General characterization of the ESTs

Suppressive subtractive hybridization, or SSH [14], results in the isolation and amplification of mRNAs present in one population (the tester population) and absent in the other (the driver population). Using SSH, we created two subtracted libraries, one putatively enriched in juvenile-specific transcripts and one putatively enriched in adult-specific transcripts (Additional file 1). From now on, we will refer to these libraries as the "juvenile library" and the "adult library", respectively.

To test the differential expression of the ESTs, 96 clones from each library were randomly chosen and spotted in the same pattern onto two nylon membranes. Each replicate membrane was hybridized with one of two probes, created either from adult or juvenile mRNA samples (Fig. 2). Out of the 96 randomly-chosen, putative juvenile clones, 53 were only expressed in juveniles, 13 were expressed at a higher level in juveniles than in adults, 5 were expressed at similar levels in adult and juveniles and 25 did not generate any signal with either probe (Fig 2., top panels). Out of the 96 randomly-chosen, putative adult clones, 44 were expressed only in adults, 14 were expressed at a higher level in adults than in juveniles, 10 were expressed at similar levels in adults and juveniles, 5 were expressed at a higher level or only in juveniles and 23 did not generate any signal with either probe (Fig 2., bottom panels). In addition, differential expression of three clones was confirmed by virtual northern blots (data not shown). Virtual Northern blots differ from Northern blots in that phase-specific cDNA is blotted on the nylon membranes instead of mRNA [15]. These data provide solid preliminary evidence that the SSH was successful at isolating many transcripts differentially expressed in these two phases.

In total, 604 and 601 ESTs were sequenced from the adult and juvenile libraries respectively. Sequences containing no insert or unreliable data (as evidenced by the sequence trace) were excluded, leaving 478 ESTs from the adult library and 463 ESTs from the juvenile library for further analysis. Sequences were cleaned in silico of contaminating fragments (vector and primer sequences; see Materials and Methods). For 87% (411) of the adult clones and 83% (392) of the juvenile clones, this single-pass sequencing provided the complete sequence of the insert, i.e. vector sequence bordered both ends of the insert. ESTs ranged from 68 bps to 855 bps in length. On average, juvenile clones were longer than adult clones, averaging 474 bps and 408 bps respectively.

Due to the way the libraries were created (Additional file 2), some ESTs in the final library contained either a polyA or polyT tract [16]. These tracts originated from the polyA tail of the corresponding original mRNAs, indicating that these ESTs probably contained untranslated regions. Because the ESTs were not cloned directionally, sequences containing polyA or polyT tracts were obtained according to which strand was sequenced. The occurrence of these tracts was significantly higher in the juvenile than in the adult library (40 vs 15% respectively).

Redundancy and overlap between the two libraries

All ESTs were aligned and partitioned into clusters (see Material and Methods for details). All ESTs that were not part of a cluster remained singletons (279 adult and 233 juvenile ESTs). A consensus sequence was derived from clusters containing two or more ESTs. The juvenile ESTs formed 77 clusters and the adult ESTs formed 84 clusters. Clusters contained up to 14 ESTs (Fig. 3). This probably over-estimates the true number of clusters, as non-overlapping ESTs would be placed into two or more separate clusters or remain singletons even if they originated from the same initial mRNA. In addition, it is possible that sequences that only differed because of sequencing errors or regions of poor sequence quality were not clustered together.

Clusters containing ESTs from both libraries were labeled as "mixed clusters". Only 2 such mixed clusters were found, representing a mere 0.3% of the total number of clusters (Fig. 3). Thus, the overlap between the two libraries is minimal, providing additional evidence that SSH probably successfully isolated ESTs specific to each developmental phase.

Gene functions of the ESTs

All singletons and clusters were also analyzed for homologies using BLASTX against the Arabidopsis thaliana protein database [17] and using TBLASTX against the current draft of the Chlamydomonas reinhardtii genome [18]. For both searches, hits with E values < 10E-06 were considered significant. In general, the same sequences produced significant hits against each of the different databases (Table 1). This independently confirmed the results of the first searches and the low percentage of coding sequences within our ESTs. In total, 178 ESTs (only 26%) produced significant hits in at least one of the BLAST searches (Table 1).

The 178 ESTs or clusters that produced significant hits were sorted into functional categories according to the classification scheme developed for plants [19] (Fig. 4). The largest functional class from both libraries contained genes associated with photosynthesis ("energy" in Fig. 4). In general, a higher percentage of juvenile ESTs have functions related to transcription, and protein synthesis, transport and storage while a higher percentage of adult ESTs have functions related to cell structure (Fig. 4 and Additional file 3).

Phylogenetic analysis of the ESTs

To assess if the putative functions of the ESTs also occurred in land plants or in other green algae, we identified (for each of the 178 ESTs that generated significant BLAST hits) the sequence giving the lowest E value ("best match") in the BLAST searches and the organism to which this sequence belongs. As expected, most of these "best-match" sequences belong either to Chlamydomonas reinhardtii or the Streptophyta (land plants). This is not surprising, as the number of sequences available for most green algal lineages remains extremely limited. Specifically, A. acetabulum belongs to the class Ulvophyceae for which very little sequence information is available [20].

Conserved sequences within the ESTs

Some ESTs showed similarity to each other over short regions. These ESTs clustered into three groups of 5, 9 and 2 sequences respectively (Fig. 5). The length of the common stretch of sequence varies between 30–70 bps for group 1, 45–90 bps for group 3 and 170 to 250 bps for group 2 (Fig. 5). Within each group, these ESTs showed no similarity to each other outside of these regions but within these regions, the level of identity was high. Most of these ESTs belonged to the adult library. None of these 16 ESTs produced any relevant BLAST hit, making it difficult to predict whether or not they contain coding sequences. Among the sequences sharing the second consensus sequence (Fig. 5c), all but two of the ESTs ended with a polyA or polyT tract, indicating that they probably contain 3' UTRs. None of the ESTs sharing the first or third consensus sequences (Fig. 5b or 5d) ended with a polyA or polyT tract. These 3 conserved elements may be specific to A. acetabulum, because they were not found in any other sequence in Genbank (nucleotide database) except for the carbonic anhydrase 1 and 2 (CA1 and CA2) genes from A. acetabulum [13]. All three of these conserved sequences fell in the non-coding regions of the two CA genes, either in introns or the 3'UTRs (Fig. 5a).

Adult and juvenile phases in A. acetabulum differ significantly in gene expression

The expression analysis (Fig. 2) and the fact that there is virtually no overlap between the two libraries suggest that the subtraction succeeded in isolating differentially expressed transcripts.

The 941 ESTs were organized into 675 independent clusters or singletons. Although this number is probably an overestimate – non-overlapping ESTs originating from the same transcript may partition to different clusters or remain singletons – these ESTs only represent a portion of all the ESTs present in the libraries. These data provide strong evidence that the adult and juvenile phases in A. acetabulum differ significantly in gene expression and that a large number of genes are probably phase-specific.

Physiological differences between the two developmental phases

The functions of the transcripts expressed at the different phases partition differently into functional classes (Fig. 4). Given that the libraries were created such that only ESTs specific to one developmental phase would be isolated, the distribution of gene functions among the ESTs is not expected to reflect that of a typical photosynthetic cell, but merely the functions that are specific to one developmental phase or the other.

Juveniles seem to devote much of their unique gene expression to transcription, protein synthesis, transport and storage, consistent with the general idea that juveniles are fast growing, more dedicated to growth than morphogenesis. During juvenile phase, the unicells increase about 10-fold in height (from <1 mm to 1 cm) in 2 weeks and are not competent to make a cap. On the other hand, adults increase in height about 3-fold, another 3 centimeters. Adult cells are much more complex than juvenile cells, with numerous whorls of hair that are highly branched. There is thus an increased requirement for cell wall synthesis during the adult phase. Adults also are competent to execute complex cap morphogenesis, and are preparing for nuclear divisions, nuclei transport and gametogenesis [21]. There is not a large increase in cytoplasmic volume between adult and juveniles because most of the volume within an adult is occupied by a central vacuole (Ngo et al., submitted). This more complex adult development is consistent with a lower percentage of transcripts dedicated to protein synthesis and a higher percentage of transcripts involved in cell structure (e.g. cytoskeletal proteins, enzymes involved in cell wall synthesis or maintenance, histones).

A surprisingly high number of ESTs from both libraries are associated with photosynthesis (Additional file 3, class 2). For example, five ESTs were putative homologs of the rbcS protein. It is possible that these ESTs truly come from different transcripts. It is also possible that they originate from the same transcript but have not been clustered together because they do not overlap or because of regions of poor sequence quality. To address this question, we aligned these sequences to the A. acetabulum rbcS mRNA sequence present in Genbank (Fig. 6). We observed in each EST or cluster, a region of high sequence identity to the Aa-rbcS sequence (identity varied between 84 and 89% at the nucleotide level). Two of the clusters (cn115 and cn116) were identical except for sequence ambiguities, which were frequent enough for these two sequences not to be clustered together. With the exception of cn116 and cn115, the other sequences were different enough from each other and from the Aa-rbcS sequence to conclude that they did not actually originate from the same transcript. In support of this hypothesis, regions outside of these fragments of very high sequence identity could not be aligned (<50% identity). At the amino acid level, the sequence identity between these regions (white boxes in Fig. 6) and the Aa-rbcS sequence was almost complete, varying between 94 and 96%. A closer look at the sequences confirmed that most of the nucleotide differences present in the ORFs occurred at the third position of a codon, often resulting in the conservation of the amino acid sequence. Finally, one of the clusters (cn115) ends with a polyA tract, indicating the end of the 3'UTR, while the transcript that generated cluster 146 seems to possess a much longer 3'UTR that contains no polyA tract. Taken together, these results suggest that there may be at least 3 different loci coding for the rbcS protein in Acetabularia (cn169, J538 and Aa-rbcS, Fig 6). Confirmation of these surprising results with different methods and detailed expression analysis of the different loci will be of great interest. A similar analysis of the chlorophyll a/b binding proteins might also determine whether different proteins are expressed during different phases and whether they are functionally distinct and/or differentially regulated. The LHC (light-harvesting complex) binding proteins form a very large family that has been best characterized in land plants [22]. Analysis of the 22 ESTs with identity to LHC binding proteins from land plants will improve our currently poor knowledge of this gene family in algae.

Putative gene functions of particular interest

Immuno-cytochemistry has been used to visualize tubulin and actin proteins in A. acetabulum during development [23, 24]. Actin microfilaments were found in thalli of all ages forming continuous, parallel bundles along the entire stalk. Microtubules, conversely, could not be detected in the alga prior to meiosis. Microtubules were detected during reproduction, surrounding haploid nuclei as they are transported up into the cap. Consistent with these results, our data suggest that juveniles express actin transcripts (but not tubulin) while adults express both alpha and beta tubulin (but not actin) transcripts, presumably in preparation for reproduction.

Two of our ESTs were putative expansin homologues. Expansins promote cellulose walls extension in land plants. Typical of the "mannan weeds", the wall of the diplophase of A. acetabulum is predominantly a para-crystalline mannan framework (Dunn et al., submitted). Only gametangia are enclosed in a cellulosic wall, itself surrounded by the mannan wall of the cap. Consistent with this, our results indicate that the expansin gene was expressed during adult but not juvenile phase. So far, expansins have only been found in land plants [25] where expansin acts within the cell wall and is activated by an acidic pH [26]. If these transcripts code for expansin proteins that play a role in loosening walls in A. acetabulum, it would be interesting to see if their mechanism of action is similar to that in land plants, and whether their substrate is also a cellulose wall.

A. acetabulum, for most of its life cycle, contains only one nucleus, which is located in the rhizoid. This nucleus undergoes replication at the end of adult phase, during reproduction [27]. At this juncture, there is a tremendous need for nucleotides and histones to make the millions of haploid nuclei needed for gametogenesis [21]. Consistent with this, histone mRNAs were found only in the adult library (Additional file 3, class 9). It would be interesting to look more deeply into when these transcripts are expressed and how this organism is able to produce histone proteins in such high quantities in such a short period of time.

Finally, one of the ESTs has homology to an argonaute protein (Additional file 3, class 11, E value to Oryza sativa argonaute of 5e-15). Argonaute proteins are highly conserved and play a major role in RNA interference in animals (a.k.a. quelling in fungi or post-transcriptional gene silencing in plants [28]). These processes are involved in the silencing of specific genes via double stranded RNA [29] and their importance in post-transcriptional regulation is just starting to be deciphered. Argonaute proteins have been found in land plants, ciliates, animals and fungi but, to the best of our knowledge, this EST is the first identified algal sequence of an argonaute protein.

Why most of the ESTs do not correspond to any previously described sequences

We can think of three reasons why only 28.6% of the ESTs, a particularly low number, were assigned a putative homolog based on BLAST and InterPro searches. First, these are subtracted libraries, created with the objective of identifying rare, phase-specific transcripts or transcripts involved in morphogenesis, apical growth, or phase change. Hence, these ESTs should include fewer housekeeping transcripts, abundant transcripts, or transcripts common to both phases or to other organisms.

Second, the ESTs were generated by a reverse transcriptase using a poly-T primer that often does not generate full-length cDNAs. Our libraries therefore tend to be enriched in 3' ends of the transcripts, which contain non-coding sequences and which would not be recognized in homology searches. The high percentage of ESTs containing a polyA or polyT stretch supports this hypothesis.

Finally, A. acetabulum belongs to the order Dasycladales, in the green algal class Ulvophyceae, for which very little sequence data is currently available. Before the addition of our ESTs, only 73 DNA sequences from A. acetabulum were available in Genbank, representing just 37 different genes. Although complete genomes of several land plants and green algae are now at least partially available, it is plausible that most of the A. acetabulum sequences are too divergent from those of other algae or land plants to be recognized as orthologs when entered in BLAST searches [30]. To test this hypothesis, we raised the cut-off value for the BLASTN and BLASTX searches against the Genbank databases from 10E-06 to 10E-03. Most additional hits obtained originated from algal or land plant sequence as opposed to a random distribution of the organisms represented in Genbank. This supports the hypothesis that these ESTs are probably homologous to these algal or plant sequences but too divergent for the homologies to be trusted.

Do adult and juvenile transcripts differ in structure? Insights into post-transcriptional regulation

Curiously, 40% of the juvenile clones but only 15% of the adult clones end with a polyA or polyT tract. If these tracts correspond to the mRNA polyA tail, then these ESTs contain some or all of the 3' untranslated regions (3' UTR) of the transcript from which they originated. We have diagrammed hypotheses explaining the differential occurrence of these tracts in adult versus juvenile clones (Fig. 7). The first explanation presumes an artifact of the techniques used to create the libraries. ESTs result from the amplification of cDNA fragments that have been digested by RsaI, each RsaI fragment having an equal chance of being amplified and cloned. If the adult cDNAs were more completely digested than the juvenile cDNAs, then the adult cDNAs would have generated a higher number of ESTs, a lower proportion of which would contain polyA tracts (Fig. 7a). A second hypothesis presumes differential mRNA length: if adult cDNAs were, on average, longer than juvenile cDNAs, each adult cDNA would produce more ESTs, yielding a lower proportion of ESTs containing the polyA tract. Adult cDNAs could be longer if on average they have longer coding sequences (Fig. 7b) or longer 3'UTRs (Fig. 7c). If they have longer 3'UTRs, the proportion of coding sequences as well as the proportion of ESTs with polyA tracts will be higher in juvenile ESTs than in adult ESTs, consistent with our findings.

Why would 3' UTRs be longer in adult transcripts than in juvenile transcripts? In adult A. acetabulum, growth and morphogenesis occurs almost exclusively at the stalk apex, centimeters away from the unique nucleus located in the rhizoid. Therefore, aspects of post-transcriptional regulation, such as mRNA stability and mRNA localization, are probably very important to the regulation of gene expression in these unicells. Indeed, more than half of the transcripts (9/16) studied to date in A. acetabulum are localized to one end or the other of the unicell, most often to its apex [31–33]. To achieve this localization, each transcript must contain cis-acting elements within its sequence, also called 'zipcodes' [34]. In yeast and animal cells, 'zipcodes' are part of the 3' UTR of the localized transcripts [35]. Also, considering the rate at which mRNA molecules move along cytoskeletal elements along the stalk of A. acetabulum [36, 37], to reach the apex, any mRNA must be at least three days old, classifying them among the "ultra-stable" mRNA species [38]. In plants, the cis-acting elements responsible for stability of an mRNA molecule are also located in its 3' UTR [38]. Transcripts 3' UTR might therefore play an important role in the regulation of gene expression in this species, especially in adults.

To achieve these stability and localization patterns, adult mRNAs probably contain several post-transcriptional regulatory elements within their 3' UTRs, potentially explaining why these would be longer. What are these regulatory elements? The fact that three conserved elements (Fig. 5) were found within several unrelated ESTs, most of which originate from the adult library is promising. These 3 elements also appear to be specific to A. acetabulum and are located in a non-coding region of the carbonic anhydrase gene, whose transcript is apically localized [13]. Five of the ESTs containing the second conserved element also contain a polyA tract, suggesting that these ESTs may code for 3' UTRs. The first and second conserved elements fall within introns of AaCA1 (Fig. 5). It is possible that these sequence elements of AaCA1 are part of alternatively spliced introns and sometimes contained in the mature mRNAs produced from this gene. Future research will focus on elucidating the function of these conserved elements and their spatial expression during development.

Culture of A. acetabulum

Unicells were grown in artificial seawater until they reached the desired developmental age. Axenic cultures were obtained by decontaminating mature caps and then using the axenic gametangia they housed for mating [21]. Zygotes were grown in sterile artificial seawater, Ace27, which is identical to Ace25 [39] except that the KCl prestock was purified over a chelex-100 column and it contains urea hydrogen peroxide at a final concentration of 10^-15 M. Cultures were grown under cool white fluorescent lights at a photon flux density of 170 μmol m^-2 s^-1 on a 14 h light/10 h dark photoperiod, at 21°C ± 2°C and repeatedly diluted to suit their developmental age [21].

mRNA extraction

Juveniles were harvested by filtration and adults were harvested using sterile dental tools. The unicells were dried briefly on a Kimwipe, and weighed on aluminum foil. Packets of algae of the same age were flash-frozen in liquid nitrogen. 7.15 g of juveniles (approximately 18,000 unicells) and 18.2 g of adults (approximately 4,000 unicells) were ground to a fine powder under liquid nitrogen. The powder was transferred to Oakridge tubes containing extraction buffer (0.1 to 0.2 g of ground unicells/ml extraction buffer). RNA was extracted according to Chang et al. [40].

Suppressive Subtractive hybridization (SSH)

cDNA synthesis and SSH were performed according to the manufacturer's recommendations using the PCR cDNA Synthesis Kit (Clontech Laboratories, Inc.) and the PCR-Select cDNA Subtraction Kit (Clontech Laboratories, Inc.) respectively. A summary of the steps involved in SSH and a more detailed figure of the formation of the ESTs from mRNA can be found in Additional file 1 and 2.

Cloning of the ESTs to make the libraries

DNA was precipitated using a standard ethanol precipitation protocol [41]. In order to add 3' A-overhangs to the PCR products for subsequent cloning, the DNA was resuspended into 25 μl of PCR reaction cocktail (2.5 μl of 10X buffer, 1.5 μl MgCl₂, 2 μl 10 mM dNTPs, 18.875 μl water and 0.125 μl Taq polymerase (Promega)) and incubated at 72°C for 8–10 minutes. The DNA was precipitated again [41] and resuspended in TE to the starting volume of the DNA amplification reaction. Following the manufacturer's recommendations, each library was cloned into 2 different cloning vectors using the AdvanTAge™ PCR Cloning Kit (Clontech Laboratories, Inc., now a discontinued product) and the TOPO™-TA Cloning Kit (Invitrogen).

Dot blot and virtual Northern blot analysis of the libraries

The quality of subtraction was controlled as recommended by the PCR-Select protocol provided by Clontech. PCR-amplified inserts of 96 randomly picked clones from both libraries were duplica-spotted onto nylon membranes and hybridized with the radioactively labeled subtraction mix from both subtractions. In addition, differential expression of cDNA inserts of three clones was confirmed by virtual northern blots using SMART cDNA synthesis (Clontech) [15]. The clones used in these dot blots and virtual northern blots were not sequenced and are not part of the following sequence analysis.

EST sequencing

Colonies were randomly picked from each library using sterile toothpicks. Plasmid DNA from each colony was isolated and eluted with 2 × 40 μl of elution buffer (Plasmid Miniprep Kit, Qiagen).

DNA sequencing was carried out at the Plant-Microbe Genomics Facility, Ohio State University. The sequencing reactions were prepared by mixing 400 ng of plasmid DNA and 4 pmol of primer (M13F (5'-GTAAAACGACGGCCAG-3') or M13R (5'-CAGGAAACAGCTATGAC-3') with water for a total volume of 10 μl. Next, 2 μl of BigDyeTerminator mixture, version 2 (Applied Biosystems), 4 μl BetterBuffer (The Gel Company) and 4 μl water were added. The cycling parameters were those recommended by the manufacturer except that the reactions were run for 35 cycles instead of 25. The reactions were cleaned up with Millipore Multiscreen/Sephadex columns, according to the manufacturers recommendations (Millipore Technical Note TN053). The resulting 20 μl of clean sequencing reaction product (in water) was placed in an Applied Biosystems 3700 DNA Analyzer for separation and analysis.

Sequence analysis