Skip to main content
NCBI home page
eLife logoLink to eLife
. 2019 Jul 11;8:e45325. doi: 10.7554/eLife.45325

Multi-enhancer transcriptional hubs confer phenotypic robustness

Albert Tsai 1,†,✉, Mariana RP Alves 1,2,†, Justin Crocker 1,✉
Editors: David N Arnosti3, Jessica K Tyler4
PMCID: PMC6650246  PMID: 31294690

Abstract

We previously showed in Drosophila melanogaster embryos that low-affinity Ultrabithorax (Ubx)-responsive shavenbaby (svb) enhancers drive expression using localized transcriptional environments and that active svb enhancers on different chromosomes tended to colocalize (Tsai et al., 2017). Here, we test the hypothesis that these multi-enhancer ‘hubs’ improve phenotypic resilience to stress by buffering against decreases in transcription factor concentrations and transcriptional output. Deleting a redundant enhancer from the svb locus led to reduced trichome numbers in embryos raised at elevated temperatures. Using high-resolution fluorescence microscopy, we observed lower Ubx concentration and transcriptional output in this deletion allele. Transcription sites of the full svb cis-regulatory region inserted into a different chromosome colocalized with the svb locus, increasing Ubx concentration, the transcriptional output of svb, and partially rescuing the phenotype. Thus, multiple enhancers could reinforce a local transcriptional hub to buffer against environmental stresses and genetic perturbations, providing a mechanism for phenotypical robustness.

Research organism: D. melanogaster

Introduction

During embryogenesis, transcriptional regulation controls precise patterns of gene-expression, leading to cell-fate specification (Long et al., 2016; Mallo and Alonso, 2013; Reiter et al., 2017; Spitz and Furlong, 2012). This involves coordinating a complex series of interactions between transcription factors and their target binding sites on DNA, leading to the recruitment or exclusion of active RNA polymerases, which determines the transcriptional state of the gene. Live imaging experiments have shown that transcription factor binding in eukaryotic cell lines and embryos is dynamic but transient, occurring frequently but with each event lasting for at most a few seconds (Chen et al., 2014; Izeddin et al., 2014; Liu et al., 2014; Normanno et al., 2015). Additionally, recent studies have shown that many developmental enhancers harbor functionally important low-affinity binding sites (Antosova et al., 2016; Crocker et al., 2010; Crocker et al., 2015; Crocker et al., 2016; Farley et al., 2015; Farley et al., 2016; Gaudet and Mango, 2002; Lebrecht et al., 2005; Lorberbaum et al., 2016; Rister et al., 2015; Rowan et al., 2010; Tanay, 2006). One example is the Homeobox (Hox) family that is responsible for body segment identity along the anterior-posterior axis in animals. Because Hox transcription factors descended from a common ancestor, their preferences for binding sequences are very similar (Berger et al., 2008; McGinnis and Krumlauf, 1992; Noyes et al., 2008). To select for specific Hox factors, several enhancers in the shavenbaby (svb) locus make use of low-affinity binding sequences for Ultrabithorax (Ubx) (Crocker et al., 2015). Svb is a transcription factor that drives the formation of trichomes, epidermal projections on the surface of the segmented fly embryo (Chanut-Delalande et al., 2006; Delon et al., 2003; Payre et al., 1999). Thus, a key question was how low affinity binding sequences are able to drive strong transcriptional activation in developing embryos.

We have previously shown in living Drosophila melanogaster embryos that Ubx transiently but repeatedly explores the same physical locations in a nucleus, which are likely clusters of binding sites (Tsai et al., 2017). We have additionally shown that transcriptional microenvironments of high local Ubx and cofactor concentrations surround active transcription sites driven by low-affinity svb enhancers. As the distributions of many transcription factors in the nucleus are highly heterogeneous, the transcriptional activity of low-affinity enhancers would depend on the local microenvironments. Interestingly, we observed that transcriptionally active, minimalized versions of two of the three ventral svb enhancers (E3 and 7) are Ubx-responsive and preferentially appear near or overlap spatially with transcription sites of the endogenous svb gene, despite being on different chromosomes (Crocker et al., 2015; Tsai et al., 2017). This colocalization suggests microenvironments could be shared between related enhancers to increase transcriptional output, where enhancers would synergistically form a larger local trap for transcription factors than each could alone. Retaining multiple enhancers within a microenvironment could also provide redundancy in case individual enhancers are compromised and buffer negative impacts when the system is subjected to stress. This idea is consistent with the observed phenotypic robustness of svb enhancers in maintaining sufficient trichome numbers even under temperature stress (Crocker et al., 2015; Frankel et al., 2010). These results are consistent with multi-component transcriptional ‘hubs’ that are local areas enriched for components of the transcriptional machinery and transcription factors through multiple attractive and cooperative interactions (Boija et al., 2018; Cisse et al., 2013; Furlong and Levine, 2018; Ghavi-Helm et al., 2014; Lim et al., 2018; Mir et al., 2017; Mir et al., 2018). One potential building block for these ‘hubs’ are multiple, long-range, enhancer-to-enhancer interactions. However, it is not yet understood how such multivalent interactions function mechanistically, and how they contribute to phenotypic robustness.

To understand the mechanistic implications of having multiple enhancers in a shared microenvironment, here we examined the ability of the svb locus to maintain transcriptional output and produce the correct phenotype under temperature-induced stress in flies harboring a deletion of a partially redundant enhancer—the DG3 ventral enhancer. When embryos were raised at high temperatures, we observed phenotypical defects in ventral trichome formation for the DG3-deletion svb allele but not for the wild-type. At the molecular level, Ubx concentrations around transcription sites of the DG3-deletion allele decreased. The transcriptional output of svb without DG3 also decreased. To test the hypothesis that shared microenvironments modulate transcriptional output and provide buffering under stress, we sought to rescue the DG3-deletion allele through inserting the complete svb cis-regulatory region on a BAC (svbBAC) on a different chromosome. We observed that Ubx concentration around active transcription sites of the DG3-deletion allele and their transcriptional output increased when the svbBAC is physically nearby. Moreover, we found that trichome formation was partially rescued at high temperature. As a result, our findings support the hypothesis that shared microenvironments provide a mechanism for phenotypic robustness.

Results

The DG3 enhancer responds specifically to Ubx in the A1 segment

The ventral svb enhancers DG3, E3 and 7 (Figure 1A) contain low-affinity Ubx binding sites and have been shown to be transcribed in microenvironments of high Ubx concentrations in the first abdominal (A1) segment on the ventral surface of the embryo (Tsai et al., 2017). Each of these enhancers produces ventral stripes of expression along segments A1-A7 in the embryo, resembling the endogenous expression pattern of svb (Figure 1B). Each enhancer contributes to different but partially overlapping portions of the total expression pattern. Furthermore, they have different Ubx ChIP enrichment profiles (Figure 1—figure supplement 1). Whereas the interaction of E3 and 7 with Ubx had been previously explored in detail (Crocker et al., 2015), DG3 remained unexplored. Therefore, we tested the response of the DG3 enhancer to Ubx by altering Ubx levels and measuring the transcriptional output with a reporter gene (lacZ). In wild-type embryos, the DG3 reporter gene was expressed ventrally in stripes along segments A1-A7, in addition to narrow thoracic stripes in T1-T3 (Figure 1C). Expression from DG3 on the ventral surface in the T2 and T3 segments was weak with wild-type Ubx expression and was primarily seen on the sides. In the absence of Ubx, DG3 reporter expression was almost completely lost on the ventral side of A1 and reduced between A2-A7 (Figure 1D), consistent with the responses of E3, 7 and the full svb locus (Crocker et al., 2015). The incomplete loss of expression in A2-A7 suggests that additional factors influence the expression of ventral trichomes in those segments. Ubiquitous expression of Ubx increased the expression levels in A1-A7, in addition to generating ectopic expressions on the ventral side of the thoracic segments T2-T3 and A8 (Figure 1E). In summary, we showed that Ubx is necessary for DG3 expression in the ventral region of the abdominal segments (completely in A1 and partially in A2-7) and can induce ectopic expressions when overexpressed. These data are consistent with our previous observation of the localization of DG3-driven transcription sites within Ubx microenvironments (Tsai et al., 2017).

Figure 1. Ubx drives the expression of the DG3 shavenbaby enhancer along the ventral abdominal segments.

(A) The cis-regulatory region of the shavenbaby (svb) gene contains three enhancers expressing stripes on the ventral side of the abdominal segments: DG3, E3 and 7. (B) The expression patterns of the three enhancers are partially overlapping. The color scheme corresponds to (A), where DG3 is magenta, E3 is yellow and 7 is cyan. (C) Expression pattern of a reporter construct with the DG3 enhancer driving LacZ expression in an embryo with wild-type Ubx expression, as visualized using immunofluorescence staining. (D) DG3 reporter in Ubx null mutant shows no expression in A1 and significantly weakened expression in the other abdominal segments. (E) Overexpression of Ubx driven through a heat shock promoter induces overexpression of DG3 reporter in all abdominal segments and ectopic expression on the ventral surface of the thoracic segments T2 and T3.

Figure 1.

Figure 1—figure supplement 1. Ubx enrichment around the three ventral svb enhancers.

Figure 1—figure supplement 1.

ChIP experiment of whole embryos between stages 10 and 12 targeting Ubx shows different enrichment profile around the three ventral svb enhancers: DG3, E3 and 7.
Open in a new tab

Deletion of a region including DG3 enhancer causes defects in ventral trichome formation specifically at elevated temperatures

Given the clear ventral stripes that DG3 generated in the abdominal segments, we next explored the phenotypic impact of its activity in driving trichome formation. It had been previously shown that deleting a region in the svb locus containing DG3 (Df(X)svb108) led to reduced phenotypic robustness of svb under non-optimal temperatures, with reduced numbers of trichomes produced (Frankel et al., 2010). This svb DG3-deletion allele encompasses the enhancers DG2, DG3 and Z (Figure 2A)—of which only DG3 is a ventral enhancer.

Figure 2. Deletion of a region from the svb locus containing DG3 reduces ventral trichome numbers under heat-induced stress.

(A) The Df(X)svb108 allele contains a deletion in the cis-regulatory of svb spanning three enhancers: DG2, DG3 and Z. Of those, only DG3 expresses on the ventral side. (B) Wild-type phenotype of trichomes along the A1 and A2 segments at 25°C. (C) At 25°C, the Df(X)svb108 deletion allele did not show a mutant phenotype along the A1 and A2 segments. (D) Zoomed out shot of the anterior region of a cuticle preparation of a wild-type (w1118) larva. (E) Zoomed out shot of the anterior region of a cuticle preparation of a larva carrying Df(X)svb108. The lack of trichomes along the T1 segment is a recessive marker used in subsequent experiments to select for embryos/larvae carrying this deletion allele. Even at 25°C, where the overall trichome numbers in A1 and A2 for the Df(X)svb108 deletion mutant is indistinguishable from wild-type svb, the trichomes at the side of the ventral stripe for A1 and A2 were lost, as marked by the black brackets. (F) Within the overall expression pattern of svb enhancers, DG3 provides exclusive coverage in the circled regions in segments A1 and A2. These regions correspond to the black brackets in panels D and E. (G) Wild-type phenotype of trichomes along the A1 and A2 segments at 32°C. (H) At 32°C, the Df(X)svb108 deletion allele showed a mutant phenotype. (I) A1 trichomes in the dashed boxes bounded by the two sensory cells, as indicated by the arrows, as shown in panels G and H, were counted. Deficiencies of the deletion allele only become clear when the animal is subjected to elevated temperature at 32°C, showing reduced trichome numbers in the A1 segment. The number of larvae counted was: 13 for wild-type at 25°C, 19 for Df(X)svb108 at 25°C, 14 for wild-type at 32°C and 30 for Df(X)svb108 at 32°C. Two-tailed t-test was applied for each individual comparison. In box plots, center line is the mean, upper and lower limits are standard deviation and whiskers show 95% confidence intervals.

Figure 2.

Figure 2—figure supplement 1. Loss of trichomes in the A2 segment.

Figure 2—figure supplement 1.

(A–C) Phenotype of trichomes along the A2 segment for wild-type and Df(X)svb108 (shown as svb108 in the panels) at 25°C and wild-type at 32°C. (D) At 32°C, the Df(X)svb108 deletion allele showed a mutant phenotype. (E) Deficiencies of the deletion allele become clear when the animal is subjected to elevated temperature at 32°C, showing reduced trichome numbers in the A2 segment (p-value<0.001). The number of larvae counted was: 10 for wild-type at 25°C, 11 or Df(X)svb108 at 25°C, 11 for wild-type at 32°C and 10 for Df(X)svb108 at 32°C. In box plots, center line is mean, upper and lower limits are standard deviation and whiskers show 95% confidence intervals.
Open in a new tab

In the A1 and A2 segments at 25°C, deletion of the DG3 enhancer did not result in a clear change in ventral trichome formation in the abdominal segments compared to the wild-type (Figure 2B and C), perhaps due to the redundancy provided by overlapping expression patterns from other svb enhancers. However, the T1 trichomes were missing in larvae homozygous for the deletion (Df(X)svb108) allele (Figure 2D and E), which we subsequently used as a homozygous marker to select for larvae homozygous for the deletion allele when crossing Df(X)svb108 flies to other lines (See ‘Cuticle preparations and trichome counting’ in Materials and Methods). Also, we observed defects in trichome formation in the dorsal edges of the stripe pattern, which are exclusively covered by DG3 (Figure 2D and E, the black brackets at A1 and A2). This is consistent with a lack of redundancy in enhancer usage in these areas (Figure 2F, white dotted circles). The trichome number in regions covered by the overlapping expression of the E3, 7 and DG3 enhancers in the A1 segment did not significantly reduce at 25°C upon the deletion of DG3 (Figure 2B,C and I). However, larvae homozygous for the Df(X)svb108 allele developed at 32°C produced fewer trichomes compared to wild-type larvae (Figure 2G,H and I). We also observed similar effects in the A2 segment (Figure 2—figure supplement 1). These results are similar to those shown with quartenary A5 trichomes (Frankel et al., 2010). However, the mechanisms behind this loss of phenotypic robustness under heat-induced stress are yet to be understood in detail.

Transcription sites from the DG3-deletion allele have weaker Ubx microenvironment and lower transcriptional output

To address molecular sources that may lead to the reduced number of ventral trichomes we observed for the Df(X)svb108 deletion allele, we imaged Ubx distributions and the transcriptional output of the svb gene in fixed Drosophila melanogaster embryos using high-resolution confocal microscopy. We reasoned that the defect could be due to changes in the transcription factor concentration around the enhancers (input) and/or the transcriptional output of the gene (output). As Ubx is specifically needed to drive DG3 and svb expression on the ventral surface of the A1 segment, we used it as a metric for the transcription factor distributions around svb transcription sites. The samples were stained with immunofluorescence (IF) for Ubx and RNA fluorescence in situ hybridization (FISH) for svb transcription sites as previously described (Tsai et al., 2017). We imaged both embryos containing the wild-type svb allele or the Df(X)svb108 allele, raised at either 25°C or 32°C. For all imaging experiments involving the svb allele with the deletion, we selected only homozygous embryos for imaging (See ‘Imaging fixed embryos’ in Materials and Methods). Because DG3 expression in the A1 segment showed a clear link to changes in Ubx level, we focused most of our subsequent imaging quantifications in this segment.

To gauge the Ubx concentration around a transcription site, we counted the averaged intensity in the Ubx IF channel within a circle four pixels in diameter (170 nm, roughly the resolution limit of AiryScan) centered on the transcription site (Figure 3A and B, see ‘Analysis of microenvironment and svb transcription intensity’ in materials and methods). In nuclei from the A1 segment, Ubx intensities around svb transcription sites with the wild-type allele did not significantly change between 25°C and 32°C (Figure 3C, bottom right panel). Measuring Ubx intensity from random locations within nuclei of wild-type embryos expressing svb at 25°C using the same method showed that Ubx concentrations around svb transcription sites were in general higher than the nuclear average (Figure 3—figure supplement 1), consistent with our previous findings (Tsai et al., 2017). Transcription sites in embryos with the DG3-deletion (Df(X)svb108) allele had a local Ubx concentration that is slightly lower than wild-type at 25°C (Figure 3C, bottom right panel). However, there was a clear decrease in Ubx intensity compared to the wild-type when we subjected the DG3-deletion embryos to heat-stress (Figure 3B, right panel, and 3C, bottom right panel). To measure the transcriptional output of svb, we adopted the same approach, but quantified the intensity in the svb RNA FISH channel (Figure 3C, upper left panel). Interestingly, we detected clear decreases in transcriptional output when the embryos are heat-stressed at 32°C, even with the wild-type allele. The Df(X)svb108 allele at 25°C showed reduced levels of transcriptional output slightly lower than the wild-type under heat-shock. At 32°C, the transcriptional output further decreased in the mutant. When the various conditions were plotted by their svb transcriptional output and Ubx intensity (Figure 3C, center panel), they displayed a weak positive correlation with Ubx concentration (R-square = 0.4331). In sum, stress conditions reduced the transcriptional output of enhancers and the correlation to Ubx concentrations was positive but weak.

Figure 3. Deletion of the cis-region of svb containing DG3 led to defects in the Ubx microenvironment and svb transcriptional output.

(A) Panels showing a nucleus from embryos with either the wild-type (w1118) or Df(X)svb108 deletion svb allele at either at normal (25°C) or elevated temperature (32°C), imaged using confocal fluorescence microscopy. Ubx (shown in magenta) is stained using immunofluorescence (IF) and the svb transcription sites (shown in green) are stained using fluorescence in situ hybridization (FISH). (B) Zoomed-in panes centered on svb transcription sites, with the height of the surface plots representing the Ubx intensity. (C) Correlation between Ubx and svb transcription intensities at different conditions. The number of transcription sites quantified was: 71 for wild-type at 25°C, 51 for wild-type at 32°C, 50 for Df(X)svb108 at 25°C and 38 for Df(X)svb108 at 32°C. Bottom right panel: Integrating the Ubx intensity surrounding transcription sites shows a small defect in the Ubx concentration around the deletion allele. The drop in Ubx increased at elevated temperature. Upper left panel: The integrated intensity of svb transcriptional output shows that there is a drop in transcriptional output for the deletion allele compared to the wild-type at both 25°C and 32°C. Even the wild-type showed reduced transcriptional output at elevated temperature (32°C). We analyzed four embryos for each genotype/temperature combination. Two-tailed t-test was applied for each individual comparison. In box plots, center line is mean, upper and lower limits are standard deviation and whiskers show 95% confidence intervals.

Figure 3.

Figure 3—figure supplement 1. Ubx signal intensities in nuclei outside of transcription sites are consistent.

Figure 3—figure supplement 1.

Fifty random locations are selected from nuclei with svb transcription sites from three wild-type (w1118) embryos raised at 25°C used for analysis in Figure 3 and the Ubx intensity around these sites are measured using the same circle ROI. The Ubx intensity distributions from these sites in each individual embryo (cyan, orange and green) and with all three embryos pooled together (gray) are similar. The average Ubx intensity from actual svb transcription sites (purple, 61 sites) in the same three embryos is higher compared to randomly selected sites, consistent with our previous findings (Tsai et al., 2017). Two-tailed t-test was applied for each individual comparison. In box plots, center line is mean, upper and lower limits are standard deviation and whiskers show 95% confidence intervals.
Open in a new tab

Df(X)svb108 deficiencies are rescued upon insertion of the full svb cis-regulatory region in a different chromosome

Having observed in the past that transcriptional microenvironments can be shared between related svb enhancers on different chromosomes (Tsai et al., 2017), we wondered whether this phenomenon could enhance transcriptional output and thus buffer against adverse environmental conditions. Therefore, we tested the capacity of a DNA sequence containing the full svb cis-regulatory region to rescue the described molecular and developmental defects of the Df(X)svb108 allele (Figure 4A). For this purpose, we used a transgenic fly line, where a bacterial artificial chromosome (BAC) carrying the complete cis-regulatory region of svb (Preger-Ben Noon et al., 2018) was integrated into chromosome 2. To exclude svb mRNA from effecting the rescue, this svbBAC construct drives a dsRed reporter gene instead of another copy of svb. We confirmed that DsRed protein expression driven by this regulatory sequence recapitulated the svb expression patterns (Figure 4B) in D. melanogaster embryos and was responsive to Ubx—the lack of Ubx led to a decrease of expression in the A1 segment (Figure 4C).

Figure 4. Introduction of the cis-regulatory region of svb on another chromosome rescues the microenvironment deficiencies of the Df(X)svb108 deletion mutant.

(A) The svbBAC construct encompasses the entire cis-regulatory region of svb, driving the expression of dsRed. (B and C) The svbBAC driving the expression of DsRed inserted into the second chromosome drives a similar expression pattern as the wild-type svb locus and responds similarly to Ubx. (D) In nuclei having both svb (for both the wild-type and the Df(X)svb108 allele) and svbBAC-dsRed transcription sites at 32°C, the distances between them areon average closer than between that of svb and a reporter construct of an unrelated gene, diachete (diBAC-gfp, at 25°C), inserted into the same location as svbBAC on the second chromosome. The pairs of distance quantified were: 25 between diBAC-gfp and wild-type svb, 25 between svbBAC-dsRed and wild-type svb and 26 between svbBAC-dsRed and Df(X)svb108. (E) The svb FISH intensity (representing transcriptional output) in Df(X)svb108 x svbBAC-dsRed embryos at 32°C recovered to wild-type levels when the svb transcription site is close to a dsRed transcription site (colocalized). FISH intensity of svb in the same embryos in nuclei without a dsRed transcription site or where svb and dsRed transcription sites were not near each other (non-colocalized) did not recover. The center image panel was stained for dsRed but the particular nucleus does not show dsRed signal. The number of transcription sites quantified was: 49 for wild-type, 53 for Df(X)svb108 not near a dsRed transcription site (including cells where there were no dsRed transcription sites) and 12 for Df(X)svb108 near a dsRed transcription site. (F) At 32°C, Ubx concentration around svb transcription sites recovered to wild-type levels in nuclei containing colocalized svb and dsRed transcription sites (colocalized) in Df(X)svb108 x svbBAC-dsRed embryos. Ubx levels around transcription sites of svb in the same embryos in nuclei without a dsRed transcription site or where svb and dsRed transcription sites were not near each other (non-colocalized) did not recover. The number of transcription sites quantified was: 38 for wild-type, 60 for Df(X)svb108 not near a dsRed transcription site (including cells where there were no dsRed transcription sites) and 12 for Df(X)svb108 near a dsRed transcription site. Two-tailed t-test was applied for each individual comparison. In box plots, center line is mean, upper and lower limits are standard deviation and whiskers show 95% confidence intervals. (G) A surface plot (the height representing Ubx intensity) showing two svb transcription sites in a nucleus, with the one on the right overlapping with a svbBAC-dsRed transcription site and showing higher Ubx concentration. (H) At 32°C, in nuclei having both svb (from the Df(X)svb108 allele) and svbBAC-dsRed transcription sites, the distances between them are plotted against svb FISH intensity and Ubx intensity. There are two clusters separated by a threshold of 360 nm in distance. Co-localized pairs below this distance threshold present higher intensities for both Ubx intensity and nascent svb transcription. The pairs quantified were 15 for colocalized and 14 for non-colocalized between svbBAC-dsRed and Df(X)svb108. We analyzed four embryos for diBAC-gfp x w1118, four embryos for svbBAC-dsRed x Df(X)svb108 and for five embryos for svbBAC-dsRed x w1118.

Figure 4.

Figure 4—figure supplement 1. Introduction of svbBAC-dsRed to wild-type (w1118) does not change Ubx microenvironment and phenotype.

Figure 4—figure supplement 1.

(A) Ubx concentrations around svb transcription sites in the ventral region of the A1 segment in wild-type (w1118) x svbBAC-dsRed embryos developed at 32°C did not change compared to wild-type. The number of quantified svb transcription sites was: 20 for wild-type and 10 for wild-type x svbBAC-dsRed. Two-tailed t-test was applied for each individual comparison. We analyzed three embryos for each genotype. In box plots, center line is mean, upper and lower limits are standard deviation and whiskers show 95% confidence intervals. (B and C) The trichome phenotype along A1 and A2 did not change with the introduction of svbBAC-dsRed to wild-type (w1118).
Open in a new tab

To test the rescue, embryos or larvae with a svbBAC-dsRed crossed into them were incubated at 32°C. We observed that many active dsRed transcription sites were close to svb transcription sites in nuclei expressing both svb and dsRed in embryos from crosses between svbBAC-dsRed and wild-type (w1118) flies (Figure 4D). Similar observations were previously seen for the endogenous svb locus with itself and with the other two ventral enhancers (E3 and 7) (Tsai et al., 2017). This observation was also true for embryos from crosses between svbBAC-dsRed and Df(X)svb108 flies, suggesting that the co-localization of related regulatory regions could occur under stressed conditions. This effect was not observed for the unrelated regulatory region of diachete driving the expression of gfp, which was inserted on a BAC in the same chromosomal location as svbBAC.

We observed that the introduction of the svb regulatory region was able to rescue both molecular and functional defects observed from the loss of the region containing DG3. Both svb transcriptional output (Figure 4E) and local Ubx concentration around svb transcription sites (Figure 4F and G) were restored to wild-type levels, but only when they co-localized with an active svbBAC-dsRed transcription site in the same nucleus (within 360 nm from each other, see ‘Analysis of distances between transcription spots’ in materials and methods for the definition of colocalization). The pairs of transcription sites under and above this threshold clustered in two groups distinguishable also by Ubx and svb transcription levels (Figure 4H). Wild-type (w1118) x svbBAC-dsRed embryos and larvae were similar to wild-type in both trichome number and Ubx levels around svb transcription sites (Figure 4—figure supplement 1).

The phenotype, ventral trichome formation on the A1 segment (Figure 5A–C), which is reduced with the DG3-deletion allele, was partially rescued by the introduction of svbBAC (Figure 5D). The loss of the outer edge trichomes in A1 (in the black brackets in Figure 5A–C, where only DG3 provides coverage Figure 2F) with the DG3-deletion allele was not rescued with svbBAC. Additionally, introducing only the DG3 enhancer as opposed to svbBAC did not rescue trichome formation under heat-stress (Figure 5D).

Figure 5. The complete cis-regulatory region of svb rescues trichome number.

Figure 5.

Open in a new tab

(A–C) Cuticle preparations from larvae developed at 32°C with wild-type svb, Df(X)svb108 and Df(X)svb108 x svbBAC-dsRed. The bracket at the edge of the A1 stripe marks a region where trichome growth is exclusively covered by DG3, which disappeared with the deletion of DG3 and did not recover with the introduction of svbBAC-dsRed. (D) The trichome number in larvae developed at 32°C with Df(X)svb108 partially recovered to wild-type levels with the introduction of svbBAC-dsRed. The number of larvae counted was: 12 for wild-type svb, 28 for Df(X)svb108, 14 for Df(X)svb108 x svbBAC-dsRed and 13 for Df(X)svb108 x DG3-lacZ. Two-tailed t-test was applied for each individual comparison. In box plots, center line is mean, upper and lower limits are standard deviation and whiskers show 95% confidence intervals.

Discussion

Transcriptional regulation is a complex and dynamic process which requires coordinated interactions between transcription factors and chromatin. Given the transient nature of these interactions, using multiple binding sites to ensure efficient and consistent transcriptional regulation under different environmental conditions appears to be a preferred strategy among many developmental enhancers (Frankel, 2012; Perry et al., 2010). Genes such as shavenbaby add another layer of redundancy on top of this through long cis-regulatory regions containing multiple enhancers whose expression patterns overlap. Previous works have shown that this redundancy ensures proper phenotype development when systems are subjected to stress (Crocker et al., 2015; Frankel et al., 2010; Osterwalder et al., 2018). However, the mechanism underlying this phenotypic robustness was not clear.

In this work, we took advantage of the high-resolution imaging and analysis techniques we had developed to observe transcriptional microenvironments around transcription sites (Tsai et al., 2017) and investigated how the DG3 enhancer contributes to the phenotypic robustness of the svb locus at the molecular level. Deletion of the DG3 enhancer from svb did not lead to clear defects in ventral trichome formation unless the embryos were subjected to heat-induced stress, as shown here and previously with the deletion of ‘shadow enhancers’ (Hong et al., 2008) for lateral svb expression (Frankel et al., 2010). Nevertheless, we observed that the DG3-deletion allele showed reduced transcriptional output even at normal temperature. Wild-type embryos did not show phenotypic defects under either normal or stressed conditions, but heat-induced stress led to slightly lower transcriptional output from the wild-type svb allele (Figure 6A and B). However, the mutant svb locus, starting with lower transcriptional output even under ideal conditions (Figure 6C), drops below a threshold and the system fails (Figure 6D). We observed that Ubx concentration and svb transcriptional output in the A1 segment initially have a weak positive correlation that quickly dissipated at higher Ubx concentrations and svb outputs. While this data must be interpreted with some caution, as multiple stages of the transcription cycle are included (initiation, elongation or termination), Ubx concentration increases after a certain point were not clearly coupled to increases in svb transcriptional output. Additional regulatory mechanisms could be at play beyond transcription factor retention that determines the final transcriptional output of the locus. Determining the complete response function would be complicated, due to reasons such as the enhancers each having clusters of factor binding sites and the overlap of expression patterns from related enhancers. As svb also accepts inputs from many additional transcription factors (Stern and Orgogozo, 2008), especially in the other body segments, the total response of the system would also depend on many factors not observed in this study.

Figure 6. Summary of model including a sigmoidal relationship between svb transcriptional output and the number of cells fated to become trichomes.

Figure 6.

Open in a new tab

(A–F) Schematic representations of the various genotypes, temperatures and rescue conditions as tested in this work. Bottom panel: Schematic of the proposed sigmoidal relationship between svb transcriptional output and the number of cells fated to become trichomes. Note that A and C are at 25°C and the rest are at 32°C. The different conditions we have studied are represented by letters A-F along the sigmoidal curve. The part of the curve corresponding to wild-type phenotype is shaded in green.

We previously observed that transcription sites of reporter genes driven by minimal svb enhancers tended to colocalize with the endogenous svb locus when it is transcriptionally active (Tsai et al., 2017). This is true also for entire cis-regulatory regions, as we observed that the svb locus did the same with svbBAC, which implies that they potentially share a common microenvironment. Homologous regions were shown to pair over long distances, between homologous chromosomal arms (Lim et al., 2018), translocated domains and even different chromosomes (Gemkow et al., 1998; Johnston and Desplan, 2014; Peifer and Bender, 1986). Our observations could be related to them and share similar mechanisms. However, our constructs may not contain sites for structural elements such as insulators, which have been described as important supporters of trans-interactions (Postika et al., 2018). On the other hand, our observations are in line with transcription-dependent associations of interchromosomal interactions (Branco and Pombo, 2006; Joyce et al., 2016; Lomvardas et al., 2006; Maass et al., 2018; Monahan et al., 2019). It is possible that such long range interactions are driven, or reinforced, through shared microenvironments.

We were able to partially rescue the DG3-deletion svb allele with svbBAC, which contains the cis-regulatory region of svb but not the svb gene itself. High-resolution imaging showed that colocalizing with a svbBAC increases the local Ubx concentration and transcriptional output of the DG3-deletion allele (Figure 6E). This supports a mechanism where transcriptional microenvironments sequestered around large and related cis-regulatory regions in physical proximity can work in trans to increase transcriptional output of other genes, even on different chromosomes. The introduction of a shorter reporter construct containing the DG3 enhancer alone did not rescue trichome expression perhaps because it is not able to effectively pair with the svb locus. It is possible that structural elements, such as insulator proteins (Lim et al., 2018) or other topologically associated elements (Furlong and Levine, 2018) could overcome this by increasing pairing efficiency. This is consistent with recent findings for long-range interactions that are dependent on specific topologically associating domains (TADs), where pairing is necessary but not sufficient for transvection (Viets et al., 2018). Interestingly, embryos with the DG3 deletion allele and svbBAC could not produce trichomes at the dorsal edges of the ventral trichome patches, where DG3 provides exclusive coverage. As the rescue only occurred on regions where other ventral svb enhancers provided overlapping coverage, it likely is the result of compensation from the additional ventral enhancers (E3 and 7) at the svb locus instead of the DG3 enhancer on the svbBAC driving svb expression in trans. Interactions in trans may serve an auxiliary role through influencing the properties of the local transcriptional environments.

Summarizing our observations, we hypothesize that the relationship between svb transcriptional output and the number of cells fated to become trichomes is sigmoidal (Figure 6, bottom panel). Below a certain threshold, the number of trichomes would drop with decreasing svb mRNA; however, any additional transcriptional output above this threshold (the green box in the figure) would not lead to significant changes in trichome production and would appear to be wild-type in phenotype. The cis-regulatory region of wild-type svb under ideal conditions likely already saturates the system, as evidenced by the lack of change in Ubx concentration and trichome numbers even with the addition of svbBAC (Figure 6F). Although operating under saturation renders this system relatively insensitive to changes in svb transcription, the risk of developing defective phenotypes when conditions are no longer ideal likely selected for this strategy to buffer against stresses. Overall, the system remains phenotypically robust and develops the same number of trichomes despite fluctuations in transcriptional outputs. In the future, it would be important to understand mechanistically how the phenotype can tolerate significant drops in transcriptional output before defects appear. It requires the direct observation of intermediate steps between svb transcription and phenotype production to understand how this buffering is achieved. Furthermore, our current technique only allows us to probe environments around active transcription sites without knowing which phase of transcription (initiation, elongation or termination) it is in and how the gene loci positioned themselves into these locations. Future works to visualize and track genes, regardless of their transcriptional state, in fixed and living embryos would answer key questions on how they find, form and interact with transcriptional microenvironments.

We previously proposed that transcriptional microenvironments form across multiple enhancers through scaffolding interactions to ensure efficient transcription from developmental enhancers. By investigating the mechanisms of how correct transcriptional regulation is maintained under stress using a DG3-deletion allele of svb, we have shown that transcriptional microenvironments could span multiple enhancers that share similar transcription factor binding sites. These microenvironments of transcription factors could form the protein core of transcription factor ‘hubs’ that have been proposed to form through phase-separation mediated through protein-protein interactions between disordered domains (Cisse et al., 2013; Furlong and Levine, 2018; Ghavi-Helm et al., 2014; Mir et al., 2017). Thus, they add another layer of redundancy on top of using multiple enhancers with overlapping expression patterns in a cis-regulatory region to ensure a sufficient margin to buffer against the negative effects of environmental stresses (Frankel et al., 2010; Perry et al., 2010). This extra margin of safety preserves phenotypical development even when environmental conditions are not ideal. Integrating multiple noisy and low-affinity elements into a coherent and synergistic network would also reduce the variance stemming from the transient and stochastic transcription factor binding dynamics observed in eukaryotic cells (Cisse et al., 2013; Ghavi-Helm et al., 2014; Mir et al., 2017; Tsai et al., 2017). In sum, specialized transcriptional microenvironments could be a critical element to ensure that gene expression occurs specifically and consistently in every embryo. Given that shadow enhancers are widespread features of gene regulatory networks (Cannavò et al., 2016; Osterwalder et al., 2018), it is likely that high local concentrations of transcription factors are a widespread feature that provides an effective regulatory buffer to prevent deleterious phenotypic consequences to genetic and environmental perturbations.

Materials and methods

Fly strains

All fly strains used have been previously described: DG3-lacZ (Tsai et al., 2017); ubx1 (Crocker et al., 2015); HS::ubx-1: (Crocker et al., 2015); Df(X)svb108 (Frankel et al., 2010); svbBAC-dsRed (Preger-Ben Noon et al., 2018); diBAC-gfp is CH322-35A16 EGFP tagged in VK37, covering D (Venken et al., 2009). Unless otherwise noted, they are generated from w1118 stock, which is referred to as wild-type.

Preparing Drosophila embryos for staining and cuticle preps

D. melanogaster strains were maintained under standard laboratory conditions, reared at 25 °C, unless otherwise specified. For heat-shock experiments, these conditions were followed: for staining with fluorescent antibodies, flies were allowed to lay eggs on apple-juice agar plates for 5 hr at 25 °C and then kept in an incubator at 32 °C for 7 hr before fixation; for cuticle preps, dechorionated embryos were kept at 32 °C until they emerged as larvae. Df(X)svb108 embryos/larvae with svbBAC-dsRed are readily discernable by the loss of svb and trichomes in the T1 segment (see Figure 2D & E).

Cuticle preparation and trichome counting

Larvae collected for cuticle preparations were mounted according to a published protocol (Stern and Sucena, 2011). A phase-contrast microscope was used to image the slides. In the case of crosses involving the Df(X)svb108 allele, only larvae lacking trichomes in the T1 segment were imaged (Figure 2E). This is a homozygous marker for the deletion locus as larvae carrying any wild-type svb allele will produce trichomes in T1 (Figure 2D). Ventral trichomes in larval A1 or A2 segments were counted in Fiji/ImageJ by find using the find maximum function (Schindelin et al., 2012; Schneider et al., 2012).

Immuno-fluorescence staining of transcription factors and in situ hybridization to mRNA

Standard protocols were used for embryo fixation and staining (Crocker et al., 2015; Tsai et al., 2017). Secondary antibodies labeled with Alexa Fluor dyes (1:500, Invitrogen) were used to detect primary antibodies. In situ hybridizations were performed using DIG, FITC or biotin labeled, antisense RNA-probes against a reporter construct RNA (lacZ, dsRed, gfp) or the first intron and second exon (16 kb) of svb. See Supplementary file 1 for primer sequences. DIG-labeled RNA products were detected with a DIG antibody: Thermofisher, 700772 (1:100 dilution), biotin-labeled RNA products with a biotin antibody: Thermofisher, PA1-26792 (1:100) and FITC-labeled RNA products with a FITC antibody: Thermofisher, A889 (1:100). Ubx protein was detected using Developmental Studies Hybridoma Bank, FP3.38-C antibody at 1:20 dilution, DsRed protein using MBL anti-RFP PM005 antibody at 1:100, LacZ protein using Promega anti-ß-Gal antibody at 1:250 and GFP protein using Aves Labs chicken anti-GFP at 1:300.

Imaging fixed embryos

Mounting of fixed Drosophila embryos was done in ProLong Gold + DAPI mounting media (Molecular Probes, Eugene, OR). Fixed embryos were imaged on a Zeiss LSM 880 confocal microscope with FastAiryscan (Carl Zeiss Microscopy, Jena, Germany). Excitation lasers with wavelengths of 405, 488, 561 and 633 nm were used as appropriate for the specific fluorescent dyes. For imaging in embryos carrying Df(X)svb108, only embryos without svb mRNA expression in the T1 segment were imaged, following the same reason described in the section on ‘Cuticle preparations and trichome counting’. Unless otherwise stated, all images were processed with Fiji/ImageJ (Schindelin et al., 2012; Schneider et al., 2012) and Matlab (MathWorks, Natick, MA, USA).

Analysis of microenvironment and svb transcription intensity

Inside nuclei with svb transcription sites, the center of the transcription site was identified using the find maximum function of Fiji/ImageJ. A circle with a diameter of 4 pixels (170 nm, roughly the lateral resolution limit of AiryScan in 3D mode) region of interest (ROI) centered on the transcription site is then created. The integrated fluorescent intensity inside the ROI from the Ubx IF channel and the RNA FISH channel are then reported as the local Ubx concentration and the transcriptional output, respectively. The intensity presented in the figures is the per-pixel average intensity with the maximum readout of the sensor normalized to 255.

Analysis of distances between transcription spots

Inside nuclei with svb and dsRed/gfp transcription sites, the centers of the transcription site were identified using the find maximum function of Fiji/ImageJ. The distance between the transcription sites were then computed using the coordinates of the transcription sites. Two sites are considered colocalized when they are within 360 nm of each other.

Ubx ChIP profile

The ChIP profile for Ubx around the svb cis-regulatory region is from Choo et al. (2011), using whole Drosophila melanogaster embryos between stages 10 and 12.

Acknowledgements

The fly line containing diBAC-gfp (CH322-35A16 EGFP) was a gift from Schulze, Karen Lynn and Bellen, Hugo J. The Df(X)svb108 flies were a gift from Stern DL and the svbBAC flies were a gift from Preger Ben-Noon E and Frankel N (Preger-Ben Noon et al., 2018). We thank Rafael Galupa, Nicolas Frankel and Ella Preger Ben-Noon for suggestions and discussions. We thank the entire Crocker lab for discussion and feedback. We would also like to thank the reviewers for their constructive input. Albert Tsai is a Damon Runyon Fellow of the Damon Runyon Cancer Research Foundation (DRG 2220–15). Albert Tsai, Mariana R P Alves and Justin Crocker are supported by the European Molecular Biological Laboratory (EMBL).

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Albert Tsai, Email: albert.tsai@embl.de.

Justin Crocker, Email: justin.crocker@embl.de.

David N Arnosti, Michigan State University, United States.

Jessica K Tyler, Weill Cornell Medicine, United States.

Funding Information

This paper was supported by the following grants:

  • Damon Runyon Cancer Research Foundation DRG 2220-15 to Albert Tsai.

  • European Molecular Biology Organization to Albert Tsai, Mariana R P Alves, Justin Crocker.

Additional information

Competing interests

No competing interests declared.

Author contributions

Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing.

Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing.

Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing.

Additional files

Supplementary file 1. Primers for RNA-probe generation Sequences of primers for amplification of DNA to be used for generation of antisense RNA-probes.

The targets - reporter construct RNA (lacZ, dsRed, gfp) and first intron and second exon (16 kb) of svb - are indicated in the left column. Sequences are indicated for forward or reverse primers of each pair. Reverse primers include a T7 sequence for transcription with T7 RNA polymerase.

elife-45325-supp1.xlsx (11.1KB, xlsx)
DOI: 10.7554/eLife.45325.012
Transparent reporting form
elife-45325-transrepform.pdf (335.6KB, pdf)
DOI: 10.7554/eLife.45325.013

Data availability

The original images (cuticle preparations and embryo images, organized into zip files) are available for download and are indexed at: https://www.embl.de/download/crocker/svb_enhancer_colocalization/index.html. Please note that the raw AiryScan images must be processed though the Zen software from Zeiss before they can be opened/analyzed using standard image processing softwares. These files are large, totaling up to approximately 180 GB in size. We can also send these files directly if a means of transfer (hard drives, etc.) is provided.

References

  1. Antosova B, Smolikova J, Klimova L, Lachova J, Bendova M, Kozmikova I, Machon O, Kozmik Z. The gene regulatory network of Lens induction is wired through Meis-Dependent shadow enhancers of Pax6. PLOS Genetics. 2016;12:e1006441. doi: 10.1371/journal.pgen.1006441. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Berger MF, Badis G, Gehrke AR, Talukder S, Philippakis AA, Peña-Castillo L, Alleyne TM, Mnaimneh S, Botvinnik OB, Chan ET, Khalid F, Zhang W, Newburger D, Jaeger SA, Morris QD, Bulyk ML, Hughes TR. Variation in homeodomain DNA binding revealed by high-resolution analysis of sequence preferences. Cell. 2008;133:1266–1276. doi: 10.1016/j.cell.2008.05.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Boija A, Klein IA, Sabari BR, Dall'Agnese A, Coffey EL, Zamudio AV, Li CH, Shrinivas K, Manteiga JC, Hannett NM, Abraham BJ, Afeyan LK, Guo YE, Rimel JK, Fant CB, Schuijers J, Lee TI, Taatjes DJ, Young RA. Transcription factors activate genes through the Phase-Separation capacity of their activation domains. Cell. 2018;175:1842–1855. doi: 10.1016/j.cell.2018.10.042. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Branco MR, Pombo A. Intermingling of chromosome territories in interphase suggests role in Translocations and transcription-dependent associations. PLOS Biology. 2006;4:e138. doi: 10.1371/journal.pbio.0040138. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Cannavò E, Khoueiry P, Garfield DA, Geeleher P, Zichner T, Gustafson EH, Ciglar L, Korbel JO, Furlong EE. Shadow enhancers are pervasive features of developmental regulatory networks. Current Biology. 2016;26:38–51. doi: 10.1016/j.cub.2015.11.034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Chanut-Delalande H, Fernandes I, Roch F, Payre F, Plaza S. Shavenbaby couples patterning to epidermal cell shape control. PLOS Biology. 2006;4:e290. doi: 10.1371/journal.pbio.0040290. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Chen J, Zhang Z, Li L, Chen BC, Revyakin A, Hajj B, Legant W, Dahan M, Lionnet T, Betzig E, Tjian R, Liu Z. Single-molecule dynamics of enhanceosome assembly in embryonic stem cells. Cell. 2014;156:1274–1285. doi: 10.1016/j.cell.2014.01.062. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Choo SW, White R, Russell S. Genome-wide analysis of the binding of the hox protein ultrabithorax and the hox cofactor homothorax in Drosophila. PLOS ONE. 2011;6:e14778. doi: 10.1371/journal.pone.0014778. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cisse II, Izeddin I, Causse SZ, Boudarene L, Senecal A, Muresan L, Dugast-Darzacq C, Hajj B, Dahan M, Darzacq X. Real-time dynamics of RNA polymerase II clustering in live human cells. Science. 2013;341:664–667. doi: 10.1126/science.1239053. [DOI] [PubMed] [Google Scholar]
  10. Crocker J, Potter N, Erives A. Dynamic evolution of precise regulatory encodings creates the clustered site signature of enhancers. Nature Communications. 2010;1:99. doi: 10.1038/ncomms1102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Crocker J, Abe N, Rinaldi L, McGregor AP, Frankel N, Wang S, Alsawadi A, Valenti P, Plaza S, Payre F, Mann RS, Stern DL. Low affinity binding site clusters confer hox specificity and regulatory robustness. Cell. 2015;160:191–203. doi: 10.1016/j.cell.2014.11.041. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Crocker J, Noon EP, Stern DL. The soft touch: low-affinity transcription factor binding sites in development and evolution. Current Topics in Developmental Biology. 2016;117:455–469. doi: 10.1016/bs.ctdb.2015.11.018. [DOI] [PubMed] [Google Scholar]
  13. Delon I, Chanut-Delalande H, Payre F. The ovo/Shavenbaby transcription factor specifies actin remodelling during epidermal differentiation in Drosophila. Mechanisms of Development. 2003;120:747–758. doi: 10.1016/S0925-4773(03)00081-9. [DOI] [PubMed] [Google Scholar]
  14. Farley EK, Olson KM, Zhang W, Brandt AJ, Rokhsar DS, Levine MS. Suboptimization of developmental enhancers. Science. 2015;350:325–328. doi: 10.1126/science.aac6948. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Farley EK, Olson KM, Zhang W, Rokhsar DS, Levine MS. Syntax compensates for poor binding sites to encode tissue specificity of developmental enhancers. PNAS. 2016;113:6508–6513. doi: 10.1073/pnas.1605085113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Frankel N, Davis GK, Vargas D, Wang S, Payre F, Stern DL. Phenotypic robustness conferred by apparently redundant transcriptional enhancers. Nature. 2010;466:490–493. doi: 10.1038/nature09158. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Frankel N. Multiple layers of complexity in cis-regulatory regions of developmental genes. Developmental Dynamics : An Official Publication of the American Association of Anatomists. 2012;241:1857–1866. doi: 10.1002/dvdy.23871. [DOI] [PubMed] [Google Scholar]
  18. Furlong EEM, Levine M. Developmental enhancers and chromosome topology. Science. 2018;361:1341–1345. doi: 10.1126/science.aau0320. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Gaudet J, Mango SE. Regulation of organogenesis by the Caenorhabditis elegans FoxA protein PHA-4. Science. 2002;295:821–825. doi: 10.1126/science.1065175. [DOI] [PubMed] [Google Scholar]
  20. Gemkow MJ, Verveer PJ, Arndt-Jovin DJ. Homologous association of the Bithorax-Complex during embryogenesis: consequences for transvection in Drosophila Melanogaster. Development. 1998;125:4541–4552. doi: 10.1242/dev.125.22.4541. [DOI] [PubMed] [Google Scholar]
  21. Ghavi-Helm Y, Klein FA, Pakozdi T, Ciglar L, Noordermeer D, Huber W, Furlong EE. Enhancer loops appear stable during development and are associated with paused polymerase. Nature. 2014;512:96–100. doi: 10.1038/nature13417. [DOI] [PubMed] [Google Scholar]
  22. Hong JW, Hendrix DA, Levine MS. Shadow enhancers as a source of evolutionary novelty. Science. 2008;321:1314. doi: 10.1126/science.1160631. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Izeddin I, Récamier V, Bosanac L, Cissé II, Boudarene L, Dugast-Darzacq C, Proux F, Bénichou O, Voituriez R, Bensaude O, Dahan M, Darzacq X. Single-molecule tracking in live cells reveals distinct target-search strategies of transcription factors in the nucleus. eLife. 2014;3:e02230. doi: 10.7554/eLife.02230. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Johnston RJ, Desplan C. Interchromosomal communication coordinates intrinsically stochastic expression between alleles. Science. 2014;343:661–665. doi: 10.1126/science.1243039. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Joyce EF, Erceg J, Wu C-ting. Pairing and anti-pairing: a balancing act in the diploid genome. Current Opinion in Genetics & Development. 2016;37:119–128. doi: 10.1016/j.gde.2016.03.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Lebrecht D, Foehr M, Smith E, Lopes FJ, Vanario-Alonso CE, Reinitz J, Burz DS, Hanes SD. Bicoid cooperative DNA binding is critical for embryonic patterning in Drosophila. PNAS. 2005;102:13176–13181. doi: 10.1073/pnas.0506462102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Lim B, Heist T, Levine M, Fukaya T. Visualization of transvection in living Drosophila embryos. Molecular Cell. 2018;70:287–296. doi: 10.1016/j.molcel.2018.02.029. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Liu Z, Legant WR, Chen BC, Li L, Grimm JB, Lavis LD, Betzig E, Tjian R. 3d imaging of Sox2 enhancer clusters in embryonic stem cells. eLife. 2014;3:e04236. doi: 10.7554/eLife.04236. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Lomvardas S, Barnea G, Pisapia DJ, Mendelsohn M, Kirkland J, Axel R. Interchromosomal interactions and olfactory receptor choice. Cell. 2006;126:403–413. doi: 10.1016/j.cell.2006.06.035. [DOI] [PubMed] [Google Scholar]
  30. Long HK, Prescott SL, Wysocka J. Ever-Changing landscapes: transcriptional enhancers in development and evolution. Cell. 2016;167:1170–1187. doi: 10.1016/j.cell.2016.09.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Lorberbaum DS, Ramos AI, Peterson KA, Carpenter BS, Parker DS, De S, Hillers LE, Blake VM, Nishi Y, McFarlane MR, Chiang AC, Kassis JA, Allen BL, McMahon AP, Barolo S. An ancient yet flexible cis-regulatory architecture allows localized Hedgehog tuning by patched/Ptch1. eLife. 2016;5:e13550. doi: 10.7554/eLife.13550. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Maass PG, Barutcu AR, Weiner CL, Rinn JL. Inter-chromosomal contact properties in Live-Cell imaging and in Hi-C. Molecular Cell. 2018;69:1039–1045. doi: 10.1016/j.molcel.2018.02.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Mallo M, Alonso CR. The regulation of hox gene expression during animal development. Development. 2013;140:3951–3963. doi: 10.1242/dev.068346. [DOI] [PubMed] [Google Scholar]
  34. McGinnis W, Krumlauf R. Homeobox genes and axial patterning. Cell. 1992;68:283–302. doi: 10.1016/0092-8674(92)90471-N. [DOI] [PubMed] [Google Scholar]
  35. Mir M, Reimer A, Haines JE, Li XY, Stadler M, Garcia H, Eisen MB, Darzacq X. Dense bicoid hubs accentuate binding along the morphogen gradient. Genes & Development. 2017;31:1784–1794. doi: 10.1101/gad.305078.117. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Mir M, Stadler MR, Ortiz SA, Hannon CE, Harrison MM, Darzacq X, Eisen MB. Dynamic multifactor hubs interact transiently with sites of active transcription in Drosophila embryos. eLife. 2018;7:e40497. doi: 10.7554/eLife.40497. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Monahan K, Horta A, Lomvardas S. LHX2- and LDB1-mediated trans interactions regulate olfactory receptor choice. Nature. 2019;565:448–453. doi: 10.1038/s41586-018-0845-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Normanno D, Boudarène L, Dugast-Darzacq C, Chen J, Richter C, Proux F, Bénichou O, Voituriez R, Darzacq X, Dahan M. Probing the target search of DNA-binding proteins in mammalian cells using TetR as model searcher. Nature Communications. 2015;6:e7357. doi: 10.1038/ncomms8357. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Noyes MB, Christensen RG, Wakabayashi A, Stormo GD, Brodsky MH, Wolfe SA. Analysis of homeodomain specificities allows the family-wide prediction of preferred recognition sites. Cell. 2008;133:1277–1289. doi: 10.1016/j.cell.2008.05.023. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Osterwalder M, Barozzi I, Tissières V, Fukuda-Yuzawa Y, Mannion BJ, Afzal SY, Lee EA, Zhu Y, Plajzer-Frick I, Pickle CS, Kato M, Garvin TH, Pham QT, Harrington AN, Akiyama JA, Afzal V, Lopez-Rios J, Dickel DE, Visel A, Pennacchio LA. Enhancer redundancy provides phenotypic robustness in mammalian development. Nature. 2018;554:239–243. doi: 10.1038/nature25461. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Payre F, Vincent A, Carreno S. Ovo/svb integrates wingless and DER pathways to control epidermis differentiation. Nature. 1999;400:271–275. doi: 10.1038/22330. [DOI] [PubMed] [Google Scholar]
  42. Peifer M, Bender W. The anterobithorax and bithorax mutations of the bithorax complex. The EMBO Journal. 1986;5:2293–2303. doi: 10.1002/j.1460-2075.1986.tb04497.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Perry MW, Boettiger AN, Bothma JP, Levine M. Shadow enhancers foster robustness of Drosophila gastrulation. Current Biology. 2010;20:1562–1567. doi: 10.1016/j.cub.2010.07.043. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Postika N, Metzler M, Affolter M, Müller M, Schedl P, Georgiev P, Kyrchanova O. Boundaries mediate long-distance interactions between enhancers and promoters in the Drosophila bithorax complex. PLOS Genetics. 2018;14:e1007702. doi: 10.1371/journal.pgen.1007702. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Preger-Ben Noon E, Sabarís G, Ortiz DM, Sager J, Liebowitz A, Stern DL, Frankel N. Comprehensive analysis of a cis-Regulatory region reveals pleiotropy in enhancer function. Cell Reports. 2018;22:3021–3031. doi: 10.1016/j.celrep.2018.02.073. [DOI] [PubMed] [Google Scholar]
  46. Reiter F, Wienerroither S, Stark A. Combinatorial function of transcription factors and cofactors. Current Opinion in Genetics & Development. 2017;43:73–81. doi: 10.1016/j.gde.2016.12.007. [DOI] [PubMed] [Google Scholar]
  47. Rister J, Razzaq A, Boodram P, Desai N, Tsanis C, Chen H, Jukam D, Desplan C. Single-base pair differences in a shared motif determine differential rhodopsin expression. Science. 2015;350:1258–1261. doi: 10.1126/science.aab3417. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Rowan S, Siggers T, Lachke SA, Yue Y, Bulyk ML, Maas RL. Precise temporal control of the eye regulatory gene Pax6 via enhancer-binding site affinity. Genes & Development. 2010;24:980–985. doi: 10.1101/gad.1890410. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, Preibisch S, Rueden C, Saalfeld S, Schmid B, Tinevez JY, White DJ, Hartenstein V, Eliceiri K, Tomancak P, Cardona A. Fiji: an open-source platform for biological-image analysis. Nature Methods. 2012;9:676–682. doi: 10.1038/nmeth.2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Schneider CA, Rasband WS, Eliceiri KW. NIH image to ImageJ: 25 years of image analysis. Nature Methods. 2012;9:671–675. doi: 10.1038/nmeth.2089. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Spitz F, Furlong EE. Transcription factors: from enhancer binding to developmental control. Nature Reviews Genetics. 2012;13:613–626. doi: 10.1038/nrg3207. [DOI] [PubMed] [Google Scholar]
  52. Stern DL, Orgogozo V. The loci of evolution: how predictable is genetic evolution? Evolution. 2008;62:2155–2177. doi: 10.1111/j.1558-5646.2008.00450.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Stern DL, Sucena E. Preparation of cuticles from unhatched first-instar Drosophila larvae. Cold Spring Harbor Protocols. 2011;2011:pdb.prot065532. doi: 10.1101/pdb.prot065532. [DOI] [PubMed] [Google Scholar]
  54. Tanay A. Extensive low-affinity transcriptional interactions in the yeast genome. Genome Research. 2006;16:962–972. doi: 10.1101/gr.5113606. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Tsai A, Muthusamy AK, Alves MR, Lavis LD, Singer RH, Stern DL, Crocker J. Nuclear microenvironments modulate transcription from low-affinity enhancers. eLife. 2017;6:e28975. doi: 10.7554/eLife.28975. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Venken KJ, Carlson JW, Schulze KL, Pan H, He Y, Spokony R, Wan KH, Koriabine M, de Jong PJ, White KP, Bellen HJ, Hoskins RA. Versatile P[acman] BAC libraries for transgenesis studies in Drosophila Melanogaster. Nature Methods. 2009;6:431–434. doi: 10.1038/nmeth.1331. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Viets K, Sauria M, Chernoff C, Anderson C, Tran S, Dove A, Goyal R, Voortman L, Gordus A, Taylor J, Johnston RJ. TADs pair homologous chromosomes to promote interchromosomal gene regulation. bioRxiv. 2018 doi: 10.1101/445627. [DOI] [PMC free article] [PubMed]

Decision letter

Editor: David N Arnosti1
Reviewed by: Angela H DePace2, Hernan Garcia3

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for sending your article entitled "Multi-enhancer transcriptional hubs confer phenotypic robustness" for peer review at eLife. Your article is being evaluated by three peer reviewers, and the evaluation is being overseen by a Reviewing Editor and Jessica Tyler as the Senior Editor.

Given the list of essential revisions, including new experiments, the editors and reviewers invite you to respond within the next two weeks with an action plan and timetable for the completion of the additional work. We plan to share your responses with the reviewers and then issue a binding recommendation.

Essential points to address are the following:

1) The connection between enhancers, Ubx protein concentration, transcriptional output, and phenotype. Although different concentrations of Ubx are measured at the deleted vs. wild-type locus, an important point that the three reviewers agree upon is that gene expression levels and Ubx concentrations are not well correlated. There appear to be other unexplained factors that are influencing the expression of svb that do not correlate with the TF concentration. Thus, it is unclear whether the experiments described here support or reject the hypothesis laid out in the "multi-enhancer "hubs" improve robustness by increasing transcription factor retention near transcription sites."

The authors must consider further whether the global measurements shown in Figure 3 are the most effective measure of this proposed correlation, or whether a more fine-grained analysis might support the idea that boosted Ubx levels from complexed enhancers are driving transcription. As noted below, the authors might show scatter plots of Ubx concentration vs. svb transcription within each condition. A positive correlation would support their hypothesis.

The trichome phenotype is also not directly correlated to the transcriptional outputs and Ubx concentrations, and as mentioned in the Discussion, further layers of regulation are likely in play. This idea needs to be more fully fleshed out.

A finding related to "enhancer hubs boosting local concentration" is the finding that an ectopic BAC appears not to influence Ubx concentration in a wild-type svb background, only in a deletion background. Does this finding indicate that such hub formation is saturable? And if so, does this indicate that the wild-type svb locus forms a local hub with enhancers in the immediate vicinity, and trans-complementation is not a normal feature of svb function?

2) A second point raised concerned the exact cis regulatory regions in play. A minimal DG3 enhancer drives gene expression in ventral abdominal stripes, but does not rescue Ubx concentrations from a trans setting. A larger deletion that includes DG3 as well as additional Ubx binding regions (that are not sufficient for, but may be part of, ventral DG3-related activity) impacts transcription, trichome development, and robustness. A yet larger cis-regulatory domain on a BAC rescues some aspects of gene expression and Ubx concentration. The interpretation conflates DG3 with the function of the deleted region; reviewers noted that a more careful interpretation would differentiate results from each of these different cis elements. For instance, the lack of trans-rescue by the DG3 enhancer alone may be due to the inability of a short segment to transvect effectively. The interpretation should explicitly take into account known properties of transvecting regulatory loci in Drosophila.

3) Several aspects can be addressed by better justification and explanation of methods and data presentation, including

– Why sometimes either A1 or A2 trichomes are quantitatively assessed, depending on the figure;

– The use of one-tailed (vs. two-tailed) T-tests for statistical relevance;

– Recommendations for inclusion of both 25C and 32C phenotypes and data uniformly, and;

– Clarification of exact role of Ubx in T1-T3 regulation, as the conclusions drawn about DG3 and Ubx roles are difficult to know based on the single images shown;

– The number of data points in Figures 3 and 4 are limited; is there a technical limitation to more extensive sampling of the transcriptional readouts and Ubx intensities?

4) In some cases, the actual experimental approach was unclear to the reviewers:

– there was some confusion about whether the DG3 deletion mutants were homozygous, so there would be no wild-type copy of the gene in these embryos – is that indeed the case?

– Figure 4 shows svb, but not dsRed mRNA expression; is that correct? Imaging of dsRed only used to score a svb locus as "overlapped" vs. "not overlapped"?

– Choice of pixel size for ROI.

– Signal to noise for Ubx across the nucleus, as well as variation in average Ubx from sample to sample.

Reviewer #1:

This paper addresses a timely topic, namely how multiple enhancers may coordinate to regulate gene expression. This is being actively investigated in a number of systems using both case studies and genome wide methods, and the field is fueled by new capabilities in high resolution microscopy. Thus, interest in the topic is high. However, I have substantial concerns about how this work contributes to this topic.

The central hypothesis of this work is that increased levels of Ubx, provided by micro-environments created through coalescing multiple Ubx-responsive enhancers, increases the level of transcription of shavenbaby, svb; this increased level of transcription allows svb expression to be robust to high temperature variation. However, the evidence presented in this paper does not consistently support this hypothesis.

Let's first consider the set of experiments correlating the presence of the enhancer deletion (via a deficiency), the concentration of Ubx and the level of svb transcription.

– Deleting an svb enhancer (DG3) decreases resilience to high temperature stress, resulting in smaller numbers of trichomes in A1.

– In the A1 segment, svb enhancers localize to a microenvironment where Ubx concentrations are high compared to background.

– Deleting an svb enhancer decreases the concentration of Ubx in these microenvironments moderately at high temperature, but has no effect at low temperature.

– At high temperatures, transcription of svb decreases at both WT and deficiency alleles.

– At low temperatures, transcription of svb decreases under low temperature stress for the deficiency allele.

These results do not support the hypothesis because svb transcription is not consistently correlated with Ubx levels in the microenvironment.

A second concern is the interpretation of the rescue experiment, where a BAC containing the full svb locus is integrated onto another chromosome. This partially rescues the trichome phenotype of the deficiency allele. However, the Ubx levels in microenvironments were unchanged from WT. If the extra svb regulatory region associates with the endogenous svb transcription site, this should either result in above WT levels of Ubx, or there is some additional layer of regulation. This was not discussed.

Reviewer #2:

In this paper, Tsai et al. studied the general problem of how low affinity binding sites, which promote enhancer specificity, also permit robust activation. They focus on ventral "shadow" enhancers for shavenbaby/svb.

The authors start by focusing on the DG3 enhancer, which was less studied before, and responds to Ubx in A1. Indeed, ChIP-seq data shows Ubx binding in a DG3 element that drive reporter gene expression in trichome forming territories, and this reporter responds to perturbations of Ubx.

A deletion, Df(x)svb108, removes DG3, as well as several other Ubx-bound regions, and causes trichome formation defects only upon heat stress, and primarily in the region of DG3 activity.

State-of-the-art imaging of svb transcription and Ubx protein distribution show a reduction of Ubx concentration in the vicinity of active svb loci, specifically in Df(X)svb108mutant cells under heat stress, and a reduction of transcriptional output upon heat stress, even in wild-type embryos.

Remarkably, exogenous svb cis-regulatory DNA could rescue the formation of a Ubx-rich "hub" containing the endogenous svb locus, and svb transcription, in Df(X)svb108 mutants, but only when the two loci colocalized.

Taken together, the data presented is consistent with the notion that multiple enhancers form a Ubx-rich microenvironment necessary for robust svb expression and trichome formation. While the specifics are rather specialized, the problem of transcriptional hub formation and robust gene expression is a general problem of importance.

Nevertheless, there are still specific issues with the main conclusions based on the work presented, which seems somewhat preliminary.

According to Figure 2, Df(X)svb108removes more than DG3, including elements that clearly bind Ubx as per Figure 1—figure supplement 1, so it is not clear if the observed defects can be attributed to loss of DG3. Ideally, CRISPR/Cas9 should be used to remove Ubx-bound regions outside DG3 (e.g. Ubx-bound DG2) and to remove DG3 specifically, in order to support the conclusions that the observed effects of Df(X)svb108 are due to a loss of DG3.

DG3 remains active in the most dorso-lateral cells in Ubx mutant (Figure 1), in a domain that seems to be specifically marked by DG3 reporters (Figure 1). This domain is portrayed as primarily affected by reduction of trichome formation in Df(X)svb108 mutants, but then it is not clear that this defect is due to a lack of Ubx activity on DG3 (both because DG3-LacZ seems to remain active and because the region removed by Df(X)svb108binds Ubx outside DG3).

In Figure 3, the discrepancy between robust accumulation of Ubx, except in Df(X)svb108embryo under stress, and the impact of heat stress on transcription even in wild-type embryos further indicates that changes in Ubx accumulation is only part of the explanation for both the Df(X)svb108 phenotype and the response to heat stress.

The rescue presented in Figure 4 is a strong result but also difficult to interpret: for instance, it seems likely that the "hub" forms independently of Ubx, which presumably cannot accumulate on the Df(X)svb108chromosome, and then help place the svb locus in a Ubx-rich microenvironment. This is implied but not explicitly stated, or clearly demonstrated by the data.

For instance, Figure 5 adds to the notion that regulatory inputs other than DG3 and Ubx govern the formation of the svb containing micro-environment and svb transcription. As acknowledged also in the Discussion, robustness of trichome formation also involves mechanisms downstream of svb transcription, making the link between Ubx hubs and robust trichome formation even more complex.

Taken together, a few of the criticisms above question whether the accumulation of Ubx, or the lack thereof, in microenvironments containing the svb locus explains the phenotypes described for Df(X)svb108. Rescue experiments more directly combining loss of DG3 function with gain of Ubx function should help resolve these issues.

Reviewer #3:

In their previous work, Tsai, et al. observed that, during fruit fly embryogenesis, actively transcribing shavenbaby (svb) loci preferentially colocalized with regions of elevated concentration for the transcription factor Ubx, a known regulator of svb activity. Moreover, they observed that other Ubx target genes exhibited a tendency to colocalize to these same Ubx-enriched "microenvironments", even when located on different chromosomes. Based on these findings, they hypothesized that the spatial overlap between multiple distal enhancers and transcription factor microenvironments could result in increased transcriptional output and could be a mechanism for redundancy in the face of environmental stresses. In the present work, Tsai, et al. test this hypothesis directly by deleting the DG3 enhancer from the svb cis-regulatory region and assessing the effects on microenvironment formation, svb transcription, and trichome number (a phenotype tied to svb expression) under normal (25C) and heat-stress (32C) conditions. This manuscript constitutes an intriguing extension of the findings presented in the authors' earlier work on Ubx microenvironments in the vicinity of svb loci. They find that the deletion mutant results in a significant depletion in Ubx concentration around the mutated svb locus at 32C, and that this depletion coincides with a reduction both in svb transcription and trichome number. Furthermore, they find that the addition of the full cis-regulatory region of svb on a different chromosome leads to preferential colocalization between the added region and the original mutated svb locus. The authors report that local concentration of Ubx in the vicinity of these colocalized enhancer regions recovers to wild-type levels under heat shock. They argue that, consistent with their initial hypothesis, the coincident partial recovery of svb transcription and trichome number indicates that this colocalization of multiple Ubx targets to a single microenvironment is indeed functioning to counteract the embryo's mutation-induced sensitivity to heat shock conditions.

Overall, the new experimental results presented in this work are deeply intriguing and hold the potential to offer significant new insights regarding the physical nature of transcription factor hubs, as well as their functional role in the spatiotemporal control of transcriptional activity. Nonetheless, several key questions remain regarding the interpretation of the experimental data and the proposed functional role of transcriptional microenvironments in facilitating phenotypic robustness in the face of genotypic and environmental perturbations.

Perhaps most fundamentally, the authors' conclusions hinge upon the assumption that the Ubx enrichment signal they observe at active loci play a causal role in modulating transcriptional activity, yet they present no data to directly support this assumption. Indeed, the box plots in Figure 3 C and D seem to show that, while both Ubx concentration and transcription decrease in response to heat, DG3 deletion, or both, the relative impacts of these perturbations on transcription and Ubx concentration differ significantly in scale. This leaves open the possibility that the perturbations affect each feature separately, and that the correlation between transcription and local Ubx concentration are spurious, not causal. A simple but important check for this would be to show scatter plots of Ubx concentration vs. svb transcription within each condition. If the input concentration and output rate of transcription for a given locus are positively correlated when other conditions are held constant, this would constitute a solid basis for the authors' subsequent arguments. If, on the other hand, no correlation is evident, then this would fundamentally alter how the manuscripts' results are interpreted. This question can be resolved with data already in-hand and, in this reviewer's opinion, it will help clarify the significance of the reported experimental observations. Along similar lines, the strength of the work would benefit from the addition of plots that clarify not only how different variables (concentration, locus colocalization, transcription rate) change across conditions, but also with respect to one another.

On a more conceptual note, the authors' proposal that the colocalization of enhancers in transcription factor microenvironments could serve as an added layer of redundancy (akin to multiple TF binding sites and multiple enhancers) to increase the transcriptional robustness is thought-provoking. The concept of robustness is used throughout biology with different meanings, and further discussion of the implications and nuances of this proposal would also help establish what exactly the authors mean by it. For instance, what the authors cast as "robustness" could equally well be termed "desensitization". While robustness to environmental changes could be desirable, the fast pace of Drosophila development also necessitates rapid, precise transcriptional responses to changes in transcription factor concentrations. Thus, naively at least, if robustness leads to desensitization of the response of an enhancer to changes in the concentrations of inputs transcription factors, one might expect locus colocalization to actually be deleterious in the case of genes that must respond rapidly to time-varying transcription factor inputs. Augmenting the existing commentary to address this point would help clarify the implications of the present work. Does this trade-off make predictions regarding where and when this kind of clustering might manifest over the course of development (and for what genes)?

General Comments:

Can the authors comment on why the box plots in Figure 3 and Figure 4 contain so few data points? All contain 52 or fewer, yet one would expect there to be many more active svb loci per embryo at this point in development. If the full set of svb loci was not used for the analyses presented, can the authors comment on how the analysis subset was selected. Also, given that many of the effects presented are relatively subtle, the addition of data points (to the extent that it is feasible) would greatly enhance the robustness of the analysis and might lead to the discovery of additional trends within the data not evident in such small sample sizes. Given that the authors get multiple nuclei per embryos, is there a fundamental limitation to how much data they can present such as, for example, their data analysis pipeline?

As mentioned above, this work leans heavily upon the presumed causal relationship between local Ubx concentration and transcriptional output. The work would be greatly strengthened by the addition of bivariate analyses that rigorously test the relationship between Ubx concentration and transcription within each of the four conditions shown in Figure 3 C and D.

The boxplots in Figure 4C show that Ubx concentration at loci where the mutated svb site is colocalized with the svbBAC more or less recovers Ubx concentration at the endogenous (unperturbed) svb locus, yet the authors do not address whether the resulting signal is less than, equal to, or greater than the sum of the signal at each locus (Df(X)svb and svbBAC) when they are not colocalized. Showing how the enrichment signal at colocalized loci compares to the signal at each separately would indicate whether and to what degree some sort of synergy is at play, or whether the increased signal is merely a result of having more Ubx binding sites in region.

While the trends in Ubx intensity, radial separation, and transcriptional activity shown in Figure 4C, E, and F are interesting taken individually, much more could be learned from the data by making bivariate plots. How does transcriptional output at the mutated svb locus vary with the radial separation between loci? What about the Ubx intensity?

eLife. 2019 Jul 11;8:e45325. doi: 10.7554/eLife.45325.017

Author response

[Editors' note: the authors’ plan for revisions was approved and the authors made a formal revised submission.]

Essential points to address are the following:

1) The connection between enhancers, Ubx protein concentration, transcriptional output, and phenotype. Although different concentrations of Ubx are measured at the deleted vs. wild-type locus, an important point that the three reviewers agree upon is that gene expression levels and Ubx concentrations are not well correlated. There appear to be other unexplained factors that are influencing the expression of svb that do not correlate with the TF concentration. Thus, it is unclear whether the experiments described here support or reject the hypothesis laid out in the "multi-enhancer "hubs" improve robustness by increasing transcription factor retention near transcription sites."

The reviewers are correct to point out that there are multiple inputs that control the response of the svb locus (Stern and Orgogozo, 2008). Therefore the response function of svb could have a positive but complicated relationship to Ubx concentrations and would depend on more than Ubx alone. We now state this in the second paragraph of the Discussion. However, Ubx is a crucial driver of svb and, specifically, also DG3 expression in the ventral region of the A1 segment (Crocker et al., 2015) and Figure 1 of this manuscript. We specifically focused most of our analyses and quantifications in A1. We now state this in the Results section, “Transcription sites from the DG3‐ deletion allele have weaker Ubx microenvironment and lower transcriptional output”.

Within this specific body segment, Ubx intensity near svb transcription sites would be a reasonable metric of how local transcription factor concentrations changed when we perturbed the system through mutations and elevated temperature. We now state this in the Results section “Transcription sites from the DG3‐deletion allele have weaker Ubx microenvironment and lower transcriptional output”. With these caveats in mind, we did observe increased Ubx concentrations and around the wildtype svb transcription sites (Figure 3C, lower right panel) compared to the deletion allele, and increased phenotype resilience for the wildtype allele at 32 °C. In the svbBAC rescue, we also observed that colocalized svbBAC and DG3‐deleted svb allele had both higher Ubx concentrations and svb transcriptional outputs (Figure 4F). We, therefore, believe that our results support the idea that multi enhancer hubs help transcriptional factor retention and can ultimately lead to a more robust phenotype. We believe the analysis that the reviewers suggested below also supports our hypothesis of improved transcriptional factor retention with complexed enhancers.

The authors must consider further whether the global measurements shown in Figure 3 are the most effective measure of this proposed correlation, or whether a more fine-grained analysis might support the idea that boosted Ubx levels from complexed enhancers are driving transcription. As noted below, the authors might show scatter plots of Ubx concentration vs. svb transcription within each condition. A positive correlation would support their hypothesis.

We have modified the analysis in Figure 3C to show svb transcription vs. Ubx intensity. There is initially a positive correlation between svb transcriptional output and Ubx intensity, in line with the reviewers’ proposal. This trend dissipates at higher Ubx and svb intensities, indicating that the response of svb output to Ubx concentration is not a simple relationship. Capturing the exact dependence of transcriptional output on Ubx concentration is difficult for the reasons we stated in the second paragraph of the Discussion (e.g., multiple binding sites in enhancers and overlapping expression patterns). To fully address this would require future live imaging experiments that can track the gene locus regardless of its transcriptional state, in addition to reporting on its transcriptional activity and transcription factors around it. Numerous new reagents (fly lines, tagged proteins, etc.) are being developed in the lab to address this.

The trichome phenotype is also not directly correlated to the transcriptional outputs and Ubx concentrations, and as mentioned in the Discussion, further layers of regulation are likely in play. This idea needs to be more fully fleshed out.

We now have a new figure (Figure 6) summarizing our hypothesis and referred to this figure throughout the Discussion to clarify our proposed mechanism. In short, we propose in our revised Discussion that 1) the relationship between Ubx concentration and svb transcriptional output is positive but complex, likely involving additional factors (paragraph two) and 2) the relationship between svb transcriptional output and phenotype is sigmoidal (paragraph four) and the wild‐type system operates in a saturated regime under ideal conditions (paragraph five). The processes leading to the response functions in 1 and the phenotypical tolerance to a range of svb out in 2 remain to be investigated (paragraphs two and five), specifically with live imaging approaches. As mentioned in the previous paragraph, work is currently ongoing in the lab to address this.

A finding related to "enhancer hubs boosting local concentration" is the finding that an ectopic BAC appears not to influence Ubx concentration in a wild-type svb background, only in a deletion background. Does this finding indicate that such hub formation is saturable? And if so, does this indicate that the wild-type svb locus forms a local hub with enhancers in the immediate vicinity, and trans-complementation is not a normal feature of svb function?

We agree with the reviewers that hub formation is saturable, based on this observation. We also believe that local hub formation with the cis‐regulatory region of the wildtype svb is sufficiently in saturation to deal with environmental challenges as seen in trichome numbers. We now state this in the Discussion (paragraph five). Our data, however, does not provide a direct answer as to if trans‐ chromosomal interactions are a normal feature of svb or other genes. Ongoing efforts in the lab to characterize and map these long‐range interactions through imaging and genomics approaches should shed light on their functional impact during development in the future.

2) A second point raised concerned the exact cis regulatory regions in play. A minimal DG3 enhancer drives gene expression in ventral abdominal stripes, but does not rescue Ubx concentrations from a trans setting. A larger deletion that includes DG3 as well as additional Ubx binding regions (that are not sufficient for, but may be part of, ventral DG3-related activity) impacts transcription, trichome development, and robustness. A yet larger cis-regulatory domain on a BAC rescues some aspects of gene expression and Ubx concentration. The interpretation conflates DG3 with the function of the deleted region; reviewers noted that a more careful interpretation would differentiate results from each of these different cis elements. For instance, the lack of trans-rescue by the DG3 enhancer alone may be due to the inability of a short segment to transvect effectively. The interpretation should explicitly take into account known properties of transvecting regulatory loci in Drosophila.

We agree that we should take into account the efficiency of the rescue locus in finding the svb locus as an important factor in if phenotype rescue takes place. We now state that DG3 alone as the rescue locus might have failed due to its inability to pair with the svb locus in the Discussion (paragraph four). We also now compare and contrast the svbBAC‐svb interactions that we observed with transvection and hypothesize that the addition of other topological elements such as insulator to DG3 could overcome this problem (paragraph four). We further state that the svbBAC did not rescue trichomes in regions where DG3 provided exclusive coverage, suggesting that the rescue BAC rescued the phenotype by overdriving the other ventral enhancers E3 and 7 rather than directly restoring DG3 function.

3) Several aspects can be addressed by better justification and explanation of methods and data presentation, including

– Why sometimes either A1 or A2 trichomes are quantitatively assessed, depending on the figure;

This issue was an oversight on our part, and we have updated all the main figures where we counted trichomes to be from the A1 segment for consistency. We moved data from the A2 segment to figure supplements as they also show a similar but weaker trend as in A1. We speculate that this is due to additional factors at work as DG3 in the A2 segment responds to additional inputs beyond Ubx, as explained in the Results section “The DG3 enhancer responds specifically to Ubx in the A1 segment”.

– The use of one-tailed (vs. two-tailed) T-tests for statistical relevance;

We have changed our test to two‐tailed T‐tests throughout, as is the standard. This did not change of our findings.

– Recommendations for inclusion of both 25C and 32C phenotypes and data uniformly, and;

We added trichome images from 32 C to Figure 2 and for the data analysis involving Ubx intensity and svb transcriptional output in Figure 3. We did not conduct svbBAC rescue experiments at 25 C as neither the wildtype nor the deletion allele displayed reduced trichome numbers (phenotype output), and the difference between the Ubx intensities (molecular input) around transcription sites of both genotypes was small at this temperature.

– Clarification of exact role of Ubx in T1-T3 regulation, as the conclusions drawn about DG3 and Ubx roles are difficult to know based on the single images shown,

As the reviewers noticed, the role of Ubx in regulating svb, and DG3 specifically, on the ventral surface in segments outside of A1 is more complicated (Figure 1B‐E). DG3 expression in the thoracic segments T1‐T3 overlaps with other ventral svb enhancers and responds to changes in Ubx level less that in the A1 segment. T2 and T3 also show only low levels of expression from DG3 on the ventral surface with wild‐type Ubx expression. This issue potentially introduces many confounding factors for quantitation. We have thus confined ourselves to qualitative descriptions of DG3 properties outside of A1 and A2. We now state this concern at several places in the Results and Discussion.

– The number of data points in Figures 3 and 4 are limited; is there a technical limitation to more extensive sampling of the transcriptional readouts and Ubx intensities?

In the process of doing the additional analysis suggested by the review, we added more data points, and the number of transcription sites and embryos quantified is comparable to the original publication this manuscript is linked to (Tsai et al., 2017). As in the previous publication, colocalization is a relatively rare event, so we observed fewer transcription sites. The number observed for colocalized sites is also similar to our previous publication.

4) In some cases, the actual experimental approach was unclear to the reviewers:

– There was some confusion about whether the DG3 deletion mutants were homozygous, so there would be no wild-type copy of the gene in these embryos – is that indeed the case?

Yes, we selected larvae or embryos homozygous for the Df(X)svb108allele based on the following phenotype in the T1 segment: the lack of trichomes or the lack of svb mRNA expression, respectively. This is a homozygous marker for the deletion allele as the wild‐type svb allele expresses in this segment. This was mentioned in “Deletion of a region including DG3 enhancer causes defects in ventral trichome formation specifically at elevated temperatures” in Results, “Preparing Drosophila embryos for staining and cuticle preps” in Materials and methods and the legends for Figure 2C in our original manuscript (2E in the revised version,). We have now made it explicit in the Results that we use this to select for animals homozygous for the deletion for both counting trichome number and for confocal imaging. It is now also described in both “Cuticle preparations and trichome counting” and “Imaging fixed embryos” in the Materials and methods section.

– Figure 4 shows svb, but not dsRed mRNA expression; is that correct? Imaging of dsRed only used to score a svb locus as "overlapped" vs. "not overlapped"?

Yes, the data analysis in Figure 4 only used only svb mRNA output. Imaging of dsRed transcription site is only used to score colocalization with svb. We displayed image panels for individual nuclei that show a dsRed signal for illustrative purposes. We explicitly stated that we are displaying svb transcription in both the Figure 4E and the legends.

– Choice of pixel size for ROI.

The 40‐pixel ROI is too large and a mistake in our part. The actual ROI size in the previous manuscript was a 4‐pixel square ROI; this has been changed to be a circle with a diameter of 4 pixels (170 nm). We chose this size because it is the resolution limit of the AiryScan images we acquired. We have corrected this and stated our rationale in the revision when we describe the image analysis in “Transcription sites from the DG3‐deletion allele have weaker Ubx microenvironment and lower transcriptional output” in the Results and “Analysis of microenvironment and svb transcription intensity” in the Materials and methods.

– Signal to noise for Ubx across the nucleus, as well as variation in average Ubx from sample to sample.

We used exactly the same antibodies and methodology as reported in (Tsai et al., 2017) to stain for Ubx and the quality of the images acquired were the same. In the previous publication we explored the Ubx cluster characteristics. We now include a new supplemental figure (Figure 3—figure supplement 1) with Ubx variation across the nucleus within embryos, and from sample to sample (there is no significant difference). In contrast, at the sites of active svb transcription there is a significant difference (p < 0.001, two‐tailed t‐test) from randomly sampled locations in the nucleus. This has also been added to the Results “Transcription sites from the DG3‐deletion allele have weaker Ubx microenvironment and lower transcriptional output”

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    Supplementary file 1. Primers for RNA-probe generation Sequences of primers for amplification of DNA to be used for generation of antisense RNA-probes.

    The targets - reporter construct RNA (lacZ, dsRed, gfp) and first intron and second exon (16 kb) of svb - are indicated in the left column. Sequences are indicated for forward or reverse primers of each pair. Reverse primers include a T7 sequence for transcription with T7 RNA polymerase.

    elife-45325-supp1.xlsx (11.1KB, xlsx)
    DOI: 10.7554/eLife.45325.012
    Transparent reporting form
    elife-45325-transrepform.pdf (335.6KB, pdf)
    DOI: 10.7554/eLife.45325.013

    Data Availability Statement

    The original images (cuticle preparations and embryo images, organized into zip files) are available for download and are indexed at: https://www.embl.de/download/crocker/svb_enhancer_colocalization/index.html. Please note that the raw AiryScan images must be processed though the Zen software from Zeiss before they can be opened/analyzed using standard image processing softwares. These files are large, totaling up to approximately 180 GB in size. We can also send these files directly if a means of transfer (hard drives, etc.) is provided.

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES