Abstract
Free full text
The cis-regulatory logic underlying abdominal Hox-mediated repression versus activation of regulatory elements in Drosophila
Associated Data
Abstract
During development diverse transcription factor inputs are integrated by cis-regulatory modules (CRMs) to yield cell-specific gene expression. Defining how CRMs recruit the appropriate combinations of factors to either activate or repress gene expression remains a challenge. In this study, we compare and contrast the ability of two CRMs within the Drosophila embryo to recruit functional Hox transcription factor complexes. The DCRE CRM recruits Ultrabithorax (Ubx) and Abdominal-A (Abd-A) Hox complexes that include the Extradenticle (Exd) and Homothorax (Hth) transcription factors to repress the Distal-less leg selector gene, whereas the RhoA CRM selectively recruits Abd-A/Exd/Hth complexes to activate rhomboid and stimulate Epidermal Growth Factor secretion in sensory cell precursors. By swapping binding sites between these elements, we found that the RhoA Exd/Hth/Hox site configuration that mediates Abd-A specific activation can convey transcriptional repression by both Ubx and Abd-A when placed into the DCRE. We further show that the orientation and spacing of Hox sites relative to additional binding sites within the RhoA and DCRE is critical to mediate cell- and segment-specific output. These results indicate that the configuration of Exd, Hth, and Hox site within RhoA is neither Abd-A specific nor activation specific. Instead Hox specific output is largely dependent upon the presence of appropriately spaced and oriented binding sites for additional TF inputs. Taken together, these studies provide insight into the cis-regulatory logic used to generate cell-specific outputs via recruiting Hox transcription factor complexes.
Introduction
The generation of distinct cell types within the metazoan body plan requires the accurate regulation of gene products needed to make specialized cell types. At the transcriptional level, cis-regulatory modules (CRMs) use transcription factor binding sites (TFBSs) to recruit and integrate diverse transcriptional inputs that modulate the cell specific production of mRNA [1]. Recent genomic studies indicate most genes have numerous distinct CRMs, each of which contributes to the overall gene transcriptional output [2]. Much like we can read the primary genomic DNA sequence to predict intron/exon boundaries and protein coding regions, we would also like to read CRM sequences to predict both the relevant TFBSs and the transcriptional output mediated by the CRM. However, there are several challenges in predicting CRM function from primary sequence. First, metazoan genomes encode for hundreds to thousands of different sequence-specific transcription factors, many of which are members of protein families that bind highly similar sequences in a degenerate manner [3]. Second, the relatively few CRMs that have been thoroughly characterized reveal differences in the importance of TFBS organization in mediating robust outcomes. Some CRMs are able to use flexible TFBS arrangements (i.e. the billboard model) whereas others require precise TFBS organization to mediate cooperative TF complexes (i.e. the enhanceosome model) [1]. In this paper, we dissect how distinct TFBS arrangements within two conserved Hox-regulated CRMs contribute to opposing transcriptional outcomes in the Drosophila ectoderm.
Hox genes encode homeodomain transcription factors that specify distinct cell fates along the developing anterior-posterior (A-P) axis of metazoans [4]. Most animals contain at least five Hox genes that are often clustered in the genome [5]. Drosophila melanogaster encodes a single set of eight Hox genes that are split into two clusters (five in the Antennapedia complex and three in the Bithorax complex), whereas mammalian genomes have undergone Hox cluster and gene duplication resulting in four clusters that encode a total of 39 Hox genes [5]. While the number of Hox genes varies between animals, Hox genes share the property of instructing cells to adopt a "regional" (or "segment") identity within the organism by regulating the expression of target genes [6]. Genomic studies have found that Hox factors affect the expression of hundreds of downstream target genes [7-10]. Since each segment under the control of a Hox factor is composed of many cell- and tissue-types, these findings present two challenges in understanding how Hox genes sculpt the body plan: First, what makes one Hox factor different from another to specify distinct embryonic regions during development? Second, how can a regionally expressed Hox factor regulate target genes in a cell- or tissue-specific manner?
Much of the focus on how Hox factors regulate distinct cell fates has been to define the mechanisms underlying DNA binding specificity. Comparative studies between Hox factors revealed each binds similar AT-rich DNA sequences [11-13]. These findings raise a paradox: how do proteins that bind similar DNA sequences in vitro regulate distinct target genes and cell fates in vivo? A partial explanation for this phenomenon is that Hox factors form complexes with additional transcription factors. The Extradenticle (Exd, Drosophila)/Pbx (vertebrate) and Homothorax (Hth, Drosophila)/Meis (vertebrate) homeodomain proteins represent the best characterized Hox co-factors [14-18]. Exd/Pbx and Hth/Meis are widely expressed during development and cooperatively bind DNA with Hox factors. As each protein in the complex interacts with DNA in a sequence-specific manner, Hox/Exd/Hth complexes enhance both target affinity and specificity over Hox binding alone [19-21]. Moreover, a selection assay revealed that Hox factors gain discriminatory power when binding DNA with Exd, a concept called latent specificity [22]. The best example of latent specificity is the activation of forkhead (fkh) by the Sex combs reduced (Scr) Hox factor during salivary gland development [23, 24]. The fkh Exd/Hox binding site has a narrow minor groove that is only bound by Scr when in complex with Exd, and changing this site to match a generic Exd/Hox consensus (Fkhcon) resulted in loss of Hox specificity as evidenced by high affinity binding by other Hox factors [25, 26]. However, not all the Hox factors that bound the Fkhcon sequence similarly mediated activation as Scr, as a subset instead repressed transcription through unknown mechanisms [25]. Thus, different Exd/Hox sites can discriminate between Hox factors, and once bound, the Hox/Exd complexes can differ in mediating distinct regulatory outcomes.
To better understand the mechanisms of how Hox binding sites mediate distinct outcomes, we have focused on defining how the Abdominal-A (Abd-A) Hox factor contributes to both activation and repression when in complex with the Exd and Hth proteins. The RhoBAD CRM contains a highly conserved sequence (RhoA) encoding an adjacent set of Exd/Hth/Hox sites that recruits an Abd-A complex to mediate rhomboid (rho) activation in a subset of sensory organ precursor (SOP) cells [27-30]. The activation of rho, which encodes a serine protease that triggers the release of an EGF ligand, results in the induction of neighboring cells to form an essential set of hepatocyte-like cells known as oenocytes [31-34]. In contrast, the Distal-less Conserved Regulatory Element (DCRE) contains three Hox/co-factor binding sites that recruit Abd-A/Exd/Hth complexes to repress Distal-less (Dll) gene expression in the abdominal ectoderm [35-37]. Dll, an appendage selector gene that promotes leg formation in thoracic segments, is thereby restricted from the abdomen to block appendage formation in these segments [38]. Intriguingly, the RhoA and DCRE elements differ in their ability to discriminate between functional Hox complexes, as the DCRE is regulated by both Abd-A and Ultrabithorax (Ubx), whereas RhoA is regulated by only Abd-A.
What determines if a CRM is activated or repressed when bound by a specific Hox factor? Current models suggest that the RhoA and DCRE CRMs integrate additional transcription factors that help dictate the sign of transcription. For example, RhoA requires a nearby Pax2 binding site to mediate activation, whereas the DCRE contains a nearby FoxG binding site (Drosophila express two FoxG homologues, Sloppy-paired 1 (Slp1) and Slp2, which are largely redundant) to mediate repression [28, 36]. How these additional factors are integrated with the Hox transcription factor complexes and the role TFBS organization plays in mediating each cell-specific output is unclear. In this study, we use a series of quantitative reporter assays to define the underlying cis-regulatory logic and mechanisms of Hox regulatory specificity by comparing and contrasting the ability of abdominal Hox factors to affect the activity of the DCRE and RhoA CRMs in conjunction with FoxG and Pax2. Our findings provide new insights into how the organization of Hox, Exd, and Hth binding sites within CRMs contribute to achieving both Hox specificity and positive versus negative regulatory specificity.
MATERIALS AND METHODS
Transgenic Reporter Assays
Oligonucletoides for DCRE sequence variants were ordered from Integrated DNA Technologies and cloned into the pAttB-LacZ plasmid containing 3xGBE. RhoAAA sequences were similarly ordered and cloned into the pAttB-LacZ plasmid. DNA sequences for each site are found in Supplemental Data. The DNA spacer sequence was generated by PCR amplification of a portion of the kanamycin gene as previously described [39]. All plasmids were sequence-confirmed prior to injection. Transgenic flies were created using the -C31 system with each construct inserted into the same locus (51C) [40]. Injections were conducted by Rainbow Transgenics Inc.
Drosophila stocks carrying lacZ transgenes were made homozygous for reporter constructs, and embryos were collected and stained using standard procedures at 25°C. The UAS-HA-Ubx and UAS-HA-AbdA lines were a kind gift from Richard Mann, and the PrdG4;UAS-HA-Ubx and PrdG4;UAS-HA-AbdA experiments were performed at 25°C. Embryos were immunostained using the following primary antibodies: chicken anti-β-gal (1:1000) (Abcam), guinea-pig anti-Abd-A (1:500) [28], rat anti-Slp2 (1:500) [37], mouse anti-Ubx (DSHB, 1:50), rabbit anti-Salm (1:2000) [41], and rat anti-HA (1:1000). Immunostains were detected using fluorescent secondary antibodies (Jackson Immunoresearch Inc and AlexaFluor).
All samples were imaged using a Zeiss Axio Imager upright microscope with an Apotome filter for optical sectioning. A AxioCam MRm digital camera was used to capture the images. Images were quantified manually using NIH ImageJ software. For each embryo, a region-of-interest (ROI) was placed over Slp2+ cells in embryonic segments T2 through A4 (see Fig S1 for an example). An additional ROI was created to measure background fluorescence. The area and shape of the ROIs were the same for all measurements within a comparison set. For each embryo, the mean β-gal intensity of the background ROI was subtracted from the mean β-gal intensity of each foreground ROI. The resultant background-subtracted β-gal intensities were normalized to the background-subtracted β-gal intensity of the T3 (third thoracic segment) ROI. To statistically compare the abdominal activity between reporter genotypes, the resultant normalized β-gal intensities of abdominal segments were mean averaged per embryo and compared using the statistical test indicated in each Figure legend. Statistical analyses and plotting were conducted using R and the ggplot2 package.
DNA binding assays
A His-tagged Slp1 construct was made by cloning the N-terminus through the DNA binding domain of Slp1 (amino acids #1 through 216) into a pET14b vector. His-tagged Exd-Hth heterodimers, Abd-A, and Slp1 protein were purified from BL21 using Ni+ beads, as described previously [36, 42]. SDS-PAGE and Coomassie blue staining was used to confirm purification of the protein of interest. For EMSA, fluorescent DNA probes (Integrated DNA Technologies) were mixed with proteins as indicated in Figure legends, incubated for 10 min prior to running on polyacrylamide gels, and imaged using an Odyssey LiCOR cLX scanner as previously described [37, 43].
Results
The DCRE mediates short-range transcriptional repression
Dll encodes an appendage selector gene that is directly bound and repressed by both the Ultrabithorax (Ubx) and Abdominal-A (Abd-A) Hox factors to suppress abdominal leg development in Drosophila [38, 44], Dll expression in the embryonic leg primordia is mediated by the DMX cis-regulatory module that can be divided into two parts: the DMEact conveys activation in thoracic and abdominal segments, and the DCRE recruits Ubx and Abd-A to repress transcription in a compartment-specific manner in abdominal segments [35, 44]. In anterior compartment cells, Ubx and Abd-A mediate repression with the Slp FoxG factors, whereas in posterior compartment cells Ubx and Abd-A cooperatively bind and repress the DCRE with the Engrailed (En) repressor protein [36]. However, recent findings revealed the abdominal Hox factors also repress the DMEact via unknown DCRE-independent mechanisms [37]. Hence, to study the cis-regulatory logic utilized by the DCRE to mediate abdominal repression, we used an assay that isolates the DCRE from the DMEact by placing it adjacent to three copies of the Grainyhead (Grh) binding element (3xGBE). 3xGBE sites are sufficient to activate gene expression throughout the ectoderm of the Drosophila embryo [37, 45]. As previously reported, comparisons between 3xGBE-lacZ (G-lacZ) and 3xGBE-DCRE-lacZ (GD-lacZ) transgenes inserted into identical loci revealed the DCRE mediates robust repression in anterior compartment abdominal cells that co-express a Drosophila FoxG (Slp2) factor (Fig 1A-C), but not in En-positive posterior compartment cells [37]. To quantify abdominal repression in anterior compartment cells, we measured β-gal intensity in Slp2+ cells (see Methods and Fig S1) and found that G-lacZ drives equivalent reporter levels in thoracic and abdominal segments whereas GD-lacZ embryos had ~70% less activity in Slp2-positive cells of abdominal segments relative to thoracic segments (Fig 1G). Hence, the GD-lacZ assay provides a means to isolate and study DCRE-mediated repression by abdominal Hox factors and Slp FoxG proteins independent from the more complex DMX element.
(A) LacZ reporter constructs used to test DCRE activity. The 3xGBE is a pan-ectodermal enhancer. The 736 bp spacer is transcriptionally inert (Fig S1. )(B-F) Lateral view of Stage 15 embryos expressing the indicated LacZ reporter constructs. Embryos were immunostained for β-gal (green), Slp2 (red), and Abd-A (blue). (B’-F’) Close-up view of the T2-A2 segments from panels B-F. (B”-F”) Same as B’-F’ but only showing β-gal stain. (G) Quantification of β-gal immunostain intensity among Slp2+ cells in segments T2-A4 relative to intensity in T3 segment. Background color indicates the Hox expression domain as shown by the legend (top-left). Light gray lines are intensity values from individual embryos, while the dark black line displays median value at each segment. The grey ribbon displays the standard deviation at each segment. Tick marks on the right border of the plots indicate mean abdominal β-gal value per embryo (as shown by legend, bottom-right). Abdominal reporter expression was statistically compared between genotypes using ANOVA followed by post-hoc Tukey’s test. Prior to ANOVA, abdominal reporter expression in each individual embryo was mean averaged. Crosses mark statistically significant difference from G-LacZ (“n.s.” p >= 0.05, “†” p < 0.05,“††" p < 0.01, “†††" p < 0.001). Asterisks mark statistically significant difference from GD-LacZ (“n.s.” p >= 0.05,“*” p < 0.05, “**" p < 0.01, “***" p < 0.001). Additional pairwise comparisons are indicated by bars and hash-signs (“n.s.” p >=0.05 “#” p >= 0.05, “##" p < 0.01, “###" p < 0.001,).
To define the range of DCRE repression activity, we engineered a series of constructs that alter the location of the DCRE relative to the 3xGBE (Fig 1A). First, we swapped the order of the DCRE and 3xGBE (DCRE-3xGBE-lacZ, DG-lacZ) and found that in this configuration the DCRE mediates transcriptional repression, albeit weaker and predominantly in a subset of Slp2+ abdominal cells (Fig 1D, ,1G).1G). Next, we moved the DCRE further from the 3xGBE (DCRE-sp-3xGBE-lacZ, DspG-lacZ) by inserting a 736 bp sequence from the kanamycin gene which was previously found to be transcriptionally inert [39]. Consistent with this spacer DNA not having significant transcriptional activity, we found that inserting it adjacent to the 3xGBE did not significantly alter reporter expression (Fig S2). Importantly, the DCRE was unable to convey abdominal repression when it was separated from the 3xGBE sites by the spacer DNA sequence (Fig 1E, ,1G).1G). However, moving the 3xGBE adjacent to the distant DCRE (3xGBE-DCRE-sp-LacZ, GDsp-lacZ) rescued repression (Fig 1F-G). Thus, these findings are consistent with the DCRE functioning as a relatively short-range element that represses transcription when placed adjacent to activation elements.
Hox specificity and the cis-regulatory logic of RhoA activation versus DCRE repression
The Ubx and abd-A Hox genes encode nearly identical homeodomains, bind highly similar DNA sequences as monomers, and form similar transcription factor complexes with the Exd and Hth Hox cofactor proteins on DNA in vitro [11, 22]. Consistent with these findings, prior studies demonstrated that Ubx and Abd-A repress Dll via the DCRE [35, 36]. To determine if both Ubx and Abd-A repress the DCRE in the GD-lacZ assay, we first compared GD-lacZ activity in Slp2+ anterior compartment cells of abdominal A1 segments that only express Ubx versus subsequent abdominal segments that express both Ubx and Abd-A, and observed a similar degree of repression in each abdominal segment (Fig 2A-B, see quantified segment data in Fig 1G). These results are consistent with Ubx being sufficient to repress gene expression via the DCRE in the abdominal A1 segment. Moreover, genetic removal of Ubx function resulted in a specific loss of GD-lacZ repression in A1 segments, whereas abd-A mutant embryos maintained significant repression in all abdominal segments (Fig 2C-E). However, embryos mutant for both Ubx and abd-A were previously found to de-repress GD-lacZ activity in all abdominal segments [37]. Thus, both Ubx and Abd-A can repress GD-lacZ activity in Slp2+ cells, and analysis of GD-lacZ reporter activity in the A1 abdominal segment specifically tests for Ubx-dependent transcriptional repression.
(A-B) 3xGBE-DCRE-LacZ (GD-LacZ) reporter embryo immunostained for β-gal (green), Slp2 (red), and Ubx (A) or Abd-A (B) (blue). (A’-B’) Close-up view of the T2-A2 segments shown in panels A-B. (A”-B”) Same as A’-B’ but only showing β-gal. Yellow outline indicates position of Slp2+ cells. (A”’-B”’) Same as A”-B” showing only Slp2 stain with either Ubx (A’”) or Abd-A (B’”) stain. (A””-B””) Same as A”-B” showing only Ubx (A””) or Abd-A (B””) stain only. Note, only Ubx is expressed in the Slp2+ cells in the A1 segment whereas both Ubx and Abd-A are expressed in the Slp2+ cells in the A2 through A4 segmeents. (C-D) GD-LacZ reporter embryos immunostained for β-gal (green), Slp2 (red), Abd-A (blue) in abda mutant (C) and Ubx mutant (D) embryos. (E) Quantification of β-gal immunostain intensity of GD-LacZ in abd-A and Ubx mutant backgrounds. Statistical comparison of β-gal intensity between A1 and mean A2-A4 abdominal segments was conducted using a t-test (“n.s.” p >= 0.05, “*” p < 0.05, “**" p < 0.01, “***" p < 0.001). (F-G) RhoBAD-lacZ;PrdG4 embryos (Stage 11) expressing either an HA-tagged AbdA (F) or HA-tagged Ubx (G) protein. Embryos were immunostained for β-gal (blue), HA (green), and Spalt-major (Salm, red). Note, Abd-A increases β-gal levels and induces additional Salm+ cells in the T2 segment whereas Ubx does not.
Unlike Dll and the DCRE, which are repressed by Abd-A, rhomboid (rho) is activated by Abd-A in a subset of abdominal sensory organ precursor cells (SOPs) via a highly conserved Exd/Hth/Hox binding site within RhoBAD [27, 28]. rho encodes a serine protease that triggers the release of an EGF ligand and neighboring cells that receive the EGF signal are specified to form larval oenocytes [32, 33]. Previous studies showed that the loss of abd-A, but not Ubx, resulted in a failure to induce oenocytes, suggesting that only Abd-A is required for activating rho in abdominal SOP cells [31]. However, Ubx is weakly expressed in these abdominal SOP cells. To test whether Ubx is capable of activating RhoBAD, we used the PrdG4 driver to ectopically express high levels of Ubx in the thorax and found that neither RhoBAD-lacZ nor oenocytes (marked by high Spalt-major (Salm) expression) were substantially induced in thoracic segments (Fig 2G). In contrast, PrdG4;UAS-Abd-A embryos induced both RhoBAD-lacZ activity and oenocytes in the thorax (Fig 2F). Thus, while both Abd-A and Ubx can repress the DCRE to inhibit leg development, only Abd-A activates RhoBAD to induce abdominal oenocyte cells.
A notable difference between the DCRE and RhoA sequences is the organization of the Hox, Exd, and Hth sites. RhoA contains a single set of contiguous Exd/Hth/Hox sites, whereas the DCRE has multiple Hox sites that are each coupled to either an adjacent Exd or Hth binding site (Fig 3A). To determine if the organization of Hox, Exd, and Hth sites contributes to Hox specificity (Abd-A and not Ubx regulation via RhoA binding sites), we generated transgenic GD-lacZ lines in which the core Exd/Hth/Hox sites of RhoA replaced the Hox/Exd-Hth/Hox sequences within the DCRE (Fig 3B). Intriguingly, we found that the RhoA Exd/Hth/Hox sites can function in the DCRE to repress gene expression, but only in one orientation. For example, in the arbitrarily assigned forward direction (AF) the Exd/Hth/Hox sites failed to repress whereas in the reverse orientation (AR) these same sequences mediated significant repression (Fig 3E-F, ,3I).3I). Electromobility shift assays (EMSAs) using purified Exd/Hth heterodimers and Abd-A revealed no significant differences in binding patterns using probes of the DCRE-AF versus the DCRE-AR, suggesting that differences in Hox DNA binding activity cannot explain the failure of the DCRE-AF to mediate repression (Fig S3). Moreover, segment specific analysis of GD-AR-LacZ activity in Slp2+ cells revealed that the A1 segment cells that express only Ubx exhibit a similar degree of repression as the abdominal segments that also express Abd-A (Fig 3E-F, ,3I).3I). These data suggest that Ubx can bind to and mediate repression on the RhoA Exd/Hth/Hox configuration of sites in the context of the DCRE, but it cannot mediate activation in the context of the RhoBAD element (Fig 2G). Consistent with this idea, EMSA analysis revealed Ubx binds the RhoA probe with Exd/Hth (Fig S4). In addition, point mutations within the RhoA Hox site that were previously shown to disrupt Abd-A mediated activation of the RhoBAD CRM [27], resulted in a dramatic loss of abdominal repression activity in all abdominal segments of GD-AR-LacZ, including in the Ubx-specific A1 abdominal segment (Fig 3F-G, ,3I).3I). In total, these data show that the RhoA Exd/Hth/Hox sites are not strictly Abd-A specific and Ubx can utilize this configuration of sites to mediate repression when placed into the DCRE.
(A) A schematic of Ubx and Abd-A action on the DCRE and RhoA elements. Note, both Ubx and Abd-A repress the DCRE, while only Abd-A can activate the RhoA. (B) Schematic demonstrating how the Hox/Exd-Hth/Hox binding sites in DCRE were modified in the 3xGBE-DCRE-LacZ (GD-LacZ) constructs. For GD-AF the Hox/Exd-Hth/Hox binding sites in DCRE were replaced with the Exd/Hth/Hox binding sites from RhoA, whereas they were inserted in the reverse complement direction to create GD-AR-LacZ. The RhoA Hox binding site in GD-AR-LacZ was mutated to create GD-AR-HoxM-LacZ. The Hox/Exd-Hth/Hox binding sites from wildtype DCRE were placed in the reverse complement orientation to create the GD-HF-LacZ construct. (C-H) Lateral views of Stage 15 embryos expressing the indicated LacZ reporter constructs. Embryos were immunostained for β-gal (green), Slp2 (red), and Abd-A (blue). (C’-H’) Close-up view of T2-A2 segments from panels C-G. (C”-H”) Same as C’-G’ but only showing β-gal stain. (I) Quantification of β-gal immunostain intensity among Slp2+ cells in segments T2-A4 relative to intensity in T3 segment. Abdominal reporter expression was statistically compared between genotypes using ANOVA followed by post-hoc Tukey’s test. Prior to ANOVA, abdominal reporter expression in each individual embryo was mean averaged. Crosses mark statistically significant difference from G-LacZ (“n.s.” p >= 0.05,“†” p < 0.05, “††" p < 0.01, “†††" p < 0.001). Asterisks mark statistically significant difference from GD-LacZ (“n.s.” p >= 0.05, “*” p < 0.05, “**" p < 0.01, “***" p < 0.001). Additional pairwise comparisons are indicated by bars and hash-signs (“n.s.” p >= 0.05, “#” p < 0.05, “##" p < 0.01, “###" p < 0.001).
The orientation and spacing of Hox sites relative to FoxG sites is critical for Abdominal Hox-mediated repression of the DCRE
The findings that the RhoA Exd/Hth/Hox sites can mediate abdominal repression in Slp+ cells when inserted into the DCRE in only one direction (DCRE-AR) suggests that Hox binding site orientation relative to the FoxG sites may be a critical factor in conveying output. Consistent with this idea, the Hox/Exd sites in the DCRE-AR (but not in the DCRE-AF) are in a similar orientation and spacing relative to the FoxG sites as in the wild type DCRE (Fig 3B). Moreover, we found that when the entire Hox/Exd-Hth/Hox sequence within the DCRE was "flipped" over in the opposite orientation (DCRE-HF), it placed the Hox/Hth site in a similar orientation and spacing relative to the FoxG site as the original Hox/Exd site and repressed Slp2+ abdominal gene expression as well as the wild type DCRE (Fig 3H-I).
To further test the importance of spacing between the DCRE FoxG and Hox sites, we inserted short DNA sequences between these sites. Care was used to ensure the inserted sequences did not code for additional Hox, Exd, Hth, or FoxG binding sites (Fig 4A). Five nucleotide intervals were used to systematically alter the DNA phasing of binding sites along the DNA helix (10 nucleotides = ~1 turn of the DNA helix). Intriguingly, we found that inserting +5 nucleotides (a half phase of the DNA helix) resulted in a complete loss of repression even though all three Hox sites and the FoxG sites are present (compare Fig 4B with with4C,4C, quantified in in4G).4G). In contrast, inserting +10 nucleotides partially rescued abdominal repression in Slp2+ cells, and GD-lacZ reporters with +15 or +20 nucleotide insertions were also able to mediate repression as well or even better than the +10 spacer (Fig 4D-4G). Since the +5bp insertion resulted in a complete loss of repression, we used EMSA analysis to compare DNA binding activity to wild type DCRE, DCRE+5 and DCRE+10 probes and found no significant difference in Abd-A and Exd/Hth binding (Fig S5). These findings suggest that the repression activity mediated by Hox factors is constrained to a specific spacing/orientation when in close proximity to the nearby FoxG sites within the DCRE, but when the distance between these sites is increased, binding site phasing is less critical in mediating transcriptional repression.
(A) Schematics of DCRE variants with +5, +10, +15, or +20 bp inserted between the dFoxG and Hox binding sites. (B-F) Lateral views of Stage 15 embryos expressing the indicated LacZ reporter constructs. Embryos were immunostained for β-gal (green), Slp2 (red), and Abd-A (blue). (B’-F’) Close-up view of T2-A2 segments from panels B-F. (B”-F”) Same as B’-F’ but only showing β-gal stain. (G) Quantification of β-gal immunostain intensity among Slp2+ cells in segments T2-A4 relative to intensity in T3 segment of the indicated LacZ reporter embryos. Abdominal reporter expression was statistically compared between genotypes using ANOVA followed by post-hoc Tukey’s test. Prior to ANOVA, abdominal reporter expression in each individual embryo was mean averaged. Crosses mark statistically significant difference from G-LacZ (“n.s.” p >= 0.05, “†” p < 0.05, “††" p < 0.01, “†††" p < 0.001). Asterisks mark statistically significant difference from GD-LacZ (“n.s.” p >= 0.05, “*” p < 0.05, “**" p < 0.01, “***" p < 0.001).
Altering the RhoA Hox and Pax2 site configuration disrupts gene activation in abdominal SOP cells
Our studies revealed that the three Hox sites within the DCRE that are coupled to either a Exd or Hth site can recruit abdominal Hox complexes to mediate repression [37]. In contrast, the RhoA element contains a single contiguous set of Exd/Hth/Hox sites that mediate activation with a nearby Pax2 site (Fig 3A). To determine if the Hox, Exd, and Hth sites from the DCRE can similarly mediate RhoA activation in abdominal SOP cells, we used a previously established transgenic reporter assay based on three copies of the RhoA element (RhoAAA-lacZ) being sufficient to mediate activation in abdominal SOP cells [28, 30] (Fig 5B). To do so, we made RhoAAA-DF (forward) and RhoAAA-DR (reverse) constructs in which the Pax2 site was maintained, but the RhoA Exd/Hth/Hox sites were replaced with the DCRE Hox/Exd-Hth/Hox sites in the "forward" and "reverse" orientations (Fig 5A). Analysis of embryos carrying these transgenes revealed neither RhoAAA-DR-lacZ nor RhoAAA-DF-lacZ were capable of activating transcription in abdominal SOPs (Fig 5C-D). Given that inserting the DCRE Hox/Exd-Hth/Hox sites into RhoA alters the spacing between Hox and Pax2 sites, we next tested how spacing between the Pax2 and RhoA Exd/Hth/Hox sites affects Abd-A mediated activation. For this purpose, we inserted +5 or +10 nucleotide sequences between the Pax2 and Exd/Hth/Hox sites (Fig 5E). Since RhoA also encodes an overlapping Senseless (Sens) binding site that can repress thoracic gene expression, care was taken to ensure that a low-affinity Sens site was maintained and that no new Pax2, Exd, Hth, or Hox sites were created within the RhoA element [46]. In fact, EMSA analysis revealed that adding the +5 or +10 sequences into the RhoA probe did not significantly alter Abd-A/Exd/Hth binding (Fig S6). Like the DCRE, we found that insertion of 5 bp sequences between the Pax2 and Exd/Hth/Hox sites resulted in a loss of Hox mediated activation (Fig 5G). However, unlike the DCRE, the RhoA element did not regain activity when a full helical phase of DNA sequence (+10 bp) was inserted between these sites (Fig 5H). These findings are consistent with two interpretations: First, the Pax2 and Exd/Hth/Hox sites may be highly constrained in order to mediate transcriptional activation in abdominal SOPs. Second, the re-engineered RhoA elements disrupt additional unknown binding sites that are required for abdominal SOP activation. Currently, we favor the former as a previous scanning mutagenesis assay only uncovered the Pax2 binding site [28]. However, additional experiments are needed to conclusively distinguish between these two possibilities.
(A) Schematic showing how the Exd/Hth/Hox binding sites in RhoA were replaced with the Hox/Exd-Hth/Hox binding sites from the DCRE to create the RhoA-DF and RhoA-DR constructs. (B-D) Lateral view of Stage 10 embryos expressing the indicated LacZ reporter construct. Embryos were immunostained for β-gal (green), and Abd-A (magenta). (E) Schematic of RhoA variants with an additional 5 or 10 bp between the Pax2 and Exd-Hth-Hox binding sites. (F-H) Lateral view of Stage 10 embryos expressing the indicated LacZ reporter construct. Embryos were immunostained for β-gal (green) and Abd-A (magenta).
The orientation of the FoxG sites is critical for Abdominal Hox-mediated repression of the DCRE
The ability to re-engineer functional DCRE repression elements using distinct Hox binding site configurations in the GD-lacZ assay also provides an opportunity to assess if altering the orientation of FoxG sites similarly affects the ability of the DCRE to mediate transcriptional repression. Previous studies identified two other experimentally confirmed CRMs that are repressed by the Drosophila FoxG factors (Slp1 and Slp2): an even-skipped (eve) enhancer in the early embryonic ectoderm and a bagpipe (bap) enhancer in the embryonic visceral mesoderm [47-49]. Unlike the DCRE, both the eve and bap enhancers are unlikely to directly integrate Hox inputs, as Slp represses eve prior to Hox gene expression and Slp represses the bap enhancer in all segments [47-49]. Sequence comparisons between the DCRE, eve, and bap elements reveals each contains at least two FoxG binding sites, but in distinct orientations. The DCRE FoxG sites are in a head-to-head (HH) orientation, the eve FoxG sites are in a head-to-tail (HT) orientation, and the bap FoxG sites are in a tail-to-tail (TT) orientation (Fig 6A). To determine if the eve and bap FoxG binding sites can mediate repression in the 3xGBE assay, we replaced the native DCRE FoxG sites with those from bap and eve (Fig 6A). Comparative EMSAs using DCRE probes with either the bap or eve FoxG sites revealed purified Slp1 protein binds both sequences as well or better than the wild type FoxG sites (Fig 6B). In contrast, point mutations within the wild type FoxG sites (SlpM) weakens Slp1 binding to the DCRE (Fig 6B). Surprisingly, expression analysis of GD-lacZ embryos containing either the eve or bap FoxG sites revealed a significant loss of repression that was comparable to the GD-SIpM-lacZ embryos (Fig 6C-E and and6H).6H). Similar results were also seen when the FoxG sites from the eve enhancer were inserted in the reverse complement orientation of the DCRE relative to the Hox sites (EveRC, Fig 6F andand6H).6H). Importantly, comparative EMSA assays revealed that DCRE probes containing each FoxG binding site configuration retained the ability to bind the Abd-A/Exd/Hth complex, indicating that a loss of Hox binding cannot explain the loss of repression activity (Fig S7). Given that one of the differences between the FoxG sites in the DCRE versus the eve and bap enhancers is the orientation of binding sites, we re-engineered the bap FoxG sites into a Head-to-Head orientation within the DCRE (BapC) and found that it repressed as well as the wild type DCRE (Fig 6G-H). In all cases, similar behaviors were observed in the A1 segment that only expresses Ubx as compared to abdominal segments that express both Ubx and Abd-A. Taken together, these findings indicate that the orientation and spacing of the Hox and FoxG sites are critical for mediating robust abdominal transcriptional repression by Ubx and Abd-A in Slp+ cells.
(A) Schematics of DCRE variants: wildtype (DCRE-WT), weakened dFoxG sites (DCRE-SlpM), or differing dFoxG site compositions (DCRE-Eve, DCRE-Bap, DCRE-EveRC, DCRE-BapC). Orientations of the dFoxG binding sites are noted as: HH = head to head sites; TT = tail to tail sites; HT = head to tail sites. (B) EMSAs comparing two concentrations of Slp1 protein (50ngs and 250ngs) binding to DCRE-WT vs DCRE-SlpM as well as DCRE-WT vs DCRE-Eve and DCRE-Bap. (C-G) Lateral view of Stage 15 embryos expressing the indicated LacZ reporter constructs. Embryos were immunostained for β-gal (green), Slp2 (red), and Abd-A (blue). (C’-G’) Close-up view of T2-A2 segments from embryos in panels C-G. (C”-G”) Same as C’-G’ but only showing the β-gal stain. (H) Quantification of β-gal immunostain intensity among Slp2+ cells in segments T2-A4 relative to intensity in T3 segment of the indicated LacZ reporter embryos. Abdominal reporter expression was statistically compared between genotypes using ANOVA followed by post-hoc Tukey’s test. Prior to ANOVA, abdominal reporter expression in each individual embryo was mean averaged. Crosses mark statistically significant difference from G-LacZ (“n.s.” p >= 0.05, “†” p < 0.05, “††" p < 0.01, “†††" p < 0.001). Asterisks mark statistically significant difference from GD-LacZ (“n.s.” p >= 0.05, “*” p < 0.05, “**" p < 0.01, “***" p < 0.001). Additional pairwise comparisons are indicated by bars and hash-signs (“n.s.” p >= 0.05, “#” < 0.05, “##" p < 0.01, “###" p < 0.001).
Discussion
In this study, we interrogated how the Hox, Exd, and Hth binding site configurations within the DCRE and RhoA regulatory elements contribute to specific transcriptional outcomes. Our results indicate that: A) The DCRE mediates transcriptional repression over a relatively short distance; B) The Exd/Hth/Hox configuration of binding sites from RhoA, which recruit Abd-A/Hth/Exd complexes to mediate activation in abdominal SOP cells, can also recruit Abd-A Hox repression complexes when placed into the context of the DCRE – with the outcome dependent upon the presence of nearby TFBSs for either the Pax2 (activation) or FoxG factors (repression). C) While the Ubx Hox factor does not regulate activation via the RhoA Exd/Hth/Hox sites in the context of the RhoBAD enhancer, a Ubx/Exd/Hth Hox complex can bind to and repress the RhoA configuration of binding sites when inserted into the DCRE; D) The orientation and spacing of FoxG binding sites relative to the Hox sites is critical to mediate transcriptional repression in Slp2+ abdominal cells. In total, these findings reveal new insights into the TFBS grammar underlying how CRMs yield cell- and segment-specific outputs by recruiting specific Hox transcriptional complexes.
cis-regulatory grammar: The role of orientation and spacing between TFBSs in mediating proper transcriptional output.
cis-regulatory modules (CRMs) integrate transcriptional inputs to mediate cell-specific output. Current models of CRM function include the billboard, enhanceosome, and transcription factor (TF) collective [1]. The billboard model simply requires binding sites to be present within the CRM and their spacing/orientation/order has little effect on transcriptional outcomes [50]. Hence, transcription output is largely additive and the loss of any one binding site often has a modest impact on overall transcription levels and pattern. In contrast, the enhanceosome requires precisely spaced and oriented sites to mediate cooperative complex formation. Hence, the loss of any one site can disrupt both complex formation and synergistic output [51]. Lastly, the TF collective model similarly stipulates cooperative complex formation on DNA, but the arrangement of TFBSs needed to mediate cooperativity isn't highly constrained because protein-protein interactions between transcription factors can compensate for changes in DNA sequence [52]. CRMs have been identified that are representative of each model; the billboard model by the even-skipped enhancers in Drosophila [50], the enhanceosome model by the interfernon-β enhancer in mammalian cells [51], and the TF collective model by a set of Drosophila heart CRMs and neural enhancers in C elegans [52, 53]. Of course, these models are not mutually exclusive and CRMs can have aspects consistent with the flexibility associated with the billboard/TF collective models and the constraints associated with the enhanceosome model [39]. Here, we highlight how our current understanding of the integration of transcriptional inputs by the DCRE and RhoA elements reveals new insight into how TFBS organization affects Hox-mediated outputs in light of these three CRM models.
Prior studies revealed the DCRE encodes two Hox/Exd sites and one Hox/Hth site that contribute to abdominal Hox mediated repression via a mechanism largely consistent with the TF collective model of CRM function; i.e. numerous protein-protein and protein-DNA contacts between Exd, Hth, abdominal Hox factors, and the DCRE ensure robust complex formation and transcriptional outcomes [37]. For example, point mutations in any one Hox, Exd, or Hth site had only a modest effect on both DNA binding and transcriptional repression. Moreover, a Hth isoform that completely lacks its DNA binding domain can still mediate cooperative protein-protein interactions to form abdominal Hox complexes that repress the DCRE. Altogether, these findings indicate that multiple distinct Hox TF complexes can yield functional repression [54]. Consistent with these results, we found that the RhoA configuration of Exd/Hth/Hox sites can replace a set of Hox/Exd and Hth/Hox sites in the DCRE to mediate cooperative complex formation and robust transcriptional repression within the context of the DCRE. However, they do so in only one orientation relative to nearby FoxG binding sites. Moreover, we found that the spacing between the Hox and FoxG sites as well as the orientation of the FoxG sites contribute to transcriptional repression. Altogether, these data are congruent with the DCRE having aspects of both the TF collective and enhanceosome models of CRM function: while some different combinations of nearby Hox, Exd, Hth, and FoxG sites can mediate robust complex formation and appropriate outputs, not all combinations are functional indicating that the binding sites are at least somewhat constrained. This later point is also in agreement with the high degree of DCRE sequence and TFBS conservation observed between different Drosophilid species [37].
Comparative studies on the highly conserved RhoA element similarly revealed constrained TFBS configurations as well as additional TFBS affinity and competition features that are required for accurate cell- and segment specific output. Five transcription factors have been shown to directly impact RhoA output: Pax2 and Exd/Hth/Hox sites promote activation whereas an overlapping binding site for the Sens transcription factor represses gene expression in thoracic segments [27, 28]. We recently defined two key TFBS properties required for proper abdominal SOP output. First, the Pax2 and Sens binding sites are required to be low affinity, as a high affinity Pax2 site resulted in ectopic RhoA activity in additional SOP cells, whereas a high affinity Sens site resulted in repressed RhoA activity in all SOP cells [46]. Second, uncoupling the Sens and Pax2 sites so that they no longer overlap and compete for binding disrupted segment-specific RhoA output [46]. Both of these features provide insights into additional constraints on TFBS organization that have not been formally incorporated into the billboard, enhanceosome, or TF collective models. In this study, we further show that changing the spacing of the Pax2 site relative to the Exd/Hth/Hox sites disrupts activity in abdominal SOP cells; data that suggests enhanceosome-like activity. Like the DCRE, this idea is supported by the high degree of RhoA sequence conservation observed across numerous Drosophilid species in terms of both TFBS sequence and organization [46]. However, it should be pointed out that while Pax2 and Abd-A can interact in cells, cooperativity has not been detected between these factors on the RhoA element [28]. In addition, since all of the tested RhoA TFBS manipulations failed to reconstitute transcriptional activation in abdominal SOP cells, it is possible that the introduced sequence changes disrupt additional, unknown binding sites within the RhoA element required to mediate activation. Hence, future studies are needed to determine if the constraints on the Hox and Pax2 binding sites represent an enhancesome-like activity or are indicative of constraints consistent with other TFBS features such as overlapping and/or low affinity sites of additional transcription factor inputs.
Hox specificity: Target activation vs target repression
While genome-wide ChIP assays define where a transcription factor binds, they do not determine if the binding event mediates a change in regulatory activity. Hence, a fundamental question is: What defines whether a specific Hox/Exd/Hth complex activates, represses, or fails to regulate transcription once bound to DNA? Our comparative studies of the DCRE and RhoA elements show that the presence of additional TFBSs is critical to mediate appropriate output. For example, the RhoA CRM contains both a set of Exd/Hth/Hox binding sites and a Pax2 binding site to mediate Abd-A specific gene activation in abdominal SOP cells. However, if the RhoA Exd/Hth/Hox sites are moved into the DCRE, they can mediate abdominal repression by not only Abd-A but also the Ubx Hox factor in a manner dependent upon nearby FoxG binding sites. Hence, the RhoA configuration of Exd/Hth/Hox sites is neither an intrinsically activating configuration of sites nor is it strictly an Abd-A specific configuration of binding sites. Instead, Hox regulatory specificity is encoded and dependent upon the presence of adjacent binding sites within the CRM.
Given that both Pax2 and the Slp FoxG factors are expressed in all Drosophila segments and not exclusively in the abdomen, these tissue-specific factors must be selectively integrated with specific abdominal Hox factors via nearby DNA binding sites on the RhoA and DCRE elements, respectively. Consistent with this idea, previous studies demonstrated that Abd-A and Pax2 could be co-immunoprecipitated in cell culture, whereas a thoracic Hox factor (Antennapedia, Antp) that fails to activate RhoA failed to form complexes with Pax2 [28]. Moreover, the vertebrate Hox11 proteins were also found to form complexes with Pax2 to regulate target gene expression in the mammalian kidney [55]. While less is known about how Hox factors interact with Slp/FoxG, recent bimolecular-fluorescence (BiFC) assays in Drosophila found that Slp2 interacts with both Abd-A and Ubx in embryos, and Abd-A does so in a manner dependent on its ability to bind DNA [56]. Moreover, the thoracic Hox factor, Antp, failed to interact with Slp2 in BiFC assays, and we previously found that instead of mediating repression Antp stimulates the DMX leg enhancer in a DCRE-dependent manner via unknown mechanisms [37]. Altogether, these findings are consistent with the Pax2 factor selectively working with an Abd-A Hox complex to mediate gene activation and FoxG/Slp factors working with Ubx and Abd-A on the DCRE to mediate abdominal repression.
Acknowledgments
Funding:
This work was supported by a National Institutes of Health grant (GM079428) to BG and a National Instititutes of Health grant (GM063483) to AZ. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Bibliography
Citations & impact
Impact metrics
Citations of article over time
Alternative metrics
Article citations
Homeodomain complex formation and biomolecular condensates in Hox gene regulation.
Semin Cell Dev Biol, 152-153:93-100, 12 Dec 2022
Cited by: 2 articles | PMID: 36517343
Review
Mechanisms Underlying Hox-Mediated Transcriptional Outcomes.
Front Cell Dev Biol, 9:787339, 16 Nov 2021
Cited by: 11 articles | PMID: 34869389 | PMCID: PMC8635045
Review
Free full text in Europe PMC
Enhancers with cooperative Notch binding sites are more resistant to regulation by the Hairless co-repressor.
PLoS Genet, 17(9):e1009039, 24 Sep 2021
Cited by: 2 articles | PMID: 34559800 | PMCID: PMC8494340
Cell-type-specific Hox regulatory strategies orchestrate tissue identity.
Curr Biol, 31(19):4246-4255.e4, 05 Aug 2021
Cited by: 21 articles | PMID: 34358443 | PMCID: PMC8511240
Conserved Gsx2/Ind homeodomain monomer versus homodimer DNA binding defines regulatory outcomes in flies and mice.
Genes Dev, 35(1-2):157-174, 17 Dec 2020
Cited by: 16 articles | PMID: 33334823 | PMCID: PMC7778271
Go to all (6) article citations
Data
Data behind the article
This data has been text mined from the article, or deposited into data resources.
BioStudies: supplemental material and supporting data
Similar Articles
To arrive at the top five similar articles we use a word-weighted algorithm to compare words from the Title and Abstract of each citation.
A Hox Transcription Factor Collective Binds a Highly Conserved Distal-less cis-Regulatory Module to Generate Robust Transcriptional Outcomes.
PLoS Genet, 12(4):e1005981, 08 Apr 2016
Cited by: 34 articles | PMID: 27058369 | PMCID: PMC4825978
Hox repression of a target gene: extradenticle-independent, additive action through multiple monomer binding sites.
Development, 129(13):3115-3126, 01 Jul 2002
Cited by: 113 articles | PMID: 12070087
Antagonism versus cooperativity with TALE cofactors at the base of the functional diversification of Hox protein function.
PLoS Genet, 9(2):e1003252, 07 Feb 2013
Cited by: 20 articles | PMID: 23408901 | PMCID: PMC3567137
Cellular analysis of newly identified Hox downstream genes in Drosophila.
Eur J Cell Biol, 89(2-3):273-278, 16 Dec 2009
Cited by: 8 articles | PMID: 20018403
Review
Funding
Funders who supported this work.
NIGMS NIH HHS (2)
Grant ID: R01 GM079428
Grant ID: T32 GM063483
National Institutes of Health (2)
Grant ID: GM079428
Grant ID: GM063483