Enhancer Biology: Holding Out for the Right Promoter

Summary

Some transcriptional enhancers work best with one type of promoter, while ignoring others. How widespread is such specificity across the genome? A new study finds that, in a fair fight, most enhancers prefer to activate promoters resembling those of their parent genes.

To developmental biologists, there are two kinds of genes. Most of our time is spent thinking about what we might call the “interesting” category, which covers all genes whose expression is differentially regulated—that is, genes that are more strongly expressed in cell A compared to cell B, or at stage X compared to stage Y. Differential gene expression, after all, is the cornerstone of development and many other complex biological processes: it’s what makes a cell different from its neighbor. The second category—genes whose expression is consistent across cell types and developmental stages—receives the pejorative label of “housekeeping genes,” even though many of these genes do more than just maintain a tidy cell. (Not that cellular housekeeping is an unimportant function, or an unregulated one for that matter.) The origin of the term “housekeeping gene” is obscure, but it seems to have arisen to describe the set of genes that are required for cell viability and basic functions such as protein translation, mitosis, and cellular metabolism. Today, the word is most often used to refer to a control group in measurements of gene expression—the underlying assumption being that such genes are constitutively expressed at the same level in all cells and under all conditions.

Perhaps not surprisingly, differentially regulated genes—and the genes that regulate them—get the lion’s share of researchers’ attention. Little is known about how ubiquitous expression patterns are generated, but the promoters of constitutively expressed RNA polymerase II (Pol II)-transcribed genes tend to share certain sequence motifs, DNA-binding factors, Pol II occupancy patterns, and chromatin states that distinguish them from “regulatory” or “developmental” gene promoters [1–8]. So the separation of genes into two classes—constitutive/housekeeping vs. regulated/regulatory/developmental—is not just a product of the human tendency to think in dichotomies, but rather, to some extent at least, a real biological distinction.

A new study by Muhammad Zabidi, Cosmas Arnold, and colleagues in Alexander Stark’s group [9] reveals that “housekeeping” and “developmental” genes differ not only in their core promoters, but in their enhancers as well. Enhancers are cis-regulatory DNA sequences containing clusters of transcription factor-binding motifs; they “loop in” to interact with core promoters and activate transcription [10, 11]. The Zabidi/Arnold study, which uses an innovative high-throughput method for testing enhancer function across the genome, shows that enhancers of housekeeping vs. developmental genes differ, not only in the kinds of regulatory factors that bind them, but also in their ability to interact with core promoters of these two categories of genes.

Long before the genomics era, it was proposed that some enhancers are choosy about which promoters they prefer to interact with, and that DNA sequences in the vicinity of the promoter are responsible for these preferences [12]. This hypothesis was supported by data from cleverly designed transgenic reporter assays in flies and mice, demonstrating that certain enhancers will shun a nearby promoter in the presence of another, more preferred, promoter sequence, even if the favored promoter is more distant [13–19].

More recently, genomic studies of chromatin accessibility and long-range intrachromosomal interactions (reviewed in [10]) have indicated that a large number of enhancers do not interact with the nearest promoter, but instead seek out more a distant partner. This could be taken to imply widespread enhancer-promoter specificity across the genome, but the possibility remains that the passed-over promoters in these studies were in a “closed” state and simply not accessible by nearby enhancers. To more conclusively determine whether most enhancers prefer one type of promoter over another in a fair contest, a new experimental approach was needed.

In a recently published study [9], Zabidi, Arnold and colleagues tested the enhancer-promoter specificity hypothesis on a genome-wide scale, using STARR-seq (Self-Transcribing Active Regulatory Region-sequencing), a technique developed by the Stark group [20] for high-throughput, genome-wide screening of enhancer activity. Randomly sheared genomic DNA fragments were cloned into transfection vectors downstream of a promoter, so that any resulting transcripts will contain the sequence of the enhancer that stimulated their expression. Vector-derived cDNA sequences are then mapped to the genome, identifying enhancers that are active in a specific cell line, on a specific promoter, across the entire genome (Figure 1A). One limitation of the STARR-seq approach is that it only identifies enhancers that are capable of activating the promoter used in the transfection vector; another is that it will only find enhancers that are active in the particular cell line used for the experiment. Zabidi, Arnold and colleagues turned these limitations into strengths by testing their library of potential enhancers in two cell lines, and in cis to either a ‘housekeeping’ or ‘developmental’ core promoter, to assess the cell- and promoter-specificities of enhancers throughout the genome.

Figure 1.

Using the STARR-seq technique to measure enhancer-promoter specificity. (A) Summary of STARR-seq. Random genomic DNA fragments were cloned into vectors containing either a developmentally regulated core promoter (dCP, green), based on a developmentally regulated gene, or a housekeeping core promoter (hkCP, pink), taken from a constitutively expressed gene. When transfected into cultured cells, enhancers “self-transcribed” if they were compatible with the core promoter in the vector. Vector-specific cDNAs were sequenced and mapped onto the genome: a representative UCSC Genome Browser view of a region containing unique dCP and hkCP enhancer peaks is shown. (B) Summary of key differences between dCP- and hkCP-preferring enhancers and their target genes.

The housekeeping core promoter (hkCP) used in this study contained the TCT motif, which is enriched in promoters of housekeeping genes, while the developmental core promoter (dCP) contained TATA, Inr, MTE, and DPE motifs, which are commonly found in promoters of developmentally regulated genes. Of the 11,364 enhancers identified by STARR-seq in S2 cells, a widely used Drosophila cell line, 72% were at least twice as active on one core promoter than on the other. Enhancers that preferred hkCP shared several properties, relative to those that preferred dCP (summarized in Figure 1B), which together suggested that the former category is enriched in enhancers of housekeeping/constitutive genes, while the latter group is enriched in developmentally regulated enhancers. For example, hkCP-favoring enhancers lie within 200 bp of the nearest promoter; those promoters are enriched in housekeeping promoter motifs, including DRE and TCT; and they tend to be found in genes with housekeeping-like functions. dCP-preferring enhancers, on the other hand, tend to be much more distal from the nearest promoter; those promoters are enriched in motifs characteristic of developmental genes (TATA, Inr, MTE, and DPE); and they are more often found near developmental regulatory genes. In other words, when given a choice, most enhancers preferentially activate core promoters that resemble those of their parent genes.

By performing the same assay in a second Drosophila cell line, derived from a developmentally unrelated source tissue, the authors could crudely estimate the tissue-specificity of each enhancer’s activity. Enhancers with comparable activity in both cell lines tended to prefer hkCP (which was taken from a constitutive gene), while enhancers with a strong cell-type bias in their activity generally preferred dCP (which was based on a developmentally regulated gene). All of these correlative data, taken together, suggest that tissue-specific enhancers of developmentally regulated genes are most active when paired with a developmental core promoter, while enhancers of constitutively expressed genes work best with a housekeeping-like core promoter.

What molecular mechanisms are responsible for such widespread enhancer-promoter specificity? DNA motif analysis of “housekeeping” vs. “developmental” enhancers revealed an enrichment of Dref-binding DRE motifs in the former category, while Trithorax-like (Trl)-binding GAGA motifs were enriched in the latter group. Mutational analyses confirmed functional roles for these regulatory sites: four hkCP enhancers depended (to varying degrees) on DRE motifs, while two dCP enhancers required GAGA motifs for some of their activity. It would be interesting to see whether extending this analysis can identify functional sub-categories within these two classes of promoters and enhancers.

The results presented by Zabidi, Arnold et al. show clearly that motif-based enhancer-promoter specificity is widespread in the fly genome. But the authors’ repeated use of the verb “separate,” as in the article’s title, “Enhancer–core-promoter specificity separates developmental and housekeeping gene regulation,” could be interpreted to mean that housekeeping and developmental genes use different enhancer-promoter recognition mechanisms in order to separate housekeeping gene regulation from developmental gene regulation—that is, to keep developmental enhancers from regulating housekeeping genes, and vice versa. But this isn’t necessarily the case. It may simply be that enhancers of housekeeping genes, which evolve under selective pressure to activate their local hkCPs, acquire regulatory motifs (such as DREs) that best facilitate that activation, while developmental enhancers evolve along a different path to optimally activate their target dCPs. In other words, each enhancer may be shaped by evolution to most effectively regulate the particular promoter it has to talk to, rather than to keep developmental and housekeeping programs separate. In this view, most of the work of enhancer trafficking in the genome could be performed by other proposed mechanisms, such as higher-order partitioning of chromatin into transcriptional domains (reviewed in [11]). Direct manipulation of enhancer and promoter sequences in situ will be needed to confirm the functional significance of motif-based enhancer-promoter specificity—which, as this exciting new study shows, is a surprisingly common feature of cis-regulatory elements throughout the genome.