Ribonuclease II preserves chloroplast RNA homeostasis by increasing mRNA decay rates, and cooperates with polynucleotide phosphorylase in 3′ end maturation


  • Arabidopsis seed stocks: SALK_013306 (pnp1-1), CS85511 (P184L), CS87183 (pnp1-3), CS86422 (S202N ), CS93456 (A263V ), SALK_090294 (rnr1-3).

(e-mail ds28@cornell.edu).


Ribonuclease R (RNR1) and polynucleotide phosphorylase (cpPNPase) are the two known 3′→5′ exoribonucleases in Arabidopsis chloroplasts, and are involved in several aspects of rRNA and mRNA metabolism. In this work, we show that mutants lacking both RNR1 and cpPNPase exhibit embryo lethality, akin to the non-viability of the analogous double mutant in Escherichia coli. We were successful, however, in combining an rnr1 null mutation with weak pnp mutant alleles, and show that the resulting chlorotic plants display a global reduction in RNA abundance. Such a counterintuitive outcome following the loss of RNA degradation activity suggests a major importance of RNA maturation as a determinant of RNA stability. Detailed analysis of the double mutant demonstrates that the enzymes catalyze a two-step maturation of mRNA 3′ ends, with RNR1 polishing 3′ termini created by cpPNPase. The bulky quaternary structure of cpPNPase compared with RNR1 could explain this activity split between the two enzymes. In contrast to the double mutants, the rnr1 single mutant overaccumulates most mRNA species when compared with the wild type. The excess mRNAs in rnr1 are often present in non-polysomal fractions, and half-life measurements demonstrate a substantial increase in the stability of most mRNA species tested. Together, our data reveal the cooperative activity of two 3′→5′ exoribonucleases in chloroplast mRNA 3′ end maturation, and support the hypothesis that RNR1 plays a significant role in the destabilization of mRNAs unprotected by ribosomes.


Substantial regulation of chloroplast gene expression is achieved at the post-transcriptional level by a variety of nuclear-encoded endo- and exoribonucleases, along with a diverse population of RNA-binding proteins (reviewed by Stern et al., 2010; Barkan, 2011; Stoppel and Meurer, 2011). Inefficient transcription termination (Stern and Gruissem, 1987) creates long RNA precursors of varying lengths, which are modified by intercistronic RNA processing (reviewed by Barkan, 2011), including endonucleolytic cleavage and 3′ and 5′ end exonucleolytic trimming. These events, combined with additional post-transcriptional events such as splicing and editing, give rise to an accumulating RNA population. The catalytic and RNA-binding proteins that mediate these activities have a mixture of prokaryotic origins resulting from endosymbiosis, and eukaryotic origins representing acquired functions.

Two enzymes whose importance has been demonstrated in chloroplasts are the prokaryotic-type 3′→5′ exoribonucleases ribonuclease R (RNR1) and polynucleotide phosphorylase (PNPase). Both enzymes degrade single-stranded RNA processively, but PNPase releases ribonucleoside-5′-diphosphate by catalyzing a reversible phosphorolytic reaction, whereas RNR1 hydrolyzes RNA to ribonucleoside-5′-monophosphates (reviewed by Nicholson, 1999). Two isoforms of PNPase are encoded by the nucleus of plants and targeted to either the mitochondria (mtPNPase) or the chloroplast (cpPNPase). While only mtPNPase is essential for embryo development, both proteins have been shown to have major roles in maturation and/or degradation of mRNA, rRNA and tRNA, as well as in biogenesis and/or accumulation of non-coding RNAs (Walter et al., 2002; Perrin et al., 2004a,b; Holec et al., 2006; Marchive et al., 2009; Germain et al., 2011; Hotto et al., 2011). Recombinant cpPNPase is inhibited by RNA secondary structures (Yehudai-Resheff et al., 2001), and cannot progress through regions protected by RNA-binding proteins such as PPR10, a pentatricopeptide repeat protein that defines RNA ends in the atpIatpH intergenic region (Prikryl et al., 2011).

RNR1 was first reported to be a mitochondrial protein, where it was named AtmtRNaseII, because it was shown to trim the 3′ end of atp9 mRNA initially generated by mtPNPase (Perrin et al., 2004b). This enzyme was subsequently found in the chloroplast, and T-DNA-induced null mutants were shown to be chlorotic with severe chloroplast rRNA defects (Kishine et al., 2004; Bollenbach et al., 2005). In vitro analysis of the PPR10 protein, however, showed that the native 3′ end of atpI could not be fully generated by cpPNPase, and it was suggested that RNR1/RNase II might complete the process (Prikryl et al., 2011). Ambiguity over the name of this protein derives from the fact that it is apparently the sole organellar representative of the superfamily of exoribonucleases that includes both RNase R and RNase II, which have distinct sensitivities to RNA secondary structure and consequently different biological roles (Matos et al., 2011). To avoid confusion with previous literature, we will use RNR1 here to refer to RNR1/RNase II, although the results presented below suggest that it would more appropriately be named RNase II in future research.

To define the functions of plant RNR1 in detail, and to determine whether it cooperates with PNPase in maturation and/or degradation of chloroplast mRNA, we have taken a genetic approach. Escherichia coli strains lacking both RNase II and RNase R are essentially normal, whereas strains depleted for either enzyme as well as PNPase are non-viable, illustrating the non-redundant and essential roles of the two types of exoribonuclease (Donovan and Kushner, 1986; Cheng et al., 1998). Here, we demonstrate the embryo lethality of the pnp1-1/rnr1-3 double mutant in Arabidopsis, and circumvent it by taking advantage of weak cpPNPase mutant alleles (Germain et al., 2011). In those double mutants, we find global reduction in chloroplast RNA accumulation, suggesting that correct RNA maturation is required for RNA stability. We also find that, as in plant mitochondria, RNR1 completes mRNA 3′ maturation initiated by PNPase. Finally, we expand our understanding of RNR1 in chloroplast RNA metabolism, specifically its role in the decay of non-polysomal mRNA. This points to a conflict of the current model of chloroplast RNA metabolism with our data, and we suggest a possible explanation linked to the sublocalization of cpPNPase and RNR1 within the chloroplast.


Analysis of cpPNPase–RNR1 double mutants

To investigate whether cpPNPase and RNR1 had completely distinct roles, we attempted to obtain a pnp/rnr1 double mutant by crossing the previously described T-DNA insertion null mutant pnp1-1 (Germain et al., 2011) to three different rnr1 alleles (rnr1-1–rnr1-3; Bollenbach et al., 2005). Because DNA analysis did not reveal any doubly homozygous mutants in any of the three original F2 populations, we checked for embryo lethality by analyzing siliques from crosses homozygous for pnp1-1 and segregating for rnr1 mutant alleles. As shown in Table 1, statistical analysis showed a significantly higher embryo abortion rate where pnp1-1 rnr1-1 or pnp1-1 rnr1-3 double mutants were segregating, in comparison to progeny segregating only for rnr1-1 or rnr1-3. Progeny segregating only for rnr1-2 had an unexpectedly high abortion rate, although lower than the cognate double mutant, for reasons we cannot immediately explain. We can conclude, however, that the complete absence of both cpPNPase and RNR1 leads to a failure to complete embryo development in Arabidopsis.

Table 1.   Embryo lethality caused by pnp1 and rnr1 mutations
Parental genotypeProgeny embryo phenotypeP-value
  1. Plants of the parental genotypes shown were self-pollinated and immature siliques were dissected. Embryos were scored as aborted or survived. Comparison of abortion rates was performed using binary logistic regression analysis to determine if plants A had significantly lower abortion rates than plants B. P < 0.05 is considered significant.

PNP1/PNP1 RNR1/rnr1-12621623A0.0004
pnp1-1/pnp1-1 RNR1/rnr1-161473534B
PNP1/PNP1 RNR1/rnr1-2109430539A0.3049
pnp1-1/pnp1-1 RNR1/rnr1-2153372525B
PNP1/PNP1 RNR1/rnr1-313619632A0.0119
pnp1-1/pnp1-1 RNR1/rnr1-326520546B

To learn more about the effects on chloroplast gene expression of depleting both major 3′→5′ exoribonucleases, we took advantage of two single amino acid cpPNPase mutants with reduced PNPase activity, P184L and S202N, and a third allele with a wild type (WT) phenotype, A263V (Germain et al., 2011). Most chloroplast transcripts in P184L have an intermediate pattern between the null mutant and the WT, and the recombinant P184L protein displays a lower catalytic activity. The RNA phenotype of S202N is closer to the WT, with a small subset of transcripts showing intermediate phenotypes consistent with reduced cpPNPase activity. Figure 1 shows the growth phenotypes of progeny segregating for rnr1-3 and homozygous for P184L, S202N or A263V. The progeny giving rise to P184L/rnr1-3 plants display the most severe phenotypes, including the highest percentage of seeds failing to germinate (Table 2), numerous plants that do not proceed past early development and accumulate anthocyanins (pink arrowheads in Figure 1), as well as the most chlorotic plants compared with the rnr1-3 single mutant (black arrowheads in Figure 1). The germination failure and anthocyanin accumulation phenotypes were observed to a lesser extent in the progeny segregating S202N/rnr1-3, while the progeny segregating A263V/rnr1-3 only displayed WT and rnr1-3 phenotypes, as expected.

Figure 1.

 Segregating populations with viable pnp1/rnr1 double mutants. Plants indicated by black arrowheads were used for RNA analysis.

Table 2.   Segregation of pnp1/rnr1 double mutants
Parental genotypeProgeny seedling phenotype
No germination rnr1-3 WTTotal
  1. Plants of the parental genotype were self-pollinated and seeds germinated on MS medium. Seeds were scored as no germination, wild type (WT) or rnr1-3 phenotype if the cotyledons were pink upon germination, according to the phenotype described by Bollenbach et al. (2005).

P184L/P184L RNR1/rnr1-350 (11.5%)24 (5.5%)361 (83%)435
S202N/S202N RNR1/rnr1-317 (5.3%)56 (17.3%)250 (77.4%)323
A263V/A263V RNR1/rnr1-31 (0.3%)92 (21.9%)327 (77.8%)420

Accumulation of chloroplast rRNA is severely affected in cpPNPase–RNR1 double mutants

To examine how a combination of cpPNPase and RNR1 deficiency affects the metabolism of chloroplast RNA, total RNA was examined by ethidium bromide staining, with gel loading based on approximately equal nuclear 28S rRNA amounts (Figure 2). This revealed apparently similar chloroplast rRNA patterns in rnr1-3 and the double mutants, with a general reduction in their accumulation and slightly less distinct banding. The rRNAs, which are transcribed from a polycistronic operon, were then examined in more detail by RNA gel blot, as shown in Figure 3a. When probing for 16S rRNA, all rnr1-3 mutants accumulate lower levels of the mature species (marked 16S in Figure 3b), a phenotype previously described for rnr1-3 by Bollenbach et al. (2005), along with the accumulation of the 16S precursor observed in two of the mutants, rnr1-3 and A263V/rnr1-3 (pre-16S in Figure 3b). The accumulation of pre-16S is, however, similar to WT for P184L/rnr1-3, and intermediate in S202N/rnr1-3, indicating a role for cpPNPase in regulating the abundance of pre-16S. Analysis of the 23S and 4.5S rRNAs yielded similar results, with a reduced accumulation of mature rRNAs for the rnr1-3 mutants coupled with an overaccumulation of precursors except for P184L/rnr1-3, in which the precursor levels are similar to those of the WT (Figure 3c, d).

Figure 2.

 Gel electrophoresis to examine ribosomal RNAs.
Total RNA was isolated from whole plants and analyzed by ethidium bromide staining, revealing the reduced accumulation of chloroplast rRNAs for all mutants containing the rnr1-3 T-DNA insertion.

Figure 3.

 Analysis of the rrn operon.
(a) Diagram of the rrn operon and locations of the probes used for RNA gel blot analysis (black bars). Bars in the dashed box relate to the bands observed in panel (e).
(b)–(e) Results for the four specific probes as indicated at the bottom right of each panel. For 4.5S and 5S rRNAs, an additional shorter exposure is shown to facilitate comparison of mature transcript levels.

We were particularly intrigued by the accumulation pattern of 5S rRNA (Figure 3e). In contrast to the reduction in mature 16S, 23S and 4.5S rRNAs for all mutants homozygous for rnr1-3, the level of mature 5S transcript (band D in Figure 3e) is enhanced in P184L/rnr1-3 relative to other mutants homozygous for rnr1-3, to approximately 20% of the WT level (Figure S1 in Supporting Information). This result implies that cpPNPase plays a role in degradation of 5S rRNA. The rnr1-3 mutants also accumulate some precursors such as bands C, B and A as described in Sharwood et al. (2011b). The accumulation pattern of bands A and B is clearly different between rnr1-3 and the P184L/rnr1-3 and S202N/rnr1-3 double mutants, with an inverted ratio in favor of band A. This result suggests that PNPase has some involvement in the conversion of band A to band B, most likely 3′ end trimming of band A until encountering the secondary structure of trnR (see the right end of panel a in Figure 3).

Overall, analysis of the rRNA operon frequently reveals an additive effect in the double mutants with a general reduction of rRNA precursors after cpPNPase depletion on top of the already reduced amounts of mature rRNAs carried over from the lack of RNR1. The strongest phenotype was observed for P184L/rnr1-3 and was less noticeable for S202N/rnr1-3.

pnp/rnr1 double mutants display a global reduction in chloroplast mRNA levels, proportional to the decline in cpPNPase activity

To establish the collective importance of cpPNPase and RNR1 in mRNA maturation and stability, several polycistronic gene clusters as well as the monocistronic rbcL and psbA transcripts were analyzed by RNA gel blot (Figures 4 and S2). As previously described in detail (Germain et al., 2011), the pnp1-1 RNA pattern is very distinct from the WT, with many species migrating more slowly due to 3′ extensions as typified by the psbD/psbC/ycf9 gene cluster (Figure 4a). Another pnp1-1 RNA phenotype is the barely detectable bands in the higher molecular weight range for the atpI/atpH/atpF/atpA gene cluster, while the monocistronic atpH bands are still distinct (Figure 4b), albeit with 3′ extensions. The point mutants P184L and S202N display an intermediate phenotype between pnp1-1 and WT, while A263V is similar to WT. The rnr1-3 mutant has been previously described as possessing no obvious changes in mRNA patterns compared to WT, but our present analysis reveals higher accumulation of most mRNA species (e.g. petD, atpI, atpH and atpA probes). This observation is discussed further below.

Figure 4.

 Analysis of the psbD and atpI gene clusters.
(a) Results for the three specific probes as indicated in the bottom diagram (black bars) for the psbD gene cluster.
(b) Results for the three specific probes as indicated in the bottom diagram (black bars) for the atpI gene cluster.

The P184L/rnr1-3 double mutant stood out because of the sharp reduction in the abundance of all mRNA species, with the exception of certain atpH and petD transcripts (Figures 4 and S2). This reduction is particularly striking for the psbD gene cluster, where no distinct band is detected, especially when compared to the other mutants homozygous for rnr1-3. Nonetheless, a comparison of S202N/rnr1-3 to rnr1-3 shows a similar trend in some cases, for example with the atpI and atpA probes (Figure 4b). To summarize the RNA gel blot data presented so far: lack of cpPNPase increases transcript length and causes 3′ end heterogeneity; lack of RNR1 increases RNA abundance; deficiency of both leads to decreased transcript abundance. We interpret these observations as evidence that the presence of 3′→5′ exoribonuclease activity is required for transcript stability, perhaps because at least a minimal level of 3′ processing is essential.

To test whether the phenotypes shown in Figure 4 held for simpler transcription units, we examined accumulation of the psbA, rbcL and matK monocistronic transcripts (Figure S2c–e). These results recapitulate the trend seen for complex gene clusters, i.e. differential accumulation of mRNA species depending on the level of 3′→5′ exoribonuclease activity. Exceptions such as the smallest atpH band, which is derived from the middle of a gene cluster, indicate that RNA instability rather than lower transcription rates underpin these observations.

Mature RNA 3′ ends are generated by the cooperative action of cpPNPase and RNR1

As mentioned in the Introduction, mtPNPase and RNR1 cooperatively generate the 3′ end of the atp9 transcript (Perrin et al., 2004b), and in bacteria the two enzymes are complementary in the sense that PNPase is sensitive to secondary structures while RNase R is not (Arraiano et al., 2010). In addition, a double mutant hyperaccumulates small structured RNA (REP) fragments (Cheng and Deutscher, 2005). To delineate their respective roles in chloroplasts, we precisely mapped the 3′ ends of the atpI, atpH and petD transcripts using circular RT-PCR (Figure 5a–c). The atpI transcript was chosen because Prikryl et al. (2011) had suggested that RNR1 might trim the 3′ end of atpI to abut the PPR10-binding site. As shown in Figure 5a, we confirmed that the WT atpI 3′ end is located 3–5 nucleotides (nt) downstream of the minimal PPR10 binding site. In contrast, the 3′ ends found in the pnp1-1 mutant are heterogeneous and extend into the coding region of the neighboring atpH gene. This is consistent with the general finding that chloroplast RNAs are 3′-extended in this mutant. When the rnr1-3 single mutant was analyzed, no WT 3′ ends were found. Instead, we found extensions of 2–16 nt, suggesting that cpPNPase cannot fully resect the RNA immediately downstream of bound PPR10. As the activity level of cpPNPase is depleted from the rnr1-3 mutant, we witnessed an increase in the length of the 3′ extensions with those in S202N/rnr1-3 being 5–12 nt, while the majority of the extensions are approximately 12 nt in P184L/rnr1-3.

Figure 5.

 Precise 3′ end mapping of atpI, atpH and petD transcripts for wild type (WT), pnp1-1, rnr1-3, S202N/rnr1-3 and P184L/rnr1-3.
The 3′ ends are mapped below the sequences with each horizontal line representing a cloned circular RT-PCR product. The number in parentheses indicates the total number of clones sequenced. The 5′ ends are not shown because they were unaffected by genotype. UTR, untranslated region.
(a) The atpI-atpH intergenic region, with the minimal PPR10-binding site shaded in grey. The PPR10-dependent atpH 5′ end as well as the beginning of the atpH ORF, shaded in light purple, are also shown.
(b), (c) Putative stem–loops defining the 3′ ends of the atpH and petD transcripts.

For atpH and petD, stem–loop structures are believed to define 3′ ends by blocking further 3′→5′ activity. For these genes, analogous results to atpI were obtained, with progressively longer 3′ extensions as cpPNPase activity was depleted from the rnr1-3 mutant background. In the case of pnp1-1, only the 3′ ends that mapped close to the WT are displayed, although a population of transcripts extending much further downstream also exists (Figures 2c and 3c in Germain et al., 2011). In contrast to atpI and petD, atpH 3′ ends were heterogeneous for both the WT and pnp1-1, including sequences that are part of the predicted stem–loop structure. We suspect that the 9-nt bulge leads to a dynamic secondary structure, allowing RNR1 to process the transcript further than cpPNPase.

Non-polysomal mRNAs over-accumulate in the absence of RNR1

As shown above, examination of the rnr1-3 single mutant revealed overaccumulation of most mRNA species, especially precursors, compared with WT (Figures 4 and S2). Enhanced mRNA accumulation is particularly striking when the fact that rRNAs are much less abundant in rnr1-3 is taken into account. Because rRNAs are underrepresented in rnr1-3, we decided to determine the translational status of some mRNAs by assessing their polysomal association. Figure 6 shows that the amount of psbD, psbC, atpI and atpH mRNAs assembled with ribosomes (fractions 7–12) is generally equivalent between WT and rnr1-3, whereas mRNA accumulation of polycistronic versions of these transcripts, is much higher in the non-polysomal (1–6) fractions of rnr1-3. When polysome loading was examined for rbcL, psaB, atpE and petD (Figure S3), atpE displayed a similar phenomenon, whereas differences for rbcL, psaB and petD were more minor: in these cases lower polysome loading could potentially be ascribed to the reduced rRNA concentration already described for rnr1-3. It is intriguing that two of the transcripts with the most pronounced effects, psbD/C and atpE, include overlapping translation units. Taken together, the above observations suggest that the excess mRNA in rnr1-3 is not engaged in translation, in agreement with the previously documented protein deficiencies (Kishine et al., 2004; Bollenbach et al., 2005).

Figure 6.

 Polysomal loading of psbD, psbC, atpI and atpH mRNAs for wild type (WT) and rnr1-3.
The RNA from 15–55% sucrose density gradient fractions was analyzed by gel blot with the probes indicated at the bottom of each panel. Equal proportions of each fraction were loaded. The bottom panels are ethidium bromide-stained gels to reflect loading, and identical samples were used for each blot presented here and in Figure S3. The position of ribosomal (<80S) and polysomal fractions indicated at the bottom were determined by running puromycin-treated samples in parallel to experimental samples.

Depletion of RNR1 leads to a substantial increase in the stability of most chloroplast mRNA species

The excess accumulation of mRNAs in rnr1-3 non-polysomal fractions raised the question of whether increased stability was responsible, which would suggest that RNR1 recycles mRNAs that cannot be translated due to limited ribosome capacity. To measure mRNA stability in vivo, we treated 2-week-old Arabidopsis WT and rnr1-3 seedlings with Actinomycin D (ActD) over a period of 24 h. Actinomycin D is a general inhibitor of RNA polymerases, thus it arrests both nuclear-encoded polymerase or plastid-encoded polymerase-directed transcription within the chloroplast, and has been used previously for this purpose (Shu and Hong-Hui, 2004; Narsai et al., 2007). Figure 7a shows the profiles of atpH mRNAs in ActD-treated WT and rnr1-3 plants. Of the mRNA species designated 1–4, atpH-1, atpH-2 and atpH-4 decay during the WT time course, whereas atpH-3 does not appreciably change in abundance. In rnr1-3, however, atpH-2 and atpH-4 appeared to be more stable. To quantify these results, we re-hybridized the blots with a nuclear 28S rRNA probe as an internal control, since this rRNA has been shown to be stable for up to 24 h (Narsai et al., 2007). The amount of each atpH mRNA species was normalized to the amount of 28S for each time point and plotted on the graphs displayed in Figure 7b, which confirmed the initial conclusions drawn from the RNA gel blots. Using linear regression, the half-life of each transcript was calculated within the 24-h period or extrapolated up to a maximum of 96 h (Table 3). Under our experimental conditions, only atpH-3 had a half-life longer than 24 h in WT, whereas all atpH mRNA species had half-lives longer than 24 h in rnr1-3, with changes most pronounced for atpH-2 and atpH-4. Of note was the higher confidence (R2 values) for the WT compared with rnr1-3, reflecting the difficulty in calculating accurate decay rates for extremely stable species.

Figure 7.

 Accumulation of atpH mRNA in Actinomycin D-treated wild type (WT) and rnr1-3 plants over a 24 h time course.
(a) Results using the atpH probe indicated in Figure 4b. The numbered arrowheads refer to the transcripts that were quantified and used to derive the graphs in panel (b). The bottom panels are ethidium bromide (EtBr)-stained gels to reflect loading of the nuclear 28S rRNA.
(b) Graphs representing linear regression analysis plotted from the gels in panel (a), after normalization to the 28S rRNA signal from the same blot. The regression analysis was used to calculate or extrapolate the mRNA half-lives presented in Table 3.

Table 3.   Half-lives of atpH, psbC, psbD, psbA, rbcL and RBCS transcripts in wild type (WT) and rnr1-3
Band nameWTrnr1-3
Half-life (h) R 2 Half-life (h) R 2
  1. Half-lives were calculated or extrapolated from the graphs of Figures 7, S6 and S7 with a maximum extrapolation at 96 h. R2 represents the scattering of the data around the linear regression and therefore the confidence of the deduced values for half-lives.

psbA 330.98>960.004
RBCS 0.80.93 1.10.88

The analysis of psbC, psbA and rbcL transcript decay kinetics gave equivalent results with a sometimes substantial increase in the stability of mRNA species in rnr1-3 as compared with the WT (Table 3; Figure S4 shows identities of individual mRNA species). Other mRNAs, including the petD intron and the intron-encoded matK transcript, showed the same trend (Figure S5).

As an additional control, we probed for the nuclear RBCS transcript and found similar half-lives in WT and rnr1-3 (Table 3, Figure S6). This result was expected since RNR1 is organellar and should not influence cytosolic mRNA decay rates, but the experiment largely excludes pleiotropic effects on the results we obtained with rnr1-3. Because RNR1 colocalizes to mitochondria, we looked at the influence of RNR1 depletion on the mitochondrial transcripts cox1, atp9 and rps4 (Figure S7). Surprisingly, we observed virtually no decrease in mRNA accumulation following the addition of ActD, implying that the half-lives of mitochondrial mRNAs in Arabidopsis are much greater than those of chloroplast mRNAs. To our knowledge, no mRNA decay kinetics have been published for plant mitochondria, but ActD has been shown to block mitochondrial transcription in animal cells (Ostronoff et al., 1995; Piechota et al., 2006), with tissue-specific stability of mitochondrial mRNAs varying from half-lives of 2 h to no detectable decline over a period of 6 h (Connor et al., 1996). Because we did observe mRNA decay for RBCS and many chloroplast mRNAs, our results for mitochondria would appear valid. The results do not, however, allow evaluation of a potential role for RNR1 in mitochondrial mRNA decay.


Chloroplast 3′→5′ exoribonucleolytic activity is essential for embryo development

Even though the depletion of both PNPase and RNase II or RNase R in E. coli leads to cell non-viability (Donovan and Kushner, 1986; Cheng et al., 1998), embryo lethality of the pnp/rnr1 double mutant was not necessarily expected in Arabidopsis since seed formation is supported by the metabolism of the parent plant. There is precedent, however, for impaired chloroplast gene expression resulting in embryo-defective mutants (e.g. Schmitz-Linneweber et al., 2006; Bryant et al., 2011). In one study, Bryant et al. (2011) identified mutations that resulted in embryo defects, and 24 of these were in genes encoding chloroplast RNA synthesis and modification components, including PPR, ribosomal and RNA-binding proteins. In addition, the absence of RNase J, a chloroplast exo- and endoribonuclease (Sharwood et al., 2011a), is known to cause embryo lethality (Tzafrir et al., 2004). Possible bases for a requirement for plastid gene expression during embryo development are discussed at length in Bryant et al. (2011).

Accumulation of RNA is dependent on 3′→5′ exoribonucleolytic activity

Using the weak cpPNPase mutant alleles P184L and S202N, we were able to reduce cpPNPase activity in the rnr1-3 mutant background. Chloroplast polynucleotide phosphorylase functions predominantly in mRNA maturation, and no major differences had been observed for overall mRNA accumulation in cpPNPase mutants. On the other hand, previous reports on rnr1 had focused on the loss of mature rRNA in favor of precursors. Adding pnp mutations caused modest changes to the accumulation of rRNA precursor (Figure 3), but much more dramatic was the reduced accumulation of mRNA in double mutants (Figures 4 and S2). This phenotype contrasts with both E. coli pnp/rnr1 and pnp/rnb (RNase II) mutants, which accumulate RNA degradation products but not at the expense of mature species (Donovan and Kushner, 1986; Arraiano et al., 1988). On the other hand, the E. coli double mutants were induced by shifting to a non-permissive temperature, as compared with the acclimated state of the Arabidopsis mutants. Escherichia coli mutants lacking PNPase and another 3′→5′ exoribonuclease, RNase PH, are also non-viable (Zhou and Deutscher, 1997). In this case, 23S rRNA is highly unstable, more closely resembling our results for mRNAs in the Arabidopsis mutant. In contrast, even a quadruple 3′→5′ exoribonuclease mutant in Bacillus subtilis remains viable (Oussenko et al., 2005). These differing results show that the importance of specific ribonucleases can vary depending on which enzymes are present and which activity dominates. In the double mutants described here, we speculate that 3′ end maturation by either cpPNPase or RNR1 is required for normal RNA stability, potentially reflecting an RNA quality control mechanism. Incorrect 3′ end maturation is linked to the degradation of nuclear Pol II transcripts (Schmid and Jensen, 2010), and E. coli exerts quality control of tRNA through a degradation mechanism targeting abnormal 3′ ends (Li et al., 2002).

The maturation of mRNA 3′ ends is achieved cooperatively

The precise mapping of atpI, atpH and petD 3′ ends in the rnr1-3 mutant corroborates the atpI/PPR10 assay using recombinant cpPNPase (Prikryl et al., 2011), where cpPNPase alone left 4–7 nt extensions, that we show here can be removed by RNR1 (Figure 5a). HCF107, another chloroplast RNA-binding protein that defines RNA termini, shows similar properties when bound to its RNA substrate in vitro and challenged with commercial prokaryotic PNPase and RNase R (Hammani et al., 2012). An analogous result was observed for petD and atpH, where 3′ ends are probably protected by stem–loops rather than bound proteins (Figure 5b and c). In all cases, the double mutants accumulate progressively longer 3′ ends as cpPNPase activity is depleted in the rnr1-3 background, demonstrating their cooperative action in mRNA maturation. RNR1 is probably able to progress more closely than cpPNPase to the structures or bound proteins defining transcript 3′ ends, because the cpPNPase catalytic site lies within a bulky hexamer (Baginsky et al., 2001) whereas RNase II is a monomer (Shen and Schlessinger, 1982) which in E. coli has been shown to leave no overhang (Marujo et al., 2000).

The importance of RNR1 in chloroplast mRNA homeostasis

A feature of the rnr1 mutant that was not previously documented is the widespread excess accumulation of mRNA precursors with enhanced stability, without a concomitant decrease in mature mRNAs. This implicates RNR1 in maintaining a balance between RNA processing and degradation. In E. coli, RNase II protects mRNAs by removing the 3′ overhang necessary for PNPase to bind, or for PAP1 to bind and add the oligo(A) tail that triggers degradation by either PNPase or RNase II (reviewed by Matos et al., 2011). Moreover, because RNase II preferentially degrades poly(A) tails, it also stabilizes those transcripts by preventing their engagement in the polyadenylated-triggered degradation pathway (Hajnsdorf et al., 1994; Marujo et al., 2000; Mohanty and Kushner, 2000). On the other hand, because of its ability to progress through extensive secondary structures, RNase R is an important participant in mRNA degradation (Cheng and Deutscher, 2005). If chloroplast RNR1 were playing mainly a protective role, its absence should lead to increased RNA decay in the presence of PNPase, which is the opposite of what we observed. Analogy to bacterial RNase R also fails, however, because mature mRNA 3′ ends, rather than unprocessed precursors, should feature strong secondary structures that would be sensitive to RNR1.

Sublocalization of cpPNPase and RNR1 may lead to their specialization

Competition must exist between the chloroplast RNA processing and degradation pathways, since many actors are shared. Hints as to how these opposing pathways may be balanced come from a recent proteomic study of chloroplast nucleoids, which are protein–DNA complexes and the site of DNA and RNA synthesis. It was found that cpPNPase is strongly enriched in the nucleoid, being about 20 times less abundant in the stroma (Majeran et al., 2012). On the other hand, RNR1, which is less abundant overall, was found only in the stroma. Taking into account the ready reversibility of cpPNPase, which responds to the nucleoside diphosphate (NDP) to inorganic phosphate (Pi) ratio, a model emerges that would explain why the major role of cpPNPase is in RNA processing whereas RNR1 plays a key role in RNA decay. A high NDP:Pi ratio favors the polymerase activity of cpPNPase, while a low ratio favors degradation (Yehudai-Resheff et al., 2001). If the local nucleoid NDP:Pi ratio is low, due to the use of (deoxy)nucleoside triphosphates for DNA replication and RNA synthesis, cpPNPase would be mostly degradative in the nucleoid, the activity required for 3′ end trimming. On the other hand, if the stromal NDP:Pi ratio is high, due to the degradation of RNA or from other metabolic pathways, stromal cpPNPase would tend to exert polymerase activity. This would be consistent with a role in RNA degradation, where cpPNPase adds heteropolymeric tails to RNA fragments; such tails are absent from the pnp1-1 mutant (Germain et al., 2011). If these assumptions are correct, RNR1 would be the only 3′→5′ exoribonuclease with degradative activity in the stroma, explaining its strong association with the RNA degradation pathway. The implication of this hypothesis is that long RNA 3′ extensions are removed in the nucleoid, then following export or diffusion of RNA from the nucleoid, stromal RNR1 completes 3′ end maturation.

Even though mRNA stability in chloroplasts has been shown to be dependent on the developmental stage and on a number of diverse RNA-binding proteins (e.g. Klaff and Gruissem, 1991; Nakamura et al., 2001; Pfalz et al., 2009), current models still feature ribosomes to be an important factor in protecting actively translating mRNAs from degradation (Figure 8 in Pfalz et al., 2009). Our results agree with this in the sense that RNR1 appears to have a disproportionate effect on the non-polysomal RNA pool, perhaps to balance the mRNA population with available ribosomes. On the other hand, our results do not support the concept that ribosomes protect RNAs from endonucleolytic cleavages that would otherwise initiate degradation. Indeed, the chloroplast endoribonuclease mutants analyzed to date (Walter et al., 2010; Qi et al., 2012) do not overaccumulate transcripts to the degree seen in rnr1. Finally, nuclear mutants defective in translation of individual mRNAs, do not additionally impact the stability of these mRNAs (e.g. Zerges and Rochaix, 1994; McCormac and Barkan, 1999; Wostrikoff et al., 2001).

Arabidopsis RNR1 behaves as an RNase II enzyme

The main difference between E. coli RNase II and RNase R is their sensitivity to secondary structures. Ribonuclease R is capable of processing through double-stranded RNAs with at least seven nucleotide overhangs, while RNase II creates blunt 3′ ends by rapidly degrading extensions, impeding the binding of other exonucleases, including PNPase, thereby stabilizing RNAs (reviewed by Matos et al., 2011). Our results show that Arabidopsis RNR1 is impeded in vivo by secondary structures, as well as RNA-binding proteins, and shortens the 3′ end extensions left by cpPNPase, similar to the characteristics of RNase II. In vitro analysis had already shown that RNR1 is arrested by 3′ end secondary structures (Bollenbach et al., 2005). Hence, we assert that RNase II is a more appropriate name than RNR1 for this particular gene product.

Experimental Procedures

Plant material and growth conditions

Arabidopsis thaliana Col-0 plants were grown on MS media under long days (16-h light/8-h dark) at 22°C. The pnp1-1, P184L, S202N, A263V and rnr1-3 mutants were previously described (Bollenbach et al., 2005; Germain et al., 2011).

Transcription inhibition with Actinomycin D treatment

Two-week-old plants grown as described above were transferred to six-well tissue culture plates lined with chromatography paper imbibed with 1 ml of liquid MS containing 200 μg ml−1 of Actinomycin D (Sigma-Aldrich, http://www.sigmaaldrich.com/). Treatment was started at the beginning of the afternoon and samples were taken every 4 h for 24 h under constant illumination.

RNA analysis

Total RNA was isolated and analyzed by RNA gel blots, quantitative RT-PCR and circular RT-PCR as described in Germain et al. (2011). For polysome analysis, crude cell lysates from 6-week-old plants grown on MS media were loaded on 15–55% sucrose gradients as described by Barkan (1998). Primers are listed in Table S1 except for those used for rRNA gel blot analysis (Bollenbach et al., 2005), mRNA analysis (Germain et al., 2011) and 5S rRNA qRT-PCR (Sharwood et al., 2011b).


This work was supported by Binational Scientific Foundation (BSF) and Binational Agriculture Research and Development Foundation (BARD) awards 2009253 and IS 4152-08, respectively.