Target gene responses differ when transcription factor levels are acutely decreased by nuclear export versus degradation

Morphogens are proteins that form concentration gradients in developing organisms to regulate the expression of target genes (Kicheva and Briscoe, 2023; Rogers and Schier, 2011). Key morphogens include signaling molecules like sonic hedgehog (SHH) and bone morphogenic protein (BMP), as well as transcription factors (TFs) such as Bicoid and Dorsal (DL) (Briscoe and Small, 2015; Greenfeld et al., 2021). While the spatial regulation of morphogen gradients is well understood, investigating their time of action has been more challenging (Kutejova et al., 2009; Nahmad and Lander, 2011; Rushlow and Shvartsman, 2012). Some studies indicate that morphogen signals can be integrated over time to influence target gene expression, highlighting the importance of signal duration (Sagner and Briscoe, 2017; Huang et al., 2017). However, determining the temporal role of a morphogen using genetic knockouts and epistasis can be difficult, as multiple roles may complicate the gene regulatory network progression. Techniques that allow for precise spatiotemporal control of protein levels are therefore essential. Optogenetics enables the control of genetically encoded proteins with light and has proven valuable for studying morphogen patterning (Rogers et al., 2020; Huang et al., 2017; McDaniel et al., 2019; Singh et al., 2022; Johnson et al., 2017). In this study, we have used two optogenetic tags – blue light inducible degradation domain (BLID; Bonger et al., 2014) and a light inducible nuclear export system (LEXY; Niopek et al., 2016) – to assess how different perturbations to nuclear DL levels affect target gene expression during Drosophila melanogaster embryogenesis.

The early stages of Drosophila embryogenesis consist of rapid DNA replication and nuclear divisions called nuclear cycles. Many zygotic genes are activated during nuclear cycle (nc) 14, which is the longest nc of early embryogenesis. During this time, DL forms a nuclear concentration gradient across the dorsal-ventral (DV) axis leading to expression of high threshold target genes in ventral regions and low threshold target genes more dorsally in ventrolateral and lateral regions (Reeves and Stathopoulos, 2009). While DL levels are instrumental in specifying target gene expression in a spatially-instructive manner consistent with morphogen outputs, nuclear DL levels are also dynamic and increase during these early stages, which raises the questions when and for how long does DL concentration have an effect on target gene expression (Rushlow and Shvartsman, 2012).

In a previous study, we used BLID to degrade DL at certain points in time to understand how the loss of nuclear DL during certain nuclear cycles affects target gene expression (Irizarry et al., 2020). We found that snail (sna), which is thought to require high nuclear DL levels, can be activated independently of peak DL levels achieved in late nc14. This suggests that DL input is necessary earlier, but once the DL-target gene twist (twi) is expressed, DL is no longer required. Supporting this model, we discovered that the late activation of sna depends on the Twi TF. Although BLID allows for spatiotemporal control of TF removal under blue light, the reversibility of the system relies on nascent protein synthesis occurring after the light is removed. For DL, no nascent DL-BLID is produced during these embryonic stages. Therefore, while DL-BLID can indicate when DL is no longer needed for sna expression (i.e. late 14), it does not allow us to distinguish the early temporal role of DL from its later role.

To investigate this, the LEXY sequence (Niopek et al., 2016; Kögler et al., 2021; Singh et al., 2022; Zhao et al., 2023 preprint) was added to DL, enabling reversible nuclear protein depletion through blue-light inducible nuclear export. The effects of DL depletion were investigated by live imaging target gene expression using the MS2/MCP system, which detects RNA stem-loops in nascent transcripts (Garcia et al., 2013a; Lucas et al., 2013). By combining BLID, LEXY and live imaging, we found that the sna regulatory system responds differently to nuclear DL loss when comparing DL-LEXY to DL-BLID. Although DL-LEXY is associated with higher nuclear DL levels than DL-BLID, sna is lost with LEXY but retained with BLID under blue light. This difference indicates that import-export kinetics influence target gene expression and are crucial regulatory factors, alongside concentration, for patterning the DV axis.

To develop a system where we could control nuclear DL levels reversibly, we used CRISPR/Cas9 genome editing to construct in-frame fusions of DL and DL-mCherry to the LEXY tag (Niopek et al., 2016), generating the Drosophila stocks dl-LEXY and dl-mCh-LEXY (Fig. 1A,B, ‘LEXY’). To determine whether DL-LEXY and DL-BLID have an effect on target gene expression in the dark, embryos were stained for sna and the normalized widths (i.e. width divided by the individual embryo circumference) were computed. Although the mean sna expression domain width was slightly narrower in dl-LEXY and dl-BLID compared to control, these differences were not significant and suggest the fusions are similarly functional (Fig. S1A,C; Bothma et al., 2015; Irizarry et al., 2020). We also compared the DL gradients in dl-LEXY and dl-BLID with control (yw) to determine whether the gradients were strongly affected by addition of LEXY or BLID in the dark (Fig. S1B). When the DL gradient was fitted to a Gaussian function (Fig. S1D,E; Trisnadi et al., 2013), the peak levels of DL and scaling factor (σ) in dl-LEXY and control were similar (Fig. S1F,G). In dl-BLID, the levels were reduced and there was also a change in the scaling factor (σ) (Fig. S1F,G). One caveat of this fixed data is that embryos were exposed to ambient light during processing, which may have a greater effect on DL-BLID. While dl-LEXY and dl-BLID both supported relatively normal sna expression, the DL gradient in dl-LEXY was more similar to that in control, suggesting that the dl-LEXY background may be more useful for analysis of dynamic requirements for DL in supporting target gene expression.

When embryos with DL-LEXY were treated with blue light, it resulted in export of the fusion protein from the nucleus as detected by following the mCherry signal associated with DL-mCh-LEXY during live imaging (Fig. 1D,E; Movies 1,2). The levels of DL were quantified and tracked over time for each individual nucleus in three embryos (Fig. 1E,F; Movie 2). After the blue light was removed, DL-mCh-LEXY again entered nuclei with an inverse time scale of 0.24±0.03 min⁻¹, 0.31±0.03 min⁻¹ and 0.24±0.03 min⁻¹ for three individual embryos averaged across nuclei. This process of export and reimport can be repeated multiple times (Movie 1), which makes it possible to use live imaging to test how temporarily removing DL affects target gene expression.

To determine whether loss of nuclear DL-LEXY through blue light illumination has a phenotypic effect, embryos were kept in the dark or exposed to blue light and imaged to determine if gastrulation occurred. Under blue light, three of four control embryos (with no DL-LEXY) gastrulated, whereas six of six embryos gastrulated in the dark for the homozygous dl-LEXY stock. This suggests that blue light leads to some lethality. However, when the dl-LEXY stock was exposed to blue light, only two of nine embryos gastrulated. These results demonstrate that loss of nuclear DL results in embryos that fail to undergo gastrulation, which is lethal (Fig. 1C). With this knowledge, we sought to test how two different approaches to perturb DL levels, transiently with DL-LEXY or permanently with DL-BLID, affect gene expression.

DL activates expression of an array of target genes including the high threshold target sna in the ventral region (Fig. 2A,B). To test the effect of removing DL transiently at mid-nc14, when nuclear concentration of DL in ventrally-positioned nuclei is highest (Reeves et al., 2012), embryos from dl-LEXY mothers were illuminated with blue light for 10 min or 20 min and compared to embryos that were kept in the dark during a similar period of time (Fig. 2C). To monitor expression of sna, we used a previously published MS2 reporter that was inserted at an exogenous location (Fig. S2A; Bothma et al., 2015; Irizarry et al., 2020). Using the MS2-MCP imaging system, we were able to observe expression of sna over time from a ventral viewpoint (Fig. 2D). We previously found that sna transcription at later stages of nc14 can be independent of high levels of DL, as nascent transcription is detected in embryos laid by dl-BLID mothers that develop under blue light at mid/late-nc14 (Irizarry et al., 2020). To compare dl-LEXY and dl-BLID, we reimaged dl-BLID under the same conditions as dl-LEXY.

In the dark, sites of active transcription were detected continuously for sna in ventral nuclei during mid-nc14 in dl-LEXY (Fig. 2E; Movie 3) and dl-BLID (Fig. 2F; Movie 3). After 10 min of blue light exposure, sites of nascent sna transcription were undetectable in most nuclei in dl-LEXY (Fig. 2G 00:11; Movie 3) or were detected in only a few nuclei in dl-BLID (Fig. 2H 00:11; Movie 3). In both dl-LEXY and dl-BLID, sites of active sna transcription were detected after returning to the dark (Fig. 2G,H 00:22; Movie 3). Under 20 min of blue light exposure, sites of nascent sna transcription remained undetectable in dl-LEXY (Fig. 2I 00:11 and 00:22; Movie 3). In contrast, when dl-BLID was exposed to 20 min of blue light, the number of active sna transcription sites initially decreased before increasing (Fig. 2J 00:12 and 00:23; Movie 3), similar to the case of 10 min of blue light (Fig. 2H). These results agree with our previous study for dl-BLID in that sna expression is supported despite prolonged blue light illumination (Irizarry et al., 2020), and highlight that dl-LEXY results in a different outcome. Specifically, these data demonstrate that nuclear DL is required to maintain sites of active sna transcription at mid/late-nc14 in dl-LEXY but not in dl-BLID.

To support these experiments, at least six movies were obtained for each of the three conditions: dark, 10 min blue light, and 20 min blue light, for both dl-LEXY and dl-BLID. To quantify the changes in the number of active transcription sites, foci were detected and counted over time, and normalized by the starting number of active sites. Upon blue light illumination, the quantification shows the initial decrease in the number of active transcription sites in both dl-LEXY and dl-BLID (Fig. 2K,L; Fig. S2C,D), and how this number increases at later time points when blue light remains on in dl-BLID, but not in dl-LEXY (Fig. 2K,L, ‘Light 20 min’; Fig. S2C,D). The number of active sna transcription sites in dl-BLID increases after ∼10 min, regardless of whether the embryo was exposed to blue light for 10 min or 20 min (Fig. 2L; Fig. S2D). Since the starting levels of DL vary and could potentially affect the expression of target genes, we compared the mean starting number of transcription sites between dl-BLID and dl-LEXY to determine if there was a statistically significant difference. This is possible because no light has been applied during the first time point of the movies. We found no significant difference in the mean number of starting sna transcription sites for dl-LEXY and dl-BLID before blue light was applied (Fig. S2B), demonstrating that the starting number of transcription sites does not explain the difference we see between dl-BLID and dl-LEXY. The variability likely relates to stochastic uncaging of the optogenetic tags. Importantly, these results demonstrate that blue light illumination results in different trends in sna transcription for dl-BLID versus dl-LEXY, showing that transcription can recover when DL is degraded (BLID) but not when it is exported (LEXY) (Fig. 2K,L; Fig. S2C,D).

As dl-LEXY and dl-BLID result in a different number of active sna transcription sites after blue light illumination (Fig. 2), we sought to determine whether the phenomenon was unique to sna, or if other targets of DL behaved similarly. Another target gene is short gastrulation (sog), which is a low threshold target and is expressed broadly in lateral regions (reviewed by Reeves and Stathopoulos, 2009). Sna acts as a transcriptional repressor to repress sog expression in the ventral domain, refining sog expression to two lateral stripes. DL also acts directly as a repressor to limit the expression of the gene zerknüllt (zen) to dorsal regions. Expression of zen is thought to respond to a similar threshold as sog, which results in zen having an inverse pattern compared to sog (Fig. 2A,B). To test the effects of changing nuclear DL levels using dl-LEXY and dl-BLID on these target genes, we created sog-MS2 and zen-MS2 at their endogenous loci using CRISPR/Cas9 (Fig. S2A). Previously, changes in the sog boundary in fixed samples were only detected if we illuminated dl-BLID embryos for a prolonged period (Irizarry et al., 2020). Thus, we illuminated continuously from the end of nc12 until late nc14 when germ-band extension was observed (Fig. 3B).

To detect differences in the sog and zen expression domains, we imaged in a dorsal-lateral field of view, which captures the dorsally-positioned boundary of sog and the boundary of zen and quantified the change in domain (i.e. area of expression) between nc13 and nc14 (Fig. 3A,C). For sog-MS2, signal was detected in the dark at nc13 and nc14 in dl-LEXY (Fig. 3D; Movie 4) and dl-BLID (Fig. 3E; Movie 4). When illuminated during nc13-14, there was not a significant difference in the position of the dorsal boundary of the sog domain in dl-LEXY (Fig. 3C,F; Movie 4), but there was a ventral retraction of the sog boundary leading to fewer detected sites of transcription in the field of view in dl-BLID (Fig. 3C,G; Movie 4). In contrast to sog, the zen boundary position remained unchanged by any of these similar perturbations (Fig. S3; Movie 7; see Discussion).

To potentially explain the differences we observed in dl-LEXY and dl-BLID for sna and sog, we assayed the levels of nuclear DL in these lines under blue light. In the traditional threshold model, sna expression requires the highest levels of DL (Fig. 4A). In this model, we would predict that DL levels are higher in dl-BLID than in dl-LEXY when under blue light. To test this, embryos laid by dl-mCh-LEXY and dl-mCh-BLID mothers were illuminated with 10 min and 20 min of blue light (Fig. 4B), as done in the previous sna experiment (Fig. 2). In the dark for both dl-mCh-LEXY and dl-mCh-BLID, there was continuous nuclear DL present (Fig. 4C,D,I,J ‘Dark’; Movie 5). Upon 10 min of blue light illumination, DL-mCh-LEXY was exported out of the nucleus, resulting in a decrease in nuclear DL levels and an increase in cytoplasmic DL levels (Fig. 4E,I 00:12; Movie 5). Upon returning to the dark, nuclear DL levels began to increase. Nuclear DL levels in the dark after being illuminated for 10 min were similar to the nuclear DL levels of embryos kept in constant darkness (Fig. 4E 00:24 compare with Fig. 4C 00:24 and quantified in Fig. 4I; Movie 5). On the other hand, DL-mCh-BLID levels decreased in both the cytoplasm and the nucleus under 10 min blue light (Fig. 4F,J 00:12). This trend continued even after the blue light was removed, suggesting that there is a slight delay in ending the degradation of DL (Fig. 4F,J 00:24; Movie 5). Upon 20 min of blue light illumination, DL-mCh-LEXY stayed cytoplasmic and maintained low levels of nuclear DL (Fig. 4G,I; Movie 5), whereas DL-mCh-BLID levels decreased even more than the 10 min exposure (Fig. 4H,J; Movie 5). This is expected, as the longer the degron is uncaged, the more degradation of DL should occur, resulting in lower nuclear levels.

These data show that the resulting levels of nuclear DL in DL-mCh-BLID were lower than the levels in DL-mCh-LEXY when both were under 20 min of blue light. Similarly, the levels of nuclear DL under 10 min of blue light were higher in dl-LEXY than dl-BLID. While this difference in nuclear DL levels can explain the differences in sog expression (i.e. only dl-BLID but not dl-LEXY levels may fall below the threshold required to support sog), it fails to explain the difference in the number of sna transcription sites observed under 20 min of blue light. Since DL is a positive input to sna, it would be expected that sna expression would occur where DL levels are higher, which would be in dl-LEXY. However, we found that there are a greater number of sna transcription sites in dl-BLID where DL levels are lower.

As nuclear DL levels cannot explain all the differences we see between dl-LEXY and dl-BLID, another possible explanation is that blue light-induced export of DL may have an effect on Twi protein levels. We have previously shown that Twi is responsible for DL-independent activation of sna at late nc14 in dl-BLID under blue light. As long as Twi is produced, then DL input can be lost and sna expression is retained (Irizarry et al., 2020). To test whether Twi levels are differentially affected in dl-LEXY versus dl-BLID under 20 min blue light and possibly account for differences observed in sna (Fig. 2), we imaged Twi using a previously published Twi-LlamaTag fly stock (Bothma et al., 2018). Twi-LlamaTag allows detection of Twi protein localization live in vivo when maternally deposited mCherry is available. We found that the levels of Twi did not change upon 20 min blue light illumination and appeared to be similar between dl-LEXY and dl-BLID when assaying Twi-LlamaTag (Fig. S4).

An alternative explanation for how DL-LEXY and DL-BLID differentially affect sna expression is that DL-LEXY sequesters potential cofactors in the cytoplasm under blue light. Implicitly, this means that when DL is exported from the nucleus, any transcription factors or cofactors bound to DL would also shuttle into the cytoplasm. There is some evidence that DL and Twi might physically interact (Shirokawa and Courey, 1997). If Twi was exported with DL under blue light, a change in localization of the Twi protein using the Twi-LlamaTag should be detected; however, as stated above, we did not observe changes in Twi levels under blue light in dl-LEXY (Fig. S4C; Movie 8). This demonstrates that DL nuclear export does not result in an appreciable reduction of nuclear Twi.

While Twi is not exported with DL, this does not rule out that other potential cofactors might be exported with DL in dl-LEXY under blue light. For example, since Twi is required for sna transcription in dl-BLID under blue light, it is possible that DL and Twi share a cofactor, and it is this cofactor that is sequestered with DL in the cytoplasm in dl-LEXY under blue light. However, it would be difficult to test the protein localization of every DL cofactor individually, especially if the cofactor in question is unknown. Thus, we sought to explore and rule out other possible explanations for how dl-LEXY and dl-BLID result in differences in the number of sna transcription sites.

Another model that could explain the differences between dl-LEXY and dl-BLID relates to the dynamics of how nuclear protein levels decrease with these perturbations. DL-LEXY is likely rapidly transiting in and out of the nucleus under blue light, and while degradation of DL-BLID does not change its localization, both cytoplasmic and nuclear levels decrease. When DL-LEXY is exposed to blue light, the normally-buried nuclear export sequence (NES) is revealed as the Jα helix is unfolding (Niopek et al., 2016). However, the nuclear localization sequence (NLS) of DL should be unaffected. Thus, DL would be rapidly transiting between the cytoplasm and the nucleus, due to its own NLS and a strong NES being present. Since DL is predominantly cytoplasmic under blue light, the export rate is likely much higher than the import rate, due to the strong nature of the exposed NES contained in LEXY. We hypothesized that this rapid import and export may act to disrupt DL-independent activation of sna. Specifically, DL might act transiently at the sna locus before being exported, causing a disruption in sna activation through other factors, such as Twi. This model suggests that sna needs DL to remain in the nucleus for a sustained amount of time.

To test whether DL is transiting in and out of the nucleus rapidly, we performed fluorescence recovery after photobleaching (FRAP; Koulouras et al., 2018; Williamson et al., 2021; Bulinski et al., 2001; Cardarelli et al., 2009) on DL-mCh-LEXY in the dark and under blue light (Movie 6). Using FRAP, we bleached a large region of interest (ROI), shown by the black circle, and quantified the nuclear levels in the center-most bleached nucleus, either in the dark (Fig. 5A,B) or under blue light (Fig. 5D,E). This was repeated with three embryos for each condition, and the recovery was fit to a single exponential (Fig. 5C,F; Koulouras et al., 2018; Cardarelli et al., 2009). The parameter β, which relates to the inverse of the recovery time, was estimated as 0.23 min⁻¹, 0.25 min⁻¹ and 0.17 min⁻¹ for the dark and 0.84 min⁻¹, 0.80 min⁻¹ and 1.14 min⁻¹ for the light. The larger β values in the light indicate that steady state is reached faster in the light and that the flux of DL in and out of the nucleus is greater. Thus, FRAP of DL-mCh-LEXY demonstrates that DL-mCh-LEXY levels recover in the nucleus even when export is favored under blue light, suggesting that DL-mCh-LEXY is rapidly moving in and out of the nucleus.

To provide support for the idea that sog expression does not change in dl-LEXY because export cannot reduce DL levels low enough, we sought alternate ways to increase nuclear export. One way of altering the nuclear export rate is by making mutations in the DL native C-terminal NES (NES4; Xylourgidis et al., 2006). In NES4, a single serine residue (S665) has been identified as a site of phosphorylation through a large mass spectrometry screen (Hilger et al., 2009). As export sequences are largely hydrophobic (Kosugi et al., 2014), phosphorylation of NES4 might act directly to decrease the export rate (McGehee and Stathopoulos, 2024). Other ways phosphorylation could affect the export rate, indirectly, is by masking a NES through regulation of a conformational change or interaction with a binding partner to support nuclear retention. In either case, by blocking phosphorylation, we hypothesized that cytoplasmic DL would increase due to an increase in export (Nardozzi et al., 2010). To be sure all possible sites of phosphorylation in NES4 were removed, we mutated S665 and three other local serine residues. Specifically, the four serine residues were changed to alanine residues (dl NES S>A) that cannot be phosphorylated or to aspartic acid residues (dl NES S>D) to mimic a constitutively phosphorylated state due to their negative charge (Fig. 6A). These mutations were made using large rescue constructs that include the known regulatory sequences of DL and were assayed at one copy in a dl¹/dl⁴ mutant background. Venus was included at the C-terminal end of the dl gene in the context of these transgenes to support visualization.

We assayed the dl NES S>A and dl NES S>D lines using antibody staining and fluorescent in situ hybridization. Nuclear DL levels in dl NES S>A are clearly decreased compared to the control dl-Venus as shown by quantification of DL antibody staining (Fig. 6C; Fig. S5). To confirm that total levels of DL were not affected, we detected DL levels in embryo extracts by western blot using DL antibody (Fig. 6B). We found that the total levels of DL are similar, although the distribution of the migrating band changed in dl NES S>A. This loss of differentially migrating bands supports the idea that phosphorylation is disrupted in dl NES S>A. The distribution of molecular weights in dl NES S>D was observed to shift slightly higher, although it is not as clear as the change in the dl NES S>A band.

To determine what effect these changes in the DL gradient had on target genes, we also performed fluorescent in situ hybridization against sna, sog, zen, and the laterally expressed gene intermediate neuroblasts defective (ind), at early nc14 (Fig. 6D) and late nc14 (Fig. 6E) and quantified the results (Fig. 6F). The early and late stages were combined for quantification of the domain width except for zen, which changes expression pattern over nc14 (Jaźwińska et al., 1999), and ind, which was only detected in late nc14 (Fig. 6D-F; Garcia et al., 2013b). Changes were observed for sna, ind and sog expression in the dl NES S>A background (Fig. 6F sna, ind, sog). We did not observe any significant differences in the expression domain widths for the dl NES S>D mutant compared to the control. This suggests that the phosphorylated state may be the normal functional state of DL in the ventral region, but DL might still be localized to the cytoplasm by Cactus in more dorsal regions.

Specifically, we observed that the sna expression domain is clearly reduced in dl NES S>A compared to control (Fig. 6D-F). In addition, sog expression expands (Fig. 6D-F,dl NES S>A), which is expected as Sna represses sog and sna expression is reduced. ind width is slightly larger in the dl NES S>A mutant when compared to the control (Fig. 6D-F). These data suggest that affecting the function of a native nuclear export sequence, by mutating putative phosphorylation sites to non-charged residues (Fig. 6A), results in decreased nuclear DL (Fig. 6C and Fig. S5dl NES S>A) that in turn is associated with a reduction in the sna domain and expansions of sog and ind (Fig. 6D-F dl NES S>A).

The observed differences in target gene responses between dl-LEXY and dl-BLID provide additional insights into how DL action, specifically DL nuclear level, is interpreted by cis-regulatory systems to pattern the DV axis. We observed that the lowest levels of DL achievable in dl-LEXY under blue light are high enough for sog expression; whereas with dl-BLID, DL is degraded and likely goes below the threshold necessary for sog expression under blue light, but only in the very tails of the DL gradient. In contrast, sna expression is lost at late nc14 in dl-LEXY; while in dl-BLID, sna expression can remain on (Fig. 2E-J) even though DL levels are higher in dl-LEXY than dl-BLID (Fig. 4G-J; Movie 5). Lastly, a C-terminal NES serine residue dl mutant, resulting in low nuclear DL levels, exhibits decreased sna and increased sog domain widths but does not affect zen domain width. Collectively, these results support the view that a threshold model is insufficient to explain the observed differences in dl-LEXY, dl-BLID and the S>A mutant and that, in addition to levels, the kinetics of DL import-export play a role in determining the expression of its target genes (Fig. 7A).

Specifically, our results suggest that the kinetics of DL transiting between the nucleus and cytoplasm in dl-LEXY disrupts sna expression (Fig. 7B ‘Disruption due to Kinetics’). In support of this model, FRAP shows that DL is rapidly transiting in and out of the nucleus in dl-LEXY under blue light (Fig. 5). Several other potential explanations were also ruled out. We found no evidence that DL sequesters Twi (Fig. S4). While it is possible that another cofactor is sequestered with DL in dl-LEXY under blue light (Fig. 7B ‘Cofactor Sequestration’), this seems unlikely due to a lack of effect on its known interaction partner Twi. One could also imagine that an indirect role could explain these differences. For example, since levels of DL-LEXY are higher under blue light than DL-BLID, a potential neurogenic repressor repressed by Sna that in turn represses sna could be active in DL-LEXY but not DL-BLID. In this model, sna transcription is lost in DL-LEXY when blue light is applied, and this loss of sna results in derepression of a neurogenic repressor, which then represses sna and prevents it from reactivating. However, this scenario seems improbable within a 10-20 min timescale. It takes ∼3 min for half the nuclei to stop transcribing sna transcripts, which have a half-life of 13.6 min and estimated copy number of ∼180 copies/cell (Boettiger and Levine, 2013). Although Sna protein half-life is unknown, based on estimated rates of translational elongation (Chen et al., 2024; Dufourt et al., 2021) and initiation (Dufourt et al., 2021), it can be produced in about 24.1-110.5 s given its length (390 aa). Therefore, Sna protein would likely still be present in this timeframe. Additionally, transcriptional bursts yield an average of tens of transcripts per minute (Bothma et al., 2014; Fukaya et al., 2016), meaning it would take 10+ minutes to make 100 or more transcripts of this invoked neurogenic repressor. Thus, the timeline for removing Sna and activating a neurogenic repressor seems unlikely within 10-20 min. For all these reasons, we favor the interpretation that DL nuclear import-export kinetics can influence target gene expression.

While our study focused on two threshold responses relating to sna and sog target genes, there is still more to learn with regard to other target genes. For example, we found that zen does not respond to DL levels in either dl-BLID or dl-LEXY when illuminated with blue light from nc13 to nc14 (Fig. S3). This is surprising because sog and zen are thought to share the same threshold. One potential explanation is that zen responds to DL at a timepoint before we imaged (e.g. nc12). Alternatively, DL-repressor activity may be fundamentally different from DL-activator activity despite similar concentration-thresholds. For example, if DL binds more tightly to sites where it acts as a repressor, DL-LEXY and DL-BLID might be unable to support export or degradation at such sites, especially if degradation is primarily happening in the cytoplasm. In future studies, it would be of interest to focus on how export and degradation affect DL repressor activity, which may require earlier perturbations or have no effect at all.

This study sheds light on the timing of morphogen action and highlights the importance of nuclear import-export kinetics, supporting recent models (Barros et al., 2021; Schloop et al., 2020). These findings suggest a broader principle regarding how gene expression responds to morphogen dynamics, emphasizing that it is influenced not only by concentration but also by the temporal presentation of morphogens. For example, transient levels required for low-threshold targets may actively prevent the transcription of high-threshold targets, ensuring the correct positioning of expression domains. Current interpretations often assume a uniform concentration of TFs throughout the nucleus. Therefore, future super-resolution imaging of DL could reveal how variations in local DL concentrations, influenced by import/export dynamics, affect gene expression.

Additionally, future research should explore TF binding site affinity and cooperative binding using optogenetics (Kim et al., 2024; Rao et al., 2021). Previous studies in yeast have examined how TF dynamics encode and decode information, with one optogenetic study indicating that certain targets are regulated by high thresholds (low affinity sites) for greater robustness (Sweeney and McClean, 2023). Combinatorial control may also make genes less sensitive to concentration changes; for example, sna lacks many linked DL and Twi sites, unlike DL target genes in lateral regions (Markstein et al., 2004; Reeves and Stathopoulos, 2009). Furthermore, studies have focused on whether stochastic gene expression is instructive or buffered against variability (Exelby et al., 2021; Naqvi et al., 2023; Boettiger, 2013). sog expression appears to be stochastic (Reeves et al., 2012; MacNamara, 2014), while sna exhibits coordinate expression (Lagha et al., 2013; Boettiger and Levine, 2013). Optogenetic tools could further elucidate the possibly differing temporal requirements for inputs that support stochastic versus coordinate gene expression.

In addition to categorizing gene expression responses supported by morphogens in Drosophila and other higher animals by their thresholds (high, medium, low), these responses should also be classified by their dynamical requirements, such as sustained, transient or absent/reduced morphogen input. Cis-regulatory systems may be tuned to detect sustained levels for high-threshold targets, transient levels for low-threshold targets, and reduced levels for targets repressed by the morphogen (reviewed by Irizarry and Stathopoulos, 2021). Moreover, enhancers with varying dynamic requirements have been shown to coordinate gene expression during Tribolium development (Mau et al., 2023). Future studies could also explore whether dynamics are sensed by enhancers and/or promoters (Hoppe et al., 2020; Delás et al., 2023). In the future, we anticipate the field will discover more dynamic regulatory paradigms through the use of optogenetic tools, providing further insights into how morphogens work.

While optogenetic tools offer precise control of transcription factor levels, they have limitations. Both DL-LEXY and DL-BLID produce different peak levels and widths of the gradient, with significant differences noted for DL-BLID when assayed by antibody staining. This variability is likely due to their different sensitivities to light, as tags may be leaky even in the dark and because embryos are inadvertently exposed to some ambient light during fixation. Despite these limitations, both tags are able to support high threshold target gene expression. Thus, the effects of light perturbations always have to be interpreted in the context of the dark condition, which – to be clear – is not exactly wildtype. Another limitation is that transcription and DL levels are measured in different embryos/lines and not within the same nucleus. Future experiments using far red fluorophores, white light lasers or two-photon microscopy could help address these limitations related to the activation wavelengths of the optogenetic tags (∼400-500 nm).

All D. melanogaster stocks were kept at 22°C in standard medium. Experimental crosses were kept in cages with apple juice agar plates supplemented with yeast paste and were kept at 18°C. Embryos were collected for less than 2 weeks, such that parents were 1-14 days old at time of collection. w; dl-LEXY/CyO; PrDr/TM3 and w; dl-BLID/CyO; PrDr/TM3 were crossed to Sp/Cyo; MCP-mCherry (w+, NLS)/TM3 (Bothma et al., 2018,) to generate w; dl-LEXY/CyO; MCP-mCherry (w+, NLS)/TM3 and w; dl-BLID/CyO; MCP-mCherry (w+, NLS)/TM3, which were grown in bottles, and virgin dl-LEXY; MCP-mCherry (w+, NLS)/TM3 or dl-BLID; MCP-mCherry (w+, NLS)/TM3 were selected. These virgins were crossed to males bearing the MS2. MS2 lines included sna-MS2 BAC (III), sog-MS2 (I); Sp/Cyo, and Sp/Cyo; zen-MS2/TM3 (III). sna-MS2 is a large reporter construct of ∼25 kB with MS2 inserted at the 5′ end of the transcript following the 5′ untranslated region (UTR) and the coding sequence replaced by the gene yellow (Perry et al., 2010; Bothma et al., 2015). Plasmid DNA from (Bothma et al., 2015) was inserted on the third chromosome at 65B2; 3L:6442676 (Irizarry et al., 2020). sog-MS2 and zen-MS2 contain insertions of MS2 within introns. In addition, dl-mCherry-LEXY/CyO and dl-mCherry-BLID/CyO were grown in bottles and homozygous mothers and fathers were added to experimental cages. y² cho² v¹ P{nos-phiC31\int.NLS}X; attP2 (III) (NIG-FLY TBX-0003) was used to make y² cho² v¹ P{nos-phiC31\int.NLS}X; P{dl-gRNA}attP2 (III). y² cho² v¹; Sp/CyO, P{nos-Cas9, y+, v+}2A (NIG-FLY CAS-0004) virgins were crossed to y² cho² v¹ P{nos-phiC31\int.NLS}X; P{dl-gRNA}attP2 (III) for injection. See Table S1 for a complete list of lines used. The CRISPR/Cas9 lines generated in this study (dl-LEXY, dl-mCh-LEXY, sog-MS2 and zen-MS2) contain a DsRed marker.

LEXY (Bonger et al., 2014) was codon optimized and, along with MS2 (Yamada et al., 2019), synthesized by GenScript in pUC57. The dl-LEXY, dl-mCh-LEXY, sog-MS2 and zen-MS2 homologous repair templates were generated by editing pHD-DsRed (Gratz et al., 2014). The right homology arm for dl-LEXY and dl-mCh-LEXY was generated by PCR using a dl-Venus-BAC (Reeves et al., 2012) as a template, and was inserted into pHD-DsRed downstream of the DsRed using BglII and XhoI sites. The left homology arm was generated by overlap PCR, combining three fragments: the C-term of dl, the LEXY domain, and the dl 3′ UTR. The left homology arm of dl-mCh-LEXY was made by overlap PCR, combining PCR products that used dl-mCherry HDR and the dl-LEXY HDR as a template. This PCR product was inserted into pHD-DsRed upstream of the DsRed using EcoRI and NheI sites. The sog-MS2 homologous repair template was made by PCR, using a BAC as the template (BacPac Resource Center, BACR25D05). Overlap PCR was used to mutate the gRNA binding site in the repair template. The left homology PCR product was cut with NheI and AseI and the pHD-DsRed plasmid was cut with NheI and NdeI to make compatible sticky ends, which were ligated together. The right homology arm PCR product and the pHD-DsRed were digested with AscI and XhoI and ligated. The zen-MS2 homologous repair template was made the same way as the sog-MS2 template but used NheI and NdeI on both the insert and the backbone, and the right homology arm also used overlap PCR to mutate the gRNA sequence. The MS2 sequence was added using NotI and AvrII, which were added to the reverse primer used to generate the left homology arm of both sog-MS2 and zen-MS2.

The zen-MS2 gRNA was made by BbsI digestion of pCFD5 and Gibson assembly was used to combine the vectorized backbone and the PCR product. In both sog-MS2 and zen-MS2, the MS2 sequence was inserted into the first intron, as annotated on Flybase. See Table S1 for a complete list of primers used.

For dl-LEXY and dl-mCh-LEXY, y² cho² v¹; Sp/CyO, P{nos-Cas9, y+, v+}2A virgins were crossed to y² cho² v¹ P{nos-phiC31\int.NLS}X; P{dl-gRNA}attP2 (III). The HDR template for dl-LEXY and dl-mCh-LEXY were injected into embryos from this cross. The sog-MS2 HDR was co-injected with a previously made gRNA (Dunipace et al., 2019) into w[1118]; PBac{y[+mDint2]=vas-Cas9}VK00027 (Bloomington Drosophila Stock Center, #51324). For zen-MS2, gRNAs were found using flyCRISPR Target Finder (Gratz et al., 2014). The zen-MS2 HDR was co-injected into y2 cho2 v1; attP40{nos-Cas9}/CyO (NIG-FLY CAS-0001). For both sog-MS2 and zen-MS2, Rainbow Transgenics performed the injections. All HDR templates included DsRed as a selectable marker, and transgenics were screened for DsRed expression.

Embryos from crosses between dl-LEXY females and males from the MS2 line were collected for 4 h or overnight, both at 18°C. To prepare the embryos for live imaging, embryos were hand dechorionated in the dark, using a red film (Neewer, 10087407). Embryos were transferred to an agar square and oriented so that the face that would be imaged was facing the agar. Preprepared slides were made by adding heptane glue (heptane plus double sided tape) to a coverslip that was taped to the slide and allowing it to sit overnight. This slide was used to pick the embryos up from the agar. Embryos were then checked to make sure the orientation had not been disrupted and oriented again if necessary. Water was then added to prevent desiccation of the embryos. Embryos were transferred to the microscope in a covered box. Imaging occurred on a Zeiss LMS 800 using a 25× immersible objective (LCI Plan-Neofluar 25×/0.8 Imm Korr DIC M27) at 1.7 zoom. The MCP-mCherry signal was detected using a 561 nm laser at 1% laser power with 800 V gain on a GaAsP PMT detector. The 488 nm laser at 4.5% laser power was used to perform blue light illumination with 500 V gain on a GaAsP PMT detector to protect the detector from the high laser power. Z-stacks were taken, with 30 z-planes per time point at 1 µm thickness. Images were taken every ∼25 s, starting as soon as the previous z-stack finished. Images were captured as 16 bit images, and each z-slice was 512×512 pixels, with each pixel being 0.29 µm in length and width. DL-mCh-LEXY and DL-mCh-BLID were imaged using the same settings as the MS2/MCP imaging, except the laser power of the 561 nm laser was 2% instead of 1%. Upon beginning imaging, embryos were staged under red light (white light covered with a red filter). Embryos that were the right stage and orientation were imaged. Orientation was determined by the signal being imaged. For DL-mCh-LEXY, DL-mCh-BLID and sna-MS2, this meant that the signal domain was centered. For sog-MS2 and zen-MS2 this meant detecting the boundary of the expression domain. FIJI/ImageJ was used to visualize images and to save movie files (Schindelin et al., 2012).

FRAP was imaged similarly to previous imaging setups. Images were 512×512 pixels, 16 bit, and taken on a Zeiss LMS 800 using a 25× immersible objective (LCI Plan-Neofluar 25×/0.8 Imm Korr DIC M27) at 5.0 zoom with a pixel size of 0.100 µm. The DL-mCh-LEXY signal was detected using a 561 nm laser at 1% laser power with 900 V gain on a GaAsP PMT detector. The 488 nm laser at 4.5% laser power was used to perform blue light illumination with 500 V gain on a GaAsP PMT detector to protect the detector from the high laser power. Time points were acquired at a rate of one frame per 2 s. Embryos were illuminated with blue light starting at time point six (∼18 s) and remained on until the end of imaging at ∼20 min. Bleaching was performed starting at time point 21 and ending at time point 40 (∼1 min) in an ROI using the 561 nm laser at 20% laser power and performing 50 iterations of bleaching at each time point.

For fixed sample preparation of the dl-Venus, dl NES S>A and dl NES S>D, embryos were collected for 1 h, aged 2-3 h and then were dechorionated in bleach, fixed in 4 ml of 9.25% formaldehyde and 4 ml of heptane for 20 min and then rinsed and stored in methanol at −20°C. For in situ hybridization, protocols were followed as described previously (Kosman et al., 2004) using riboprobes generated for sna, ind, sog and zen. Sheep anti-digoxigenin (Life Technology, PA185378), rabbit anti-FITC (Invitrogen, A889) and mouse anti-Biotin (Invitrogen, 03-3700) were used (1:400). Fluorescently conjugated secondary antibodies, Alexa 555, 488 and 647, from Thermo Fisher Scientific were used (1:400). See Table S1 for a complete list of reagents used. Mouse anti-dorsal 7A4 [Developmental Studies Hybridoma Bank (DSHB); Whalen and Steward, 1993] was used for DL antibody staining (1:10), following the same protocol as FISH. Embryos that were the right stage based on visualization of a nuclear stain (DAPI) were imaged and staging was confirmed by zen and ind expression. Fixed sample preparation for DL-LEXY and DL-BLID were performed the same way. During processing, embryos were shielded from ambient light as much as possible; however, embryos were exposed to light, which may have affected DL levels in these embryos. This effect would be greater in DL-BLID, which cannot recover, unlike DL-LEXY, which is reversible.

For western blot analysis, embryos were collected and staged under white light. Embryos at mid nc14 were then added to a tube containing SDS buffer, ruptured using a fine needle and homogenized. Embryo extracts were then run on a discontinuous SDS-PAGE gel and transferred to 0.45 µm Immobilon-P PVDF. Chemiluminescent detection was performed using a DL antibody (anti-Dorsal 7A4, DSHB, final concentration of 480 ng/ml in 4 ml total volume; Whalen and Steward, 1993), and β-tubulin (E7, DSHB, final concentration of 590 ng/ml in 4 ml total volume; Chu and Klymkowsky, 1989) antibody was used as a loading control.

To quantify the number of MS2 foci, or sites of transcription, in the images/movies captured, three custom MATLAB functions (https://github.com/StathopoulosLab/McGehee_2024) were used. The first function opens the image, including relevant metadata, and performs the foci detection. First ‘salt and pepper’ noise is removed using a median filter. The background is subtracted by using a median filter over a larger area to blur the image and then subtracting the blurred image from the original. After background subtraction, the image is blurred with a Gaussian filter. The image is then thresholded by a user defined threshold, tiny objects of only one pixel are removed, and objects detected on the edge are removed. The entire embryo is segmented by projecting all the time points together, blurring the image with a Gaussian filter and using thresholding. The detected embryo is then morphologically closed to smooth the edges and small objects less than 100 pixels are discarded so only one object, the embryo, is detected. Any focus detected outside the embryo is removed. We observed that the background intensity of nuclei increased over time and, to account for this, we increased the threshold by a small amount using a user defined value that increases logarithmically during nc14. To increase the detection of foci, we segmented the unprocessed image a second time using a user defined threshold and retained only the foci detected with both thresholds. The algorithm works by setting the first threshold low, detecting both real foci and noise, and then removing the noise based on a second, higher threshold. The centroid coordinates for these foci are saved for further analysis.

A second function displays the images in a graphical user interface and allows overlapping the mask of the segmented foci on the image. In addition to overlaying, the centroids can be used to plot points at the detected foci. These were used to evaluate the success of the thresholding. Comparable imaging conditions used the same empirically determined threshold. Specifically, when comparing dark and light or dl-LEXY and dl-BLID the same threshold was used. The threshold was only changed for different MS2 signals (i.e. sna versus sog) or different lengths of imaging (i.e. nc12-nc14 versus 25 min of nc14). A third function was used to quantify the number of foci and plot the results. To plot the averages, the data was interpolated using the MATLAB built in interp1 function and the Modified Akima cubic Hermite interpolation method. The mean number of foci and standard error of the mean were calculated from the interpolated data and plotted. In addition, this function also approximated the area of expression. This was done by concatenating all the centroids in given time windows, corresponding to nc13 or the first 100 time points of nc14, and removing foci that were two median absolute deviations (MAD) from the median for the centroids of all foci detected. We used a conservative approach because this removed points that tended to be isolated, were not detected in multiple frames, or were actually noise and not a real focus. To determine the area, a convex hull was drawn around the remaining points and the area for this convex hull was determined. The area at nc13 was then subtracted from the area at nc14 to determine the change in nc13 to nc14 and account for potential discrepancies in the orientation of the embryo. This was only done for sog-MS2 and zen-MS2.

To determine whether the differences in area were statistically significant, we performed one way ANOVA and Tukey's honestly significant difference (HSD) test for multiple comparisons to compare dl-LEXY dark, dl-LEXY light, dl-BLID dark and dl-BLID light. We performed this analysis for the areas determined for both sog-MS2 and zen-MS2. A P-value less than 0.05 was considered statistically significant.

To quantify DL-mCh-LEXY, DL-mCh-BLID (Fig. 4) and Twi-LlamaTag-mCh (Fig. S4) nuclear levels, images were segmented using the MATLAB edge function and a Laplacian of Gaussian with a standard deviation of four for the filter. Water-shedding was performed using the MATLAB watershed function to disconnect any nuclei that were connected. Objects were filtered by size to remove large (i.e. two connected nuclei) or small (i.e. non nuclei or partial nuclei) objects. During time points when no nuclei were detected (such as when blue light was turned on or off), the image segmentation from the previous time point was used. The embryo was detected by blurring the image and using a normalized threshold of 0.005, followed by morphological opening and closing. The boundaries of the embryo were detected and used to fit an ellipse. From the ellipse, the midline was determined and only nuclei within 100 pixels above or below the midline were included for analysis. The average intensity was calculated for each individual nucleus at a given time point and then was averaged together for each time point and plotted as the mean±s.d. To quantify individual nuclei (Fig. 1) a similar procedure was used, except the standard deviation for the Laplacian of Gaussian was ten and the embryo and midline were not detected due to the high level of zoom. In addition, individual nuclei were tracked by computing the distance between nuclei centers between time points. The minimum distance under a threshold (set as the radius of a single nucleus) was used to join tracks. In the event a nearest neighbor was not found, previous time points were used to find similar positioned nuclei up to ten time points away. Tracks that were missing the first ten time points, last ten time points, or had less than half the total number of time points were removed. This ensured that enough data was provided to the fitting algorithm. Fitting was performed as described below for FRAP.

To quantify the DL gradient in fixed samples, images were inputted into a previous developed function which measures the intensity of DL along the DV axis and fits the DL gradient to a Gaussian function (Trisnadi et al., 2013).

To quantify the gene expression patterns, first the ring of nuclei was segmented by using the MATLAB edge function and a Laplacian of Gaussian with a standard deviation of 20 for the filter on the DAPI or histone channel. The two largest filled objects were taken, which are the outer and inner ring of the nuclei. Ellipses were fit to both of these rings, and the average ellipse parameters were taken to get the ellipse centered between these two rings. The channels for the gene expression were segmented using a threshold and the largest objects were used to remove background. The domain width was determined as the points where the ellipse crossed the segmented objects of gene expression. To calculate arc length, the equation of an ellipse was integrated between these two points. The perimeter was calculated by integrating around the entire ellipse. The domain width was then normalized by the perimeter. The widths were averaged together across embryos, and if a pattern contained two domains, both widths were included in the average. Tukey's HSD for multiple comparisons after performing one way ANOVA was used to determine whether the means were significantly different between experimental and control conditions.

To determine the sample size, a power analysis was done for each experiment. For the number of MS2 transcription sites in sna-MS2 a power analysis using a predicted mean of 200 transcription sites, standard deviation of 40, power of 0.80 and significance value of 0.05 for a two-sample unpaired t-test was carried out in MATLAB using sampsizepwr. This resulted in a sample size of n=3. For the change in area of sog-MS2 and zen-MS2, this procedure was repeated using a mean area of 2000, standard deviation of 1500, power of 0.80 and significance value of 0.05 for a two-sample unpaired t-test. This also resulted in a sample size of n=3. Similarly, we performed this analysis for the quantification of the DL gradient using a mean amplitude of 100, standard deviation of 25, power of 0.80 and significance value of 0.05 for a two-sample unpaired t-test. This resulted in a sample size of n=6.

We are grateful to Hernan Garcia and Mike Levine for sharing fly stocks and plasmids, and Vince Stepanik and Anil Ozdemir for technical support. We also thank Leslie Dunipace and the Lagha group for helpful discussions. We acknowledge the use of ChatGPT, an AI language model developed by OpenAI, for assisting in shortening text in the final version of this manuscript.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202775.reviewer-comments.pdf