Maize defective kernel5 is a bacterial TamB homologue required for chloroplast envelope biogenesis

Plastids are essential organelles for plants. Higher plants differentiate specialized plastids distinguished by structure, pigmentation, and function, such as photosynthetic chloroplasts in leaves and starch accumulating amyloplasts in the cereal endosperm (Jarvis and López-Juez, 2013). Plastids originated through endosymbiosis ∼1.5 billion years ago, when cyanobacteria were acquired by eukaryotic cells (Yoon et al., 2004). Extant cyanobacteria are Gram negative, with inner and outer plasma membranes. Chloroplasts also have a double-membrane structure, with inner and outer envelopes likely corresponding to bacterial membranes (Gould et al., 2008; Gross and Bhattacharya, 2009). The vast majority of chloroplast proteins are nuclear encoded, synthesized on cytosolic ribosomes, and imported into plastids post-translationally (Jarvis, 2008). These precursors are imported through the protein translocons of the outer and inner chloroplast envelope membranes, termed TOC and TIC, respectively (Keegstra and Cline, 1999; Cline and Dabney-Smith, 2008).

The plastid has a major role in primary metabolism (Bowsher and Tobin, 2001). Transport of solutes and metabolites across the envelope is important to integrate chloroplast metabolism with the cytosol and other cellular organelles. Chloroplast envelopes exchange ions, carbohydrates, nucleotides, and amino acids to support metabolic pathways in which the chloroplast has unique enzymatic activities (Block et al., 2007; Facchinelli and Weber, 2011).

The inner envelope has multiple solute translocators and is considered the primary metabolite permeability barrier (Flügge, 1999; Fischer, 2011). Inner envelope translocators are integral membrane proteins with two pathways for insertion. During protein import, some inner envelope membrane (IEM) proteins are transferred to the membrane through a stop-transfer mechanism. Other IEM proteins complete import into the stroma and are inserted similar to posttranslational translocation of secreted bacterial proteins (Li and Schnell, 2006; Tripp et al., 2007; Viana et al., 2010).

The outer envelope is thought to be permeable to solutes of <10 kD, which is similar to outer membranes of Gram-negative bacteria (Flügge and Benz, 1984). Porins facilitate this nonspecific diffusion of small solutes in Gram-negative bacteria (Nikaido, 1994). Many chloroplast outer envelope proteins (OEPs) have a β-barrel structure similar to porins and were hypothesized to facilitate nonspecific diffusion; however, biochemical analyses show more selective transport. Pea OEP21 transports P_i, triose phosphates, and 3-phosphoglycerates (Hemmler et al., 2006). OEP24 allows diffusion of triose phosphates, dicarboxylic acids, charged amino acids, ATP, and P_i (Pohlmeyer et al., 1998). OEP40 is permeable to glucose, glucose-1-phosphate, and glucose-6-phosphate (Harsman et al., 2016). OEP16 and OEP37 are selective for amino acids and peptides and even have tissue specific expression patterns (Pohlmeyer et al., 1997; Goetze et al., 2006; Pudelski et al., 2012). Thus, OEP channels studied so far show specificity for distinct metabolites, challenging the notion that the outer envelope is a nonspecific molecular sieve.

Relatively little is known about the biogenesis pathways of β-barrel OEPs (Huang et al., 2011). In Gram-negative bacteria, most β-barrel outer membrane proteins require the β-barrel assembly machinery (β-BAM) for correct folding (Hagan et al., 2011; Selkrig et al., 2014). The translocation and assembly module (TAM) is also important for bacterial outer membrane biogenesis. TAM is composed of TamA, localized to the outer membrane, and TamB, localized to the inner membrane (Selkrig et al., 2012). Tam mutations in different bacterial species can alter membrane morphology or block secretion of toxins (Selkrig et al., 2012; Shen et al., 2014; Iqbal et al., 2016). Phylogenetic analysis showed that TamA is restricted to Proteobacteria, whereas TamB is widely distributed across Gram-negative bacteria (Heinz et al., 2015). In Borrelia burgdorferi, TamB interacts with a BAM subunit, suggesting that TamB participates in assembly of β-barrel proteins through multiple outer membrane protein assembly machineries (Iqbal et al., 2016).

Here, we show that the maize defective kernel5 (dek5) mutant has plastid division defects that disrupt endosperm starch accumulation. Mutant dek5 seedling leaves have fewer and larger chloroplasts with defects in chloroplast membranes. Molecular identification of the dek5 locus demonstrated that it encodes a predicted TamB homologue. Contrary to a prior report for the rice DEK5 orthologue (Matsushima et al., 2014), the maize DEK5 protein is localized to the chloroplast envelope with analogous topology to TamB and the Arabidopsis thaliana DEK5 orthologue (Chen et al., 2018). The dek5 mutant alters envelope ultrastructure, reduces OEP accumulation, alters inner envelope protein levels, and blocks P_i uptake. These data suggest that Dek5 has a role in plastid envelope biogenesis and illustrate the importance of selective solute transport across the plastid envelope.

The Maize Genetics Cooperative Stock Center maintains six mutant alleles of the recessive dek5 locus: dek5-N874A, dek5-N961, and dek5-N1339A isolated from ethyl methanesulfonate mutagenesis; dek5-PS25 and dek5-MS33 isolated from a Robertson’s Mutator transposon-tagging population, and the spontaneous dek5-Briggs allele isolated at a commercial breeding company (Neuffer and Sheridan, 1980; Scanlon et al., 1994; Sachs, 2009). At maturity, each allele conditions shrunken or collapsed endosperm, while embryo development is frequently normal (Fig. 1 A). Among the alleles, dek5-Briggs and dek5-N1339A have more severe phenotypes. The visual phenotype of mature kernels resembles that of brittle1 (bt1), brittle2 (bt2), and shrunken2 (sh2) mutants blocked in starch biosynthesis (Fig. 1 B). Unless noted, dek5-MS33 was used for phenotype analysis.

Near-infrared reflectance spectroscopy analyses of mature kernels indicated dek5 mutants have reduced starch and increased oil content compared with normal siblings (Fig. 1 C). The dek5 starch deficiency is apparent in developing endosperm at 12 d after pollination (DAP) and extends to maturity (Fig. 1 D). Hand sections of developing kernels had uneven starch staining, even though mutants enlarged to a similar extent as normal siblings (Fig. 1 E). Like bt1, bt2, and sh2, the dek5 endosperm collapses as mature kernels dry down. Relative zein content, particularly α-zein, is also reduced in dek5 mutants (Fig. 1 F). Both bt1 and bt2 have reduced zein content associated with reduced starch (Lee and Tsai, 1984).

Soluble sugar levels are consistent with reduced accumulation of starch. At 20 and 24 DAP, dek5 endosperm has elevated levels of sucrose, glucose, and maltose as well as higher levels of ADP-glucose at 24 DAP (Fig. 1 G). Sucrose, glucose, and ADP-glucose are soluble precursors of starch. On a fresh weight basis, sucrose and glucose are elevated similarly in dek5, sh2, and bt1 with dek5 sucrose and glucose being 4.2-fold and 3.4-fold higher than normal siblings, respectively (Tobias et al., 1992). The higher levels of ADP-glucose in dek5 are similar to sh2; bt1 double mutants (Shannon et al., 1996).

Endosperm starch granules have a normal diameter at 12 DAP but increase significantly in dek5 at 24 DAP (Fig. 2, A and B). Mature endosperm granule size is 20% larger in dek5 (Fig. 2 C). Reduced total starch concomitant with increased granule size implies dek5 endosperm accumulates fewer granules than normal.

The average chain length distribution of the amylopectin component of endosperm starch was determined after enzymatically converting the branched polymer to a population of linear α(1→4)-linked glucan chains. The frequency of each length from 6 to 44 glucosyl units is essentially identical in dek5 and normal siblings (Fig. 2 D). Despite enlarged granules, there is no apparent change in assembly of semicrystalline amylopectin.

Depending upon the allele, 5–40% of dek5 kernels germinate and seedlings have nonclonal, white variegation. Mutant seedlings are stunted at 7 d after sowing (DAS) and usually die in 2–3 wk (Fig. 3 A). Mo17 and W22 genetic backgrounds enhance dek5, resulting in a nearly empty pericarp kernel, lower germination, and more severe seedling phenotypes. The dek5-Briggs allele was the most severely modified (Fig. S1 A). By contrast, introgression of dek5 into B73 improved grain-fill, germination, and mitigated seedling lethality (Fig. S1 B). The mild dek5-MS33 allele could complete a life cycle to produce all mutant kernels (Fig. S1 C).

The pale green, variegated dek5 leaves are due to multiple chloroplast abnormalities. Photosynthetic pigments and subunits of the electron transport chain are reduced (Fig. 3, B and C). Leaf protoplasts showed dek5 mutants have larger and fewer chloroplasts compared with normal siblings; a subset of mutant cells have one chloroplast or no chloroplasts (Fig. 3 D). Z-stack imaging of chlorophyll florescence showed dek5 protoplasts have 7.5-fold fewer chloroplasts with a 6.6-fold increase in chloroplast volume (Fig. 3, E and F; and Video 1).

Mutant protoplasts isolated in standard buffers had a low recovery rate, while doubling mannitol to 0.8 M increased recovery nearly twofold (Fig. 3 G; Student’s t test, P = 0.02). The dek5 cells isolated in high osmotic buffer frequently contained multiple vacuolar compartments, which is a morphological characteristic of autophagy or programed cell death (Fig. 3 H; van Doorn et al., 2011). These data suggest that dek5 cells are more likely to lyse with standard buffers and have altered osmotic pressure.

Transmission EM (TEM) revealed multiple ultrastructural defects in dek5 chloroplasts. Chloroplasts in expanding maize seedling leaves develop in a gradient. At the base, differentiating plastids have few internal membranes, while chloroplasts at the middle and tip of the leaf have developed thylakoids (Pogson et al., 2015). Sections from the base, middle, and distal tip of dek5 seedling leaves had much larger mesophyll chloroplasts with similar thylakoid development to normal siblings (Fig. 4, A–F). Normal chloroplasts have two thylakoid centers defining the long axis of the organelle (Fig. 4, B and C). In dek5, there are multiple thylakoid centers and a more spherical organelle shape (Fig. 4, E and F).

C4 plants, like maize, localize starch biosynthetic enzymes and starch granules in bundle sheath chloroplasts (Majeran et al., 2005; Leegood, 2008; Weise et al., 2011). Large starch granules were observed in normal bundle sheath chloroplasts at the tip of the leaf (Fig. 4 I). Mutant bundle sheath chloroplasts were much larger; however, these contained many vesicular structures with a few, small starch granules (Fig. 4 L). Some dek5 mesophyll chloroplasts also had starch granules, which was not found in normal siblings (Fig. 4 F). These observations suggest defects in regulating starch synthesis or transporting fixed carbon to bundle sheath cells.

Consistent with protoplast imaging, dek5 mesophyll chloroplasts have fourfold larger cross-sectional area than normal (Fig. 5 A). Although thylakoid grana appear normal, dek5 grana are 40% larger in diameter, suggesting a higher proportion of the chloroplast develops as thylakoid (Fig. 5 B). Stromal area is threefold larger in dek5 chloroplasts (Fig. 5 C). Comparing total and stromal area, the fraction of stroma in dek5 is reduced by 20% relative to normal (Fig. 5 D).

The ratio of envelope to other chloroplast compartments is also altered in dek5. The cross-sectional area of the chloroplast could be modeled as an ellipse from long and short axis measurements (Fig. 5, E–H), suggesting that volume and surface area could be modeled as an ellipsoid. Estimates from TEM indicate a 14.7-fold increase in chloroplast volume and an 8.6-fold increase in stromal volume, but surface area only increases by 4.7-fold (Fig. 5, I–K). The surface area to volume ratios for dek5 chloroplasts or stroma is reduced by threefold or twofold, respectively (Fig. 5, L and M). The increase in thylakoid with concomitant reductions in surface area suggests that dek5 has a relative reduction in envelope membrane.

The more spherical shape and increased volume of dek5 chloroplasts is not commonly reported in chloroplast division mutants. TEM analyses of division mutants show filamentous or buckled chloroplasts with a relatively normal short axis (Itoh et al., 2001; Ii and Webber, 2005; Karamoko et al., 2011; Kamau et al., 2015). Filamentous chloroplasts with a constant short axis can be roughly modeled as a cylinder, and surface area increases linearly with increased volume instead of the nonlinear relationship for ellipsoid-like dek5 chloroplasts (Fig. 5 N). By contrast, the A. thaliana tic40 mutant causes a more spherical chloroplast with a similar long axis to short axis ratio as dek5 (Kovacheva et al., 2005). However, only the long axis of tic40 chloroplasts is reduced, and tic40 has predicted volume and surface area in the range of WT chloroplasts (Fig. 5 N). For ellipsoid-shaped organelles, the long/short axis ratio has relatively little impact on the surface area to volume ratio at a given volume (Fig. 5 O). Even though dek5 and tic40 have similar shapes, the low volume of tic40 chloroplasts results in a typical surface area to volume ratio (Fig. 5 O). A chloroplast volume increase without forming filamentous or buckled organelles indicates a relative decrease in envelope.

TEM also revealed a range of membrane defects. Mutant chloroplasts had unusual envelope ingrowths as well as an expanded peripheral reticulum (Fig. 6, A–C; Szczepanik and Sowiński, 2014). Internal thylakoid membranes appeared disorganized in some chloroplasts (Fig. 6, D and E; and Fig. S2). In some cases, internal vesicles and completely disorganized plastids were observed (Fig. S2, E and F). The more severe phenotypes were observed immediately adjacent to cells with intact chloroplasts indicating that the defective membranes are unlikely to be an artifact from fixation or sample processing. Mutant chloroplasts with a loss of internal chlorophyll autofluorescence were also observed by laser confocal scanning microscopy in 16 of 51 dek5 protoplasts that had full Z-stack imaging (Fig. 6, F–H; and Video 1). The data indicate that dek5 has pleiotropic plastid membrane biogenesis defects.

The dek5 locus was mapped to the short arm of chromosome 3 using B-A translocations (Neuffer and Sheridan, 1980). We fine-mapped the dek5-PS25 allele to a genomic interval of 460 kbp containing 13 gene models in the B73 RefGen_v3 genome assembly (Fig. 7 A). Mu-Seq analysis of Mutator transposon insertions in dek5-PS25 and dek5-MS33 identified an insertion in GRMZM2G083374 (Fig. 7 B). There is a Mu8-like transposon in dek5-PS25 and a Mu1 insertion in dek5-MS33 at the identical position in exon 1, indicating that these are independent alleles. The dek5-Briggs allele has a frameshift mutation with a single base insertion at exon 10 that results in a downstream premature termination codon. Two additional Mu insertion alleles from the UniformMu genetic resource failed to complement dek5-PS25 (Fig. 7, C and D). Molecular identification of five noncomplementing alleles shows GRMZM2G083374 is the dek5 locus.

The dek5 genomic locus is predicted to be a single-copy gene that is >30 kbp. We amplified and sequenced the B73 cDNA to validate the predicted gene model. The sequenced Dek5 mRNA has a different structure than GRMZM2G083374. It consists of 23 exons with a genomic transcriptional unit of 30,833 bp, which is identical to the current B73_v4 transcript model, Zm00001d039612_T001 (Fig. 7 B). Dek5 encodes a predicted protein of 232.9 kD that has an N-terminal chloroplast transit peptide (TP), a transmembrane domain (TMD), and a C-terminal TamB domain, formerly DUF490 (Fig. 7 E). In rice, a missense mutation in the Dek5 orthologue, Substandard Starch Grain4 (SSG4), results in enlarged starch granules and chloroplasts similar to dek5 (Matsushima et al., 2014). Loss of function of the A. thaliana dek5 orthologue, emb2410, causes an embryonic lethal phenotype, while weak alleles (tic236) have been reported to reduce chloroplast protein import (Chen et al., 2018). The DEK5 domain structure is similar to the Escherichia coli TamB (Selkrig et al., 2012). Phylogenetic analysis indicates that DEK5 has additional conserved sequence domains within land plants (Matsushima et al., 2014; Figs. 8 and S3).

Chloroplast targeting of DEK5 was tested by fusing the 5′ 1,374 bp of the Dek5 ORF with EGFP (Fig. 9, A and B; and Fig. S3). Transient expression of DEK5N-GFP in Nicotiana benthamiana leaves resulted in a punctate ring of GFP signal surrounding red chlorophyll fluorescence from thylakoid membranes (Fig. 9 B). This pattern is identical to known chloroplast envelope membrane proteins, such as CHUP1, OEP7, Tic40, and Tic20 (Oikawa et al., 2008; Breuers et al., 2012; Machettira et al., 2012). However, it contrasts with stromal localization reported for a differing length N-terminal fragment of SSG4 fused to GFP (Fig. S3; Matsushima et al., 2014).

Deletion of the predicted TMD sequence from DEK5N-GFP resulted in DEK5NΔTMD-GPF showing punctate distribution of GFP fluorescence overlapping with chlorophyll fluorescence, indicating targeting to plastids (Fig. 9 C). Fusing only the DEK5 TP and TMD was sufficient for envelope localization of GFP signal (Fig. 9 D). These results support the conclusion that the DEK5 TMD is necessary and sufficient for envelope localization.

Polyclonal antibodies raised against the DEK5 TamB domain cross-react with an ∼100-kD polypeptide that copurifies with WT chloroplasts as well as a larger band in total leaf protein (Figs. 9 E and S4 A). Preincubation of the α-DEK5 sera with the TamB recombinant antigen causes loss of the 100-kD signal, showing that this band is specific for the DEK5 protein despite migrating much faster than the full-length protein (Fig. S4 B). The A. thaliana TIC236 orthologue is easily degraded, suggesting the maize signal is a partially degraded protein (Chen et al., 2018). The specific α-DEK5 signal is reduced in dek5 leaf extracts, and the amount of DEK5 protein correlates with the strength of the mutant phenotype (Figs. 9 E and S4 C). The severe dek5-Briggs and dek5-N1339A alleles have little to no detectable DEK5 protein as well as severely reduced outer envelope (Toc75 and OEP80) and thylakoid (cytochrome f [Cyt f]) membrane proteins (Fig. S4 C). These data indicate that α-DEK5 reports a DEK5 protein that is most likely a C-terminal fragment without a TMD.

Chloroplasts fractions of soluble, thylakoid membrane, IEM, and outer envelope membrane (OEM) showed that DEK5 cofractionated with envelopes and soluble proteins (Fig. 9 F). Compartment-specific protein markers have clean separation of soluble proteins from envelope membranes. The DEK5 membrane association can be disrupted by washes with 0.1 M Na₂CO₃ or 1 M NaCl, but not with 6 M urea (Fig. 9 G). The thylakoid membrane-associated oxygen evolving complex 23-kD (OEC23) subunit has similar sensitivity, while the integral membrane Cyt f protein is resistant to all salt and chaotropic agents. These data are consistent with a C-terminal degradation product of DEK5 lacking the TMD.

A dual-protease protection assay was used to determine the topology of DEK5 (Froehlich, 2011). Thermolysin only digests proteins on the outer face of the OEM. DEK5, 6PGDH, and OEP80 were all resistant to thermolysin when chloroplasts are intact (Fig. 9 H). Each of these proteins show some sensitivity to thermolysin when chloroplasts are lysed under hypotonic conditions. By contrast, trypsin digests protein domains in the intermembrane space between the OEM and IEM. The OEM protein OEP80 is sensitive to trypsin, while stromal 6PGDH is resistant (Fig. 9 I). DEK5 was partially digested by trypsin in five biological replicates (Figs. 9 I and S4 D). All three proteins are fully digested in hypotonic chloroplast lysates. These results suggest that DEK5 protein is sensitive to trypsin and likely to face the intermembrane space between the OEM and IEM. This is the topology of TamB in E. coli and TIC236 in A. thaliana but is contrasting to the stromal localization reported for rice SSG4 (Selkrig et al., 2012; Matsushima et al., 2014; Chen et al., 2018).

DEK5 envelope localization and intermembrane space topology support an orthologous function to TamB in facilitating membrane protein insertion. Proteomics of purified chloroplast envelope fractions showed broad changes in envelope protein composition (Fig. 10 A). This analysis identified 68 known chloroplast envelope protein homologues. We compared envelope protein abundance by normalizing MaxQuant intensity values for the envelope proteins detected (Table S1). Nearly 33% of these proteins were reduced or absent in dek5. Among 22 proteins with reduced abundance, 91% are predicted to function in membrane transport.

Half of the OEM integral membrane proteins (5/10) were down-regulated in dek5, with the remaining OEM proteins being unchanged in relative abundance (Fig. 10 A). OEM proteins with reduced levels included the chloroplast protein targeting machinery, Toc34, Toc75, and Toc159, as well as OEP16 and OEP80. These results are consistent with immunoblots of dek5 leaf protein showing that Toc75 and OEP80 are reduced (Fig. S4 C).

Seven membrane proteins had increased relative abundance in the IEM. However, 14 IEM proteins were reduced in dek5, including a Na⁺-dependent P_i transporter and two P_i/phosphoenolpyruvate translocators (Table S1). Both α-helical and β-barrel transmembrane proteins are reduced in dek5, and we conclude that there are pleiotropic defects in envelope protein composition.

Reduced P_i transporter levels in dek5 envelopes suggest that P_i transport may be affected. P_i uptake was assayed by incubating purified chloroplasts with ³²P_i (Fig. 10 B). Normal chloroplasts reach equilibrium ³²P_i levels between 3 and 5 min in 0.1 mM P_i. Mutant dek5-N961 chloroplasts show no appreciable uptake over the 5-min incubation at 0.1 mM P_i. Higher P_i concentrations of 0.5 mM (5×) or 10 mM (100×) did not increase uptake to normal levels. A longer incubation time at 0.5 mM P_i showed no appreciable uptake.

P_i accumulates in the stroma, and uptake is limited by total stromal volume (Fliege et al., 1978). A dilution series of normal chloroplasts found that the detection limit of ³²P_i uptake is ∼20–50% of the stromal volume in the standard assay (Fig. 10 C). Although dek5 volume is larger than in normal chloroplasts, thylakoids account for ∼10% more of total chloroplast volume, leading to a relative reduction in stroma (Fig. 5 M). To normalize P_i uptake on stromal volume, we counted chloroplast number and estimated the average stroma volume. A dilution series from 1.7- to 5.8-fold larger stromal volume relative to the standard assay with normal chloroplasts showed no appreciable P_i uptake (Fig. 10 C). Based on the normal and dek5 chloroplast dilution series, we expected to observe uptake if dek5 membrane had as little as 17% of normal P_i uptake.

To control for chloroplast integrity throughout the assay, we completed P_i uptake assays without radioactive labeling and purified the chloroplasts in an isosmotic sorbitol solution. Phase-contrast microscopy of the recovered chloroplasts showed that dek5 chloroplasts were intact (Fig. 10, D and E). Immunoblots of purified chloroplasts loaded on an equal chlorophyll content showed equivalent levels of thylakoid membrane proteins CF1-β and Cyt f, as expected for this normalization (Fig. 10 F). DEK5 and OEP80 proteins were reduced in the mutant chloroplasts relative to thylakoid membrane proteins. This result is consistent with TEM and proteomics observations, from which dek5 and normal chloroplasts are predicted to have different amounts of envelope membrane relative to chlorophyll content or thylakoid membrane proteins. Together, the data argue that multiple envelope membrane proteins and P_i transport are greatly reduced in dek5 chloroplasts.

Cloning of dek5 revealed that the locus is orthologous to rice ssg4 and A. thaliana tic236 (Matsushima et al., 2014; Chen et al., 2018). However, we have come to different conclusions about the biochemical function of plant TamB proteins. Matsushima et al. (2014) suggested that SSG4 has a novel function in chloroplast division. Chen et al. (2018) suggested that TIC236 evolved a novel function in protein import. We propose that the DEK5/SSG4/TIC236 proteins have analogous envelope biogenesis roles to bacterial TamB proteins.

A series of alleles with differing severity is critical to develop an accurate the model for plant TamB homologues. The rice ssg4 locus is defined by a single mutant allele with a missense mutation (Matsushima et al., 2014). Compared with dek5, ssg4 shows mild starch granule and chloroplast phenotypes with no ultrastructural changes observed in plastid membranes. The severe alleles of A. thaliana emb2410/tic236 arrest seed development at the globular embryo stage (Tzafrir et al., 2004). Weak tic236 alleles cause leaf morphology defects, but chloroplast ultrastructure was not characterized (Chen et al., 2018). Maize alleles expressing some DEK5 protein give more normal development, while strong alleles lacking DEK5 protein cause severe seed defects and embryo lethality. Much of our phenotypic analysis focused on the mild dek5-MS33 and dek5-N961 alleles, which enable sufficient plastid development to observe a range of organelle defects. From these phenotype data, we conclude that dek5 is required for envelope biogenesis and plastid metabolite transport.

Multiple lines of evidence suggest dek5 disrupts intracellular metabolite transport. Sucrose is the carbon source for endosperm starch synthesis. Endosperm starch is decreased in dek5 with concomitant increase of sucrose, suggesting that carbon is not limiting. Sucrose is converted to ADP-glucose in the cytoplasm of endosperm cells (Hannah, 1997). The BT1 inner envelope transporter selectively transports ADP-glucose into the amyloplast (Shannon et al., 1998). Like bt1, dek5 mutants have higher levels of ADP-glucose, suggesting that reduced transport into amyloplasts may be the cause of increased sugars (Tobias et al., 1992; Shannon et al., 1996). This inference is further supported by nearly identical starch-branch chain lengths in dek5 and normal endosperm starch, indicating normal polymerization and debranching within the amyloplast.

There is also evidence for solute transport defects in dek5 leaves. Maize fixes CO₂ in mesophyll chloroplasts and shuttles malate into bundle sheath chloroplasts for starch synthesis (Majeran et al., 2005; Weise et al., 2011). TEM of dek5 bundle sheath chloroplasts found few starch granules but extensive thylakoid vesicular structures consistent with an osmotic imbalance between the chloroplast and cytosol. Moreover, dek5 mesophyll chloroplasts accumulate starch, potentially indicating reduced flux of malate from the mesophyll.

Proteomics found reduced metabolite and ion transporters in dek5 envelopes. P_i uptake was not detected for dek5 chloroplasts in vitro. The sensitivity of the assay indicates that dek5 chloroplasts have ±80% reduction in P_i transport. These data support a role for DEK5 in envelope protein accumulation and metabolite transport. Transport of exogenous ATP into purified chloroplasts is required for efficient in vitro chloroplast import (Theg et al., 1989). Although we did not specifically measure ATP uptake, reduced transport could explain reduced protein import efficiency observed for tic236 mutants (Chen et al., 2018).

Proteomics and P_i uptake data are consistent with dek5 chloroplast size and ultrastructure. Mutant chloroplasts have much larger chloroplasts, but envelope surface area is reduced relative to volume. A. thaliana chloroplast division mutants have fewer and larger chloroplasts but do not limit envelope membrane biogenesis (Pyke and Leech, 1992; Pyke et al., 1994; Robertson et al., 1996; Shimada et al., 2004; Maple et al., 2007). TEM of division mutants shows elongated chloroplasts that maintain a surface area to volume ratio closer to WT.

The peripheral reticulum is also expanded in dek5, which is a tubular network derived from the inner envelope (Wise, 2007). This ultrastructural change may indicate differential development of inner and outer envelopes analogous to a mutation in a TamB homologue, morC, in Aggregatibacter actinomycetemcomitans. The morC mutation reduces outer membrane relative to inner membrane, resulting in a smooth surface, when WT cells have convoluted, rugose outer membranes (Gallant et al., 2008; Azari et al., 2013). Like dek5 chloroplasts, morC mutants have a more spherical cell morphology, suggesting that loss of TamB function can lead to similar membrane imbalance in both bacteria and chloroplasts.

TamB mutants in E. coli, A. actinomycetemcomitans, and Deinococcus radiodurans all disrupt localization of a subset of outer membrane proteins (Selkrig et al., 2012; Azari et al., 2013; Shen et al., 2014; Yu et al., 2017). In the dek5 outer envelope, half of the membrane proteins are reduced, while a subset of inner envelope protein have increased relative levels. This observation argues that the outer envelope is more compromised in dek5 relative to the inner envelope.

Thylakoids can also be disrupted in dek5, but the loss of thylakoid organization is likely secondary to defects in envelope proteins. Toc34, Toc75, and Toc159 are the core protein translocation complex (Schleiff et al., 2003; Baldwin et al., 2005). Based on proteomics, all three of these proteins have reduced levels in dek5 envelopes. Toc75 levels in immunoblots correlate with the amount of DEK5 protein in the dek5 allelic series. Toc75 is a β-barrel integral membrane protein that is the channel for translocation of nuclear-encoded proteins into the chloroplast (reviewed in Bölter and Soll, 2016). Loss of Toc75 in A. thaliana causes early embryo arrest, while knockdown reduces protein import, chloroplast size, and thylakoid development (Schleiff et al., 2003; Hust and Gutensohn, 2006). Variations in import machinery levels in dek5 is expected to limit chloroplast protein accumulation and cause defects in thylakoid membranes.

The function of TamB in bacteria is to promote integration of β-barrel membrane proteins, including Omp85 family members like Toc75 (Selkrig et al., 2012; Shen et al., 2014; Yu et al., 2017). In E. coli, TamB interacts with TamA, the BAM subunits, and client proteins (Babu et al., 2018). The A. thaliana DEK5 orthologue, TIC236, interacts with Toc75, and weak tic236 alleles reduce in vitro protein import (Chen et al., 2018). The tic236 mutant did not impact Toc75 protein levels, and TIC236 was interpreted as having a direct role in import. Based on this postulated role, tic236 chloroplasts should be reduced in size, similar to a knockdown of toc75. Chloroplast structure was not investigated in tic236, but both ssg4 and dek5 show greatly enlarged chloroplasts, robust thylakoid development, and a relative reduction in envelope. These chloroplast phenotypes are opposite to predicted for a protein import mutant. We propose it is more likely that DEK5 plays a direct role in targeting β-barrel membrane proteins to the outer envelope.

Biochemical analysis of DEK5 topology further supports the model that plant TamB proteins have an analogous function with bacterial homologues. TamB is an inner plasma membrane protein that bridges the periplasmic space with an N-terminal TMD and a C-terminal TamB domain (Selkrig et al., 2012; Shen et al., 2014). Plant TamB proteins are much bigger than homologous proteins in proteobacteria. Matsushima et al. (2014) proposed that SSG4 has a distinct biochemical function from TamB, in part, based on protein length differences. However, Gram-negative bacteria have a 2–6-nm periplasm, while cyanobacteria and chloroplasts have thicker periplasm and intermembrane spaces of 10–35 nm (Hoiczyk and Hansel, 2000). Plant DEK5 proteins would need to be larger to bridge the intermembrane space of the plastid.

We used two experimental approaches to investigate DEK5 topology. An N-terminal region of DEK5 fused to GFP encompasses chlorophyll fluorescence identical to other envelope proteins (Okawa et al., 2008; Singh et al., 2008; Kasmati et al., 2011). Antibodies that cross-react with the C-terminal region of DEK5 support a model in which DEK5 is an intermembrane space protein associated with both outer and inner envelopes. This is the same topology as observed for E. coli TamB and TIC236 (Selkrig et al., 2012; Shen et al., 2014; Chen et al., 2018).

Reconciling all available data for plant TamB proteins argues for a function in chloroplast envelope biogenesis. Based on TamB function, we propose that DEK5, SSG4, and TIC236 have conserved roles in targeting a subset of β-barrel proteins to the outer envelope. The proteomics experiment identified Toc75 and OEP80 as likely client proteins. OEP37 levels were just above the twofold reduction cutoff in the proteomics analysis and may also be a DEK5 client. The correlation between the levels of chloroplast import machinery and the severity of dek5 mutant alleles suggests that severe alleles display extreme pleiotropy with plastids being unable to support import of nuclear-encoded proteins. Weak alleles that support protein import would also retain sufficient DEK5 activity to support β-barrel protein insertion. Consequently, direct proof of the proposed functions in OEP insertion or protein import will require biochemical methods to block DEK5 action in WT chloroplasts.

All maize plants were grown during April–July or August–December at the University of Florida Institute of Food and Agricultural Sciences Plant Science Research and Education Unit in Citra, Florida. The dek5 and normal siblings were germinated in a greenhouse with Metro-Mix 300 (Scotts-Sierra). The dek5-N874A, dek5-N961, and dek5-N1339A alleles were originally isolated by Neuffer and Sheridan (1980) from ethyl methanesulfonate mutagenesis. Scanlon et al. (1994) isolated dek5-PS25 and dek5-MS33 from Mutator transposon-tagging populations. The dek5-Briggs allele is a spontaneous mutation (www.maizegdb.org). Stocks for dek5 alleles were ordered from the Maize Genetics COOP Stock Center. Each dek5 allele was crossed with B73, W22, or Mo17 inbred lines five times to develop BC₄ introgression stocks to characterize phenotypes. The dek5-umu1 (mu1021358) and dek5-umu2 (mu1047944) UniformMu alleles were maintained in the W22 inbred background. We confirmed all stocks were dek5 alleles by genetic complementation tests using reciprocal crosses between each stock and dek5-PS25. First ears were used for reciprocal crosses. Second ears of both male parents and female parents in the crosses were self-pollinated to determine plant genotypes.

Endosperm composition was determined from dissected tissues. Developing endosperm was flash frozen in liquid nitrogen and stored at −80°C. Frozen endosperm tissues were cracked into small pieces with mortar and pestle, lyophilized for 3 d, and ground to a fine powder with a bead mill. For mature dry kernels, endosperm was isolated by cutting the kernel in the sagittal plane and cutting away the embryo with a utility knife. The pericarp was removed by scraping the surface of the mature dry endosperm with a razor blade. The remaining endosperm tissue was ground to a fine powder with a bead mill.

Zein and nonzein protein was extracted from the dissected, dry endosperm tissue as described by Wu and Messing (2012). Protein extracts were loaded on equal dry weight basis for SDS-PAGE and stained with Coomassie Blue R-250. Endosperm starch was measured from developing kernels using the Megazyme Total Starch Assay Kit (AA/AMG). Sucrose, glucose, and maltose were measured with the Megazyme Maltose/Sucrose/D-Glucose Assay Kit (K-MASUG). ADP-glucose was measured with a glycerol-3-phosphate oxidase/glycerol-3-phosphate dehydrogenase cycling assay as described previously (Gibon et al., 2002). Briefly, frozen endosperm was ground and extracted in 200 mM tricine/KOH, pH 8, containing 10 mM MgCl₂. 30 μl of endosperm extracts or ADP-glucose standards (0–20 pmol) was dispensed into a microplate with 20 µl tricine buffer containing 0.05 U UGPase, 0.05 U glycerokinase, 2.5 U Gly3POX, 130 U catalase, 0.05 U triose phosphate isomerase, 0.1 U GAPDH, 0.02 µmol P P_i, 0.15 µmol NaF, 0.25 µmol glycerol, 0.01 µmol NAD⁺, and 0.1 µmol sodium arsenate. Reactions were incubated for 40 min at room temperature, and 20 µl of 1.2 mM NADH solution containing 0.4 U glycerol-3-phosphate dehydrogenase was added. Absorbance at 340 nm was recorded over 20 min. Next, 0.3 U AGPase was added, and the absorbance read 340 nm for 20 min. The difference in the two rates was used to calculate the amount of ADP-glucose in the sample.

Developing kernels were collected and fixed in FAA solution (3.7% formaldehyde, 5% glacial acetic acid, and 45% ethanol) at 4°C overnight. Kernels were dehydrated in an ethanol series and kept in 100% ethanol. Additional sample preparation and scanning EM observations were performed at Interdisciplinary Center for Biotechnology Research Electron Microscopy core at the University of Florida, Gainesville, FL. Scanning EM data were collected with a Hitachi SU5000 Schottky field emission scanning electron microscope. Starch granule diameter was measured using ImageJ for three scanning EM images from each genotype and developmental stage. For particle size analysis, mature endosperm tissue was dissected and ground into fine powder with a bead mill. Endosperm powder (0.1 g) was mixed with 1 ml of ethanol and vortexed immediately before measurement with a Beckman/Coulter LS 13 320 laser diffraction particle size analyzer.

Single endosperms were homogenized on ice in a mortar and pestle in 4 ml H₂O. Starch from 1 ml lysate was pelleted in a 2-ml polypropylene tube by centrifugation at 1,400 g for 10 min at 4°C. The starch pellet was washed twice in ice cold 80% ethanol with centrifugation at 1,400 g for 5 min. Starch was dissolved by boiling in 1 ml DMSO for 20 min and then quantified by enzymatic conversion to glucose (Lin et al., 2012; Huang et al., 2014). For the debranching reaction, 1.5 mg of dissolved starch polymers was precipitated with 5 vol of ethanol at −20°C for ≥1 h. Glucans were collected by centrifugation at full speed in a microfuge at room temperature and then suspended in 0.5 ml 50 mM sodium acetate, pH 4.5, containing 0.05 U Pseudomonas isoamylase (Megazyme no. E-ISAMY) and incubated overnight at 50°C. The solution was then diluted with 1.5 ml H₂O, filtered through a 0.2-µm syringe-tip filter, and 25 µl was applied to a high-pressure ion-exchange chromatography system with pulsed amperometric detection (HPAEC-PAD) that resolves and quantifies α(1→4)-linked (linear) glucan chains ranging from 2 to 40 glucosyl units. Separation was on a Dionex CarboPac PA-100 4 × 250–mm column with a CarboPac PA-100 4 × 50–mm guard column at a flow rate of 1 ml/min. Elution was in 0.1 M NaOH for 4 min followed by a 40-min gradient of 0–0.4 M sodium acetate in 0.1 M NaOH. Peak areas for each degree of polymerization (DP) were normalized to the total peak area, and the normalized values for each DP were compared between samples.

Maize protoplasts were prepared essentially as described previously (Yoo et al., 2007), with some modifications. Kernels were planted in a greenhouse. After 10 d, 0.5 g seedling leaves was harvested, chopped with a razor blade, and digested in enzyme solution at 28°C for 4–5 h with a gentle shaking at 60 rpm in the dark. The enzyme solution consisted of 1.5% (wt/vol) cellulase R10, 0.5% (wt/vol) macerozyme R10 in 0.4 M mannitol, 20 mM KCl, 10 mM CaCl₂, 5 mM β-mercaptoethanol, 0.1% BSA, and 20 mM MES, pH 5.7. Cell debris was filtered using 35-µm nylon mesh, and collected protoplasts were suspended in W5 solution (2 mM MES, pH 5.7, 154 mM NaCl, 125 mM CaCl₂, and 5 mM KCl) and kept on ice for observation. Protoplast isolation at high osmotic concentration replaced 0.4 M mannitol with 0.8 M mannitol in all solutions. Protoplast yield was determined by counting intact protoplasts in W5 solution using a hemocytometer and a Zeiss Standard WL microscope with a Neofluor 40×/0.75 objective at room temperature. Light microscopy imaging used an AMScope MU500 digital camera with the AMscope acquisition software version 3.7 for a Windows 64-bit operating system.

Chloroplast number within each cell was determined with Z-stack images of individual protoplast cells using a Leica TCS SP5 laser-scanning confocal microscope (Leica Microsystems). Images were captured at 37°C using a 100× PL APO NA 1.40 oil-immersion objective (Leica) and a filter cube set for FITC/TRITC for multicolor fluorescence. Leica LAS AF software was used to acquire images with water as the imaging medium. Chlorophyll was excited at 488 nm and detected with an emission band of 650–700 nm. A 3D image was reconstructed for each cell and manually scored for the number of chloroplasts. Chloroplast cross-sectional surface areas and volume were measured using ImageJ (Schindelin et al., 2012).

Fresh seedling leaf tissue (0.5 g) from both dek5 mutant and normal siblings was harvested and chopped with a razor blade. The leaf tissue was extracted in 15 ml of 96% ethanol with gentle agitation for 12 h at room temperature in aluminum foil wrapped bottles. Ethanol extracts were separated from tissue debris with filter paper. The tissue was rinsed with 96% ethanol to completely extract residual pigment. The extract and rinse were combined and brought to 25 ml final volume in 96% ethanol. Chlorophyll a, chlorophyll b, and carotenoid absorbance was measured at 665, 649, and 470 nm using a spectrophotometer. Pigment content was calculated according to Lichtenthaler and Wellburn (1983).

Chloroplast protein or total seedling proteins were separated by 10% SDS-PAGE and transferred to nitrocellulose membrane using a Bio-Rad Trans-Blot SD semi-dry transfer cell at 15 V for 30 min. Membrane was first blocked with blocking solution containing 3% BSA and 0.2% Tween 20 with PBS (pH 7.4) at room temperature for 30 min and then incubated with primary antibody at 1:2,000 dilution in blocking solution for 2 h at room temperature or overnight at 4°C. Membranes were washed four times for 5 min using PBS with 0.2% Tween 20. Primary antibody signal was detected after incubating goat anti-rabbit secondary antibody (Sigma) at 1:1,000 dilution in blocking solution for 2 h at room temperature followed by four 5-min washes using PBS with 0.2% Tween 20. Chemiluminescent was produced by applying 1 ml ECL substrate (Thermo Scientific Pierce) per gel membrane and imaging with a FOTODYNE FOTO Luminary/FX digital imaging system.

Seedling leaf tissues were harvested 10 DAS in a growth chamber. Sample fixation, sectioning and TEM were completed by Electron Microscopy Services, as described previously (Turgeon and Medville, 2004). Briefly, leaf samples were fixed for 4 h at 4°C in 2% (wt/vol) paraformaldehyde and 2.5% (vol/vol) glutaraldehyde in 70 mM sodium cacodylate buffer, pH 6.8, followed by overnight incubation in 1% (wt/vol) osmium tetroxide in fixation buffer. Fixed tissues were subsequently dehydrated and embedded in resin. Ultrathin sections were stained with uranyl acetate and lead citrate for observations at 60 kV with a Philips EM-300 transmission electron microscope.

Chloroplast area, long (a) axis, and short (b) axis were measured in ImageJ from TEM images using freehand selections for area and straight line selections for axes (Schindelin et al., 2012). Estimated cross-sectional area was calculated using the formula for an ellipse (⁠$A=πab$⁠). Stroma and thylakoid areas were estimated by first determining the average auto threshold value for all TEM images. The average threshold value was applied to all images to measure thylakoid area, which was subtracted from chloroplast area to estimate stroma area. Calculations for ellipsoid volume,

and surface area,

(P = 1.6075), assumed the chloroplast long (a) axis represents length and the short (b) axis represents height and width. Thylakoid and stroma volumes were calculated for each chloroplast based on the relative fraction of the total chloroplast area. All traits were calculated for each chloroplast individually with mean and standard deviation statistics calculated using the estimated traits. Models for a cylinder volume,

and surface area,

assumed the chloroplast long (a) axis was the height and the short (b) axis was the diameter. Models for a sphere volume,

and surface area,

assumed a single radius (r).

An F₂ mapping population was generated by self-pollinating F₁ plants from a cross of the Mo17 inbred and dek5-PS25/+. Genomic DNA was extracted from individual F₂ mutant kernels and analyzed with simple sequence repeat and insertion-deletion PCR markers as described previously (Settles et al., 2014). Table S2 gives primer sequences of the markers used for linkage analysis. For the transposon flanking sequence experiment, homozygous dek5-PS25 and dek5-MS33 mutant genomic DNA was isolated from pools of six F₂ mutant kernels. Mu flanking sequences were amplified, purified, and adapted to the Illumina sequencing platform as described previously (McCarty et al., 2013). Reads were quality filtered and mapped to the B73_v3 genome assembly using the UniformMu informatics pipeline (McCarty et al., 2013).

Total RNA was isolated using Trizol (Invitrogen) from 10 DAS B73 maize seedlings. Approximately 1 mg of total RNA was reverse transcribed with either ThermoScript Reverse transcription (Invitrogen) for Dek5 cDNA cloning. Both oligo (dT) and random hexamer primers were used for first-strand synthesis to ensure full-length coverage of the Dek5 mRNA. The Dek5 cDNA was amplified with Phusion high-fidelity DNA polymerase (New England Biolabs) and cloned into pENTR/D-TOPO Vector (Invitrogen) as three overlapping fragments: Dek5-FN (1802 bp), Dek5-FM (2330 bp), and Dek5-FC (3006 bp). The full-length Dek5 cDNA was assembled from the sequence of these clones. Specific PCR primers are listed in Table S2.

Protein sequences were aligned using ClustalW (Larkin et al., 2007). The phylogenetic tree was produced with MEGA7 using the neighbor-joining method with 2,000 bootstrap replications and the Poisson substitution model (Hedges, 1992). The DEK5 homologue sequences used for phylogenetic analysis were Zea mays (GenBank accession no. MF034077), Sorghum bicolor (UniGene accession no. XP_002457235.1), Glycine max (NCBI Protein accession no. XP_003527803.1), Oryza sativa Japonica (Protein accession no. XP_015619226.1), Setaria italic (Protein accession no. XP_004971719.1), Fragaria vesca subsp. Vesca (Protein accession no. XP_011459147.1), Vitis vinifera (Protein accession no. XP_010648561.1), A. thaliana (Protein accession no. NP_180137.3), Theobroma cacao (GenBank accession no. EOY31352.1), Amborella trichopoda (Protein accession no. XP_006844393.1), Selaginella moellendorffii (Protein accession no. XP_002982313.1), Ostreococcus tauri (Protein accession no. XP_003079877.1), Nostoc punctiforme (Protein accession no. WP_012412249.1), Cyanobacterium PCC 7702 (Protein accession no. WP_017320409.1), Pseudomonas stutzeri (Protein accession no. WP_003289734.1), Methylobacterium sp. Leaf91 (Protein accession no. WP_082490488.1), Rhizobium leguminosarum (Protein accession no. WP_003543422.1), Arhodomonas aquaeolei (Protein accession no. WP_018716326.1), and E. coli (Protein accession no. WP_040089969.1).

The DEK5N-GFP fusion protein was constructed by amplifying the initial 1,374 bp of the Dek5 ORF from B73 seedling cDNA using Phusion high-fidelity DNA polymerase (New England Biolabs) with the Dek5N-L and Dek5N-R primers (Table S2). The N-terminal ORF product was cloned into pENTR vector (Invitrogen) and recombined into the binary vector, pB7FWG2, to generate a C-terminal GFP fusion (Karimi et al., 2002, 2007). The TMD deletion was generated with the Q5 Site-Directed Mutagenesis Kit (New England Biolabs) to delete the peptide sequence, LFLVRCSVFAAVVSVAGALSW, in-frame from the DEK5N clone and then recombined into pB7FWG2. The binary constructs were transformed into Agrobacterium tumefaciens strain GV3101 by the freeze–thaw method (Weigel and Glazebrook, 2006). Each construct was transferred to N. benthamiana leaves via agroinfiltration as described previously (van Esse, 2012), with some modifications. Before infiltration, the agrobacteria were suspended in buffer containing 10 mM MgCl₂, 10 mM MES, and 200 µM acetosyringone and adjusted to an OD₆₀₀ of 0.4. The N. benthamiana plants were grown in a growth chamber at 23°C with 16-h/8-h day/night light cycle. Leaves from 4–5-wk-old plants were infiltrated and examined by epifluorescence microscopy between 40 and 50 h after infiltration. Representative images were captured at 37°C using a 100× PL APO 1.40-NA oil-immersion objective (Leica) on an inverted Leica SP5 microscope equipped with a filter cube set for FITC/TRITC for multicolor fluorescence. Leica LAS AF software was used to acquire images with water as the imaging medium. GFP was excited at 488 nm and detected with an emission band of 500–530 nm. Chlorophyll was detected with an emission band of 650–700 nm. Captured images were processed with ImageJ to add scale bars to split and merged color channels.

The coding sequence for the DEK5 TamB domain was amplified with Phusion high fidelity DNA polymerase from B73 cDNA with the DUF490-L and DUF490-R primers (Table S2). The PCR product was digested with NcoI and NotI, purified, and cloned into the pMAL-c5X vector (New England Biolabs). Recombinant protein was expressed in the E. coli Top10 strain by inducing 12 liters of culture with 10 mM IPTG for 1 h at 30°C. Cells were sonicated using a 5-s/5-s pulse/pause cycle for a total of 3 min in an ice water bath. Recombinant protein was purified using amylose resin according to the manufacturer’s instructions (New England Biolabs). The recombinant protein was concentrated in Amicon Ultra-15 Centrifugal Filters (EMD Millipore), and 2.5 mg of purified recombinant protein was sent to Thermo Scientific Pierce for rabbit polyclonal antibody production.

Other marker proteins were detected with rabbit polyclonal antibody sera raised against human histone H3 C-terminal peptide (ab1791; Abcam), spinach ATP synthase CF1-β (Fristedt et al., 2014), A. thaliana 6PGDH (Hölscher et al., 2016), sorghum OEC23 (Voelker and Barkan, 1995), maize Cyt f (Barkan et al., 1986), A. thaliana Toc75 (Ma et al., 1996), and A. thaliana OEP80 (Inoue and Potter, 2004).

Normal chloroplasts were isolated essentially as described previously (Voelker and Barkan, 1995). Leaf tissue from 7-d-old plants was cut and homogenized in grinding buffer (50 mM Hepes/KOH, pH 7.5, 0.33 M sorbitol, 1 mM MgCl₂, 1 mM MnCl₂, 2 mM EDTA, 5 mM Na-ascorbate, and 1% BSA) using a Warring blender with five bursts of ∼5 s at max speed. The homogenate was filtered through two layers of Miracloth, and the filtrate was centrifuged at 3,000 g for 5 min in a swinging bucket rotor at 4°C. The organelle pellet was gently suspended in 5 ml grinding buffer and layered onto a 35%/75% (8 ml/16 ml) Percoll step gradient. The gradient was centrifuged at 6,000 g for 15 min, and the lower, intact chloroplast, band was recovered. Intact chloroplasts were washed with import buffer (0.33 M sorbitol and 50 mM Hepes-KOH, pH 8.0) and pelleted by centrifugation at 1,000 g for 5 min. Pelleted chloroplasts were suspended in import buffer and adjusted to 0.5–1.0 mg chlorophyll per ml based on OD₆₅₂.

Mutant dek5 chloroplasts were purified from protoplasts. Isolated dek5 mutant protoplasts were washed and suspended in a protoplast hypotonic lysis buffer of 0.7 M sorbitol supplemented with 50 mM Hepes-KOH (pH 7.5). The cell lysate was passed through a 15-µm nylon mesh filter (Tisch Scientific) to further lyse remaining intact protoplasts. Released dek5 mutant chloroplasts were purified in a 25%/75% Percoll step gradient centrifuged at 6,000 g for 15 min, and the lower, intact chloroplast band, was recovered. Isolated dek5 chloroplasts were washed with 0.7 M sorbitol and 50 mM Hepes-KOH, pH 7.5, pelleted, and adjusted to 0.5–1 mg/ml chlorophyll.

Chloroplasts were fractionated as described by Smith et al. (2003), with some modifications. Briefly, intact isolated chloroplasts were pelleted from import buffer and resuspended to 1–2 mg/ml chlorophyll in TE buffer (10 mM tricine and 2 mM EDTA, pH 7.5) with 0.6 M sucrose. Chloroplasts were lysed with two cycles of freezing and thawing and then diluted with TE buffer containing protease inhibitor cocktail (P9599; Sigma) to 0.2 M sucrose. Lysed chloroplasts were centrifuged at 45,000 g for 60 min to collect the crude envelope membranes. The soluble supernatant was collected and stored at −80°C for further analysis. The membrane pellet was resuspended in TE buffer containing 0.2 M sucrose to a concentration of 2–3 mg/ml chlorophyll, and 1 ml of the membrane suspension was layered onto the top of a sucrose gradient. The gradient consisted of 1.5 ml of 1 M sucrose, 1 ml of 0.8 M sucrose, and 1 ml of 0.46 M sucrose in a 5-ml polyallomer tube. The gradient was centrifuged for 1.5 h at 270,000 g at 4°C using low acceleration and deceleration rates. The yellow-green band from the 0.8 to 1 M sucrose interface was collected as the IEM fraction. The light yellow band from the 0.46–0.8 M sucrose interface was collected as the OEM faction. The green thylakoid pellet was resuspended in TE. All membrane fractions were washed using 3–5 vol of TE buffer supplemented with protease inhibitor and centrifuged for 1.5 h at 270,000 g at 4°C.

Normal and dek5 chloroplasts were isolated and diluted to 1 mg chlorophyll per milliliter in import buffer. For each wash condition, 100 µl of chloroplasts was pelleted and resuspended in 0.1 M Na₂CO₃, pH 11.5, 6 M urea in 10 mM Hepes/KOH, pH 7.6, or 1 M NaCl. Each wash was held at room temperature for 30 min. Freeze–thaw lysis, as described for chloroplast fractionation, was used as a control. Treated chloroplasts were centrifuged at 40,000 g for 1 h at 4°C. Pellets were resuspended in TE on an equal volume basis with supernatants before SDS-PAGE and immunoblotting.

The dual protease protection assay was done according to Froehlich (2011) with some modifications. Intact chloroplasts were isolated and suspended in import buffer (0.33 M sorbitol and 50 mM Hepes/KOH, pH 8.0) to a concentration of 1 mg/ml chlorophyll. Intact chloroplasts (150 µl) were incubated with either 25 µl thermolysin or trypsin (1 mg/ml) for 50 min on ice. Thermolysin was stopped using 10 mM EDTA and kept on ice for 5 min. The trypsin digestion was stopped using 50 µl of protease inhibitors containing 1 mg/ml trypsin inhibitor (Sigma), 10 µM/ml leupeptin, and 10 mM PMSF. Intact chloroplasts were recovered using 30% Percoll cushion in import buffer and centrifuged at 1,500 g for 5 min. Pelleted intact chloroplasts were washed and resuspended in import buffer for further analysis.

To collect enough chloroplast envelope proteins for proteome analysis, six independent chloroplast preps from dek5 and four independent chloroplast preps from normal siblings were pooled separately and resuspended to 1–2 mg/ml chlorophyll in suspension buffer (0.6 M sucrose, 10 mM tricine, and 2 mM EDTA, pH 7.5). Both the dek5 and normal chloroplasts were lysed with two cycles of snap freezing and thawing and then diluted with TE buffer containing protease inhibitor cocktail to 0.2 M sucrose. A Dounce homogenizer was used to further rupture the dek5 chloroplasts. Lysed chloroplasts were centrifuged at 4,000 g for 15 min to remove most of the thylakoid membranes. The supernatant was transferred to a new tube and centrifuged at 40,000 g for 60 min to collect crude envelope membrane fraction. Enriched chloroplast envelope proteins were solubilized in 8M urea, 100 mM Tris-HCl, pH 7, and 5 mM Tris-(2-carboxyethyl) phosphine, hydrochloride (TCEP). Proteins were then subjected to four rounds of acetone precipitation with probe sonication between each round. Pelleted protein was solubilized in UA buffer (8 M urea, 50 mM Tris-HCl, pH 7, and 5 mM TCEP) and processed using the filter-aided sample preparation method (Wiśniewski et al., 2009; Song et al., 2017) with microcon YM-30 centrifugal filters (Millipore). Proteins were digested on the filter overnight with 1 µg trypsin (Roche) at 37°C. Proteins were then further digested with 0.1 µg of trypsin and 0.1 µg of LysC (Wako) for 2 h. Recovered peptides were acidified to a pH of ∼2–3 with formic acid and desalted with 50 mg Sep-Pak C18 cartridges (Waters). Eluted peptides were dried in a vacuum centrifuge, resuspended in 0.1% formic acid, and quantified using the Pierce BCA Protein assay kit.

1 μg of peptides were separated on a 20-cm nanospray column, which was pulled and packed in-house with 2.5 µm C18 (Waters), using an acetonitrile gradient of 5–30% for 120 min, 30–80% for 25 min, and 0% for 5 min (150 min total) that was delivered via an Agilent 1260 quaternary HPLC at a flow rate of ∼500 nl/min. The HPLC system was coupled with a Thermo Scientific Q-Exactive Plus high-resolution quadrupole Orbitrap mass spectrometer using a custom-fabricated nanospray source. Data-dependent acquisition was obtained using Xcalibur 4.0 software in positive-ion mode with a spray voltage of 2.00 kV, radio frequency of 60, and a capillary temperature of 275°C. MS1 spectra were measured at a resolution of 70,000 with an automatic gain control of 3 × 10⁶, a maximum ion time of 100 ms, and a mass range of 400–2,000 m/z. Up to 15 MS2, with a charge state of 2 to 4, were triggered at a resolution of 17,500, an automatic gain control of 10⁵ with a maximum ion time of 50 ms, a 1.5-m/z isolation window, and a normalized collision energy of 28. MS1 that triggered MS2 scans were dynamically excluded for 25 s. We performed two runs for each sample to generate two technical replicates.

The raw data were searched against the B73 RefGen_v2 5b Filtered Gene Set using MaxQuant version 1.5.8.3 (Tyanova et al., 2016). Methionine oxidation and protein N-terminal acetylation were set as variable modifications. The digestion parameters were set to “specific” for “Trypsin/P;LysC” with a maximum of two missed cleavages. The match between runs feature was turned off. Default settings were used for the remaining parameters including a peptide spectrum match and protein false discovery rate of 0.01, which was determined using a reverse decoy database.

The dek5 mutant and normal sibling envelope protein extracts had different levels of nonenvelope contaminants. To compare relative levels of envelope proteins, we curated a set of 96 envelope proteins that were validated experimentally. The MaxQuant intensity for the 85 envelope peptides within this subset found in at least one technical replicate were normalized relative to one normal technical replicate. Proteins identified in a single technical replicate with a normalized MaxQuant intensity value <10⁷ were removed to leave 68 proteins. Relative fold change between dek5 and normal sibling samples were calculated from the average of the two technical replicates or the single MaxQuant intensity value if the protein was detected only once. TMDs for envelope proteins were predicted using TMHMM Server V2.0 (http://www.cbs.dtu.dk/services/TMHMM/), and envelope localization was annotated using the Plant Proteome Database (Sun et al., 2009).

Uptake of ³²P_i into isolated chloroplasts was assayed as described with some modifications (Fliege et al., 1978). Purified, intact chloroplasts from B73 and dek5-N961 mutants were adjusted in concentration to OD₆₅₂ of 0.2 and 90 µl (∼2.3× 10⁷ normal and 4.6× 10⁶ dek5) chloroplasts were mixed with 10 µl 1 mM KH₂PO₄, pH 7.5. KH₂PO₄ was prepared in a 19:1 ratio of unlabeled to radiolabeled 37 MBq/mM KH₂³²PO₄ (PerkinElmer). Chloroplasts were incubated in a time course from 5 to 300 s. Phosphate uptake was stopped by adding 2 mM 4,4′-diisothiocyanatostilbene-2,2′-disulfonic acid (DIDS; Sigma) to the reaction. Relative uptake was calculated from background radiolabel as determined by preincubating chloroplasts with DIDS for 5 min before incubating with ³²P_i for 300 s. Intact normal chloroplasts with radiolabel were repurified by transferring 100 µl of the chloroplast suspension to a centrifuge tube containing a lower layer of 30 µl of 10% perchloric acid and an upper layer of 70 µl AP 150 silicone oil (Wacker). The density of dek5 mutant chloroplasts was lower than normal, and intact chloroplasts were purified using a 4:1 mixture of AP 150 and AR 20 silicone oil (Sigma). The radiolabeled chloroplast suspension was overlaid onto the silicone oil and then centrifuged for 30 s at 14,000 g. The radiolabel associated with intact chloroplasts was assayed by sampling 15 µl of the 10% perchloric acid layer for scintillation counting using a Beckman LS6500SE scintillation counter.

Detection limits were determined after a 300-s incubation with normal chloroplasts diluted from 5% to 50% of an OD₆₅₂ of 0.2. Mutant chloroplasts were concentrated by centrifugation of two- and fourfold larger volumes at 500 g for 1 min followed by resuspension in 90 µl of 0.7 M sorbitol and 50 mM Hepes-KOH, pH 7.5. Stromal volume was estimated by counting intact chloroplasts and multiplying by the average stromal volume estimated from TEM. Chloroplasts were counted in a hemocytometer using a Zeiss Standard WL microscope with a Neofluor 40×/0.75 objective at room temperature.

Assay effects on chloroplast integrity was determined by incubating chloroplasts with 0.1 mM KH₂PO₄, pH 7.5, for 300 s. Chloroplasts were repurified through silicone oil with a bottom layer of 30 µl of 50 mM Hepes-KOH, pH 7.5, and 0.3 M or 0.7 M sorbitol for normal and dek5 mutants, respectively. Pelleted, intact chloroplasts were washed twice with the sorbitol buffer to remove residual silicone oil, observed at room temperature under a Zeiss Standard WL microscope with a Neofluor 40×/0.75 objective with a phase-contrast condenser. Microscopy images were captured with an AMScope MU500 digital camera with AMscope acquisition software version 3.7 for a Windows 64-bit operating system.

Fig. S1 shows the influence of different genetic backgrounds on dek5 kernel and plant phenotypes. Fig. S2 provides additional TEM images showing the range of chloroplast defects in dek5. Fig. S3 is a multiple sequence alignment of N-terminal DEK5 protein sequences showing conserved regions used to determine sites for GFP fusion protein constructs. Fig. S4 shows control experiments for the DEK5 antibody and biological replicates of experiments in Fig. 9. Table S1 lists data from the proteomic analysis of chloroplast envelope proteins. Table S2 lists the primers used in this study. Video 1 shows Z-stack images of chlorophyll autofluorescence in normal and dek5 protoplasts to provide evidence for a lack of chlorophyll in the interior of some dek5 chloroplasts.

We thank Richard Medville at the Electron Microscopy Services (Colorado Springs, CO) for TEM analysis. We thank the following collaborators for providing antibodies: Dr. Alice Barkan at the University of Oregon (Eugene, OR; OEC23, Cyt F, and CF1-β), Dr. Carole Dabney-Smith at Miami University (Miami, FL; Toc75), Dr. Antje von Schaewen at the University of Münster (Münster, Germany; 6PGDH), and Dr. Kentaro Inoue at University of California, Davis (Davis, CA; OEP80). We also thank Drs. L. Curtis Hannah and Ken Cline at the University of Florida for intellectual input.

This work was supported by the National Institute of Food and Agriculture (awards 2011-67003-30215 and 2018-51181-28419), the China Scholarship Council, and the Vasil-Monsanto Endowment.

The authors declare no competing financial interests.

Author contributions: J. Zhang and A.M. Settles designed the research, performed experiments, analyzed data, and wrote the manuscript; S. Wu completed Mu-Seq, and D.R. McCarty analyzed the data; S.K. Boehlein measured ADP-glucose; G. Song and J. Walley performed the chloroplast envelope proteomics. A. Myers measured amylopectin chain distribution.