding. These results suggest a parallel action between Arp6 and proneural proteins in the activation of proneural protein target genes. Genetic analyses showed that Ac overexpression failed to rescue timely SOP differentiation in Arp6 mutant clones, indicating that the absence of Arp6 could not be replaced by enhanced Ac expression. Our genetic and biochemical data support the suggestion that Arp6 acts both downstream and in parallel with proneural proteins to facilitate proneural protein-induced gene activation.

Studies in the past two decades have demonstrated that nuclear actin and nuclear ARPs are crucial components of chromatin remodeling and modifying complexes (reviewed by Klages-Mundt et al., 2018). The nuclear actin-Arp4 dimer participates in chromatin complexes, such as SWR1, Ino80 and BAF. We have shown that nuclear actin is important for proneural protein-mediated gene activation in SOPs and S2 cells; knockdown of both Actin 5C and Actin 42A disrupted SOP formation and overexpression of nuclear β-actin increased Ac/Da-dependent transcription in S2 cells (Hsiao et al., 2014). Given that nuclear actin, Arp4 and Arp6 are the core components of SWR1, our results suggest that one of the major functions of nuclear actin family proteins in neurogenesis involves H2A.Z exchange. A recent study also reveals that the β-actin level is particularly important for enhancing chromatin accessibility at the promoter region of genes involved in the regulation of neuron differentiation in mouse cells (Mahmood et al., 2021). Thus, our analyses of nuclear actin and ARPs in SOPs suggest a highly conserved mechanism of the nuclear actin family protein in the modulation of gene expression in neural development.

Proneural proteins are master regulators of ES organ development. Their spatial and temporal expression patterns determine where and when SOP fate determination and differentiation occur. In developing nota, Ac and Sc are expressed in longitudinal proneural stripes at 6-9 h APF in response to patterning cues from Delta-Notch signaling (Couturier et al., 2019). Our results showed that proneural patterning and ac sc expression in cultural cells was largely unaffected by loss of Arp6-mediated H2A.Z incorporation. At later stages, proneural protein is crucial for precursor selection from proneural cells; single SOPs are selected and arranged in regular arrays by several proneural protein-mediated mechanisms: (1) the proneural target gene phyl is expressed at a high level in SOPs to facilitate lateral inhibition by Notch signaling (Pi et al., 2004); (2) proneural protein target gene sca is expressed in proneural clusters and SOPs to mediate long-range precursor spacing; and (3) Notch target genes E(spl)m7 and m8 were activated by proneural proteins in proneural stripes to mediate SOP selection and spacing (Couturier et al., 2019). Following SOP selection, proneural proteins in single SOPs are gradually downregulated via negative feedback regulation by phyl to promote G2-M progression (Chang et al., 2008). SOPs also require proneural protein target gene sens for timely SOP division (Jafar-Nejad et al., 2006). Therefore, insufficient target gene expressions, caused either by reduced H2A.Z replacement in Arp6 and H2A.Z mutants or by low ac sc expression in ac^sbm mutants, lead to retarded SOP formation and division, and, ultimately, smaller ES organs.

Our results showed that phyl and sens are activated by Arp6 in SOPs. Although phyl and sens are also required for binary cell fate determination of SOP daughter cells, cell fate transformation was almost not observed in Arp6 and H2A.Z mutants. There are two possible explanations. First, phyl and sens were expressed in lower but sufficient levels in SOP daughters to sustain normal cell fate determination in Arp6 and H2A.Z mutants. Second, Arp6 and H2A.Z do not modulate phyl and sens expression in SOP daughter cells. We observed that once phyl and Sens reached sufficient levels in mature SOPs in Arp6 mutants, their levels in SOP daughter cells (pIIa and pIIb cells) were not reduced in mutant clones. Combined with the transcriptomic result that H2A.Z is largely dispensable for the expression of non-Ac-Da-inducible genes, we hypothesized that Arp6-mediated H2A.Z incorporation has little to no effect on phyl and sens expression in SOP daughter cells.

The transcriptomic analysis revealed that H2A.Z is preferentially required for Ac-Da-induced gene activation. By contrast, recent studies also showed that H2A.Z is essential for the expression of constitutively activated genes but not for transcriptional activator Zelda-dependent transcription during the zygotic genome activation (Ibarra-Morales et al., 2021). In addition, H2A.Z plays diverse roles in regulating DNA binding by transcriptional factors. H2A.Z could have positive (Murphy et al., 2020), negative (Cole et al., 2021), or no effect (this study) on the DNA binding.

Analyses of H2A.Z incorporation among Ac-induced genes support the hypothesis that differential H2A.Z replacement at the +1 nucleosome contributes to proneural target gene activation by H2A.Z. Several mechanisms have been proposed to mediate differential incorporation by SWR1. The yeast SWR1 complex preferentially binds promoters containing a more extended nucleosome-free region located between the +1 and −1 nucleosomes (Ranjan et al., 2013). DNA methylation prevents H2A.Z incorporation in Arabidopsis (Coleman-Derr and Zilberman, 2012; Zilberman et al., 2008). Histone H3 acetylation contributes to preferential SWR1 recruitment from yeast to human (Hsu et al., 2018; Klein et al., 2018; Ranjan et al., 2013). Therefore, intrinsic chromatin architecture and modification might further determine the responsiveness of proneural target genes to H2A.Z-dependent transcriptional activation.

Although Arp6 is essential for maintenance of H2A.Z levels in developing cells, as shown by the elimination of H2A.Z in a strong loss-of-function Arp6¹ clone, H2A.Z incorporation was only partially compromised in S2 cells by the knockdown of Arp6 (this study) and YL-1, another core component of the SWR1 complex (Weber et al., 2014). We hypothesize that RNAi-mediated silencing of Arp6 and YL-1 is not efficient enough to abolish most SWR1 activity in S2 cells, leading to incomplete elimination of H2A.Z from nucleosomes in knockdown cells.

In mammals and yeast, the catalytic subunits of the SWR1 complex and NuA4 complex are encoded by distinct genes. In Drosophila, recent studies have revealed that the core ATPase of SWR1 complex and NuA4 complex are encoded by different isoforms of dom, thus complicating the previous molecular interpretation of dom mutant phenotypes. Analysis of the SWR1 complex component has revealed that Arp6 is the sole SWR1-specific component from yeast to human (Scacchetti and Becker, 2021). Thus, our Arp6 mutants are an excellent tool with which to verify and further investigate SWR1-specific function in Drosophila.

The following fly stocks were obtained from stock centers: ac^sbm [Bloomington Drosophila Stock Center (BDSC) #36539], His2Av⁸¹⁰ (BDSC #9264), UAS-His2Av RNAi (BDSC #28966), UAS-RFP RNAi (BDSC #35785), sca^A2–6 (sca-lacZ, BDSC #5403), Act5C-Gal4 (BDSC #4414), tubP-Gal4 (BDSC #5138), sca-Gal4 (BDSC #6479), Ubi-mRFP-nls FRT19A (BDSC #31416) and Ubx-FLP tubP-Gal80 FRT19A (BDSC #42731). sc-GFP was a gift from Dr Xi (Chen et al., 2018). dom-B-GFP was a gift from Dr Becker (Börner and Becker, 2016). phyl^4.1-nGFP has been described by Pi et al. (2004). UAS-3XHA-Arp6 was generated in this study (see below for plasmid construction). Transgenic lines were generated by standard P-element mediated transformation. All fly stocks were raised at 18°C or 25°C.

Negatively marked mutant clones were generated using the FLP/FRT technique (Xu and Rubin, 1993). For Arp6 mutant clones generated by hs-FLP, larvae were heat-shocked twice at 37°C for 1 h at the first and the second instar and returned to 25°C until dissection at the indicated time. For ac^sbm clones and Arp6 mutant clones generated by ubiquitously expressed ubi-FLP, larvae were raised at 25°C until dissection.

The UAS plasmids for Arps (Arp4, 5, 6 and 8), proneural proteins (Sc and Ato) and H2A.Z were constructed by cloning the open reading frames (ORFs) into Drosophila Gateway vectors pTFW or pTMW. The Ac^R29G substitution (codon change from AGA to GGA) was generated using the QuikChange site-directed mutagenesis kit (Stratagene).

The Arp6¹ mutant allele was generated by gRNA-1-guided DNA break followed by homology-directed repair (HDR) (see Fig. 1C). HDR donor plasmids harbor a DNA cassette containing the visible marker 3XP3-RFP and SV40 terminator. Homology sequences of Arp6 (∼500 bp in length) flank each side of the marker DNA cassette. We obtained an insertion allele Arp6¹ in which the 3XP3 marker cassette was inserted at the 11th residue of Arp6. For the Arp6^D allele, the whole ORF was deleted via gRNA-1 and gRNA-2 (see Fig. 1C). The HA knock-in allele Arp6^HA was generated by gRNA-2 guided DNA break followed by HDR via the donor plasmid containing the in-framed 3XHA tag at the 3′end of the ORF (see Fig. S2A). The gRNAs and HDR donor plasmids used for microinjection were purified using the Plasmid Midiprep kit (Invitrogen). The Arp6¹, Arp6^D and Arp6^HA alleles were validated by genomic PCR followed by Sanger sequencing.

The rescue experiment was performed by a cross between Arp6¹/FM7C; Act5C-Gal4 adult females and UAS-3HA-Arp6 adult males. The numbers of progeny in each genotype group were examined.

Pupal nota were dissected in PBS, fixed in 4% paraformaldehyde, and permeabilized in 0.1% Triton X-100 in PBS (PBST). Pupal nota were incubated with primary antibodies overnight at 4°C and secondary antibodies for 2 h at room temperature. The following primary antibodies were used: mouse anti-Achaete [1:5, Developmental Studies Hybridoma Bank (DSHB)], guinea pig anti-Sens (1:2000, gift from Dr Bellen, Baylor College of Medicine, Houston, TX, USA), mouse anti-Pros MR1A (1:10, DSHB), mouse anti-Elav 9F8A9 (1:10, DSHB), mouse anti-Hnt 1G9 (1:5, DSHB), mouse anti-Cut 2B10 (1:25, DSHB), rabbit anti-Drosophila H2A.Z (1:2500, a gift from Dr Becker, Ludwig-Maximilians-University, Munich, Germany), mouse anti-FLAG M2 (1:1000, Sigma-Aldrich, F3165) and rabbit anti-GFP (1:2000, Thermo Fisher Scientific, A11122). All immunofluorescence images were obtained with a Zeiss LSM 780 microscope.

Drosophila S2 cells were maintained in Schneider's Drosophila medium (Invitrogen) supplemented with 10% heat-inactivated fetal bovine serum (Gibco BRL). S2 cells were from the Drosophila Genomics Resource Center. UAS-based expression plasmids and the driver, pWA-GAL4, were transfected using TransIT-Insect transfection reagent (Mirus Bio) for gene expression. The transfected cells were harvested 48 h after transfection for Arp6 knockdown assays and 72 h after transfection for H2A.Z knockdown assays.

Primers for dsRNA templates of Arp6, H2A.Z and firefly luciferase were designed using SnapDragon (https://www.flyrnai.org/snapdragon). For the myc-H2A.Z ChIP assay, the dsRNA template for H2A.Z was designed at the H2A.Z 5′ and 3′ UTRs. Templates for in vitro transcription were generated by PCR amplification using KOD FX Polymerase (Toyobo). dsRNAs were synthesized in vitro using the T7 RiboMAX Express RNAi System (Promega) according to the manufacturer's instructions. For Arp6 knockdown, cells were treated with 30 µg (12-well plate) or 300 µg (10-cm dish) dsRNA for 4 days. For H2A.Z knockdown, cells were treated with 5 µg (12-well plate) or 50 µg (10-cm dish) dsRNA for 6 days.

Total RNA from Drosophila S2 cells was harvested using TRIzol reagent (Invitrogen) and purified as per the manufacturer's instructions. cDNA was synthesized from ∼2-3 µg of total RNA using random hexamers and MMLV reverse transcriptase (Thermo Fisher Scientific). qPCR was performed with 2X SYBR Green Master Mix (Bio-Rad) in a Bio-Rad iQ5 gradient Real-Time PCR system. Gapdh2 was the reference gene for expression quantification in all RT-PCR experiments. Results are shown as mean±s.e.m. of at least three biologically independent experiments. Sequences of the primers are listed in Table S2.

For next-generation sequencing analysis, total RNA was extracted from S2 cells and 100 ng purified mRNA was used for cDNA library construction by NEBNext^® UltraTM RNA Library Prep Kit. Raw sequencing data were analyzed by STAR (v2.7.3a) two-pass mapping strategy to align the raw FASTQ reads against the Drosophila melanogaster reference genome (BDGP6.32) downloaded from the Ensembl database. Ac-Da-induced genes were defined as genes with induction fold change ≥1.5 by Ac-Da relative to Ac-Da^R29G. Non-Ac-Da-induced genes were genes with fold change between 0.8 and 1.2 by Ac-Da vs Ac^R29G-Da and expressed at 15-100 reads per kilobase of transcript per million reads mapped (RPKM), the levels at which Ac-Da-induced genes are usually expressed.

Cells were washed twice with ice-cold PBS and fractioned into cytosolic and nuclear extracts as described (Hsiao et al., 2014). Protein extracts were diluted in mRIPA buffer [50 mM Tris-HCl (pH 7.8), 150 mM NaCl, 5 mM EDTA (pH 8.0), 0.5% Triton X-100, 0.5% NP-40] supplemented with cOmplete EDTA-free Protease Inhibitor (Roche). For immunoprecipitation, 20 µl FLAG-M2 affinity gel (Sigma-Aldrich, A2220) was incubated with 2 mg of lysate in 1 ml mRIPA buffer for 2 h. Beads were collected by centrifugation (2000 g for 3 min at 4°C) and washed five times with mRIPA buffer. The associated protein complexes were analyzed by SDS-PAGE followed by western blotting. The primary antibodies for western blot were: mouse anti-Drosophila H2A.Z (1:1000, Active Motif, 61752), mouse anti-c-Myc 9E10 (1:5000, Santa Cruz Biotechnology, sc-40), mouse anti-FLAG M2 (1:10,000, Sigma-Aldrich, F3165) and mouse anti-α-Tubulin (1:25,000, Sigma-Aldrich, T9026).

The ChIP assay was performed as described (Chen et al., 2013; Riddle et al., 2011; Tettey et al., 2019) with the following modifications. S2 cells were cross-linked by adding formaldehyde (final concentration, 1%) directly to the culture for 10 min on ice. Freshly prepared 2.5 M glycine was added to stop the fixation at a final concentration of 0.125 M and incubated for 10 min on ice. Cells were pelleted at 500 g for 10 min at 4°C and washed twice with ice-cold PBS, followed by the preparation of cytosolic and nuclear protein extracts. Nuclei were suspended in RIPA buffer [50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 5 mM EDTA (pH 8.0), 0.5% sodium deoxycholate, 0.1% SDS, 1% NP-40, 1× cOmplete EDTA-free Protease Inhibitor] for 30 min. Chromatin samples were sheared for ten cycles of 30 s ON/OFF with the Bioruptor^® Plus sonication device to obtain 200-250 bp chromatin fragments, and insoluble chromatin was removed by centrifugation at 13,500 rpm (12,000 g) for 30 min at 4°C. An equal protein quantity (1 mg/ChIP) of soluble chromatin was pre-cleared by incubation with 15 µl Dynabeads™ protein G (Thermo Fisher Scientific, 10004D) for 1 h at 4°C. We added mouse anti-c-Myc 9E10 (3.5 µg/ChIP, Santa Cruz Biotechnology, sc-40), mouse anti-c-Myc 9E10 (3.5 µg/ChIP, Abcam, ab32), mouse anti-FLAG M2 (2 µg/ChIP, Sigma-Aldrich, F3165) or control normal mouse IgG (Millipore, CS200621); 30 µl protein G-coupled magnetic beads were added to each immunoprecipitation followed by rotation for 3 h at 4°C. After extensive washes, the immuno-complexes were treated with proteinase K and heated for de-crosslinking. Bound DNA in the ChIP and input DNA was extracted, purified, and subjected to quantitative real-time PCR analysis. Primer sequences are shown in Table S2.

The shaft length of the ES organ was analyzed by measuring the distance of the shaft from the tip to the socket by ImageJ. In Figs 1I and 3B, ten ES organs in each notum were measured. In Fig. S2B, all ES organs in the dorsal-central region of five to eight nota were examined.

The number of positive cells in the RFP-negative, Arp6¹ area was considered to be the cell number in the mutant clones. For comparison, the RFP-negative, Arp6¹ mutant area was first outlined, and the outlined area was superimposed on the neighboring heterozygous tissue. The cell numbers in the superimposed area were counted as the numbers in the controls. The ratio was the numbers in Arp6¹ clones normalized to those in the controls.

The Sens intensity in SOPs was calculated as the fluorescence intensity of anti-Sens staining in SOPs subtracted by the background fluorescence intensity in the neighboring area. The intensity of Sens in Arp6¹ SOPs was normalized to the mean intensity of control SOPs in the neighboring heterozygous tissue.

mRNA levels of phyl, sca and endogenous ac were normalized to Gapdh in the qRT-PCR analysis (Figs 4A and 6A). In Fig. 6C, each dot is the average of the two independent RNA-seq analyses.

The binding percentage (% input) is shown as the percentage of DNA bound by the specific antibody subtracted by that bound by IgG.

Statistical analyses were performed using one-way ANOVA with Tukey's multiple comparison test for Figs 1I and 6C-J, Figs S2E and S4, and unpaired two-sample t-tests in Fig. 3B and Fig. S1D. Data were using by GraphPad Prism and Microsoft Excel. Statistical significance was defined as P<0.05. Data are represented as mean and s.e.m., except for Figs 1I and 3B in which data are represented as mean and s.d.

We thank Dr Cheng-Ting Chien for his critical reading of the manuscript. We thank Dr Peter B Becker for sharing the anti-His2Av (H2A.Z) antibody. The authors also thank the Microscopy Center at Chang Gung University for technical assistance.