PBX1 and PBX3 have complementary expression patterns that delineate SHH expression in the FEZ

To begin, we analyzed expression of SHH relative to PBX1 and PBX3 by in situ hybridization in chick embryonic faces at HH22 (n = 6). At this developmental stage, the FEZ is active, and the FNP is undergoing patterned outgrowth to form the upper jaw. SHH is strongly expressed in the roof of the developing mouth (Fig 1A). At this same time, PBX1 expression overlaps the SHH expression domain (Fig 1B). In contrast, PBX3 is expressed in the globular process indicated by a red circle (Fig 1C) and the dorsal surface of the FNP tip just outside of the SHH expression domain, in a manner that appears to define the dorsal boundary of SHH expression in the FEZ (Fig 1C). Ontogeny of expression patterns of SHH, PBX1, and PBX3 between HH20 and HH23 is shown in S1 Fig. Over the time course, expression of SHH intensifies on the ventral surface of the FEZ covering the mouth roof, and expression of PBX1 also intensifies at the tip of the FNP. Expression of PBX3 at earlier stages appears along the tip of the FNP, similar to PBX1, but becomes confined to the globular process and the dorsal surface of the growing FNP tip (Fig 1C). This spatial pattern prompted us to assess whether PBX1 and PBX3 play complementary roles in regulating SHH expression in the FEZ and help delineate the boundaries of SHH expression in this region.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Spatial expression patterns of SHH, PBX1, and PBX3 in the normal chicken embryo FNP at HH22.

(A) At HH22, in situ hybridization reveals that SHH is strongly expressed in the chick FEZ on the ventral surface of the tip of the growing FNP (yellow arrow). (B) PBX1 is co-expressed with SHH throughout the FEZ, and there is strong expression at the tip of FNP (yellow arrow). (C) In contrast, PBX3 is expressed in the globular process (red circle) and the dorsal surface of the tip of the FNP (red arrow); these regions define the boundaries of SHH in the FEZ. n = 6; Scale bar, 500 μm.

https://doi.org/10.1371/journal.pgen.1011315.g001

PBX1 and PBX3 exert opposite effects on SHH expression in the FEZ

To assess the extent to which PBX1 and PBX3 may regulate the pattern of SHH expression in the FEZ, we used both gain- and loss-of-function approaches. First, we over-expressed each gene using a replication competent retroviral vector (RCAS) encoding PBX1 or PBX3 (RCAS-PBX1 and RCAS-PBX3; Fig 2A, 2B, 2F, and 2G). Second, we knocked down each gene using an RCAS vector that encodes a microRNA targeting either PBX1 or PBX3 transcripts (RCAS-mirPBX1 and RCAS-mirPBX3; Fig 2C, 2D, 2H, and 2I). In both experiments RCAS-AP (expressing alkaline phosphatase) was used as a control for retroviral infection and spread (Fig 2E and 2J). Embryos (n = 12) were infected at HH10 by injecting viral supernatant (titer, 10⁸) into the mesenchyme adjacent to the anterior neural tube. This time point was chosen as it precedes the onset of SHH expression and leads to widespread infection of the face. Embryos were allowed to develop for additional 72 hours (~HH22) and then collected for whole mount in situ hybridization and real-time polymerase chain reaction (quantitative PCR; qPCR). Overexpression of PBX1 led to upregulation of SHH expression and expansion of the SHH expression domain (Fig 2A), while knockdown of PBX1 reduced SHH expression in the FEZ (Fig 2C). In contrast, overexpression of PBX3 led to downregulation of SHH in the FEZ as well as facial malformations including a smaller head (Figs 2B and S2 Fig), and knockdown of PBX3 led to an expansion of SHH expression and premature/ectopic expression of SHH in the globular process (Fig 2D; red arrow). At the same time, both overexpression and knockdown of PBX1 or PBX3 did not affect SHH expression in the brain (Fig 2F–2I). Expression of SHH in the FEZ from each treatment group was quantified by qPCR, and the results were concordant with the in situ hybridization data (Fig 2K). Both RCAS-AP and RCAS-mirLacZ controls did not affect SHH expression compared to the normal control (Fig 2K). We confirmed with qPCR that RCAS-mirPBX1 decreased expression of PBX1 and RCAS-mirPBX3 decreased expression of PBX3 (Fig 2L and 2M). Similar to previous observations [27], where loss-of-function experiments utilizing conventional shRNAs with a chicken U6 promoter resulted in knockdown efficiency ranging 26 – 47% in chicken ovary-derived cell culture, our in vivo experiments showed knockdown efficiency from RCAS-mirPBX1 at ~27% and from RCAS-mirPBX3 at ~45% (Fig 2L and 2M). Although the efficiency was relatively low, likely due to mosaicism or suboptimal in vivo electroporation, repression of PBX1 and PBX3 significantly altered expression of SHH compared to the normal controls (Fig 2K). To ensure that the gene expression changes did not result from apoptosis after RCAS-miRNA virus infection, we used terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assays in chick embryonic faces and determined that no cell death was apparent 24 hours after infection (~HH14/15, n = 6; S3 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Expression of SHH in the FEZ after RCAS-virus infection to induce overexpression and knockdown of PBX1 and PBX3.

RCAS-PBX1, RCAS-PBX3, RCAS-mirPBX1, RCAS-mirPBX3, and RCAS-AP/RCAS-mirLacZ control viruses were used to infect embryos at HH10. After 72 hours (HH22), the resulting expression patterns of SHH were evaluated by whole mount in situ hybridization (in the FNP, A-E; in the brain, F-J), and relative expression levels of SHH in the FEZ were quantified by qPCR (K). (A) Overexpressing PBX1 expanded expression domains of SHH within the FEZ. The green arrow shows ventral extent of the expanded SHH domain. (B) Overexpressing PBX3 and (C) repressing PBX1 reduced expression of SHH in the FEZ. (D) Repressing PBX3 induced ectopic expression of SHH (red arrow). (E) RCAS-AP control embryo illustrates that the SHH expression domain is in the roof of the mouth. The green arrow points to the ventral extent of the normal expression domain of SHH in the FEZ. (F-J) In contrast, altering PBX1 and PBX3 expression in the face did not affect SHH expression in the brain. n = 12; Scale bar, 1 mm. (K) Relative expression levels of SHH in the FEZ after overexpressing PBX1 (n = 6); overexpressing PBX3 (n = 3); repressing PBX1 (n = 6); and repressing PBX3 (n = 3) compared to the normal control (n = 6). RCAS-AP (n = 6) or RCAS-mirLacZ (n = 6) treatments did not affect SHH expression compared to the normal control. In the FNP (including the mesenchyme and the ectoderm), RCAS-mirPBX1 (n = 6) decreased PBX1 expression compared to the normal control (n = 6; L), and RCAS-mirPBX3 (n = 6) decreased PBX3 expression compared to the normal control (n = 4; M). Statistical significance was determined at P-value < 0.05; *, P < 0.05; **, P < 0.01; ****, P < 0.0001; ns, not significant.

https://doi.org/10.1371/journal.pgen.1011315.g002

Together, these data suggest that PBX1 is a transcriptional activator and PBX3 is a transcriptional repressor of SHH expression in the FEZ. However, whether PBX1-PBX3 bind DNA directly, or within a complex with cofactors, to modulate SHH transcription remained unexplored and became the focus of the rest of the study.

Profiling epigenomic landscapes at the SHH locus by high-throughput sequencing

We assessed whether PBX1 and PBX3 directly participate in transcriptional regulation of SHH expression in cells comprising the FEZ at HH22 (Fig 3A). First, we performed genome-wide assays for transposase-accessible chromatin using sequencing (ATAC-Seq, n = 2) to profile open chromatin regions near the SHH locus. Second, we performed chromatin immunoprecipitation followed by sequencing (ChIP-seq, n = 2) to identify PBX1 and PBX3 binding to DNA genome-wide and specifically to the SHH locus. We then intersected these datasets to identify regions of open chromatin that were bound by PBX1 and/or PBX3.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Genome-wide analyses on molecular mechanisms in the chick FEZ at HH22 by ATAC-seq and ChIP-seq targeting PBX1 and PBX3.

The high-throughput sequencing data were generated from cells comprising the chick FEZ at HH22 (n = 2). (A) This illustrates the region of the embryo where the ectoderm was used for data generation. (B) The KEGG pathway analysis [32,33] was conducted on differentially bound PBX1 and PBX3 peaks (adjusted P-value < 0.05). (C) Known motifs were identified from the open chromatin profiles (ATAC-seq peaks with IDR < 0.05), and both known motifs (D, E) and de novo motifs (F, G) were analyzed from the PBX1 and PBX3 binding profiles (PBX1 and PBX3 ChIP-seq peaks with IDR < 0.05) by HOMER [34]. (H) Identified ATAC-seq and ChIP-seq peaks (simple overlaps between biological replicates; overlapping peaks that have IDR < 0.05 between biological replicates; and enriched binding sites comparing PBX1 and PBX3 peak profiles) and overall peak profiles are visualized on the IGV viewer [77]. The yellow rectangle indicates the 400 bp PBX1-enriched locus (chr2:8,173,222-621). (I) Three putative MEIS-PREP/PBX1-PREP binding sites [35,36] (colored with magenta, CCCGCCC, and blue, GGCGG; located at chr2:8,173,380-386; chr2:8,173,425-429; and chr2:8,173,497-503) on the 400 bp PBX1-enriched peak (colored with red; referred to as “SFE1”) which is shown in the yellow rectangle in Fig 3H. (J) The 400 bp SFE1 region within SHH (chr2:8,173,222-621) contains some of the top ranked motifs that were discovered from the ChIP-seq dataset. (K) Two regions within SFE1 (SFE1_1 and SFE1_2) and the promoter region of PBX1 were amplified by PCR from the ChIP-seq libraries targeting PBX1 and IgG as well as the input control. The final products were run through a 2% agarose gel along with the 100 bp DNA ladder. (L) The PCR products amplified from the ChIP-seq PBX1 libraries with SFE1_1 and SFE1_2 primers were confirmed by DNA sequencing. SFE1_1 primer targets the first CCCGCCC motif within SFE1 (SFE1A) and SFE1_2 primer targets ~80 bp upstream of SFE1A.

https://doi.org/10.1371/journal.pgen.1011315.g003

First, ATAC-seq profiles were used to evaluate open chromatin. For these datasets, the percentage of reads mapped onto the reference genome (mapping rate) were greater than 95%, and the number of unique reads (sequencing depth) ranged from 585,502,859 – 707,110,287 reads (S1 Table). Open chromatin peak profiles from two biological replicates resulted in 65,556 peaks that have an Irreproducibility Discovery Rate (IDR) [28] less than 0.05 (S1 Table). These open chromatin regions were used for subsequent intersections with ChIP-seq datasets.

Next, we performed ChIP-seq also on the FEZ tissue at HH22 using specific antibodies directed against PBX1 and PBX3. To verify the efficiency and specificity of our ChIP-seq experiments and of each antibody used, the ChIP-seq libraries were assessed by qPCR. We amplified the promoter regions of PBX1, PBX3, and MEIS2 that contain known PBX1/3 target sites as positive controls [29], as well as two random regions on different chromosomes that do not contain PBX1 or PBX3 consensus sites and were therefore used as negative controls. The ChIP-seq libraries targeting PBX1 and PBX3 amplified the target regions of PBX1 and PBX3, respectively, as well as their traditional partner, MEIS2, while the random regions in both the libraries resulted in minimal amplification (S4 Fig). For the ChIP-seq data, mapping rates ranged from 88.56 – 98.18%, and the sequencing depth ranged from 25,146,384 – 90,585,798 reads (S1 Table). PBX1 peak profiles from two biological replicates resulted in 13,430 overlapping peaks (IDR < 0.05), and PBX3 peak profiles from two biological replicates resulted in 36,122 overlapping peaks (IDR < 0.05; S1 Table). PBX1 and PBX3 peaks (IDR < 0.05) were then analyzed to identify differentially bound sites by DiffBind [30,31]. Differentially bound peak profiles of PBX1 and PBX3 underwent the KEGG pathway enrichment analysis [32,33], and each PBX protein’s binding profiles resulted in its own exclusive overrepresented pathways (Fig 3B). From their top 12 overrepresented biological processes, PBX1 and PBX3 binding profiles shared seven biological processes, including regulation of actin cytoskeleton, focal adhesion, FoxO signaling pathway, and TGF-beta signaling pathway. The other pathways were exclusive to either PBX1- or PBX3-specific binding signatures, including cell cycle, mitophagy, apoptosis, and cellular senescence for PBX1 profiles; and cytokine-cytokine receptor interaction, MAPK signaling pathway, and WNT signaling pathway for PBX3 profiles.

Sequentially, we conducted motif analysis on ATAC-seq peaks with IDR < 0.05 by Hypergeometric Optimization of Motif EnRichment (HOMER [34], version 4.11) to identify known motifs. The top three motifs were AYAGTGCCMYCTRGTGGCCA (CTCF, CCCTC-Binding Factor), CNNBRGCGCCCCCTGSTGGC (BORIS, CCCTC-Binding Factor Like), and GKVTCADRTTWC (SIX1, SIX homeobox 1; Fig 2C). PBX3 motif (SCTGTCAMTCAN) ranked 17^th, and PBX1 motif (GSCTGTCACTCA) ranked 39^th (S2 Table). Also, the motif analysis was conducted on PBX1 and PBX3 peaks with IDR < 0.05 to discover both known and de novo motifs (Fig 2D–2G). The top three known motifs from PBX1 peak profiles were VGCTGWCAVB (MEIS1, Myeloid Ecotropic Integration Site 1), YTGWCADY (TGIF1; TGFB Induced Factor Homeobox 1), and SCTGTCAMTCAN (PBX3). The top three known motifs from PBX3 peak profiles were YTGWCADY (TGIF1), VGCTGWCAVB (MEIS1), and TGTCANYT (TGIF2). The top three de novo motifs discovered from PBX1 peak profiles were NRNCTGWCAG, TGATTGRCNG, and RDRGGMGGDR. The top three de novo motifs discovered from PBX3 peak profiles were NCTGWCAGNH, TGATTGRCGG, and NAGGAATGYG. Notably, PREP1/PKNOX1 (PBX regulating protein 1, PREP1; also called PBX/Knotted Homeobox 1, PKNOX1) motif was the second best match for de novo motif discovery from both PBX1 and PBX3 ChIP-seq datasets. The full lists of motif discovery are shown in supplemental tables: ATAC-seq data (known motif discovery, S2 Table), ChIP-seq data targeting PBX1 (known motif discovery, S3 Table; de novo motif discovery, S4 Table), and ChIP-seq data targeting PBX3 (known motif discovery, S5 Table; de novo motif discovery, S6 Table).

Finally, we intersected the replicated peaks from the ATAC-seq and ChIP-seq datasets. All PBX1 and PBX3 binding sequences on the SHH locus are located within intronic regions and span ATAC-seq peaks (Fig 3H). In intron 1, PBX1 and PBX3 are bound to the same region, suggesting a potential competitive binding within that DNA fragment. Among PBX1 and PBX3 differentially bound sites, one 400 bp long sequence was identified as a PBX1-enriched region within the first intron of SHH (chr2:8,173,222-621; Fig 3H, yellow rectangle, and Fig 3I). Within this 400 bp PBX1-enriched sequence, the TGIF2 motif (the third known motif from the PBX3 binding profile) was found once (Fig 3J). The third de novo motif from the PBX1 binding profiles (RDRGGMGGDR, best match POLII – GC box) was found four times within the 400 bp PBX1-enriched sequence (Fig 3J). This third de novo motif from the PBX1 peak profile and the second de novo motif from the PBX3 peak profile (TGATTGRCGG) contain GGMGG and GRCGG sequences that match putative MEIS-PREP/PBX1-PREP binding sites [35,36] (CCCGCCC and GGCGG). These sequences appear at three regions within the 400 bp PBX1-enriched fragment identified within intron 1 of SHH (referred to as “SFE1” for SHH FEZ Enhancer 1; Fig 3H and 3I). Altogether, these genome-wide high-throughput sequencing data suggest potential interactions between PBX proteins and MEIS-PREP complexes at this 400 bp sequence within intron 1 of SHH in the chick FEZ at HH22.

Since SFE1 does not have the traditional PBX binding motif (TTGATTGAT), we reasoned that the GGCGG/CCCGCCC motif might be a potential alternative PBX binding motif within this fragment. To provide biochemical evidence that PBXs bind to the GGCGG/CCCGCCC motifs within SFE1, we conducted PCR assays from the ChIP-seq libraries targeting PBX1, using two primers around the first CCCGCCC motif (SFE1A) within SFE1. Both primer sets amplified the SFE1 fragments around the first CCCGCCC motif from the ChIP-PBX1 libraries, pointing to interaction of PBX1 with this DNA fragment (Fig 3K). We confirmed that SFE1 sequence is present in the PCR products by DNA sequencing (Fig 3L). To further assess whether PBX proteins bind to the three GGCGG/CCCGCCC motifs within SFE1, we conducted electrophoretic mobility shift assay (EMSA) with three 25 bp biotinylated sequences that contain each of the GGCGG/CCCGCCC motifs within SFE1 (SFE1A, B, and C; S5A Fig). Due to the lack of chicken PBX1 recombinant protein availability, we overexpressed PBX1 in DF1 cell culture (chicken fibroblast) by the RCAS-virus system and extracted nuclear proteins. When PBX1 was over expressed (RCAS-PBX1), SFE1A and SFE1C resulted in strong shifted bands while SFE1B did not show protein-DNA binding (S5B Fig), suggesting that the interaction between PBX proteins and SFE1 requires the entire 7 bp CCCGCCC motif. These results further confirm that PBX1 binds to this element (within the SF1A and SF1C fragments) either directly or indirectly, by forming a complex with other transcription factor partners, potentially PREP/MEIS [35,36] or other transcription factors that may interact with the CCCGCCC motif.

Activity of the regulatory sequence within the SHH locus

The 400 bp PBX1-enriched sequence (chr2:8,173,222-621) containing multiple PBX motifs became our region of interest. To test whether this 400 bp genomic sequence has enhancer activity, we cloned a 1.8 kb (chr2:8,172,590-8,174,390) fragment that included our region of interest (Fig 4A) into the TKP-LacZ vector. This vector has a minimal promoter with no basal reporter (LacZ) activity and can be used to assess enhancer activity via electroporation into chick embryos [3]. This construct was electroporated into the ectoderm covering the FNP at HH20 (n = 20). After 24 hours (HH24), this fragment exhibited transcriptional activity in the SHH expression domain of the FEZ suggesting that this element has cis-regulatory functions (Fig 4B). Then, we repeated the experiments with a narrower, 600 bp fragment containing the 400 bp sequence (n = 15; Fig 4C), and the transcriptional activity was still detected in the SHH expression domain of the FEZ (Fig 4C). Also, both the 1.8kb fragment and the narrower 600 bp fragment containing intact SFE1 showed transcriptional activities outside of the SHH expression domain in the FEZ (Fig 4B and 4C). This could result from one or a combination of the following reasons. First, these fragments are isolated sequences and out of context of the regulatory regions that control SHH expression. Second, the transcriptional activity might also localize to areas where activators that interact with SFE1, such as PBX1, are expressed, even outside of the SHH-expression domain. Indeed, PBX1 can form complexes with PREP and MEIS and its interaction with the three putative MEIS-PREP/PBX1-PREP binding sites (chr2:8,173,380-386; chr2:8,173,425-429; and chr2:8,173,497-503; Fig 3I) might occur within SFE1. Therefore, X-gal staining could also localize to additional domains where the genes encoding PREP and MEIS are expressed. Notably, our experiments show that both the deletion (SFE1 null, n = 10; Fig 4D) and the mutation (SFE1 mutation, n = 10; Fig 4E) of the three putative PBX binding sites from SFE1 eliminated all transcriptional activities, suggesting that the X-gal reporting activity is driven by PBX, either alone or in a complex. The X-gal staining patterns outside of the FEZ were broader from the experiments using the 1.8kb fragment (Fig 4B) compared to the experiment employing the shorter 600 bp fragment (Fig 4C). This finding suggests that the 1.8kb fragment might contain additional regulatory elements that could activate transcription outside of the FEZ. To demonstrate the efficiency of electroporation into the facial epithelium, a heat shock protein promoter (HSP68-LacZ) was used as a positive control and displayed non-specific X-gal staining patterns in the ectoderm (Fig 4F–4H), concordant with our previous observations. A TKP-LacZ vector lacking any cloned elements was used as a negative control and resulted in no basal expression (Fig 4I).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. In vivo functional assays for the potential regulatory sequence regulating SHH expression.

(A) A 1.8kb fragment (chr2:8,172,590-8,174,390) and a narrower, 600 bp fragment (Middle 600 bp; chr2:8,173,096-688) containing the PBX1-enriched sequence (SFE1; 400 bp, chr2:8,173,222-621) were cloned into the TKP-LacZ vector and electroporated into the chick FEZ at HH20. (B) After 24 hours (HH24), X-gal staining showed that the 1.8kb fragment exhibited transcriptional activity in the SHH expression domain of the FEZ (n = 20). (C) The narrower fragment (Middle 600 bp) also activated transcription in the SHH expression domain (n = 15). Transcription was not activated when three putative PBX binding sites in SFE1 were either deleted (D; n = 10), or mutated (E; n = 10). (F, G, H) The positive control (HSP68-LacZ) displays broad X-gal staining patterns in the area where the electroporation was performed (n = 20). (H) Sagittal section from positive control embryo shows that the electroporation was confined to the facial ectoderm. (I) A TKP-LacZ vector containing no cloned fragment was electroporated as the negative control (n = 20). Scale bars for A-G and I, 1 mm; Scale bar for H, 200 µm.

https://doi.org/10.1371/journal.pgen.1011315.g004

In parallel to the in vivo experiments, we conducted luciferase assays in vitro to confirm the transactivating function of the minimal region with enhancer activity. We divided the 1.8kb segment (chr2:8,172,590-8,174,390) into three 600 bp fragments (the 3’ end, 3’-600 bp; the middle fragment containing SFE1, Middle 600 bp; the 5’ end, 5’-600 bp; Fig 5A). Each of these fragments were cloned into a luciferase expression vector and transfected into DF1 cells along with RCAS-PBX1 (3’-600 bp, n = 7; Middle 600 bp, n = 6; 5’-600 bp, n = 4; triplicate technical replicates). Luciferase intensity of each experimental group was normalized to the RCAS-AP controls, of which the luciferase expression was normalized to 1 (shown as a red dashed line in Fig 5B and 5C). Luciferase activity was only detected when the fragment containing SFE1 (Middle 600 bp; chr2:8,173,096-688) was inserted (P < 0.001; Fig 5B). Accordingly, we made one construct completely deleting all three binding motifs within SFE1 (Middle 600 bp SFE1 null) and a second construct where we mutated all three binding motifs (Middle 600 bp SFE1 mutation; n = 3/each). Both deletion and mutation of the putative binding sites significantly decreased luciferase activity compared to the intact 400 bp element, SFE1 (P < 0.001) and resulted in no significant luciferase activity compared to the RCAS-AP controls (P > 0.05; Fig 5C). The results from both the in vivo and the in vitro experiments indicate that PBX1 physically interacts with all or at least one of the three binding motifs.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. In vitro functional assays for the potential regulatory sequence regulating SHH expression.

(A) The 1.8kb sequence (chr2:8,172,590-8,174,390) containing the 400 bp PBX1 enriched sequence (chr2:8,173,222-621) were divided into three 600 bp segments; the 3’ end; the middle (containing SFE1); and the 5’ end. (B) The three 600 bp segments were individually transfected (3’-600 bp, n = 7; Middle 600 bp, n = 6; 5’-600 bp, n = 4; triplicate technical replicates) along with a luciferase reporter construct and a PBX1 expression vector (RCAS-PBX1). Compared to the RCAS-AP controls (normalized to the intensity of 1; red dashed line), the only insert resulted in increased luciferase intensity was the middle fragment (chr2:8,173,096-688), which contains the 400 bp PBX1 enriched region (SFE1). (C) Three PBX consensus binding sites in SFE1 (chr2:8,173,380-386; chr2:8,173,425-429; and chr2:8,173,497-503; Fig 3I) were deleted or mutated to eliminate the putative PBX binding sequences (n = 3/each). These fragments were transfected into cell cultures with a luciferase reporter construct and a PBX1 expression vector (RCAS-PBX1). Both simple deletion (SFE1 null) and mutation (SFE1 mutation) resulted in no significant luciferase activity compared to the RCAS-AP controls (red dashed line). (D) Fluorometric β-galactosidase assay was conducted to overcome the low basal expression of luciferase reporter for the RCAS-PBX3 experiment. The intact middle 600 bp fragment and SFE1 null fragment were cloned into the HSP68-LacZ vector and transfected into DF1 cell cultures with RCAS-PBX3 (n = 4; triplicate technical replicates). Overexpression of PBX3 reduced β-galactosidase fluorescent intensity of the HSP68-LacZ vector with intact SFE1 by half compared to the RCAS-AP controls (normalized to the intensity of 1; red dashed line). On the other hand, when the putative PBX binding motifs were deleted (SFE1 null), overexpression of PBX3 did not change β-galactosidase fluorescent intensity compared to the RCAS-AP controls (P = 0.47). Statistical significance: **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

https://doi.org/10.1371/journal.pgen.1011315.g005

In order to confirm the repressive role of PBX3 on SFE1, we transfected DF1 cells with the Middle 600 bp fragment cloned in a luciferase reporter construct with either PBX3 (n = 9) or mirPBX3 (n = 5) expression vectors. While repression of PBX3 using RCAS-mirPBX3 increased luciferase activity almost two-fold (P < 0.0001; S6 Fig), overexpression of PBX3 (RCAS-PBX3) did not decrease intensity of luciferase expression (P > 0.05; S6 Fig), possibly due to lower basal expression of the luciferase constructs. To overcome the minimal basal expression of the luciferase reporter, we then individually cloned the SFE1 fragment and the SFE1 null fragment into the HSP68-LacZ vector that has strong basal expression and transfected it into DF1 cells along with RCAS-PBX3 (n = 4; triplicate technical replicates). Fluorometric β-galactosidase assays demonstrated that when PBX3 was overexpressed (RCAS-PBX3) the HSP68-Lacz vector with intact SFE1 (Middle 600 bp) resulted in significantly reduced β-galactosidase fluorescent intensity compared to the RCAS-AP controls (shown as a red dashed line; P < 0.0001), while deletion of the three GGCGG/CCCGCCC motifs within SFE1 (Middle 600 bp SFE1 null) did not change the reporter expression compared to the RCAS-AP controls (P = 0.47; Fig 5D). Together, these data confirm that PBX3 acts as a repressor of SHH when it binds to SFE1, while PBX1 activates transcription from this element.

SHH as a key molecular regulator in the FEZ and PBX1/3 as potential regulators of SHH expression

The FEZ was first identified in the chicken embryo based on the observation of a transient boundary between SHH and FGF8 expression domains in the ectoderm covering the FNP [10]. In the chick FEZ, SHH expression begins around HH20 [1,2] at which time it forms a boundary with a previously established domain of FGF8 expression. At this time, FGF8 expression becomes down-regulated across the midline of the FNP and restricted to the nasal pits. In contrast, SHH expression is maintained in the FEZ for a long period of time and is required for the FEZ to function. This has been observed in other amniote embryos as well [7,11]. Interestingly, our data revealed that PBX1 was co-expressed with SHH in the FEZ, while PBX3 was expressed at the boundary of the SHH expression domain. By using gain- and loss-of-function approaches in the chick, we demonstrated that PBX1 induces and PBX3 represses SHH transcription. Further, these opposite effects on SHH transcription appear to be mediated by a regulatory element in an intronic region of SHH that acts as a switch. The same element behaves as an enhancer when associated with PBX1, and as a repressor when associated with PBX3. Hence, PBX1 and PBX3 play complementary roles in the regulation of SHH expression in the FEZ during early stages of its function.

PBX genes in development

The PBX gene family comprises four transcription factors that have essential roles in embryonic development. PBX1, the gene encoding the Pre-B-cell leukemia homeobox transcription factor 1, is the mammalian homologue of the extradenticle (exd) gene in Drosophila melanogaster and shares over 70% identity to exd [37–39]. PBX1 encodes a homeodomain transcription factor of the three amino acid loop extension (TALE) family. The PBX1 protein and its related family members PBX2–4 dimerize with other TALE class homeodomain proteins from the MEIS and PREP families through a PBC domain to form nuclear complexes that can enhance the binding specificity of HOX proteins to DNA, as demonstrated in some embryonic contexts (reviewed by Moens and Selleri [29]). In the mouse, PBX-PREP or PBX-MEIS complexes regulate target genes that control segment identity and organ patterning during embryogenesis [29,40–42]^. Indeed, the discovery from the present motif analysis supports that PBX1 and PBX3 binding patterns within the SHH locus in the FEZ at HH22 are substantially associated with binding of PREP and MEIS complexes. Also, these transcription factors can act upstream of Hox genes, or even independently of HOX proteins in collaboration with other cofactors, in various developmental processes [29]. In addition to forming heterodimers with HOX proteins and with TALE partners MEIS/PREP, PBX proteins can form multimeric complexes with other transcription factors, such as MYOD, EN, PDX1 (reviewed by Selleri et al. [43]) and HAND2 [44]. Lastly, PBX proteins can regulate transcription by interacting with basic transcription regulators, such as histone acetyltransferases (HATs) and CBP coactivators, and with histone deacetylases (HDACs) and the corepressor N-CoR/SMRT (reviewed by Selleri et al. [43]). Thus, PBX transcription factors can be part of activating or repressing transcriptional complexes.

In mice Pbx1 is widely transcribed throughout the developing embryo, and Pbx1-deficient embryos that are homozygous for a null allele (Pbx1−^/−) develop pleiotropic developmental defects. These abnormalities include perturbed morphogenesis, severe hypoplasia, or aplasia of multiple organ systems, including the nervous system, craniofacial skeleton, ear pinnae, branchial arch-derived structures, limbs, heart, hematopoietic system, lungs, diaphragm, liver, stomach, gut, pancreas, spleen, kidneys, and gonads [40,45–49]. Heterozygous mice are viable and fertile but are smaller in size than wild type mice [40]. Overall, our previous studies have established that transcription factors of the PBX family are critical developmental regulators. Accordingly, Pbx genes play critical roles during morphogenesis of the murine midface, especially in the FNP epithelium, which expresses Pbx1/2/3 transcripts. Interestingly, PBX transcription factors are essential in the cephalic epithelium but dispensable in the mesenchyme for the proper fusion of the upper lip and primary palate [26]. The Pbx2 gene is absent from the avian genome suggesting that there may be differences in the roles of members of this gene family among organisms. Evolutionary-developmental biology approaches will be needed to elucidate potential unique roles of these transcription factors in different organisms.

Epigenomic landscapes in the FEZ and a regulatory element within the SHH locus

Regions of high sequence conservation across species and with elevated DNaseI hypersensitivity [50,51] usually correspond to accessible regions of the genome that are functionally related to transcriptional activity, since this remodeled state is necessary for the binding of proteins such as transcription factors. Accessible regions have also been shown to comprise many types of cis-regulatory elements including promoters, enhancers, silencers, and locus control regions. In the present study, we assessed open chromatin sites from the FEZ of chick embryos at HH22 using ATAC-seq. We combined the obtained dataset with results from the ChIP-seq assays that profiled genome-wide binding of PBX1 and PBX3 to DNA. These combined high-throughput analyses enabled us to: 1) broadly demarcate the chromatin landscape during FEZ morphogenesis in the chick embryo; and 2) locate a candidate regulatory element within the SHH locus that is bound by PBX proteins and controls facial development by regulating SHH expression.

In the epigenomic landscapes, all PBX1- and PBX3-bound sites within the SHH locus were located at the open, accessible chromatin regions uncovered by ATAC-seq peaks (Fig 3H). Both PBX1 and PBX3 bound to the same region in intron 1 of the SHH locus. However, the differentially bound peak analysis determined that PBX1 enrichment at the 400 bp element is greater than PBX3 binding (Fig 3H), suggesting that this 400 bp sequence, SFE1, is a candidate regulatory region that is bound preferentially by PBX1. Indeed, enhancer activity was observed in vivo by electroporating a reporter construct containing SFE1 (Fig 4B and 4C), and this was also observed in in vitro  luciferase reporter assays (Fig 5B). In addition, the reporter assay data also suggest preferential binding of PBX1 to this sequence. Overexpression of PBX3 in cultured cells transfected with the PBX1-bound element cloned upstream of the luciferase reporter neither increased nor decreased luciferase intensity (S6 Fig). This result may indicate that 1) the luciferase assay was simply not sensitive enough to detect reduced expression and/or 2) expression did not change by the additional PBX3 over expression in the presence of endogenous PBX1 proteins. In other words, increased levels of PBX3 proteins did not overcome the binding competition by endogenous PBX1 factors. However, downregulation of PBX3 in this cell culture system increased the luciferase intensity (S6 Fig), indicating that expression was enhanced when competition between PBX1 and PBX3 was altered by decrease of PBX3 protein levels. Reduced PBX3 expression appeared to allow greater enrichment of PBX1 binding to this regulatory sequence, increasing in turn the expression of luciferase. The repressive role of PBX3 was further confirmed by the fluorometric β-galactosidase assay (Fig 5D).

To elucidate the mechanisms of interaction between PBX1/3 and SFE1, we located putative PBX binding sequences within SFE1. As SFE1 does not contain the traditional PBX binding motif (TTGATTGAT) [52,53], we sought other possible motifs that would interact with PBXs within SFE1. As a result, an alternative PBX binding motif, GGCGG/CCCGCCC [35,36], was found three times within SFE1. Our data derived from ChIP-seq, ChIP-PCR and EMSA suggest that PBX1 and PBX3 interact with SFE1 possibly through this 7 bp CCCGCCC motifs either directly or indirectly. Specificity of transcription factor-DNA binding is determined by numerous factors, such dimer orientation, spacing preferences, base-stacking interactions, structural DNA features, shape-based DNA recognition, and formation of complexes with other cofactors that may confer higher binding specificity [54,55]. Therefore, we cannot unequivocally prove direct binding of the CCCGCCC motif by PBXs. Indeed, PBX proteins may directly bind to this motif or bind indirectly via complexes with their traditional partners PREP/MEIS [35,36], or other transcription factors. Interestingly, the GC box sequence (GGGCGG) is comprised of known binding sequences for SP2, suggesting that SP2 may be a potential binding partner [56]. Additionally, we cannot rule out the possibility that overexpression or repression of PBXs might also alter expression of other transcription factors that could ultimately affect activation of SFE1. Further research with specific antibodies against PBX1, PBX3, PREP, and SP2 will identify the transcription factor that directly contacts DNA within the SFE1 fragment via the CCCGCCC motif.

PBX mutations are associated with craniofacial birth defects in humans

There is significant clinical relevance in understanding the roles of PBX transcription factors in regulating facial morphogenesis. Indeed, we reported in a previous study eight patients with craniofacial dysmorphology who had de novo, deleterious sequence variants in the PBX1 gene [57]. The cases exhibited varying expressivity and severity of facial dysmorphology, as well as other affected organ systems. The sequence variants in these cases included missense substitutions adjacent to the PBX1 homeodomain or within the homeodomain, and mutations yielding truncated PBX1 proteins. Functional studies on five PBX1 sequence variants revealed perturbation of intrinsic, PBX-dependent transactivation ability and altered nuclear translocation, suggesting abnormal interactions between mutant PBX1 proteins and wild-type TALE or HOX cofactors. These mutations may directly affect transcription of PBX1 target genes to impact development [57]. Also, genomic analyses revealed evidence of gene-gene interactions between the human PBX1/2 genes and SNPs in the ARHGAP29 locus, a candidate for CL/P [58].

In the present study, using the chick embryo, we explored the specific roles of PBX1 and PBX3 in regulating FEZ morphogenesis by controlling SHH expression. In mammals, Pbx2 may compensate for some aspects of Pbx1 loss, thus confounding the potential effect of Pbx1 loss on Shh expression [26,59,60]. As discussed above, birds do not appear to have the PBX2 gene, making the chick embryo an ideal model system to address the functions of PBX1 and PBX3. With the complementary expression patterns of PBX1 and PBX3 around the SHH domain in normal chick embryos at HH22, we examined how altered PBX1/3 expression prior to the onset of SHH expression would affect the spatial SHH domain in the FEZ by HH22. By gain-of-function and loss-of-function experiments in the chick, we showed that over-expressing PBX1 resulted in increased and expanded SHH expression, while repressing PBX1 decreased SHH expression in the FEZ, suggesting that PBX1 is a positive regulator of SHH transcription in the FEZ. We can envision that PBX1 binding to the 400 bp SHH regulatory element identified in this study could play: 1) an ‘instructive’ role in target gene transcriptional activation; or 2) a ‘permissive’ (pioneer factor) role [61,62] as a recruiter of cofactors that modify the chromatin, marking control regions for activation by other transcription factors to target enhancer elements of craniofacial developmental genes. In contrast, repressing PBX3 increased expression of SHH in its original expression domains on the ventral surface of the FNP around the mouth cavity and induced premature ectopic expression of SHH in the globular process at HH22, where SHH expression is observed only by HH25 [13]. Overexpression of PBX3 resulted in decreased SHH expression in the FEZ at HH22 and malformations including a smaller head (S2 Fig), which is associated with reduced SHH signaling [1,2,8]. Together, our results demonstrate complementary functions of PBX1/3 in the regulation of SHH expression in the developing FNP. Future research will aim at identifying co-factors that partner with PBX homeoproteins to activate or repress SHH expression in the FEZ of avian embryos during craniofacial development.

Experimental design

The aim of the present study was to assess regulatory mechanisms involving PBX1 and PBX3 that induce and maintain SHH expression in the FEZ during early craniofacial development. The study is divided into two major parts: 1) demonstrating expression patterns of PBX1, PBX3, and SHH in normal chick embryos at HH22 and evaluating SHH expression patterns after manipulating PBX1 and PBX3 expression and 2) profiling open chromatin configurations and PBX binding sites in the SHH locus of the normal chick FEZ at HH22 and testing activity of a potential regulatory sequence.

First, we demonstrated normal expression patterns of SHH, PBX1, and PBX3 in the FNP of normal HH22 chick embryos (n = 6) by whole mount in situ hybridization. Next, we demonstrated how overexpression and repression of PBX1 and PBX3 affect SHH in the FEZ by standard gain- and loss-of-function experiments with RCAS virus infection (n = 12).

Assessment of accessible chromatin signatures and the binding patterns of PBX1 and PBX3 to regulatory regions of the chick SHH locus was done in order to determine the regulatory mechanisms of PBX1 and PBX3 on SHH expression in the FEZ. Open chromatin configurations were evaluated by utilizing ATAC-seq (n = 2), and PBX binding sites were evaluated by utilizing ChIP-seq (n = 2). Then, a potential regulatory sequence analyzed from the high-throughput sequencing data was tested both in vivo and in vitro by electroporation (n = 15 – 20) and luciferase assays (n = 3 – 9), respectively.

Chick embryos

The experimental procedures involving chick embryos met the institutional and national regulatory standards applied for vertebrate embryos. Fertile White Leghorn chicken (Gallus gallus) eggs (Petaluma Farms, Petaluma, CA) or SPAFAS Flock -C17 RCAS free eggs from Charles River (Wilmington, MA) were incubated at 38°C until either sample collection or experiments. The embryos were staged using a strategy that relies on external morphological characters and that is independent of body size and incubation time [63,64]. Specifically, we applied the Hamburger and Hamilton staging system, which was originally developed for chick [12].

Whole mount in situ hybridization

In situ hybridization was performed on whole embryos as described [65]. Briefly, subclones of chick SHH, PBX1 and PBX3 were linearized to transcribe digoxygenin-labeled antisense riboprobes. The chick embryos underwent hybridization with 0.5 – 1.0 μg/ml digoxygenin-labeled cRNA probes, and, after washing, the embryos were incubated with an alkaline phosphatase-conjugated anti-digoxygenin antibody (Boehringer Ingelheim, Ingelheim, Germany). Nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate substrate (NBT/BCIP substrate; Roche, Basel, Switzerland) was used for color detection. Stained embryos were observed and imaged by the Leica MZ FLIII Stereomicroscope.

Gain-of-function and loss-of-function experiments by RCAS virus infection

Chick embryos were incubated until HH10. Then, 1.0 ml of albumin was removed from the egg, and a small window was cut out on the top of the shell to expose the embryo. At HH10, overexpression or knockdown of PBX1 and PBX3 in the face was achieved by RCAS virus infection of the migrating mesenchyme adjacent to the anterior neural tube at HH10. This leads to widespread infection of the facial tissues. For overexpression, we engineered RCAS virus to encode each gene (RCAS-PBX1, RCAS-PBX3, and RCAS-AP as a control). For knockdown experiments, we modified the BLOCK-iT Pol-II RNAi expression vector (Cat #, K493500; Invitrogen, Waltham, MA) to clone into RCAS virus. We designed multiple miRNAs targeting the open reading frame and the 3’UTR of PBX1 and PBX3, as well as the mirLacZ control using Invitrogen’s BLOCK-iT RNAi Designer. In our experience, these constructs significantly reduced toxicity of knockdown by employing an artificial miRNA that is processed as endogenous miRNAs rather than as shRNA. The miRNA sequences used for repressing expression of PBX1 and PBX3 and the mirLacZ control were as below:

1. mirPBX1 –
2. forward 5’-TGCTGAAGAGAAGGAGTCTTCTTCTGGTTTTGGCCACTGACTGACCAGAAGAACTCCTTCTCTT-3’ and
3. reverse 5’-CCTGAAGAGAAGGAGTTCTTCTGGTCAGTCAGTGGCCAAAACCAGAAGAAGACTCCTTCTCTTC-3’;
4. mirPBX3 –
5. forward 5’-TGCTGTTTACAGTAAAGCTGTAGGTGGTTTTGGCCACTGACTGACCACCTACATTTACTGTAAA-3’ and
6. reverse 5’-CCTGTTTACAGTAAATGTAGGTGGTCAGTCAGTGGCCAAAACCACCTACAGCTTTACTGTAAAC-3’;
7. mirLacZ –
8. forward 5’-TGCTGTTCAGACGTAGTGTGACGCGAGTTTTGGCCACTGACTGACTCGCGTCACTACGTCTGAA-3’ and
9. reverse 5’-CCTGTTCAGACGTAGTGACGCGAGTCAGTCAGTGGCCAAAACTCGCGTCACACTACGTCTGAAC-3’.

After RCAS infection, embryos were returned to the incubator, and the whole heads were collected at HH22 for whole mount in situ hybridization to detect SHH expression (conducted as described above). Also, SHH expression from the FEZ at HH22 was quantified by qPCR as described below.

Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay on embryos infected with RCAS-miRNA virus

To assess apoptotic cells in embryos that were infected with RCAS-miRNA virus, the embryo heads were fixed in 4% paraformaldehyde (PFA) solution, paraffin-embedded, and sectioned at 8 μm. On the sections, DNA fragmentation was visualized by using an in situ cell death detection kit (Fluorescein; Cat #, 11684795910; Roche, Basel, Switzerland) following the manufacturer’s instructions.

Total RNA preparation and real-time polymerase chain reaction (qPCR)

The FEZ was isolated as described below. RNA was then extracted from the FEZ using the RNeasy Kit (Cat #, 74104; Qiagen, Hilden, Germany). cDNA was synthesized by the Invitrogen Superscript III kit following the manufacturer’s instructions. Quantitative PCR was performed using a Bio-Rad CFX 96 real-time PCR machine. The qPCR primers for SHH, PBX1, and PBX3 were as below:

1. SHH – forward 5’-GCTGACAGACTGATGACTCA-3’ and reverse 5’-TCGTAGTGCAGCGATTCCTC-3’;
2. PBX1 – forward 5’-GGCTACGGAAATCCTGAATGAG-3’ and reverse 5’-ACCAGTTTGATACCTGTGAGAC-3’;
3. PBX3 – forward 5’-GCATCGATATGGACGAGCAGTCC-3’ and reverse 5’-GCATCGATTTAGTTAGAGGTATCA-3’.

Relative gene expression was calculated based on the ΔΔCt method. ΔCt was calculated between each target gene and GAPDH as an endogenous control (forward 5’ CTGGTATGACAATGAGTTTGG 3’; reverse 5’ ATCAGTTTCTATCAGCCTCTC 3’), and non-infected, normal chick FEZ was used as a control sample for calculating ΔΔCt. The relative quantity (RQ) was then calculated by the equation: RQ = 2^-ΔΔCt. At least three biological replicates were prepared, and one-way analysis of variance (ANOVA) test was used to assess the statistical significance. (Prism 10, GraphPad).

Frontonasal ectodermal zone isolation

The isolation of the chick FEZ was conducted as previously described [66]. Chick embryo heads at collection points were removed and placed into serum free DMEM, and the FNP was dissected (Fig 3A) so that only the region covered by the FEZ remains. Then, the tissue was digested in dispase (2.4U/ml) in DMEM on ice for 20 minutes. Digestion was quenched by transferring the tissue to DMEM with 1% BSA. Subsequently, using a sharpened tungsten needle, the surface ectoderm was gently lifted off of the FNP. Fresh tissues were used immediately for ATAC-seq, since fixation of cells has been found to reduce transposition frequency and is not recommended [67]. In contrast, tissues were cross-linked and snap-frozen to be stored at -80^oC for later use in ChIP-seq assays [68–70].

ATAC-Sequencing

The FEZ tissues from 10-15 normal chick embryos at HH22 were pooled for each replicate for ATAC-seq analysis (n = 2). ATAC-seq was performed on the FEZ cells employing an established two-step protocol that can use a low number of cells (as low as 50,000 cells) [67]. DNA samples were incubated with the Tn5 transposome, which performs both adaptor ligation and fragmentation of open chromatin regions. Paired-end reads (2 × 150 nt) were generated by HiSeq 4000 (Illumina) at the UCSF Center for Advanced Technology (CAT) Facility. The quality of sequencing was assessed by FastQC (version 0.11.9) [71]. The raw reads were trimmed and mapped onto the latest NCBI chicken reference genome (galGal6) by the local alignment of Bowtie2 (version 2.4.5) [72]. After read alignment, duplicates and unmapped reads were removed by Sambamba (version 0.6.8) [73], and mitochondrial reads were also removed by Samtool (version 1.11) [74]. The filtered reads were then processed by MAC2S [75] to call narrow peaks. For peak calling, the effective genome size of 1.03E + 9 was applied (for 150 bp k-mer, calculated with the NCBI chicken genome size), and the statistical threshold was set at q-value (minimum false discovery rate) < 0.05. Simple overlaps of peaks from both biological replicate were determined by the intersect function from BEDTools (version 2.31.1) [76]. Irreproducibility discovery rate (IDR) for every peaks between two replicates were calculated, and only peaks with IDR < 0.05 were merged as overlapping peaks [28]. The analyzed peaks and overall peak profiles were then visualized on the Integrative Genomics Viewer (IGV viewer, version 2.15.4) [77]. Lastly, the motif analysis on ATAC-seq peaks (IDR < 0.05) was conducted by the findMotifsGenome.pl function in the HOMER [34] package to identify known motifs from the open chromatin profiles. The region size setting used for motif discovery was given peak size.

ChIP-Sequencing targeting PBX1 and PBX3

Since ChIP-seq requires a substantially greater number of cells compared to ATAC-seq, the FEZ tissues from ~80–85 normal chick embryos were pooled for each replicate (n = 2). For immunoprecipitation, cells from the FEZ were crosslinked with 1% PFA for 10 min at room temperature and sonicated by 30 cycles (1 min/cycle; reset every 5 min). An aliquot was removed and precipitated for input controls, and another aliquot was incubated with Pbx1 antibody (5 µg/reaction; Cat #, 4342; Cell Signaling Technology, Danvers, MA), PBX3 antibody (3.9 µg/reaction; Cat #, 12571–1-AP; Proteintech, Rosemont, IL), or IgG antibody (5 µg/reaction; Cat #, 2729S; Cell Signaling Technology, Danvers, MA). DNA libraries were constructed using the TruSeq ChIP Sample Preparation Kit (Cat #, IP-202–1012; Illumina, San Diego, CA). To confirm that the ChIP reactions were properly performed, qPCR was conducted. The ChIP-seq libraries targeting PBX1, PBX3, and IgG were amplified with primers that amplify target regions of PBX1, PBX3, and MEIS2 (as a positive control) as well as negative control primers targeting random regions of different chromosomes, where PBX1 and PBX3 do not bind:

1. PBX1 promoter region – forward 5’-GGGCTCCAAACTTTCCCCCACC-3’ and reverse 5’-GAGGAGTCCGGAGCCGAACAA-3’;
2. PBX3 promoter region – forward 5’-AAACTTCTGATCGCCGCCGT-3’ and reverse 5’-CACCCCGGATCGCGACATAA-3’;
3. MEIS2 promoter region – forward 5’-CCGTCACTTCATTCTCCCGCC-3’ and reverse 5’-GCTCGGAGATGAGCGAGTGTC-3’;
4. Negative control 1 – forward 5’-ACCGTTAACACCACCTTGGGC-3’ and reverse 5’-CTGGCTGGCCCAACATACCA-3’;
5. Negative control 2 – forward 5’-ACAGTGGTCTCTGGCCTTAGGT-3’ and reverse 5’-AGGCTGACTTCAGGGGTAGAGC-3’.

Single-end reads (50 nt) were generated by HiSeq 4000 (Illumina) at the UCSF CAT Facility. The same sequencing data analysis pipeline (including quality check, read trimming, alignment, peak calling, and merging peaks between biological replicates) that was used for ATAC-seq was also applied for the ChIP-seq datasets with a few modifications described below. The effective genome size of 1.02E + 9 (for 50 bp k-mer, calculated with the same NCBI chicken genome size) was used for peak calling by MAC2S [75] to call narrow peaks (q < 0.05). Peaks from each sample were normalized by an input control during peak calling. After merging IDR < 0.05 peaks from each experimental group, PBX1- and PBX3-enriched peaks were further analyzed by an R package, DiffBind [30,31], to identify differential binding affinity. The pathway analysis was conducted based on the KEGG database [32,33]. Lastly, we conducted the motif analysis on PBX1 and PBX3 peaks with IDR < 0.05 by the findMotifsGenome.pl function in the HOMER [34] package to identify both de novo and known motifs in our datasets. The region size setting used for motif discovery was 200 bp as recommended for transcription factors by the developers to identify both primary and co-enriched motifs.

Both ATAC-seq and ChIP-seq data are openly accessible through National Center for Biotechnology Information Sequence Read Archive (NCBI SRA; accession number: PRJNA1111924) and FaceBase (https://doi.org/10.25550/5G-0TNJ).

To confirm that PBX proteins are bound to SFE1 elements, we used PCR to amplify SFE1 sequences in the ChIP-seq libraries targeting PBX1 with the following primers targeting the SFE1 around the first CCCGCCC motif (SFE1A):

SFE1_1 (product length: 191 bp, which includes SFE1A) – forward 5’-GCCGGATCCAGCCAATG-3’ and reverse 5’-GAATGCATTAGCTGCGGGTT-3’;

SFE1_2 (product length: 191 bp, which is ~ 80 bp upstream from SFE1A) – forward 5’-CTCTGAGTTCCTCCCATCGCG-3’ and reverse 5’-TGACAACTGCGGCTTGCTGA-3’.

The PCR products were then run through a 2% agarose gel along with the 100 bp DNA ladders, and DNA sequencing was conducted to confirm the presence of amplified SFE1 sequence in the final products.

Electrophoretic mobility shift assay

Three 25 bp double stranded DNA oligos that contain each of the GGCGG/CCCGCCC motifs within SFE1 were biotin labeled at the 5’ end:

1. SFE1A – forward 5’-CCCCCCCCCCCCGCCCCCCTCTGAA-3’ and reverse 5’-TTCAGAGGGGGGCGGGGGGGGGGGG-3’;
2. SFE1B – forward 5’-TAATGCATTCGGCGGCTATTCCCCG-3’ and reverse 5’-CGGGGAATAGCCGCCGAATGCATTA-3’;
3. SFE1C – forward 5’-CGAGGCGAAGGGCGGGGGGCTGCGG-3’ and reverse 5’-CCGCAGCCCCCCGCCCTTCGCCTCG-3’.

Nuclear extracts from DF1 cell cultures that were transfected with RCAS-PBX1 were prepared using the NE-PER Nuclear and Cytoplasmic Extraction Reagents (Cat #, 78833; ThermoFisher Scientific). Binding reactions of DNA oligos and cell extracts were conducted using the LightShift Chemiluminescent EMSA Kit (Cat #, 20148; ThermoFisher Scientific). Final concentration of each reaction was comprised of 50 ng/ul of Poly (dI*dC), 2.5% of glycerol, 0.05% of NP-40, 5mM of MgCl₂, 20 fmol of biotin-labeled DNA oligos, and 0.6 ug of cell extracts in total volume of 20 ul. For unlabeled DNA controls, 4 pmol of label free DNA oligos that have the same sequence as the biotin labeled oligos were added to the reaction mix. Binding reactions were incubated at room temperature for 20 min and run through a 5% polyacrylamide gel (Cat #, 4565015; Bio-Rad) in a 0.5X Tris-boric acid-EDTA (TBE) buffer for 45 min. The gel was transferred onto a positively charged nylon membrane (Cat #, 11209299001; Roche) by using Trans-Blot Turbo Transfer System (Cat #, 1704150; Bio-Rad). The transferred blot was cross-linked with a UV-transilluminator for 15 min. The blot was then blocked with DNA blocking buffer, incubated with streptavidin-HRP conjugate for 15 min, washed, and finally incubated with luminol/enhancer and peroxide solution for 5 min. The labeled DNA oligos were detected and imaged by ChemiDoc MP (Cat #, 12003154; Bio-Rad).

Electroporation

A reporter vector comprised of a minimal Thymidine kinase promoter, LacZ encoding β-galactosidase, and a polyadenylation sequence (TKP-LacZ) was used for the analyses as the previously established approach [3]. The 1.8kb sequence (chr2:8,172,590-8,174,390) and the middle 600 bp (chr2:8,173,096-688) spanning the 400 bp-long PBX1-enriched region on intron 1 of SHH was cloned by Long-Range PCR, and then the cloned genomic fragments were individually ligated into the TKP-LacZ vector. This construct (TKP-LacZ-1.8kb or TKP-LacZ-Middle 600 bp) was electroporated into the developing FEZ of embryos at HH20 (n = 15 – 20). As a negative control, we electroporated the TKP-LacZ vector without any cloned inserts, and as a positive control to assess the extent of transfection, we electroporated a constitutively activated vector comprised of heat shock protein promoter that encodes β-galactosidase (HSP68-LacZ). Embryos were then allowed to develop for 24 hours (~HH24), fixed, and assessed for β-galactosidase activity by a standard X-gal reaction as previously described [13].

Luciferase assay

The 1.8kb sequence (chr2:8,172,590-8,174,390) spanning the 400 bp PBX1-enriched locus was divided into three 600 bp segments. Each of these segments were transfected into DF1 cells along with a luciferase reporter construct and either of an expression vector (RCAS-PBX1, RCAS-PBX3, RCAS-mirPBX1, RCAS-mirPBX3, and RCAS-AP as control). Luciferase intensity was measured by the Luciferase Assay System kit (Cat #, E1500; Promega, Madison, WI) on Promega GloMax Explorer (Cat #, GM3500) and normalized with the RCAS-AP control. One assay includes three technical replicates, and total 3 – 9 inter-assays were conducted for each combination. Outliers were determined by Grubbs’ test (extreme studentized deviate method) and then removed for further data analyses. Then, unpaired t-test was conducted, and the significance was determined when P < 0.05 (Prism 10, GraphPad).

Fluorometric β-galactosidase assay

The middle 600 bp fragments (chr2:8,173,096-688) with intact SFE1 and SFE1 null sequences (three putative PBX binding sites were deleted from SFE1) were individually cloned into the HSP68-LacZ vector with ApaI/HindIII cloning sites. DF1 cell cultures were transfected with 300 ng of RCAS-PBX3 or RCAS-AP control along with 50 ng of either of the following three constructs: HSP68-LacZ-no fragment, HSP68-LacZ-Middle 600 bp, or HSP68-LacZ-Middle 600 bp SFE1 null. The transfected cells were cultured for 48 hours. For quantification of β-galactosidase, we used the sensolyte-mug-beta-galactosidase-assay-kit-fluorimetric (Cat #, AS-72132; Eurogentec, Seraing, Belgium). Fluorescence signal was recorded on Promega GloMax Explorer (Cat #, GM3500) at EX/EM = 365/445 (four inter-assays with triplicate technical replicates). The fluorescent intensity was corrected with the HSP68-LacZ-no fragment controls and then normalized by RCAS-AP controls. The normalized relative fluorescent intensity underwent unpaired t-test, and the statistical significance was determined when P < 0.05 (Prism 10, GraphPad).