The transcriptomes of two heritable cell types illuminate the circuit governing their differentiation

Abstract

The differentiation of cells into distinct cell types, each of which is heritable for many generations, underlies many biological phenomena. White and opaque cells of the fungal pathogen Candida albicans are two such heritable cell types, each thought to be adapted to unique niches within their human host. To systematically investigate their differences, we performed strand-specific, massively-parallel sequencing of RNA from C. albicans white and opaque cells. With these data we first annotated the C. albicans transcriptome, finding hundreds of novel differentially-expressed transcripts. Using the new annotation, we compared differences in transcript abundance between the two cell types with the genomic regions bound by a master regulator of the white-opaque switch (Wor1). We found that the revised transcriptional landscape considerably alters our understanding of the circuit governing differentiation. In particular, we can now resolve the poor concordance between binding of a master regulator and the differential expression of adjacent genes, a discrepancy observed in several other studies of cell differentiation. More than one third of the Wor1-bound differentially-expressed transcripts were previously unannotated, which explains the formerly puzzling presence of Wor1 at these positions along the genome. Many of these newly identified Wor1-regulated genes are non-coding and transcribed antisense to coding transcripts. We also find that 5' and 3' UTRs of mRNAs in the circuit are unusually long and that 5' UTRs often differ in length between cell-types, suggesting UTRs encode important regulatory information and that use of alternative promoters is widespread. Further analysis revealed that the revised Wor1 circuit bears several striking similarities to the Oct4 circuit that specifies the pluripotency of mammalian embryonic stem cells. Additional characteristics shared with the Oct4 circuit suggest a set of general hallmarks characteristic of heritable differentiation states in eukaryotes.

Figure 1. RNA sequencing of white and opaque cells.

(A) White and opaque cells have distinct morphologies. (B) Summary of experimental design. (C) Summary statistics for alignments of RNA sequence reads. Read counts listed are expressed in millions (left column) or as a percentage of the total reads processed (right column) for each sample.

Figure 2. Defining a new transcript annotation for C. albicans.

(A) Summary of computational workflow. (B) Summary statistics comparing the old ORF-based and new RNA-Seq-based transcript annotations.

Figure 3. Transcripts differentially expressed between white and opaque cell types.

(A) Expression and Wor1 enrichment at the WOR1 locus as visualized in the MochiView Genome Browser . In this and all other genomic plots presented here, Wor1 ChIP-Chip data are plotted in the top row (red-curves are from biological replicates of the Wor1 IP in opaque cells and orange curves are from IPs in wor1Δ Δ strains; normalized log₂ IP DNA/Input DNA enrichment values are plotted), followed by RNA-Seq data for white and opaque cells (colored green on the plus and blue on the minus strand; values plotted are log₂ sequence coverage), followed by transcript definitions in our new annotation (gray regions are coding and white are un-translated), and finally regions determined to be Wor1-bound by the peak finding algorithm (gray boxes). For interested readers, a MochiView database export of all the data presented in this work is provided at http://johnsonlab.ucsf.edu/mochi_files/Tuch_et_al_2010_PLoS_Genetics.cvw. (B) Comparison of RNA-Seq and microarray measurements of differential transcript expression (for previously annotated transcripts only). Transcripts are colored by their mean abundance across samples as measured by RNA-Seq: purple indicates mean RPKM ≤30 and orange indicates mean RPKM >30. (C) The expression of STE4 is anti-correlated with the expression of its antisense transcript. (D) The expression of sense-antisense transcript pairs is only modestly anti-correlated (ρ = −0.25; P-value = 0.05).

Figure 4. A selection of sense-antisense gene pairs with the most strongly anti-correlated expression.

Each row lists a sense-antisense transcript pair, the differential expression in opaque versus white cells for each transcript in the pair, and whether or not each transcript is Wor1 bound.

Figure 5. Three clusters of novel Candida-specific ORFs are strongly up-regulated in opaque cells.

(A) Expression and Wor1 binding at cluster A, the NTAR_1176 locus (others shown in Figure S6A and S6B), containing 7 novel ORFs on the positive strand, 6 of which are expressed only in opaque cells. (B) Partial multiple sequence alignment of all members of cluster A (see Figure S6C for alignment of all 46 homologs) in C. albicans and C. dubliniensis. The predicted secondary structure is noted in the final row (E  =  β sheet and H  =  α helix) . (C) Compressed neighbor-joining phylogeny of all 46 NTAR_1176 homologs found in C. albicans and C. dubliniensis (see Figure S6D for full tree).

Figure 6. Association of Wor1 binding with white-versus-opaque differential expression when different transcript annotations and measurement platforms are employed.

An RNA-Seq-based annotation with RNA-Seq-based differential expression measurements (top row) provides the strongest concordance between differential expression and Wor1 binding. Footnotes: 1 The transcript annotation derived from RNA-Seq data in this work. 2 The previous ORF-based gene annotation from Candida Genome Database (CGD). 3 Differential expression measurements from RNA-Seq data reported in this work. 4 Differential expression measurements from hybridization to custom Agilent 8x15k microarrays reported in this work. 5 Differential expression measurements from hybridization to spotted cDNA microarrays reported previously . 6 Indicates whether or not a gene expression detection threshold was employed to filter putatively dubious transcripts from the annotation prior to computing the association between binding and differential expression. 7 Indicates whether or not the genes detected as having UTR length changes between the cell types are considered “differentially-expressed.” Note that such genes may or may not be differentially expressed in the traditional sense (i.e., when considering the entire transcript or just the coding region of the transcript).

Figure 7. Properties of transcripts in the Wor1 circuit.

(A) The distribution of lengths for all intergenic regions and Wor1-bound intergenic regions. The distribution of lengths for the (B) 5′ UTRs and (C) 3′ UTRs of all transcripts, transcripts associated with Wor1 binding, and transcripts that are associated with Wor1 binding and differentially expressed (“dEx-ed”) between white and opaque cells. Expression and Wor1 binding at three genes with apparent changes in UTR length between the cell types: (D) ORF19.2049, (E) EFG1, and (F) ORF19.7060. (G) Expression and Wor1 binding at the NTAR_913 locus, an example of a gene for which down-regulation in opaque cells is correlated with overlapping binding of Wor1. (H) The distribution of differential expression (opaque versus white fold-changes) for all transcripts, transcripts associated with Wor1 binding, and transcripts that are directly overlapped at least 50% or 100% by Wor1 binding. The gray dashed oval highlights an enriched subset of transcripts for which overlapping Wor1 binding is correlated with down-regulation in opaque cells.

Figure 8. Properties of transcripts in the Oct4 circuit.

(A) The distribution of lengths for all intergenic regions and intergenic regions that are associated with Oct4 binding. The distribution of lengths for the (B) 5′ UTRs and (C) 3′ UTRs of all transcripts, transcripts associated with Oct4 binding, and transcripts that are associated with Oct4 binding and differentially expressed (“dEx-ed”) during differentiation.