Role of Architecture in the Function and Specificity of Two Notch-Regulated Transcriptional Enhancer Modules

In Drosophila melanogaster, cis-regulatory modules that are activated by the Notch cell–cell signaling pathway all contain two types of transcription factor binding sites: those for the pathway's transducing factor Suppressor of Hairless [Su(H)] and those for one or more tissue- or cell type–specific factors called “local activators.” The use of different “Su(H) plus local activator” motif combinations, or codes, is critical to ensure that only the correct subset of the broadly utilized Notch pathway's target genes are activated in each developmental context. However, much less is known about the role of enhancer “architecture”—the number, order, spacing, and orientation of its component transcription factor binding motifs—in determining the module's specificity. Here we investigate the relationship between architecture and function for two Notch-regulated enhancers with spatially distinct activities, each of which includes five high-affinity Su(H) sites. We find that the first, which is active specifically in the socket cells of external sensory organs, is largely resistant to perturbations of its architecture. By contrast, the second enhancer, active in the “non-SOP” cells of the proneural clusters from which neural precursors arise, is sensitive to even simple rearrangements of its transcription factor binding sites, responding with both loss of normal specificity and striking ectopic activity. Thus, diverse cryptic specificities can be inherent in an enhancer's particular combination of transcription factor binding motifs. We propose that for certain types of enhancer, architecture plays an essential role in determining specificity, not only by permitting factor–factor synergies necessary to generate the desired activity, but also by preventing other activator synergies that would otherwise lead to unwanted specificities.


Sequence of ASE5
ASE5 was originally defined by a 372-bp Bsu36I-AseI genomic DNA fragment downstream of the 3' UTR of Su(H) [10]. In the present study, we included an additional 19-bp sequence at the 3' end (underlined) because this segment is part of a highly conserved sequence block. Later we found that adding this 19-bp sequence also allowed us to include the entirety of a Vvl type 2 binding site (AATTAA). Mutant variants of ASE5 were created by introducing non-complementary transversions (A <-> C and G <-> T) into target regions. As an example, the sequence of ASE5M2, in which all bases between the five Su(H) binding sites were mutated, is shown below. The sequences of other ASE5 variants are available on request.
In the sequences below, Su(H) binding sites are shaded in green, box A in red, and Vvl binding sites in yellow. Mutated bases in ASE5M2 are shown in lower case; the proneural protein binding site introduced by the mutagenesis is shown in bold. The Bsu36I and AseI restriction sites are shown in italics.

Design of primer sequences for synthesizing ASE5
To accelerate the introduction of large blocks of mutations (non-complementary transversions: A <-> C and G <-> T) into ASE5, we synthesized the enhancer fragment de novo by recursive PCR [22], using nine forward primers and nine reverse primers as listed below. The forward primer sequences do not overlap with each other, but each primer overlaps with two adjacent reverse primers by 20 bases each. The 5' and 3' end primers include EcoRI and BamHI restriction sites, respectively. To generate ASE5 mutants or ASE5-shuffle variants, primers containing mutated or shuffled sequences were used instead of their wild-type counterparts in the recursive PCR.
The primers used for synthesizing wild-type ASE5 are shown below; primer sequences used for making ASE5 mutants and other variants are available upon request.
Lower-case letters denote restriction sites (bold) and end-protecting nucleotides.

Yeast one-hybrid screens
We performed two separate yeast one-hybrid screens, with either Fragment X or Fragment Y of ASE5 (see Figure S1A) as bait, using the Matchmaker Yeast One-Hybrid System (Clontech). The cDNA library used in the screen was prepared from stage-13 embryos, when embryo socket cells are newly born.
The Fragment Y bait was generated by PCR; the Fragment X bait by annealing two synthetic complementary oligonucleotides (Integrated DNA Technologies). A cDNA library was generated from 2 µg of total RNA from stage-13 embryos, and was cloned into the prey plasmid as a fusion with GAL4AD coding sequences by in vivo recombination according to the manufacturer's instructions. High-efficiency yeast transformation was performed as described [29]. Positive clones were selected on synthetic defined media -Trp -Leu -His, plus 10 mM 3-AT. In total, 1.2 million clones were screened with bait Y, and 0.6 million clones with bait X. For each screen, 100 independent clones from selection plates were serially streaked three times on fresh selection plates, after which the cDNA insert was amplified from individual clones by PCR and sequenced. Identity of the clones was determined by BLAST analysis against the Drosophila melanogaster genome.
With Fragment Y as bait, a significant fraction of positively selected clones (13/75) contained a cDNA fragment encoding Ventral veins lacking (Vvl), a POU-HD transcription factor [30,31]. By contrast, the screen using Fragment X as bait did not yield obvious candidates. See Table S1 and Table S2 for details.

Identification of Vvl binding motifs in ASE5
We found that an eight-nucleotide motif (ATGCAAAT) located in ASE5 Fragment Y perfectly matches a previously characterized Vvl binding site [12]. Using an electrophoretic mobility shift assay (EMSA), we confirmed that this motif, designated V1, is bound specifically in vitro by a purified GST-Vvl fusion protein (see below and Figure S3). To determine if other Vvl binding sites are contained within ASE5, we used double-stranded oligonucleotides covering the entire ASE5 as competitors in the EMSA, which led to the discovery that the AATTAA motif is also bound strongly by GST-Vvl (see Figure S3).

Electrophoretic mobility shift assays (EMSAs)
The complete vvl coding sequence was obtained by PCR from a cDNA library of stage-13 embryos (strain w 1118 ), and cloned into vector pGEX-5X-2 (GE Healthcare Life Sciences). The GST-Vvl fusion protein was expressed in E. coli strain BL21, and purified according to the manufacturer's instructions. The concentration of the purified protein was estimated by comparing it with BSA standards in a Coomassie-stained 4-20% SDS-PAGE gel (Bio-Rad Laboratories, Inc.).
Oligonucleotides used for EMSA probes were labeled with Biotin-11-UTP and annealed in vitro according to the manufacturer's instructions (Thermo Scientific). For each EMSA reaction, approximately 10 µg of purified GST-Vvl was incubated with 20 pmol of labeled probe for 20 minutes at room temperature. Free and bound probes were separated on 5% non-denaturing polyacrylamide gels (Bio-Rad) and detected using the LightShift Chemiluminescent EMSA kit according to the manufacturer's instructions (Thermo Scientific). For competition assays, 10 nmol of unlabeled oligonucleotides (500-fold excess) were included in each reaction mix.

Sequences of synthetic enhancers based on ASE5
The synthetic enhancer ASE5-shrink is composed of the essential sequences of ASE5: The segments underlined in the mα enhancer sequence below were joined to make the mα-shrink enhancer; the "E box" segment was omitted in the construction of the mα-shrink∆E and mα-shrink∆E-Vm variants. Both the mα-Vm and mα-shrink∆E-Vm variants include the same point mutations in the Vvl binding motifs; these are shown in lower case in the mα-shrink∆E-Vm sequence below. TCGCTTTGCACACACTTTCTCCCTGCGCCTTATCGGGTAATCCCTCTCCCTTGAAACAATATTG  AAGAAACCCTCCCTATCGCTCCACACGTTCAGCACACTTGGGCCGCAAGAAAGGGAATTAATTA  AAATATCTATAAGATTCCCAGCTCACCCCTGGGGAGTGTTTCCAAATTGAAGACAAGTGGCAAG  TCCGTATTTTATTGGTGGTGGTTTCGATGGTGCTAGTGAATAGTGGTAAATGGATTCATCGAGC  CCTGTGGGAAAGTTGGAAATCAAAACACCATAACAAGTGATTCGAGATGCCTATACAAATAGAA  GATCCAACCCCAAGATCCCTTATGCCCTTTCATATGCACGAAACCAGAGCCAGGACGAAGCAAT  GTGTGGGAATGCGTGGGAATGGTCCTGGGGATTCGAAACTCAGAAACGGTCCCCTATCCCTGCT  TTTATGTTGATTGCCCATTAGGAATACAATTTGCAACTCCTTTTCCATGATGTGATCCCCTTGT  TGATCCTGGGCTTCCTTCGGTGTCGTGAGAAATTTTACCAAGGAACACCTGCCCCGTATCGGTG  TGTCCTGGATTGCCAGTGTGTTGGTGATTTTTTTCATGTCCTCTGGGGATTTCACGCAGCACTT  GAATACGTTTCCGAAAATTTTGAGGCCAGCGAAAAATTGTTTCCCACACTCGTCGGACAGTTGC  AGTTTTTGTAGGGGTTTGTTTCCTTTGCCGATGCACTCTTCTCTCGTCTCATCCCGTCTCGTCT  TTACCTGCTCCTGGAACTTCCCCCGACTCCTGGCGAGCGTCGTATTTCAGGTTCTCCGCCCCGA  TTATCTCGGCTCTTTGAGGGCGAGGGTCTGCGTACTCCGTTTCTGTCTCTGTTTGTATTCCCCT