Arabidopsis deadenylases AtCAF1a and AtCAF1b play overlapping and distinct roles in mediating environmental stress responses

To maintain homeostasis in an ever-changing environment organisms have evolved mechanisms to reprogram gene expression. One central mechanism regulating gene expression is mRNA degradation, which is initiated by poly(A) tail shortening (deadenylation). The carbon catabolite repressor 4-CCR4 associated factor 1 (CCR4-CAF1) complex is the major enzyme complex that catalyzes mRNA deadenylation and is conserved among eukaryotes. However, the components and functions of this global regulatory complex have not been well characterized in plants. Here we investigate the CAF1 family in Arabidopsis thaliana . We identify eleven AtCAF1 homologs and show that a subset of these genes are responsive to mechanical wounding, among them are AtCAF1a and AtCAF1b whose expression levels are rapidly and transiently induced by wounding. The differential expression profiles of the various AtCAF1s suggest that not all AtCAF1 genes are involved in stress-responsive regulation of transcript levels. Comparison of mis-expressed genes identified via transcript profiling of Atcaf1a and Atcaf1b mutants at different time points before and after wounding suggests that AtCAF1a and AtCAF1b target shared and unique transcripts for deadenylation with temporal specificity. Consistent with the AtPI4K γ 3 transcript exhibiting the largest increase in abundance in Atcaf1b , AtCAF1b targets AtPI4K γ 3 mRNA for deadenylation. Stress tolerance assays demonstrate that AtCAF1a and AtCAF1b are involved in mediating abiotic stress responses. However, AtCAF1a and AtCAF1b are not functionally redundant in all cases, nor are they essential for all environmental stresses. These findings demonstrate that these closely related proteins exhibit overlapping and distinct roles with respect to mRNA deadenylation and mediation of stress responses.


INTRODUCTION
Communication of developmental and environmental information is mediated through alteration of gene expression. The steady-state level of messenger RNA (mRNA) within a cell is determined by the combination of the rate of transcription and the rate of post-transcriptional processes regulating mRNA decay. Genome-wide approaches are revealing that in response to environmental stimuli there is a large reprogramming of gene expression (Eulgem, 2005;Swindell, 2006;Kilian et al., 2007;Ma and Bohnert, 2007;Walley et al., 2007;Walther et al., 2007;Shalem et al., 2008). While changes in the rate of mRNA degradation provide a rapid mechanism to alter mRNA abundance, research has largely focused on changes in transcriptional regulation as a means to control mRNA levels (Gutierrez et al., 2002;Yamashita et al., 2005;Narsai et al., 2007;Belostotsky and Sieburth, 2009;Chiba and Green, 2009;Lee and Glaunsinger, 2009).
In eukaryotes, degradation of mRNA begins with shortening of the poly(A) tail, referred to as deadenylation, which is the rate limiting step (Meyer et al., 2004;Belostotsky and Sieburth, 2009;Chiba and Green, 2009). The process of deadenylation has been best studied in yeast where it has been shown that the CCR4-CAF1 (also called Pop2p) complex serves as the major deadenylase complex (Tucker et al., 2001). In addition to the CCR4-CAF1 complex, both the poly(A) ribonuclease (PARN), for which mutants in Arabidopsis are embryo-lethal, and poly(A) nuclease (PAN) are also active deadenylases (Chiba et al., 2004;Meyer et al., 2004;Reverdatto et al., 2004;Belostotsky and Sieburth, 2009). Following deadenylation, mRNA decay proceeds via two distinct pathways. In one pathway mRNA is degraded in a 3'-to-5' manner via the exosome (Chekanova et al., 2007;Belostotsky and Sieburth, 2009). Alternatively, deadenylated mRNA can also undergo 5'-to-3' decay, which first requires that the 5' cap is removed from the deadenylated mRNA (Goeres et al., 2007;Belostotsky and Sieburth, 2009).
The biological role of CAF1 has been examined in a range of eukaryotic species.
In yeast, caf1 deletion strains are sensitive to high temperatures and caffeine (Hata et al., 1998). Analysis of CAF1 in C. elegans using RNAi and deletion alleles demonstrates that CAF1 is essential for embryonic and larval development (Molin and Puisieux, 2005).

6
In plants, limited experiments have investigated components of the CCR4-CAF1 complex. However, the studies that have been conducted focus on CAF1 and demonstrate a role for CAF1 in both development as well as response to biotic stresses.
In Arabidopsis, AtCAF1 homologs have been shown to be induced by hormone treatment as well as abiotic and biotic stresses (Lee et al., 2005;Zheng et al., 2006;Ferrari et al., 2007;Walley et al., 2007;Liang et al., 2009). Additionally, one study found that overexpression of a yeast CAF1 homolog from pepper (Capsicum annuum; CaCAF1) in tomato resulted in growth enhancement and resistance to the oomycete pathogen Phytophthora infestans. Conversely, when CaCAF1 is silenced in pepper there is significant growth retardation and plants are susceptible to the pathogen Xanthomonas axonopodis pv. vesicatoria (Sujon et al., 2007). Furthermore, in Arabidopsis it has been demonstrated that two yeast CAF1 homologs, AtCAF1a (At3g44260) and AtCAF1b (At5g22250), are active deadenylases in vitro and partially complement growth defects of a yeast caf1 deletion strain. Finally, Arabidopsis mutant lines Atcaf1a-1 and Atcaf1b-1 are susceptible to Pseudomonas syringae pv. tomato DC3000 (Liang et al., 2009).
In this study we examine the role of the Arabidopsis CAF1 family of deadenylases in abiotic stress responses. We identify a total of eleven AtCAF1 homologs that exhibit differential patterns of expression in response to wounding. Transcriptional profiling of Atcafa-1 and Atcaf1b-2 suggests that AtCAF1a and AtCAF1b target unique transcripts for deadenylation with temporal specificity. Additionally, we show that AtCAF1b targets a putative phosphatidylinositol 4-kinase mRNA for deadenylation.
Finally, stress tolerance assays suggest that while AtCAF1a and AtCAF1b are involved in mediating responses to a number of environmental stresses, they do not act redundantly in all cases nor are they required for all stresses.

Identification of the CAF1 Family in Arabidopsis
We previously identified two Arabidopsis homologs of the yeast CAF1 protein, AtCAF1a and AtCAF1b, via transcript profiling, as rapid wound response (RWR) genes (Walley et al., 2007). In order to determine the full complement of AtCAF1 homologs present in the Arabidopsis genome, we used the AtCAF1a sequence to identify all 7 proteins with sequence similarity using a Basic Local Alignment Search Tool (BLAST) (Altschul et al., 1990) at Phytozome (www.phytozome.net). This search yielded nine additional sequences, which we named AtCAF1c-k (Figures 1&2), resulting in a small family of eleven members. Alignment the AtCAF1 proteins with CAF1 sequence from other eukaryotes demonstrates that the RNaseD domain (Daugeron et al., 2001;Sujon et al., 2007) is well conserved in the AtCAF1 family ( Figure 1). We also found two closely related sequences in moss (Physcomitrella patens), eight sequences in rice (Oryza sativa) and ten sequences in poplar (Populus trichocarpa) (data not shown), demonstrating that there has been a significant expansion of the CAF1 gene family in angiosperms, possibly due to gene duplication events. In other eukaryotes such as yeast, fly, mouse and human there are only one or two CAF1 genes (Figure 2 and (Albert et al., 2000;Daugeron et al., 2001;Nakamura et al., 2004;Temme et al., 2004;Yamashita et al., 2005)).
In order to determine which of the AtCAF1 sequences are most closely related to the yeast CAF1 protein we first performed an alignment of all eleven AtCAF1 protein sequences with the yeast, fly, mouse and human CAF1 sequences using Se-Al and CLUSTALW. The alignment was trimmed to include only conserved regions (data not shown). Using this alignment we ran a parsimony analysis in PAUP* (Harrison and Langdale, 2006). We obtained ten trees from this analysis, which were then computed into a consensus tree using the 'strict' option in PAUP* ( Figure 2). The tree was then rooted with S. pombe CAF1 and bootstrap values were calculated in PAUP*. The obtained consensus tree places all of the AtCAF1s as sister to S. cervisiae CAF1, showing that while AtCAF1a-k are related to the yeast protein orthology is equivocal ( Figure 2). The Arabidopsis homologs group into three well supported clades: AtCAF1cg; AtCAF1h-k; and AtCAF1a and AtCAF1b ( Figure 2). These phylogenetic relationships suggest that there could be functional overlap between clade members. Given the large number of AtCAF1 homologs it is also possible that either sub-or neo-functionalization has occurred among some of the family members.

In order to determine if stress-induced transcriptional activation of AtCAF1a and
AtCAF1b was unique or a general property of all eleven family members, we examined the expression of AtCAF1a-k in response to mechanical wounding by qRT-PCR. These data show that five of the eleven AtCAF1 genes are upregulated in response to mechanical stress, while four of the eleven AtCAF1 genes do not respond to mechanical wounding and one is down-regulated ( Figure 3). Finally, we could not detect expression of AtCAF1f under the conditions we examined. The differential expression patterns exhibited by the AtCAF1 family members suggests that a subset of the AtCAF1 genes may be involved in rapid stress-responsive regulation of transcript levels while others may regulate deadenylation of mRNAs responsive to various developmental and environmental cues.

Transcript Profiling of AtCAF1a and AtCAF1b Mutants
For further investigation into the role of the Arabidopsis CAF1 homologs in regulating gene expression and stress tolerance we focused on AtCAF1a and AtCAF1b.
These two AtCAF1 family members were selected because 1) they were identified from previous studies as RWR genes (Walley et al., 2007) and 2) they form a monophyletic clade, suggesting that they may exhibit functional overlap. Many of the RWR genes are also categorized as abiotic and biotic stress-responsive genes, suggesting that they likely comprise a set of core components in the general stress-response network (Walley et al., 2007). We were therefore interested in elucidating what role(s) AtCAF1a and AtCAF1b may play during signaling upon environmental stress. Towards this aim we obtained Salk T-DNA insertion lines for AtCAF1a and AtCAF1b and confirmed that they do not produce full-length transcripts and are likely functional nulls ( Figure S1). These mutant lines are visually indistinguishable form the WT plants.
We hypothesized that AtCAF1a and AtCAF1b could interact with target mRNAs to result in altered gene expression profiles. In order to test this hypothesis we performed transcript profiling using the Agilent V3 oligo microarray, which contains probes for 28,500 genes from TIGR's ATH1 v.5 annotation as well 10,000 probes based on MPSS data (Meyers et al., 2004) and identified genes which are mis-expressed in the absence of AtCAF1a and AtCAF1b. For this experiment we examined three-week-old rosette leaves from wild-type, Atcaf1a-1 and Atcaf1b-2 plants, both before and intervals after wounding, using three independent biological replicates for each genotype and timepoint.
The full set of the raw intensity microarray data are deposited at http://www.ncbi.nlm.nih.gov/geo/, under GEO accession GSE17022. Mis-expressed genes were classified as probes that exhibited an increase or decrease in signal of at least 2-fold and had a p-value ≤ 0.05. Using this approach, we found that between 62 and 136 genes were mis-expressed in Atcaf1a-1 compared to wild-type, depending on the timepoint ( Figure 4 and Table S1). While, a smaller number of genes (12-28) were misexpressed in Atcaf1b-2 compared to wild-type ( Figure 4 and Table S1). The number of up and down regulated genes in each of the caf mutant lines at various intervals post wounding are shown (Table S2). Atcaf1a-1 and Atcaf1b-2 mis-expressed genes were classified according to gene ontology (GO) terms in order to provide insight into their biological function (Berardini et al., 2004). The two largest defined classes of GO terms involve response to stress or abiotic/biotic stimuli ( Figure S2), suggesting that these two abundance observed for either of the Atcaf1 mutants encodes a putative phosphatidylinositol 4-kinase (AtPI4Kγ3) (Mueller-Roeber and Pical, 2002). In Atcaf1b-2 transcript levels of AtPI4Kγ3 were increased 42 to 63 fold in our microarray suggesting that it may be a direct target of AtCAF1b ( Figure S3). We confirmed by qRT-PCR that AtPI4Kγ3 levels were indeed increased exclusively in Atcaf1b-2 ( Figure 6A). To test if AtCAF1b is acting directly on AtPI4Kγ3 mRNA we performed poly(A) tail length (PAT) assays (Salles and Strickland, 1999). As shown in Figure 6B the length of AtPI4Kγ3 mRNA increased specifically in Atcaf1b-2. Additionally, the increase in length of AtPI4Kγ3 mRNA in Atcaf1b-2 is not a result of a general increase in poly(A) tail length as other transcripts tested, including UBQ10, did not increase in size ( Figure 6C and data not shown). Taken together these results strongly suggest that AtCAF1b targets AtPI4Kγ3 mRNA for deadenylation. We also tested, via the PAT assay, a number of transcripts that were increased in abundance in Atcaf1-a, albeit to a much lower level than AtPI4Kγ3 was in Atcaf1b-2, and were unable to uncover any potential targets of CAF1a (data not shown), likely due to the low sensitivity of the PAT assay.

AtCAF1a and AtCAF1b Regulate Abiotic Stress Responses
AtCAF1a and AtCAF1b were originally identified as RWR genes, which are enriched for abiotic and biotic stress responsive genes, suggesting that these CAF1s may be involved in mediating multi-stress resistance. Additionally, it was recently shown that  S4). We also tested their ability to cope with salt stress by germinating seeds on salt plates. Atcaf1a plants show an increased germination rate on 200mM NaCl compared to wild-type and Atcaf1b ( Figure 7B&C).
No differences in survival or germination were observed for plants grown on control plates ( Figure 7C and data not shown). However, Atcaf1a and Atcaf1b mutants did not differ in tolerance to water logging or the pathogen Botrytis cinerea when compared to wild-type ( Figure 7D&E). These data suggest that while AtCAF1a and AtCAF1b are involved in mediating response to both abiotic and biotic stresses they do not act redundantly in all cases nor are they required for all stresses.

DISCUSSION
In eukaryotes control of mRNA degradation is an essential component in the Given the large number of AtCAF1 homologs in Arabidopsis it is possible that sub-or neo-functionalization has occurred among some of the family members, resulting in distinct CAF1 proteins functioning under specific conditions and/or targeting unique transcripts. In support of this idea qRT-PCR analysis of the AtCAF1 family revealed that AtCAF1 genes exhibit differential expression patterns in response to mechanical wounding ( Figure 3). This differential expression behavior exhibited by AtCAF1 family members is consistent with the processes of sub-and/or neo-functionalization.
Second, the expression of AtPI4Kγ3 is increased specifically in Atcaf1b. The increased expression of AtPI4Kγ3 is consistent with the observed increase in poly(A) tail length specifically in the absence of AtCAF1b, further demonstrating that AtCAF1a and AtCAF1b are capable of targeting specific transcripts ( Figure 6). Lastly, while Atcaf1a and Atcaf1b exhibit similar responses to most environmental stresses, Atcaf1a is more tolerant to salt treatment. These results demonstrate a functional specificity between these two paralogs, which would be consistent with a sub (or neo) functionalization event ( Figure 7). Taken together these expression and stress analyses show that not only do the different types of deadenylation (PARN versus CCR4-CAF1) complexes target unique transcripts but also that closely related deadenylases are able to target unique mRNAs, resulting in unique biological responses. AtCAF1s are not simply required for stress tolerance but in certain cases they can actually act as a negative regulator of stress tolerance ( Figure 7B&C). We also show that AtCAF1a and AtCAF1b play a role in abiotic stress responses ( Figure 7A-C & S4).
Finally, while AtCAF1a and AtCAF1b mediate resistance towards both abiotic and biotic stresses, they are not involved in mediating the response towards all environmental stresses ( Figure 7D&E).
The observed increase in salt tolerance exhibited specifically by Atcaf1a is supported at the molecular level by our microarray profiling experiments. We observed that 29% of transcripts which increase in abundance in non-wounded Atcaf1a-1 plants are genes previously shown to be induced by salt treatment (Ma et al., 2006). This enrichment in salt induced genes is highly significant at a p-value of 5.5 x 10 -11 , as calculated by the hypergeometric test (Sokal and Rohlf, 1995;Covington and Harmer, 2007).
Included in this common gene set is ALDH7B4, which has been shown to confer salt tolerance when overexpressed (Kotchoni et al., 2006). Consistent with Atcaf1b mutants behaving like wild-type plants in response to salt treatment, there was no overlap between transcripts increased in Atcaf1b-2 and salt induced transcripts. The transcript profiling experiment also provides insight into the increased sensitivity of both Atcaf1a and Atcaf1b to MV. One gene, galactinol synthase2 (GolS2), which is decreased in abundance in non-wounded Atcaf1 as well as Atcaf1b, confers tolerance to MV when overexpressed (Nishizawa et al., 2008).
In conclusion, we show that the CAF1 family has expanded in angiosperms and that a subset of the AtCAF1 family responds to mechanical stress. Significantly, we also immunoprecipitation followed by sequencing (RIP-seq) will be useful in identification of the complete suite of AtCAF1a and AtCAF1b direct mRNA targets. Additionally, biochemical characterization of AtCAF1a and AtCAF1b associated mRNA binding proteins will provide further insight into how these deadenylases act with such a high degree of specificity.

Plant Materials and Growth Conditions
Arabidopsis

BLAST Analyses to Identify CAF1 Homologs in Plants
Unless otherwise stated, all computer analyses were performed on a single Macintosh iMac running Mac OS 9 or OS X. Using the Arabidopsis CAF1a (AtCAF1a) sequence as a query we performed a BLAST search to the Viridiplantae node at Phytozome (www.phytozome.net). After filtering out any repetitive sequences we downloaded all of the retrieved sequences. Yeast, fly, mouse and human CAF1 sequences were identified by performing a BLAST search at NCBI. Locally installed versions of Se-Al and CLUSTALW were used to generate alignments using all the obtained amino acid sequences, which were subsequently edited to include only conserved regions of the proteins.

Phylogenetic Analysis
All steps for phylogeny reconstruction were performed according to (Harrison and Langdale, 2006). Using the amino acid sequence alignment generated above we performed a 'parsimony' and 'heuristic' search to generate a phylogenetic tree in PAUP* (Swofford, 1999). The ten trees generated by this analysis were subsequently computed into a 'strict' consensus tree and bootstrap values were calculated in PAUP*.

RT-PCR
Total RNA from rosette leaves was isolated by TRIzol extraction ( Table S3.

Microarray Analyses
The Agilent Arabidopsis V3 oliogoarray chip containing 60-mer oligos probes for 28,500 genes from TIGR's ATH1 v.5 annotation as well 10,000 probes based on MPSS data (Meyers et al., 2004) was utilized for transcript profiling (Agilent Technologies, Wilmington, Delaware). Total RNA was prepared from three independent biological replicates for each genotype x treatment combination as described for RT-PCR. Prior to hybridizations, the quality and quantity of the total RNA sample was confirmed by running on an Agilent 2100 bioanalyzer, and by using a NanoDrop ND-1000 Wilmington, Delaware) to analyze for mis-expressed genes according to the manufacturers protocol. Genes with a fold-change ≥ 2.0 and a p-value ≤ 0.5 were selected as mis-expressed.

Comparison of transcript profiles
The statistical significance of the observed overlap in transcript profiles, as well as enrichment of GO terms, was analyzed using hypergeometric tests (Sokal and Rohlf, 1995;Covington and Harmer, 2007;Rivals et al., 2007).

Poly A-Tail (PAT) Assay
Poly A-Tail (PAT) assays were performed as previously described (Salles and Strickland, 1999) Figure S4. Survival of WT, Atcaf1a-2 and Atcaf1b-3 on 5 µM MV (paraquat). Data are means of 9 independent biological replicates ± SEM. Asterisks denote a significant difference of both Atcaf1a and Atcaf1b from WT (P<0.05) as determined by t-tests. Table S1. List of mis-expressed genes in Atcaf1a-1 and Atcaf1b-2 compared to wild-type plant, at different timepoint post wounding (Table S1). Table S2. Numbers of mis-expressed genes in each mutant relative to WT at each timepoint post wounding.  (B) PAT assay measuring the poly(A) tail length of AtPI4K?3 in WT, Atcaf1a-1 and Atcaf1b-2 before and in response to mechanical wounding. (C) PAT assay measuring the poly(A) tail length of UBQ10 in WT, Atcaf1a-1 and Atcaf1b-2 before and in response to mechanical wounding.