Generation and timing of graded responses to morphogen gradients

Morphogen gradients manifest positional information by graded activation of signaling pathways that culminate in the activation of transcription factors. This activation leads to the induction of target genes that establish the fate of the tissue. The most common outcome of morphogen gradients is the division of a target tissue into distinct zones of gene expression, as originally proposed by the ‘French flag’ model (Rogers and Schier, 2011; Wolpert, 1969). The genes triggered within each zone respond to the level of the activating transcription factor, such that different sets of genes are induced accordingly (Sagner and Briscoe, 2019).

Although the morphogen dictates the division of a naive field into distinct uniform zones, there may be instances in which additional patterning within these domains is necessary. For example, in response to the morphogen, coordinated cell movement and tissue morphogenesis may require manifestation of the graded positional information. This output is inherently distinct, in that cells within a given zone would respond differently according to the level of the morphogen. To examine whether a morphogen gradient can also elicit patterning within a designated zone, we have investigated zygotic patterning along the dorsoventral (DV) axis of the Drosophila embryo, where both aspects are operating.

Polarity along the DV axis of the Drosophila embryo is established by the Spaetzle/Toll signaling pathway (Moussian and Roth, 2005). The ligand Spaetzle (Spz) is distributed in a graded manner within the extracellular perivitelline fluid, which surrounds the embryo, with peak levels at the ventral midline (Haskel-Ittah et al., 2012; Rahimi et al., 2019). Following binding by Spz, activation of the Toll receptor leads to nuclear targeting from the embryo cytoplasm of the NFκB-related transcription factor Dorsal, a process that is graded along the DV axis (Roth et al., 1989; Rushlow et al., 1989; Steward, 1989). The Dorsal nuclear-localization gradient then dictates subdivision of the DV axis into three main zones within the blastoderm embryo (Liberman et al., 2009; Moussian and Roth, 2005).

The Drosophila embryo is a syncytium during the initial 13 synchronous nuclear division cycles that follow fertilization. The crucial manifestation of zygotic DV axis target-gene expression takes place during the initial 45 min of nuclear cycle (NC) 14, in parallel with embryo cellularization, during which all ∼6000 cortical nuclei are enclosed in individual cells. Live measurements have shown that graded Dorsal nuclear localization is re-established within 10-15 min of the onset of NC 14 (Rahimi et al., 2019).

Within the mesoderm, gastrulation involves coordinated three-dimensional cell shape changes and actomyosin-driven movement of mesodermal cells (Fig. S1) (Heer and Martin, 2017; Martin et al., 2009). After completion of cellularization, the ventral cells expressing zygotic twist (twi) and snail (sna) invaginate by actomyosin-mediated constriction of their apical surfaces to form the ventral furrow (VF). This constriction is driven by activation of the small GTPase Rho1, which leads to recruitment of the actin nucleator Diaphanous (Dia), as well as Rho kinase (Rock; also known as Rok), which activates myosin II (Fig. 1A). VF formation initiates as a wedge-shaped internalization of the ventral-most cells of the domain, a pattern that results from the graded (ventral to lateral) apical recruitment of myosin (Denk-Lobnig et al., 2021; Heer and Martin, 2017). In mutant backgrounds in which apical myosin recruitment is uniform, the entire cell cohort invaginates simultaneously. Thus, it is necessary to control gene expression within the ventral domain in a manner that will lead to graded myosin recruitment.

Most of the proteins comprising the contractile machinery are provided maternally and, therefore, are ubiquitous and uniformly distributed within the early embryo. Thus, the ability of the Dorsal-nuclear gradient to shape the spatial pattern of mesoderm internalization relies on directing the patterned expression of zygotic target genes that will drive actomyosin contractility. Three key zygotic target genes that are expressed in the ventral part of the embryo, folded gastrulation (fog), mist (also known as mthl1) and T48, are involved in myosin recruitment. They trigger the key initiating event, activation of Rho1, via its activator RhoGEF2 (Mason et al., 2016). Fog is a secreted ligand that activates G protein-coupled receptors (GPCRs), and Mist is the corresponding GPCR. Binding of Fog to Mist in ventral cells results in the activation and release of the membrane-tethered Gα protein Concertina, leading, in turn, to recruitment from the cytoplasm and activation of RhoGEF2 (Manning et al., 2013; Perez-Vale and Peifer, 2020). In addition, the transmembrane protein T48, which is localized on the apical membrane, mediates RhoGEF2 recruitment to this cellular domain (Kolsch et al., 2007), where myosin II is subsequently required for apical cell constriction (Fig. 1A; Fig. S1).

We examined the impact of the Dorsal-nuclear gradient on the endogenous expression of the zygotic transmembrane regulators of actomyosin contractility, T48 and mist (Fig. 1B). We found that the Dorsal gradient leads to a corresponding graded accumulation of mRNA for both genes through two complementary mechanisms. Activation of gene expression expands over time from the ventral region with high nuclear Dorsal to lateral regions, where the levels are lower. Thus, nuclear Dorsal levels dictate the duration of gene activation. In addition, once a transcription site is activated, the rate of RNA Polymerase II (Pol II, also known as Polr2F) loading depends on the level of nuclear Dorsal. A combination of these two responses contributes to the graded accumulation of T48 and mist mRNA in the cytoplasm. In addition to the expression pattern of these two regulators, the timing of their activity is also crucial because the cell movements of gastrulation should begin only after the completion of cellularization. We found that such timing is dictated by long introns in T48 and fog, which delay the appearance of mature mRNAs that can undergo translation.

twi, encoding a bHLH transcription factor, is one of the two cardinal zygotic genes (along with sna) that define the ventral mesodermal zone downstream of the Dorsal gradient and subsequently drive the expression of a plethora of tissue-specific genes. To monitor the expression of twi quantitatively, we used single-molecule fluorescence in situ hybridization (smFISH), which allows visualization of the active transcription sites (TSs) in the nucleus, as well as the accumulation of individual mRNA molecules in the cytoplasm. Given that the analysis is carried out on fixed embryos, each embryo represents a single time point, and the combined analysis of such ‘snapshots’ of multiple staged embryos allows for the reconstruction of the dynamic expression profile.

We found that twi expression displays a switch-like behavior, manifesting a sharp on/off threshold in response to Dorsal-nuclear localization levels. Active twi TSs appeared simultaneously throughout the entire mesoderm region. The spatial distribution of these sites within the mesoderm was uniform along the DV axis (Fig. 2A-C). Most of the mesodermal nuclei displayed two TSs, indicating that both alleles are transcriptionally active. The median fluorescence intensity of twi TSs along the DV axis was also uniform, indicating a similar rate of Pol II loading onto all twi loci throughout the mesodermal domain (Fig. 2D). The uniform activation of twi transcription is also corroborated by a recent study that followed live accumulation of Twi and identified an early (within ∼5 min) and even appearance of Twi throughout the mesoderm during NC 14 (Dufourt et al., 2021). The rapid activation of twi expression by Dorsal could be facilitated by the presence of multiple binding sites for the Zelda (Zld) pioneer factor, which clears the regulatory region from nucleosomes (Sun et al., 2015) (Fig. S2). Moreover, binding of Zld to the twi regulatory region is initiated as early as nuclear cycles 8 to 9 (Harrison et al., 2011). In addition, binding of Twi to its own regulatory region (Jiang and Levine, 1993; Zeitlinger et al., 2007) may represent a ‘feed forward’ mechanism, which enhances robust twi expression in nuclei in which the levels of Dorsal are above a designated threshold.

To examine the transcriptional basis for graded actomyosin contractility within the mesoderm, we followed the expression patterns of T48 and mist, key zygotic target genes that encode transmembrane proteins involved in eliciting a cell autonomous actomyosin response. We followed the expression of T48 and twi in parallel in the same embryos. Similar to twi, the T48 regulatory region also contains multiple sites bound by Dorsal (Sun et al., 2015) (Fig. S2). However, whereas twi expression encompassed the entire mesoderm from the outset, the expression of T48 was initially observed only in the central (ventral-most) part of the mesoderm (Fig. 2E). By contrast, in older embryos, the distribution of TSs for twi and T48 showed a complete overlap (Fig. 2F-H″). We conclude that, whereas twi expression is triggered simultaneously in all mesodermal nuclei, the onset of T48 expression expands over time. A similar temporal expansion has also been observed in live imaging of a MS2-tagged T48 reporter (Keller et al., 2020; Lim et al., 2017).

Twi is necessary for activation of T48 expression, because no transcription of T48 takes place in twi-mutant embryos (Kolsch et al., 2007; Leptin, 1991). However, the distinct temporal profiles of twi and T48 expression imply that additional factors besides Twi regulate T48 expression. The most likely candidate for such a factor is Dorsal, which exhibits graded nuclear levels within the mesoderm. Binding sites for both Twi and Dorsal have been identified on the T48 enhancer (Zeitlinger et al., 2007; J. Zeitlinger, personal communication). In addition, prominent binding of Zld to the T48 regulatory region occurred during nuclear cycles 13 to 14 (Harrison et al., 2011) (Fig. S2).

From our analysis of the number of T48 TSs that were activated in each nucleus at different time points, we concluded that the graded activation reflected the independent response of individual T48 loci, because many nuclei displayed only one active TS. In other words, although all loci may be primed for expression by binding of maternal Zld, there is a separate probability, dependent on the level of nuclear Dorsal, of activating each T48 locus to make it permanently accessible for the recruitment of transcription factors and Pol II. Over time, nuclei started to display two active TSs, first in the ventral region and subsequently in more lateral domains. This pattern was distinct from that of twi TSs, which mainly occurred as two per nucleus throughout the future mesoderm (Fig. 3A-C). Similar to other early developmental genes, twi and T48 displayed prominent pausing of Pol II prior to transcription activation (Fig. S3) (Saunders et al., 2013). The initial activation event of the transcription start site should entail the release of paused Pol II before the recruitment of subsequent Pol II complexes.

The embryo in Fig. 3A-C is ∼25 min into cycle 14, as estimated from the size and shape of the nuclei (Lecuit and Wieschaus, 2000). Given that establishing a graded pattern of nuclear Dorsal requires 10-15 min following the onset of NC 14 (Rahimi et al., 2019), the nuclei of this embryo were exposed to the Dorsal gradient for ∼15 min. The fraction of nuclei displaying activated T48 TSs was measured along the ventral domain of this embryo. This value was then used to derive the activation probability per minute at different positions along the DV axis (Fig. 3D,E).

Using this approach, we found that the probability of T48 TS activation peaked at the ventral midline and declined towards the lateral region. The probability distribution resembles the shape of the Dorsal-nuclear gradient. This tight correlation (Pearson's r=0.94) indicates that the probability of TS activation declined linearly with the reduction in Dorsal levels. Such a monotonic response to Dorsal is in sharp contrast to the nonlinear response of the twi enhancer to Dorsal, which generates the sharp borders of twi transcription. The activation probability of twi expression was indeed less correlated with Dorsal nuclear levels (Fig. 3F).

The instructive role of nuclear Dorsal levels in determining the probability of triggering the T48 loci was examined in embryos expressing the constitutively active Toll^ΔLRR construct, resulting in uniformly high levels of nuclear Dorsal along the length of the DV axis (Rahimi et al., 2016). Nuclei around the entire embryo circumference exhibited a similar response (Fig. 3G). In younger embryos, two twi TSs per nucleus appeared in most nuclei, whereas mostly one or (in a minority) two T48 TSs were observed (Fig. 3H,H′). This intermediate response of T48 TSs reflects the probability of activation, which requires some time to result in a significant fraction of active TSs. Indeed, in older Toll^ΔLRR embryos, most nuclei presented two active TSs for both twi and T48 (Fig. 3I,I′).

Median intensity of the individual twi TSs across the mesoderm appeared constant (Fig. 2D), indicating a comparable rate of Pol II loading. Measuring and analyzing the intensity of T48 TSs as a proxy for the number of polymerases traversing the gene is more complex, given the gradual, time-dependent progression of expression initiation of this gene along the DV axis. Therefore, it is imperative to only consider those T48 loci that are already at a steady-state and are producing mature transcripts. Thus, we used distinct smFISH probes for the 5′ and 3′ domains of T48 and limited our T48 TS intensity analysis to loci that displayed both a 5′ and a 3′ probe signal, indicating that some Pol II molecules had reached the 3′ region of the gene (Fig. 4A-D). Monitoring the intensity of the 5′ probe for these sites provided a cumulative measure of the number of polymerases along the gene. A graded median intensity gradient was observed across the ventral region (Fig. 4C), with a tight correlation of the Pol II loading rate with the position of nuclei (and, hence, with nuclear Dorsal levels) (Pearson's r=0.47) (Fig. 4E).

To determine the average number of Pol II complexes that are loaded on T48, we compared the intensity of nuclear TSs with that of individual mRNA spots in the cytoplasm, using a probe covering the entire coding sequence (Fig. 4F). The cumulative and maximal pixel intensity and volume of each spot were monitored. The rationale for this approach was that the cumulative spot intensity indicates the number of mRNAs, whereas the maximal intensity distinguishes highly concentrated mRNA molecules at the TS in the nuclei from the more-dispersed mRNA clusters in the cytoplasm. A histogram across the spot intensity (in Log scale) for each of these groups identified two peaks, corresponding to the median intensity of a single mRNA and the TS, respectively (Fig. 4G,H). The median intensity of the TSs was 70 times higher than that of the single mRNAs, providing an estimate of the average extent of Pol II loading onto T48 loci (Fig. 4G,H). Given that T48 is 29 kb long, this translates to a density of one Pol II complex every ∼400 bp, which is about tenfold lower than the maximal loading that is physically possible. Thus, control of the number of Pol II molecules loaded on the T48 locus is likely within the linear dynamic range.

A combination of the two mechanisms outlined above (graded onset of TS activation and graded polymerase loading rates of the gene across the ventral region) should give rise to a graded distribution of T48 mRNA. Indeed, quantitation of T48 mRNA in the cytoplasm showed a graded pattern that was highly correlated with nuclear Dorsal levels (Pearson's r=0.92±0.02) (Fig. 5A-C). Given that Twi distribution in the mesoderm was uniform, we assume that nuclear Dorsal underlies the graded expression of T48.

The quantitative measurement of T48 TS activation (Fig. 3D,E) allows simulation of the dynamic pattern of production and accumulation of mRNA (Fig. S4). The fraction of activated TSs increases over time in a graded manner. If every activated site is transcribing at a similar rate, this would still lead to the graded accumulation of mRNA, because the ventral TSs were activated earlier (Fig. 5D,E). We assume that the transcribed mRNAs are stable within the relevant time window.

In addition to the temporal order of TS activation, the actual rate of Pol II loading depends on the level of nuclear Dorsal. When this aspect was added to the simulation, the shape of the accumulating mRNA peak became sharper (Fig. 5F). Given that the activation of TSs reached saturation at the midline (i.e. almost all TSs become activated), this additional feature contributes to the shape of the mRNA peak within the relevant time window.

Expression of mist was analyzed in a similar way to that of T48 and exhibited comparable features. Binding by Twi, Dorsal and Zld has been identified on the mist enhancer (Sun et al., 2015) (Fig. S2). The mist enhancer also displayed paused Pol II (Fig. S3) (Saunders et al., 2013). Graded activation of TSs was monitored over time, with the initial activation taking place along the ventral midline (Fig. 6A,B). A gradient of mist TS intensities, reflecting a position-dependent Pol II loading rate, was apparent (Fig. 6A′,B′). Of note, because mist is a relatively short gene (9.3 kb), we assumed that transcription had reached a steady-state on all TSs in older embryos. mist mRNA accumulated in the cytoplasm in a graded manner (Fig. 6A″,B″). Monitoring the number of activated TSs per nucleus revealed more nuclei with two TSs in the ventral region (Fig. 6C). Finally, uniform nuclear targeting of Dorsal following expression of Toll^ΔLRR resulted in a uniform level of mist activation along the DV axis (Fig. S5).

To determine the Pol II loading rate onto the mist loci, the intensity of TSs and single mRNA spots was compared (Fig. 6D,E). From this, a median number of ∼30 Pol II complexes per gene was estimated. Taking the size of the gene (9.3 kb) into account, this distribution was 1.4 times denser than that calculated for T48, but still well below saturation levels.

The graded accumulation of T48 and mist mRNAs is effectively driven early in NC 14, following completion of the final syncytial cycle of nuclear divisions. However, the protein products that trigger apical actomyosin contractility, including RhoGEF2 (Mason et al., 2016) and Myosin II (Dawes-Hoang et al., 2005), are recruited apically and function only after cellularization is complete, 30-40 min after the onset of NC 14 (see also Movie 1), necessitating a delay. An indication that regulation of mature mRNA production may contribute to this timing mechanism was observed when comparing the cytoplasmic mRNA accumulation patterns of twi and T48. As reported above (Fig. 2), early NC 14 embryos displayed prominent, active TSs for both twi and T48. However, whereas twi mRNA was readily detected in the cytoplasm of such embryos, no T48 mRNA was observed in an embryo that was ∼26 min into NC 14 (embryo age was determined throughout from the apical-basal length of DAPI-stained nuclei) (Lecuit and Wieschaus, 2000) (Fig. 7A-A″; Fig. S6). Given that twi is a short gene of 2.2 kb, no delay in completion of transcription is expected to take place. In contrast, T48 is 29 kb long, because of the presence of large introns encompassing ∼25 kb. This organization is exceptional for early zygotic genes, which are typically intron-less or have only short introns (Artieri and Fraser, 2014). Given a transcription rate of ∼2 kb/min (Fukaya et al., 2017), such long introns could delay the production of mature mRNA and protein until after the completion of cellularization. Interestingly, fog also displayed similarly sized large introns. Although the presence of Fog is key for the timing of Mist activation, its distribution may not be instructive because it is secreted to the fluid extracellular milieu.

If the intron prolongs T48 transcription, using distinct T48 5′ and 3′ probes we should expect temporal and spatial differences (because of the expansion of TS activation) in the detection of TSs. The onset of T48 5′ transcription was observed in the ventral-most nuclei by 7 min into NC 14, with no apparent detection of TSs displaying the 3′ probe. The positive nuclei predominantly displayed a single 5′ TS (Fig. 7B-B‴; Fig. S6). By 24 min, 5′-probe-displaying TSs expanded laterally and more nuclei displaying two spots appeared. TSs that also showed the 3′ probe began to appear in the ventral-most nuclei (Fig. 7C-C‴; Fig. S6). The ∼17 min delay between the two panels provides an estimate of the time required for Pol II to complete T48 transcription at 20°C. By 38 min, TSs displaying both 5′ and 3′ probes were detected throughout the mesoderm, with most nuclei showing two spots (Fig. 7D-D‴; Fig. S6). T48 mRNA was then observed by both probes in the cytoplasm.

If the delay is caused by transcription through the T48 intron, then transcription of intron-less T48 within a given nucleus should be completed earlier than that of the endogenous locus. To test this prediction, we fused a T48 cDNA construct to its 5′ regulatory region (Lim et al., 2017) and generated transgenic flies that integrated the construct at a locus that was different from the endogenous one. By using distinct intron and 3′ probes, we could distinguish the completion of transcription at endogenous loci by their reactivity with both probes, whereas the cDNA locus reacted only with the 3′ probe (Fig. 7E). Analysis of the distribution of both TS types demonstrated that completion of transcription for the cDNA TSs occurred earlier than that of the endogenous T48 locus. Whereas the mature endogenous TSs were detected mostly in the ventral domain, TSs exhibiting the cDNA 3′ domain were distributed approximately uniformly throughout the mesoderm (Fig. 7F). Accordingly, the fraction of cDNA TSs was higher at the lateral domain (Fig. 7G).

The primary hallmark of pattern formation based on morphogen gradients is the ability to use a graded morphogen distribution for the induction of distinct zones of gene expression. We explored whether the information conveyed by the morphogen gradient can also be used to generate a graded response within defined domains of the target tissue. In situations in which the morphogen gradient is the only symmetry-breaking cue, morphogenetic movements of distinct zones in the target tissue may also require a graded transcriptional response to the morphogen.

We examined this paradigm during the establishment of the DV axis in the Drosophila embryo, in which the gradient of nuclear localization of the Dorsal protein represents the only asymmetric cue for patterning of the embryo along this axis. The future mesoderm undergoes graded constriction by an actomyosin network in order to fold and internalize in a wedge-shaped pattern. If this behavior is guided by a zygotic mechanism in the embryo, its origins may well lie in a mesoderm-specific transcriptional response to the Dorsal nuclear gradient.

To examine the underlying transcriptional mechanism for graded actomyosin constriction, we focused on two zygotic genes encoding transmembrane proteins that regulate the activation of the small GTPase Rho1, a key initiating event in the process. T48 was shown to recruit RhoGEF2 to the apical surface (Kolsch et al., 2007), and Mist is a GPCR that is activated apically by the ligand Fog, releasing the Gα protein Concertina, which activates RhoGEF2 (Manning et al., 2013; Mason et al., 2016; Perez-Vale and Peifer, 2020) (Fig. 1).

We found two distinct Dorsal-dependent mechanisms that generate graded mRNA accumulation for T48 and mist. Initially, expression of both genes was triggered only in the central (ventral-most) domain of the mesoderm, expanding laterally over time. This expansion was consistent with a TS activation response to the Dorsal nuclear gradient, which is maximal at the ventral midline. Distribution of the number of activated TSs per nucleus indicated that the probability of activation is independent for each allele. Thus, two alleles within the same nucleus that experience similar levels of nuclear Dorsal may be switched on at different times. The probability of TS activation was linearly correlated with the levels of nuclear Dorsal.

The molecular nature of the Dorsal-dependent mechanism that leads to activation of T48 and mist TSs is not known. Transcription of early zygotic genes in the embryo is facilitated by binding of the pioneer factor Zld, which removes nucleosomes from the regulatory regions (Sun et al., 2015). Enrichment of Zld binding to these loci indicates that nucleosomes are cleared from these regulatory regions. Furthermore, enhancement of Zld binding was shown to increase the transcription rate of a T48 reporter construct (Keller et al., 2020). These enhancers also display binding sites for Twi, which is more prominent and uniform across the mesoderm and, thus, stably associates with the loci across the future mesoderm. Whereas nuclear Dorsal levels are limiting, once Dorsal is recruited to the locus, a stable association between Twi and Dorsal proteins may stabilize Dorsal binding and permanently alter the future accessibility of the site to incoming Pol II. Binding of Dorsal and Twi to target genes, such as single minded, was shown to facilitate their transcription by enhancing the probability of DNA binding by the Notch intracellular domain (Falo-Sanjuan et al., 2019).

The different transcription onset times are expected to generate graded mRNA accumulation, with higher levels adjacent to the nuclei that were induced earlier. The effectiveness of this mechanism is restricted to a defined time window. When most of the transcription sites in the ventral region are triggered over time, the differences in mRNA production diminish. Indeed, monitoring apical Myosin II recruitment showed that lateral mesodermal cells eventually recruit similar levels to the ventral-most cells (Bhide et al., 2021). For optimal function, the probability of transcription activation should be high enough to prime a significant number of TSs and yet not plateau within the critical time window.

Once a TS is activated, the rate of Pol II loading is also dependent upon nuclear Dorsal levels. We can imagine a scenario in which binding of Dorsal to the enhancers has on and off rates, such that the fraction of time that Dorsal is bound would depend on its concentration. Analysis of a reporter T48 construct indicated that Dorsal-dependent Pol II activity is modulated via the on rate, whereas the off rate is fairly constant (Keller et al., 2020). The different Pol II loading rates would also contribute to the generation of an mRNA gradient. Furthermore, this mechanism would continue to operate even after the activation of TSs begins to plateau. Simulations based on our measurements showed that a combination of the two mechanisms can lead to a robust gradient of mRNA accumulation. Graded alteration in transcriptional burst size was also recently demonstrated for BMP target genes in the dorsal ectoderm of the Drosophila embryo (Hoppe et al., 2020). However, it is not clear whether this graded response has a biological role in this setting. The proposed mechanisms underlying graded activation of transcription by Dorsal are summarized in Fig. 8.

As for any biological mechanism that employs  probability, the question of the precision and robustness of the outcome arises. Wedge-shaped mesoderm folding relies on relative differences in contractility between the cells, not on absolute levels, such that the system is less restricted. Furthermore, the connectivity between cells, mediated by cell junctions, may average out local perturbations. The tension along the anterior-posterior axis of the embryo was shown to be greater than that along the DV axis, leading to the distinct elongated shape of mesoderm invagination (Chanet et al., 2017). This global tension may also act as a buffering element for local perturbations.

Timing is a central issue when coordinating the processes of cellularization and mesoderm invagination in Drosophila embryos during NC 14. The Dorsal nuclear gradient is stabilized within 10-15 min, shortly after onset of NC 14 (Rahimi et al., 2019). The generation of the mRNA gradient for T48 and mist, encoding proteins regulating actomyosin constriction, takes place early in NC 14, whereas the process of cellularization, which is driven by elongation of the cell membranes towards the basal side, is completed 15-30 min later. Therefore, it is imperative to ensure that the graded apical recruitment of Myosin II driving apical constriction of the gastrulating cells is delayed until this time (Dawes-Hoang et al., 2005), because premature apical constriction before cellularization is completed may be deleterious.

Most early zygotic genes, including those expressed in the future mesoderm, are intron-less or exhibit introns of only a few kb (De Renzis et al., 2007). T48 and fog stand out, displaying introns of 25 kb. This large intron size is conserved in the homologous genes of different Drosophila species. Given that the measured speed of Pol II progression is ∼2 kb/min (Fukaya et al., 2017), these introns could provide a crucial delay of 10-15 min in the production of the mature mRNA. We have demonstrated that this delay does indeed take place, by comparing the detection timing of 5′ versus 3′ T48 probes and by demonstrating that an mRNA produced by an intron-less T48 matured before the transcript produced from the endogenous locus (Fig. 7). It is interesting that the intron sizes of T48 and fog are similar, facilitating a coordinated timing of mature mRNA appearance. T48 may generate a graded pattern, whereas the production of Fog, a diffusible extracellular ligand, could provide a temporal switch. If the Fog receptor, Mist, is produced in a graded manner, it will lead to graded activation. Given that mist is only 9.3 kb long, Mist may appear before the presentation of Fog. The manifestation of intron delay in timing was previously demonstrated for kni related (also known as knirps-like) in the Drosophila embryo (Rothe et al., 1992), and for her7/Hes7 in the segmentation clock in zebrafish and mice (Lewis, 2003; Takashima et al., 2011). Intron length may provide an effective temporal design mechanism in situations in which intricate control of timing is necessary (Swinburne and Silver, 2008).

In conclusion, the Dorsal nuclear gradient is used not only for subdivision of the embryonic DV axis into distinct domains, but also for specification of graded responses leading to tissue shape changes that culminate in gastrulation. The convergence of two distinct mechanisms for generating transcriptional gradients of genes regulating actomyosin contractility sharpens the resulting response. Finally, temporal coordination between the graded transcriptional response and execution of cell movement is achieved by a delayed maturation of the relevant transcripts, dictated by intron size.

The bulk of experiments were performed using wild-type (yw) embryos.

For ectopic Toll activation, we analyzed embryos generated by females carrying UASp-Toll^ΔLRR and the nos-Gal4 driver (Rahimi et al., 2016). Intronless T48 was generated by fusing the cDNA to the T48 mesodermal enhancer and the eve minimal promoter (Lim et al., 2017). The construct was generated by GenScript, cloned into pATTB and injected into attP40 flies by BestGene. Movie 1 shows the visualization by lightsheet microscopy of a Myosin II (Zipper)-GFP embryo (BDSC line B-51564).

Probe sets were designed by Stellaris Probe Designer and purchased from LGC Biosearch Technologies (see Table S1). Three hours after egg lay (AEL), embryos were fixed for 25 min in 4% formaldehyde, washed in methanol and kept at −20°C. The following day, embryos were washed in methanol and then in ethanol, rocked in 90% xylene:10% ethanol for 1 h, post-fixed in 4% paraformaldehyde/PBS/0.05 M EGTA for 25 min, washed three times (10 min each), then incubated for 6 min with 10 µg/ml Proteinase K recombinant, PCR Grade (Roche, RPROTKSOL-RO) and post-fixed again. Embryos were transferred gradually to 10% Formamide Deionized (FA; Thermo Fisher Scientific; AM9342) in 2× SSC+10 μg/ml ssDNA preheated to 37°C and prehybridized for 30 min at 37°C. Hybridization buffer included 10% FA, 10% dextran, 2 mg/ml bovine serum albumin (Thermo Fisher Scientific; AM2616), ribonucleoside vanadyl complex (RVC; New England Biolabs; S1402) and ssDNA+ tRNA in 2× SSC, containing the probe set (1 ng/μl) (Trcek et al., 2017). Hybridization was carried out overnight at 37°C. Next morning, the embryos were shaken gently and incubated for another 30 min. Embryos were then washed twice for 30 min each at 37°C with 10% FA in 2× SSC+10 μg/ml ssDNA and gradually transferred to PBS-0.5% Tween 20, Molecular Biology Grade (Promega, H5151) and mounted with Vectashield+DAPI Mounting Medium Vectashield+DAPI Mounting Medium (Vector Laboratories, H-1200-10). Fluorescence was visualized with a Zeiss LSM800 confocal microscope or Nikon Eclipse Ti2 microscope (Rahimi et al., 2019).

After acquisition, raw images were processed in MATLAB (MathWorks). First, and for each channel, the image stack was maximally projected along the z-axis and the TS spots in each channel were then detected using the method of Raj et al. (2008). In brief, z-projected images were processed with a log-filter (size=15 px, σ=2.5 px), and the detection threshold was defined as the first threshold in the environment of which the number of detected TSs stayed relatively constant [change (in log) smaller than half the maximal relative change across all thresholds (in log)]. Then, eight-connected pixels above the threshold were defined as distinct TSs and their intensity defined as the sum of their pixel intensities. For each embryo, the midline was then defined according to twi fluorescence (or T48 if twi was not available) and the distance (in pixels) between each TS and the midline was calculated. To convert the pixel measurement into nuclei columns, the DAPI image was used to determine the average column distance for each embryo and all distances were converted accordingly. For comparisons with the Dorsal nuclear level, Dorsal levels in each nuclear column were taken from Ambrosi et al. (2014) and rescaled so that the maximal Dorsal level at the midline corresponded to 1. For each embryo, only the TSs (and mRNAs) inside a manually selected region of interest (∼400 nuclei) were analyzed. mRNA levels at each distance from the midline were calculated as the median fluorescence of all pixels with this midline distance. All codes for TS and mRNA analyses can be found on GitHub (https://github.com/barkailab).

After acquisition, raw image stacks were processed in MATLAB. For each fluorescence channel, the ‘local’ background fluorescence (median fluorescence in a moving sphere) was first determined and subtracted from the raw signal. The background-corrected image stacks were processed with a three-dimensional log-filter, and the spots on the processed image stack detected with the mRNAcount function from Raj et al. (2008). For each 18-connected spot (mRNA and TSs), the cumulative and maximal pixel intensities were determined from the fluorescence value of all corresponding pixels. The spots were then classified as mRNA spots (single, cytoplasmic molecules) or TS spots (multiple, transcribed mRNAs at the TS) based on their maximal intensity (Fig. 4A).

Based on our results, TS activation was assumed to be a stochastic, independent process and, thus, the activated promoter fraction (at a certain nuclear column) can be described by 1−exp(−T*p), with T being the time in NC 14 and p being the activation probability. This formula was used to calculate the activation probabilities in Fig. 3E. This experimentally derived probability was used as an input for the dynamic model in Fig. 5. In addition, the activation probability outside the midline was assumed to be linearly related to Dorsal levels [as measured by Ambrosi et al. (2014)] and the mRNA production rate after TS activation was assumed to be either constant (=1) between the different TSs or fixed at the midline (=1) and linearly related to the Dorsal levels at other positions. After transcription, mRNAs were assumed to be stable and, thus, to accumulate over time.

To calculate the correlation between nuclear dorsal level and the activated fraction or activation probability (Fig. 3D-F), we first calculated the activated fraction and activation probability from the number of active TSs in the different cell rows (compare with Fig. 3B). Pearson’s correlation was used to determine the linear correlation between these parameters and the nuclear Dorsal level in the corresponding cell rows, according to Ambrosi et al. (2014). To calculate the correlation between nuclear dorsal level and intensity for individual TSs (Fig. 4E), we first linearly interpolated the Dorsal values of individual TSs (and nuclei) from the row-based measurement (Ambrosi et al., 2014), based on their exact distance from the midline. Pearson’s correlation was then used to calculate the linear correlation between each spot intensity of the TSs and the interpolated nuclear Dorsal level. For visualization, the median TS intensity in the different cell rows and its fit to the nuclear Dorsal level is shown.

We thank Melanie Weilert and Julia Zeitlinger for sharing information and fruitful discussions; Shalev Itzkovitz for input and discussions; and Neta Rahimi for the Zipper-GFP lightsheet movie.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199991.