CLAMP regulates zygotic genome activation in Drosophila embryos

Abstract

Embryonic patterning is critically dependent on zygotic genome activation (ZGA). In Drosophila melanogaster embryos, the pioneer factor Zelda directs ZGA, possibly in conjunction with other factors. Here, we have explored the novel involvement of Chromatin-Linked Adapter for MSL Proteins (CLAMP) during ZGA. CLAMP binds thousands of sites genome-wide throughout early embryogenesis. Interestingly, CLAMP relocates to target promoter sequences across the genome when ZGA is initiated. Although there is a considerable overlap between CLAMP and Zelda binding sites, the proteins display distinct temporal dynamics. To assess whether CLAMP occupancy affects gene expression, we analyzed transcriptomes of embryos zygotically compromised for either clamp or zelda and found that transcript levels of many zygotically activated genes are similarly affected. Importantly, compromising either clamp or zelda disrupted the expression of critical segmentation and sex determination genes bound by CLAMP (and Zelda). Furthermore, clamp knockdown embryos recapitulate other phenotypes observed in Zelda-depleted embryos, including nuclear division defects, centrosome aberrations, and a disorganized actomyosin network. Based on these data, we propose that CLAMP acts in concert with Zelda to regulate early zygotic transcription.

Introduction

In most organisms, maternally deposited proteins and RNAs control early pattern formation while the zygotic genome remains dormant. As maternal determinants begin to be degraded, zygotic transcription is gradually upregulated during the maternal to zygotic transition (MZT), a developmental process conserved from insects to mammals (Tadros and Lipshitz 2009; Vastenhouw et al. 2019). This transition depends upon a select group of uniformly distributed transcriptional regulators that act genome-wide to drive zygotic gene expression across the embryo. Consistent with their broad influence, these proteins are thought to sculpt the burgeoning chromatin landscape of an embryo. This gradual transformation is ultimately responsible for reprogramming embryonic cells to generate specific tissues that are vital throughout the entire life cycle.

Following fertilization, embryonic development in Drosophila begins with a series of rapid synchronous nuclear division cycles (NC) in a shared cytoplasm. Initially, these divisions take place in the center of the embryo; however, at NC4, nuclei begin to expand along the anterior-posterior axis (Zalokar and Erk 1976; Deneke et al. 2019). And from NC7-9, nuclei progressively migrate toward the cortex of the embryo. Once they reach the cortex, they undergo four more division cycles before cellularizing during NC14 (Baker et al. 1993; Deneke et al. 2019). Unlike the earlier division cycles which last only about 8–10 minutes, the duration of cycles at the surface of the embryo gradually increases such that NC13 takes about 25 minutes while NC14 is asynchronous and lasts for more than an hour (Farrell and O’Farrell 2014).

During the early division cycles, development transitions from being primarily governed by maternally deposited RNAs and proteins to being controlled by gene products synthesized from the zygotic genome. Zygotic transcription commences around NC8 when a small collection of genes is transcribed. From this point through NC13, there is a gradual increase in transcriptional activity until the genome becomes fully activated in NC14 with the major onset of transcription. Zygotic genome activation (ZGA) thus constitutes a crucial component of the MZT. The MZT also involves a mass degradation of maternally deposited transcripts which is synchronized with ZGA, resulting in a smooth transition from maternal control to zygotically driven development (Farrell and O’Farrell 2014; Hamm and Harrison 2018; Vastenhouw et al. 2019).

The process of genome activation is thought to require de novo construction of a chromatin organization that permits spatiotemporally regulated transcription (Hug and Vaquerizas 2018; Xu and Xie 2018; Steensel and Furlong 2019). The available evidence has led to the view that the 3D architecture of chromosomes in the early zygote is vastly different from that later in development and that many key proteins associated with chromosomes later in development are largely missing during the early nuclear division cycles (Hamm and Harrison 2018; Hug and Vaquerizas 2018; Lefebvre and Lécuyer 2018; Weide and de Wit 2019).

In Drosophila, one protein has been implicated in ZGA and is involved in the establishment of embryonic chromatin architecture: the C₂-H₂ zinc finger protein Zelda (Zld) (Bosch et al. 2006; Staudt et al. 2006; Liang et al. 2008; Harrison et al. 2010, 2011; Nien et al. 2011; Hug et al. 2017). Zld recognizes a heptamer motif, called TAGteam, which was originally identified in the promoter regions of three sex determination genes, scute, sis-A, and Sex lethal (Sxl) as well as the dorsal-ventral (DV) polarity gene zerknullt (zen). The TAGteam motif has since been found genome-wide at many Zld-bound zygotic targets (Erickson and Cline 1998; Bosch et al. 2006; Nien et al. 2011). zld mRNAs are deposited in the egg during oogenesis and translated into early embryos even before the expression of these genes is activated (Liang et al. 2008; Nien et al. 2011). Subsequent studies showed that Zld binds more than 2000 sites located near promoters or other regulatory regions in early embryos (Harrison et al. 2011; Nien et al. 2011). Moreover, over 500 genes were identified which had reduced expression during the late nuclear division cycles in zld^- embryos (Schulz et al. 2015). For a substantial subset of these genes, Zld protein was bound in promoter regions or nearby regulatory elements, lending credence to the idea that Zld has a direct role in facilitating their transcription (Nien et al. 2011). Consistent with this idea, binding of Twist, Dorsal, and Bicoid (Bcd) to their target enhancers depends upon the presence of nearby Zld binding sites (Yáñez-Cuna et al. 2012; Foo et al. 2014; Xu et al. 2014; Sun et al. 2015). This was further confirmed by genome-wide analysis of histone H3 occupancy and chromatin accessibility in embryos from zld germline clone mothers. These studies showed that when Zld was depleted, H3 occupancy increased and general accessibility decreased in the immediate vicinity of Zld binding sites (Li et al. 2014; Schulz et al. 2015).

In their studies on chromatin accessibility in 2–3 hours embryos, Schulz et al. found that Zld was associated with about 40% of the ∼5000 sites classified as accessible (Schulz et al. 2015). Of these Zld associated sites, only about 20% had reduced accessibility after Zld depletion. One plausible explanation for why Zld depletion only affects a small subset of the total accessible sites in 2–3 hours embryos is that, in most cases, other factors also contribute to accessibility, and these factors are able to compensate for the loss of Zld protein. Schulz et al. also found that many Zld bound sites that remained accessible after Zld depletion are enriched in GA_N sequences (see also Moshe and Kaplan 2017). Thus, maternally derived DNA binding proteins with GA_N target sequences could help generate accessible regions of chromatin during ZGA. One of these is the GAGA factor (Trl or GAF) which has been implicated in the formation and maintenance of nucleosome-free regions spanning the promoter regions of fly heat shock genes (Glaser et al. 1990; Lu et al. 1993; Tsukiyama et al. 1994). More recent genome-wide studies indicate that GAF is required to maintain chromatin accessibility—not only for many different promoters, but also for other classes of regulatory elements (Fuda et al. 2015). Importantly, like Zld, GAF has also been shown to be required for proper gene expression during ZGA (Bhat et al. 1996; Lagha et al. 2013; Gaskill et al. 2021).

Another potential candidate that binds GA_N motifs is the zinc finger DNA binding protein, Chromatin-Linked Adaptor for MSL Protein (CLAMP) (Kuzu et al. 2016). CLAMP was first identified for its role in X-chromosome dosage compensation in males (Soruco et al. 2013). CLAMP binds to special cis-acting elements on the X-chromosome called chromatin entry sites and helps recruit the Male-Specific Lethal complex (MSLc), which associates with actively transcribed genes on the male X chromosome to upregulate their expression. This process is necessary in males but deleterious to females. Importantly, CLAMP is an essential transcription factor and, unlike components of the MSLc, is required for viability in both sexes (Urban et al. 2017a). ChIP-seq experiments show that CLAMP binds to thousands of sites throughout the genome, which are distributed across not only on the X chromosome but also the autosomes (Rieder et al. 2017). Moreover, though both GAF and CLAMP bind GA-rich repeats, they overlap at only a subset of sites, with CLAMP having more independent targets (Urban et al. 2017c). Amongst the CLAMP targets is the histone locus, where it plays a critical role in the transcription of histone genes (Rieder et al. 2017). Like Zld and GAF, CLAMP binding enhances chromatin accessibility, both on the X chromosome and autosomes (Urban et al. 2017b). Furthermore, suggesting a role for CLAMP in ZGA, thousands of genes are misregulated in embryos maternally compromised for clamp (Rieder et al. 2017).

Given the widespread binding of CLAMP and its similarities to GAF and Zld, we have investigated the possible involvement of CLAMP in ZGA. Using zygotic-specific knockdown, we show that loss of CLAMP results in similar phenotypes to those observed for the two known ZGA regulators Zld and GAF, including the disruption of zygotic transcription, aberrant nuclear migration, asynchronous mitotic cycles, and disorganized cytoskeletal elements. Furthermore, global transcription profiles during early development of clamp knockdown embryos show that expression of many genes activated in both the minor and major waves of ZGA are affected by partially compromising clamp. Taken together, our data demonstrate that CLAMP functions as a general regulator of ZGA and cooperates with the pioneer factor Zelda.

Materials and methods

Fly stocks

The following Drosophila melanogaster stocks were used: w¹, maternal-tubulin-Gal4 line 67.15, maternal triple driver-Gal4 (31777), egfp RNAi (41552), clamp RNAi (27080 for clampi¹ and a combination of 57163 and 57008 for clampi^2x), zld RNAi (42016), and bcd RNAi (33886 for bcdi¹ and 35478 for bcdi²) from Bloomington Stock Center. These stocks are also detailed in the Reagent Table.

Single molecule fluorescence in situ hybridization (smFISH) and immunostaining

We collected embryos by mating virgin mat-tubulin-Gal4 females with males carrying RNAi transgenes (indicated in fly stocks, above). Embryos were dechorionated with bleach, then fixed in one of two ways: in 4% formaldehyde in PBS with an equal volume of heptane for 20 minutes or heat fixed in a saline solution (NaCl, Triton X-100). Subsequently, embryos were vortexed for 1 minute in a vial containing an equal volume of heptane and methanol to remove the vitelline membrane. They were then rinsed and stored in methanol at −20°C until being used for smFISH or immunostaining. All experiments were performed with at least three independent biological replicates.

smFISH was performed as previously described (Little and Gregor, 2018) . osk (Little et al., 2015) probes were coupled to NHS-Ester 633 (Sigma) and purified using HPLC . eve FISH probes were a gift from Shawn Little, and Sxl, run, and btd FISH probes were a gift from Thomas Gregor and Sergey Ryabichko. Samples were mounted using Aqua Poly/mount (Polysciences) on slides and imaged as described below.

Antibodies used are as follows (and are also detailed in the Reagent Table): mouse anti-Sxl (M18), mouse anti-Even-skipped (3C10), mouse anti-Peanut (4C9H4), and mouse anti-Neurotactin (BP 106) from Developmental Studies Hybridoma Bank, each used 1:10; rabbit anti-phospho-histone H3 (Ser10) used 1:500 from Millipore-Sigma (06-570); rabbit anti-CLAMP used 1:1000 (Novus/SDIX) (Larschan et al. 2012); rat anti-Bicoid used 1:1000, guinea pig anti-Giant used 1:1000, and rat anti-Buttonhead used 1:500 were gifts from Eric Wieschaus; rabbit anti-Centrosomin used 1:500 was a gift from Thom Kaufmann. Secondary antibodies used 1:500 for fluorescent visualization were goat anti-mouse IgG Alexa 488, goat anti-rabbit Alexa 546 or 647, goat anti-rat Alexa 546, and goat anti-guinea pig Alexa 647 (Molecular Probes). Fluorescently labeled embryos were co-labeled using Hoescht (3 µg/mL, Invitrogen). For DAB staining, a standard immunohistochemical protocol was used as described previously, employing a secondary goat anti-peroxidase antibody (Jackson Immunoresearch Laboratories 115-035-166) (Deshpande et al. 1999). All samples were mounted using Aqua Poly/mount (Polysciences) on slides, and then were imaged as described below.

Microscopy and image processing

Confocal imaging for all samples was performed on a Nikon A1 inverted laser-scanning confocal microscope. Unless noted in the figure legend, all images were of single sections, not maximum intensity projections from Z stacks through the embryonic cortex. Images were assembled using Fiji (ImageJ, NIH) and Adobe Photoshop software to crop regions of interest, adjust brightness and contrast, and separate or merge channels.

Statistical analysis

CLAMP protein was measured using mean fluorescence intensity of 200x200 pixel squares from cortexes of individual embryos, and significance was tested using a One-Way ANOVA with a Student’s t-test for pairwise comparisons.

For Sxl smFISH experiments, total number of blastoderm nuclei and total transcription puncta were counted using Fiji (ImageJ). The proportions of Sxl foci/nucleus were then analyzed using a one-way ANOVA with a Student’s t-test for pairwise comparisons.

For experiments in which we categorized defects in embryos by classes, we used pairwise comparisons of the proportion of embryos in each class (as defined in text/figure legends and examples shown in Supplementary Figure S1) using Fisher’s exact test. Likewise, Sxl protein levels were analyzed similarly, though classes were based on the amount of DAB signal. For all patterning genes examined, defects in embryos NC13/14 were classified compared to their corresponding wild-type pattern at the same stage.

Overlap in the target genes decreased at least twofold in knockdown embryos was analyzed using hypergeometric tests for pairwise comparisons and Cochran–Mantel–Haenszel tests for comparisons of three or more datasets.

Data were plotted and statistical analyses were performed using Microsoft Excel and R Project software.

ChIP-seq analysis

ChIP-seq data for CLAMP (Rieder et al. 2017) and Zelda (Harrison et al. 2011) were trimmed using Cutadapt (Martin 2011), mapped to the Drosophila genome (dmel_r6.08) using Bowtie2 (Langmead and Salzberg 2012), converted to a bam file using SAMtools (Li et al. 2009), and peaks were called using MACS 1.4 (Zhang et al. 2008) with the parameters “-f BAM -g dm -bw 300 -w -S.” Promoter and TSS localization and functional enrichment analysis were carried out using ChIPseeker (Yu et al. 2015).

Embryo RNA-seq and analysis

Embryos from the crosses described above for 0–1.5 hours and 1.5–3 hours were collected and dechorionated. Total RNA was extracted using phenol-chloroform, and then cleaned using the Qiagen RNeasy Mini Kit. Three separate biological replicates were performed for each time point and genotype. Libraries were prepared by the University of Minnesota Center for Genomics.

2 × 50 bp FastQ paired-end reads (n = 31.2 Million average per sample) were trimmed using Trimmomatic (v 0.33) enabled with the optional “-q” option; 3 bp sliding-window trimming from 3′ end requiring minimum Q30.

Quality control on raw sequence data for each sample were performed with FastQC. Read mapping was performed via Hisat2 (v2.1.0) using the D. melanogaster genome (BDGP6) as reference. Gene quantification was done via Feature Counts for raw read counts. Differentially expressed genes were identified using the edgeR (negative binomial) feature in CLCGWB (Qiagen, Redwood City, CA, USA) using raw read counts. We filtered the generated list based on a minimum 2x Absolute Fold Change and FDR (Benjamini-Hochberg) corrected P < 0.05.

Data availability

The authors affirm that all data necessary for confirming the conclusions of this article are represented within the article and its tables and figures (including supplemental material found at figshare), except for the RNA-seq data that are deposited and accessible at BioProject ID PRJNA69385 at NCBI Sequence Read Archive. Supplementary material is available at figshare: https://doi.org/10.25386/genetics.14823885.

Results

Zygotic-specific knockdown to study CLAMP function during early embryogenesis

Though clamp is an essential gene, the lethal effects of null alleles are delayed until after embryogenesis is complete (Urban et al. 2017a). Since clamp mRNA is deposited in the egg during oogenesis, it is thought that the mutants survive until the larval stage because there are sufficient amounts of maternally derived clamp to sustain embryonic development. Consistent with this idea, when clamp expression is knocked down during oogenesis, the resulting progeny do not complete embryogenesis and exhibit a range of patterning defects (Rieder et al. 2017; Duan et al. 2020). Accompanying these patterning defects are global alterations in the accumulation of mRNAs during the pre-cellular blastoderm stages of embryogenesis (Rieder et al. 2017). These findings, together with more recent studies showing that maternal knockdown of clamp disrupts the formation of accessible regions of chromatin in the early embryo (Duan et al. 2020) provide support for the idea that CLAMP has a central role in ZGA.

There is, however, a potential complication with depleting CLAMP during oogenesis: if clamp is required not only for embryogenesis but also for normal oogenesis, then some of the developmental abnormalities of pre-cellular blastoderm embryos might be indirect consequences of disruptions in egg development rather than due to the loss of CLAMP activity specifically in the zygote. This possible complication is supported by the finding that clamp is required for the formation of the histone body locus. When clamp is knocked down by RNAi during oogenesis, histone mRNA levels in early embryos are substantially reduced (Rieder et al. 2017). As earlier studies have shown that depletion of histone mRNAs in yeast causes cell cycle defects and ectopic gene expression (Han et al. 1987; Han and Grunstein 1988), it is possible that insufficient levels of maternally deposited histone mRNAs could have similar effects during the rapid nuclear divisions in fly embryos.

For this reason, we decided to knock down clamp expression specifically in the zygote by mating mothers carrying the GAL4 67.15 mat-tub driver to a paternally derived UAS-clamp RNAi transgenes (Staller et al. 2013). We used two different UAS-clamp RNAi transgenic lines. The first was, clampi¹ (BDSC #27080) while the second was a line, clampi^2x, carrying clamp RNAi transgenes on the 2nd (BDSC #57163) and 3rd (BDSC #57008) chromosomes. The three RNAi transgenes target different coding sequences in the clamp mRNA, and both #27080 and #57008 have been shown to reduce clamp mRNA levels when expressed in tissue culture cells or the female germline (Soruco et al. 2013; Rieder et al. 2017). As a negative control, we mated 67.15 mothers to a UAS-egfp-RNAi transgene. As illustrated in the egfpi control embryo (Figure 1A, a), and a′, uniformly high levels of CLAMP protein are typically observed in all nuclei in syncytial blastoderm embryos. This is not true for clampi¹ and clampi^2x embryos. While about half of the RNAi embryos appeared similar to controls, CLAMP protein expression was abnormal in the remaining embryos (Figure 1, B and C; see also DAB stained embryos at different NC in Supplementary Figure S2). In about 40% of the syncytial blastoderm embryos, protein accumulation was irregular (Figure 1, a, b, and b’) or modestly reduced across the entire embryo. In the remaining 10–15% of the embryos CLAMP protein expression was substantially reduced or nearly absent (Figure 1, A, c and c′).

Figure 1.

Zygotic-specific knockdown depletes CLAMP protein in early embryos. (A) Immunostaining of CLAMP protein in egfpi (a–a’), clampi¹ (b–b’), and clampi^2x (c–c’) cellular blastoderm embryos is shown on the left. Reduced CLAMP expression in a fallen in nucleus is marked with an asterisk (*) in subpanel b’. Embryos were co-labeled with Hoescht. Scale bar represents 10 µm. (B) Quantification of CLAMP protein fluorescence intensity of embryos displaying knockdown in a representative experiment. n = 20 for each genotype. Significance was tested using a One-way ANOVA (P = 0.000945) with pairwise comparisons (t-test) (P = 0.00942 and 0.00098 for clampi¹ and clampi^2x compared to egfpi embryos, respectively, and P = 0.32747 for clampi¹ compared to clampi^2x). (C) Bar plot showing classification of embryos stained for CLAMP protein according to severity of knockdown using DAB signal. n = 29–32 for each genotype. Medium: Embryos with an overall modest reduction in CLAMP protein and embryos that small sections in which there is little or no detectable protein. Severe: Embryos with a substantial reduction in protein and embryos with large regions that lack protein. Significance was tested using Fisher’s exact test (P = 0.000032 and 0.000287, for clampi¹ and clampi^2x compared to egfpi embryos, respectively).

While we observed reproducible disruptions in CLAMP protein expression in slightly more than half of clampi¹ and clampi^2x embryos, both RT-PCR and RNA-seq (see below) analysis indicated that zygotically expressed clampi¹ and clampi^2x shRNAs did not significantly destabilize clamp mRNAs in pooled pre-cellular blastoderm embryos. As we do not understand the molecular mechanism underlying the observed reductions in CLAMP protein expression in clampi¹ and clampi^2x embryos, we sought to validate our zygotic knockdown strategy. For this purpose, we chose bicoid (bcd). Like clamp, bcd mRNA is maternally derived and is translated during early embryogenesis; however, instead of being uniformly distributed in the embryo, it is localized at the anterior. Translation of the bcd mRNA generates a well-characterized Bcd protein gradient (Driever and Nüsslein-Volhard 1988). As shown in Supplementary Figure S3, abnormalities in the pattern of Bcd protein accumulation were observed for two different bcd RNAi lines, bcdi¹ (BDSC #33886) and bcdi² (BDSC #35478). Moreover, as was the case for clampi, the disruptions in Bcd expression were variable not only between embryos but even within embryos. These data indicate that our zygotic knockdown strategy can deplete proteins encoded by maternal mRNAs, though the degree of knockdown is variable and incomplete.

CLAMP affects a multitude of zygotic transcriptional targets

The defects in CLAMP accumulation induced by zygotic knockdown encouraged us to test whether there are alterations in gene expression in pre-cellular blastoderm embryos. For this purpose, we examined global transcriptional profiles of clampi¹ and clampi^2x embryos as well as egfpi control embryos. RNA-seq was performed on NC 0-14 embryos using three biological replicates with two collection time points (0–1.5 hours and 1.5–3 hours).

We found that a large number of transcripts had at least a twofold change in their levels in the two clamp RNAi knockdowns. For clampi¹ there were a total of 2536 transcripts that were at least twofold reduced in the 0–1.5 hours collection, while in the 1.5–3 hours collection there were 880 (Supplementary Table S1). For clampi^2x, there were 2362 transcripts that were reduced at least twofold in the 0–1.5 hours collection while in the 1.5–3 hours collection there were 1133 (Supplementary Table S1). Of these, 1204 were common between the two clamp RNAi knockdowns in the 0–1.5 hours time point (Supplementary Figure S4A) and 523 were common in the 1.5–3 hours time point (Supplementary Figure S4B). As anticipated from the incomplete depletion of CLAMP protein in the knockdown embryos (Figure 1 and Supplementary Figure S2), only a small subset of transcripts was altered to a statistically significant degree. It is worth noting, however, that there was a significant overlap of transcripts affected by either clampi¹ or clampi^2x knockdown (Supplementary Figure S4), and these are the best candidates for CLAMP targets during ZGA. In addition to the transcripts whose levels decreased at least twofold, there were also transcripts whose levels increased at least twofold. While we cannot exclude the possibility that some of these might be direct targets for CLAMP regulation, we decided to focus on the decreased transcripts. We further classified our identified target transcripts as being of maternal, zygotic, or maternal/zygotic origin based on previously published data (Lott et al. 2011). While many of our identified targets have not yet been classified, about half of the mRNAs whose levels decreased in the clamp RNAi 0–1.5 hours knockdowns are classified as zygotic or maternal/zygotic (Supplementary Table S3). Because our knockdown is zygotic, the remaining maternally derived transcripts may be affected as an indirect consequence of clamp knockdown, or perhaps these transcripts may also be expressed zygotically.

Early embryonic patterning is disrupted in embryos with reduced levels of CLAMP

One seemingly counterintuitive result in the clamp RNAi knockdowns is that the number of transcripts that were reduced by at least twofold is greater in the 0–1.5 hours collection than in the 1.5–3 hours collection (Supplementary Figure S4, A and B). However, this is not altogether surprising as our antibody staining experiments indicated that reductions in CLAMP protein in the knockdown are incomplete and likely to be increasingly less effective as transcription is activated genome-wide. Though there were clear-cut defects in CLAMP accumulation in about half of the syncytial blastoderm embryos, the levels of protein in the remaining embryos were not obviously different from the control. Moreover, even in the embryos that had defects, reductions were often regional and/or not all nuclei were affected. In RNA-seq experiments, the presence of many normal or nearly normal embryos/nuclei would be expected to obscure effects of the knockdown on transcription. On the other hand, if CLAMP is important for upregulating transcription in 1.5–3 hours pre-cellular blastoderm embryos, one might expect to observe expression defects in a subset of the embryos.

To explore this possibility, we selected four segmentation genes that are expressed in the 1.5–3 hours window and are known to be important ZGA genes. These include two pair-rule genes, even-skipped (eve) and runt (run), and two gap genes, buttonhead (btd) and giant (gt). run is X-linked and also functions as an X-linked counting element (XCE) in the sex determination pathway. All four genes have CLAMP binding sites in close proximity to their promoters and thus are plausible candidates for CLAMP dependent activation (Supplementary Figure S5). They are also known Zld targets and have Zld binding sites (Supplementary Figure S5). Supplementary Figure S4C shows the average counts per million reads (compared to the egfpi control) in the clamp¹ and clamp^2x 0–1.5 hours and 1.5–3 hours collections. The figure also shows the average number of reads for two other triplicate RNAi samples, a zygotic zld RNAi knockdown, and a zygotic clampi¹+zldi double knockdown (zld/clampi¹). In the two clamp RNAi knockdowns, there is an apparent reduction in transcript levels in the 0–1.5 hours sample; however, as was observed for other transcripts, the clamp knockdowns had minimal effects on the levels of mRNA for these four genes in the 1.5–3 hours collections. Likewise, even though zld is known to be important for the expression of these four genes during the major wave of ZGA, the effects of the zld RNAi knockdown are also much stronger in the earlier collection.

If CLAMP has functions during ZGA analogous to those of Zld and is important for activating transcription of these four segmentation genes during the major wave in NC13/14 embryos, then one would predict that a subset of the knockdown embryos would show obvious abnormalities in their expression patterns. We used single-molecule fluorescent in situ hybridization (smFISH) and antibody staining to test this prediction. All four genes have a well-characterized pattern of expression in NC13/14 embryos. Based on the WT pattern of expression, we classified the RNAi embryos into different categories as follows (examples shown in Supplementary Figure S1): None-level and pattern of expression resembles wild type; Medium—readily discernable but not uniform reductions in levels of expression or alterations in one more pattern elements; Severe—more complete and/or widespread reductions in the level of expression, or the loss of normal pattern elements. In addition, we analyzed embryos co-stained for CLAMP and putative targets and observed frequent correlation between levels of nuclear CLAMP and targets (like Btd) in clampi embryos (Supplementary Figure S6). This confirms that WT levels of CLAMP correspond to WT patterns and levels of the target genes we examined.

We first examined the expression of the two pair-rule genes, eve and run, in clamp and zld knockdown NC13/14 embryos. In WT and the egfpi knockdown embryos, high levels of cytoplasmic eve mRNA are observed in a 7-stripe pattern (Frasch and Levine 1987; Irvine and Wieschaus 1994). In the egfpi embryo shown in Figure 2A, the level of expression and width of the first stripe is slightly greater than the six remaining stripes. As would be expected if CLAMP is important for eve expression during ZGA, there are obvious defects in more than 2/3rds of the clamp¹ embryos. As illustrated by the example in Figure 2B, the levels of eve mRNA are greatly reduced in the posterior 6 stripes of most clamp¹ embryos. The effects of the knockdown are not as pronounced in the first stripe, though the levels of mRNA are slightly lower and the stripe borders more irregular. Supporting a role for CLAMP during the upregulation of transcription in NC13/14, we observed a similar disruption in the eve stripe pattern in the positive control, the zldi embryos (Figure 2C). To confirm the effects of the clamp¹ knockdown on eve transcription, we used antibody staining to examine the expression of Eve protein. As was observed for eve mRNA, the disruption in protein accumulation is most pronounced in the six posterior stripes. In this case, over 80% of the clamp¹ embryos show detectable defects (Figure 2, F and H). A similar effect on Eve expression is observed in zldi embryos (Figure 2, G and H). Like eve, CLAMP is required for transcription of another pair-rule gene, run. In WT embryos, run is also expressed in seven stripes (Kania et al. 1990), but this transcription pattern is abnormal in clamp¹ and zldi embryos (Supplementary Figure S7).

Figure 2.

CLAMP and Zelda are required for expression of the pair-rule gene eve and the gap gene btd. (A–C) Pattern of eve mRNA accumulation in blastoderm egfpi (A), clampi¹ (B), and zldi (C) embryos. In WT or egfpi, eve is expressed in seven stripes at the blastoderm stage. Note the reduction in eve mRNA in the 6 posterior stripes in both clampi¹, and zldi embryos. (D–F) Expression of Eve protein in egfpi (D), clampi¹ (E), and zldi (F) embryos. Note that, like eve mRNA, Eve protein levels are reduced in the posterior six stripes in the clampi¹ and zldi embryos. (G–I) btd transcription visualized using smFISH in egfpi (G), clampi¹ (H), and zldi (I) embryos. Embryos were co-labeled using Hoescht. Scale bar represents 10 µm. In WT or egfpi embryos btd is expressed in broadband at the anterior. In clampi¹ and zldi embryos, btd expression is reduced and the width of the anterior band is reduced. In some zldi embryos (like the example shown), there is a slight posterior shift of the btd expression band. (J–L) Classification of defects for eve transcription (J), Eve protein (K), and btd transcription (L). For eve, medium defects included embryos showing the modest overall reduction in eve transcript levels or a blurring of borders among the 7 stripes. Embryos were classified as having severe defects if entire stripes were absent (as shown) or overall reductions obscured the stereotypical 7 stripe pattern altogether. n = 12–15 per genotype. Significance was determined using Fisher’s exact test (P = 0.021765 and 0.070471 for clampi¹ and zldi, compared to egfpi embryos, respectively). Defect classes for Eve protein were the same as stated above for eve smFISH experiments. n = 8–12 per genotype. Significance was determined using Fisher’s exact test (P = 0.004389 and 0.020362 for clampi¹ and zldi compared to egfpi embryos, respectively). For btd, embryos were determined to have medium defects if the border of btd expression was blurred or levels were modestly decreased, whereas they were classified as severely defective if the region of expression was shifted or decreased to be indistinguishable from noise in the same channel (as shown in the figure). n = 21–23 embryos per genotype. Significance was determined using Fisher’s exact test (P = 2.60e-05 and 0.000165 for clampi¹ and zldi compared to egfpi embryos, respectively).

clamp RNAi also affected expression of the gap genes btd and gt. In WT and egfpi embryos (Figure 2I), btd is expressed in a single anterior band (Wimmer et al. 1993). In nearly 2/3rds of the clamp¹ embryos, btd transcript levels in this anterior band are reduced compared to the egfpi control (Figure 2, J and L). A similar percentage of zldi embryos showed reduced btd expression (Figure 2, K and L). The effects of the clamp RNAi on btd mRNAs are recapitulated in antibody staining experiments. Like the mRNA, nuclear Btd protein accumulates in a wide anterior band in WT and in the egfpi control (Supplementary Figure S8A). In clampi¹, the overall levels of Btd in the anterior band are reduced, and the band of nuclei that have readily detectable levels of Btd protein is usually narrower (mirroring btd mRNA). In embryos in which we observe a nearly uniform reduction in the expression of the Btd protein, we also observe a reduction in levels of CLAMP protein in the same region of the embryo (Supplementary Figure S6). In other clampi¹ embryos, a patchy pattern of expression is observed with some nuclei having nearly normal levels of Btd while other, neighboring nuclei, are almost completely devoid of protein (asterisks, Supplementary Figure S8, B and C). This would be consistent with the patchy patterns of CLAMP accumulation that are often observed in the RNAi knockdowns (Figure 1 and Supplementary Figure S2). For gt, we examined protein expression. In WT or the egfpi control, Gt accumulates in two anterior bands (one of which is largely restricted to the dorsal side of the embryo) and in a single posterior band (Supplementary Figure S9A) (Kraut and Levine 1991). In the clampi¹ knockdown, the level of protein is reduced, and the borders of the bands are blurred (Supplementary Figure S9B). Similar alterations in the level of Gt protein and the edges of the expression domain were also observed in zldi (Supplementary Figure S9C) and clampi^2x (Supplementary Figure S9D). For clampi¹ and zldi, we quantitated the frequency of syncytial blastoderm embryos showing obvious defects in Gt protein expression. It both cases, it was close to 70% (Supplementary Figure S9E).

Sxl transcription is reduced in clampi embryos

In addition to being the time when cell identity along the A-P and D-V axes is initially selected, ZGA is also the stage in development when sexual identity is chosen. In D. melanogaster, somatic sexual identity depends on the activity state of the master determinant, Sex lethal (Sxl). Embryos with two X chromosomes (females) activate transcription of Sxl from a special establishment promoter Sxl-Pe, while this promoter is kept off in embryos with a single X chromosome (males). The Sxl-Pe promoter is turned on in female embryos only briefly between NC11 and NC14, but this fleeting transcription initiates a positive autoregulatory loop in which Sxl proteins maintain the female determined state. In males, there are no Sxl-Pe mRNAs, and consequently no functional Sxl protein is produced, ensuring male identity (Bell et al. 1988; Bopp et al. 1991; Keyes et al. 1992; Estes et al. 1995; Salz and Erickson 2010). Previous studies have shown that in addition to XCEs (which correspond to transcription factors like run), maternally derived factors like Zelda are required for Sxl-Pe activity (Bosch et al. 2006). Like Zelda, CLAMP protein localizes near the Sxl-Pe TSS (Rieder et al. 2017). This association raised the possibility that activation of Sxl-Pe in females might require CLAMP. In addition, amongst the genes whose expression was at least marginally reduced in the clampi zygotic knockdowns in the 0–1.5 hours window are two XCEs, run and upd (see below). Though small, this reduction in XCE expression might also impact Sxl-Pe activity. In addition, ChIP-seq data (Rieder et al. 2017) shows that CLAMP binds the four primary XCEs—sisA, sisB/sc, run, and upd1—beginning at early stages (NC0-10), further supporting the idea that CLAMP can affect expression from Sxl-Pe, either directly or indirectly.

As was the case for the pair-rule and gap genes, defects in Sxl-Pe activity are evident when individual clampi embryos are examined using smFISH. In control female embryos (egfpi), puncta corresponding to nascent Sxl-Pe transcripts can be detected in all somatic nuclei while they are absent from pole cell nuclei (Figure 3A). In a subset of nuclei, the Sxl-Pe promoter appears to be active on both X chromosomes, and two puncta can be observed. The clamp¹ and clamp^2x RNAi knockdowns differ in at least two respects. First, there are many nuclei in which Sxl-Pe transcripts cannot be detected. Second, in those nuclei in which Sxl-Pe transcripts are present, the puncta are much dimmer, which would indicate that the number of nascent transcripts is lower than in the egfpi control (Figure 3, B and D). Quantfication of the average number of puncta/nucleus in female embryos indicates that the average for the egfpi control is slightly greater than 1, while for both clamp RNAi knockdowns the average number is about 0.5 (Figure 3E). Similar results are observed for the zld RNAi knockdown (Figure 3C). To confirm this analysis, we stained white¹ and clampi¹ blastoderm stage embryos for Sxl protein. At the blastoderm stage in WT, about 50% of the embryos (females) express high levels of Sxl protein (Supplementary Figure S10A), while Sxl protein cannot be detected in the remaining 50% (males). The pattern of Sxl protein accumulation is altered in clampi¹ blastoderm stage embryos (Supplementary Figure S10B). As shown in Supplementary Figure S10C, the levels of Sxl protein are consistently reduced in clampi¹ female embryos compared to the white¹ control.

Figure 3.

Zygotic knockdown of clamp or zelda decreases Sxl transcription in blastoderm embryos. (A–D) Sxl transcription visualized using smFISH with an intron probe in egfpi (A), clampi¹ (B), zldi (C), and clampi^2x (D) embryos. Embryos were co-labeled using Hoescht. Scale bar represents 10 µm. Note the reduced intensity and number of transcription foci in the clampi¹, zldi, and clampi^2x compared to the egfpi control (A’–D’). (E) Quantification of Sxl transcription puncta per nucleus. n = 10–12 embryos per genotype. Significance was determined using a One-Way ANOVA with pairwise comparisons (t-test): P = 0.042, 0.035, and 0.035 for clampi¹, zldi, and clampi^2x compared to egfpi, respectively.

CLAMP and Zelda show distinct binding profiles, suggesting nonoverlapping roles during early embryogenesis

Our data indicate that zygotic transcription is affected by loss of clamp. Consequently, in blastoderm stage embryos compromised for clamp, position-specific expression of some of the candidate target genes (eve, runt, Sxl, and so on.) is disrupted. These putative target genes harbor CLAMP binding sites in the 5’ cis-regulatory regions. As CLAMP is uniformly distributed in early embryonic nuclei, we sought to analyze its genome-wide distribution during the early nuclear division cycles using previously published ChIP-seq data from staged embryos (Rieder et al. 2017). The first dataset from embryos NC0-10 could be compared to our 0–1.5 hours RNA-seq collection while NC11-14 will correspond to the 1.5–3 hours collection. We found that CLAMP occupancy undergoes a dramatic rearrangement during the early nuclear division cycles. In pre-NC11 embryos, there are 1622 peaks. These peaks are fewer in number, weaker, and more dispersed than those observed in tissue culture cells or in later nuclear division cycles. At this stage, over 1/3rd of CLAMP binding sites is located in intergenic regions, while another 30% is located in intron sequences, either the 1st intron (7%) or downstream introns (26%) (Figure 4A). Only 27.49% of the CLAMP bound genes have peaks within 1 kb of the transcription start sites (TSS) (Figure 4E, Supplementary Table S4). A quite different occupancy profile is evident in NC11 embryos. First, instead of only about 1600 peaks, there are over 4000, close to the number reported in tissue culture cells (Figure 4B). Second, there is relatively little overlap between genes bound by CLAMP between NC0-10 and NC11 (Supplementary Figure S11A). Third, the distribution of binding sites is also dramatically different. While only 27.49% of the CLAMP peaks are located within 1 kb of TSS in NC0-10, 95.91% of the CLAMP-bound genes have peaks within 1 kb of TSS in NC11 embryos (Figure 4F, Supplementary Table S4). In subsequent NC, the preferential localization at promoter regions and gene occupancy is maintained (Supplementary Table S4, Supplementary Figure S11B). This behavior is even clearer when examining individual target genes (Supplementary Figure S11C). These findings are consistent with the CLAMP binding profile in tissue culture cells, which also shows enrichment near TSS (Soruco et al. 2013). Supporting the claim that CLAMP binding near TSS is functionally relevant, RNAi knockdown of clamp alters chromatin accessibility and nucleosome occupancy at the TSS in cultured cells and maternally compromised clampi embryos (Urban et al. 2017b; Duan et al. 2020).

Figure 4.

Unlike Zld, CLAMP binding changes during the nuclear division cycles. (A,B) Pie charts showing the distribution of CLAMP occupancy in NC0-10 and NC11 embryos. (C,D) Pie charts showing distribution of Zld occupancy in NC8 and NC13. Color code for the classification of binding sites is on right. (E,F) CLAMP occupancy profiles near TSS at NC0-10 and NC11. (G,H) Zld occupancy profiles at NC8 and NC13. Y axes are adjusted based on peak counts in each dataset (with more peaks in later-stage embryos) (I) Unique and overlapping CLAMP and Zld binding sites located within 1.0 kb of TSS at NC13.

Some features of CLAMP binding occupancy are reminiscent of the other two ZGA regulators Zld and GAF (Harrison et al. 2011; Nien et al. 2011; Moshe and Kaplan 2017). Zld, a pioneer factor, also binds to regulatory elements and genes activated during MZT as early as cycle 8, the start of the minor wave of ZGA (Harrison et al. 2011). But unlike CLAMP, Zld does not appear to undergo a major change in genome-wide occupancy as ZGA progresses (Figure 4, C, D, and G–H). While the timing of their binding may differ, CLAMP and Zld show a substantial overlap in binding sites that are within 1 kb of a TSS. In NC13, there are 1501 genes bound by CLAMP within 1 kb of a TSS, while for Zld there are 3163. Of these genes with promoter-associated binding sites, 991 are occupied by both CLAMP and Zld (Figure 4I). This shared occupancy suggests that these two proteins perform overlapping yet not completely redundant functions during ZGA.

CLAMP and Zelda may act in a cooperative manner during ZGA

ChIP-seq experiments indicate that a subset of CLAMP-associated genes is also bound by Zld. Furthermore, we found that embryos depleted for either CLAMP or Zld display similar defects in the expression of several different genes. If both CLAMP and Zld function in establishing a chromatin environment that supports the transcriptional activation/differential regulation of these genes, one might expect that the reduction in transcript levels would be greater when zygotic RNAi is used to deplete both factors. To explore this possibility, we selected a collection of critical early zygotically activated genes involved in segmentation, D-V polarity, and sex determination that have both CLAMP and Zld binding sites and compared the transcript levels for these genes in the single clampi¹ or zldi knockdowns with that in the double clampi ¹+zldi knockdown (zld/clampi¹). We focused on the 0–1.5 hours collection as the zygotic RNAi knockdowns were most efficacious in this interval. As can be seen in Figure 5A, the extent of reduction was enhanced in the double knockdown compared to the single knockdown. For example, the reduction in eve expression in the single knockdowns was less than twofold, while in the double knockdown the level was reduced fourfold. Similarly, for odd, the single knockdowns reduced transcript levels less than twofold, while in the double knockdown, odd transcripts were reduced to less than 20% of the control.

Figure 5.

Simultaneous knockdown of CLAMP and Zld reveals exacerbated effects on the embryonic transcriptome compared to compromise of either CLAMP of Zld alone. (A) Heatmap showing expression changes (relative to egfpi control) in 0–1.5 hours collections from clampi¹, zldi, and zldi/clampi¹ embryos. Values in the heat map (and represented in the color key) are fractional counts per million compared to egfpi control, making 1 equal to amount of RNA detected in egfpi sample. Any number below one (warmer colors) represents decrease while numbers greater than one indicate increase. (B) Comparison of all zygotic genes decreased by at least twofold in clampi¹ (yellow), zldi (orange), and zld/clampi¹(gray) datasets. (C) P values indicating significance of overlap among datasets illustrated in panel B. Hypergeometric tests were used to analyze significance of overlap between pairs, while a Cochran-Mantel-Haenszel test was used to analyze significance of the overlap among all three datasets.

We then compared the zygotic genes [classified by previously published data (Lott et al. 2011)] that were decreased at least twofold among the early clampi¹, zldi, and zld/clampi¹ datasets (Supplementary Table S5). We found significant overlap between each pair of datasets, as well as shared targets identified from all three genotypes (Figure 5, B and C). It is notable that the number of zygotic genes with reduced expression (by at least twofold) in the zld/clampi¹ collection (1245) was greater than those identified when either clamp or zld alone was knocked down (573 and 740, respectively) (Supplementary Table S5). Furthermore, the majority of decreased genes in the zld/clampi¹ dataset were shared with clampi¹ and/or zldi collections (Figure 5B).

clampi embryos display nuclear and mitotic defects during early nuclear division cycles

Studies on the two other proteins implicated in ZGA, Zld and GAF, indicate that in addition to defects in transcriptional activation, other key steps in early development are disrupted when their activity is compromised (Bhat et al. 1996; Staudt et al. 2006; Liang et al. 2008). For example, embryos injected with dsRNA for zld display a variety of nuclear division defects, including chromosomal bridges and divisions perpendicular to the cortex, which results in a second layer of nuclei. There are also defects in nuclear spacing and morphology, mitotic synchrony, and chromosome segregation (Staudt et al. 2006). A similar range of nuclear division phenotypes were found in embryos compromised for maternally derived GAF (Trl^13C mothers) (Bhat et al. 1996). For these reasons, we tested whether depletion of CLAMP in embryos is accompanied by a comparable spectrum of nuclear division cycle defects.

We first examined nuclear morphology using Hoescht staining. The nuclei in egfpi embryos resemble WT (Figure 6A). They are distributed evenly at the cortex in blastoderm stage embryos and have a quite homogenous morphology. In contrast, in many clampi¹ embryos, nuclear spacing is irregular, with some nuclei clustered closely together, while other nuclei are separated by abnormally large gaps (Figure 6B). In addition, instead of a homogenous morphology, some nuclei look condensed, while others do not. There are also small satellite nuclei that have less than a full complement of DNA, while other nuclei are larger than normal, appearing to have excess DNA. Similar nuclear phenotypes were evident in zldi blastoderm embryos (Figure 6C). We classified embryos according to the extent of nuclear abnormalities (Figure 6, D-F): none (all nuclei in the same mitotic phase and spaced uniformly at the cortex), medium (a few nuclei show abnormal morphology or spacing), and severe (most of the nuclei in an embryo show abnormal morphology or spacing). We found that only a small percentage of control embryos (14%) show any nuclear defects, but both clampi¹ and zldi embryos show significant nuclear abnormalities that are similarly penetrant (Figure 6G). While the images in Figure 6 show syncytial blastoderm stage embryos, a similar spectrum of defects was also observed during earlier NCs, before migration of nuclei to the embryonic cortex.

Figure 6.

Nuclear morphology is disrupted upon clamp or zelda knockdown. (A–C) Confocal images of DNA visualized using Hoescht in egfpi (A), clampi (B), and zldi (C) blastoderm embryos. (D–F) Examples of nuclear defects illustrating classes represented in bar chart on right. A representative nucleus with too much DNA is indicated with an asterisk (*) while a nucleus with too little is indicated with an arrowhead (>). (G) Classification of nuclear defects as indicated. n = 150–226 per genotype. Significance was determined using Fisher’s exact test (P = 0.000 and 0.000 for clampi¹ and zldi, respectively, compared to egfpi embryos). Scale bars represent 10 µm.

To further investigate nuclear division defects in clamp or zld knockdown embryos, we analyzed the mitotic marker histone-H3 phospho-Ser10 (pH3). pH3 first appears in late G2/M and persists until late anaphase when it is dephosphorylated (Hendzel et al. 1997). In WT or control egfpi pre-cellular blastoderm embryos, essentially all nuclei are at the same point in the nuclear division cycle. Therefore, all nuclei in WT/control embryos are either negative for pH3 or are stained by pH3 antibody (Figure 7A). By contrast, in clampi¹ (Figure 7B) or zldi (Figure 7C) embryos, we observe many instances in which pH3-specific signal is absent in some nuclei but present in others. Some embryos also exhibited neighboring pH3 positive nuclei in different phases of mitosis based on the state of condensation and orientation of the pH3-positive chromosomes. We classified the mitotic defects in embryos as none (all nuclei show pH3 signal and are in the same stage of mitosis), medium (a small patch of nuclei or few scattered nuclei throughout an embryo in a different phase or not undergoing mitosis), and severe (a large proportion of nuclei in a different phase of the cell cycle, either in noticeable groups or scattered all over giving a patchy appearance of pH3-positive nuclei), and examples of these classifications are given in Supplementary Figure S1. Based on this classification, about 80% of the clampi¹ embryos have detectable abnormalities in the pattern of pH3 staining, while an even greater percentage is seen in zldi (Figure 7D).

Figure 7.

clampi¹ or zldi blastoderm embryos display mitotic asynchrony. (A–C) Images of embryos immunostained for Histone-H3 phosphorylation on Ser10 (pH3) in dividing nuclei in egfpi (A), clampi¹ (B), and zldi (C) blastoderm embryos. Embryos were co-labeled using Hoescht. Each image is a maximum intensity projection of three slices at 1 µm intervals at the embryonic cortex. Scale bar represents 10 µm. Note the asynchrony in clampi¹ (B and B’) and zldi (C and C’) embryos. Asterisks mark adjacent nuclei that are pH3 negative. (D) Classification of mitotic defects in pH3-positive nuclei as follows: Medium: only a few, scattered nuclei in a different condensation state or pH3 negative (as shown in clampi¹ representative image); Severe: Many nuclei in different condensation states/mitotic phases (as shown in example zldi embryo). n = 10–13 per genotype. Significance was determined using Fisher’s exact test (P = 0.001046 and 0.000115 for clampi¹ and zldi, respectively, compared to egfpi embryos).

An equivalent set of abnormalities is evident in centrosome organization. Centrosomes duplicate early in the nuclear division cycle and then immediately migrate to opposite poles of the nucleus where they remain until mitosis. For this reason, one rarely observes duplicated but not yet separated centrosomes. While egfpi control embryos have two centrosomes at opposite poles of each nucleus (Figure 8A), clampi¹ and zldi embryos exhibit a variety of defects in centrosome duplication and migration (Figure 8, B and C). In some cases, the duplicated centrosomes remain in close proximity, failing to migrate apart (as in the zldi embryo shown, Figure 8C’). In other cases, centrosomes initiate but do not complete migration to opposite poles, resulting in mispositioned centrosomes. An even more striking defect is the presence of multiple orphaned centrosomes that are not associated with a nucleus in both clamp and zld knockdown embryos (highlighted with asterisks in Figure 8, B′–C′). Altogether, about 2/3rds of the clampi¹ and zldi embryos have one or more of these centrosome defects (Figure 8D).

Figure 8.

Centrosomal and cytoskeletal defects in syncytial blastoderm embryos compromised for CLAMP or Zld. (A–C) Immunofluorescent images of embryos co-immunostained for Centrosomin to assess integrity of the mitotic apparatus and Pnut (Septin) to visualize the cortical cytoskeleton. Scale bar represents 10 µm. Cnn panels show examples of orphan centrosomes—marked with an asterisk(*)—and centrosomes that have duplicated but not separated (B’ and C’). Arrowheads (>) highlight examples of Pnut accumulation in regions containing orphaned centrosomes (B″ and C″). (D) Embryos were classified according to severity of Cnn defects (medium: only a few centrosomes improperly duplicated and localized; severe: many centrosomes inappropriately duplicated or localized, or a large patch of orphaned centrosomes, as evident in panels B’ and C’). n = 34–45 per genotype. Significance was tested using Fisher’s exact test (P = 0.000033 and 0.000086, for clampi¹ and zldi compared to egfpi embryos, respectively).

Embryos compromised for clamp or zelda exhibit actin-myosin cytoskeletal defects

Cellularization during NC14 is a multistep process which begins in earlier cycles after the nuclei have migrated to the surface of the embryo (Mavrakis et al. 2009). Components of the actomyosin-based cytoskeleton initially form a cap above the regularly spaced blastoderm. At the onset of cellularization, these caps reorganize into hexagonal actomyosin rings encircling the nuclei, and these rings drive the process of membrane invagination. This cytoskeletal architecture is critical for ensuring that each nucleus is properly encapsulated by a cell membrane during NC14 (Warn and Magrath 1983; Mazumdar and Mazumdar 2002; Schmidt and Grosshans 2018).

The centrosomal defects in clampi¹ (and zldi) embryos suggest that there may be related aberrations in the embryonic actomyosin cytoskeleton that could impact the process of cellularization. To investigate this possibility, we examined the distribution of the Septin Peanut (Pnut) and a transmembrane glycoprotein Neurotactin (Nrt). In wild type blastoderm embryos, these proteins are distributed in a hexagonal pattern surrounding each nucleus, producing a regular lattice-like array. This is shown for Pnut in the control egfpi embryo in Figure 8A, while it is shown for Nrt in the egfpi embryo in Figure 9A. For both clampi¹ and zldi, the regular hexagonal lattice is disrupted in many of the embryos. In the case of Pnut, high concentrations of protein accumulate in regions of the embryo that have multiple orphan centrosomes with no associated nuclei (see arrowheads in Figure 8, B’’–C’’). For Nrt, the lattice often encircles several nuclei (asterisk) instead of individual nuclei or encircles regions that have no nuclei (arrowhead) (Figure 9, B and C). With either clamp or zld knockdown, at least half of the embryos showed cytoskeletal defects, which we classified similarly to other phenotypes (examples of defect classes are shown in Supplementary Figure S1).

Figure 9.

The Neurotactin network of syncytial blastoderm (NC14) embryos is disrupted upon clamp or zld knockdown. (A–C) Immunofluorescent images of cortical cytoskeleton from embryos immunostained for Nrt and co-labeled using Hoescht. Scale bar represents 10 µm. Examples of Nrt encompassing multiple nuclei are marked with an asterisk(*) while examples of Nrt encompassing a domain devoid of nuclei are marked with an arrowhead (>). (D) Embryos were classified according to severity of defect in Nrt network. Defects were determined to be medium if only a few rings of cytoskeleton displayed no or multiple nuclei, while they were deemed severe if much of the embryo exhibited an irregular lattice pattern with asymmetrical distribution of nuclear material (as shown in panels B and C). n = 20–30 per genotype. Significance was tested using Fisher’s exact test (P = 0.000167 and 0.001086, for clampi¹ and zldi compared to egfpi embryos, respectively).

Discussion

ZGA, a universal feature of early embryogenesis, is ultimately responsible for converting pluripotent cells into differentiated cell types. This is achieved by spatially restricted transcription, which is essential for acquiring distinct cellular identities across the embryo (Ahmed et al. 2010; Vastenhouw et al. 2010; Li et al. 2014; Perino and Veenstra 2016; Eckersley-Maslin et al. 2018). Ensuring that key transcriptional regulators have access to their target sequences in a temporally and spatially coordinated manner is thus a crucial function of factors that provide the chromatin platform for ZGA. The initial chromatin architecture established during ZGA also provides the template for subsequent alterations in the local chromatin organization of genes encoding determinants that are critical in defining and maintaining unique cell identities.

We show here that the polydactyl C₂H₂ zinc finger protein CLAMP functions similarly to (and probably coordinates with) the zinc finger DNA binding protein, Zld (and likely also GAF), in establishing the chromatin architecture essential for ZGA. Though these factors are maternally deposited as mRNAs and are required during embryogenesis, both the CLAMP and GAF proteins also play an important role during oogenesis. In contrast, Zld protein does not seem to have a role during oogenesis as zld mRNAs are only translated following fertilization (Staudt et al. 2006; Liang et al. 2008; Eichhorn et al. 2016; Hamm et al. 2017). A requirement during oogenesis potentially complicates interpreting experiments in which CLAMP (or GAF) is depleted in the mother. For this reason, we devised a strategy to deplete CLAMP exclusively zygotically. For this purpose, we used maternally deposited GAL4 to drive the expression of UAS-clamp RNAi transgenes inherited from the father in early embryos. Though this strategy circumvents the potential complications of depleting CLAMP during oogenesis, it has the drawback that it results in only a partial knockdown of CLAMP in early embryos. In slightly more than half of the embryos, there is detectable reduction in CLAMP protein levels, and in many cases this reduction is limited to a subset of nuclei. Consequently, we expect that our findings may significantly underestimate the scope of CLAMP functions at this stage of development. Despite this limitation, our studies reveal that CLAMP is a key participant in establishing a chromatin organization that provides a foundation for the genome-wide activation of transcription during ZGA. Our results also indicate that generating this permissive chromatin organization impacts processes beyond gene expression.

Not unexpectedly given the partial disruption in CLAMP accumulation, our RNA-seq analysis of both clamp RNAi knockdowns (clampi¹ and clampi^2x) did not show massive reductions in zygotic gene expression. However, a substantial number of genes showed at least a twofold reduction in expression in 0–1.5 hours and 1.5–3 hours collections in the clampi¹ and clampi^2x knockdowns. Of these genes, the most probable CLAMP targets are the ones in common in the two knockdowns. This is likely an underestimate of the number of genes that require CLAMP during ZGA, given our observations of expression patterns of four patterning genes—two pair-rule genes and two gap genes—that are known Zld targets and exhibit CLAMP binding near the TSS. In the RNA-seq experiments, all four of these genes had rather modest reductions in expression in the 0–1.5 hour collection while there was little or no difference in levels from the control egfpi in the 1.5–3 hours collection. However, when we examined their expression in individual syncytial blastoderm embryos, obvious reductions in the level of expression and/or alterations in the patterns of expression were evident in the majority of the embryos. For example, in clampi¹ embryos, expression of eve in the posterior six stripes was substantially reduced, while there was only a minimal effect on eve expression in the most anterior stripe. Moreover, similar effects on the patterns of these four genes were observed when we knocked down Zld using the same zygotic RNAi strategy employed for CLAMP. Patterning genes are not the only ones affected by reducing CLAMP. We also found that expression of Sxl transcripts from the Sxl-Pe establishment promoter is suppressed in clampi embryos. The most obvious change is in the frequency of nuclear Sxl-Pe foci in female embryos (Sxl-Pe is not active in males). In the control, nearly all nuclei in female embryos have Sxl foci and in many cases both genes on the X chromosome are active as two puncta can be detected. In contrast, in the clampi knockdowns, less than half of the nuclei in female embryos had foci, and, in those that did, two puncta were not observed. While CLAMP is associated with the Sxl gene, the effect of the CLAMP knockdown on Sxl-Pe activity could be both direct and indirect. In particular, we found that CLAMP binds all four primary XCES, though transcription of only two (run and upd) decreased in the clamp knockdowns.

Also supporting a function for CLAMP in establishing a permissive chromatin organization for gene expression during ZGA is the distribution of CLAMP protein across the genome. Like Zld and GAF, CLAMP is enriched in sequences located within less than one kb of promoters in early embryos. However, unlike Zld, which appears to be preferentially associated with promoter regions during most of the nuclear division cycles, CLAMP localization is dynamic. Prior to NC11, CLAMP is found largely in intergenic sequences and introns. After NC11, the total number of CLAMP-associated sites jumps more than twofold, and about 2/3rds of these sites are near promoters. Interestingly, a subset of the CLAMP-associated promoter regions is also bound by Zld. This co-localization of CLAMP and Zld suggests that, at least in some instances, both factors may be important in preparing the local chromatin organization for transcriptional activation. We tested whether CLAMP might facilitate Zld activity by simultaneously knocking down both CLAMP and Zld and examining the expression of a collection of ZGA genes known to depend on Zld. In this sample, the extent of reduction in transcript levels in the double knockdown was greater than that in either single knockdown alone. This finding is consistent with recent studies which have shown that CLAMP can facilitate the formation of regions of chromatin accessible for Zld binding and vice versa (Duan et al. 2020). For this particular group of genes, our results suggest that CLAMP and Zld have similar roles in their transcriptional activation; however, this is not the only possible relationship between these two zinc finger proteins. In addition to the sets of genes whose transcriptional activity depends on both factors, there may also be genes in which the two proteins play antagonistic rather than complementary roles.

Alterations or reductions in the patterns of gene expression during ZGA are not the only defects evident in the zygotic clamp RNAi knockdowns. There are also abnormalities in the nuclear division cycles and in the cytoskeleton. The nuclear division defects include mitotic asynchrony, abnormal chromosome segregation, and chromosome fragmentation. Accompanying these chromosomal defects, there are abnormalities in centrosome function including incomplete separation after centrosome duplication and the accumulation of orphaned centrosomes. The organization of the cytoskeleton is also anomalous. The regular hexagonal lattice that envelops blastoderm nuclei is disrupted in clamp RNAi embryos. In some cases, a single ring encompasses multiple nuclei instead of a single nucleus, while in other cases, an actomyosin ring is formed around a region devoid of DNA.

At this point, it is not clear why these abnormalities in nuclear division and cytoskeletal organization arise. However, since similar defects have been observed when Zld or GAF is depleted during early embryogenesis (Bhat et al. 1996; Staudt et al. 2006; Liang et al. 2008), it seems likely that the mechanisms responsible might be shared in all three cases. One idea is that zygotic gene expression essential for nuclear division and/or cytoskeletal organization is altered in the absence of CLAMP, Zld, or GAF. While there are likely genes in this category that could impact nuclear division and/or cytoskeletal organization, it also seems possible that failing to establish an appropriate chromatin architecture when one of these factors is depleted could have deleterious effects not only on transcriptional activation but also on processes such as replication, chromosome compaction, sister chromatid decatenation, and chromosome segregation. For example, topoisomerases are known to utilize open regions of chromatin to gain access to DNA, and they might not be able to efficiently disentangle catenated sister chromatids during the rapid divisions in the absence of these factors (Udvardy and Schedl 1991; Sperling et al. 2011). Likewise, access to replication origins during the rapid nuclear divisions might require open regions of chromatin (Cayrou et al. 2015). Further studies will be required to draw connections between chromosome organization and these other cellular processes.

Emerging evidence has indicated that CLAMP, like Zld and GAF, displays the canonical traits of a pioneer factor: binding nucleosomal DNA, establishing open chromatin regions, and permitting recruitment of additional transcription factors (Zaret and Carroll 2011; Schulz et al. 2015; Sun et al. 2015; Hamm and Harrison 2018; Fernandez Garcia et al. 2019; Duan et al. 2020; Gaskill et al. 2021). In Drosophila embryos, the burgeoning chromatin architecture is sculpted by the early transcription initiated during syncytial nuclear division cycles leading to the establishment of topologically associated domain (TAD) boundaries (Hug et al. 2017; Hug and Vaquerizas 2018). Interestingly, however, only a subset of TAD boundaries are affected upon depletion of Zld (Hug et al. 2017). It is thus conceivable that CLAMP acts in conjunction with Zld to sculpt early embryonic chromatin architecture. Future experiments will thus focus on the mechanistic underpinnings of how CLAMP functions during establishment and/or maintenance of TAD boundaries during ZGA.