Transcriptional regulation^*

Regulation of Polymerase II (Pol II) transcription in C. elegans can be described as typical for eukaryotes. Pol II appears to act in concert with TATA Binding Protein (TBP) and TBP-Associated Factors (TAFs) at the promoter of protein coding genes (Dantonel et al., 2000; Kaltenbach et al., 2000; Lichtsteiner and Tjian, 1993; Walker et al., 2004). Active Pol II is phosphorylated on the C-terminal domain (CTD) at serine 2 and 5 like other eukaryotes (Seydoux and Dunn, 1997; Wallenfang and Seydoux, 2002; Zhang et al., 2003). The functions of these proteins at the core of transcription are beginning to be defined and are reviewed in Transcription mechanisms. However, there are many things about transcription in C. elegans that we do not yet know for sure. For example, putative TATA and CAT boxes upstream of the coding region are often described, but this is largely done subjectively without any firm experimental evidence for the function of these elements. We also have not fully explored the role of histone modifications and chromatin organization in somatic cell transcription. Progress on these fronts has primarily been made in the areas of dosage compensation (see X-chromosome dosage compensation) and germline chromatin organization (see Germline chromatin). For these cases, the evolutionary conservation suggests that somatic cell transcription will similarly be influenced by typical eucaryotic mechanisms of chromatin organization. In many ways then, our understanding of transcription in C. elegans is still in its infancy, reflecting the fact that C. elegans, as a model biological system, is still a growing field that has primarily been exploited for its genetics.

The purpose of this chapter is to provide an overview of transcriptional regulation in C. elegans. It is geared towards an audience that is naïve in the ways of C. elegans gene regulation, but, it also includes information that should be helpful even to the seasoned veteran. The tools for studying transcription in C. elegans will be described in an effort to illustrate successful approaches and highlight techniques that, while useful in other systems, are challenging in the nematode. A review of the general trends in regulatory elements is followed by specific examples of spatial and temporal regulatory strategies. The hope is that this information will serve as both a useful review and an entry point into literature appropriate for specific applications.

Reporter genes are the most commonly used method to study transcriptional regulation in C. elegans. It is straightforward to generate transgenic lines (see Transformation and microinjection), and, as C. elegans is transparent throughout its life, it is easy to visualize reporter gene expression in all cells. Early studies of gene expression relied on lacZ reporter genes and were aided by the development of a set of vectors by the Fire lab (Fire et al., 1990). These lacZ reporters were very useful for determining cis-acting transcriptional control elements (Fire and Waterston, 1989; MacMorris et al., 1994; Okkema et al., 1993) and lacZ continues to be a robust marker that can be pushed to detect even low levels of expression (Wilkinson and Greenwald, 1995).

More recently, Green Fluorescent Protein (GFP), or one of its variants, serves as the common reporter. A GFP coding cassette can be inserted in different locations within a large genomic clone (tens of kilobases) to generate transcriptional and/or translational fusions. These constructs provide the greatest chance of capturing required cis-acting regulatory elements. However, it is more common to make the assumption that genomic sequences 5′ to the coding region represent the core promoter. This region, usually a few kilobases in length, can be PCR amplified and easily cloned into the reporter gene backbone of choice. Alternatively, the Promoterome Project can serve as a source for many promoter regions and is useful if the reporter gene cloning strategy is Gateway (Invitrogen) compatible (Dupuy et al., 2004). These constructs are commercially available through Open Biosytems.

There are several considerations to take into account when making reporter genes. One is the distinction between transcriptional and translational reporters; often one would like to have both. For transcriptional reporters, expression can be engineered to highlight the cytoplasm, nucleus or other cellular compartments in the expressing cell. Nuclear localized reporters are useful for embryonic cell identification whereas cytoplasmic reporters are often more useful for larval cells, particularly in neurons where they highlight the axonal and dendritic tracks. The Chalfie lab is developing a two-part fluorescent expression system that has the potential to simplify cell type identification (Zhang et al., 2004). Translational reporter genes can provide information on the subcellular localization of the endogenous gene product. In this case, it is advisable to use the fusion protein to rescue a mutant phenotype, thus demonstrating that some or all aspects of the expression pattern and subcellular localization are biologically relevant.

Once a pattern of expression is determined, promoter analyses can be used to home in on important regulatory elements. Sequential deletions of putative promoter regions linked to a GFP reporter gene are easily made by traditional cloning or Splicing by Overlap Extension (SOEing) Polymerase Chain Reaction (PCR) amplification (Horton et al., 1990). The latter technique allows high throughput and convenience as PCR reactions can be injected directly into animals without purification or cloning (Hobert, 2002). As the control elements become localized to small genomic regions (several hundred base pairs or less), they can be placed upstream of “basal” promoters to assay for enhancer activity. These approaches are common in the C. elegans literature and can be very successful in defining important cis-acting regulatory sequences.

The use of reporter genes has several important caveats. First and foremost is that they are artificial and can easily misrepresent the pattern of gene expression. Important positive- and negative-acting control elements can be excluded by assuming promoter location leading to mosaicism, loss of expression, or ectopic expression (Krause et al., 1994). For example, reporter genes that lack sufficient control elements for fidelity are often expressed in the anterior and posterior intestinal cells or a small set of head neurons. Moreover, small changes in promoter regions can dramatically alter expression patterns as illustrated by the studies of the ges-1 gene (Egan et al., 1995). It is critical to confirm reporter gene expression patterns with an independent technique such as in situ hybridization, antibody staining, or mutant phenotype.

A second concern of reporter genes is the nature of any additional modules in the construct or in a co-injected marker that may have an effect on expression. For example, many of the standard constructs available from the Fire Lab have a 3′ untranslated region (UTR) derived from the unc-54 gene encoding muscle myosin heavy chain. These sequences may not be neutral when combined with promoters from other cell types. There are also reports of dramatic effects of co-transformation markers on expression levels suggesting that you should not rely on a single co-transformation marker when exploring the expression of a novel gene (Fukushige and Siddiqui, 1995). Similarly, “basal” promoters (e.g., pes-10) may be biased in working with certain types of genomic elements. This may cause them to fail to respond in certain cell types resulting in false information about a particular enhancer element (Natarajan et al., 2004). As long as you keep these limitations in mind, reporter genes can be very helpful in characterizing transcriptional control elements for the gene of interest.

Genome-wide approaches provide a more global assessment of transcriptional regulation and have begun to become more common in C. elegans. The availability of both spotted cDNA (see Kim Lab) and oligonucleotide microarrays (e.g., Affymetrix) for C. elegans has given birth to a large amount of gene expression data in response to tissue type (e.g., germline-enriched; Reinke et al., 2004), growth conditions (e.g., dauer; Liu et al., 2004), or mutant background (e.g., DAF-16; McElwee et al., 2003). Much of this data is available on web sites (e.g., http://genome-www5.stanford.edu/cgi-bin/login.pl) or is linked in Wormbase to individual genes. A second global approach for gene expression profiling is Serial Analysis of Gene Expression (SAGE) that has recently been combined with tissue isolation or cell type sorting (McKay et al., 2003; http://elegans.bcgsc.ca/home/sage.html). These approaches give an overview of expression. For specific genes, this data should be validated by an independent method, such as reverse transcriptase (RT)-PCR or reporter genes.

Bioinformatics provides another way to study gene regulation, either alone or in combination with other methods. Currently, the genome sequence of two Caenorhabditis species are finished (elegans and briggsae), one is in draft (remanei), and two are planned (japonica and CB5161). Interspecific comparisons of non-coding regions provides a powerful tool in identifying important cis-acting regulatory elements controlling gene expression, as functional elements will remain constrained through evolution. Comparisons between C. elegans and C. briggsae revealed important cis-acting sequences controlling the vitellogenin genes and helped to identify GATA-type transcription factors as likely regulators (MacMorris et al., 1994; Spieth et al., 1991; Winter et al., 1996; Zucker-Aprison and Blumenthal, 1989). Such comparison continue to provide valuable information about cis-acting sequences within gene promoter regions with many examples in the literature (Culetto et al., 1999; Kirouac and Sternberg, 2003; Marshall and McGhee, 2001; Natarajan et al., 2004; Teng et al., 2004). The power of these comparisons is increased as the number of species is increased and will thus become more informative as sequences of additional species are finished. Recently developed programs, such as FamilyJewels, provide methods for sophisticated multiple alignments (Brown et al., 2002). This approach will become widely exploited in coming years to pinpoint regulatory promoter elements.

Bioinformatic analysis of known transcription factor binding sites upstream of coding regions has also been successful. Given a known binding consensus site of sufficient length, the CisOrtho program can be used to ferret out a list of potential genes sharing expression patterns (Bigelow et al., 2004). Bioinformatic comparisons of promoters from genes with the same or overlapping expression patterns can also be informative to home in on potential regulatory elements (for example, see Chang et al., 2004; Guhathakurta et al., 2004). A nice combination of bioinformatics and in vitro studies used the DNA binding properties of DAF-12, a regulator of dauer development and lifespan, to define potential binding sites and gene targets (Shostak et al., 2004). Regardless of the method used, candidate elements and gene targets should be validated experimentally by an independent means.

There are also several techniques for studying gene expression that, while commonplace in other organisms, are not routinely used in the worm. For example, one would ideally isolate pure populations of cells and tease apart transcriptional regulation at a biochemical level. At only 1mm in length as an adult, C. elegans makes tissue dissections tedious or impossible for generating enough homogeneous tissue for biochemical analysis. The recent development of cell culture techniques (see Methods in Cell biology), coupled with cell sorting, may make biochemical approaches more feasible in the future. However, the technique is still challenging enough that most researchers have opted for other methods to study transcription.

In situ hybridization is another common technique for cataloging transcriptional profiles in many organisms but it is less often used in C. elegans studies. The impermeable egg shell of C. elegans embryos and the cuticle of larvae and adults often lead to background hybridization or partially permeabilized animals, making it difficult to get in situ hybridization signals that are reproducible or trustworthy. Despite these difficulties, a genome-scale effort to catalog gene transcription profiles using in situ hybridization by the Kohara group is now underway. Their protocols and data are useful and can be accessed at http://nematode.lab.nig.ac.jp/db2/index.php.

The majority of protein coding genes in C. elegans are within gene-dense regions of the genome. Consequently, cis-acting regulatory regions are usually close to the coding region. The minimal promoter region required for proper expression of most Pol II transcripts lies within a couple of kilobases upstream of the start codon. There are notable exceptions to this compact view of cis-acting sequences. For example, egl-1 expression is controlled, in part, by an element located greater than 2 kb downstream of the coding region and beyond an unrelated, intervening gene (Thellmann et al., 2003). For lin-39, proper reporter gene expression required inclusion of ~30 kb of genomic DNA that extended upstream and downstream of the protein coding region (Wigmaister and Eisenmann, personal communication). Clearly C. elegans genes can have complex and distant control regions. However, a rule-of-thumb of 2 kb upstream of the ATG works well as a starting point in the search for cis-acting control elements.

It is important to remember that the minimal promoter region is not synonymous with the natural promoter. The natural promoter may span a much larger region due to redundancy in the function of regulatory elements that ensure proper and robust regulation of the endogenous gene. One common site of additional control elements is within introns. Most C. elegans introns are small (e.g., <100 bp; see Alternative splicing in C. elegans) and are thus unlikely to contain elements controlling expression. However, introns larger than several hundred base pairs do often have such elements (e.g., Nam et al., 2002; Okkema et al., 1993). Therefore, intron size can provide a clue in searching for transcriptional control sequences. Large introns, particularly at the beginning of a coding region, may also provide a clue to promoter organization and the presence of multiple transcriptional initiation sites. For example, nhr-23 has a 1.8 kb intron at the start of the gene that is included in one transcript and absent in a second (Kostrouchova et al., 1998). In cases such as this, the presence of a trans-spliced leader (see Trans-splicing and operons) on two or more different transcripts from a single gene can be an indicator of multiple messages, possibly encoding different protein isoforms.

A simple promoter is defined here as one in which the cis-acting control elements necessary for proper expression are confined to a small region (a few hundreds of bp) of the genome. Housekeeping genes expressed in all tissues might be good candidates for regulation by simple promoters, Unfortunately, few housekeeping genes in C. elegans have been characterized. Among the best characterized simple promoters are those of the hsp-16 family of genes. This family consists of pairs of divergently transcribed genes with promoter regions sufficient for heat-regulated expression contained within the short (~350 bp) intragenic regions (Jones et al., 1986; Russnak and Candido, 1985; Stringham et al., 1992). Despite these compact promoters, distinct tissue expression patterns are induced from different hsp-16 promoters (Stringham et al., 1992), suggesting the presence of multiple regulatory sites within these simple promoters. Another excellent example of simple promoters are in the vitellogenin (vit) genes, which exhibit stage-, tissue- and sex-specific expression controlled, in the case of vit-2, by a 247 bp promoter (MacMorris et al., 1992; MacMorris et al., 1994). vit-2 promoter activity depends on GATA-factor binding sites and a novel VPE2 site (TGTCAAT) conserved in vit gene promoters in C. elegans and C. briggsae (Spieth et al., 1985; Zucker-Aprison and Blumenthal, 1989). Certain cell cycle promoters have also been shown to be remarkably simple. Analysis of several genes expressed only in proliferative cells and encoding G1 phase regulators (e.g., cyclin D) revealed that proper regulation minimally required a 67 bp region of the promoter (Brodigan et al., 2003; Park and Krause, 1999). How could genes with such dynamic expression profiles throughout development be regulated in an apparently simple way? The answer is likely that they are end effectors of a cell's decision to divide rather than integrating lineage or temporal information governing proliferation.

The term complex is used here to describe a promoter in which the overall pattern of gene expression is the result of the composite action of several dispersed elements, each influencing or contributing to the overall expression pattern. This piecemeal organization has been described for the promoter region of several genes, including myo-2, hlh-1 and lin-26. These studies reveal examples in which spatial control of transcription is regulated by elements active in groups of cells related by cell-, tissue- and organ-type and by lineage history.

The completion of the C. elegans genome makes it possible, in theory, to define all transcription factors in the worm. In practice, this effort is more difficult because of several uncertainties when surveying the properties of a given gene. For example, zinc finger motifs can bind DNA but also can serve other functions including RNA binding and protein-protein interactions. It is therefore difficult to conclude that a given gene product is indeed a transcription factor based solely on the presence of signature motifs. For factors that modify chromatin or participate in a transcription complex, the definition of a transcription factor often lies in the eyes of the investigator. A first-pass attempt at defining a list of C. elegans transcription factors is presented in Table 1. Originally compiled in the Sternberg Lab (courtesy of T. Ririe and J. Fernandes), we present a modified version of their list with the understanding that it will necessarily need refinement over time to correct inaccuracies and omissions. The current list includes 664 genes representing only about 3.5% of the predicted genes in C. elegans. This number is surprisingly low and about one half the number of transcription factors estimated previously (McGhee and Krause, 1997).