Module TSS4: Annotation of Broad Transcription Start Sites

In the previous modules in this series, we learned how to access information used to annotate a Transcription Start Site (TSS) for genes with a peaked promoter (Figure 1). However, genes with intermediate or broad promoters are often more difficult to annotate. These promoters have multiple TSSs and several lines of evidence must be used to define a search region for the TSS. This module will use several examples to illustrate strategies that you can use to define a search region and locate (annotate) the TSS or TSSs for these genes using the available evidence. In particular, F element genes tend to have a higher probability of having broad promoters, making the challenge of annotating their TSSs more difficult. Moreover, on occasion you will run into a peaked promoter that is similarly troublesome to annotate and the strategies presented in this module can be applied to these genes.

As part of the GEP TSS annotation projects, we marshal different lines of evidence to define the boundaries of the TSS search region for any given gene. Ideally, the data from these different sources should be in agreement. However, not all lines of evidence are equal.

The different lines of evidence that can be used to establish TSS search region boundaries are ranked below from most to least reliable:

blastn alignment of the 1^st transcribed D. melanogaster exon: The gold standard evidence for the location of the TSS for a GEP project gene is homology to the 1^st transcribed exon of the ortholog in D. melanogaster. The blastn match should have a low E-value with a high sequence identity between the 1^st transcribed exon in D. melanogaster and the GEP project sequence. In addition, the TSS shouldn’t need to be extrapolated by more than 150 bp from the end of the aligned sequence; and other lines of evidence (such as those described below) should agree with the proposed TSS position.

If blastn succeeds at finding the location of the first exon in the project sequence, then define the boundaries of the TSS search region as +/- 300 bp from the initial 5' nucleotide in exon 1. Typically, this will be the narrow search window annotation. Using the information contained in the tracks listed below may allow you to define a narrower search window, but most of the time if you find a blastn match, your annotation adventure can stop here!

Of course, blastn won’t always identify the location of the 1^st transcribed D. melanogaster exon (when it fails one or more of the criteria listed above). Fortunately, other lines of evidence can be utilized to define a TSS search region.

RNA Pol II X-ChIP-Seq: RNA Polymerase II (RNA Pol II) is the enzyme that transcribes protein-coding genes in eukaryotes and is often enriched near the TSS of a gene isoform. Researchers have used a technique called ChIP-Seq to identify regions in the genome where RNA Pol II is in fact enriched. Very briefly, chromatin (DNA and proteins) is fragmented to small size. RNA Pol II and the DNA that it is bound to is precipitated using an antibody specific to RNA Pol II. The DNA is then sequenced, and the reads are mapped back to the genomic sequence. Because the sequenced DNA corresponds to sites where RNA Pol II was bound, locations where RNA Pol II is enriched can be identified by a pile-up of reads called peaks (see the RNA Pol II Enrichment track).

The program that identifies peaks reports both the boundaries of the peak (grey bar), and a peak apex (vertical red line) where the pile-up is the highest (Figure 7). These peaks are a good proxy for the region that includes the TSS and can be used as very high-quality landmarks for defining a TSS search region. When defining a search region, the boundaries of the RNA Pol II peaks should be used rather than the peak apex.

RNA-Seq (including TopHat junctions): In RNA-Seq all of the messenger RNA is sequenced, and the reads are mapped back to the genome to identify the location of protein-coding genes. Furthermore, the program TopHat identifies reads that span intron exon junctions. These reads are displayed in the TopHat reads tracks.

The RNA-Seq tracks can be very useful for determining a TSS search region. Often when blastn fails to identify the location of the 1^st transcribed D. melanogaster exon, the RNA-Seq and TopHat tracks can be used to identify the approximate location of the first exon. Moreover, these tracks can be used to define a TSS search region. A tail of continuous RNA-Seq enrichment upstream of the initial (predicted) exon can be used to define the upstream boundary of a search region because this is indicative of where transcription for the gene began. In addition, TopHat junctions can be used to define the downstream boundary of a search region because, by definition the TSS must be upstream of the 3' splice site of the first exon. See examples of both of these types of regions in the boxed areas in Figure 8.

Conservation: Sequences in the genome that serve a function, like the sequence of a gene, tend to be conserved over evolutionary time. Therefore, we would expect that promoter and exon sequences would tend to be more highly conserved than surrounding sequences. The conservation tracks highlight conserved nucleotides between Drosophila species. If a nucleotide is conserved it is shaded black, so the more black nucleotides, the higher the degree of conservation (Figure 9). The boundaries of highly conserved regions can serve as landmarks for defining the boundaries of a TSS search region. However, this is the lowest quality evidence you can use for this task and should only be used if the other lines of evidence are completely lacking.

You can find more technical information about the evidence tracks described above by clicking on the text links for each track to read a brief summary of the track, and by reading the primary literature sources found on the track description page.

We use combinations of these evidence tracks to define the narrowest TSS search window possible (given the available evidence). A broad search window (600 bp or more) can also be defined by using landmarks in the different evidence tracks that are further apart from each other. In any event, both the narrow and broad search regions should be anchored to clear landmarks in the evidence tracks.

A quick note before we move on to some practice: In general, the alignments to D. melanogaster proteins and transcripts (i.e., the "D. mel Proteins" and "D. mel Transcripts" evidence tracks) make poor landmarks for annotation. The sequences of the untranslated regions of genes tend to diverge faster over evolutionary time than the coding regions of genes. For this reason, alignments are often only partial — particularly at the 5' end of the first transcribed exon where we expect to find the gene promoter. Hence, these alignments should not be used as annotation landmarks.

In exercise 1 we determined that the TSS of the B isoform of the CG46466 gene in D. melanogaster had a broad shape. Now that we have learned how to use evidence-based landmarks to set the boundaries of a TSS search region, let’s apply that knowledge to define a TSS search region for the CG46466-RB gene in the D. biarmipes GEP project sequence.

First, we will determine if we can locate the 1^st transcribed D. melanogaster exon using BLAST.

Now we can see displayed the beginnings of the coding DNA sequence (CDS) of the CG46466 isoforms (red tracks at the top) along with the Spaln alignment of D. melanogaster transcripts (purple tracks) (Figure 12). Spaln is a program that aligns the RNA sequences of many genes from a species (in this case D. melanogaster) against the genome of another (in this case the GEP D. biarmipes project sequences). The Spaln alignments can help us zero in on a region that may contain the TSS for a gene, but as mentioned above should not be used to set the boundaries of a search region.

Now we can clearly see the region surrounding the CG46466-RB exon (Figure 13).

Finally, a nice feature of the D. biarmipes projects is that all of the consensus promoter motifs have been identified and assembled into an individual track that can be displayed in the browser. Therefore, we can visualize all of these motifs in a single step, without needing to search for each of them individually using Short Match (Figure 15). There are individual tracks for the motifs found on either the plus or minus strand. Since the CG46466 gene is encoded on the minus strand of contig38 we will only display the motifs on that strand. Bring up this track by configuring the browser display modes as follows:

In this module we have learned how to utilize different lines of evidence to annotate a TSS search region for broad promoter type genes, particularly when blastn fails to locate the 1^st transcribed D. melanogaster exon in your project sequence. The strategy presented here relies on making use of the RNA Pol II X-ChIP-Seq data, RNA-Seq and TopHat data and Conservation to rationally assign the boundaries of a TSS search region using evidence-based landmarks in these data types. This strategy should be applicable to a wide range of GEP project TSS annotation problems.

If you would like further practice annotating this type of TSS, try doing CG33941 in either the D. biarmipes or D. elegans projects.