The Arabidopsis thaliana T-DNA insertion mutant pool (CS76508) was purchased from the Arabidopsis Biological Resource Center (ABRC, Ohio State University, Columbus, OH, United States). All A. thaliana plants used in this study were in Col-0 wild-type (WT) background. After 3 days stratification at 4°C in the dark, seeds were germinated on Murashige-Skoog (MS) plates containing 2% sucrose at 22°C under a light intensity of 120 μmol photons m^-2 s^-1 with a 16 h/8 h light/dark photoperiod. Two-week-old seedlings were moved to grow in soil (a mixture of peat moss, vermiculite and perlite at 1:1:1) under the same conditions (Wang et al., 2016).

Mutants were screened on MS plates containing 25 mg/L kanamycin. Seedlings with albino, yellowish, or pale green cotyledons were moved to grow in soil, and their seeds were individually collected.

Genomic DNA was extracted from rosette leaves using the CTAB method (Green and Sambrook, 2012). For identifying T-DNA insertion position of each mutant, genome walking was performed using a Genome Walking Kit (TaKaRa, Shiga, Japan) following the manufacturer’s manual. Homozygous progenies of the mutants were screened according to the SIGnAL iSect tool¹. All primers used in this study are listed in Supplementary Table S1.

RNA was isolated using RNAiso Plus Reagent (TaKaRa) according to manufacturer’s instruction. Total RNA (1 μg) was reverse transcribed using a PrimeScript 1st Strand cDNA synthesis Kit (TaKaRa). Transcript abundance of each gene studied was determined by quantitative real-time PCR (qPCR) in a Thermal Cycler Dice Real Time System TP800 (TaKaRa) using SYBR Premix ExTaq II (TaKaRa), following the manufacturer’s manuals, and calculated using the comparative C_T method (Schmittgen and Livak, 2008). ACT 2 (At3g18780) was used as a reference. At least three biological replicates, each with three repeats, were analyzed for each sample.

For genetic complementation, a 4332 bp genomic DNA fragment of DY1 (At5g19540), ranging from -2044 bp upstream of the translation initiation codon (ATG) to 290 bp downstream of the stop codon, was amplified using primers DY1-GF and DY1-GR, and cloned into the BamH I site of pCAMBIA1300 (CAMBIA, Canberra, ACT, Australia) to generate the construct DY1:DY1. For subcellular localization study, full-length open reading frame (ORF) of DY1 or At4g20360 was amplified from the 1st strand cDNA pool using primers DY1-HF and DY1-ER for DY1, or RABE1b-HF and RABE1b-ER for At4g20360, and subsequently cloned into the BamH I site of pA7-eYFP to generate the construct 35S:DY1-eYFP or 35S:At4g20360-eYFP. To generate the construct for expressing DY1 with dual fluorescent proteins, the coding sequences for the chloroplast transit peptide (cTP) of DY1 (cTP^{DY 1}) (using primer pair DY1-cTP-HF and DY1-cTP-ER), eYFP (eYFP-HF and eYFP-ER), DY1 without cTP (DY1^ΔcTP) (DY1-ΔcTP-HF and DY1-ΔcTP-ER) and mCherry (mCherry-HF and mCherry-ER), were separately amplified and joined together into the BamH I and Sac I sites of pA7-eYFP using In-Fusion technique (TaKaRa). The entire fragment encoding the cTP^{DY 1}-eYFP-DY1^ΔcTP-mCherry fusion protein was further amplified (DY1-mature-HF and DY1-mature-ER) and then subcloned into the Nco I site of pCAMBIA1300-RTL2 using In-Fusion technique for transient expression in tobacco leaves and genetic transformation of Arabidopsis. A. thaliana was transformed by the floral dip method (Clough and Bent, 1998).

For transient transformation, protoplasts were isolated from rosette leaves according to Yoo et al. (2007). About 20 μg of purified plasmid DNA was applied for each transformation. The protoplasts were incubated in the dark at 22°C overnight before microscopy observation.

For all PCR amplifications, high-fidelity PrimeSTAR DNA polymerase (TaKaRa) was used according to the manufacturer’s instruction.

Subcellular localization of DY1 were predicted using online programs ChloroP and TargetP (Emanuelsson et al., 1999; Emanuelsson et al., 2007). Deduced amino acid sequence of DY1 was used as a query to search its homologs in different plant species of which full genomes have been sequenced in GenBank using the BlastP algorithm. Sequences were aligned using the ClustalX program and a maximum-likelihood phylogenetic tree was constructed using MEGA 6 with a bootstrap replication value of 1,000 (Chenna et al., 2003; Tamura et al., 2013).

To perform Northern blot, total RNA was extracted from leaves of 3-week-old dy1 mutant and WT plants. Ten microgram of total RNA was loaded for each lane and separated on 1.5% agarose gel. After separation, the RNA was blotted onto a positively charged nylon membrane (GE Healthcare, Pittsburgh, PA, United States). The probe for detecting DY1 transcripts was amplified using primer pair DY1-Probe-F and DY1-Probe-R, with DY1 cDNA as a template, and then labeled using DIG DNA Labeling Mix (Roche, Basel, Switzerland). Probing, washing, and detection of DY1 were performed according to the DIG High Prime DNA Labeling and Detection Starter Kit II user’s manual (Roche).

Total protein was extracted from rosette leaves using the trichloroacetic acid (TCA) precipitation method. Approximately 1 g of rosette leaves was homogenized in liquid nitrogen. Ten milliliter of cold 10% TCA in acetone was then added and mixed well by vortex and sonication. The mixture was incubated at -20°C for at least 4 h and then centrifuged at 4°C, 15,000 g for 20 min. The pellet was washed several times with cold acetone, air dried briefly, and solubilized in 9 M urea (in 100 mM phosphate buffered saline, 1 mM DTT) with sonication. The precipitation was repeated once, and the pelleted protein was finally dissolved in 9 M urea solution for direct use or storage.

Protein samples were mixed with equal amounts of 2 × SDS loading buffer (Green and Sambrook, 2012), heated at 65°C for 10 min, separated on 12 % SDS–polyacrylamide gel and subsequently blotted onto nitrocellulose membrane (GE Healthcare) for immunodetection. A peptide (WSIRNPTDLETSSY) was synthesized based on deduced amino acid sequence of DY1 and used as an antigen to immune rabbits by GenScript (Nanjing, China). Antiserum was purified from blots according to Harlow and Lane (1988). Antibody against Actin 11 was purchased from Agrisera (Vännäs, Sweden). Horseradish peroxidase (HRP)-conjugated secondary antibody against rabbit IgG was from Promega (Madison, WI, United States). Common protocols (Green and Sambrook, 2012) and the manufacturers’ manuals for electrophoresis, semi-dry blotting and Western detection using the ECL Western Blotting Substrate (Promega, Madison, WI, United States) were followed.

A FluoView FV1000 (Olympus, Tokyo, Japan) laser scanning confocal microscopy system was used for fluorescence observation. The eYFP fluorescent was excited with 488 nm laser and the emitted light was recorded from 500 to 530 nm. The mCherry fluorescent was excited with 543 nm laser, recorded from 580 to 620 nm. 543 nm laser excitation and 680 to 720 nm recording range were used for chlorophyll auto-fluorescence observation.

For transmission electron microscopy (TEM) analysis, leaves from 3-week-old seedlings were fixed, embedded and sectioned according to Faso et al. (2009). A Hitachi-7700 transmission electron microscope (Hitachi, Tokyo, Japan) was used for observation and image capturing.

About 10 g of rosette leaves were homogenized briefly in cold chloroplast isolation buffer (0.3 M sorbitol, 5 mM MgCl₂, 5 mM EGTA, 5 mM EDTA, 20 mM HEPES, 10 mM NaHCO₃, pH 8.0) and then filtered through two layers of miracloth (EMD Millipore, Billerica, MA, United States). Crude chloroplasts were collected by centrifugation for 5 min at 4°C, 1,000 g. After the separation of crude chloroplasts on a 2-layer Percoll gradient (40 and 80%), intact chloroplasts were obtained from the interface between two layers. Purified chloroplasts were further lysed in TE solution (10 mM Tris–HCl, 1 mM EDTA, pH 7.5) for 30 min on ice, followed by a centrifugation for 5 min at 3,000 g to collect thylakoid membranes.

For BN-PAGE, intact chloroplasts were solubilized in Solubilization Buffer (50 mM Bis–Tris, pH 7.0, 0.5 M aminocaproic acid, 10% glycerol, 1% n-dodecyl β-D-maltoside and 1 mM PMSF) on ice for 10 min before separation on a 4% ∼ 14% gradient gel. After the first dimension electrophoresis, the gel was either directly blotted onto a PVDF membrane (Millipore) or further separated by SDS–PAGE for a second dimension (Zhou et al., 2017).

Chlorophylls and carotenoids were isolated from rosette leaves according to Pogson et al. (1996). One hundred milligram of leaves was ground in liquid nitrogen into fine powder and mixed thoroughly with 250 μl 80% acetone containing 10 μg deuteroporphyrin IX dimethyl ester (Sigma–Aldrich, St. Louis, MO, United States) as an internal standard. The extract was then mixed with 250 μl ethyl acetate, followed by 200 μl of water. After a centrifugation at 4°C, 15,000 g for 10 min, the upper organic phase was collected and dried under a stream of nitrogen. The pigment was re-dissolved in Solvent A (acetonitrile:water:trimethylamine = 9:1:0.01) for high-performance liquid chromatography (HPLC) analysis. All steps were carried out under dim light.

Total pigment samples were analyzed by HPLC (Waters 2695 separation module) with a Waters ODS2 C18 analytical column (5 μm, 4.6 mm × 250 mm) and 2998 photodiode array detector (Waters, Milford, MA, United States) according to Wang et al. (2016).

Chlorophyll fluorescence of 3-week-old dy1 mutant and WT plants were measured by a MINI-PAM fluorometer (Walz, Effeltrich, Germany) according to Wang et al. (2016). For light-response curves of PSII quantum yield (ΦPSII) and non-photochemical quenching (NPQ), dark-adapted plants were illuminated at a series of photosynthetically active photon flux densities (PPFD) (3, 19, 43, 145, 300, 387, 611, 1,006, and 1,312 μmol photons m^-2 s^-1). The minimal yield of fluorescence (F₀) under 650 nm was measured at 0.8 μmol photons m^-2 s^-1. To estimate the maximum fluorescence yield (F_m), a saturating pulse (0.8 s, 5,000 μmol photons m^-2 s^-1) was applied.

The yeast two-hybrid screening was performed largely according to the Matchmaker^TM GAL4 Two-Hybrid System 3 & Libraries User Manual (TaKaRa). Full length ORF of DY1 was cloned into pDEST32 (Invitrogen, Carlsbad, CA, United States) and then transformed into the yeast (Saccharomyces cerevisiae) strain AH109. A. thaliana cDNA library constructed in pDEST22 (Invitrogen) was maintained in the yeast strain Y187. After mating, diploid yeast cells were screened on SD/-Leu/-Trp/-His triple dropout medium. From positive colonies, cDNA insertions encoding the prey proteins were amplified by PCR and sequenced. Candidate genes were further tested for pair-wise interaction using a pGAD-T7/pGBK-T7 system (TaKaRa). Briefly, the coding sequences of DY1 and candidate genes were cloned into pGAD-T7 and pGBK-T7, respectively. Plasmids were co-transformed into yeast strain AH109 and selected on SD/-Leu/-Trp double drop out plates. Colonies were serial diluted and further tested on SD/-Leu/-Trp/-His/-Ade quadruple dropout plates for detecting protein–protein interactions.

Statistical significance was tested using GraphPad Prism6 (GraphPad Software). Data are shown as the means ± SD of at least three replications.

We screened 10,000 individuals of the T-DNA insertion mutant library. At least 20 lines showed pale green, yellowish or completely albino cotyledons. The T-DNA insertion positions of these lines were identified by genome walking. One of these lines that showed dwarfed growth and yellowish cotyledons (Figure 1A) was proved to harbor a T-DNA insertion in front of the third exon of At5g19540 (Figure 1C). We named this uncharacterized gene Dwarf and Yellow 1 (DY1). Northern blot (Figure 1B) and genomic complementation (Figure 1A) experiments confirmed that At5g19540 is responsible for the corresponding phenotypes. DY1 encodes a 53-kDa protein with a predicted N-terminal transit peptide (cTP), which was predicted to target DY1 to chloroplasts (Supplementary Figure S1). The dy1 mutant seedlings were defective in de-etiolation, i.e., their cotyledons stayed yellow after illumination. Vegetative growth of the dy1 seedlings was severely retarded. Juvenile true leaves of the dy1 mutant also had defects in their pigmentation, whereas mature leaves of the mutant were greener (Figure 1A).

We did not find functional characterization reports nor annotation for DY1, and therefore, we searched GenBank for homologs of DY1. In each plant species for which a full genome has been sequenced, only one copy of a DY1 homolog gene was found. We did not find any close homologs of DY1 in cyanobacteria, green algae or diatoms, although there are some sequences in certain organisms with weak similarities to DY1 (Figure 2).

Comparisons of the deduced amino acid sequences of DY1 homologs identified highly conserved regions spanning the entire sequence beyond the homologs’ cTPs (Figure 2A). Overall sequence identity among DY1 homologs is higher than 40%, and it is over 60% if we exclude the most variable cTP regions. On a phylogenetic tree, DY1 homologs formed two major clades, one of which was composed of members from angiosperms and the other of which consisted of homologs from the remaining land plants, including gymnosperms and bryophytes. The angiosperm clade can be further divided into 2 sub-clades, one each for dicots and monocots (Figure 2B).

Leaves of the dy1 mutant showed a distinctly lighter color compared to those of the WT plants. Therefore, we postulated that the dy1 mutant might accumulate lower amounts of pigments. In our quantification of the pigments in 3-week-old seedlings, both chlorophyll a and b contents in the dy1 mutant were significantly lower than in the WT plants. For carotenoids, dy1 leaves also accumulated less neoxanthin, violaxanthin, lutein and β-carotene compared with the WT plants. However, under normal growth conditions, two xanthophylls, antheraxanthin and zeaxanthin, were found to accumulate in dy1 leaves, although at trace amounts (Figure 3). The quantities of all pigment contents are listed in Table 1.

We then quantified transcript abundances of genes for chlorophyll and carotenoid biosynthesis. In 3-week-old dy1 leaves, the expression of most of the genes for chlorophyll biosynthesis, e.g., CRD1, CHLH, PORB, GUN4 and CAO, were down-regulated (Figure 4A). Interestingly, genes for chlorophyll turnover, including PAO and RCCR, were also down-regulated in dy1 (Figure 4B). Moreover, the expression of most of the genes for chlorophyll a/b-binding proteins (CABs) was also down-regulated in dy1 (Figure 4C). Most of the genes for carotenoid biosynthesis and catabolism were down-regulated in dy1 as well (Figure 4D). In brief, no genes for pigment metabolism were up-regulated in dy1 in our determination.

Because dy1 had lower amounts of leaf pigments, we postulated that it might also have lower photosynthetic capacity. This was confirmed by measurements of chlorophyll fluorescence parameters. In our study, F_v/F_m, which reflects the efficiency of PSII photochemistry, was reduced from 0.83 in the WT plants to 0.73 in dy1 (Table 2). We also measured light-responsive curves to assess ΦPSII and NPQ of both dy1 and WT plants under a series of light intensities. WT plants had higher ΦPSII than did dy1 plants under all of the light intensities that we measured (Figure 5A). However, dy1 had a higher NPQ than did the WT plants when light intensity was lower than 300 μmol photons m^-2 s^-1 (P < 0.05 at 43 and 145 μmol photons m^-2 s^-1) (Figure 5B).

We dissected true leaves of 3-week-old dy1 mutant and WT plants for TEM observations. The dy1 mutant generally showed normal chloroplast shape and thylakoid formation. However, in the dy1 mutant, thylakoid membranes were loosely packed into grana (with an average thickness of each grana thylakoid layer at 21.3 ± 1.9 nm, comparing with the WT level of 14.3 ± 1.4 nm) and possessed more plastoglobules (Figure 6).

DY1 encodes a protein with 462 amino acids. Both online programs ChloroP and TargetP predicted chloroplast localization. To confirm chloroplast localization, full-length DY1 protein with an enhanced yellow fluorescent protein (eYFP) fused to its C-terminus was transiently expressed in Arabidopsis protoplasts. The fluorescence of DY1-eYFP strictly overlapped with chlorophyll auto-fluorescence, indicating chloroplast localization (Figure 7A). From the amino acid sequence of DY1, we identified a twin-arginine (R⁸⁴R⁸⁵) motif (Supplementary Figure S1), which indicates that DY1 is probably imported into chloroplasts through the ΔpH-dependent pathway (Robinson and Bolhuis, 2001). It was reported that, for proteins imported through this pathway, their C-terminal fragments prior to the twin-arginine motif are to be cleaved after importing (Williams et al., 2000; Robinson and Bolhuis, 2001). Therefore, we isolated intact chloroplasts from WT leaves, sub-fractioned them into stroma and thylakoid fractions, and performed SDS–PAGE and Western blot analyses. From our immunoblots probed with a specific antibody raised against DY1, a band at 42 kDa was found in the total-chloroplast protein sample and the stromal fraction (Figure 7B). This finding agreed with the calculated size of DY1 after the cleavage (DY1^ΔcTP) and proved the stroma localization of DY1.

To our surprise, an additional smaller band of 24 kDa was frequently detected in the stroma fractions of the WT plants (Figure 7B). We found that the protein level of DY1^ΔcTP was more stable throughout developmental stages, while the content of the smaller form tended to accumulate over time (Figure 7C). Thus, we postulated that after being imported into chloroplasts, DY1 might be further processed during the maturation of leaves. To test this visually, we fused eYFP immediately after the cTP and mCherry at the end of the C-terminus of DY1 (Figure 8A), and we expressed this fusion protein with dual fluorescent tags in tobacco and Arabidopsis. In tobacco, we performed transient expression in juvenile leaves that were close to the apical shoot and in mature leaves from the bottom of the plant. After 3 days, signals of both eYFP and mCherry could be clearly observed within chloroplasts in juvenile leaves. However, in mature leaves, most observed chloroplasts had similar strengths of the eYFP signal but their mCherry signals were much weaker (Figure 8C). In Arabidopsis plants that were genetically transformed to express this fusion protein, it was clear that only eYFP signal could be detected in mature leaves (Figure 8D). When we probed the protein samples of 1- and 2-week-old transgenic Arabidopsis seedlings, antibody against eYFP could detect the DY1 fusion protein in both samples, whereas that against mCherry showed only a faint band on the blot, suggesting a removal of the C-terminus from DY1 (Figure 8B). These results demonstrated that DY1 matures with leaf development.

To determine the distributions of the two forms of DY1 in chloroplasts, we performed BN-PAGE to separate different protein complexes in chloroplasts (Figure 9A), followed by SDS–PAGE as a second dimension (Figure 9B). After electrophoresis, gels of each dimension were blotted and then probed with an antibody against DY1. Our Western blot clearly showed that both the 42- and 24-kDa forms of DY1 could be found at the position of about 300-kDa, which is around the PSII core + Cyt b₆f complex (Timperio et al., 2007; Chen et al., 2010), and the 42-kDa form could also be probed at the position of approximately 100 kDa (Figure 9B).