Noncoding regulatory sequences of Ciona exhibit strong correspondence between evolutionary constraint and functional importance

Abstract

We show that sequence comparisons at different levels of resolution can efficiently guide functional analyses of regulatory regions in the ascidians Ciona savignyi and Ciona intestinalis. Sequence alignments of several tissue-specific genes guided discovery of minimal regulatory regions that are active in whole-embryo reporter assays. Using the Troponin I (TnI) locus as a case study, we show that more refined local sequence analyses can then be used to reveal functional substructure within a regulatory region. A high-resolution saturation mutagenesis in conjunction with comparative sequence analyses defined essential sequence elements within the TnI regulatory region. Finally, we found a significant, quantitative relationship between function and sequence divergence of noncoding functional elements. This work demonstrates the power of comparative sequence analysis between the two Ciona species for guiding gene regulatory experiments.

Figure 1.

Five prime regions of eight Ciona loci. The black plots represent sequence identity between C. intestinalis and C. savignyi across each region. C. savignyi is the reference sequence for a–d, and C. intestinalis is the reference sequence for e–h. Green vertical bars correspond to annotated C. intestinalis exons. Blue arrows indicate direction of transcription of each gene. The red arrows indicate the native sequence present in the promoter fusion construct. Each promoter fusion construct contains significant regions of similarity between the two species. Expression patterns for C. savignyi constructs (i–o). Troponin I is strongly expressed in the tail muscle (i), Synaptotagmin is expressed in various neural lineages, including epidermal neurons (EN) (j) and spinal cord (SC) (k), α-tubulin is expressed in various tissues, including central nervous system (CNS) (l) and spinal cord (SC) (m), and ectopic expression frequently occurs in the mesenchyme (MC) (m), Noto9 is expressed specifically in the notochord (n,o). All embryos were fixed at the mid–late tailbud stages. Electroporated embryos always exhibit mosaic staining patterns. A variety of typical images are archived at http://mendel.stanford.edu/supplementarydata/johnson.

Figure 2.

Ciona Troponin I gene expression. (a–h) Expression of Troponin I transcript as determined by in situ hybridization. Expression is weak as early as gastrulation(a), is very strong by initial tailbud (b), and continues in the muscle through mid-tailbud (c), late tailbud (d), tadpole (e), and juvenile stages (f–h). Note the expression in the trunk ventral cells (TVC; adult heart precursors), during the late tailbud stage in e. (i–n) Expression of lacZ reporter construct tn.pro.1793 as determined by X-gal staining. Expression is strong in mid tailbud (i,j), and continues through late tailbud (k), early tadpole (l), and juvenile stages (m). Ectopic expression in the mesenchyme is common (arrowhead, i). Expression continues in the tail muscle from the late tailbud (k) to the early tadpole (l). Expression also occurs in the TVC (arrowhead, m) and juvenile heart (arrowhead, n).

Figure 3.

Activity of Troponin I–lacZ promoter constructs bearing various deletions, in context with sequence identity and exon structure of the locus. (a) Plot of sequence identity between C. savignyi and C. intestinalis in the promoter construct pro.1793 (cf. Fig 1a). Green shading represents exons, the blue bar represents the MSRR, and the red bars indicate highly constrained regions (90% id; >20 bp). (b) Activity of deletion constructs, with tn.pro.1793; black lines denote sequence present in the constructs. The table lists activities of the constructs as explained below. Truncation of 329 bp from the 5′-end of the tn.pro.1793 construct completely eliminates expression in the tail muscle. Deletion of internal regions, including annotated exon 1, exons 1 and 2, and exons 1–3, has no effect. The gray line (tn.ci.pro.921) denotes the construct derived from C. intestinalis. (c) Activity of minimally sufficient regulatory region (MSRR) constructs. Fusion of the 5′-most 363 bp of tn.pro.1793 to a heterologous Forkhead basal promoter has strong activity. Deletion of the Forkhead basal promoter has no effect on lacZ expression. The reversed insert has no activity. (d–f) Relative activities of constructs. We classify constructs as (d) +++, representing 50%–100% animals staining, with many of the animals staining strongly in a majority of the tail muscle cells; (e) ++, representing 25%–50% of the animals staining in the tail muscle, and the minority of these staining a majority of the tail muscle cells; and (f) +, with <25% of the animals staining, and none of the animals staining the majority of the tail muscle cells; “-” indicates a construct that never showed staining. Constructs that were mostly “-” but stained weakly on a single occasion are denoted +/-.

Figure 4.

Deletion analysis of the minimally sufficient regulatory region. (a) The CHAOS algorithm returns four major regions of local conservation between C. savignyi and C. intestinalis. Note that this level of conservation stands out over neutrally evolving regions. (b) Plot of sequence identity between the C. savignyi and C. intestinalis TnI enhancer regions. The yellow shaded boxes represent the four CHAOS matches. (c) Horizontal lines denote sequence present in the constructs, activity is noted as in Figure 3. CHAOS matches CH1, CH2, and either of CH3/CH4 are necessary for expression of lacZ reporter in larval tail muscle. Concatenation of the four CHAOS matches is sufficient for weak lacZ expression.

Figure 5.

Conservation, predicted binding sites, and activity upon mutagenesis in the TnMSRR. (a) Plot of predicted binding sites for four common vertebrate muscle regulators for the TnI locus. The right y-axis shows the log-odds score of the site. The black line plots sequence identity (left y-axis) between C. savignyi and C. intestinalis. The location of the enhancer (tn.msrr.363) is shown. (b) Plot of conservation versus activity upon mutagenesis, with predicted binding sites. The left y-axis is sequence identity (black) and activity on mutagenesis (red), and the right y-axis is the log-odds score of predicted sites. The black line plots sequence identity between the C. savignyi and C. intestinalis enhancers. The red line represents the activity (number of embryos staining/total embryos) of the lacZ reporter when a 20-bp window spanning that position is randomized. The yellow shading represents CHAOS hits, the green line is a conserved TATA-box, the pink arrow is the predicted transcription start site, and the blue lines are the three myf predictions. (c) Scatterplot of activity upon mutagenesis versus distance. The distance within each 20-mer is calculated by Kimura's two-parameter model. Mutagenized windows with activity >50% are not included, nor are poorly aligned sequences. (d) Predicted myf sites in C. savignyi and C. intestinalis, aligned with the two relevant scrambled windows. A predicted site remains in scrambled window 1, which retains activity, and no sites are predicted in scrambled window 2, which disrupts activity. Reverting one of the predicted sites back to wild type restores function (reverted window 2). Sequence of C. savignyi (Cs) is shown on top, with capital letters indicating identity with C. intestinalis. Blue sites are predicted on the forward strand, red sites on the reverse. All images and sequences are archived for retrieval at http://mendel.stanford.edu/supplementarydata/johnson.