Microarray Analysis Confirms the Specificity of a Chlamydomonas reinhardtii Chloroplast RNA Stability Mutant

Abstract

The expression of chloroplast and mitochondrial genes depends on nucleus-encoded proteins, some of which control processing, stability, and/or translation of organellar RNAs. To test the specificity of one such RNA stability factor, we used two known Chlamydomonas reinhardtii nonphotosynthetic mutants carrying mutations in the Mcd1 nuclear gene (mcd1-1 and mcd1-2). We previously reported that these mutants fail to accumulate the chloroplast petD mRNA and its product, subunit IV of the cytochrome b₆/f complex, which is essential for photosynthesis. Such mutants are generally presumed to be gene specific but are not tested rigorously. Here, we have used microarray analysis to assess changes in chloroplast, mitochondrial, and nuclear RNAs, and since few other RNAs were significantly altered in these mutants, conclude that Mcd1 is indeed specifically required for petD mRNA accumulation. In addition, a new unlinked nuclear mutation was discovered in mcd1-2, which greatly reduced chloroplast atpA mRNA accumulation. Genetic analyses showed failure to complement mda1-ncc1, where atpA-containing transcripts are similarly affected (D. Drapier, J. Girard-Bascou, D.B. Stern, F.-A. Wollman [2002] Plant J 31: 687–697), and we have named this putative new allele mda1-2. We conclude that DNA microarrays are efficient and useful for characterizing the specificity of organellar RNA accumulation mutants.

Chloroplast and mitochondrial genes are dependent on nucleus-encoded proteins that activate and regulate their expression (for review, see Grivell et al., 1999; Barkan and Goldschmidt-Clermont, 2000; Monde et al., 2000; Zerges, 2000). This intracellular genetic circuitry provides mechanisms by which nuclei and organelles can coordinate production of proteins involved in photosynthesis and respiration. As part of this coordination, numerous nucleus-encoded regulatory genes are known to promote RNA processing and stability in chloroplasts and mitochondria.

Genetic studies in vascular plants and the green alga Chlamydomonas reinhardtii have identified many such genes that regulate chloroplast transcripts. Although a few of these mutations affect the accumulation of RNAs from more than one chloroplast gene (Levy et al., 1997; Meurer et al., 1998), the majority appear to be specific for one transcript, as determined by RNA gel blots testing a small number of transcripts (Kuchka et al., 1989; Goldschmidt-Clermont et al., 1990; Drapier et al., 1992, 2002; Monod et al., 1992; Gumpel et al., 1995; Drager et al., 1998). Studies in Saccharomyces cerevisiae have shown that mitochondria are similar to chloroplasts in that nuclear gene products stabilize specific mitochondrial transcripts (Mittelmeier and Dieckmann, 1993; Wiesenberger et al., 1995). Despite these significant efforts, the specificity for any chloroplast RNA stability gene has not been reported using genome-wide approaches. Knowing the RNA substrates for such regulatory proteins is necessary to fully understand their mechanisms of action and their roles in controlling photosynthesis and/or organellar biogenesis.

We have previously reported the characterization of the Chlamydomonas nuclear Mcd1 gene, whose product stabilizes the chloroplast petD mRNA (Drager et al., 1998, 1999). The petD gene encodes subunit IV (SUIV) of the cytochrome b₆/f complex, and it is essential for photosynthetic electron transport (Wollman et al., 1999). The Mcd1 gene product interacts directly or indirectly with nucleotides 2 to 9 of the 362-nucleotide petD 5′ untranslated region to block degradation by a 5′-3′ exoribonucleolytic activity (Drager et al., 1998, 1999; Higgs et al., 1999). The phenotypes of two Mcd1 mutant alleles (mcd1-1 and mcd1-2) have been characterized; both are nonphotosynthetic (PS−) and fail to accumulate petD mRNA despite normal levels of transcription (Drager et al., 1998). This 5′-3′ degradation may also function to process petD pre-mRNA and form the mature 5′ end. The RNA instability phenotype of mcd1-2, but not mcd1-1, can be partially suppressed by a mutation in the unlinked Mcd2 gene (mcd2-1; Esposito et al., 2001).

The recent availability of organellar genomic resources for Chlamydomonas makes it possible to test broadly the effects of nuclear mutations on organellar RNAs. These resources include the chloroplast (Maul et al., 2002), mitochondrial (GenBank accession no. NC_001638), and nuclear (Kathir et al., 2003; http://genome.jgi-psf.org/chlre2/chlre2.home.html) genome sequences as well as DNA microarrays (Im et al., 2003). Here, we report the use of a Chlamydomonas organellar DNA microarray to test the specificity of the mcd1-1 and mcd1-2 mutations. Changes in RNA levels could either be a direct result of the mcd1 mutations or an indirect result from the failed expression of petD and subsequent PS− phenotype. Our data show that Mcd1 is specific for the chloroplast petD mRNA. Surprisingly, a second, unlinked mutation was discovered in mcd1-2, which we subsequently found to be an allele of Mda1, which itself is required for atpA mRNA stability (Drapier et al., 1998).

RESULTS

DNA Microarrays Were Used to Test the mcd1-1 and mcd1-2 Nonphotosynthetic Mutants

Chlamydomonas DNA microarrays were prepared (see “Materials and Methods”) and used to test for changes in RNA accumulation from 47 chloroplast, 9 mitochondrial, and 15 nuclear genes. As a final check on DNA spot identity, all DNA samples were resequenced and only those that confirmed identity were included in microarray analyses. Where appropriate, the minimum information about microarray experiments standards were followed (Brazma et al., 2001). The data reported here have been submitted to the Gene Expression Omnibus (GEO) database (Edgar et al., 2002) at the National Center for Biotechnology Information (see “Materials and Methods”).

Three independent microarray hybridizations were performed for each experiment, including a dye-swap control to reduce dye-specific effects. To further improve quantification, each gene was spotted three times per slide. Because the intent was to identify RNAs directly dependent on Mcd1, total RNA was isolated from mutants and wild-type cells grown under low-intensity light in Tris-acetate phosphate (TAP) medium that included acetate as a carbon source. These conditions were used to minimize the effect on RNAs that otherwise might be altered due to the PS− state and light sensitivity (Drager et al., 1998). Cy3 and Cy5-labeled cDNAs were generated from total RNA and hybridized to gene-specific PCR products spotted on the glass microarray slides (Hegde et al., 2000). The green Cy3 (wild-type RNA) and red Cy5 (mutant RNA) fluorescent images were quantified. It should be noted that a change in RNA abundance may be a direct response to the loss of Mcd1 or, alternatively, may be an indirect response to one or more of the following: failure to express petD, the PS− state, or genetic background differences that could not be controlled in this experiment.

Figure 1A shows three replicate features (DNA spots) for six different genes from a single microarray hybridization. Fluorescence data for each feature were normalized to adjust for nonuniform labeling so that the overall Cy5/Cy3 ratio across all features was 1.0. This common normalization strategy assumes that RNAs that deviate up or down from the ratio of 1.0 will balance (Hegde et al., 2000). To test if normalization maintained data integrity, control rRNAs predicted to be unaffected in these mutants (chloroplastic 16S and 23S rRNAs, mitochondrial L7 rRNA, and cytosolic 25S rRNA) were checked and found to have <2-fold deviation from the ratio 1.0 (Table I). From these normalized data, RNAs with a >2-fold increase (red) or decrease (green) are highlighted in Table I. RNA gel blots were used to confirm the microarray data (Fig. 1B). These included the chloroplast petD, atpA, cemA, and psbA mRNAs and the mitochondrial nad2 mRNA. In addition, the cytosolic 25S rRNA (ethidium bromide stained) was included as a loading control in RNA gel blots. Quantitative microarray data are presented as scatter plots in which each data point is the average from three hybridized slides with each slide having three replicate spots for each gene (Fig. 2).

Figure 1.

Differential RNA accumulation in mcd1-1, mcd1-2, and mcd1-2/mcd2-1. A, Representative fluorescent images of three replicate spots for the indicated genes from a single microarray experiment. Test cDNAs (mcd1-1, mcd1-2, and mcd1-2/mcd2-1) were labeled with Cy5 (red), while the reference cDNAs (wild type [WT] and mcd1-2 when compared to the suppressor) were labeled with Cy3 (green). B, RNA gel blots were used to test for petD, atpA, cemA, psbA, and nad2 mRNAs. Ethidium bromide-stained 25S rRNA is the loading control. Polycistronic transcripts for atpA and cemA are indicted and numbers correspond to polycistronic transcripts in Figure 3A. C, Hierarchical clustering of expression ratio data for the two mcd1 mutants (mcd1-1, mcd1-1 versus wild type; mcd1-2, mcd1-2 versus wild type). Ratio data are averages from three independent hybridizations. Green represents a decrease in RNA abundance in the mutant strains and red represents an increase in RNA abundance. The color key for log2 fold changes in expression (−3 to +3) is shown. Gene names are shown at the right, and arrows indicate RNAs tested in B. Vertical lines marked 1 and 2 highlight RNAs discussed in the text.

Table I.

Chlamydomonas microarray data^a

^aRatio values are the mean ± sd of the mean determined from three independent hybridizations each with at least three replicates for each gene. Ratios ≥2.0 (2-fold increase) are shown in red. Ratios ≤0.50 (2-fold decrease) are shown in green.

^bLocation of gene. CP, Chloroplast; MT, mitochondrial; NL, nuclear.

^cOrganellar targeted location of nucleus-encoded proteins shown in parentheses. CP, Chloroplast; MT, mitochondria.

^dmcd1-1 to wild-type ratio.

^emcd1-2 to wild-type ratio.

^fmcd1-2/mcd2-1 to mcd1-2 ratio.

Figure 2.

Scatter plots of the three microarray experiments. Each point indicates the average fluorescent of test cDNA (y axes) plotted against the reference cDNA (x axes) in log scale. Diagonal lines 10 and 2 indicate the 10-fold and 2-fold increase cut-off lines, and −10 and −2 diagonal lines indicate the 10-fold and 2-fold decrease cut-off lines. The diagonal line 1 indicates the ratio of 1.0 with no change in RNA between two cells. Selected RNAs with significant differences between strains are labeled.

In microarrays and RNA gel blots petD mRNA was reduced by >25-fold in both mcd1-1 and mcd1-2 mutants, consistent with our previous report (Drager et al., 1998). petD was the only RNA that had a significant change in abundance in both mutants (Fig. 2; Table I). However, we cannot rule out that some of the untested mRNAs might be affected in the mutants. The mcd1-2 mutant had additional transcripts altered in abundance that were not changed in mcd1-1. In mcd1-2, the chloroplast atpA mRNA, which encodes the α-subunit of ATP synthase, was reduced by almost 10-fold as compared to wild type. The chloroplast cemA mRNA (also known as ycf10), which is involved in carbon assimilation (Rolland et al., 1997), had an apparent increase in abundance in mcd1-2. The extent of this increase varied among the three hybridizations, as is reflected in the sd (Table I). The cemA gene is part of the coordinately expressed atpA/psbI/cemA/atpH gene cluster (Drapier et al., 1998). Wild-type cells accumulate multiple polycistronic transcripts for the atpA and cemA chloroplast mRNAs, four of which are labeled in Figure 1B. In RNA gel blots the di- and tricistronic cemA transcripts appeared slightly more abundant in mcd1-2 as compared to wild type, an observation consistent with the microarray data (Table I). These data indicated that the mitochondrial nad2 and nad4 mRNAs, which encode NADH dehydrogenase subunits, were also more abundant in mcd1-2. This was confirmed for nad2 by RNA gel blots, although the relative abundance of this transcript was quite low making it impossible to quantify accurately (Fig. 1B).

To determine the sensitivity of these microarray experiments to subtle changes in RNA abundance a suppressor strain was tested. This strain is a double mutant, carrying both the mcd1-2 petD instability mutation and the mcd2-1 mutation that suppresses mcd1-2 and restores photosynthesis (Esposito et al., 2001). In the double mutant approximately 10% of wild-type levels of petD mRNA and approximately 50% of SUIV accumulate. To test if these microarray experiments were sufficiently sensitive to detect this small increase of petD mRNA, the suppressor was compared to mcd1-2. The faint but red appearance of the petD features in the microarray image (Fig. 1A) was consistent with a small increase of petD in the suppressed strain, as shown by RNA gel blots (Esposito et al., 2001; Fig. 1B). Although the average petD expression ratio for the mcd1-2/mcd2-1 versus mcd1-2 was 4.37 (Fig. 2; Table I), the relatively high sd (±1.84) suggests that this was near the limit of detection. Additional RNAs (nuclear Oee2 and PetE and the chloroplastic ORF2971B, rpoC2, and trnM2) had expression ratios >2 in the mcd1-2/mcd2-1 versus mcd1-2 experiment, although some of these ratios have a relatively high sd of the mean. In contrast, the ratios for these same RNAs were near one in the mcd1-1 and mcd1-2 versus wild-type experiments (Table I). Finally, the small increase of atpA mRNA (ratio = 2.79 ± 1.63) in the mcd1-2/mcd2-1 versus mcd1-2 microarray experiment was not clearly observed in RNA gel blots (Fig. 1, A and B). This is likely due to the relatively low abundance of atpA in these strains.

To identify genes with possible common regulatory responses, unsupervised hierarchical clustering was done with The Institute for Genome Research's (TIGR) Multiple Experiment Viewer software (Saeed et al., 2003) using normalized and log2 transformed microarray ratios. Figure 1C shows a dendogram of genes grouped by RNA expression patterns in the PS− mcd1 mutants as compared to photosynthetic (PS+) wild type. Two gene groups are highlighted in Figure 1C. First, the petD and atpA mRNAs (1) were distinct and separate due to the dramatic decrease (green) of petD in both strains (mcd1-1 and mcd1-2) and decrease of atpA in mcd1-2, as previously described. A second group of RNAs (2) appeared more abundant (red) in mcd1-2. This cluster includes four mitochondrial transcripts (cob, cox1, nad2, and nad4) as well as the chloroplast cemA and rps18 transcripts.

atpA and cemA Polycistronic mRNAs Are Altered in mcd1-2

Complex transcription units complicate interpretation of microarray data since the labeled cDNAs from different transcripts can hybridize to the same DNA feature on a slide. Moreover, a net change in hybridizable signal for a gene could result from the alteration of one or a subset of overlapping transcripts. For these reasons microarrays need to be supplemented by a second method that distinguishes each transcript.

Since the atpA and cemA mRNAs in mcd1-2 had a predicted >2-fold change in abundance with microarrays and these transcripts are expressed from the atpA/psbI/cemA/atpH gene cluster, RNA gel blots were used for further analysis. In wild-type cells this gene cluster generates eight detectable transcripts, including a tetracistronic RNA with all four coding regions, three monocistronic RNAs (cemA does not accumulate a monocistronic RNA), and multiple di- and tricistronic transcripts (Drapier et al., 1998; Fig. 3A). Figure 3B shows RNA gel blots for wild type, the two mcd1 mutants, and mcd1-2/mcd2-1. In wild type and mcd1-1 the four normal cemA-containing transcripts were detected; their identities are indicated at the right of section B. In contrast, both mcd1-2 and mcd1-2/mcd2-1 failed to accumulate the tetracistronic (no. 1) and atpA-containing tricistronic (no. 2) transcripts. Yet, in these cells the non-atpA containing cemA transcripts (nos. 5 and 6) accumulated to an amount higher than in the wild type. As shown in Figure 1B, monocistronic atpA mRNA (no. 4) was significantly reduced, while the faint dicistronic atpA-psbI transcript (no. 3) appeared to be only slightly reduced in mcd1-2, consistent with previous reports (Drapier et al., 2002). To summarize these findings, the transcripts affected in mcd1-2 are indicated by asterisks in Figure 3A. Ethidium bromide stained 25S rRNA was used as a loading control and psbA was tested as an apparently independent chloroplast mRNA (Fig. 3B).

Figure 3.

RNA gel blots and immunoblots to test accumulation of transcripts and proteins from the chloroplast atpA/psbI/cemA/atpH gene cluster. A, Diagram of the eight transcripts produced from this gene cluster as previously reported (Drapier et al., 1998). Filled boxes are gene coding regions, and numbered heavy arrows represent known transcripts with the corresponding coding regions. Small bent arrows show the known promoters. Asterisks indicate those atpA-containing transcripts that are reduced in mda1-2-containing strains. B, RNA gel blots were probed for chloroplast cemA and psbA mRNAs with ethidium bromide-stained 25S rRNA as the loading control. Transcripts labeled at right correspond to those labeled in A. C, Immunoblots testing accumulation of chloroplast SUIV protein and the ATP synthase α- and β-subunits. Proteins were visualized by enhanced chemiluminescence.

To assess if the reduction in atpA transcripts resulted in a corresponding decrease for the encoded ATP synthase α-subunit, immunoblots were performed. Proteins were separated by SDS-PAGE and reacted with antibodies specific to chloroplast α- and β-subunits of ATP synthase, as well as SUIV (Fig. 3C). In both mcd1 mutants SUIV failed to accumulate, but in mcd1-2/mcd2-1 SUIV accumulated to near-normal levels, consistent with our previous report (Esposito et al., 2001). All strains accumulated to what appeared to be wild-type levels of α- and β-subunits, despite the reduction of atpA mRNA in mcd1-2 and mcd1-2/mcd2-1 (Fig. 3C). This inconsistency between RNA and protein levels is not uncommon in chloroplasts, and a large reduction in mRNA does not necessarily cause a decrease in protein (Eberhard et al., 2002). This observation was also reported for α-subunit in the Chlamydomonas mda1-ncc1 mutant in which the atpA mRNA is significantly reduced, but there is not a corresponding decrease in α-subunit (Drapier et al., 1998, 2002).

mcd1-2 Contains a Second Mutation That Causes atpA Reduction

The mcd1-1 and mcd1-2 mutant alleles are known to have distinct lesions due to the fact that mcd2-1 can suppress mcd1-2 but not mcd1-1 (Esposito et al., 2001). The reduced accumulation of atpA mRNA in mcd1-2 but not in mcd1-1 could be due to an allele-specific phenotype in which mcd1-2 affects both petD and atpA mRNAs. Alternatively, the reduced atpA phenotype could be due to a second mutation in the mcd1-2 strain that is not present in the mcd1-1 strain. Such a second mutation, if present, could be in the known Mda1 gene. Genetic crosses and RNA analyses of the progeny could distinguish these two possibilities. If two unlinked mutations exist, one that affects petD mRNA (mcd1-2) and a second that affects atpA mRNA (e.g. a new allele mda1-2), then a mcd1-2 strain × wild-type cross would generate four possible progeny: wild type, atpA and petD double mutant (mcd1-2/mda1-2), petD single mutant (mcd1-2), and an atpA single mutant (mda1-2). In contrast, if mcd1-2 is a pleiotropic mutation, then one-half of the progeny would be mutant (fail to accumulate atpA or petD) and the other one-half would be wild type.

To test the hypothesis that mcd1-2 also carries mda1-2, we crossed it to wild type. Progeny from 12 tetrads were analyzed for accumulation of petD, atpA, cemA, and psbA mRNAs using RNA gel blots, with 25S rRNA as a loading control. Data for the parents and progeny from two representative tetrads are shown in Figure 4A, and these clearly indicate that the defects in atpA and petD mRNA accumulation are from unlinked mutations and are consistent with the hypothesis that the mcd1-2 parental strain is a double mutant. Based on RNA phenotypes, tetrad 43 is a nonparental ditype with two progeny (43-1 and 43-3) that failed to accumulate petD but had normal levels of all four atpA transcripts. These included the low abundance tetracistronic (no. 1) and dicistronic (no. 2) RNAs detected with the cemA probe (Fig. 4A). In contrast, the other progeny (43-2 and 43-4) had reduced levels of monocistronic atpA (no. 4) but normal levels of petD mRNA. Tetrad 21 is a tetratype and all four possible phenotypes and thus presumed genotypes were observed among these progeny (21-1 = wild type; 21-2 = mcd1-2/mda1-2; 21-3 = mda1-2; 21-4 = mcd1-2; progeny 21-1 RNA was overloaded as revealed by the loading controls). Of the 12 tetrads, two were parental ditypes, four were nonparental ditypes, and six were tetratypes. Photosynthetic growth phenotypes of these progeny were tested and, as expected, progeny without petD mRNA were PS− and progeny with petD mRNA were PS+, despite having reduced atpA mRNA (Fig. 4A). As we show below, the second nuclear mutation that causes reduced atpA mRNA accumulation is indeed an allele of Mda1, justifying our tentative reference to it as mda1-2.

Figure 4.

RNA gel blots of parents and representative progeny from genetic crosses. A, Parents and tetrad numbers 43 and 21 from the cross shown at the top were subjected to RNA gel-blot analysis for atpA, cemA, petD, and psbA chloroplast mRNAs, with ethidium bromide-stained 25S rRNA as a loading control. Numbered lanes (1–4) correspond to four progeny within a tetrad, and the inferred genotype of each progeny is shown at bottom, along with photosynthetic phenotypes (+ or −). B, Parents, tetrad numbers 2 and 3 progeny, and a wild-type control for the cross shown at the top were tested for atpA and petD mRNAs, with ethidium bromide-stained 25S rRNA as a loading control.

As discussed above, the mda1-ncc1 mutation confers the same phenotype as that reported here for certain segregants (43-2, 43-4, and 21-3) of the cross shown in Figure 4A. To test if these two mutations are allelic, the putative mda1-2 strain was crossed to the known mda1-ncc1 strain, and progeny were tested for atpA and petD mRNAs (Fig. 4B). Because detection of low abundance tetra- and tricistronic atpA transcripts can be variable in mda1-ncc1 (Drapier et al., 1998), the primary diagnostic for mda1 is the relative abundance of the monocistronic atpA (no. 4) to dicistronic atpA-psbI (no. 3); these transcripts were approximately equal in abundance in mda1-ncc1 and mda1-2 (Figs. 1B and 4B). Wild-type cells, in contrast, have 5 to 10 times more monocistronic atpA than dicistronic atpA. If putative mda1-2 is allelic to mda1-ncc1, then all progeny from this cross would inherit a mutant allele of Mda1 and have the reduced monocistronic atpA mRNA phenotype; otherwise some tetrads would contain one or two progeny with wild-type atpA levels.

To test this, progeny with the mda1-2 genotype (43-3 and 21-3) from the cross in Figure 4A were crossed to mda1-ncc1. Note all parents for this cross carry the wild-type Mcd1 allele and thus accumulate wild-type levels of petD mRNA. Twelve complete tetrads were tested, and all 48 progeny from these tetrads had the mutant phenotype with roughly equal amounts of monocistronic and dicistronic atpA mRNAs, as exemplified in Figure 4B. For these RNA gel blots petD was also detected, as well as the ethidium bromide-stained 25S rRNA loading control. In summary, genetic data indicate that mda1-2 is not linked to mcd1-1, but is tightly linked to mda1-ncc1 and confers the same phenotype. Thus, it is most likely a second mutant allele of the Mda1 gene. We cannot absolutely rule out the possibility that the mda1-2 mutation is in a second gene tightly linked to Mda1; based on the number of progeny analyzed, the maximum distance would be 4.2 cM. Due to the improbable coincidence of mutations in different genes being tightly linked and giving the same phenotype, we conclude this to be a new allele of Mda1 and name it mda1-2. Thus, microarray analysis led to the identification of a potentially valuable second mutant allele of a known RNA stability locus.

DISCUSSION

We conclude that DNA microarrays are efficient and useful for characterizing mutants that affect organellar RNA accumulation. This method makes it possible to evaluate many chloroplast and mitochondrial RNAs simultaneously, improving the characterization of regulatory genes of interest. The mcd1-1 and mcd1-2 mutants, shown previously to lack petD mRNA, were tested with microarrays to identify organellar RNAs dependent on Mcd1. These mutants were grown in the presence of a carbon source and low light, allowing them to generate ATP and lessening the stress due to high light. Interestingly, no large-scale change in chloroplast and mitochondrial RNAs was evident in response to the PS− condition. In fact, petD is the only RNA significantly altered in mcd1-1. The mcd1-2 strain, in contrast, had an additional near 10-fold decrease in atpA mRNA. Furthermore, four mitochondrial RNAs (cob, cox1, nad2, and nad4) were slightly up-regulated in mcd1-2, perhaps a response to the mda1-2 allele or perhaps reflecting additional genetic differences with respect to mcd1-1. Chlamydomonas mitochondrial genes appear to be coordinately expressed by two promoters, one per strand (Gray and Boer, 1988), and mcd1-2 could carry a genetic variant that affects transcription from these promoters or possibly RNA stability.

Based on genetic linkage and phenotypes the reduced atpA mRNA in mcd1-2 is due to a second mutant allele of the characterized Mda1 nuclear gene, which we have named mda1-2. This allele may prove helpful for cloning Mda1 using either map-based cloning or complementation, since both known mutant alleles are spontaneous and hence likely to represent point mutations. Having multiple alleles may also be useful in characterizing functional domains of the gene and/or encoded protein. Since mcd1-2 is really a double mutant, RNAs altered in this strain but not in mcd1-1 are likely caused by mda1-2 or the combination. Array analysis of mda1-ncc1 would help to clarify this issue. The fact that mda1-2 was present in both mcd1-2 and mcd1-2/mcd2-1 indicates that mda1-2 arose antecedent to the suppressor screen. However, the mutation was not discovered until microarrays were employed, validating this method for screening such mutants.

Most laboratories keep Chlamydomonas strains on culture plates and passage them regularly under vegetative growth conditions. Indeed, the main Chlamydomonas culture collection is largely stored in this way (http://www.biology.duke.edu/chlamy/). Like any organism maintained in a growth state, such cells will tend to adapt to the laboratory environment and may acquire additional and undesired mutations. In a case such as mcd1, the PS− growth phenotype means that an additional mutation affecting photosynthesis would be masked. Since mda1-2 does not strongly affect photosynthesis in a wild-type chloroplast DNA background, it would not be detected even upon outcrossing to a wild type unless the proper RNA analysis was performed. This demonstrates the power of microarrays to find the needle in the haystack, and allows for easy purification of mcd1-2 not carrying mda1-2. Similar confounding mutations have been discovered before in Chlamydomonas. For example, the nuclear mutant F34 was originally reported to be specifically deficient in the chloroplast psbC gene product but subsequently, another laboratory reported that psbA was additionally affected (Jensen et al., 1986; Rochaix et al., 1989). Through genetic crosses it was revealed by a second laboratory that F34 had accumulated two additional mutations: a suppressor of F34 and a second, unlinked mutation that conferred the psbA phenotype (Girard-Bascou et al., 1992). In some cases, microarray analysis may offer shortcuts to resolve such ambiguities.

The two mcd1 mutants analyzed in this study have a block in photosynthetic electron transport (PET) due to the loss of SUIV from the cytochrome b₆/f complex (Drager et al., 1998). Consistent with this, the chlorophyll fluorescence F_max values in the mutants are identical to wild-type cells treated with the electron transport inhibitor 3-(3,4-dichlorophenyl)-1,1-dimethylurea (DCMU; Drager et al., 1998). The mcd1 mutations would cause reduction/oxidation (redox) changes in chloroplasts. Such redox changes have been reported to control the transcription and stability of chloroplast RNAs and nucleus-encoded mRNAs that encode chloroplast proteins (Pfannschmidt, 2003). The chloroplast redox state can be changed using different wavelengths of light that favor formation of either PSI or PSII complexes or with PET inhibitors, and this was reported to affect transcription rates and accumulation of endogenous chloroplast mRNAs in the mustard plant (Pfannschmidt et al., 1999) or mRNAs from chloroplast reporter genes in tobacco (Nicotiana tabacum; Pfannschmidt et al., 2001). Using a nylon membrane array of Arabidopsis (Arabidopsis thaliana) nucleus-encoded cDNAs thought to encode chloroplast proteins, it was similarly reported that different redox and photosynthetic conditions caused large-scale changes in RNA accumulation (Richly et al., 2003). In Chlamydomonas, redox changes due to DCMU or altered light conditions were reported to affect the stability of a heterologous GUS mRNA as well as endogenous atpB, psaB, and rbcL mRNAs (Salvador and Klein, 1999).

In contrast to the redox studies, the microarray data presented here do not support a photosynthesis-driven redox state that alters accumulation of RNAs. Aside from petD, no other chloroplast RNAs are significantly altered in mcd1-1 (Fig. 2; Table I). Additional RNAs were apparently affected in mcd1-2, but these turned out to be due to mda1-2. There are several possible explanations for this discrepancy in observed redox effects. One, the redox state in algal or vascular plant chloroplasts might not play a significant role in controlling RNA accumulation and instead other signals such as light or ATP/ADP levels might be involved (Danon and Mayfield, 1994). Two, blocking different steps along the PET chain might affect RNA accumulation differently, and this may vary among photosynthetic eukaryotes. Three, redox effects on RNA accumulation might be too subtle (<2-fold change) to be detectable by these microarrays. In support of the first explanation, Matsuo and Obokata (2002) concluded that Chlamydomonas photosystem I mRNAs are dependent on light induction but not on the redox state of plastoquinone. While in tobacco, a chloroplast psbA deletion mutant had no change in PSI, cytochrome b₆/f, or PSII light harvesting complex proteins, and no changes were observed for chloroplast mRNAs that encode the NAD(P)H dehydrogenase or plastid terminal oxidase enzymes (Baena-Gonzalez et al., 2003).

This study expands the list of microarray experiments that investigate RNA stability and degradation to include organellar RNA stability factors. Previously, either glass slide or nylon arrays were used to investigate transcriptional regulation, for example in an Arabidopsis mutant defective for the chloroplast SIG2 protein (Nagashima et al., 2004), in tobacco plants engineered to lack the prokaryotic-type chloroplast RNA polymerase (Legen et al., 2002), or in mitochondria of wild-type Arabidopsis plants (Giege et al., 2000). These studies together illustrate the wide application of arrays that is possible in organelle molecular biology.

With respect to post-transcriptional control, arrays have been used successfully in several systems. In Escherichia coli, global RNA half-live estimates were made using RNA extracted from rifampicin-treated cells (Selinger et al., 2003). A correlation between gene function and relative RNA half-life was discovered in these bacteria, as well as a novel rifampicin-insensitive promoter. A similar study was done with yeast using the temperature-sensitive RNA polymerase II mutant to block transcription and obtain global RNA stability estimates (Grigull et al., 2004). This group also used deadenylase mutants to show that Ccr4p is the major yeast mRNA deadenylase. Arabidopsis microarrays were used to identify the plant RNA targets for two types of RNA regulatory proteins. First, Pérez-Amador et al. (2001) identified the targets of the Dst regulator gene that promotes destabilization of inherently unstable cytosolic mRNAs that contain the 3′ untranslated region DST regulatory sequence and more recently, this group investigated targets of the cytosolically localized AtXrn4 exoribonuclease (Souret et al., 2004).

In summary, Chlamydomonas organellar microarrays provided an efficient and useful method to further characterize Mcd1, confirmed the gene-specific RNA substrate (petD) of Mcd1, identified the new mda1-2 allele, and confirmed the Mda1 RNA substrates to be atpA-containing transcripts. In the future, this genomic tool will likely prove useful to investigate other organellar RNA metabolism mutants and even translational regulation by comparing polysomal versus nonpolysomal RNAs.

MATERIALS AND METHODS

Strains and Culture Conditions

The following strains were used in this study: P17 wild-type cells (Stern et al., 1991) that were derived from the strain CC373 (Shepherd et al., 1979). mcd1-1 (also known as F16) and mcd1-2 (also known as 670.1) are petD mRNA stability mutants that have different genetic backgrounds (Drager et al., 1998), mcd1-2/mcd2-1 (also known as SUP670) is a double mutant that is PS+ (Esposito et al., 2001), and the mda1-ncc1 atpA stability mutant (also known as ncc1) is also PS+ (Drapier et al., 1998). All cells were grown in TAP medium (Harris, 1989) under constant low-level fluorescent light.

RNA Isolation and RNA Gel Blots

Total RNA was isolated from 10 mL of cells grown to 2 × 10⁶ cells/mL as previously described (Drager et al., 1999). RNA was separated in 1% agarose/3% formaldehyde gels, blotted to GeneScreen membranes (PerkinElmer-NEN, Boston), and cross-linked to the membrane by exposure to UV light (UV Stratalinker 1800, Stratagene, La Jolla, CA). RNA gel blots were hybridized as previously described (Church and Gilbert, 1984), using ³²P-labeled DNA probes for petD (Higgs et al., 1998), psbA (Higgs et al., 1998), atpA (Drapier et al., 1998), nad2 (GenBank accession no. NC_001638), and cemA (Drapier et al., 1998). Radioactive bands were visualized by exposure to x-ray film at −75°C with an intensifying screen (Fisher Scientific, Chicago).

Microarrays, Fluorescent Labeling, and Hybridization

Chlamydomonas organellar microarrays spotted with gene-specific PCR products were generated at the Center for Gene Expression Profiling (CGEP) located at the Boyce Thompson Institute for Plant Research, using methods previously described (Hegde et al., 2000). Gene-specific PCR primer sequences for chloroplast, mitochondrial, and nuclear genes related to organellar biology were obtained from the Chlamydomonas genomic sequences: chloroplast (Maul et al., 2002), mitochondrial (GenBank accession no. NC_001638), and nucleus (http://genome.jgi-psf.org/chlre2/chlre2.home.html). Primers were designed to amplified products from 150 to 1,500 bp in length and contain coding sequence but little to no noncoding sequence. Organellar genes were amplified using standard PCR with Taq DNA polymerase (Fisher Scientific) and total DNA. For nuclear genes, reverse transcriptase-PCR with the Access PCR System (Promega, Madison, WI) was used to amplify cDNA from total RNA. All PCR products were subjected to agarose gel electrophoresis, size confirmed, and single bands were excised from the gel using the Qiaex II (Qiagen, Valencia, CA) and subcloned into pGEM T-tailed PCR vector (Promega) or PCR-II (Invitrogen, Carlsbad, CA). Positive bacterial colonies were selected and plasmid DNA extracted. To confirm correct PCR inserts, DNA was sequenced with vector primers (M13 forward/reverse).

To generate large amounts of DNA for spotting onto the GAPII (Corning, Corning, NY) glass microarrays slides, standard PCR with universal vector primers (M13 forward/reverse) and Taq DNA Polymerase (Fisher Scientific) were performed in replicate for all clones. Replicate PCR products were combined and purified using PCR filter plates (Millipore, Bedford, MA) in a 96-well format. Purified PCR products were resuspended in 20 μL of water, and 1 μL was checked for amplification quantity and purity by agarose gel electrophoreses. PCR and gel analyses were repeated until all products were obtained. Once confirmed, products were diluted with 20 μL dimethyl sulfoxide and stored at −80°C. Each DNA fragment was spotted three independent times on each GAPII slide using the BioRobotics Microgrid Pro Arrayer (BioRobotics, Woburn, MA) with a 16-split-pin spotting tool. Because of the relatively small size of the whole array, two complete arrays were spotted per glass slide. After spotting, slides were allowed to dry on the spotting stage overnight, then cross-linked with 300 μJ of 254-nm UV light and baked at 80°C for 4 h. Slides were blocked with succinic anhydride according to the manufacturer's directions (Corning), then stored under desiccation for up to 3 months. As a final check, all DNA samples used for making microarrays were resequenced a second time to confirm identity, since similar array studies were previously published (Lilly et al., 2002) and found to contain numerous incorrect annotations after the earlier study was withdrawn (Lilly et al., 2004). Only those genes determined to be correct were included for the data analyses reported here.

Fluorescently labeled Cy3 and Cy5 cDNA probes were synthesized from 30 μg of DNase-treated total RNA using a direct incorporation reverse transcription method as previously described (Hegde et al., 2000). In addition to oligo(dT) primer, 6 μg of random hexamers (Invitrogen) were added to prime synthesis from nonpolyadenylated chloroplast and mitochondrial RNAs. Labeled cDNA was passed through a PCR purification column (Qiagen) and eluted with 40 μL of sterile water. Corresponding Cy3 and Cy5 samples were combined and then vacuum dried. Pellets were resuspended in 22 μL hybridization solution [25% formamide, 5× SSC, 0.25% SDS, 100 ng/μL sheared and denatured herring sperm DNA, and 1 μL poly(A) blocker (Invitrogen)].

Hybridizations were conducted similarly to (Hegde et al., 2000) but with the following changes. Prepared and blocked slides were incubated for 1 h at 42°C in prehybridization buffer (25% formamide, 5× SSC, 0.25% SDS, 0.25% bovine serum albumin), washed with water, rinsed with 100% ethanol, and spun dry. Experimental and reference fluorescent-labeled cDNAs were combined in 22 μL hybridization buffer (see above), denatured at 95°C for 5 min, cooled to room temperature for 2 min, and then placed over the microarray, covered with a 22 × 22 mm coverslip (Z36, 590-4 Hybri-Slips, Sigma, St. Louis), and incubated in a sealed chamber (Corning, CMT Hybridization chambers 2551) at 42°C for 20 h. Slides were washed at room temperature in 2× SSC, 0.25% SDS to remove coverslips, followed by a 10-min wash in the same buffer at 42°C with vigorous agitation. Slides were transferred to 0.1× SSC and 0.2% SDS for 10 min at room temperature, followed by four 1-min rinses in 0.1× SSC. Slides were then plunged three times into sterile water for no more than 3 s total, rinsed in 100% ethanol, and spun dry.

Microarray Scanning and Data Analyses

Microarray slides were scanned for Cy5 and Cy3 fluorescent signals using either the GenePix 4000B (Axon Instruments, Union City, CA) or ScanArray 5000 (Packard BioScience, Billerica, MA), both set to a resolution of 10 μm. Data obtained from the two scanners were quite similar, indicating no clear scanner-specific biases. Laser and photomultiplier tube voltages were adjusted to minimize background and number of saturated spots and to balance the overall Cy3 and Cy5 signals. Fluorescent signal at each spot was measured, and the Cy5 to Cy3 ratio was determined for each spot as described in “Results.”

Normalized and averaged ratio data were used to generate scatter plots. Hierarchical clustering was done using TIGR Multi-Experiment Viewer (TMeV) software version 2.2 (Saeed et al., 2003). The normalized and averaged ratio data were log2 transformed then imported into the TMeV software. A clustered dendogram was generated using the Euclidean metric with complete linkage, and the clustered image was colored to represent the fold increase (red) or decrease (green). Microarray data were submitted and can be accessed from the GEO database (Edgar et al., 2002) at NCBI under the GEO Platform ID number (GPL1395), GEO Series ID number (GSE1651), and individual GEO samples (GSM28409 to GSM28417).

Protein Isolation and Immunoblots

Total proteins were isolated and size fractionated by SDS-PAGE (Chen et al., 1993). Antibodies specific for the chloroplast ATP synthase α- and β-subunits (Drapier et al., 1998) and SUIV (Drager et al., 1998) were reacted with immunoblots and detected by enhanced chemiluminescence.

Genetic Analysis

Crosses and dissection of tetrads were performed as previously described (Harris, 1989). Photosynthetic phenotypes were determined by growth on TAP versus minimal medium (Harris, 1989). From this cross, progeny (43-3 and 21-3) with reduced atpA but normal petD RNA levels were obtained and termed mda1-2.