AT2G21280 in Arabidopsis has 12 exons and 11 introns (Figure 1). In order to investigate the function of this gene, we took the advantage of CRISPR/Cas9-mediated gene-editing technique. Four constructs were designed to target four different sites in the gene, respectively and transformed into the wild-type plants. As expected, in the T1 and T2 generation, a part of the transgenic plants were edited at the targeting sites, and sgRNA mutants, such as sgRNA5#7-14, sgRNA6#6-1, sgRNA2#62-5, sgRNA1#13-5 (Figure 1), which have mutations in the 8th, 9th, and 12th exons, were obtained (Figure 2).

Homozygous mutants were verified by DNA sequencing for further analysis (Figure 2). In sgRNA5#7-14, a single base pair was inserted, causing a frame shift from the 184th amino acid and premature termination of the protein soon after. Similarly, in sgRNA6#6-1, a single base pair insertion caused a frame shift from the 206th amino acid and a premature stop codon 46 amino acids downstream. In sgRNA2#62-5, five base pairs were missing, which resulted in a frame shift and premature stop of the protein. In sgRNA1#13-5, a single base pair insertion caused a frame shift and premature stop codon 6 amino acids after. BLAST analysis indicated that the protein sequence of AT2G21280 is well-conserved in plants with 347 amino acids, except the N-terminal region, which is a chloroplast transit peptide. So, the premature stop of the protein in these mutants should have a severe effect on the function of the gene. Even for sgRNA1#13-5, the mutation site is in the last exon and only 63 amino acids upstream of the stop codon, the function of the gene is also very likely to be affected (Figure 2).

At first, we studied the chloroplast division phenotypes of the 3-week-old plants of these sgRNA mutants (Figures 3A,B). The results indicated that the chloroplast sizes of these mutants are similar to that of the wild type. Only in rare cases, slightly enlarged chloroplasts were found in sgRNA5#7-14 (4 out of more than 800 cells) and sgRNA2#62-5 (1 out of more than 800 cells) (Figure 3A). We also analyzed the chloroplast division phenotype of 5-week-old plants, which have larger cell and chloroplast sizes and may give a stronger chloroplast division phenotype. In these mutant plants, chloroplast sizes are also similar to that of the wild type (Figures 3C,D). Furthermore, statistical analysis of the numbers of chloroplasts per cell and cell area indicated that there was no obvious difference between the mutant and the wild type, both in 3- and 5-week-old plants (Figures 3B,D). This result is completely different from the previous reports that GC1 cosuppression lines contained giant chloroplasts (Maple et al., 2004).

To further explore this discrepancy, transfer DNA (T-DNA) insertional mutants (SALK_100683 and SALK_039726) were obtained for analysis. The homozygous SALK_100683 and SALK_039726, which contain T-DNA insertions in the 8th exon and the 9th exon, respectively, were verified by PCR (Figure 1 and Supplementary Figure 1). Mutant plants of SALK_100683 and SALK_039726 at the stages of 3 and 5 weeks were analyzed for the chloroplast division phenotypes (Figure 4). We found that the sizes of most of the chloroplasts in the mutants are similar to that of the wild type, and only in rare cases (5 out of more than 800 cells), larger chloroplasts could be found (Figures 4A,C). Statistical analysis further indicated that the number of chloroplasts per cell of the mutants and the wild type are similar (Figures 4B,D). Thus, the results of T-DNA insertion mutants are similar to that of sgRNA mutants. At the same time, chloroplast division mutants pdv2-3 and arc6-6 were used as controls for comparison (Vitha et al., 2003; Miyagishima et al., 2006; Chang et al., 2017; Wang et al., 2017). These mutants contains only a few giant chloroplasts in the cell (Figures 4A,C).

The transcriptional level of AT2G21280 in different sgRNA and T-DNA mutant lines was analyzed by semi-quantitative reverse transcription (RT) PCR (Supplementary Figure 2). The results showed that the levels of AT2G21280 were reduced in all of these mutant lines. Especially, PCR product was undetectable in T-DNA insertion mutants.

To further analyze the protein levels in these mutant lines, we generated the antibodies of AT2G21280. As shown in Figure 5, a band of approximately 33 kD was detected in the wild type, which is close to the expected size of AT2G21280. While in all of these mutants, this band was missing. Moreover, no band smaller than this size was detected in these mutants. These results indicated that the protein of AT2G21280 was either not translated or degraded. Therefore, these sgRNA and T-DNA mutants are true knockout mutants.

Taken together, these results showed that AT2G21280 only has a minor role in chloroplast division.

To test whether overexpression of AT2G21280 can affect chloroplast division as reported before (Raynaud et al., 2004), we transformed Arabidopsis wild-type plants with constructs containing CaMV35S-driven full-length cDNA (c), or full-length genome DNA (g) individually. Moreover, we also obtained transgenic plants expressing 35S-g-YFP and 35S-g^ΔH-YFP, respectively. The latter had a truncation of the last C-terminal 20 amino acids (for details see below) of the protein. The protein level of AT2G21280 in the transgenic plants with various constructs as mentioned above was analyzed by immuno-blot. Most of the transgenic plants have a protein level much higher than that of the wild type (Figure 6C). We chose the plants with a very high level of AT2G21280 for phenotypic analysis. The chloroplasts in these plants are very similar to those in the wild-type plants (Figure 6A). Then we analyzed chloroplast division phenotypes of these plants by statistical analysis and still found no obvious differences between these overexpression lines and the wild type (Figure 6B).

In addition, in some of the lines, the endogenous protein level of AT2G21280 was undetectable, possibly due to cosuppression. The chloroplast phenotype of these lines was also analyzed (Supplementary Figure 3). Phenotypic and statistical analysis results indicated that there was no significant difference between them and the wild type.

Thus, the previous report that overexpression of AtSulA could cause a chloroplast division defect (Raynaud et al., 2004) cannot be verified by us. Based on these results and the results shown above, we think the previous names of AT2G21280, GC1 and AtSulA, are not appropriate.

Previous study suggested that the nine amino acids at the C-terminal end is an amphipathic helix which may anchor AT2G21280 to the chloroplast inner envelope in tobacco leaf cells with a transiently expression of 35S-AT2G21280-YFP fusion protein (Maple et al., 2004). We studied the subcellular localization of AT2G21280 with GFP fused to the full-length protein 35S-g-YFP (or AT2G21280-YFP), and a protein with a 20 amino acids truncation at the C-terminal end, 35S-g^ΔH-YFP (or AT2G21280^ΔH-YFP) in Arabidopsis. The results showed that, consistent with previous results, AT2G21280-YFP was indeed localized to the envelope of chloroplasts (Figure 7). However, we found that AT2G21280^ΔH-YFP had two types of distribution, one is dot-like aggregation in the stroma as before (Maple et al., 2004), the other is a region near the chloroplast envelope, which is wider than and different from that of AT2G21280-YFP (Figure 7), suggesting when missing the last C-terminal 20 amino acids, AT2G21280 cannot bind well to the envelop membrane. The latter localization result is different from the previous study (Maple et al., 2004).

Previous results suggest that AT2G21280 is similar to slr1223 protein of Synechocystis (SSulA) and All2390 protein of Anabaena sp. PCC7120, which were annotated as SulA homologs in Cyanobacteria (Maple et al., 2004; Raynaud et al., 2004). These proteins are well-conserved in Cyanobacteria and plants. However, a BLAST search with the real SulA protein of E. coli found no homologs in Cyanobacteria and plants. Moreover, BLAST searches with AT2G21280, slr1223, All2390, or their homologs all suggest these proteins are NAD dependent epimerase/dehydratase family enzymes, which is totally different from SulA.

To resolve this problem, we retrieved homologous sequences of SulA in bacteria and AT2G21280 in bacteria and plants based on BLAST search results and carried out a phylogenetic analysis. The result clearly indicated that SulA and AT2G21280 are proteins of distinct families (Figure 8). AT2G21280 belongs to a family widely distributed in bacteria and plants, while SulA belongs to a family in E. coli in its close relatives. Therefore, it is a mistake to name AT2G21280 as AtSulA.

All Arabidopsis plants used in this study are in Col ecotype background. T-DNA insertion mutants of AT2G21280, SALK_039726 and SALK_100683, were obtained from ABRC (Arabidopsis Biological Resource Center, United States). Seeds were sterilized and sowed on 1/2MS (Murashige and Skoog) solid medium containing 0.8% agar and 1% sucrose. After being placed in a refrigerator at 4°C for 2 days, plates were moved to a growth chamber at 22°C with 16-h-light/8-h-dark cycles. Ten days later, seedlings were transferred into soil and grown in the same growth chamber.

Leaf fixation and chloroplast phenotype analysis were performed as described previously (Gao et al., 2013; Chang et al., 2017). Briefly, a piece of 4-week-old leaf was immersed into a tube containing 1 mL 3.5% glutaraldehyde for 1 h in the dark. Then the fixative solution was replaced with 0.1 M Na₂EDTA (pH9.0) and the tube was incubated in water bath at 55°C for 2 h. Chloroplast phenotype was observed with an Olympus CX21 microscope (Olympus, Tokyo, Japan) equipped with a USB 2.0 digital camera (Changheng, Beijing, China). Statistical analysis of chloroplast phenotypes was done as before (Gao et al., 2013). Fluorescence images of YFP and chlorophyll were obtained with a TCS SP8 confocal laser scanning microscope (Leica Microsystems, Germany).

A series of constructs were made based on the CRISPR/CAS9 plasmid, pHEE401E (Wang et al., 2015), in order to edit AT2G21280. Targeting sites were chosen with the tools at the website CRISPRscan¹ and CAS-OFFinder². Primers used for targeting four different sites of AT2G21280 (sgRNA5, sgRNA6, sgRNA2, and sgRNA1) were shown in Supplementary Table 1.

For constructs overexpressing AT2G21280, OE g (genomic DNA) and OE c (cDNA), the full length genomic DNA and cDNA were amplified by PCR with primers 2g21280-5 and 2g21280-6 (Supplementary Table 1), digested with NcoI and MluI, and cloned into 3302Y2 vector, respectively.

To construct 35S-g-YFP and 35S-g^ΔH-YFP, full length genomic DNA (g) of AT2G21280 and AT2G21280 lacking of the last 20 amino acids at the C terminus were amplified using primers GC1-5 and GC1-6 and GC1-5 and GC1-7 (Supplementary Table 1). The PCR products digested with MluI were cloned into pCAMBIA 3302Y3 vector.

Overexpressing constructs were introduced into Agrobacterium tumefaciens and then transformed into Arabidopsis by floral dipping method (Feldmann and Marks, 1987; Clough and Bent, 1998; Bent, 2006). Transgenic plants of T1 generation were screened with Basta. T2 plants were used for the analysis.

Transgenic plants of AT2G21280 sgRNA were obtained as described above. The T1 transgenic plants were screened on 1/2 MS medium containing 20 μg/mL hygromycin and 10 μg/mL carbenicillin. The plates were placed in the dark for 3∼4 days and then moved to the light. One week later, seedlings selected by the antibiotic were transferred into soil and grown in a growth chamber.

For AT2G21280 sgRNA mutants, sgRNA5#7-14, sgRNA6#6-1, and sgRNA2#62-5 were identified by amplifying genomic DNA sequences using primers GC1-8 and GC1-2 and analyzing the sequencing results of the PCR products (Supplementary Table 1). For sgRNA1#13-5, primers GC1-9 and GC1-3 were used instead (Supplementary Table 1).

The sequences flanking the insertion sites of T-DNA mutants, SALK_039726 and SALK_100683, were amplified by PCR with primers GC1-8 and LBC1. The accurate insertion sites were deduced by DNA sequencing. The genomic DNA sequence spanning the insertion sites in the two T-DNA mutants was amplified by primers GC1-8 and GC1-2 (Supplementary Table 1 and Figure 1).

RNA was isolated from the leaves of 4-week-old plants grown in soil under white light, using an RNApure Total RNA Isolation Kit (Aidlab, Beijing, China). The RNA samples (3 μg each) were used as templates for first-strand cDNA synthesis (Thermo Fisher Scientific, United States). Semi-quantitative RT-PCR analysis was performed according to Pan et al. (2013). AT2G21280 were amplified with specific primers GC1-5 and GC1-7. The PP2AA3 gene was taken as a control and amplified with primers PP2AA3-1 and PP2AA3-2.

To generate AT2G21280 antibodies, a fragment of AT2G21280, which is from the 46th amino acid to the end, was expressed in E. coli and purified as antigen. Antibodies were produced in rabbit and purified. For immunoblot analysis, proteins from 5 mg of leaves were separated by SDS–PAGE and transferred to PVDF membrane (Bio-rad). After being blocked with 5% milk for 2 h, the PVDF membrane was incubated with anti-AT2G21280 polyclonal antibodies at a dilution of 1:1000 in 3% milk for 1 h, then washed with 3% milk for four times, and incubated with HRP-labeled goat anti-mouse IgG secondary antibody at a dilution of 1:10,000. Finally, an eECL Western Blot kit (Beijing ComWin Biotech Company, China) were used for the film development.

Homologous sequences of AT2G21280 and SulA in various species were searched with NCBI BLAST³ and downloaded (Supplementary Table 2). Phylogenetic analysis of AT2G21280, SulA and their relatives was carried out by DNAman software (Version 7, Lynnon Biosoft Inc., United States).