Saccharomyces cerevisiae can divide asymmetrically so that the mother and daughter cells have different fates. We show that the RNA-binding protein Khd1 regulates asymmetric expression of FLO11 to determine daughter cell fate during filamentous growth. Khd1 represses transcription of FLO11 indirectly through its regulation of ASH1 mRNA. Khd1 also represses FLO11 through a post-transcriptional mechanism independent of ASH1. Cross-linking immunoprecipitation (CLIP) coupled with high-throughput sequencing shows that Khd1 directly binds repetitive sequences in FLO11 mRNA. Khd1 inhibits translation through this interaction, establishing feed-forward repression of FLO11. This regulation enables changes in FLO11 expression between mother and daughter cells, which establishes the asymmetry required for the developmental transition between yeast form and filamentous growth.

ASYMMETRIC cell division produces two cells with different developmental fates (Horvitz and Herskowitz 1992). The unequal inheritance of cell fate determinants establishes this asymmetry in many systems through diverse mechanisms that ultimately produce asymmetric gene expression between cells (Macara and Mili 2008). In multicellular eukaryotes, this process directs a cell lineage down a developmental path. In Saccharomyces cerevisiae, each mitotic division requires a new decision to determine the fate of the daughter cell, providing a tractable model to study the underlying mechanisms of asymmetric cell division.

The RNA-binding protein Khd1 (KH-domain protein 1) regulates the asymmetric expression of ASH1 in budding yeast to control mating-type switching, a key developmental event in haploid cells (Irie et al. 2002; Paquin et al. 2007; Hasegawa et al. 2008). Ash1 protein accumulates specifically in the nuclei of daughter cells (Bobola et al. 1996; Sil and Herskowitz 1996). Genetic and biochemical analysis led to the model that Khd1 represses translation of ASH1 mRNA during transport to the bud tip, where phosphorylation by Yck1 reduces the affinity of Khd1 for the transcript, relieving repression and allowing translation to occur (Long et al. 1997; Chartrand et al. 2002; Irie et al. 2002; Paquin et al. 2007). As Ash1 is a transcription factor that represses mating-type switching, translational repression of ASH1 mRNA in the mother but not the daughter leads to asymmetry—the mother can switch mating type, but the daughter cannot (Strathern and Herskowitz 1979; Chartrand et al. 2002; Paquin and Chartrand 2008).

ASH1 has also been implicated in the regulation of filamentous growth, another developmental event in S. cerevisiae (Chandarlapaty and Errede 1998). Under conditions of nitrogen starvation, diploid cells enact a specialized growth program characterized by an elongated morphology and unipolar budding that leads to the formation of filaments (Gimeno et al. 1992). The transition to filamentous growth requires an asymmetric cell division, as a yeast-form mother cell produces a filamentous daughter cell. ASH1 regulates filamentous growth by activating expression of FLO11 (Pan and Heitman 2000), which encodes a cell wall protein required for this growth form (Lambrechts et al. 1996; Lo and Dranginis 1998). Cells induce FLO11 expression to activate filamentation in response to nitrogen starvation (Lo and Dranginis 1998). Deletion of ASH1 prevents both FLO11 expression (Pan and Heitman 2000) and the transition to filamentous growth (Chandarlapaty and Errede 1998).

Khd1 has no known role in regulating filamentous growth. However, since Khd1 represses ASH1 in the context of mating-type switching, it may regulate ASH1 during filamentation as well. Given that RNA-binding proteins can coordinate the expression of mRNAs encoding functionally related proteins (Keene 2007), Khd1 may regulate additional genes in the filamentation pathway. Microarray analysis following immunoprecipitation of Khd1 has been used to identify its mRNA targets (Hasegawa et al. 2008; Hogan et al. 2008), but the strains used do not transcribe FLO11 mRNA (Liu et al. 1996) and the binding of Khd1 to mRNAs of the filamentation pathway such as FLO11 would not have been detected.

The ability to comprehensively define post-transcriptional regulatory networks has been enormously advanced by the cross-linking immunoprecipitation (CLIP) method. CLIP utilizes UV radiation to crosslink an RNA-binding protein to its direct RNA targets in vivo, providing a snapshot of binding interactions. Direct sequencing of the RNAs following RNAse treatment localizing binding sites to a 60- to 100-nucleotide region within target transcripts (Ule et al. 2003). CLIP has been used in combination with high-throughput sequencing to comprehensively identify RNA targets of mammalian RNA-binding proteins (Licatalosi et al. 2008; Sanford et al. 2009; Yeo et al. 2009), but has not been previously applied to yeast.

In this report, we use genetic analysis and CLIP coupled with high-throughput sequencing to determine the role of Khd1 in regulating filamentous growth. We find that Khd1 regulates both transcription and translation of FLO11 to repress filamentation. Khd1 represses FLO11 at the transcriptional level through its inhibition of ASH1, as we predicted based on published regulatory interactions (Chandarlapaty and Errede 1998; Pan and Heitman 2000; Irie et al. 2002; Paquin et al. 2007; Hasegawa et al. 2008), and at the post-transcriptional level by directly repressing translation of FLO11 mRNA. The feed-forward regulation of FLO11 by Khd1 provides a dynamic mechanism for generating asymmetric expression and determining daughter cell fate following cell division. FLO11 mRNA is the predominant unique transcript bound by Khd1, indicating that this regulation is a primary function of the protein. Khd1 binds to repeated sequences in the coding region of FLO11 mRNA and mRNAs encoding many other cell surface proteins, suggesting that this RNA binding protein may coordinate the synthesis of many disparate proteins that assemble into the cell wall.


Yeast strains, media, and growth conditions:

All yeast strains used in this study are derived from Σ1278b and listed in supporting information, Table S1. Standard yeast media, yeast transformations, and genetic manipulations were performed as previously described (Guthrie and Fink 2001). To induce filamentation, strains were grown on nitrogen-poor SLAD media (Gimeno et al. 1992). Approximately 20 cells per strain were spotted onto a SLAD plate in 50 μl of water to compare filamentation under comparable conditions. To assay agar adhesion, 106 cells were spotted onto a YPD plate in 5 μl and grown for 3 days at 30° prior to washing. Yeast strains carrying gene deletions were constructed by PCR amplification of kanamycin-resistance gene cassettes from the yeast deletion library (Winzeler et al. 1999) with approximately 200 bases of flanking sequence and transformation into Σ1278b. Yeast strains carrying TAP-tagged Khd1 were similarly constructed by PCR amplification of the KHD1-TAP∷HIS3 construct from the TAP-tag library (Ghaemmaghami et al. 2003) and transformation into Σ1278b. Strains carrying PADH or PCYC1 (Janke et al. 2004) were constructed by PCR amplification with primers containing 50 bp of homology to the target locus and transformation into Σ1278b. Strains carrying GFP∷ADH 3′ UTR∷URA3 or ADH 3′ UTR∷URA3 were similarly constructed using a plasmid provided by Sherwin Chan. See Table S2 for primer sequences.

Plasmid construction:

The Khd1 overexpression construct was made by amplifying the gene using PCR, with oligonucleotides that added restriction sites (NotI at the 5′ end, XhoI at the 3′ end) to the final product (Table S2). Amplified DNA was digested using NotI and XhoI and cloned into p413TEF (Mumberg et al. 1995).

Flow cytometry and immunofluoresence:

Single colonies were picked after 2 days of growth on YPD plates and resuspended in 1.5 ml liquid YPD. Cells were inoculated into 10 ml liquid YPD and grown for 18 hr to OD600 0.13–0.16, washed twice with PBS, and resuspended in 50 μl PBS containing 1 μl Alexafluor 488-conjugated anti-hemaglutinin antibody (Molecular Probes A-21287) per 200 μl PBS. Cells were incubated 30 min at 4° and washed three times in PBS prior to flow cytometry using the BD FACSCalibur, or imaging with the Nikon Eclipse TE2000-S.


Total RNA was obtained by standard acid phenol extraction from 1 ml of cultures grown to OD600 0.9–1.1 in YPD. The Qiagen QuantiTect Reverse Transcription Kit was used to remove residual genomic DNA and reverse transcribe the RNA templates to generate cDNAs. Aliquots of cDNA were used in Real Time PCR analyses with reagents from Applied Biosystems and the ABI 7500 real-time PCR system.

Immunoprecipitation for measuring RNA enrichment:

TAP tag immunoprecipitation and RNA isolation was performed as previously described (Gerber et al. 2004), using 200 ml of starting culture rather than 1 liter and proportionately fewer reagents.

Cross-linking immunoprecipitation:

Khd1–TAP was purified from 1 liter of cells grown to OD600 2.5 and UV crosslinked three times at 400 mJ/cm2. Purification using calmodulin sepharose was followed by binding to magnetic IgG beads (File S1). The CLIP protocol was then followed as previously described (Ule et al. 2005). The resulting cDNA was amplified using PCR with oligonucleotides containing sequences for hybridization to the Illumina flow cell (Table S2).

Illumina sequencing:

Samples were sequenced using Illumina sequencing with a custom primer (Table S2), returning 16,026,920 36-nucleotide-long reads. Reads containing unresolved bases (N) were ignored. The complete set of reads contained 6,324,854 unique sequences. All reads were mapped to the Σ1278b genome (Dowell et al. 2010) using Novoalign (v1.05; second September 2008) with default settings. All mappings are included, weighted inversely by the number of genomic locations to which a read maps. The reads have been deposited in the Sequence Read Archive under accession no. SRA012416.

Peak calling:

The peak caller uses a rolling window approach (10-base windows; 5-base offset) to compare the observed reads to those expected from a Poisson background model. Adjacent enriched windows are combined into peaks. Peaks are assigned to genes on the basis of overlap with existing annotation, extending 500 nucleotides in each direction (unless the extension overlaps adjacent annotation) to account for UTRs.

A local (5 kb) window is used to parameterize the background model. A visual examination of the read mappings relative to available tiled expression data (Danford et al. 2010) indicates that reads are strand specific and show perfect correspondence with expressed segments, indicating the background of possible RNA binding sites is the transcriptome, not the genome. A weak correlation is observed between the expression levels of a transcript and the number of observed reads.

We set a peak cutoff by maximizing the correspondence of gene targets predicted relative to the targets reported by Hasegawa et al. (2008). The peaks are weighted by the corresponding expression level of each transcript, as determined from tiled expression data (Danford et al. 2010). Only peaks containing at least 50% of the reads of the transcript's maximal peak size are considered.

Motif discovery:

Three methods were utilized to identify the motif recognized by Khd1. First, MEME (v4.1; Bailey and Elkan 1994) was utilized on the sequences under the peaks, filtering to remove highly identical sequences (80% identity). Second, all k-mers were evaluated (for k = 1, 2, 3, 4) to identify overrepresented sequences under the peaks. Random nonpeak windows of matching length were selected from the same set of transcripts as the peaks to calculate the distribution of background k-mers. Finally, “RNApromo” (Rabani et al. 2008) and CMfinder (Yao et al. 2006) were applied to the peaks to search for potential secondary structure. The structure motifs returned were single-strand loops with sequence patterns consistent with the primary sequence motif identified by MEME. Presence of the discovered MEME motif within the peak list was determined using MAST (v4.1; Bailey and Gribskov 1998) with default parameters.

Western blot analysis:

Protein was prepared using TCA precipitation from 3 ml of culture grown to OD600 of 0.9–1.1, resuspended in 150 μl SDS loading buffer, and boiled for 10 min. Ten microliters were run on a 10% SDS–polyacrylamide gel and transferred to nitrocellulose filter paper. Blotting against GFP was performed with mouse anti-GFP primary antibody (Roche 11814460001) and HRP-conjugated sheep anti-mouse secondary antibody (Amersham NA931V), and against tubulin using rat anti-tubulin (Accurate Chemicals MCA77G) and HRP-conjugated goat anti-rat antibody (Jackson ImmunoResearch 112-035-062). Blots were detected using SuperSignal West femto substrate (Thermo Scientific 34095).


Khd1 has ASH1-dependent and ASH1-independent functions in repressing FLO11:

Given that ASH1 promotes filamentous growth (Chandarlapaty and Errede 1998) by activating transcription of FLO11 (Pan and Heitman 2000) and that Khd1 represses ASH1 in the context of mating-type switching (Irie et al. 2002; Paquin et al. 2007; Hasegawa et al. 2008), we hypothesized that Khd1 regulates filamentous growth. Genetic analysis shows that Khd1 represses filamentation. The khd1Δ/khd1Δ mutant is hyperfilamentous relative to wild type, and cells fail to filament when Khd1 is overexpressed (Figure 1A). The hyperfilamentation phenotype of the khd1Δ/khd1Δ mutant requires FLO11. As is the case with the flo11Δ/flo11Δ mutant, the khd1Δ/khd1Δ flo11Δ/flo11Δ mutant is nonfilamentous (Figure 1B). These findings are consistent with our prediction that Khd1 regulates filamentation by repressing transcription of FLO11 indirectly through its translational repression of ASH1 mRNA.

Figure 1.—

Khd1 represses FLO11-dependent phenotypes independent of ASH1. (A) Khd1 represses filamentous growth in diploid cells. PTEF–KHD1 is an overexpression construct. (B) Khd1 represses filamentation independent of ASH1. (C) ASH1-independent repression of haploid agar adhesion by Khd1. Picture taken of the same plate before and after washing.

However, Khd1 represses filamentation at least in part through an ASH1-independent pathway. The khd1Δ/khd1Δ ash1Δ/ash1Δ double mutant is filamentous, unlike the ash1Δ/ash1Δ single mutant, indicating that Khd1 represses filamentation independent of ASH1 (Figure 1B). This finding extends to haploid agar adhesion, another FLO11-dependent phenotype. Cells deleted for KHD1 adhere more than wild-type cells, and khd1Δ ash1Δ double mutants adhere more than ash1Δ single mutants (Figure 1C). As is the case for filamentation, adhesion of both wild-type and khd1Δ cells requires FLO11 (Figure 1C; Lambrechts et al. 1996; Lo and Dranginis 1998). These data show that Khd1 represses FLO11-dependent phenotypes independent of ASH1.

Given the repression of FLO11-dependent phenotypes by Khd1, we tested whether Khd1 regulates FLO11 expression. To quantify FLO11 expression, we employed a FLO11∷HA allele that permits the measurement of Flo11 protein in individual cells (Guo et al. 2000). Flo11 protein is expressed in a subset of cells in a clonal population because of variegating transcription from the FLO11 promoter (Halme et al. 2004; Bumgarner et al. 2009). Mutations that affect FLO11 mRNA levels and filamentation show a corresponding change in the number of cells containing the FLO11∷HA allele that stain positive using an anti-HA antibody (Halme et al. 2004).

Flow cytometry shows that Khd1 represses Flo11 protein expression. Deletion of KHD1 increases the percentage of diploid cells expressing Flo11 protein (Table 1). In addition, the khd1Δ/khd1Δ cells that express Flo11 protein do so at a higher level than wild-type cells that express Flo11. Similar to its regulation of filamentous growth, Khd1 represses Flo11 protein expression independent of ASH1. Although the populations of khd1Δ/khd1Δ ash1Δ/ash1Δ and ash1Δ/ash1Δ cells that express Flo11 display similar levels of the protein, a higher percentage of khd1Δ/khd1Δ ash1Δ/ash1Δ cells express Flo11 (Table 1). The Flo11 expression data, together with the filamentation and agar adhesion phenotypes, point to an ASH1-independent function for Khd1 in repressing FLO11.

View this table:

Khd1 represses Flo11 protein expression independent of ASH1

To explore the regulation of FLO11 by Khd1, we used qPCR to measure FLO11 mRNA levels. khd1Δ/khd1Δ mutants have increased FLO11 mRNA levels relative to wild type (Figure 2), which indicates that Khd1 represses FLO11 mRNA accumulation. In contrast to its ASH1-independent repression of filamentation and Flo11 protein expression, Khd1 represses FLO11 mRNA levels exclusively through its regulation of ASH1. khd1Δ/khd1Δ ash1Δ/ash1Δ double mutants display the same FLO11 mRNA levels as ash1Δ/ash1Δ single mutants, which are below that of wild type (Figure 2). We conclude that Khd1 represses transcription of FLO11 mRNA through its regulation of ASH1. The restoration of filamentation and increased Flo11 protein expression in khd1Δ/khd1Δ ash1Δ/ash1Δ relative to ash1Δ/ash1Δ, without a concomitant increase in FLO11 mRNA levels, suggests that Khd1 represses FLO11 through a post-transcriptional mechanism as well.

Figure 2.—

Khd1 represses FLO11 mRNA levels through ASH1. FLO11 mRNA levels normalized to ACT1 mRNA. Values are average of four independent experiments. Error reported as standard deviation.

Khd1 binds repeated sequences in the FLO11 open reading frame:

The post-transcriptional regulation of FLO11 by Khd1 suggested that Khd1 might interact with FLO11 mRNA. To address this possibility, we tested whether FLO11 mRNA co-immunoprecipitates with a TAP-tagged version of Khd1. qPCR shows that immunoprecipitation of Khd1–TAP enriches FLO11 mRNA more than 50-fold (Figure 3A). The same immunoprecipitation does not enrich FLO11 mRNA when Khd1 is untagged. Immunoprecipitations testing for an interaction between Khd1 and constructs containing different combinations of the FLO11 open reading frame and untranslated regions indicate that Khd1 interacts with the FLO11 coding sequence (Figure S1).

Figure 3.—

Khd1 binds repetitive sequences in the FLO11 open reading frame. (A) Enrichment of FLO11 mRNA following immunoprecipitation from cells expressing either Khd1–TAP or untagged Khd1. (B) Khd1 target sequences from CLIP map to the FLO11 repetitive element. Histogram of read mappings overlaid on a dot plot highlighting the repetitive region of the FLO11 open reading frame from the ∑1278b genome (http://www.vivo.colostate.edu/molkit/dnadot/, window size = 11, mismatch limit = 1). (C) Enrichment of constructs following immunoprecipitation of Khd1–TAP. Enrichments expressed as the level of the transcript relative to ACT1 mRNA in the immunoprecipitate divided by the level of the transcript relative to ACT1 mRNA in the input. Values are average of four independent experiments. Error reported as standard deviation.

To examine the interaction between Khd1 and FLO11 mRNA further, we identified in vivo RNA binding sites for Khd1 using CLIP in conjunction with high-throughput sequencing (File S2, Figure S2, and Table S3). The CLIP analysis shows that Khd1 interacts directly with repetitive sequences in FLO11 mRNA (Figure 3B). FLO11 mRNA is the most frequently represented unique mRNA in the data set; of the 16 million sequences we generated, 1.97 million derive from Khd1 binding to FLO11 mRNA.

To determine whether the repeated sequences in FLO11 mRNA are sufficient for recognition by Khd1, we generated a construct that isolates the FLO11 repetitive element. Immunoprecipitation of Khd1–TAP enriches a transcript with the FLO11 repeats fused to GFP driven by the ADH promoter (Figure 3C). Because the repeats cause a 10-fold decrease in GFP mRNA levels relative to the ADH promoter driving GFP alone (Figure S3), we used the weaker CYC1 promoter to express comparable levels of GFP without the repeated sequences. GFP mRNA does not enrich in the Khd1–TAP immunoprecipitation when driven by either promoter in the absence of the FLO11 repetitive element (Figure 3C). We conclude that the repeated sequences in FLO11 mRNA are sufficient for recognition by Khd1.

Khd1 represses translation through the FLO11 repetitive element:

We used the construct with GFP fused to the FLO11 repetitive element to test the effect of Khd1 binding to this region. Western blotting shows that GFP protein levels from this fusion construct increase 12-fold in khd1Δ relative to wild type (Figure 4A, compare lanes 1 and 2). qPCR measurements show that Khd1 expression causes a 2-fold decrease in mRNA levels from this construct (Figure 4B compare lanes 1 and 2). We attribute the remaining 6-fold difference in GFP protein levels relative to GFP mRNA levels between wild type and khd1Δ to translational repression that results from Khd1 binding the FLO11 repetitive element. Khd1 overexpression further represses the construct with the FLO11 repeats fused to GFP, reducing the amount of GFP protein below that seen with the empty vector, without decreasing GFP mRNA levels (Figure 4, A and B, compare lanes 1 and 3). Neither deletion nor overexpression of Khd1 affects protein or mRNA levels from constructs lacking the FLO11 repetitive element (Figure 4, A and B, lanes 5–8, and Figure S4). In addition to repressing transcription of FLO11 by regulating ASH1 expression, Khd1 represses translation through its interaction with repeated sequences in FLO11 mRNA.

Figure 4.—

Khd1 represses translation through the FLO11 repeats. (A) Western blot analysis of GFP protein levels from constructs expressing GFP alone, or GFP fused to the FLO11 repetitive sequences. PTEF–KHD1 is an overexpression construct. The only visible band detected from wild type, and the predominant band from khd1Δ, migrate at the same molecular weight as GFP alone, suggesting that translation initiated at the GFP start codon. The higher migrating band from khd1Δ may result from low levels of translation initiation inside the repetitive element that become visible after derepression. (B) GFP mRNA levels normalized to TUB1 mRNA levels for the strains shown in A. Values are average of four independent experiments. Error reported as standard deviation.

Translational repression of the fusion construct is consistent with the post-transcriptional repression of Flo11 protein expression by Khd1. Although Khd1 does not appear to regulate endogenous FLO11 mRNA levels independent of ASH1 (Figure 2), mRNA levels from the construct with the FLO11 repeats fused to GFP increase in the khd1Δ mutant (Figure 4B). The fusion transcript may be subject to different regulation than FLO11 mRNA independent of Khd1. Alternatively, low levels of FLO11 mRNA in the ash1Δ/ash1Δ mutant may preclude detection of small changes in stability. To test FLO11 mRNA stability, we used the ADH promoter to transcribe full-length FLO11 mRNA and measured its steady-state levels, similar to our measurement of mRNA from the fusion construct. In the khd1Δ mutant, FLO11 mRNA levels from this construct are 63% of those in wild type. Changes in mRNA stability alone do not explain the differences between mRNA and protein levels for either the fusion construct or endogenous FLO11 in the absence of Khd1. Therefore, translational repression through the repeats is the predominant post-transcriptional regulation of FLO11 mRNA by Khd1.

Khd1 regulates Flo11 asymmetry:

Flo11 protein expression determines daughter cell fate during filamentous growth. To determine whether the transcriptional and translational regulation by Khd1 affects Flo11 expression between mother and daughter cells, we scored Flo11 expression patterns using the FLO11∷HA allele and fluorescence microscopy. The four possible expression patterns between mother and daughter cells were each observed (Figure 5A). Mother cells that express Flo11 can give rise to daughter cells that also express the protein, or those that switch Flo11 expression off. Reciprocally, mother cells that do not express Flo11 can produce daughter cells that similarly do not express the protein, or those that switch Flo11 expression on. We calculated probabilities for daughter cell Flo11 expression given the Flo11 expression of the mother cell based on the frequencies of these expression patterns

Figure 5.—

Khd1 regulates mother–daughter Flo11 expression. Fluorescence microscopy was used to visualize Flo11 protein expression from the FLO11∷HA allele. (A) Flo11 expression patterns in mother–daughter pairs. (B) Khd1 affects the frequency at which daughter cells express Flo11 protein. The chance that a mother cell gives rise to a daughter cell expressing Flo11 protein increases when KHD1 is deleted, independent of ASH1 and whether or not the mother cell expresses Flo11 protein. The frequency of a daughter cell expressing Flo11 protein being produced from a mother cell that expresses Flo11 protein was determined by dividing the number of these mother–daughter pairs by the total number of pairs in which the mother expresses Flo11 protein. The frequency of a daughter cell expressing Flo11 protein being produced from a mother cell that does not express Flo11 protein was determined by dividing the number of these mother–daughter pairs by the total number of pairs in which the mother does not express Flo11. Two hundred and fifty mother-daughter pairs were analyzed per genotype in each of nine separate trials. Error reported as standard deviation.

Repression by Khd1 reduces the frequency of Flo11 expression in daughter cells. Compared to wild-type daughter cells, khd1Δ/khd1Δ daughter cells are more likely to express Flo11 protein whether or not it is expressed in the mother (Figure 5B). These increases result from the loss of the combined transcriptional and translational repression of FLO11 by Khd1. More khd1Δ/khd1Δ ash1Δ/ash1Δ daughter cells than ash1Δ/ash1Δ daughter cells also express Flo11 protein whether or not it is expressed in the mother (Figure 5B). These increases result solely from the loss of translational repression by Khd1, since the deletion of ASH1 inactivates the transcriptional regulation. Although the loss of Khd1-mediated translational repression of FLO11 mRNA increases the expression of Flo11 protein in daughter cells, maximal induction of Flo11 expression in daughter cells, seen in the khd1Δ/khd1Δ mutant, requires the dual relief of both the transcriptional and translational repression of FLO11 by Khd1.

Khd1 binds many mRNAs that encode cell wall proteins:

Khd1 binds a number of mRNAs encoding cell wall proteins in addition to FLO11 mRNA. Fifty-four of the Khd1 target mRNAs we identify using CLIP (Table S4) encode proteins that play a role in cell wall function, nearly half of the 114 genes with this annotation (p = 5.85 × 10−15) (Beissbarth and Speed 2004). Similar to FLO11 mRNA, many of the Khd1 targets that encode cell surface proteins contain repeated sequences. When target genes are sorted by the number of sequences that map to their binding sites, 9 of the top 10—FLO11, SED1, YIL169C, AGA1, SCW10, MSB2, RPO21, CRH1, and YNL190W—contain repeats (reported in Verstrepen et al. 2005 or determined by visual inspection) and 8 of these 9 encode cell surface proteins, with the lone exception being RPO21. With the exception of CRH1 mRNA, Khd1 binds these nine transcripts through their repetitive elements (Figure S5 and Figure 3B), implying that Khd1 frequently binds repeated sequences. Khd1 appears to have a bias for messages with repeated sequences as it binds mRNAs transcribed from 32 of the 44 S. cerevisiae genes previously reported to contain intragenic repeats (Verstrepen et al. 2005).

However, the presence of repeats is not the only determinant of Khd1 binding. First, not all mRNAs bound by Khd1 have repeated sequences. Second, in some cases where Khd1 binds to messages with repeated sequences, the binding is not in the region of repeats (Figure S3, CRH1). Third, Khd1 does not bind all mRNAs that contain repeated sequences.

To understand the determinants of recognition by Khd1, we analyzed the sequences within its binding sites. MEME analysis (Bailey and Elkan 1994) produces a degenerate octamer motif (Figure 6) that occurs in 12% of the Khd1 binding sites. This result is consistent with the CNN repeats found to mediate Khd1 binding in a previous study (Hasegawa et al. 2008). Examination of our motif reveals additional features that may contribute to the interaction between Khd1 and its target RNAs. The repeating CA pattern is similar to the one found in RNAs recognized by the mammalian RNA-binding protein Nova (Buckanovich and Darnell 1997; Jensen et al. 2000; Ule et al. 2003; Licatalosi et al. 2008). Khd1 and Nova both contain three K-homology RNA-binding domains (Currie and Brown 1999; Buckanovich et al. 1993), and structural studies indicate that the third KH domain in Nova makes specific contacts with the internal CA in a YCAY (where Y indicates a pyrimidine, U or C) tetramer (Lewis et al. 2000). CA is the most enriched dinucleotide (1.8-fold relative to background) in the Khd1 binding sites. Two of the four tetranucleotides with the highest enrichments relative to background—CAAC, CUCC, CAUC, and CUAC are enriched 3.3-, 3.0-, 2.9-, and 2.6-fold, respectively—contain CA in the first and second position, but not internally as in the YCAY motif. All four contain C in the first and last position. This analysis identifies new possible determinants of recognition by Khd1, but despite our high-resolution detection of in vivo binding sites, we do not find a motif to explain the specificity of Khd1 for all of its RNA targets.

Figure 6.—

Motif recognized by Khd1. MEME result (Bailey and Elkan 1994) from the sequences within the binding sites identified by CLIP.


Our genetic and biochemical studies show that Khd1 acts post-transcriptionally on two mRNAs to repress FLO11 expression and filamentation. Previous studies showed that ASH1 activates FLO11 expression (Pan and Heitman 2000) and filamentous growth (Chandarlapaty and Errede 1998) and that Khd1 represses translation of ASH1 mRNA in the context of mating-type switching (Irie et al. 2002; Paquin et al. 2007; Hasegawa et al. 2008). Our results demonstrate that Khd1 represses FLO11 expression both through its regulation of ASH1 and by directly inhibiting translation of FLO11 mRNA through repetitive sequences in the open reading frame. This dual inhibition places Khd1 at the head of a feed-forward loop regulating FLO11 (Figure 7) and raises the question of why cells employ this regulatory architecture.

Figure 7.—

Feed-forward regulation of FLO11 by Khd1. Khd1 regulates transcription of FLO11 through its repression of ASH1 mRNA and directly represses translation by binding repeated sequences in the open reading frame of FLO11 mRNA.

The answer may reside in the biology of FLO11, whose function is required to switch from the yeast form to the filamentous form (Lambrechts et al. 1996; Lo and Dranginis 1998; Halme et al. 2004). In the first cell cycle under conditions of nitrogen starvation, over 90% of yeast-form cells produce a filamentous bud (Ahn et al. 1999). The immediate relief of Khd1-mediated translational repression on an existing pool of FLO11 mRNA would allow for the rapid production of Flo11 protein in the first daughter cell even if the mother cell did not express the protein during yeast-form growth. This effect is seen in comparing Flo11 protein expression between ash1Δ/ash1Δ and khd1Δ/khd1Δ ash1Δ/ash1Δ. More khd1Δ/khd1Δ ash1Δ/ash1Δ cells express Flo11 protein than ash1Δ/ash1Δ cells (Table 1), resulting from the higher likelihood that a daughter cell expresses Flo11 protein whether or not it is expressed in the mother cell (Figure 5B). Given that there is not a concomitant increase in FLO11 mRNA levels (Figure 2), this change represents increased translation of FLO11 mRNA upon the loss of Khd1-mediated repression. The rapid inductive response leading to filamentation in the daughter suggests that repression by Khd1 may be quickly relieved under conditions of nitrogen starvation.

A filamentous cell expressing Flo11 protein can divide to produce a yeast-form cell that does not express Flo11 protein (Halme et al. 2004). Such a rapid transition may require inhibition of both transcription and translation of FLO11 mRNA. This dual control would repress preexisting FLO11 mRNA from the mother and prevent the daughter from transcribing new FLO11 mRNA. Khd1 can execute both of these functions to produce asymmetric Flo11 protein expression. Since Flo11 protein is required in the daughter cell to maintain filamentous growth, the increase in Flo11 protein expression when repression by Khd1 is lost in the khd1Δ/khd1Δ mutant (Table 1, Figure 5C) likely explains its hyperfilamentation phenotype (Figure 1A).

This model for asymmetric FLO11 expression and developmental switching posits differential Khd1 activity between cells. This heterogeneity would explain a surprising aspect of the changes in Flo11 protein expression between the ash1Δ/ash1Δ and khd1Δ/khd1Δ ash1Δ/ash1Δ mutants. In the absence of ASH1, the loss of Khd1 enables a higher percentage of cells to express Flo11 protein, but not more of it (Table 1). Individual cells can therefore express Flo11 protein at the same level whether or not they can express Khd1. Because ASH1 is deleted, deletion of KHD1 relieves translational repression on FLO11 mRNA, but does not affect FLO11 transcription (Figure 7). If Khd1 repressed translation of FLO11 mRNA uniformly across all cells, its absence in khd1Δ/khd1Δ ash1Δ/ash1Δ cells would result in increased levels of Flo11 protein. Instead, it appears that some cells containing Khd1 fail to repress translation of FLO11 mRNA, and deletion of KHD1 simply expands this population. Phosphorylation of Khd1 by Yck1 regulates its repression of ASH1 mRNA during mating-type switching (Paquin et al. 2007). Although deletion of YCK1 does not affect filamentous growth (data not shown), post-translational modifications may regulate Khd1 to generate heterogeneous activity and enable the rapid changes in FLO11 expression that underlie asymmetry during filamentous growth.

The asymmetry that arises when a yeast-form mother cell produces a filamentous daughter cell has similarities to the asymmetry of mothers and daughters with respect to mating-type switching. In both morphogenetic events, the mother and daughter have different developmental outcomes dependent on asymmetric gene expression. The two processes also have some differences. One striking difference is that Ash1 activates filamentation but represses mating-type switching, which could reflect the different potentials of the mother and daughter cells between the two processes. The asymmetric expression of ASH1 allows the mother to switch mating type, but prevents the daughter from doing so (Strathern and Herskowitz 1979; Chartrand et al. 2002; Paquin and Chartrand 2008). However, an elliptical yeast-form mother cell already encased in a cell wall of defined structure does not elongate. Instead it is the daughter cell that must express Flo11 protein to develop into a filamentous cell.

The developmental potential of the mother cell is constrained because filamentous growth requires a different program for construction of the cell wall. In this context it may be significant that Khd1 binds 54 mRNAs that encode proteins annotated to function in this macromolecular structure. Post-transcriptional regulation of these genes by Khd1 could provide a unifying mechanism for constructing this organelle. One mechanism for coordinating translational control of these messages would be to have a signature binding site in the mRNAs dedicated to this function. Although we observe a motif consistent with a previous report that used other methods to identify Khd1 binding sites (Hasegawa et al. 2008), we do not identify a sequence that comprehensively explains recognition by Khd1. These data suggest that although the motif we identify contributes to target recognition by Khd1, there must be additional recognition determinants.

Our studies identify a new biological role for Khd1. Its bipartite repression of FLO11 provides dynamic regulation that controls the expression of a cell fate determinant in the daughter cell. Given the prevalence of sequences derived from FLO11 in the CLIP experiment, this likely represents a major function for Khd1. Khd1 binds a number of transcripts that encode cell wall proteins through repetitive sequences in addition to FLO11 mRNA, and Khd1 may regulate the synthesis of many proteins that play a role in this structure. The documented expansion and contraction of the repeats bound by Khd1 (Verstrepen et al. 2005) would generate target sequences of diverse lengths that could be bound differentially, and as a consequence produce altered levels of these cell surface proteins. These changes could have important consequences for the structure and function of the yeast cell wall.


We thank the members of the laboratories of G.R.F. and D.K.G. for discussions; A. Mele and R. Darnell for advice and technical support with CLIP; A. Rolfe for assistance with computational methods; B. Chin for critical reading of the manuscript. This work was supported by National Institutes of Health Grants GM040266, GM069676, and GM035010 and the Abraham Siegel Fellowship and National Science Foundation Graduate Research Fellowship to J.J.W. G.R.F is an American Cancer Society Professor.


  • Received January 11, 2010.
  • Accepted April 7, 2010.


View Abstract