We sought to determine whether eve enhancers regulate transcription by modulating the same set of bursting parameters in endogenous and ectopic eve expression regions. Specifically, we aimed to compare how eve enhancers drive transcriptional bursting in and out of the well-known seven endogenous eve stripes (Frasch and Levine, 1987; Hare et al., 2008).

As our starting point, we took a previously established BAC-based eve-MS2 reporter system (Berrocal et al., 2020) that carries an ~20 kb DNA fragment around the D. melanogaster eve coding region containing the five eve enhancers responsible for regulating the expression of the seven eve stripes, other cis-regulatory elements such as neuronal and muscular regulatory elements (Fujioka et al., 1999; Fujioka et al., 2013) that might influence eve stripe expression in early development (Fujioka et al., 1999; Fujioka et al., 2013), and the late element (LE) that upregulates eve expression in all stripes in response to the EVE protein (Fujioka et al., 1996; Jiang et al., 1991; Figure 2A). We will refer to this construct as eveMS2-BAC (see SI section: DNA constructs and fly lines in Materials and methods). The MS2 reporter system fluorescently labels nascent mRNA molecules resulting in sites of nascent transcription appearing as puncta whose fluorescence is proportional to the number of active RNA Pol II molecules. As a result, the system allows for the visualization of transcriptional bursting at single locus resolution, in real-time, in living embryos (Chubb et al., 2006; Ferguson and Larson, 2013; Garcia et al., 2013; Golding et al., 2005; Golding and Cox, 2004). When inserted into the D. melanogaster genome, eveMS2-BAC expresses in seven stripes that recapitulate the wild-type expression of eve (Figure 2B; Berrocal et al., 2020) as observed by FISH and live-imaging experiments (Lammers et al., 2020; Lim et al., 2018; Luengo Hendriks et al., 2006).

To establish an ectopic eve expression pattern, we modified the eve reporter locus (Figure 2A; Berrocal et al., 2020). Specifically, we aimed to create an anterior expansion of eve stripe 2 beyond its endogenous expression domain and into ectopic regions where we could study transcriptional bursting under inputs foreign to an eve enhancer, for example higher levels of the activator Bicoid and the repressor Giant (Gt) (Figure 1D). To make this possible, we leveraged the fact that the anterior boundary of eve stripe 2 is established through repression by Giant (Small et al., 1992). Classic work by Small et al. identified a minimal regulatory element of the eve stripe 2 enhancer (eveS2-MRE; Figure 2A) and found that deleting three Giant binding sites within this minimal enhancer produced a strong anterior expansion of eve stripe 2 in the context of a reporter driven by eveS2-MRE (Small et al., 1992).

We generated an eveMS2-BAC variant, where the three binding sites for Giant identified in the eveS2-MRE were disrupted on the complete eve stripe 2 enhancer (eveS1wt-eveS2Gt^-; Figure 2A and C). Live imaging experiments on eveS1wt-eveS2Gt^- embryos showed only transient ectopic expression at the inter-stripe region between eve stripes 1 and 2. This transient inter-stripe expression lasts until 30–35 min into nc14; while inter-stripe expression between eve stripe 1 and eve stripe 2 disappears after ~20 min in wild-type embryos (compare Figure 2B and C; compare Figure 2—figure supplement 1A and B). These eveS1wt-eveS2Gt^- embryos did not produce the robust anterior expansion of eve stripe 2 described for the eveS2-MRE alone (Small et al., 1992). We attribute this muted anterior expansion in eveS1wt-eveS2Gt^- embryos (Figure 2C) to the regulatory sequences not present in the original minimal eve stripe 2 reporter construct which might provide a buffering effect to the disruption of the three Giant binding sites (López-Rivera et al., 2020).

In an attempt to expand the anterior ectopic domain of eveS1wt-eveS2Gt^-, we sought to free its expression domain from any potential interference from eve stripe 1 expression. To make this possible, we deleted endogenous expression corresponding to the eve stripe 1 enhancer. Specifically, we generated a mutant version of eveMS2-BAC with the eve stripe 1 enhancer deleted (eveS1∆-eveS2wt; Figure 2A and D; Figure 2—figure supplement 1C). Unexpectedly, these embryos still exhibited a dim eve stripe 1 (~30% of embryo length) after ~30 min into nc14, perhaps due to the activity of the eve late element; and a dim additional anterior stripe that we refer to as eve stripe 0 (~20% embryo length) after ~25 min into nc14. In a previous study, Small et al., 1992 identified a ‘head patch’ of gene expression when assaying the regulation of the minimal regulatory element of the eve stripe 2 enhancer. It is tempting to identify our eve stripe 0 with their observed head patch. (Small et al., 1992) speculated that this head patch was the result of sequences in the P-transposon system used for their genomic insertions, which are not present in our experimental design. Thus, the appearance of eve stripe 0 indicates a repressive role of eve stripe 1 enhancer beyond the anterior boundary of eve stripe 1 (Figure 2D), and it may imply that the minimal regulatory element of the eve stripe 2 enhancer can indeed drive expression in this head patch when eve stripe 1 enhancer is not present.

Finally, we coupled the three deletions of Gt-binding sites in the eve stripe 2 enhancer from eveS1wt-eveS2Gt^- with the complete deletion of the eve stripe 1 enhancer in eveS1∆-eveS2wt to create eveS1∆-eveS2Gt^- (Figure 2A and E; Figure 2—figure supplement 1D). Surprisingly, eveS1∆-eveS2Gt^- embryos revealed large ectopic regions of eve expression more complex than the sum of patterns displayed by the independent mutants described above. Beyond a stronger and longer-lasting inter-stripe expression between eve stripe 1 and eve stripe 2 than observed in eveS1wt-eveS2Gt^-, eveS1∆-eveS2Gt^- embryos exhibited the following notable features: a stronger-than-wild-type eve stripe 2 (located at ~40% of embryo length); the presence of eve stripe 1 (~30% of embryo length) and eve stripe 0 (~20% embryo length); and many eve-active nuclei in normally silent inter-stripe regions between eve stripe 2 and eve stripe 0 (Figure 2E). The fact that the knock-out of eve stripe 1 enhancer coupled with the disruption of Gt-binding sites in eve stripe 2 enhancer renders more ectopic expression on the anterior half of fruit fly embryos than the independent disruptions in eveS1∆-eveS2wt and eveS1wt-eveS2Gt^- implies that the repressive activity of the eve stripe 1 enhancer synergizes with the repression exerted by Giant—and potentially with other unidentified transcription factors that bind in the vicinity of Gt-binding sites—on the eve stripe 2 enhancer. The hypothesis that Gt binding sites in eve stripe 2 enhancer may recognize other transcription factors was proposed by Small et al., 1992, who observed that the anterior expansion of eve stripe 2 that results from disrupting Gt-binding sites in eve stripe 2 enhancer is ‘somewhat more severe’ than the expansion observed in Gt^- embryos.

Taken together, our results suggest that the eve stripe 1 enhancer plays a repressing role in the anterior half of fruit fly embryos which synergizes with the Giant repressor and likely with other transcriptional regulators bound to Gt binding sites or their vicinity in the eve stripe 2 enhancer. This argues in favor of cross-activity between the eve stripe 1 and 2 enhancers that impacts eve expression in the anterior half of the embryo. eve stripe 1 enhancer might be also playing a role in the regulation of eve stripe 2, as Giant-binding site deletions in the eve stripe 2 enhancer alone do not result in the stronger-than-wild-type eve stripe 2 observed in eveS1∆-eveS2Gt^- embryos. In summary, coupling the disruption of Giant-binding sites in the eve stripe 2 enhancer with the deletion of the eve stripe 1 enhancer produces different mutant patterns than the sum of the individual mutants. Finally, regardless of the complex regulatory interactions uncovered by our enhancer mutants, our results indicate that the ectopic gene expression patterns driven by our eveS1∆-eveS2Gt^- reporter provide an ideal scaffold for our investigations of the regulation of transcriptional bursting outside of endogenous embryo regions.

We determined the position of nuclei displaying active eve transcription and labeled them as endogenous if they were positioned within the boundaries of wild-type eve stripes (eve stripe 1, eve stripe 2, eve stripe 3, eve stripe 4); or as ectopic if they were located in the inter-stripe region between eve stripe 1 and eve stripe 2 (eve stripe 1–2) or in eve stripe 0 (in the far anterior; Figure 3A) as described in Materials and methods. eve stripe 1 expression in embryos with disrupted eve stripe 1 enhancer was considered endogenous, as we believe that this expression results from activity of the late element. All active nuclei in wild-type embryos were labeled as endogenous. Overall, ectopic regions show lower levels of mean MS2 fluorescence than endogenous regions, as is evident by comparing eve the interstripe 1–2 and eve stripe 0 against eve stripe 1, eve stripe 2, and eve stripe 3 in eveS1∆-eveS2Gt^- embryos (Figure 2E, Right). This is perhaps due to the unavailability of optimal concentrations of transcription factors; for example a lack of activators or an excess of repressors with respect to the concentrations found in endogenous regions (Figure 1C and D).

To uncover which bursting parameters are modulated to create each endogenous and ectopic stripes and interstripe regions, we need to extract the bursting parameters in each region. We computed bursting parameters for nuclei grouped by stripe and binned by transcriptional output (Figure 3—figure supplement 1) in our four fly lines, with the following rationale. In the bursting model, the average rate of transcription initiation is described by the formula rkonkon+koff, where konkon+koff indicates the fraction of time the promoter spends in the ON state (Lammers et al., 2020). As enhancers and their inputs (e.g. transcription factors, chromatin state) define bursting parameters (k_on, k_off, r), nuclei of similar average transcriptional output within the same stripe should be driven by similar inputs acting over the same enhancer. Thus, these nuclei should show similar values of the bursting parameters k_on, k_off and r that satisfy the equation above. On the other hand, our model predicts that nuclei with different fluorescence must differ in at least one of their bursting parameter values (k_on, k_off and/or r).

The average MS2 fluorescence is a direct reporter of the average rate of transcriptional initiation. Single-cell MS2 fluorescence measurements reflect the transcriptional dynamics of individual promoters as they undergo transcriptional bursting (Figure 3B). However, the actual promoter states, or bursting parameters, underlying the transcriptional bursting remain ‘hidden’, as RNA Pol II molecules engage in elongation for several minutes (~140 s for the MS2::yellow transcriptional unit in our reporter system) (Berrocal et al., 2020). As a result, MS2 fluorescence is observable even after the promoter switches to the OFF state, convolving the promoter switching dynamics with those of transcriptional elongation. Thus, we can only compute promoter states by inferring them from MS2 fluorescence over time. To infer hidden promoter states, we used a compound-state Hidden Markov Model (cpHMM) developed by Lammers et al., 2020. By inferring the succession of activity states, cpHMM estimates rates of transitioning between the OFF and ON states (k_on and k_off) and the rate at which ON promoters load active RNA Pol II molecules (r).

Consistent with our previous work (Berrocal et al., 2020), we find that endogenous stripes in eveMS2-BAC wild-type embryos modulate their transcriptional output (mean MS2 fluorescence in wild-type embryos ranges from 2 to 15 AU) by tuning the average k_on (from 0.5 to 1.5 OFF to ON promoter transitions per minute) and r (from an average fluorescence increase at a rate of 5 AU per minute to 10 AU per minute). The average k_off value remains largely constant (0.5 ON to OFF promoter transitions per minute), with only a minor downregulation at high transcription outputs (Figure 3C). Thus, we confirm that eve-active nuclei in all wild-type stripes achieve higher levels of transcription by upregulating average bursting frequency (k_on) and amplitude (r), while average burst duration (k_off^–1) remains largely the same.

eveS1wt-eveS2Gt^- (Figure 3D) and eveS1∆-eveS2wt (Figure 3E) embryos did not yield enough ectopic nuclei for cpHMM inference. However, their endogenous stripes followed the same bursting strategy observed in wild-type embryos, regardless of whether stripes were activated by wild-type or mutant enhancers (see SI Section: Complementary analysis of bursting parameters in Materials and methods). We inferred bursting parameters across regions of endogenous and ectopic nuclei in eveS1∆-eveS2Gt^- embryos (eve stripe 1–2 and eve stripe 0), as they yielded sufficient ectopic eve-active nuclei to support cpHMM inference. As noted previously, these embryos feature an eve stripe 2 with nuclei of higher transcriptional output than wild-type embryos (compare Figure 2B and E), and a large region of ectopic expression towards the embryo anterior. Despite these differences in transcriptional output, bursting parameters in endogenous and ectopic eve-active nuclei in eveS1∆-eveS2Gt^- embryos follow the same trends as wild-type embryos (Figure 3—figure supplement 2). In all regions–both endogenous and ectopic–enhancers increase transcription by increasing in k_on and r, while k_off remains largely constant (Figure 3F).

We performed an orthogonal cpHMM inference of bursting parameters by grouping nuclei in only two categories (endogenous and ectopic) (Figure 3—figure supplement 3), instead of grouping them according to their stripe, and we observed that this approach renders the same results (see SI Section: Complementary analysis of bursting parameters in Materials and Methods).

Taken together, our results show that all eve enhancers modulate their transcriptional output by increasing burst frequency (k_on) and amplitude (r). k_off, which shapes burst duration, remains largely constant, and shows a subtle drop as the mean MS2 fluorescence of nuclei increases. A wide range of transcriptional outputs result from these parameters. eve strategies of bursting control are robust to mutations on eve enhancers, and remain consistent in the presence of a myriad of inputs, including ectopic inputs different from those that shape the transcription of the seven canonical eve stripes.

We generated four reporter constructs based on a previously established Bacterial Artificial Chromosome (BAC) carrying the ~20 Kb DNA sequence around eve (Venken et al., 2006; Venken et al., 2009), and whose eve coding sequence has been replaced by an MS2::yellow transcriptional unit (Berrocal et al., 2020). We used wild-type eveMS2-BAC from Berrocal et al., 2020. The other three BAC constructs were derived from wild-type eveMS2-BAC. These constructs carried mutant versions of eve stripe 1 and eve stripe 2 enhancers. Vector Builder (https://en.vectorbuilder.com/) generated the mutant versions through ccdB-amp cassette mediated recombineering. These mutant BACs are available on Vector Builder’s website. SnapGene (.dna) files with eveMS2 BAC sequences are in the repository https://github.com/aberrocal/BurstingStrategies-eve, folder BurstingStrategies-eve/_DataSubmission/BACSequences/.

We crossed male flies from lines carrying eveMS2-BAC constructs (w-; +; MS2::yellow) and female flies carrying His::RFP and MCP::GFP fusion proteins (yw; His::RFP; MCP::GFP; Garcia et al., 2013). His::RFP allows for visualization of nuclei, MCP::GFP binds MS2 nascent transcripts to form fluorescent puncta at sites of nascent MS2 transcription. We set embryo-collection cages with ~30 male and ~100 female fruit flies, and collected offspring embryos after 1 hr 30 min. All movies in the same dataset were recorded within ~1 week. We mounted embryos on a slide for confocal imaging, as described in Berrocal et al., 2020; Bothma et al., 2014. Aging embryos for 1 hr 30 min allows us to capture the entire interval between the 14th synchronous cell cleavage and the beginning of gastrulation. We recorded a total of 22 live embryos as shown in Table 1. All imaging was done in a Zeiss-800 scanning-laser confocal microscope. Movies of embryonic development were captured under a 63 x oil objective, in windows of 202.8 µm x 50.7 µm, at pixel size of 0.2 µm, zoom 0.5 x. Movies were recorded in two channels, EGFP for MS2 signal, and TagRFP for His::RFP signal. Imaging parameters were 16 bits per pixel, scan mode frame, bidirectional scanning, scan speed 7, pixel dwelling 1.03 µs, laser scanning averaging 2, averaging method mean, averaging mode line, laser power EGFP 30 µW and TagRFP 7.5 µW, master gain in EGFP channel 550 V and in TagRFP channel 650 V, digital offset in both channels 0, digital gain in both channels 1, pinhole size 44 µm (1 Airy unit - 0.7 µm/section) at 63 x objective, laser filters EGFP:SP545 and TagRFP:LBF640. Data points consist of Z-stacks of 21 slices separated by intervals of 0.5 µm, to span a range of 10 µm across the Z axis. Z-stack mode full stack. Whole Z-stacks were recorded every 16.8 s (wild-type, eveS1wt-eveS2Gt^-, eveS1∆-eveS2Gt^-) and 19.5 s (eveS1∆-eveS2wt). The difference in time resolution between datasets does not impact our analysis, as the cpHMM analyzes interpolated data points at 20 s intervals. These parameters are based on the imaging protocol and settings in Berrocal et al., 2020. We stopped live imaging of individual embryos after 50 min into nuclear cycle 14, before the cell rearrangements of gastrulation, and took mid-sagittal and surface images of the whole embryo to localize our 202.8 µm x 50.7 µm window along the embryonic anterior-posterior axis. Raw data from confocal microscope imaging is publicly available in Zenodo (https://zenodo.org/, https://doi.org/10.5281/zenodo.7204096; see SI section: Data and Code) (Table 1).

We tracked MS2 foci from movies and segmented them using the MATLAB based analysis pipeline developed by Berrocal et al., 2020; Garcia et al., 2013; Lammers et al., 2020. Specifically, for segmentation of MS2/MCP::GFP foci across stacks on the Z-axis, we combined the MATLAB pipeline mentioned above with Fiji-Weka Segmentation 3D software, as described in Berrocal et al., 2020. The MATLAB/Fiji-Weka pipeline extracts the position of nuclei and the fluorescence intensity and position of individual MS2 foci over time. The final result of the MATLAB based analysis pipeline are CompiledParticles.mat files that contain the position of nuclei, as well as their MS2 fluorescence intensity over time (see Data and Code).

We manually segmented nuclei from eveS1∆-eveS2Gt^- and eveS1wt-eveS2Gt^- fly lines, as their stripes were not always clearly discernible. For these embryos, we assigned nuclei to individual stripes based on the position of stripes at 45 min into nc14, when they became separated from the background. The boundary between eve stripe 1–2 and eve stripe 2 in eveS1∆-eveS2Gt^- embryos was set at 36% of embryo length, according to the kymograph of MS2 fluorescence over time. On the other hand, eveS1∆-eveS2wt and wild-type embryos showed defined stripes after 25 min into nc14. Thus, we used a MATLAB k-means clustering algorithm to dynamically assign eve-active nuclei to individual stripes, tracking nuclei by the accumulation of MS2 fluorescent output in windows of five-minutes. Nuclei active between 0 and 25 min into nc14 were assigned to stripes based on their position at 25 min into nc14. We generated movies of segmented MS2 spots assigned to individual stripes in windows of ~5 minutes. MATLAB scripts for manual and k-means-automated segmentation of stripes, as well as scripts to generate movies of segmented stripes are available in github (see Data and Code).

We used traces of MS2 fluorescence intensity over time, which reflect transcriptional activity, to generate heatmap/kymographs of MS2 transcription datasets. We generated heatmaps (Figure 2, Figure 2—figure supplement 1) by collapsing data points from all embryos of the same genotype into a single kymograph plot. We started by adjusting the position of nuclei in each embryo relative to nuclei in other embryos of the same genotype. As we had assigned MS2 active nuclei to individual stripes, we measured the distance along the anterior-posterior axis from each MS2 focus to the center of its corresponding stripe. We inferred the position of pseudo-stripes formed by the combined data from all embryos of the same genotype. We calculated the position of pseudo-stripes along the anterior-posterior embryo axis by averaging the position of the center of stripes along the anterior-posterior axis in individual embryos of the same genotype. Finally, we assigned a position to all nuclei of the same genotype relative to pseudo-stripes by positioning them at the same distance from the center of pseudo-stripes as they were from the center of the stripe where they originated. We followed the same procedure to locate the position of inactive nuclei.

To compare the bursting parameters between endogenous and ectopic regions of eve activity, we segmented MS2-active nuclei and assigned them to individual regions that were deemed to be either endogenous or ectopic. We labeled regions as endogenous if their position overlapped within the boundaries of wild-type eve stripes (eve stripe 1, eve stripe 2, eve stripe 3, eve stripe 4); or as ectopic if their position overlapped with the inter-stripe region between eve stripe 1 and eve stripe 2 (eve stripe 1–2) or with the novel eve stripe 0 (~20% embryo length). All stripes in wild-type embryos were labeled as endogenous.

We selected a three-state model of promoter activity (OFF, ON₁, ON₂) based on the following argument. Transcription in pre-gastrulating Drosophila embryos occurs after DNA replication, and sister chromatids remain paired. However, most of the time, paired MS2-tagged sister loci cannot be resolved independently using diffraction-limited microscopy (Lammers et al., 2020). Therefore, each fluorescent spot in our data results from the combined activity of two promoters, each of which, in the simplest possible model of transcriptional bursting, may be ON or OFF (Lammers et al., 2020). To account for this, the cpHMM infers three states from the observed MS2 data: OFF (both sister promoters inactive), ON₁ (one sister promoter active), and ON₂ (two sister promoters active). For ease of presentation, we aggregated ON₁ and ON₂ states into a single effective ON state, as we did in our previous work (Berrocal et al., 2020). This leads to an effective two-state model with one OFF and one ON state and three burst parameters: k_off^–1 (the burst duration), k_on (the burst frequency), and r (the burst amplitude). k_on is defined as the sum of the transition rates from OFF to any of the two active states described above: OFF → ON₁ and OFF → ON₂. k_off is defined as the rate at which the system returns to the OFF state upon leaving it, which is described by the formula k_off^–1 = (1poff - 1) k_on^–1, where poff is the fraction of time the system spends in the OFF state. k_off is the inverse of mean burst duration. r is defined by the average of the rates of transcription initiation in the two ON states (r₁ and r₂) weighted by the fraction of the time that the system spends on each state (p₁ and p₂) as described by the formula r = p1r1+p2r2p1+p2 (Lammers et al., 2020). The outputs of the three state model of promoter activity (k_on, k_off, and r) were used for downstream analyses.

The three-state model of promoter activity is the simplest model compatible with our current understanding of transcription at the eve locus in early fruit fly embryos. However, we do not dismiss the possibility that more complex processes, not captured by our model, define eve transcription. Promoters, for instance, may exhibit more than two states of activity, beyond a simple ON and OFF mechanism. Nevertheless, as pointed out by Lammers et al., 2020 - SI Section: G. the cross-validation of cpHMM inference sensitivities between different model schemes (two, three, or multiple state Hidden Markov Models) do not yield consistent results regarding on which one is more accurate; and for the time being, there is no alternative to a HMM for inference of promoter states from MS2/PP7 fluorescence signals obtained using laser-scanning confocal microscopy (Lammers et al., 2020; Syed et al., 2023; although other approaches exist using state-of-the-art microscopy and deconvolution algorithms to improve signal-to-noise ratio). Furthermore, orthogonal approaches to quantify transcription that rely on static methods, such as smFISH, have a limited ability to capture temporal dynamics. Due to these considerations, we selected a HMM based on an effective two-state model (derived from a three-state model) of promoter activity to describe our live MS2 imaging data.

We used a cpHMM approach (Lammers et al., 2020) to extract average bursting parameters (k_on, k_off, r) from different sets of MS2-active nuclei. We input MS2 fluorescent traces over time from these sets into the cpHMM. Specifically, we combined nuclei from same-genotype embryos, sorted them by stripe and distributed them across bins of varying fluorescence. To ensure reliable inference, we enforced each bin to contain ~40 nuclei, equivalent to ~2500 time points at a 20 s resolution (Lammers et al., 2020). The number of bins was determined by the amount of data available (Table 2).

Wild-type embryos yielded sufficient nuclei to support the cpHMM inference of bursting parameters for various endogenous stripes (eve stripe 1, 2, 3, 4). eveS1wt-eveS2Gt^- and eveS1∆-eveS2wt did not yield enough ectopically active nuclei for cpHMM analysis (eve stripe 1–2 in eveS1wt-eveS2Gt^-; eve stripe 0 in eveS1∆-eveS2wt). These fly lines did exhibit endogenous eve stripes with enough active-nuclei for further analysis on the cpHMM (eve stripe 1, 2, 3, and 4 in eveS1wt-eveS2Gt^-; eve stripe 1, 2, and 3 in eveS1∆-eveS2wt). eveS1∆-eveS2Gt^- embryos did yield sufficient eve-active nuclei (297 nuclei) to support cpHMM inference of the bursting parameters of ectopic eve stripe 1–2 and eve stripe 0. It also resulted in enough active nuclei for the cpHMM inference of bursting parameters of endogenous stripes (eve stripe 1, 2, 3, and 4).

The output of the effective two-state cpHMM described above are the bursting parameters (k_on, k_off, r) for each set of nuclei input into the model. Thus, Figure 3, Figure 3—figure supplement 2 are plots of mean k_on, k_off, r, and their standard deviations σ_kon, σ_koff, σ_r, computed from sets of nuclei binned by stripe. For Figure 3—figure supplement 3, we followed a similar approach, but grouping active nuclei by their endogenous or ectopic location. Nuclei grouped in endogenous and ectopic categories were distributed across 6–13 bins of increasing fluorescence (Table 3). Their mean k_on, k_off, r, and standard deviations, σ_kon, σ_koff, σ_r were plotted in Figure 3—figure supplement 3.

Raw data, Movies, and CompiledParticles files are stored in the Zenodo dataset ‘Unified bursting strategies in ectopic and endogenous even-skipped expression patterns - Supplemental Data’ (https://doi.org/10.5281/zenodo.7204096; Berrocal et al., 2023). Specific paths in this dataset are listed below. Raw confocal-imaging data from embryos of each of the genotypes used in this work are located in [Genotype]_rawData/[Date]/[Dataset] as czi files (Zeiss file format) of sequential Z-stacks recorded over two channels, and whole embryo stills, as described above. Maximum Z-projection movies of all recorded embryos are in Movies/[Genotype]/Composite. Movies of MS2-foci assigned to stripes are in Movies/[Genotype]/Segmentation. The outcome of Garcia et al., 2013 MATLAB pipeline to analyze MS2 data from each embryo are mat files named CompiledParticles, they are stored in the folder CompiledParticles/[Genotype].

MATLAB scripts and data for this analysis are stored in the github repository https://github.com/aberrocal/BurstingStrategies-eve. The code for the segmentation of our live imaging data of eve transcription in embryonic development is in BurstingStrategies-eve/_DataSubmission/DataSheetsAndCode/StripeSegmentation/. We generated csv files containing the position of active and inactive nuclei over time for each of four genotypes (see BurstingStrategies-eve/_DataSubmission/DataSheetsAndCode/Heatmaps/singleTraceFits_Heatmaps/). In these files, active nuclei have fluorescence values associated with each time point. These datasets also contain the promoter state of active nuclei at each time point. We considered three promoter states: 1=OFF, 2=one sister promoter ON (ON₁), and 3=two sister promoters ON (ON₂); see SI section: Inference of bursting parameters in Materials and methods. The heatmaps in this work (Figure 2, Figure 2—figure supplement 1) were generated with MATLAB scripts and datasets in BurstingStrategies-eve/_DataSubmission/DataSheetsAndCode/Heatmaps/. We generated mat files (compiledResults_[Stripe/ectopicFlag].mat) that contain mean values of k_on (frequency), k_off^–1 (duration), r (amplitude), their standard deviations, and mean fluorescence bin values. compiledResults_Stripe.mat files and scripts to generate Figure 3, Figure 3—figure supplement 2 are sorted by genotype in BurstingStrategies-eve/_DataSubmission/DataSheetsAndCode/KineticsPlotStripes_Color/. compiledResults_ectopicFlag.mat and scripts to generate Figure 3—figure supplement 3 are sorted by genotype in BurstingStrategies-eve/_DataSubmission/DataSheetsAndCode/KineticsPlotsEndogenousEctopic/. Data to generate Tables 2 and 3 is located in BurstingStrategies-eve/_DataSubmission/DataSheetsAndCode/BinStats/particle_counts/. Data sheets with detailed features of individual data points (identity and position of nuclei and MS2 foci; MS2 fluorescence; cpHMM-inference of fluorescence; cpHMM-inferred promoter state) are located in BurstingStrategies-eve/_DataSubmission/DataSheetsAndCode/BinStats/singleTraceFits/. Adobe Illustrator ai, eps, and png files for all figures are stored in BurstingStrategies-eve/_DataSubmission/Figures/.

Fly line expressing wild-type eveMS2-BAC is available at Bloomington Drosophila Stock center (stock #92368). Wild-type and mutant eveMS2-BAC constructs can be obtained through Vector Builder, as described in ‘DNA constructs and fly lines’, or by requesting them from the corresponding authors. For ChIP-seq data of enhancer associated-Zelda refer to ‘Data Availability’ section of Harrison et al., 2011. Details on sample preparation for confocal imaging are provided in the ‘Embryo collection and mounting’ section of Berrocal et al., 2020. Raw confocal imaging data, movies, and compiled particles files are stored in Zenodo, as described in ‘Data and Code’. MATLAB code to segment MS2 signals in fruit fly embryos and generate compiled particles files, as described in ‘Segmentation and quantification of movies’, is publicly available in https://github.com/GarciaLab/mRNADynamics (Reimer and Garcia, 2022). For details on cpHMM scripts and methods refer to ‘Data Availability’ section in Lammers et al., 2020. DNA sequences, figures, and MATLAB code for data analysis are stored in https://github.com/aberrocal/BurstingStrategies-eve, as described in ‘Data and Code’.