Bicoid-nucleosome competition sets a concentration threshold for transcription constrained by genome replication

SUMMARY

Transcription factors (TFs) regulate gene expression despite constraints from chromatin structure and the cell cycle. Here, we examine the concentration-dependent regulation of hunchback by the Bicoid morphogen through a combination of quantitative imaging, mathematical modeling, and epigenomics in Drosophila embryos. By live imaging of MS2 reporters, we find that, following mitosis, the timing of transcriptional activation driven by the hunchback P2 (hbP2) enhancer directly reflects Bicoid concentration. We build a stochastic model that can explain in vivo onset time distributions by accounting for both the competition between Bicoid and nucleosomes at hbP2 and a negative influence of DNA replication on transcriptional elongation. Experimental modulation of nucleosome stability alters onset time distributions and the posterior boundary of hunchback expression. We conclude that TF-nucleosome competition is the molecular mechanism whereby the Bicoid morphogen gradient specifies the posterior boundary of hunchback expression.

In brief

The mechanism whereby the Bicoid morphogen gradient regulates target gene expression domains has been challenging to explain at the level of DNA binding. Here, Degen et al. demonstrate how transcription factor-nucleosome competition at key binding sites determines the concentration-dependent regulation of target gene expression.

Graphical Abstract

INTRODUCTION

Transcription factors (TFs) bind to sites within cis-regulatory DNA sequences (enhancers) to modulate the expression of associated genes.^1,2 How TFs regulate target gene transcription depends on enhancer binding site composition, contemporaneous TF concentration, as well as competition between DNA binding factors.³ Theoretical models have highlighted the importance of TF concentration and chromatin context in determining TF binding.^4–6 The activity of the TF Bicoid (Bcd) in the Drosophila embryo presents an ideal system for studying concentration-sensitive TF-chromatin interactions in vivo. Long appreciated as a canonical morphogen, Bcd is expressed in an anterior-posterior (AP) concentration gradient that positions the expression of zygotic segmentation genes early in development.^7–13 Target gene expression domains span different ranges of Bcd levels, suggesting that the enhancers mediating these domains differ in their affinity for Bcd.¹⁴ However, Bcd binding site strength and enhancer affinity fail to predict the Bcd concentration thresholds associated with target gene expression boundaries.^15,16 Prior work has established that the concentration sensitivity of chromatin accessibility at a locus can explain the degree to which Bcd binding at that locus depends on Bcd concentration.¹⁵ Yet, we do not fully understand how Bcd-nucleosome competition for occupancy at a locus contributes to precise target gene transcriptional outputs across the gradient.

Maternally supplied morphogens like Bcd initiate pattern formation despite constraints on transcription imposed by intense cell cycle activity.^17,18 In the syncytial Drosophila embryo, 13 synchronous nuclear divisions occur over the first 2 h of development. During each mitosis, nuclei export TFs and RNA polymerase II, limiting transcriptional activity. In each following interphase, nuclei must rapidly complete DNA replication, reorganize chromatin, and attempt to activate transcription.^19,20 While intense cellular proliferation characterizes early development, few studies have considered the impact of the cell cycle on transcription prior to large-scale zygotic genome activation. Recent work in early Drosophila embryos has shown a coupling between replication and transcription such that inhibition of DNA replication reduces the activity of a Bcd-driven transcriptional reporter.²¹ In addition to TF-chromatin interactions, the influence of the cell cycle on the gene expression outputs of the Bcd morphogen gradient has not been fully addressed.

Computational models have built up our understanding of how Bcd regulates patterns of gene expression in vivo.^22–31 Many studies have focused on Bcd’s activity at hunchback P2 (hbP2), an enhancer that contributes to the anterior domain of zygotic hunchback (hb) expression.¹¹ Analyses of static expression measurements have suggested that Bcd binds cooperatively to determine the position and steepness of the anterior hb domain’s posterior boundary.^32,33 Bcd’s cooperative binding may require energy expenditure or a higher-order mechanism than pairwise protein-protein interactions.^24,29,31,33 Additionally, the TF Zelda (Zld) and hb itself can influence hbP2 reporter gene expression.^23,25,33,34 Models describing transcriptional dynamics measured with the MS2-MCP RNA labeling system have helped detangle potential regulatory roles of Bcd and additional TFs at hbP2.^{23,25,29,35,36} A recent model suggests that energy-consuming remodeling of chromatin may serve as a mode of Bcd-dependent transcriptional regulation at hbP2.²³ Modeling hbP2-mediated transcriptional dynamics therefore provides a path toward understanding how in vivo context dictates TF concentration-sensitivity.

Here, we investigate how the initiation of transcription of a highly Bcd-dependent hbP2-MS2 reporter during nuclear cycle 13 (NC13) relies on Bcd concentration. By building a stochastic model of transcriptional activation, we propose how Bcd-nucleosome competition and DNA replication determine the concentration sensitivity of Bcd-dependent transcription. We validate the model by testing its predictions through live-imaging and genome-wide sequencing experiments and propose common regulatory mechanisms for Bcd target genes.

RESULTS

The activity of the hbP2 enhancer as a case study for TF-nucleosome competition

To understand how TF-chromatin interactions influence transcription, we consider the regulation of the zygotic patterning gene hb. Early in development, a proximal promoter and pair of enhancers (P2 and shadow) regulate an anterior domain of hb expression.^11,37,38 Later, the distal promoter and the stripe enhancer take over the regulation of hb, mediating the anterior and posterior stripes of expression.³⁹ As measured by ChIP-seq, Bcd binds to the P2 and shadow enhancers and, to a lesser extent, the stripe enhancer (Figure 1A). Genome-wide measurements of Bcd binding and chromatin accessibility suggest that Bcd competes with nucleosomes to bind the hbP2 enhancer.¹⁵ ChIP-seq measurements in embryos expressing Bcd at low, medium, or high uniform (non-graded) concentrations along the embryo show that Bcd binds to hbP2 less at the lower concentration than it does at the higher concentrations (Figure 1A).¹⁵ Chromatin accessibility as measured by ATAC-seq at hbP2 also depends on Bcd concentration, suggesting Bcd may reorganize nucleosomes at hbP2 (Figure 1B). Using NucleoATAC to predict dyad centers from ATAC-seq fragment size distributions,⁴⁰ we find that two nucleosome dyads are predicted to overlap the hbP2 enhancer in embryos mutant for bcd (Figure 1B, bcd^E1 arrowheads).¹⁵ The bcd^E1 mutation represents a complete loss of function, and the ATAC measurements, therefore, reflect the chromatin state observed in the absence of the TF. Increasing the concentration of Bcd gradually reduces the likelihood of dyad prediction, suggesting that a high Bcd concentration reduces nucleosome occupancy at hbP2 (Figure 1B). The likelihood that a nucleosome occupies a site can also be predicted from the underlying DNA sequence.³⁰ A sequence analysis using the Widom-Segal statistical positioning model predicts the occupancy of two nucleosomes within hbP2 at similar positions to those measured in bcd^E1 mutants (Figure 1C).⁴¹ These predicted nucleosomes obscure all available Bcd binding sites. ChIP-nexus footprinting of Bcd binding identifies nine Bcd binding sites in hbP2 that overlap with the observed distal and proximal nucleosomes and that correspond at least in part to the expected Bcd motif position weight matrix (Figure 1C; Table S1).⁴² Our in vivo and in silico observations both support a model where Bcd must outcompete nucleosomes in order to access its sites in hbP2, where the outcome of Bcd-nucleosome competition depends on Bcd concentration.

Figure 1. Nucleosome positioning at the hbP2 locus is sensitive to Bcd concentration and cell cycle timing.

(A) A 15-kb region flanking the hb locus, including the P2, shadow, and stripe regulatory elements (indicated at top), and counts-per-million (CPM) normalized ChIP-seq for the Bcd protein. Mean binding measurements for wild-type embryos (top row) is shown in comparison to three lines that express Bcd uniformly across the AP axis at low, medium, or high concentrations ($n=3$ averaged per genotype). The locations of the minimal P2 enhancer and extended P2+promoter regions are shown in green (bottom). Bracketed numbers indicate displayed y-axis range in CPM.

(B) A 0.6-kb detail of the region upstream of the transcription start site of the hbP2 isoform including the entire minimal P2 enhancer element (green, bottom). ATAC-seq data corresponding to accessible regions and nucleosome dyad positions predicted by NucleoATAC (nuc probability) are shown for five genotypes ($n=3$ averaged per genotype). Predicted dyad positions referred to in the text are highlighted with arrowheads.

(C) Bcd ChIP-nexus data shown for the same genomic region as in (B). Positive and negative contributions to Bcd binding are plotted (blue)⁴², and the positions are shown of six binding sites identified through DNase footprinting (A1–3, X1–3), in addition to three extra motifs with PWM scores greater than 80% (E1–3). Two footprints are observed overlapping motifs for Zld (“z”). It is unclear if this represents Bcd binding at these sites or footprinting of co-immunoprecipitated Zld proteins. The lower portion of the panel shows the positions of two nucleosomes predicted by NuPoP and a schematic of the P2 enhancer with nine Bcd binding sites.

(D) Schematic of a minimal hbP2-MS2 reporter: the minimal hbP2 enhancer (green) drives the expression of a chimeric MS2(24)-yellow intron-Gal4 reporter gene through the heterologous Drosophila synthetic core promoter (DSCP, gray). Sequences encoding 24 MS2 (v5) hairpins (blue) were introduced to the 5′ end of the yellow intron (yellow) upstream of the gal4 coding sequence (orange).

(E) Mean reporter activity of hbP2-MS2 (blue) and mean nuclear intensity of EGFP-Bcd (green) measured over NC13. All three measurements were performed in separate embryos. The timeline (top) highlights the cell cycle landmarks relative to the period of NC13, timed from the beginning of anaphase 12.

(F) ATAC data over the same region as in (B) but for wild-type embryos collected at mitotic metaphase (top, NC13 + 0′) and late interphase (bottom, NC13 + 15′).²⁰

See also Table S1.

Live transcriptional reporters allow for the measurement of stages of the transcriptional cycle regulated by upstream TF activity

To read out the effects of Bcd-nucleosome competition on transcription, we created an MS2 reporter for the previously defined hbP2 enhancer (hbP2-MS2).¹¹ hbP2 and a Drosophila synthetic core promoter in our reporter drive the expression of a series of 24 MS2 stem loops, facilitating live imaging with the MS2-MCP system (Figure 1D).^35,36 This reporter does not contain the hbP2 promoter or any additional genomic sequence from the hb locus, allowing us to directly evaluate the impact of Bcd’s binding to its sites in hbP2 and minimize regulation by additional factors (Figure 1D). Importantly, NuPoP predicts that a pair of nucleosomes overlap the P2 enhancer within the reporter construct at similar positions to the nucleosomes predicted in the genomic context. While prior studies have analyzed the expression patterns of hbP2-MS2 reporters, these reporters reflected the combined activities of hbP2 and the P2 promoter.^{23,25,29,35,36} We measured hbP2-MS2 expression during NC13, an on-average 20-min syncytial cell cycle bounded by near-synchronous nuclear divisions. Quantification of hbP2-MS2 expression across NC13 reveals that average reporter transcription is delayed until around 5 min into NC13, after which mean reporter activity increases linearly until reaching a maximum at around 10 min into the nuclear cycle (Figure 1E, blue). Similar patterns of transcription have been produced by hbP2+P2 promoter reporters.^35,36 We aimed to evaluate how hbP2-MS2 expression depends on Bcd concentration.

To test the hypothesis that a feature of transcription reflects Bcd’s concentration-sensitive role at hbP2, we characterized our hbP2-MS2 reporter’s activity by live imaging a series of 19 embryos over NC13. We measured the transcriptional dynamics of 1904 individual nuclei spanning the anterior 60% of the embryo, with 1384 of these nuclei expressing the reporter (Figure 2A). We extracted features of the dynamics, including the time at which an MCP-GFP spot appears within a nucleus (the “transcriptional onset time” of that nucleus; Figure 2B), as well as the rate of increase in MCP-GFP spot fluorescence over the first minute following the transcriptional onset of a nucleus (the “Pol II loading rate” of that nucleus; Figure 2B). Additionally, we quantified the fraction of nuclei at Bcd concentrations along the AP axis (AP position) that initiate transcription in NC13 (the “fraction of active nuclei”). A heatmap representation of per-nucleus MS2 signal indicates that nuclei vary in the time when they first initiate reporter transcription (Figure 2A). We hypothesized that one or more of the features of transcriptional dynamics—transcriptional onset time, Pol II loading rate, and fraction of active nuclei—would correlate with the change in Bcd concentration across the AP axis.

Figure 2. Transcriptional onset time of hbP2-MS2 varies as a function of AP position.

(A) A heatmap of hbP2-MS2::MCP-GFP fluorescence intensity profiles for ~1400 individual nuclei scored as positive for hbP2-MS2 ($n=19$ independent movies) over NC13. Nuclei are sorted by onset time.

(B) The MS2::MCP-GFP measurements for two representative nuclei, one from the anterior (dark blue) and another from the posterior (light blue) of the hbP2 expression domain. The portions of the data that yield onset time and loading rate measurements are indicated in red.

(C) Mean fraction of hbP2-MS2 active nuclei per 2.5% AP bin (blue ± SD). The gray box in this plot and in Figure 2 D–E indicates that no movies were taken within this region.

(D) The onset time of hbP2-MS2 as a function of AP position (gray points). A rolling average of onset times ($n=100$) is shown (blue ± SD) across the 95% CI of active nuclei AP positions.

(E) The loading rate of hbP2-MS2 as a function of AP position (gray points). A rolling average of the measurements is shown as in Figure 2D.

See also Figures S1 and S2.

The timing of transcriptional onset following mitosis reads out changes in Bcd concentration

Out of the three features of transcriptional dynamics, we find that hbP2-MS2 transcriptional onset time following mitotic exit (“onset time”, hereafter) exhibits the highest sensitivity to Bcd concentration (Figures 2C–2E). In contrast, the fraction of active nuclei is not highly sensitive to Bcd concentration, as it plateaus in the anterior of the embryo; although Bcd concentrations drop by ~55% between 20 and 35% egg length (EL), almost 100% of nuclei initiate transcription across these AP positions (Figure 2C). Similarly, the mean Pol II loading rate does not change substantially across the hbP2-MS2 expression domain and displays a low sensitivity to Bcd concentration (Figure 2E). This finding fits with prior observations of little correlation between the transcription rates and gene expression boundaries of Bcd targets.⁴³ The mean onset time, however, correlates more with AP position, supporting this feature’s higher sensitivity to changes in Bcd concentration (Figure 2D). These results are reproducible across biological replicates (Figure S1). As onset times control the duration of transcription and have other downstream effects, we also find that transcriptional duration, amplitude, and total output correlate to varying degrees with AP position (Figure S2). We focus on onset times in this work, as they most likely reflect the direct and primary influence of Bcd concentration on transcription. To test the effects of changing Bcd concentrations on onset times, we generated mutant embryos that express Bcd uniformly.¹⁵ Comparison of the average fluorescence intensity of uniform EGFP-Bcd (tub>uEGFP-Bcd) to wild-type (graded) reveals that the uniform line expresses Bcd at a concentration equivalent to the Bcd level at 38% EL in wild-type (Figure 3A). By in situ hybridization, hbP2-MS2 is uniformly expressed in embryos expressing this uniform level of Bcd (Figure 3B). By live imaging, onset times are also uniform and overlap with the wild-type distribution at 37% EL, closely matching the 38% EL overlap position of the graded and uniform Bcd concentration profiles (Figure 3C). As changes in onset times precisely follow changes in Bcd concentrations, we conclude that onset time likely reflects the direct consequences of Bcd-chromatin interactions on transcription.

Figure 3. Transcriptional onset time of the hbP2-MS2 reporter is a direct reflection of Bicoid concentration.

(A) NC13 nuclear fluorescence of graded EGFP-Bcd (2× EGFP-Bcd, $n=6$ embryos, gray) and uniform EGFP-Bcd (1× tub>uEGFP-Bcd, $n=8$ embryos, green). A rolling average of fluorescence within a 10% AP window is shown ± SEM. The dotted line indicates the approximate AP position where uniform Bcd expression coincides with graded Bcd (38%).

(B) Left: midsagittal NC13 EGFP-Bcd fluorescence (gray) in representative live embryos expressing either 2× EGFP-Bcd (top) or 1× tub>uEGFP-Bcd (bottom). Right: hybridization chain reaction images for hbP2-MS2-Gal4 (cyan) in either 2× EGFP-Bcd (top) or 1× tub>uEGFP-Bcd (bottom) fixed embryos. Nuclear DNA (DAPI) is shown in gray. Embryo lengths are approximately 500 μm.

(C) Onset time distributions for hbP2-MS2 measured in uniform Bcd (blue, $n=949$ nuclei) and graded Bcd (red, $n=413$ nuclei) as described in (A). The vertical dashed line indicates the computed AP axis position where the uniform matches the distribution of graded Bcd hbP2-MS2 onset times (37% EL). Plotted gray points indicate individual measurements from the uniform Bcd experiment only; lines represent rolling averages of onset times ($n=50\pm \mathrm{S}\mathrm{D}$) over the 95% CI of AP positions with active nuclei.

(D) Left: representative hbP2-MS2 reporter activity measured with hybridization chain reaction (cyan) in a wild-type (Kr/+) embryo (top) compared with reporter activity in a homozygous Kr mutant embryo (bottom), both staged at early NC14. Right: Kr transcript expression (red) in the same embryos pictured at left. DAPI staining is shown (gray).

See also Figure S3. Embryo lengths are approximately 500 μm.

Transcriptional repressors do not contribute to the hbP2-mediated expression domain

While it has been shown that the posterior boundaries of several Bcd targets are shaped by opposing repressor gradients,⁷ it is likely that the posterior boundary of hbP2 is set directly by a Bcd concentration threshold. To confirm that hbP2-MS2 onset times can read out the direct effects of changing Bcd concentrations at hbP2, we sought to rule out possible repressive feedback from maternal and zygotic factors in the patterning gene network. We assert that Capicua (Cic), Runt (Run), Knirps (Kni), and Krüppel (Kr), repressive TFs expressed in domains either overlapping or bordering anterior hb expression, do not contribute to hbP2 activity. cic mutants impact hb expression associated with the stripe regulatory element and not the early hb expression related to the P2/shadow enhancers.⁴⁴ Furthermore, Cic does not bind to the hb locus.⁴⁵ In addition, run mutants do not affect the hbP2 expression domain.⁷ kni and Kr mutants also do not significantly affect early hb expression that corresponds to P2 promoter activity.⁴⁶ While the hbP2 enhancer does not contain PWM matches to the Kni motif, we find a potential Kr binding site that overlaps the A3 Bcd motif on the opposite strand (Figure S3). To confirm that Kr does not regulate hbP2-MS2 expression, we measured the hbP2-MS2 reporter in Kr¹ mutant embryos. Loss of Kr does not change the position of hbP2-MS2 expression, demonstrating that transcriptional repression by Kr does not limit the posterior boundary of the hbP2 domain (Figures 3D and S3). Therefore, the Bcd concentration gradient likely serves directly as the primary source of positional information for hbP2-mediated transcription.

Modeling the determination of transcriptional onset times

Transcription in early development takes place during a time of intense proliferation, where both TF-nucleosome competition and features of the cell cycle can impact gene regulation. Nuclei export TFs and Pol II during mitosis and then reimport the factors in the following interphase, exemplified by the nuclear EGFP-Bcd concentration dynamics during NC13 (Figure 1E, green). Accessibility at hbP2 increases between mitosis 12 and the interphase of NC13, potentially reflecting Bcd’s competition with nucleosomes as nuclear Bcd levels rise (Figure 1F). Genome replication following mitosis could also influence the transcriptional process, as transcription in syncytial-stage Drosophila embryos requires the competition of DNA replication.²¹ We developed a mathematical model to investigate how these components of the dynamic genomic environment contribute to the Bcd concentration-sensitive expression of the hbP2-MS2 reporter.

The sensitivity of hbP2-MS2 onset time to Bcd concentration implies that Bcd concentration determines the rate at which the reporter’s promoter shifts from a transcriptionally inactive to an active state. We therefore built a two-state model of promoter activity that can simulate transcriptional onset times (Figure 4A). In this model, following mitosis, the promoter of the reporter starts the nuclear cycle in an “OFF” state where transcription cannot occur. As the nuclear cycle progresses, the promoter can transition to an “ON” state where Pol II transcribes the reporter. To account for the Bcd concentration sensitivity of the timing of transcriptional onset, we defined the rate at which a nucleus transitions from OFF to ON, $k_{on}$, as a function of Bcd concentration (Figure 4A). The simple two-state, one-transition structure of the model isolates the role of Bcd concentration in transcriptional activation. In the model, Bcd binding at hbP2 leads to an increase in $k_{on}$ through a mechanism described by $k_{on}\left(\right[Bcd\left]\right(x,t\left)\right)$. By simulating onset times with different forms of $k_{on}\left(\right[Bcd\left]\right(x,t\left)\right)$, we tested hypotheses about how the hbP2 enhancer governs the Bcd concentration sensitivity of hbP2-MS2 transcription. In the following, we assess whether different definitions of $k_{on}$—including a Michaelis-Menten formulation, Hill function, and a model with competitive nucleosome binding—allow the model to reproduce the measured hbP2-MS2 transcriptional onset time distribution. We first sought to determine whether Bcd-nucleosome competition alone regulates the hbP2-MS2 expression domain. We employ a version of Gillespie’s stochastic simulation algorithm to simulate the OFF-ON transition in individual nuclei (see Data S1: modeling supplement for details) and then compare simulated onset time measurements to the in vivo data.

Figure 4. An allosteric stochastic model for onset time prediction produces a posterior boundary for simulated hbP2-MS2 activity.

(A) Cartoon schematic of the initial simulation strategy demonstrated for two example nuclei at NC13. Two nuclei (A and B, lower left) encounter spatiotemporal differences in Bcd concentration throughout the cell cycle (top left, $\mathrm{B}\mathrm{c}\mathrm{d}(x,t)$). The transition to an “on” state is defined by $k_{on}$ and depends solely on whether Bcd binds to the target. Beginning at anaphase 12, $\mathrm{B}\mathrm{c}\mathrm{d}(x,t)$ is evaluated, and $k_{on}$ is posited for each nucleus at timepoint $t_i$ until a nucleus is scored as “on”. If, at timepoint $t,k_{on}$ yields a wait-time for binding less than the time-step ($\Delta t$), the nucleus is scored as “on” at time $t(t_{onset}$). If the calculated wait-time is greater than $\Delta t$, the timepoint is incremented to $t+\Delta t$, and the process repeats. This process yields a $t_{onset}$ for each nucleus in the field $x$.

(B) Onset times modeled with an open chromatin mechanism reflecting Michaelis-Menten kinetics. Gray points represent individual modeled onset times. The red trace is the rolling mean ± SD of modeled onset times over the 95% CI of AP positions with modeled active nuclei.

(C) Calculated fractional Bcd occupancy predicted by a TF-nucleosome competition model over the region covered by the proximal (brown) and distal (yellow) nucleosomes. The TF-nucleosome competition model was applied using $K_O=9\mathrm{n}\mathrm{M},K_N=2610\mathrm{n}\mathrm{M}$, and $L=700$ for each nucleosome and using $n=4$ and $n=5$ for the proximal and distal nucleosomes respectively.

(D) Onset times modeled with an allosteric mechanism for Bcd binding to hbP2. Gray points represent individual modeled onset times. The red trace is the rolling mean ± SD of modeled onset times over the 95% CI of AP positions with modeled active nuclei. This simulation was performed using $K_O=9\mathrm{n}\mathrm{M},K_N=2610\mathrm{n}\mathrm{M},L=700$, and $V_{max}=1$.

See also Figure S4 and Data S1.

TF-nucleosome competition allows the model to predict a Bcd concentration threshold

We assessed whether chromatin accessibility influences the Bcd concentration sensitivity of hbP2-MS2, first by ruling out that an interaction between Bcd and open chromatin could produce the observed onset time distribution. To simulate Bcd binding non-cooperatively to open chromatin, we defined $k_{on}$ according to Michaelis-Menten reaction kinetics (Data S1). We discuss this model in detail in Data S1: modeling supplement. We find that, with this definition of $k_{on}$, the model produces poor onset time predictions. The open chromatin model notably fails to predict a posterior boundary of expression and predicts transcriptional onset substantially prior to the 4.5-min onset time minimum observed in the in vivo data (Figure 4B; Data S1). The model’s simulation of onset times in the far posterior suggests that, if Bcd binding sites are entirely accessible, very-low Bcd concentrations can activate transcription. We performed parameter sweeps across one to three orders of magnitude and found that no tested parameter values allow the open-chromatin model to accurately simulate both the position and steepness of the hbP2-MS2 posterior boundary, as we describe in Data S1. As we expect no repressive contribution from other maternal or zygotic factors, we propose instead that the nucleosomes at hbP2 provide a barrier to transcriptional activation that delays the onset of transcription. While nucleosomes likely hinder Bcd binding to hbP2, Bcd binding to open chromatin through a pairwise cooperative mechanism could theoretically also produce a posterior boundary of target gene expression. We therefore tried defining $k_{on}$ according to the Hill equation for cooperative binding.⁴⁷ With this definition of $k_{on}$, our parameter sweeps revealed that no tested parameter combination allows the model to predict the position and steepness of the hbP2-MS2 expression domain boundary within 5% of their measured values (Data S1). Similar to prior work, we assert that pairwise protein-protein interactions cannot fully explain Bcd’s activity.^24,29,31,33 Data S1: modeling supplement presents the insufficiencies of both the open chromatin non-cooperative and cooperative models in more detail.

Given the failures of the open chromatin models, we incorporated a previously described model for TF-nucleosome competition into our two-state model’s definition of $k_{on}$.⁵ Mirny’s TF-nucleosome competition model takes the form of a Monod-Wyman-Changeux model of allostery and can estimate TF and nucleosome occupancies at a locus across different TF concentrations.^5,48,49 The TF-nucleosome competition model requires a short list of biophysical parameters: a TF’s affinity for its sites in the absence of a nucleosome ($K_O$), a TF’s affinity for its sites in the presence of a nucleosome ($K_N$), a nucleosome stability parameter ($L$), and the number of TF binding sites ($n$; Figure S4). We modeled Bcd’s competition with the proximal and distal nucleosomes using their respective numbers of Bcd binding sites ($n=4$ and $n=5$; Figure 1C) and left $K_O$, $K_N$, and $L$ as free parameters that describe Bcd’s competition with both nucleosomes. To estimate Bcd occupancy per site at the proximal ($Y_{prox}$) and distal ($Y_{dist}$) nucleosomes, we applied the TF-nucleosome competition model:

$$Y_i=\frac{\left[Bcd\right]}{K_O}\frac{{\left(1+\frac{\left[Bcd\right]}{K_O}\right)}^{n-1}+L\frac{K_O}{K_N}{\left(1+\frac{\left[Bcd\right]}{K_N}\right)}^{n-1}}{{\left(1+\frac{\left[Bcd\right]}{K_O}\right)}^n+L{\left(1+\frac{\left[Bcd\right]}{K_N}\right)}^n}.$$

Here, we show the fractional Bcd occupancies at each of the nucleosomes at hbP2 calculated with the TF-nucleosome competition model presented above (Figure 4C). We wished to incorporate these predicted fractional occupancies into our two-state stochastic model of transcriptional activation to predict hbP2-MS2 onset times. To build the influence of nucleosomes into the two-state model, we set $k_{on}$ as proportional to the probability that Bcd evicts both nucleosomes, $Y_{dist}*Y_{prox}$:

$$k_{on}=V_{max}\left(Y_{dist}\mathrm{*}{Y}_{prox}\right).$$

Defining $k_{on}$ according to Bcd-nucleosome competition allows the two-state model of transcriptional activation to predict a posterior boundary of hbP2-MS2 expression (Figure 4D). With this change to $k_{on}$, we now refer to the two-state model as the allosteric model. We estimated best-fit values for the parameters $K_O$, $K_N$, and $L$ by sweeping across ranges of possible values and evaluating simulation outputs (see Data S1 for details). We find that the best fit parameters fall within the ranges of prior measurements or reasonable estimates and successfully simulate a posterior expression boundary that aligns with in vivo observations (Figure 4D; Data S1). The combination of $K_O=9\mathrm{n}\mathrm{M},K_N=2610\mathrm{n}\mathrm{M}$, and $L=700$ predicts the posterior boundary position of our hbP2-MS2 measurements, deviating by only 0.26% EL, well within the error of our measurements (Figure 4D). $K_O=9\mathrm{n}\mathrm{M}$ falls within the range of published estimates of Bcd’s affinity for its sites in vitro.^15,50–52 $K_N=2610\mathrm{n}\mathrm{M}$ and $L=700$ are within the ranges approximated in previous work, specifying $K_O/K_N=0.1–0.001$ and $L=100–1000$.⁵ As the allosteric model predicts the data’s posterior boundary with reasonable parameter estimates, these results support the hypothesis that nucleosomes play a regulatory role at hbP2. We conclude that, at low Bcd concentrations, nucleosomes prevent Bcd from binding hbP2 and activating transcription.

Nucleosome stability modulates the Bcd concentration threshold

We tested the conclusion that Bcd-nucleosome competition sets a concentration threshold for transcription by measuring and then modeling the hbP2-MS2 expression boundary in conditions of altered nucleosome stability. To alter nucleosome stability, we manipulated the activity of the TF Zld at the reporter, as Zld can disrupt nucleosomes and increase chromatin accessibility upon binding to a locus.^53,54 Prior ATAC-seq data show a slight reduction in chromatin accessibility at the hbP2 enhancer in zld germline clone mutants (Figure 5A). Furthermore, the hbP2 enhancer contains a single Zld motif occluded by a NuPoP-predicted nucleosome (Figure 5B). We, therefore, propose that Zld indirectly contributes to hbP2-mediated transcription by influencing nucleosome stability and the dynamics of Bcd-nucleosome competition. We first tested the effects of nucleosome stabilization by characterizing hbP2-MS2 expression in Zld knockdown embryos. We find that, in zld-RNAi embryos, the posterior boundary of the fraction of hbP2-MS2 active nuclei is shifted slightly toward the anterior (Figure 5C, dark blue). We next tested the effects of nucleosome destabilization by increasing the recruitment of Zld to the hbP2-MS2 reporter by adding a Zld motif to hbP2, thereby creating a [hbP2 + 1xZld] MS2 reporter (Figure 5B). As indicated by the fraction of active nuclei, we find that [hbP2 + 1xZld] MS2 is expressed farther to the posterior than the wild-type hbP2-MS2 reporter (Figure 5C, green). Additionally, the Bcd concentration sensitivity of the reporter’s onset time distribution decreases with increases in Zld activity, from the −Zld to wild-type to +Zld conditions (Figure S5). Our experimental results point to a correlation between nucleosome stability and the Bcd concentration threshold required for transcription.

Figure 5. Zelda-dependent changes in reporter activity are approximated by altering modeled nucleosome stability.

(A) ATAC measurements of accessibility and modeled nucleosome dyad positions for wild-type (top) or zelda mutant embryos (bottom). Wild-type data are replotted from Figure 1B for comparison. The orange arrowhead marks a slight increase in the occupancy of the proximal nucleosome.

(B) Schematic of hbP2 minimal elements highlighting the positions of Zelda binding sites (orange) within the domain of the proximal nucleosome (below) and relative to Bcd binding sites (blue). Top: enhancer of the hbP2-MS2 reporter. Bottom: enhancer of the [hbP2 + 1x Zld] MS2 reporter.

(C) The fraction of active nuclei in 2.5% egg length bins for hbP2-MS2 in wild-type embryos (light blue ± SD, $n=19$) and zelda-RNAi embryos (dark blue ± SD, $n=15$), as well as for [hbP2 + 1x Zld] MS2 in wild-type embryos (green ± SD, $n=17$). Dots mark the EC50 of each of the fraction of active nuclei profiles determined by Hill equation fits (Data S1).

(D) The posterior boundary positions (x axis) that are simulated by the model with specific $L$ values (y axis), as determined by the EC50s of Hill fits to the simulated fraction of active nuclei (Data S1). The solid black dot indicates the maximal AP position achievable by the model when $L$ is set to the lower bound estimate of 50. Colored dots mark the measured posterior boundaries of hbP2-MS2 in wild-type (light blue ± SD), hbP2-MS2 in zelda-RNAi (dark blue ± SD), and [hbP2 + 1x Zld] MS2 in wild-type (green ± SD). $L=640,L=900$, and $L=280$ allow for prediction of these measurements, respectively.

See also Figure S5 and supplemental information: Data S1.

We applied the allosteric two-state model to examine whether alteration of modeled nucleosome stability could result in shifts in the modeled posterior boundary of hbP2-MS2 expression. In the model, the dimensionless parameter $L\left(\right[{\mathrm{N}\mathrm{u}\mathrm{c}}_{0\mathrm{T}\mathrm{F}}]/[{\mathrm{O}\mathrm{p}\mathrm{e}\mathrm{n}}_{0\mathrm{T}\mathrm{F}}\left]\right)$ represents the equilibrium between the nucleosomal and open states while no TFs are bound to the locus.⁵ $L$ is termed the “nucleosome stability parameter,” as higher-stability nucleosomes will more often occupy DNA, increasing the nucleosome equilibrium occupancy. We estimated a lower bound for $L$ as 50–60 based on mitotic ATAC-seq measurements (e.g., Figure 1F, top) of genome-wide fractions of nucleosome-associated and open chromatin. Our estimate agrees well with prior lower bound estimates of human genome accessibility (range 10–100) based on DNase hypersensitivity measurements.⁵ We approximate the upper bound of $L$ as 1000, a value previously calculated from nucleosome occupancy measurements.⁵ The results of a parameter sweep from $L=50$ to 1000 reveals that, as $L$ increases, the predicted posterior boundary of expression shifts toward the anterior (Figure 5D). Changing $L$ allows the model to simulate the posterior boundaries observed in each of our experimental datasets while holding all other parameters at their best-fit values ($K_O=9\mathrm{n}\mathrm{M},K_N=2610\mathrm{n}\mathrm{M}$, and $V_{max}=1$; Figure 5D). The model requires a high nucleosome stability value to simulate the effects of decreasing Zld activity (Figure 5D, dark blue), while it requires a low nucleosome stability value to simulate the effects of increasing Zld activity (Figure 5D, green). An intermediate nucleosome stability value of $L=640$ predicts the boundary of hbP2-MS2 in wild-type (Figure 5D, light blue). Therefore, the range of $L$ seems to define a range of possible posterior boundary positions for a given arrangement of Bcd motifs and nucleosomes. The success of the model in simulating the results of experimentally perturbing Zld activity supports the conclusion that nucleosome stability determines the degree to which transcription depends on Bcd concentration.

DNA replication delays the onset of transcription at high Bcd concentrations

While defining $k_{on}$ according to Bcd-nucleosome competition allows the model to simulate the positioning of the posterior boundary of hbP2-MS2 expression, the model (Figure 4D) notably fails to accurately predict the observed timing of transcriptional onset in the anterior (Figure 2D). The model underestimates both the mean and variance of the onset times at high Bcd concentrations. Regardless of $K_O,K_N$, and $L$ combination tested, a portion of nuclei simulated in the anterior still initiates transcription prior to the 4.5-min onset time minimum we observe in vivo (Data S1). A substantial increase in $K_N$ or $L$ does not improve anterior onset time predictions (Data S1). Increasing $K_O$ slightly increases the predicted onset time variance in the anterior but results in an anterior shift of the predicted posterior boundary of expression that reduces the overall accuracy of the simulation (Data S1; Figure 2D). The discrepancies between the simulations and the in vivo data motivated us to dig deeper into the mechanism of concentration-sensitive transcriptional regulation. DNA replication following mitosis 12 could delay transcriptional activation, as transcription in early Drosophila embryos requires the completion of DNA replication.²¹ We therefore sought to estimate the impact of DNA replication on the OFF-ON transition of the two-state model to test whether DNA replication contributes to the timing and variance of hbP2-MS2 onset times.

Active replication origins are thought to be specified at random positions in syncytial embryos.⁵⁵ Therefore, between nuclei A and B from the same embryo, a fixed genomic position (e.g., a reporter) will have variable distances to flanking origins (ori₁, ori₂) and will therefore complete replication and—by extension—become competent to be transcribed at different times (Figure 6A, left panel, after Blumenthal et al.⁵⁵). To account for this effect, we designed a new two-state model where switching to the ON state requires both that Bcd binds the reporter and that the promoter completes replication (Figure 6A, right panel). This replication-dependent model still estimates Bcd binding using the nucleosome-dependent $k_{on}\left(\right[Bcd\left]\right(x,t\left)\right)$, as described in Figure 4. However, if Bcd evicts nucleosomes prior to the completion of replication at the promoter, the promoter does not transition to ON at that time, and the simulation sets the onset time to the DNA replication time (Figure 6A, right panel). If replication occurs before Bcd evicts the nucleosomes, then Bcd binding instead dictates the promoter’s transcriptional onset time (Figure 6A, right panel). See Data S1 for simulation details.

Figure 6. Accounting for the influence of DNA replication on transcription improves the allosteric model’s predictions.

(A) Cartoon schematic of the revised simulation strategy demonstrated on two example nuclei. The cartoon on the left shows a single genomic locus in two different nuclei for time points spanning DNA replication, initiated locally at two randomly positioned origins (ori₁ and ori₂, after Blumenthal et al.⁵⁵). Due to the dependency of transcriptional elongation on completion of DNA replication, random origin spacing is expected to impact observed onset times. On the right is a revised model schema, which now requires the enhancer both to bind Bcd and to complete replication prior to switching to the ON state.

(B) A Gamma distribution with a 9.7-kb average inter-origin spacing and origin frequency of 2 origins per 9.7 kb, plotted as a histogram (gray). Measured inter-origin distances were manually measured from Figure 4 of Blumenthal et al.⁵⁵ and reproduced here (purple).

(C) Modeled replication delays calculated from the inter-origin distance model.

(D) Onset times simulated by the allosteric + replication model. A rolling average of simulated onset times ($n=50$) is shown (blue ± SD) across the 95% CI of the AP positions of simulated active nuclei. The simulation was performed with $K_O=9\mathrm{n}\mathrm{M},K_N=2810\mathrm{n}\mathrm{M}$, and $L=600$.

(E) A set of sampled hbP2-MS2 onset times as a function of AP position, with the rolling average of observed onset times (blue ± SD), the average of 10000 onset time simulations (red, dotted) using the allosteric + replication model, and the average of 10000 onset time simulations with the allosteric-only model (black, dotted). Observed and simulated onset time averages are plotted across the 95% CI of the AP positions of active nuclei (observed and simulated, respectively). Simulations were performed with $K_O=9\mathrm{n}\mathrm{M},K_N=2810\mathrm{n}\mathrm{M}$, and $L=600$ for both the replication-dependent and independent models.

(F) The SD of onset times as a function of AP position for in vivo measurements (blue), and the average SD of 10000 simulations of the allosteric + replication (red) or allosteric-only (black) models.

(G) One instance of the simulation (the same as in (D)), with points color coded to indicate whether the onset time was limited by either replication time (blue) or Bcd concentration (red).

The model estimates DNA replication timing by drawing from a modeled distribution of inter-origin distances and calculating the time to complete replication. To model the replication timing of the reporter, we needed to estimate four parameters: global timing of replication initiation, the speed of DNA polymerase, the relative spacing of active replication origins, and the distribution of origin firing times. We estimated the timing of replication initiation to be 3.75 min following anaphase 12 based on prior imaging of PCNA-GFP, which forms foci at sites of active replication whose intensities peak at this time.^20,56 The speed of DNA polymerase was previously estimated to be 5.3 kb/min for bidirectional replication from an origin.⁵⁵ Next, while no genomic studies have mapped origin positions specifically in cleavage-stage Drosophila embryos, the average spacing between origins has been approximated to be 9.7 kb by measuring between replication fork centers observed on electron micrographs. Finally, the statistical distribution of DNA fibers with active replication forks was found to be consistent with synchronous origin firing.⁵⁵ Based on these reported measurements, we estimate that $x$ kb between origins, with average spacing of 9.7 kb, takes $x/5.3$ min to complete replication following initiation at 3.75 min past anaphase 12.

We modeled the likelihood of origin-origin distances by using a gamma distribution with $\mathrm{\alpha }=2$ origins and $\mathrm{\beta }=9.7$ kb/2 origins to estimate the probability that stretches of DNA of length $X$ contain a pair of origins:

$$x\sim \mathrm{\Gamma }\left(\alpha =2origins,\beta =\frac{9.7kb}{2origins}\right).$$

By KS test, we find that the observed origin distribution measurements are indistinguishable from this gamma distribution model at a significance threshold of $p=0.05$ (Figure 6B; Data S1). A gamma distribution describes the waiting times, or in this case distances, before a specified number of events in a Poisson process. The success of the gamma distribution in modeling the measured origin-origin distances supports the assumption of random distribution of origins along chromosomes, with origins activating according to a Poisson process with rate $\lambda =2$ origins/9.7 kb. By dividing our modeled origin-origin distances by 5.3 kb/min and accounting for the 3.75-min delay before initiation of DNA replication following anaphase 12, we calculated a distribution of possible times at which the reporter’s promoter completes replication (Figure 6C). This replication time distribution allows the two-state model of promoter activity to account for transcription’s requirement for DNA replication.

We find that the allosteric + replication model of transcriptional activation successfully recapitulates the measured hbP2-MS2 onset time distribution (Figures 6D and 6E). The OFF-ON transition’s requirement for the completion of DNA replication at the promoter introduces additional variance into the simulated onset times at high Bcd concentrations that better reflects the variance observed in our hbP2-MS2 data (Figure 6F). Furthermore, the requirement for DNA replication prevents anterior onset time predictions substantially prior to 5 min into NC13. The new model structure also accurately predicts the posterior boundary of MS2 expression; the model’s best-fit parameter set contains Bcd-nucleosome competition parameters that place the posterior boundary of simulated expression at 45% EL ($K_O=9\mathrm{n}\mathrm{M},K_N=2810\mathrm{n}\mathrm{M}$, and $L=600$; Figures 6D and 6E). The model suggests that DNA replication delays transcriptional onset at high Bcd concentrations where Bcd readily outcompetes nucleosomes, while at lower Bcd concentrations, the requirement that Bcd outcompetes nucleosomes dictates the timing of transcriptional onset (Figure 6G).

Genome replication and Bcd binding regulate RNA polymerase II pause-release

The model assumes that, following mitotic exit, Bcd can compete with nucleosomes and bind to DNA independently of the completion of DNA replication. An alternative possibility, however, is that DNA replication could itself lower the nucleosome barrier for TF binding, as passage of the replication fork disrupts nucleosomes and transiently reduces the nucleosome content of DNA by half.^57,58 A second assumption is that transcription becomes immediately detectable following the completion of both Bcd binding and DNA replication and that, therefore, the recruitment of RNA Pol II to the promoter likely occurs independently of these processes (Figure 6A). The transition of RNA Pol II from transcriptional initiation to elongation, akin to a pause-release, may require Bcd binding and the completion of DNA replication. We experimentally tested the following hypotheses that stem from these assumptions: (1) Bcd binding to hbP2 occurs independently of DNA replication, (2) Pol II recruitment to hb occurs independently of DNA replication, and (3) Pol II recruitment to hb occurs independently of Bcd binding.

To determine how DNA replication impacts Bcd-dependent transcription, we measured Bcd and Pol II occupancy in control and DNA replication-inhibited embryos. To collect DNA replication-inhibited embryos for ChIP-seq, we developed a protocol for treating embryos in bulk with hydroxyurea (HU) to block their synthesis of dNTPs (Figure S6). ChIP-seq measurements do not reveal any differences in Bcd binding at hbP2 between control and HU-treated embryos (Figure 7A, blue). Genome-wide Bcd binding also does not change substantially upon inhibition of DNA replication (Figure 7B). Therefore, Bcd binds to DNA independently of DNA replication state, and its ability to outcompete nucleosomes likely does not depend on the disruption of chromatin by replication forks. In contrast, RNA Pol II coverage across gene bodies is strongly affected when DNA replication is inhibited. At the hb locus, RNA Pol II is markedly reduced over the gene body, yet retains a small peak at the P2 transcriptional start site, consistent with a replication-sensitive block to the transition between initiation and elongation (Figure 7A, red). Genes expressed prior to large-scale zygotic genome activation (“pre-MBT” genes⁵⁹) also exhibit a higher ratio of initiated vs. elongating Pol II upon replication inhibition (Figures 7C–C′ and S6). These measurements confirm the central assumptions of our model and indicate that the completion of DNA replication precedes the transition from Pol II initiation to elongation.

Figure 7. Initiated RNA Pol II waits for the completion of replication and Bcd binding to enter into productive elongation of hb.

(A) ChIP-seq data over the hb locus for RNA Pol II (dark red), Bcd (blue), and control (gray) comparing occupancy at NC13 in control (untreated) and HU-treated embryos. Data are the average of two independent biological replicates, normalized to read depth (CPM).

(B) The average standardized Bcd ChIP-seq data for control (dark blue) and HU-treated (light blue) embryos over 1026 peak regions.

(C) Upper panels: heatmap representation of standardized RNA Pol II ChIP-seq signal over a set of genes transcribed before large-scale ZGA (pre-MBT genes).⁵⁹ Lower panel (C′): the HU-treatment (red) and control (dark blue) averages of the heatmap representation.

(D) ChIP-seq data (average of $n=3$ biological replicates, normalized to CPM) for initiated RNA Pol II (pSer5) and control in wild-type and bcd osk tsl blastoderm-stage embryos. See also Figures S6 and S7.

We next evaluated the independence of Pol II initiation from Bcd binding by performing Pol II ChIP-seq in wild-type embryos and embryos triply mutant for the maternal factors bicoid (bcd), oskar (osk), and torso-like (tsl). We chose to use bcd osk tsl rather than bcd^E1 embryos, as the triple mutant context eliminates AP and terminal maternal factors that contribute to the regulation of anterior and posterior hb stripe enhancers, reducing confounding Pol II activity at hb. In triple mutant embryos, we find that while Pol II is lost from the hb gene body, Pol II remains associated with the P2 promoter (Figure 7D). Nearly every Bcd target we assess also maintains Pol II at its promoter in the triple mutant condition (Figure S7). The preservation of Pol II at promoters indicates that transcriptional initiation at hb is independent of Bcd activity. The regulation of Pol II pause-release likely underlies how Bcd can rapidly initiate transcription within the short cell cycles of early development.

DISCUSSION

The molecular mechanisms that determine the concentration thresholds of Bcd target genes have been pursued since the discovery of Bcd in 1988.^{7–9,11,14–16} Here, we have illustrated that the pattern of Bcd binding sites and nucleosomes at an enhancer uniquely determines a Bcd concentration threshold for transcription and have provided a generalizable framework for understanding Bcd concentration sensitivity. With physically reasonable parameter values, our computational model of promoter activity accurately simulates how the timing of hbP2-MS2 transcriptional onset changes across Bcd concentrations. Our theoretical and experimental work asserts that Bcd-nucleosome competition sets a concentration threshold for transcription, while DNA replication delays transcriptional onset at high Bcd concentrations, where Bcd “wins” its competition with nucleosomes. We experimentally validate the model structure and conclude that Pol II elongation on a Bcd target gene following mitosis requires both that Bcd outcompetes nucleosomes for occupancy at an enhancer and that an associated promoter completes DNA replication. This hbP2 case study provides a molecular explanation for how the in vivo environment of early development shapes the formation of concentration-sensitive patterning gene expression domains.

TF-enhancer interactions must be understood in the context of chromatin and the cell cycle. Nucleosomes act as barriers to TF binding to genomic loci, and rapid cell divisions limit transcriptional outputs early in the development of many organisms.^60–63 This work fills in gaps in understanding highlighted by prior models that describe in the abstract how the maternal factors Bcd and Zld regulate a promoter’s OFF-ON transition.^23,25,64 A prior model found that Zld shortens the duration of inactive promoter states preceding the transcription of a reporter for a dorsal-ventral (DV) patterning enhancer.⁶⁴ Zld similarly influences the OFF-ON transition in our model, pointing to the conclusion that the previously characterized inactive promoter states correspond to high degrees of nucleosome occupancy at the enhancer. At both DV and AP target genes, Zld likely speeds the transition between inactive states by acting as a pioneer factor and by reducing nucleosome stability at enhancers. To account for the delays preceding transcription of a hbP2 + P2 promoter reporter, another model incorporated irreversible transitions through inactive promoter states driven by Bcd and Zld.²³ Here, we specify that the molecular mechanisms that account for the irreversible OFF-ON transition are Bcd-nucleosome competition and the dependence of Pol II elongation on the completion of DNA replication. Recently, a model based on synthetic enhancer measurements suggested that Zld speeds a promoter’s OFF-ON transition by increasing the rate of Bcd binding.²⁵ This finding fits within the theoretical framework we present, as Zld’s impact on nucleosome stability allows Bcd to outcompete nucleosomes more easily and bind its sites. Importantly, our work specifies that at high TF activity levels, the timing of DNA replication at a locus sets an upper limit for the rate of transcriptional activation.

This work aligns with expectations based on prior observations of Bcd dynamics in the nucleus and the formation of clusters at presumed binding sites.^65–68 A broad range of TFs have been observed to form high-concentration assemblies or clusters within the nucleus.^64–70 In the case of Bcd, while clusters contain only a small fraction of total protein, they show strong spatial and temporal correlation with target gene transcription as measured with an MS2 reporter.^67,68 How the frequencies of cluster formation and dispersal impact the dynamics of Pol II activity at a promoter remains an open question. A recent study estimated the number of Bcd molecules within clusters coinciding with the hbP2 expression domain to be greater than or equal to the number of expected binding events at hbP2 (≥9).⁶⁸ In light of our work, we propose that Bcd clustering allows for rapid delivery of the high local concentrations of Bcd necessary to outcompete nucleosomes for occupancy at the nine Bcd sites in hbP2. Therefore, the frequency of cluster formation provides an upper bound for the rate at which Bcd molecules can outcompete the nucleosomes at a locus. As our model’s OFF-ON transition rate depends on Bcd’s ability to outcompete nucleosomes, it provides a framework for translating TF cluster dynamics to Pol II behavior at a promoter. We propose that periodic TF cluster formation regulates the transcriptional process by incrementally reorganizing the chromatin at enhancers.

Finally, this work raises the possibility that the input-output relationship between any TF and a target gene could be modeled accurately, given nucleosome positions measured both in the presence and absence of key regulators. In the case of Bcd, we expect that this model will, with elaboration, explain the mechanisms producing the concentration thresholds of additional target genes. Other physiologically important Bcd targets receive competitive input from repressors and undergo more extensive chromatin reorganization by pioneers such as Zld.^7,15 With additional work, an allosteric modeling framework could incorporate the activity of repressors and pioneers that—either directly or indirectly—modulate an activator’s ability to access its sites. Further, we conceptualize nucleosome occupancy as a means to create suboptimal, effectively low-affinity TF binding conditions. Our model therefore supports previous observations of the importance of low-affinity TF binding motifs for spatially restricted enhancer activity in response to graded and uniform TF inputs.^71–73 We hypothesize that, across organisms and developmental contexts, nucleosomes play key gene regulatory roles by creating conditions that require high TF concentrations for TF-DNA binding.

Limitations of the study

The parameter ranges of our models agree well with prior estimates or direct measurements, when available. One parameter, the affinity of Bcd to a nucleosomal target, has not been measured directly; our model therefore assumes a low-affinity interaction between Bcd and nucleosomal DNA, consistent with expectations for most TFs.⁷⁴ It will be important in the future to directly measure this biophysical parameter. We also rely on an indirect approach to modulate nucleosome stability. Direct modulation of nucleosome stability while maintaining relevant features of DNA sequence for TF binding is challenging and could be addressed in future work through the design of synthetic cis-regulatory elements. Finally, we use HU to inhibit DNA replication in bulk embryo collections. To our knowledge, this is the only pharmacological replication inhibitor that can be delivered to permeabilized embryos. Our HU-treated ChIP result shows the population average of embryos with at least a partial inhibition of replication and therefore reflects a lower-bound estimate of the reliance of RNA Pol II on replication status.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources should be directed to the lead contact, Shelby A. Blythe (shelby.blythe@gmail.com).

Materials availability

Fly stocks generated in this study are available from the lead contact upon request.

Data and code availability

• Image analysis and modeling code is available through GitHub and archived through Zenodo: https://doi.org/10.5281/zenodo.15832329.

• Newly generated ChIP-seq datasets have been deposited to GEO (GSE283996).

• Any additional information required to reproduce the data reported in this paper is available from the lead contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Drosophila stocks

This study used the following Drosophila melanogaster (NCBI Taxon 7227) stocks.

w¹¹¹⁸ (source: BestGene Inc., Chino Hills, CA).

y {w⁺= His2Av-miRFP670-T2A-HO1}ZH2A w (this study)

w; {w ⁺ = RpA-70-EGFP}attP2 ¹⁹

bcd^E2 osk¹⁶⁶ tsl⁴/TM3

yw; P{w = his2AV-RFP}2; P{w = MCP-GFP}4F (kind gift of H. Garcia)²³

y w; {w⁺ = nos>MCP-mCherry}VK22 (kind gift of H. Garcia)

cn bw Kr¹/SM6 bw (Bloomington Drosophila Stock Center (BDSC) #3494)

w; UASp>zldRNAi (III) (kind gift of C. Rushlow)⁵⁴

w; P{w = alpha-tubulin67c-GAL4 VP16}67 (Princeton Stock Collection)

w; bcd^E1/TM3 (Princeton Stock Collection)

w P{w = His2Av-RFP}1; bcd^EGFP (kind gift of P. Onal and S. Small)

All fly stocks were maintained on an enriched high-agar cornmeal media as reported previously.⁷⁵

Transgenic lines

A hbP2-MS2 reporter was made using the hbP2 minimal enhancer, 245 bp (chr3R:8,694,654–8,694,898, dm6 reference assembly) from the 263 bp hb regulatory element identified previously.¹¹ The hbP2 sequence was cloned upstream of the Drosophila Synthetic Core Promoter⁷⁶ which drives expression of 24 MS2(v5) stem loop sequences⁷⁷ embedded in the intron of yellow and followed by the Gal4 sequence in an attB transgenic vector. The reporter was integrated into the attP40 landing site on the second chromosome using PhiC31 integrase-mediated transgenesis. A [hbP2 + 1xZld]-MS2 reporter was made by introducing an optimal zld motif to base pair positions 178–185 within hbP2, a location that does not disrupt any of the TF binding sites in hbP2 identified with PWM matches. Besides the zld motif alteration to hbP2, the [hbP2 + 1xZld]-MS2 line was made through an identical cloning and integration process to that performed to make hbP2-MS2.

A transgenic line that produces embryos that express Bcd uniformly was achieved as follows. First, a w; αTub67C-EGFP-Bcd/^FRTbcd 3′UTR 3xP3>dsRed/^FRTsqh 3′UTR/TM3 (III) transgenic line was made by inserting a previously published uniform Bcd construct¹⁵ into the VK33 landing site on the third chromosome using PhiC31 integrase-mediated transgenesis. The αTub67C-EGFP-Bcd/^FRTbcd 3′UTR 3xP3>dsRed/^FRTsqh 3′UTR transgene was then recombined with the bcd^E1 allele to produce a αTub67C-EGFP-Bcd/^FRTbcd 3′UTR 3xP3>dsRed/^FRTsqh 3′UTR, bcd^E1 recombinant. Uniform Bcd expression was achieved as described previously¹⁵ by excising the FRT-flanked bcd 3′UTR with transient induction of Flp expression and scoring for loss of the 3xP3>dsRed marker. This yields αTub67C-EGFP-Bcd sqh 3′UTR, bcd^E1, which expresses Bcd uniformly is henceforth referred to as αTub67C-uEGFP-Bcd, bcd^E1.

Datasets

The ChIP- and ATAC-seq data for wild-type versus uniform Bicoid, including predicted nucleosome dyad positions were previously reported (GEO: GSE86966).¹⁵ ATAC-seq data from different stages of the NC13 cell cycle were previously reported (GEO: GSE83851).²⁰ The processed Bcd ChIP-nexus data demonstrating positive and negative sequence contributions to Bcd binding⁴² was retrieved from https://zenodo.org/record/8075860 and plotted over the hb genomic locus.

Raw reads and coverage data for the HU and the bcd osk tsl ChIP experiments performed for this study have been deposited to GEO (GEO: GSE283996).

METHOD DETAILS

Live-imaging

MS2 reporters

Embryos for hbP2-MS2 and [hbP2 + 1xZld]-MS2 imaging: Embryos were collected from a cross of hbP2-MS2 or [hbP2 + 1xZld]-MS2 males to yw/w; his2AV-RFP/+; MCP-GFP/+ females. Embryos for hbP2-MS2 in uniform Bcd imaging: Embryos were collected from a cross of hbP2-MS2 males to his-miRFP670/+; MCP-mCh/+; αTub67C-uGFP-Bcd, bcd^E1/bcd^E1 females. Embryos for hbP2-MS2 in zld^RNAi imaging: Embryos were collected from a cross of hbP2-MS2 males to his2AV-RFP/+; mat-tub-GAL4 67/+; UAS-shRNA-zld/+ females. The imaging of MS2 transcription in all genotypes was performed as follows.

Pre-NC13 embryos were dechorionated in 4% bleach (50% dilution of concentrated Clorox) for 1–2 min. Dechorionated embryos were mounted in Halocarbon 27 oil (Sigma) on a gas-permeable membrane (BioFoil, Heraeus, New Jersey, USA) and overlaid with a glass coverslip. Embryos were imaged on a Leica SP8 WLL Confocal Microscope using a 63× 1.3NA glycerol objective, 1.28× zoom, and 700 Hz scan rate at a 512×256 pixel resolution. When imaging MS2 in graded Bcd embryos, a 590 nm laser (measured as 15 μW through the 10× objective) and a 488 nm laser (measured as 42 μW through the 10× objective) were used to excite RFP and GFP respectively. RFP and GFP emissions were collected at [602, 650] nm (detector set to BrightR) and [498, 531] nm (detector set to Photon Counting) respectively. When imaging MS2 in uniform GFP-Bcd embryos, a 488 nm laser (measured as 29 μW through the 10× objective) was used to excite GFP-Bcd, a 587 nm laser (measured as 18 μW through the 10× objective) was used to excite MCP-mCh, and a 670 nm laser (measured as 56 μW through the 10× objective) was used to excite His2Av-miRFP670. GFP-Bcd emission was collected at [498, 535] nm (detector set to Photon Counting), MCP-mCh emission was collected at [597, 629] nm (detector set to Photon Counting), and his-miRFP670 emission was collected at [680, 720] nm (detector set to BrightR). As the maximum laser outputs of confocal microscopes can fluctuate day-to-day, the % outputs of the lasers were adjusted prior to each imaging session to ensure that embryos were imaged with equivalent laser powers between imaging sessions. For all genotypes, per embryo, a 10 μm z stack was imaged at 10 s/frame and with a 0.5 μm z-step from pre-NC13 to at least 15 min into NC14. Immediately following live-imaging of MS2 in graded Bcd embryos, an overview image of the embryo was taken by performing a tile scan using 590 nm excitation at 700 Hz and [602, 650] nm detection with a top z-plane intersecting the center of nuclei and a bottom z-plane at the embryo’s midsagittal plane. Tiles were merged into a single image on the Leica software. Immediately following live-imaging of MS2 in uniform GFP-Bcd, an overview image was taken by following an equivalent procedure, with the exception of using a 488 nm laser and 200 Hz scan speed for exciting GFP and detecting emission at [498, 535] nm.

EGFP-bicoid

For live-imaging His2AV-RFP; bcd^EGFP embryos, pre-NC13 embryos were dechorionated and mounted for imaging through the aforementioned MS2 imaging method. Embryos were imaged on a Leica SP8 WLL Confocal Microscope using a 63× 1.3NA glycerol objective, 1.28× zoom, and 700 Hz scan rate at 512×256 pixels. A 590 nm laser (measured as 19 μW through the 10× objective) and a 488 nm laser (measured as 28 μW through the 10× objective) were used to excite RFP and GFP respectively. RFP and GFP emissions were collected at [602, 650] nm (detector set to BrightR) and [499, 531] nm (detector set to Photon Counting) respectively. Per embryo, a 10-μm z stack was imaged at 10 s/frame and with a 0.5 μm z-step from pre-NC13 to at least 15 min into NC14.

Hybridization chain reaction

Collection

To allow for measurement of hbP2-MS2 in Kr mutant embryos, the hbP2-MS2 reporter chromosome was recombined with cn bw Kr¹ to generate a hbP2-MS2 Kr¹/SM6 line. Embryos were collected from a cross of hbP2-MS2 Kr¹/SM6 females to hbP2-MS2 Kr¹/SM6 males after a laying period of 4 h on an apple juice-agarose plate. The mutant embryo genotype produced was hbP2-MS2 Kr¹/hbP2-MS2 Kr¹ and the control embryo genotype was hbP2-MS2 Kr¹/SM6. For measurement of hbP2-MS2 in embryos uniformly expressing Bcd, embryos were collected from a cross of αTub67C-uEGFP-Bcd, bcd^E1/bcd^E1 females to hbP2-MS2 males after a laying period of 4 h on an apple juice-agarose plate. Control embryos were similarly collected for the uniform Bcd experiment from a cross of white females to hbP2-MS2 males.

Fixation

Embryos were dechorionated for 1 min in 4% bleach (from an 8% concentrated stock) and then added to a combination of 1 mL 20% formaldehyde, 4 mL 1x Phosphate Buffered Saline (PBS), and 5 mL heptane. After the embryos were shaken for 20 min, the aqueous layer was removed, methanol was added, and the embryos were vortexed for 30 s to remove their vitelline membranes. The embryos were then washed 3x in methanol and stored in methanol at −20 C for at least overnight before undergoing the staining protocol. For the Kr mutant experiment, mutant and control embryos were fixed together in the same tube, as they were produced by the same cross and can be distinguished by the presence of Kr mRNA. For the uniform Bcd experiment, mutant and control embryos were fixed separately.

Probes and amplifiers

The following RNA probes were synthesized by Molecular Instruments: a probe that targets Gal4 mRNA and is compatible with B1 amplifier sets and a probe that targets Kr mRNA and is compatible with B3 amplifier sets. B1–488 nm, B1–647 nm, and B3–546 nm RNA hairpin-fluorophore conjugates were synthesized by Molecular Instruments and used as amplifiers.

Staining

The Hybridization Chain Reaction (HCR) In Situ Protocol V.1⁷⁸ was performed on fixed embryos with the following modifications. For the embryo permeabilization, embryos were permeabilized for 30 min at room temperature rocking in a detergent solution (0.1% Triton X-100, 0.05% Igepal CA-630, 500 μg/mL sodium deoxycholate, 500 μg/mL saponin, 2 mg/mL BSA Fraction V). In our hands, use of SDS for embryo permeabilization as recommended in the cited protocol resulted in undesirable expansion of the embryos and difficulty in downstream handling. We substituted the above detergent mix without loss of signal intensity as determined in a pairwise comparison. For the probe incubation, the embryos were incubated for 17 h at 37°C in a solution of probe hybridization buffer, 0.8 pmol of Gal4-B1 probe, and 0.8 pmol of Kr-B3 probe (Kr mutant experiment), or for 17 h at 37°C in a solution of probe hybridization buffer and 0.8 pmol of Gal4-B1 probe (uniform Bcd experiment). At the amplification step, an amplification reaction was performed by incubating the embryos in 0.06 μM of each of the B1–488 nm and B3–546 nm hairpin sets for 24 h at room temperature (Kr mutant experiment) or in 0.06 μM of the B1–647 nm hairpin set for 18 h at room temperature (uniform Bcd experiment). Following washes according to Bruce et al., 2021, embryos were stained with 1 μg/mL DAPI (4′,6-diamidino-2-phenylindole; Invitrogen REF: D1306) in 50% glycerol (in 1xPBS) rocking for 1 h at room temperature. The DAPI solution was replaced with 50% glycerol (in 1xPBS) and the embryos stored at 4°C prior to mounting.

Imaging

Embryos were mounted in Prolong Gold Antifade Mountant (Invitrogen P36934) and overlaid with a glass coverslip. Images were collected on a Leica SP8 WLL Confocal Microscope using a 20× 0.75 NA glycerol objective and 400 Hz scan rate at 512×256 pixels and 12-bit resolution. Images were collected at 0.85x zoom (Kr mutant experiment) or 1x zoom (uniform Bcd experiment). A z stack that spanned from the top of the embryo to the embryo’s midsagittal plane was imaged with a 1.04 μm z-step size. For the Kr mutant experiment, mutant and control embryos were imaged on the same microscope slide and distinguished by the presence of Kr mRNA, as the Kr¹ mutant is RNA-null. Gal4 staining was excited with a 488 nm laser and detected at [495, 516] nM (Photon Counting), and Kr staining was excited with a 457 nm laser and detected at [559, 573] nM (Photon Counting). Both channels were collected with a line accumulation of 3. For the uniform Bcd experiment, mutant and control embryos were mounted and imaged on different microscope slides. Gal4 staining was excited with a 650 nM laser (measured as 48 μW through the 10× objective for both mutant and control) and detected at [672, 694] nM (Photon Counting) with a line accumulation of 1. In both experiments, Dapi was excited with a 405 nm laser.

Chromatin immunoprecipitation

Hydroxyurea treatment

Embryos were collected from RpA70-EGFP parents on standard apple juice-agar plates after a 2.5-h laying period and then dechorionated 1–2 min in 50% bleach. Dechorionated embryos were transferred to a homemade plastic basket with a nylon mesh bottom, similar to reported previously.⁷⁹ Embryos were then permeabilized in 1:10 Citrasolv (CitraSolv LLC, Danbury CT, USA) in water in a glass container with constant gentle shaking. Embryos were washed extensively with 1xPBS and then transferred to either a 10 mL hydroxyurea (HU) treatment solution (1xPBS, 100 μM Rhodamine B, 100 mM HU), or a 10 mL control solution (1xPBS, 100 μM Rhodamine B). Rhodamine B was included in these incubation solutions to mark embryos that had been successfully permeabilized. Embryos were incubated in the HU treatment or control solution for 40 min to maximize the number of interphase NC12 and NC13 embryos following fixation. In the case of the HU-treatment, the 40 min-duration also maximizes the number of DNA replication-inhibited NC12/NC13 embryos. The basket was removed from the incubation solution and blotted on a paper towel before the nylon mesh was removed. The embryos were immediately transferred to a glass scintillation vial filled with a formaldehyde fixation solution (2 mL 1xPBS/0.5% Triton X-100, 6 mL heptane, 180 μL 20% formaldehyde) by dunking the nylon mesh carrying the embryos in the vial, with the transfer taking ~30 s. The embryo-free mesh was then removed from the vial and a standard embryo fixation protocol for ChIP-seq performed.

Embryo collection

For the collection of DNA replication-inhibited and control-treated embryos, following treatment and fixation embryos were sorted for ChIP-seq on 1%PBS-agarose plates in 1xPBS/0.5% Triton X-100. Sorting was performed using a Leica M165 FC Fluorescent Stereo Microscope to illuminate embryos with 488 nm light. NC12 and NC13 embryos were identified based on nuclear density indicated by Rpa70-GFP nuclear signal. During the collection of HU-treated embryos, embryos with most likely inhibited DNA replication were identified based on the distribution of Rpa70-GFP signal within nuclei. Embryos displaying punctate Rpa70-GFP within nuclei were preferentially selected, given the prior observation that replication-inhibition alters the intra-nucleus distribution of Rpa70-GFP.¹⁹ 100 NC12/NC13 embryos were collected for each ChIP-seq replicate and stored at −80°C in 1xPBS/0.5% Triton X-100.

For the collection of bcd osk tsl and control w embryos, no treatment was performed. Embryos were collected after a 4-h laying period, dechorionated, and fixed for ChIP-seq according to a standard protocol.¹⁹ Fixed embryos were hand-sorted for NC14-stage embryos on 1%PBS-agarose plates in 1xPBS/0.5% Triton X-100. Sorting was performed under white light illumination by a Leica M165 FC Fluorescent Stereo Microscope. 50 NC14 embryos were selected for each ChIP-seq replicate based on morphological indicators including signs of cellularization and the clearing of lipid vesicles from the embryo’s periphery. Collected embryos were stored at −80°C in 1xPBS/0.5% Triton X-100.

Chromatin immunoprecipitation

Chromatin immunoprecipitations were performed as described previously,¹⁹ with the following alterations. For each experiment, the embryos used for immunoprecipitation (IP) were 100 hand-sorted NC12/NC13 HU-treated Rpa70-GFP embryos and 100 hand-sorted NC12/NC13 control-treated Rpa70-GFP embryos, or 50 hand-sorted NC14 bcd osk tsl embryos and 50 hand-sorted NC14 $w$ embryos. Each genotype was sonicated 4 × 15 s at 20% Output and full duty cycle (Branson Sonifier 450) and then evenly split up into samples for the immunoprecipitations. ChIP was performed using a rabbit α-Bcd antibody (gift from J. Zeitlinger), a mouse α-Rpb1 CTD (4H8) monoclonal antibody (Cell Signaling Technology #2629), and a rabbit α-c-myc polyclonal antibody (Sigma-Aldrich #C3956) on HU-treated and control-treated embryos, with 25 embryo’s-worth of DNA in each IP. ChIP was performed using a mouse α-RNA Polymerase II CTD repeat YSPTSPS antibody (abcam #ab5408) and a rabbit α-c-myc polyclonal antibody (Sigma-Aldrich #C3956) on bcd osk tsl and $w$ embryos, with 25 embryo’s-worth of DNA in each IP. Following incubation of the samples with blocked Protein G Dynabeads (Thermo Fisher), samples were washed twice in 10 mM Tris pH7.5 5 mM MgCl₂ and then tagmented by exposure to 1 μL Nextera Tn5 Transposase (Illumina) while shaking at 1,000 rpm for 40 min at 37°C. All washes were performed after tagmentation as described in Blythe and Wieschaus, 2015.¹⁹ Following crosslink reversal and DNA cleanup, libraries were amplified as described in Blythe and Wieschaus, 2016.²⁰

QUANTIFICATION AND STATISTICAL ANALYSIS

Modeling

Model structures, simulation method, and parameter sweeps are described in detail in Data S1: modeling supplement.

Enhancer sequence analysis

Predicting binding sites

PWM scores were calculated for each bp in the hbP2 sequence using the R Biostrings package (https://github.com/Bioconductor/Biostrings) and normalized by their maximum possible score. The Bcd and Zld PWMs were obtained from the UMass Chan Medical School Database of Drosophila TF binding specificities.

Predicting nucleosome positioning

The positions of likely nucleosomes were determined for the hbP2 sequence using NuPoP software⁴¹ and identified at nucleotide positions 2–148 and 157–303 within a genomic sequence containing hbP2 flanked by 30 bp on its 5′ and 3′ ends. Similar positions were obtained by varying the flanking length of DNA for the sequence given to the NuPoP algorithm.

Live-imaging analysis: MS2

All analyses of live-imaging data were performed in MATLAB (https://www.mathworks.com/products/matlab.html, R2022a). Live-imaging data was quantified during NC13, defined as the time period between the start of anaphase 12 and the start of anaphase 13.

Nuclei segmentation

Nuclei were segmented based on the His2Av-RFP channel. On a per-frame basis, the His2Av-RFP stack was filtered with a 3-D Guassian smoothing kernel with a standard deviation of 2 pixels. The smoothed stack was then binarized using Otsu’s method. Following binarization, morphological opening was performed using a spherical structuring element with a 4-pixel radius. Connected components with a radius of fewer than 5 pixels were then removed from the binary stack. Following an extended minima transformation, the stack was segmented with the watershed algorithm.

MS2 foci detection

To identify transcriptional foci, the MCP-GFP channel was processed on a per-frame basis. GFP intensity was first contrast-enhanced by employing top-hat and bottom-hat filtration (using a spherical structuring element with a 3-pixel radius). Difference of Gaussian (DoG) filtration was then performed on the contrast-enhanced stack, with 3-D Guassian smoothing kernels of 1-pixel and 5-pixel standard deviations. Pixels outside of the nuclear mask or with values below the estimated noise floor were then eliminated before the stack was binarized. Foci were identified from the binarized stack and eliminated if they had a radius of greater than 100 pixels or if they occupied fewer than 3 z-planes. Following segmentation of the DoG image with the binary mask, foci were eliminated if their total intensity was less than 100 AU.

AP position mapping

The AP coordinates for each pixel in the imaging field were determined relative to the whole embryo as follows. Immediately following the completion of an imaging session (mid NC14), an “overview image” was generated by taking a His2Av-RFP tile-scan image of the entire embryo, capturing both the embryo surface containing the live-imaged ROI, as well as the mid-saggital plane, which shows the maximal extent of the AP and DV axes. The overview image was subsequently max-projected and an embryo mask was generated by thresholding the projected image. During image processing, the AP axis was automatically defined by finding the extrema of the embryo mask of the overview image representing the long axis of the embryo. The position of the live-imaged ROI relative to the overview image was then determined by the MATLAB normxcorr2 function, which calculates the 2D correlation coefficient between the projected overview image and the final live-imaged frame. From this information, each pixel in the live imaging dataset is assigned an AP position relative to the coordinate system defined by the AP axis and the position of the live-imaging ROI as determined by cross-correlation analysis. Imaged nuclei are thereafter assigned AP positions based on the computed AP coordinates of the pixels corresponding to nuclear mask centroids.

Nuclei tracking

Nuclei were tracked frame-to-frame based on the segmentation of the His2Av-RFP channel. Centroids were calculated for all segmented nuclei in a frame and nuclei tracked by finding their centroid’s nearest neighbor in the following frame. If nuclei were lost track of during the time of active transcription, they were excluded from subsequent analyses.

Calculation of transcriptional features

The MS2 transcriptional dynamics of each nucleus over NC13 was calculated by segmenting the MCP-GFP channel with the 3-D MS2 foci mask and summing the fluorescence within each spot volume. To account for differences in nuclear cycle length between replicates, frames were sequentially divided into 100 bins—approximately the number of frames in the shortest NC13 of the MS2 imaging datasets—and the mean MS2 fluorescence for each nucleus calculated within each bin. Following binning, the nuclear fluorescence tracks were filtered to remove spots detected for durations less than the time required for transcription of the $y$ intron of the reporter containing the MS2 stem loops (~1.6 min given the prior measurement of a ~2.5 kb/min Pol II elongation rate in NC13⁸⁰). Per-nucleus transcriptional onset times were calculated by finding the earliest bin in which each spot appeared following mitosis and converting the bin numbers into minutes given a 10-s frame time. Relative Pol II loading rates (AU/min) were calculated for each nucleus as the difference in a rolling average of spot fluorescence between the onset time of the spot and the time 1 min after transcriptional onset. The fraction of active nuclei was calculated within 1% AP bins by dividing the number of nuclei that initiated transcription in a 1% AP span by the total number of nuclei measured within the bin.

Live-imaging analysis: EGFP-bcd

Nuclei were segmented and tracked over NC13 with the methods described in the MS2 analysis section. The GFP-Bcd channel was segmented using the nuclear mask, and the mean GFP fluorescence calculated within each nucleus over NC13.

Fixed imaging analysis: HCR

For the Kr mutant experiment, hbP2-MS2 Kr¹/hbP2-MS2 Kr¹ and control embryos were distinguished based on the Kr channel, as the Kr¹ mutant is RNA-null. For visualization, images were max-projected and the maximum pixel intensities set to allow for comparison of the posterior boundary of reporter expression between genotypes. As hbP2-MS2 Kr¹/SM6 (control) embryos were heterozygous for the reporter, the maximum intensity threshold of the 488 nm channel was lowered farther than that of the mutant embryos homozygous for hbP2-MS2. For the uniform Bcd experiment, images were max-projected and the maximum intensity threshold of the 488 nm channel set to be equivalent between control (hbP2-MS2/+) and mutant (hbP2-MS2/+ in uniform Bcd) embryos. For quantification of HCR images, the 488 nm channel was segmented using a 3D embryo mask created from a smoothed and thresholded version of the DAPI channel. Embryo masks were 15-μm thick shells that captured the nuclei at the outer surface of each embryo, and were used to remove any confounding background signal from the centers of the embryos. AP intensity profiles were generated from the masked 488 nm channel by averaging all pixels ≤15 μm dorsal and ≤15 μm ventral to the AP axis at each AP position.

ChIP-seq data analysis

Unique dual-indexed ChIP-seq libraries were subjected to paired-end 150 bp sequencing on an Illumina Novaseq X sequencer (Admera Health). Barcode-split fastq files were first trimmed of adapter sequences using TrimGalore (https://github.com/FelixKrueger/TrimGalore.git). Trimmed reads were mapped to the Drosophila melanogaster dm6 reference genome using Bowtie2 (v2.4.1)⁸¹ with option -X 2000. Sorted mapped reads were passed through Picard MarkDuplicates (v2.21.4, http://broadinstitute.github.io/picard/) to mark suspected duplicate reads. Non-duplicated mapped reads with map quality scores ≥10 were merged by replicate following import to R (v4.2.2) as a Genomic Ranges object using the GenomicAlignments package.⁸² To visualize coverage of ChIP-seq experiments, the average coverage of reads per 10 bp window was calculated and normalized to the total number of reads per merged sample/one million to yield counts-per-million reads per 10 bp window. Such coverage data were plotted using the GViz (v1.42.0).⁸³

Supplementary Material

SUPPLEMENTAL INFORMATION

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.116121.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse anti Rpb1 CTD (4H8) monoclonal antibody	Cell Signaling Technology	Cat# 2629; RRID:AB_2167468
Mouse anti RNA Polymerase II CTD repeat YSPTSPS (phospho S5) monoclonal antibody– ChIP Grade	Abcam	Cat# ab5408; RRID:AB_304868
Rabbit anti Bicoid antibody	Julia Zeitlinger, Stowers Institute
Rabbit anti c-myc polyclonal antibody	Sigma-Aldrich	Cat# C3956; RRID:AB_439680
Chemicals, peptides, and recombinant proteins
Hydroxyurea	Sigma-Aldrich	H8627
Critical commercial assays
Hybridization Chain Reaction Probe Set against Gal4 for use with B1 amplifier	Molecular Instruments	PRJ578
Hybridization Chain Reaction Probe Set against Kr for use with B3 amplifier	Molecular Instruments	PRO170
Hybridization Chain Reaction B1 Amplifier Hairpin Set coupled to 488nm fluor	Molecular Instruments
Hybridization Chain Reaction B3 Amplifier Hairpin Set coupled to 546nm fluor	Molecular Instruments
Deposited data
ChIP-seq: Bcd and RNA Pol2, HU-treated embryos and bcd osk tsl mutants	This work	GEO: GSE283996
Uniform Bcd ChIP-seq	Hannon et al.15	GEO: GSE86966
Wild-type, bcdE1, and zld294 ATAC-seq	Hannon et al.15	GEO: GSE86966
Bcd ChIP-nexus	Brennan et al.42	https://zenodo.org/record/8075860
Wild-type ATAC seq, NC13 mitosis and i nterphase	Blythe and Wieschaus20	GEO: GSE83851
Image analysis and modeling code	This work	https://doi.org/10.5281/zenodo.15832329
Experimental models: Organisms/strains
w P{w+ = His2Av-RFP}1; bcdEGFP	Steve Small & Pinar Onal, New York University
y w; P{w+ = his2AV-RFP}2; P{w+ = nos>MCP-GFP}4F	Hernan Garcia, University of California, Berkeley
y w; {w+ = nos>MCP-mCherry}VK22	Hernan Garcia, University of California, Berkeley
y {w+= His2Av-miRFP670-T2A-HO1}ZH2A w	This work
y w; {3xP3>RFP = hbP2>y-MS224-Gal4}attP40	This work
y w; {w2 = alphaTub64c>EGFP-Bcd-sqh 3′UTR}VK33 bcdE1 /TM3, Sb	This work
y w; {3xP3>RFP = hbP2(+1 zld)>y-MS224-Gal4}attP40	This work
w; UASp>zldRNAi (III)	Christine Rushlow, New York University
w; bcdE2 osk166 e tsl4/TM6c, Sb	This work (newly recombined from bcdE2 osk166 and tsl.4)
w; {w+ = RpA-70-EGFP}attP2	Blythe & Wieschaus, 2015
cn bw Kr1/SM6 bw	Bloomington Drosophila Stock Center	BDSC #3494
w; P{w+ = alpha-tubulin67c>GAL4 VP16}67	Eric Wieschaus, Princeton University
Software and algorithms
MATLAB R2022A	The Mathworks	R2022A
R 4.2.2	R Foundation for Statistical	R4.2.2
	Computing, Vienna Austria
NuPoP	CRAN	v2.5.5

Highlights.

• Nucleosomes compete with Bicoid to shape morphogen gradient output

• Concentration-dependent regulation is assayed by MS2/MCP reporters

• A stochastic model of transcription incorporates nucleosomes and DNA replication

• Bicoid and DNA replication control transcription at the level of Pol2 pause release