RIP-Seq analysis of non-PPR chloroplast editing factors reveals broad RNA interactions and enrichment of less efficiently translated RNAs by OZ1 and ORRM1 complexes

SUMMARY

C-to-U RNA editing in angiosperm chloroplasts requires a large suite of proteins bound together in the editosome. The editosome is comprised of PPR proteins, RIP/MORFs, OZ proteins, and ORRM proteins that physically interact in high molecular weight complexes. The specific functions of non-PPR editing factors in the editosome are unclear, however specific subsets of editing sites are affected by absence of non-PPR editing factors. Unlike the PPR components of editosomes that have predictable nucleotide specificities, domains present in non-PPR editing factors make RNA associations difficult to predict. In this study, chloroplast extracts were isolated from juvenile maize seedlings. RNAs were immunoprecipitated using polyclonal antibodies targeting non-PPR editing factors RIP9, OZ1, and ORRM1. RNA libraries from duplicate experiments were compared. RIP9 was associated with most of the non-ribosomal RNA content of chloroplasts, consistent with a general binding function to PPR L-motifs and tethering of large ribonucleoprotein complexes. The breadth of RNA associations was greater than predicted and include mRNAs without predicted editing sites, tRNA sequences, and introns. OZ1 and ORRM1 were associated with a highly similar pool of RNAs that have a bias towards lower translational efficiency values in mature chloroplasts. Lower translational efficiency was also associated with the pool of edited RNAs compared to RNAs without editing sites. The unexpected breadth of interactions by non-PPR editing factors suggests the editosome is large, diverse, and associated with RNAs with lower relative translational efficiency in mature chloroplasts.

INTRODUCTION

RNA editing mechanisms are common in eukaryotes and include adenosine to inosine (A-to-I) deamination, uridine insertion/deletion, and cytidine to uridine (C-to-U) deamination. In the chloroplasts and mitochondria of flowering plants, C-to-U mRNA editing acts to correct inherited mutations that would otherwise be deleterious. Many organellar proteins are thus encoded accurately only after the DNA sequence has been corrected at the level of the RNA. In Zea mays chloroplasts, editing activity has been associated with large ribonucleoprotein complexes that typically exceed 413 kDa (Sandoval et al., 2019). The composition of the organelle RNA processing body (editosome) is likely diverse (Sun et al., 2016) and requires proteins from the pentatricopeptide repeat (PPR) (Shikanai, 2015), RNA editing factor interacting protein (RIP)/ multiple organellar RNA editing factor (MORF) (Bentolila et al., 2012, Takenaka et al., 2012), organellar zinc-finger protein (OZ) (Sun et al., 2015), organelle RNA recognition motif (ORRM) (Sun et al., 2013), and DEAD-box helicase (Bobik et al., 2017) families to function properly.

PPR proteins are the primary specificity factors present in the editosome that directly bind RNA through associations with hydrophilic residues at fixed positions in each PPR (Barkan et al., 2012). All RNA editing specificity factors come from the PLS subfamily defined by a series of tripartite P (pentatricopeptide), L (long), and S (short) motifs. Each motif includes two antiparallel helices and the typical PPR protein comprised of many serial PPR motifs form a solenoid structure. PPR editing factors end in a C-terminal E, E+, or DYW motif defined in Lurin et al., 2004. The E, E+, and DYW motifs all contain portions of an enzymatic domain and C-terminal truncation of the enzymatic domain past the PG box amino acid sequence disrupts editing function (Hayes et al., 2013). Genetic PPR knock-out plants exhibit complete abolition of editing at a single site or a small subset of editing sites (Shikanai, 2015). PPR proteins often have a complete C-terminal DYW-deaminase domain that acts as the necessary and sufficient enzymatic component of the editosome (Oldenkott et al., 2019, Hayes and Santibanez, 2020). When non-PPR editing factors are knocked out, typically a large subset of editing sites is affected often with more moderate changes in editing instead of complete abolition (Sun et al., 2016). Considering the wide-ranging changes in editing observed in non-PPR editing factor knock-out plants, the nature of selectivity for individual sites as well as the specific functions for ORRM1 and OZ1 remain unclear.

Compared to seed plants, non-vascular plants like Physcomitrella patens do not appear to encode homologs to the non-PPR editing factors despite performing C-to-U RNA editing in organelles (Schallenberg-Rüdinger et al., 2013). Indicative of a simple editosome, exogenously expressed PPR proteins from Physcomitrella patens that contain the DYW-deaminase domain called PPR56 and PPR65 are sufficient for editing co-expressed native targets in Escherichia coli (Oldenkott et al., 2019). Purified PPR65 (Hayes and Santibanez, 2020) and PPR56 (Takenaka et al., 2021) proteins were found capable of editing RNA substrates in vitro under defined conditions. Therefore, a single protein can act as specificity factor due to the N-terminal RNA binding PPR tract and as a C-to-U editase through the C-terminal DYW-deaminase domain. The existence of seemingly more simplified editing complexes makes the necessity of non-PPR editing factors in plants perplexing.

Of the non-PPR editing factors, the most structural and functional information has been provided for the RIP/MORFs. Crystal structures for RIP9/MORF9 and MORF1/RIP8 have been reported (Haag et al., 2017, Yan et al., 2017). Currently, they have been projected to bind the L-motif of PPRs and contribute to target recognition (Yan et al., 2017). RIP/MORFs have also been shown to bind to the E-motif of PLS PPR proteins and might contribute to productive editing complex formation (Bayer-Császár et al., 2017). Genetic knock-out plant studies indicate that subsets of editing sites require specific RIP/MORFs (Bentolila et al., 2012, Takenaka et al., 2012), though the mechanism for selectivity is unresolved. Based on the RIP/MORF crystal structure, a region with similarities to the N-terminal ferredoxin-like domain (NFLS) was identified and projected to potentially directly bind RNA (Haag et al., 2017). However, the role of this domain in RNA binding has not been confirmed biochemically and at least one report published the failure to observe a direct interaction of RIP/MORFs to RNA using electrophoretic mobility shift assay (EMSA) techniques (Huang et al., 2019). Most likely, RIP/MORFs exert their editing function in concert with PPR proteins as supported by the co-crystal structure of a synthetic PPR with RIP/MORFs (Yan et al., 2017). Given that RIP/MORFs generally bind to L-motifs, the mechanism providing RIP/MORF specificity for certain editing sites might relate to protein-protein interactions as opposed to direct RNA binding.

ORRM1 is critical for many editing sites in maize and Arabidopsis chloroplasts (Sun et al., 2013). ORRM1 is part of an organelle specific class of RRM binding protein with 20 members, however, only ORRM1 is predicted to have a RIP/MORF motif. Though its specific role in editing is not clear, the RRM domain itself is sufficient to restore RNA editing activity (Sun et al., 2013). Additional members of this family ORRM2, ORRM3, ORRM4, and ORRM6 have also been shown to be RNA editing factors (Sun et al., 2016, Hackett et al., 2017).

OZ1 was discovered through co-immunoprecipitation with ORRM1 and a physical interaction was further confirmed through yeast-2 hybrid analysis (Sun et al., 2015). Arabidopsis OZ1 knock-out plants have severe phenotypes displaying photosynthetic defects and massive, broad-based reductions in the extent of RNA editing. OZ1 is part of a family of at least 4 proteins with each having between 2–4 RanBP2-type zinc finger domains. Like ORRM1, the domains suggest a role for OZ1 in RNA binding. However, it is not clear what RNAs might be associated with these proteins or why the editosome would require additional RNA binding capabilities.

A group investigating the dynamics of maize chloroplast translation developed an approach to allow for quantitative comparisons between the total mRNA content of chloroplasts versus mRNAs associated with ribosomes (Chotewutmontri and Barkan, 2016). RNA-Seq was performed on a series of leaf sections where leaf emergence correlated with light exposure and chloroplast maturation. At the leaf tip, mature chloroplasts are present in dark green tissues. At the leaf base, proplastids are present and tissues are yellow. By taking sections of the same leaf, differences in chloroplast gene expression can be observed throughout leaf and chloroplast development. Ribosome profiles were generated by pelleting monosomes on a sucrose cushion, followed by next generation sequencing of ribosomal “footprints”. Translational efficiency was quantified by taking the ratio of ribosome footprint RPKM over the RPKM values from a total RNA library. Due to this study, a list of translational efficiency values for each mRNA in chloroplasts is available through 4 stages of chloroplast development.

In this study, the RNA associations for the non-PPR editing factors OZ1, ORRM1, and RIP9 were investigated using RNA immunoprecipitation sequencing (RIP-Seq). Antibodies that recognize OZ1 and ORRM1 enrich a highly similar pool of RNAs, whereas Anti-RIP9 antibodies enriched an overlapping but distinct pool of RNAs. Each non-PPR editing factor interacts with numerous RNAs that vary by RNA-type, RNA function, and presence/absence of editing sites. A bias toward association of RNAs inefficiently translated in mature chloroplasts by both OZ1 and ORRM1 was discovered as well as a bias toward spliced mRNAs by RIP9. Messenger RNAs with editing sites were found to be comparatively less efficiently translated than ones without editing sites with the correlation increasing throughout maize leaf development.

RESULTS

This study aimed to discover RNAs associated with non-PPR editing factors. The RNA component of the editosome was sought since prior associations were only based on sites exhibiting changes in RNA editing efficiency in knock-out plants. Figure 1a shows the procedure used to obtain the samples analyzed. Antibodies for RIP9/MORF9, OZ1, and ORRM1 were generated from a prior study investigating native complex size and composition that discovered RNA editing activity was able to survive immunoprecipitation on Anti-RIP9 bound beads (Sandoval et al., 2019). Western blots demonstrate the high specificity of the RIP9 and OZ1 polyclonal antibodies and the less specific ORRM1 antibodies (Figure 1b). These antibodies were used to immunoprecipitate respective protein targets and associated RNAs. Co-immunoprecipitations were performed under “native” conditions without addition of RNase or a cross-linking step. Contributions may be direct via RNA binding or indirect through protein complexing and tethering.

Figure 1.

RNA libraries were constructed to investigate RNA species specifically associated with RNA editing factors in maize chloroplasts. a) A diagram represents the procedure followed to create the various RNA sequencing libraries used in this manuscript. b) Three immunoblots are shown using polyclonal antibodies raised against Zea mays editing factors RIP9, OZ1, and ORRM1. Arrows highlight the prominent band that corresponds with expected molecular weights based on protein standard. c) Raw read counts for three representative libraries graphed along the chloroplast genome show the relative abundance of different RNAs displayed by the integrative genome viewer (IGV2.8.2).

To characterize the specific RNAs enriched by each non-PPR editing factor, RNA pools were generated and used to create cDNA libraries that were then sequenced using MiSeq SR 150 (Figure 1a). Libraries include total chloroplast RNA (ZmCpRNA.A and ZmCpRNA.B), RNAs enriched on magnetic beads without addition of antibody (Bead.Enriched), co- immunoprecipitation with Anti-OZ1 (Co.IP.AntiOZ1.A and Co.IP.AntiOZ1.B), co-immunoprecipitation with Anti-ORRM1 (Co.IP.Anti.ORRM1.A and Co.IP.Anti.ORRM1.B), co-immunoprecipitation with Anti-RIP9 (Co.IP.Anti.RIP9.A, Co.IP.Anti.RIP9.B), and chloroplast extracts cleared of RNAs bound during the co-immunoprecipitation experiment with Anti-RIP9 (Anti.RIP9.cleared.A, Anti.RIP9.cleared.B). Bioanalyzer RNA profiles were compared for RNAs isolated after immunoprecipitation. Differences in the Bioanalyzer profiles indicate a change in RNA content observed after isolation from Anti-RIP9 beads (Co.IP.Anti.RIP9.A, Co.IP.Anti.RIP9.B) (Figure S1). RNA profiles for other purified RIP-Seq experiments had profiles that were very similar to RNAs directly isolated from chloroplast extracts, suggesting a less robust enrichment with limited degradation.

RIP-Seq libraries were trimmed of poor-quality reads and then aligned to the maize chloroplast genome B73_RefGen_v4 (Jiao et al., 2017). Despite no attempt at clearance of ribosomal RNAs, the coverage from the RIP-Seq libraries was sufficient to examine if sequences were enriched that might indicate RNAs bound by RNA binding proteins. Using integrated genome viewer IGV2.8.2 (Robinson et al., 2011), as expected the highest number of reads mapped to the rRNA genes and psbA for all the libraries (Figure 1c). A clear difference for the OZ1 and ORRM1 profiles was observed: both libraries contain a much higher raw count of 5S rRNA reads (illustrated twice due to the inverted repeat) compared with total chloroplast RNAs or RIP9 libraries after autoscaling to account for differences in total reads. Hits to many genes were elevated in the co-immunoprecipitation samples, though RIP9 libraries exhibited a much greater enrichment of reads across a diverse pool of genes (Figure 1c).

Cladogram clustering by RPKM and fold change values reveal similarity in Anti-OZ1 and Anti-ORRM1 pools, which are distinct from Anti-RIP9.

The RNA hits were calculated using featureCounts (Liao et al., 2014) then processed to RPKM values for each gene feature of the chloroplast genome (Table S1). RPKM values were then visualized through a series of heat maps using the Bioconductor package of R (Gu et al., 2016). The optimal number of clusters was determined to be 5 for a K-mean neighborhood statistical method after comparing different maximum cluster variations in random sets of initial clusters. A heatmap was created to illustrate differences in the RPKM values, and a cladogram represents relationships between libraries (Figure 2a). Libraries associated with OZ1 and ORRM1 were placed into sister groups indicating similarity in the pattern of RPKM values across RNA features. Higher RPKM values were observed for many mRNAs, tRNAs, and introns from all libraries using antibodies that recognize non-PPR editing factors (Figure 2a). Specific gene names can be found in Table S2. The bead control library, which would be equivalent to the preclearing step performed on all samples revealed strong preference for the group IV RNAs including the 5S rRNA gene.

Figure 2.

RIP-Seq libraries were clustered using the K-mean neighborhood statistical method to compare RNA abundance and fold enrichment in various libraries. A heatmap illustrates high similarity between replicate sequence libraries, novel clustering of OZ1 and ORRM1 libraries, and defines 5 groups of RNAs based on their differential enrichment. Higher enrichment is represented in red while blue indicates lower enrichment. A cladogram (along top) compares libraries between immunoprecipitations (RIP9, ORRM1, OZ1), and control (ZmCp, Bead enriched, and RIP9-cleared) samples. Comparing RPMK values (a), Group I is composed of RNAs not abundant in all sample types, while RNAs in groups IV and V are more abundant. Groups II and III consist of RNAs more abundant in co-IP libraries versus control libraries. Gene length and transcriptional unit are shown. A heatmap comparison for fold enrichment values (b), shows significant similarity between OZ1 and ORRM1 groups. The bead enriched sample was used to calculate fold change values. Groups (I-V) show differential enrichment in the co-IP samples. The RNA names for each group can be found in Table S2.

The RPKM values were further adjusted by calculating the fold change of each library compared to the bead control (Table S1). Calculating fold change (experimental library/bead control library) based on the RPKM values reduces biases from the chloroplast extract preclearance step. Since each RNA has a different steady state RNA level, fold change values can better relate how much more represented an RNA may be after the RIP-Seq experiment versus the library. Fold change values are shown as a heatmap (Figure 2b), and gene feature names of each group can be found in Table S2. Anti-RIP9 libraries enriched a broad scion of RNAs, including highly enriched ones that encode photosystem components psaC, psbE, psaJ, psbK, psbF, psbJ, and psbM. Anti-OZ1 and Anti-ORRM1 libraries have very similar levels of enrichment for many RNAs. The clustering analysis of fold change values placed the ORRM1 and OZ1 libraries into sister groups, a computational validation of their similarity (Figure 2b).

Broad enrichment of RNAs by co-immunoprecipitation of editing factors RIP9, OZ1, and ORRM1.

Volcano plots (Figure 3) were generated using the average log₂ of the fold change values for the replicates and the -log₁₀ of the p-value calculated by comparing the co-immunoprecipitation samples and the ZmCpRNA libraries using a student’s t-test. Points above the marked line are significantly different from the chloroplast extract (p < 0.05) and points to the right of the y-axis are enriched. Eleven points along the perimeter of the distribution that are the most significantly different from the chloroplast extract are labeled for each graph. All editing factor co-immunoprecipitation volcano plots (Figure 3a–c) show enrichment of non-rRNAs. Anti-ORRM1 (Figure 3a) and Anti-OZ1 libraries (Figure 3b) have highly similar distributions supporting prior observations in the heatmap analysis (Figure 2). RIP9 (Figure 3c) was associated with more RNAs than OZ1 and ORRM1. Alternatively, the Anti-RIP9.cleared sample (Figure 3d) shows an opposite trend with most points present in the bottom left of the plot.

Figure 3.

Volcano plots for each library display RNAs and RNA features associated RIP, OZ1, and ORRM1 through relative fold enrichment and significance metrics. Each plot includes labels for eleven RNAs that are significantly enriched or not enriched. A line marking -log10(0.05), when p = 0.05, is also shown. p-values are calculated using a t-test comparing the library shown with the chloroplast extract. a) ORRM1; b) OZ1; c) RIP9; d) RIP9 cleared; e) Venn Diagram showing overlap of significantly (p <0.05) enriched (log2(fold change) > 0) RNAs compared to untreated sample.

A Venn diagram illustrates the overlap of significantly enriched RNAs in co-immunoprecipitation samples (Figure 3e). Overall, most of the significantly enriched RNAs overlap between the three co-immunoprecipitation samples. Anti-RIP9 libraries have the most enriched RNA features, at 143 compared to 123 and 129 for Anti-OZ1 and Anti-ORRM1 respectively. OZ1 was not associated with any RNA that did not overlap with another non-PPR editing factor examined, while the single tRNA trnV-GAC was specifically associated with ORRM1. RIP9 was associated with 18 unique RNAs including mRNAs for photosystem genes psaJ, psbA, psbE, psbF, psbJ, psbL, and petG.

Differences in RNA fold enrichment between RIP9 and OZ1/ORRM1 libraries for RNA type and RNA function.

The data was further analyzed by categorizing the RNA sequences by type (Figure 4) and by function of the RNAs themselves or the proteins encoded by mRNAs (Figure 5). Boxplots were generated for each category. In Figure 4a, the Anti-ORRM1 and Anti-OZ1 libraries are similar across all RNA types. Anti-RIP9 libraries have a larger range of fold change values with higher fold change values for tRNA and mRNA compared to the Anti-ORRM1 and Anti-OZ1 derived samples (Figure 4). The RIP9 libraries have a greater median enrichment for mRNAs compared to introns unlike OZ1 and ORRM1 libraries. Fold enrichment values for the three editing factors libraries were lower than 1 for rRNAs with a much lower value for RIP9.

Figure 4.

Differences in fold change values between libraries categorized by RNA type indicate enrichment of tRNAs, mRNAs, catalytic introns, complete introns, and rRNAs by Anti-RIP9, Anti-ORRM1, and Anti-OZ1. a) Boxplots of fold change values from each library for each type of RNA including tRNA (n = 31), mRNA (n = 92), catalytic intron (n = 16), intron (n = 16), and rRNA (n = 4). Boxplots represent the interquartile range and the solid line inside each box shows the median value. b) Enlarged view of the rRNA section of boxplot in panel a. c) qRT-PCR verifies no significant enrichment of the 5S rRNA in all co-immunoprecipitation samples compared to pre-cleared chloroplast extract. Error bars represent 1 standard deviation from the mean for triplicate reactions. Box plots represent the ratio of RPKM values (d) and fold enrichment values (e) for the exon sequences over the intron sequences. (d) Outliers have been excluded from the plot. (d and e) mean values are shown by an X and datapoints are depicted by filled circles.

Figure 5.

Fold change values of RIP-Seq libraries illustrate differences in enrichment of RNAs with different functions. Boxplots elucidate enriched RNAs from co-IP libraries Anti-RIP9 (a), Anti-OZ1 (b), Anti-ORRM1 (c). Some functions have only 1 associated RNA which include (ccsA = cytochrome C, cem = Envelope membrane, clpP = protease, infA = initiation factor 1, lsrR = transcriptional regulation, matK = postranscriptional splicing, rbcL = carboxylase/oxygenase).

Due to the initial observation of more hits to 5S rRNA in Anti-OZ1 and Anti-ORRM1 libraries compared to chloroplast extract libraries (Figure1c), qRT-PCR was used to investigate enrichment by coimmunoprecipitation (Figure 4c). Consistent with fold enrichment value comparisons, higher count threshold values in the 3 co-immunoprecipitation samples indicate absence of enrichment of 5S rRNA due to the antibody.

The extent of intron splicing in the RNA libraries can be estimated by the ratio of exon coding sequence (cds) RPKM values over intron RPKM values for individual genes. The median values of around 3.5 would be consistent with a theoretical ~71% spliced RNA content (Figure 4d, Table S6). The preferential enrichment of spliced RNAs by Anti-RIP9 libraries was observed with a median value of around 7 (~86% spliced). Ratios of fold enrichment for exon over intron indicate specific enrichment of spliced RNAs by RIP9 (Figure 4e, Table S6).

RNA functions were assigned using classifications listed in UniProt. Boxplots were constructed to compare fold enrichment across RNA functions (Figure 5). Photosystem components (PSI and PSII), Cytochrome C (ccsA) and ClpP protease (clpP) are more highly associated with RIP9 compared to ORRM1 and OZ1.

Fold change values for OZ1 and ORRM1 are correlated with each other and mRNAs with lower translational efficiency in mature chloroplasts.

Since RNA profiles were similar between Anti-ORRM1 and Anti-OZ1 libraries, correlations were quantified using a series of X-Y scatterplots and associated Pearson’s coefficients with calculated p-values for linear regressions. Fold change values for Anti-ORRM1 and Anti-OZ1 associated RNAs were highly correlated with Pearson’s coefficient values between 0.873– 0.827 for different library comparisons (Figure 6, Figure S3, Table S3). Fold change values for Anti-RIP9 libraries compared with Anti-ORRM1 and Anti-OZ1 were comparatively less positively correlated with values from 0.415– 0.276. Unlike Pearson’s coefficient that compares the raw values that might be compressed with unequal variance, Spearman’s rho uses a rank order for equal variance between points. Spearman’s rho values for Anti-OZ1 and Anti-ORRM1 libraries are high from 0.860– 0.811 (Table S3).

Figure 6.

Libraries constructed with Anti-ORRM1 and Anti-OZ1 antibodies are very similar and share a bias for RNAs low translational efficiency. XY scatterplots compare fold enrichment values (a) and rank ordered fold enrichment values calculated for each RNA in libraries associated with ORRM1, OZ1, and RIP9. c) A diagram of a maize leaf illustrating the differences in segments 1, 4, 9, and 14 from which the values for translational efficiency were calculated by Chotewutmontri and Barkan, 2016. Segment 14 has an * to highlight the segment used in (d) and (e). XY scatterplots compare translational efficiency and fold enrichment values (d). e) Translational efficiency values and fold enrichment values were rank ordered and compared in XY scatterplots. Black dots represent RNAs with at least one editing site. RNA features used were based on Zea_mays.B73_RefGen_v4.46.chromosome.Pt.gff3 which does not include the entire intron information to prevent overlapping data with annotated catalytic introns.

Translational efficiency values have been quantified for maize chloroplast mRNAs throughout development based on the comparison of next generation sequencing reads versus ribosome associated sequences (Chotewutmontri and Barkan, 2016). Since translating ribosomes might actively disassociate RNA processing complexes and/or unprocessed RNAs might not fully disassociate from RNA processing complexes into the polyribosome pool, it was hypothesized that non-PPR editing factors might be anticorrelated with translational efficiency.

RPKM fold change values for each RIP-Seq library were compared to translational efficiency values for leaf segments 1, 4, 9, and 14. The segments are ordered such that segment 1 (closest to the base) represents the fewest and least developed chloroplasts, with increasingly differentiated and mature chloroplasts across segments 4, 9, and 14 (closest to the tip) (Figure 6c). When fold enrichment was compared to translational efficiency of segment 14 a negative correlation was found in X-Y scatterplots for OZ1 and ORRM1 libraries (Figure 6). Pearson’s coefficients and Spearman’s rho values were calculated for segments 1 and 14 (Table 1) as well as 4 and 9 (Table S4). Translational efficiency values for immature chloroplasts were not correlated with fold change values for OZ1 and ORRM1 libraries unlike the comparisons with segments from mature chloroplasts.

Table 1:

RNAs enriched using Anti-OZ1 and Anti-ORRM1 antibodies correlate with low translational efficiency in mature chloroplasts

RIP Fold Change vs average translational efficiency of Segment 1 non-photosynthetic proplastids (Chotewutmontri and Barkan, 2016)
RIP-Seq Libraries	Pearson’s Coefficient	p-value		Spearman’s Rho	p-value
ZmCpRNA-A	−0.072	0.556		0.032	0.792
ZmCpRNA-B	−0.121	0.323		−0.050	0.683
Co-IP Anti-OZ1-A	−0.114	0.351		−0.189	0.120
Co-IP Anti-OZ1-B	−0.038	0.754		−0.086	0.480
Co-IP Anti-ORRM1-A	−0.174	0.154		−0.225	0.063
Co-IP Anti-ORRM1-B	−0.149	0.223		−0.186	0.126
Co-IP Anti-RIP9-A	−0.109	0.371		−0.085	0.486
Co-IP Anti-RIP9-B	−0.078	0.524		−0.083	0.495
Anti-RIP9 cleared-A	−0.099	0.416		0.070	0.568
Anti-RIP9 cleared-B	−0.064	0.600		0.037	0.760
RIP Fold Change vs average translational efficiency of Segment 14 mature leaf tip (Chotewutmontri and Barkan, 2016)
RIP-Seq Libraries	Pearson’s Coefficient	p-value		Spearman’s Rho	p-value
ZmCpRNA-A	0.091	0.458		0.161	0.188
ZmCpRNA-B	−0.069	0.572		0.076	0.531
Co-IP Anti-OZ1-A	−0.343	3.92E-03	**	−0.485	2.99E-05	***
Co-IP Anti-OZ1-B	−0.313	8.92E-03	**	−0.379	1.42E-03	**
Co-IP Anti-ORRM1-A	−0.360	2.41E-03	**	−0.438	1.97E-04	***
Co-IP Anti-ORRM1-B	−0.369	1.82E-03	**	−0.449	1.28E-04	***
Co-IP Anti-RIP9-A	−0.081	0.509		−0.006	0.960
Co-IP Anti-RIP9-B	−0.047	0.701		0.008	0.950
Anti-RIP9 cleared-A	0.134	0.272		0.260	0.031	*
Anti-RIP9 cleared-B	0.125	0.305		0.236	0.051

^*= p ≤ 0.05;

^**= p ≤ 0.01; and

^***= p ≤ 0.005

Translational efficiency values were then averaged across replicates for each RNA in mature chloroplasts (Segment 14) and compared against fold change values from RIP-Seq libraries in this study. Several three-dimensional scatterplots of two co-immunoprecipitation fold change values were compared to translational efficiency (Figure S5). RNAs with editing sites were found to generally have higher fold enrichment values for Anti-OZ1 and Anti-ORRM1 libraries but have decreased translational efficiency (Figure S5).

Editing factors OZ1 and ORRM1 are more associated with RNAs with editing sites but editing efficiency for all RNAs in all libraries is largely equivalent.

A box plot comparing the fold change of RNAs with and without editing sites (Figure 7a) by library reveals less enrichment of the RNAs without editing sites in the Anti-ORRM1 and Anti-OZ1 samples. However, the Anti-RIP9 sample has approximately the same high average fold change in both categories. Since OZ1, ORRM1, and RIP9 are editing factors, it was hypothesized that they might preferentially bind unedited RNAs or at least RNAs with editing sites. No general bias was observed in terms of average editing extent (Figure 7b). Though, in Anti-RIP9 libraries rpoB transcripts were slightly less edited than in the other libraries (Figure S4, Table S5).

Figure 7.

RNAs with editing sites are highly enriched by editing factor antibodies and are less efficiently translated in mature chloroplasts compared to mRNAs without an editing site. a) Boxplots display fold purification of co-immunoprecipitation samples RIP9, ORRM1, and OZ1 depending on the presence or absence of an editing site in a particular mRNA. b) Median transcript editing extent was similar cross the libraries. c) Translational efficiency values from Chotewutmontri and Barkan, 2016 generally increase throughout chloroplast maturation for RNAs lacking editing sites. Alternatively, translational efficiencies of RNAs without editing sites remain relatively stable throughout development.

Since the pool of RNAs associated with OZ1 and ORRM1 has a bias towards mRNAs with an editing site, the influence on editing site presence on translational efficiency was examined. Translational efficiency was found to increase dramatically during leaf maturation for mRNAs without an editing site unlike the very small change observed in those with an editing site (Figure 7c).

DISCUSSION

Most RNAs could be associated with editing factors RIP9, ORRM1, and OZ1.

Genetic knock-out and knock-down studies of non-PPR proteins revealed that each affects a distinct, overlapping group of RNA editing sites (Sun et al., 2016). The present study aimed to determine the native RNA specificity of non-PPR proteins OZ1, ORRM1, and RIP9. Based on the RIP-Seq results, RIP9 associates with most of the non-ribosomal RNAs present in chloroplasts including mRNAs that encode photosystem components and preferentially spliced mRNAs. OZ1 and ORRM1 were associated with a slightly smaller but homogenous cohort of RNAs anticorrelated with translational efficiency. During the immunoprecipitation procedure there was likely extensive tethering of RNA consistent with a large and complex editosome.

Yeast 2-hybrid studies determined that RIP/MORFs are capable of general binding to PLS-family PPRs and potentially the E domain (Bayer-Császár et al., 2017). Structural analysis has confirmed RIP/MORF association with the L-motif of PPRs which often comprise 1/3 of the PPR content of PLS-class PPR proteins (Yan et al., 2017). Additional support for general binding of RIP/MORF to L-motifs comes from a report of yeast 2-hybrid analysis between higher plant RIP/MORFs and PPR proteins from Physcomitrella patens, an organism with no native RIP/MORFs (Schallenberg-Rüdinger et al., 2013). PPR proteins that are not critical for RNA editing also bind RIP/MORF proteins. For example, the PLS-type splicing factor PDM1 has been shown to associate with RIP9 (Zhang et al., 2015). The interactions between RIP/MORFs and PPR proteins might be even more broad since through a yeast-2-hybrid study, the P-type PPR protein WSL5 has been shown to interact with RIP/MORF protein homologs in Oryza sativa (Liu et al., 2018). Our data indicates that RIP/MORFs are associated with more than just mRNAs with editing sites and is consistent with a model where RIP/MORF proteins generally bind a multitude of PPR proteins that are involved in extensive interactions with many RNAs. Remarkably of the 160 annotated RNA features 143 (89%) were enriched by Anti-RIP9 antibodies.

Compared with RIP9, OZ1 and ORRM1 RNA associations are not quite as broad and are comparatively less intense. This could be due to efficiency of the pulldown or specificity. Given the differences in editing in OZ1 and ORRM1 knock-out mutants (Table S7), it was anticipated there might be differences in RNA association profiles between the two non-PPR editing factors. RIP-Seq analysis indicated that effectively the same suite of RNAs were enriched with antibodies that bound either editing factor. We interpret this to mean that OZ1 and ORRM1 function as part of a heterocomplex and RNA interaction from both contribute to the RIP-Seq libraries. Evidence for heterocomplex formation comes from the initial identification of OZ1 from co-immunoprecipitants with ORRM1 followed by yeast 2-hybrid analysis (Sun et al., 2015), and coincidence in size separated chloroplast fractions (Sandoval et al., 2019).

Specific enrichment for spliced RNAs by RIP9 was observed compared to OZ1/ORRM1 libraries (Figure 4). This observation might indicate RIP9 complexes either preferentially associate with spliced transcripts or the RNA editing competent RIP9 complexes might also retain splicing activity leading to a downstream bias during the RIP procedure.

A correlation between RNAs associated with OZ1 and ORRM1 and inefficiently translated RNAs in mature chloroplasts

RNAs with editing sites are enriched by all three editing factor antibodies despite specificities suggested by single knock-out plants rip9-1, orrm1-1, and oz1-1 (Table S6). RIP-Seq libraries for Anti-ORRM1 and Anti-OZ1 indicate a bias toward RNAs with editing sites (Figure 7) and an anticorrelation with translational efficiency (Figure 6). The correlate with inefficiently translated mRNAs is greatest using translational efficiency values from plant tissues with mature chloroplasts. A correlation was also discovered between reduced translational activation during chloroplast maturation and the presence of an editing site in a mRNA (Figure 7).

During chloroplast development RNAs with editing sites seem less translationally activated in mature tissues. Inefficient translation might allow editing factors greater access to RNA sequences. Selective pressures may eliminate editing sites in mRNAs that are more translationally active during chloroplast development. RNA editing is unlikely to drive chloroplast development (Peeters and Hanson, 2002). ORRM1 and OZ1 also appear to be linked to RNAs with comparatively poor translational activation. Future experiments could investigate if ORRM1 and OZ1 repress translation, influence passage of RNAs from the editosome to polyribosomes, or are merely correlated as a core part of the editosome. We can’t exclude the possibility that ORRM1/OZ1 complexes generally prefer accessible RNAs outside of polyribosomes resulting in their specific capture by RIP-Seq.

The inefficient translation of RNAs requiring more RNA processing steps would be logical. Mechanisms to ensure completion of RNA processing might have evolved to reduce the potential cost of translation of incompletely processed RNAs. As with RNA processing, critical steps leading to translation are often controlled by PPR proteins (Prikryl et al., 2011), but it is unclear if there is a mechanism allowing the ribosome to receive signals of completion from the RNA processing machinery. Incompletely processed RNA transcripts have been observed in mitochondrial polyribosome fractions (Lu and Hanson, 1994, Lu and Hanson, 1996). Unedited chloroplast RNAs can form polyribosomes as ribosome associated transcripts maintain the same ratio of unedited to edited RNAs for petB and ndhA compared to total RNAs (Chotewutmontri and Barkan, 2016). Thus, completion of all RNA processing steps is not obligatory in organelles before the transition to polyribosomes. Protein products from unedited RNAs have been identified (Lu and Hanson, 1996, Phreaner et al., 1996), but have been shown to fail to accumulate in mitochondrial complexes (Grohmann et al., 1994, Lu and Hanson, 1994, Lu and Hanson, 1996). Also, protein products from incompletely edited transcripts could not be observed with sensitive antibodies in maize mitochondria (Williams et al., 1998). Translation products from unedited transcripts would often be affected by radical amino acid differences leading to non-functional proteins that might have short half-lives. Future work could investigate the progression of RNA from the editosome into polyribosomes and any specific role for editosome components in affecting translational efficiency.

EXPERIMENTAL PROCEDURES

Plant material, growth conditions and chloroplast isolation

Chloroplasts were isolated identically to recent publications that investigated intact native RNA editing complexes (Sandoval et al., 2019). Zea mays var. Sugar Buns (Johnny’s Selected Seeds, Fair- field, ME, USA) seeds were planted and grown for 14–21 days in a greenhouse. One-hundred grams of plant leaf tissue was immediately diced into 1-inch pieces in an ice-cold solution containing 100 mM HEPES-NaOH pH 8.0, 4 mM EDTA, and 10.94% mannitol then placed in a blender at 4°C. Homogenization involved short blending pulses to avoid overheating the material. Lysate was filtered and centrifugation at 4,000 × g resulted in a crude chloroplast pellet. The pellet was resuspended and overlaid onto a continuous Percoll gradient created through centrifugation at 43,000xg for 30 min in a buffer containing 50% Percoll (v/v), 50 mM HEPES-NaOH pH 8, 2 mM EDTA, 0.3 M mannitol and 2 mM dithiothreitol (DTT). The denser intact chloroplast band was collected compared to a less dense lysed chloroplast band. The intact chloroplasts were then washed in a solution containing 50 mM HEPES-NaOH pH 8, 2 mM EDTA, 0.3 M mannitol and 2 mM DTT. Chloroplasts were lyzed using detergent in a hypertonic solution by suspension of the chloroplast pellet roughly 1:2 w/v in a solution containing 30 mM HEPES-KOH pH 7.7, 10 mM magnesium acetate, 2 M KCl and 0.002% Triton X-100 (v/v) for 30 min. Chloroplast extracts were centrifuged at 13,000xg for 10 min, and then dialyzed twice using SnakeSkin Dialysis Tubing 7000 MWCO (Thermo Scientific, Waltham, MA, USA) against 500 ml of Dialysis buffer [30 mM HEPES- KOH pH 7.7, 3 mM magnesium acetate, 45 mM potassium acetate, 30 mM ammonium acetate, 10% glycerol (w/v), and 2 mM DTT]. Chloroplast extracts were finally aliquoted, flash frozen and stored at −80°C.

Co-immunoprecipitation

Affinity-purified polyclonal rabbit antibodies were produced for a previous study (Sandoval et al., 2019) using respective amino acid sequences RKPRQQAPAQTQTESASS, SPQYRRNLPIVRSETDEDAS and DRLSGGSNQAFRPHYQAR from RIP9, OZ1 and ORRM1. Pierce protein A/G magnetic beads (Thermo Scientific) were gently vortexed for >30 sec in their vial. Afterwards, 12.25 μL of bead solution was transferred to a clean 1.5 mL microcentrifuge tube, placed on a magnetic stand, decanted, and washed with dialysis buffer lacking DTT. Then 54 μL of chloroplast extract was diluted 1:1 to reduce the DTT concentration to 1 mM with a non DTT containing dialysis buffer to prevent unacceptable reduction of the antibodies. The diluted chloroplast extract was precleared by adding magnetic beads and gently re-suspended by pipette. This sample was then incubated with rotation for 30 min at 4°C. Next, the tube was placed on the magnetic stand, and the pre-cleared chloroplast extract was transferred to a clean 1.5 ml microcentrifuge and placed on ice, while the magnetic beads were washed and saved for further analysis (pre-cleared beads for the no antibody control bead library).

Affinity-purified polyclonal primary antibody (Ag) was diluted 1:50 in 100 μL dialysis buffer, added to a new 1.5 mL microcentrifuge tube with 12.25 μL cleaned magnetic beads, and gently re-suspended by pipette. This antibody-beads sample was then incubated with rotation for 10 min at room temperature. Next, the tube was placed on the magnetic stand, washed gently, decanted, and placed on ice. Lastly, pre-cleared chloroplast supernatant was placed in the chilled antibody-beads tube, mixed gently by pipetting, and incubated for 20 min at 4°C. Afterwards, the tube was placed on the magnetic stand, the cleared supernatant (SP) was saved for further analysis, and the pellet (Ab-Ag complex beads or BD) was washed three times using 100 μL of dialysis buffer without DTT [30 mM HEPES- KOH pH 7.7, 3 mM magnesium acetate, 45 mM potassium acetate, 30 mM ammonium acetate, and 10% glycerol (w/v)].

RNA extraction and isolation

Following co-immunoprecipitation, bead samples were gently resuspended in 1 mL RiboZol^™ RNA Extraction Reagent (AMRESCO LLC., Solon, OH) and incubated at RT° for 5 min. Next, 200 μL of chloroform (HPLC grade) was added and mixed by inverting then flicking the tube. After centrifugation for 10 min at 4°C, the aqueous layer was removed and transferred to a silica column following manufacturer’s instructions from the Zymoclean Gel RNA Recovery Kit (Zymo Research, Tustin, CA). RNA sample quantity and quality were assessed using 260/280 ration from Nanodrop-Lite and Qubit instruments. RNA samples were aliquoted and stored at −80°C.

Mi-Sequencing

RNAs were sent to the UCI Genomics High-Throughput Facility without ribosome reduction. Total RNA was monitored for quality using the Agilent Bioanalyzer Nano RNA chip and Nanodrop absorbance ratios for 260/280 nm and 260/230 nm. Library construction was performed according to the Illumina TruSeq^® Stranded mRNA Sample Preparation Guide with modification. The input quantity for total RNA was 50 ng and chemically fragmented for three minutes. First strand synthesis used random primers and reverse transcriptase was used to make cDNA. After second strand synthesis the ds cDNA was cleaned using AMPure XP beads and the cDNA was end repaired and then the 3’ ends were adenylated. Illumina barcoded adapters were ligated on the ends and the adapter ligated fragments were enriched by nine cycles of PCR. The resulting libraries were validated by qPCR and sized by Agilent Bioanalyzer DNA high sensitivity chip. The concentrations for the libraries were normalized and then multiplexed together. The multiplexed libraries were sequenced on the Miseq using version 3 single end 150 cycles chemistry. The version of Miseq control software was MCS v: 2.6.2.1 with real time analysis software, RTA 1.18.54.0.

Bioinformatic analysis

Reads were delivered as .fastq files and bioinformatic analysis was performed at CSULA. Bad quality reads and adapter sequences were trimmed from the MiSeq libraries using Trimmomatic (Bolger et al., 2014). Next the trimmed datasets were aligned to the maize reference genome B73_RefGen_v4 (Jiao et al., 2017) using bowtie2, tophat2 and cufflink (Langmead and Salzberg, 2012, Trapnell et al., 2012). Resulting files were converted into SAM and BAM format using SAMtools (Li et al., 2009). BAM files were further indexed, counted, and sorted using igvtools and visualized on the Integrative Genome Viewer (IGV2.8.2) (Busan and Weeks, 2017). In order to calculate RPKM values, reads were quantified for gene features listed in Zea_mays.B73_RefGen_v4.46.chromosome.Pt.gff3 from Ensembl Genomes (Howe et al., 2020) using FeatureCounts (Liao et al., 2014). Noncoding RNAs such as catalytic introns were annotated from Zea_mays.B73_RefGen_v4 in the Zea_mays.B73_RefGen_v4.ncrna.fa file. Positional information for intron sequences was gathered from NCBI Reference Sequence: NC_001666.2 and cross-referenced to the Zea_mays.B73_RefGen_v4.46.chromosome.Pt.gff3 file. A modified gff containing the positional information was used to calculate reads using sorted .bam files and FeatureCounts.

Reads Per Kilobase per Million (RPKM) were then calculated using the equation [feature reads / (gene length (kb)/ total mapped reads/1,000,000)]. Heatmaps and boxplots were generated using Phyton and R programming. Specifically the Complex Heatmap tool (Gu et al., 2016) from the Bioconductor package (Huber et al., 2015) of R was used to generate heatmaps. The Pearson correlation analysis and the X-Y scatterplot involved using Jupyter Notebook with Python version 3.8.2 embedded, along with Numpy, Pandas, Matplotlib, and Seaborn package to write the codes and generate the plots.

RStudio ggplot2 (Villanueva and Chen, 2019) was used to create boxplots of fold change values for various categories. Excel was used to create volcano plots by graphing the average log2(fold change) by the −10 log(p-value). The p-values were obtained using the student’s t-test assuming homoscedasticity on the two replicates for each sample, compared to the pre-cleared chloroplast extract.

Quantitative Polymerase Chain Reaction (qPCR)

Forward and reverse primers (Forward Primer 5’ - CGC CTA GGA CAC CAG AAT A – 3’/Reverse Primer 5’ - ATT CTG GCA TCG AGC TAT TTT - 3’ ) for the 5S rRNA were added for a total of 0.5 μM each in the final reaction mix. 2x GoTaq qPCR Master Mix (Promega) was added with no modifications. Two microliters of cDNA template was used per reaction, with HPLC water being added to bring the total up to 20 μL. Quantitative PCR program was as follows: 2 minutes at 95°C, then 40 cycles of 15 seconds at 95°C, 30 seconds at 50°C, and 1 minute at 60°C. All qPCR experiments were run on an Eppendorf ep gradient Realplex 2 Mastercycler. Count thresholds were obtained using the CalQPlex method from the Eppendorf ep gradient Realplex 2 Mastercycler. Values from samples co-immunoprecipitating with RIP9, OZ1, and ORRM1 were compared to the pre-cleared chloroplast extract sample using two-tailed Student’s t-tests (assuming homoscedasticity) in Excel.

Supplementary Material

Tab S3

Table S3. Pearson’s coefficient and Spearman’s Rho analysis between library fold change values.

Tab S2

Table S2. Labels of RNAs from heatmaps for comparisons.

Tab S4

Table S4. RNAs enriched using Anti-OZ1 and Anti-ORRM1 antibodies correlate with low translational efficiency.

Tab S5

Table S5. Sequence hits that correspond to edited (T) and unedited (C) mRNA sequences used to calculate percent RNA editing for each of the maize RNA editing sites.

Tab S1

Table S1. Hits and RPKM values for all RIP-Seq libraries used in this study.

Tab S6

Table S6. Raw data for exon and intron comparisons of RPKM and fold enrichment in respective libraries.

Tab S7

Table S7. Comparison of RNA editing levels in editing factor knock-out plants with RIP fold change values.

Supinfo

Figure S1. Bioanalyzer results are shown in (a) from RNA libraries total chloroplast RNA (ZmCpRNA.A), duplicate co-immunoprecipitations experiments with antibody Anti-RIP9 (Co.IP.Anti.RIP9.A/B), cleared total RNA pools left after duplicate immunoprecipitations using Anti-RIP9 (Anti.RIP9.cleared.A/B) and (b) total chloroplast RNA library ZmCpRNA.B, the bead enriched sample (Bead.Enriched), duplicate co-immunoprecipitations experiments with antibody Anti-OZ1 (Co.IP.Anti.OZ1.A/B), and duplicate co-immunoprecipitations experiments with antibody Anti-ORRM1 (Co.IP.Anti.ORRM1.A/B). Results indicate a general enrichment of non rRNA species. Ribosomal RNA bands were estimated based on sizes from report describing the 23S hidden break point (Nishimura et al., 2010). A solid black lane indicates nonsequential lanes from a single experiment.

Figure S2. Log RPKM values are represented on a heat map to illustrate differences in RIP Seq libraries, and the baseline steady-state RNA levels present for each transcriptional unit in chloroplast extracts.

Figure S3. X-Y scatterplots of Pearson’s correlation analysis between fold change values in RIP-Seq libraries compared to the bead control library.

Figure S4. Percent editing of transcripts is largely unchanged across RIP-seq libraries demonstrating little bias for edited versus unedited RNAs through co-immunoprecipitation with non-PPR editing factors except for less edited rpoB transcripts observed for Anti-RIP9 libraries. A heatmap displays percent editing for each of the RIP-Seq libraries with a cladogram relating similarities in editing extent across libraries.

Figure S5. Three dimensional scatterplots comparing two co-immunoprecipitation samples with translational efficiency. Translational efficiency values are from segment 14 (Chotewutmontri and Barkan, 2016). Co-IP libraries Anti-OZ1 and Anti-ORRM1 are compared with translation efficiency using fold change values (a) and a rank-order of fold change with translational efficiency values (b). Rank-order values were used to investigate monotonicity. Lesser correlated libraries Anti-OZ1 and Anti-RIP9 are shown that also relate translational efficiency and fold change values (c) as well as a rank-order comparison (d). Axes are referenced such that x is Anti-OZ1, y corresponds to either Anti-ORRM1 or Anti-RIP9, and z is translational efficiency.

SIGNIFICANCE.

RNA editing is a critical process in angiosperm chloroplasts that requires complexes composed of PPR and non-PPR proteins. RIP-Seq analysis indicates broad RNA associations with specific correlates for photosystem components with RIP9 and RNAs with lower translational efficiency values for OZ1 and ORRM1.