Maize endosperm-specific transcription factors O2and PBF network the regulation of protein andstarch synthesisZhiyong Zhanga,b, Xixi Zhenga, Jun Yanga, Joachim Messingb,1, and Yongrui Wua,1

aNational Key Laboratory of Plant Molecular Genetics, Chinese Academy of Sciences Center for Excellence in Molecular Plant Sciences, Institute of PlantPhysiology and Ecology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai 200032, China; and bWaksman Institute ofMicrobiology, Rutgers University, Piscataway, NJ 08854

Contributed by Joachim Messing, August 18, 2016 (sent for review August 15, 2016; reviewed by L. Curtis Hannah and Alan Myers)

The maize endosperm-specific transcription factors opaque2 (O2) andprolamine-box binding factor (PBF) regulate storage protein zeingenes. We show that they also control starch synthesis. The starchcontent in the PbfRNAi and o2 mutants was reduced by ∼5% and11%, respectively, compared with normal genotypes. In the double-mutant PbfRNAi;o2, starch was decreased by 25%. Transcriptomeanalysis reveals that >1,000 genes were affected in each of the twomutants and in the double mutant; these genes were mainly enrichedin sugar and proteinmetabolism. Pyruvate orthophosphate dikinase 1and 2 (PPDKs) and starch synthase III (SSIII) are critical components inthe starch biosynthetic enzyme complex. The expression of PPDK1,PPDK2, and SSIII and their protein levels are further reduced in thedouble mutants as compared with the single mutants. When thepromoters of these genes were analyzed, we found a prolamine boxand an O2 box that can be additively transactivated by PBF and O2.Starch synthase IIa (SSIIa, encoding another starch synthase for amy-lopectin) and starch branching enzyme 1 (SBEI, encoding one of thetwomain starch branching enzymes) are not directly regulated by PBFand O2, but their protein levels are significantly decreased in the o2mutant and are further decreased in the double mutant, indicatingthat o2 and PbfRNAimay affect the levels of some other transcriptionfactor(s) or mRNA regulatory factor(s) that in turn would affect thetranscript and protein levels of SSIIa and SBEI. These findings showthat three important traits—nutritional quality, calories, and yield—are linked through the same transcription factors.

gene regulation | zein | starch synthesis | yield | breeding strategy

Maize (Zea mays) is one of the most important food sourceson earth. Its endosperm is composed of ∼70% starch and

10% protein. Although starch is the main source of caloriesconsumed, protein provides the critical nutrients for our foodsupply. However, maize’s nutritional quality is poor because itsmain storage protein, zein, is devoid of the essential amino acidslysine and tryptophan, and corn-based animal feed must besupplemented with soy protein and synthetic methionine. Theclassic mutant opaque2 (o2) was found to improve the seed nu-tritional value by reducing the synthesis of zein proteins (1). O2encodes an endosperm-specific bZIP transcription factor andmainly regulates the expression of α- and β-zein genes by rec-ognizing the O2 box in their promoters (2–4). For practical ap-plications, however, the o2 mutant has multiple agronomicdefects, i.e., soft texture, susceptibility to disease, and yield drop.Prolamine-box binding factor (PBF), another endosperm-spe-cific transcription factor, belongs to the DOF family and regu-lates the expression of zein genes by recognizing the prolamine(P) box in their promoters (5, 6).Starch, the main contributor of kernel weight, is synthesized and

assembled into semicrystalline starch granules by a suite of well-characterized enzymes in the starchy endosperm cells, includingsucrose synthase (SUS), ADP-glucose pyrophosphorylase (AGP),soluble starch synthase (SS), granule-bound starch synthase (GBSS),starch branching enzyme (SBE), and starch debranching enzyme

(DBE) (7, 8). Waxy (GBSSI) is involved in the synthesis of amy-lose, whereas SSs are mainly involved in the synthesis of amylo-pectin, which is required for starch granule formation. Three starchsynthases, i.e., SSI, SSIIa (Sugary2), and SSIII (Dull1) are prefer-entially expressed at the filling stage of the endosperm and arethought to be primarily responsible for amylopectin synthesis in theamyloplasts (7–9). The functions of SSIIa and SSIII have beengenetically proven by mutant analysis (10, 11), but null mutants forSSI have not yet been identified. Interestingly, biochemical studiesdemonstrate that starch biosynthetic enzymes and proteins frommultiple metabolic pathways associate with each other to formhigh-molecular-weight complexes in wheat and maize endospermamyloplasts (12–14). Almost all SSIII and SSIIa exist in the com-plex form and are considered regulators of starch biosynthesis aswell as of enzymatic functions (14). Pyruvate orthophosphatedikinase (PPDK) is a key enzyme for CO2 fixation, which catalyzespyruvate (PYR) to phosphoenolpyruvate (PEP) conversion in C4photosynthesis (15). This protein is also abundant in the non-photosynthetic tissue of endosperm in C3 and C4 cereal grasses(16, 17). Although the exact biological function of endospermPPDK is still unclear, a small percentage of PPDK that exists inamyloplasts can associate stably with starch biosynthetic enzymes(14), suggesting that endosperm PPDK might be involved in starchor other reserve synthesis. In addition to the starch biosyntheticpathway, the oxidative pentose phosphate pathway (oxPPP) is alsothought to play an important role in endosperm starch synthesis(18). A recent report showed that loss of 6-phosphogluconate de-hydrogenase (PGD3) in oxPPP leads to severely reduced grain-fillphenotypes with reduced starch accumulation in maize (19).

Significance

Nutritional quality and yield are equally important considerationsin crop breeding, although they sometimes appear at odds. In thiswork we made the discovery that these traits are linked throughregulation by two transcription factors. Mutations that affect theexpression of these transcription factors can improve the nutri-tional quality of the seed but also can reduce kernel yield andhardness. Therefore future corn-breeding programs should si-lence zein genes directly, not by blocking transcription factors.

Author contributions: Z.Z., J.M., and Y.W. designed research; Z.Z., X.Z., and J.Y. performedresearch; Z.Z., J.M., and Y.W. analyzed data; and Z.Z., J.M., and Y.W. wrote the paper.

Reviewers: L.C.H., University of Florida; and A.M., Iowa State University.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

Data deposition: The sequence reported in this paper has been deposited in the NationalCenter for Biotechnology Information Gene Expression Omnibus database (accession no.GSE79513).1To whom correspondence may be addressed. Email: messing@waksman.rutgers.edu oryrwu@sibs.ac.cn.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1613721113/-/DCSupplemental.

10842–10847 | PNAS | September 27, 2016 | vol. 113 | no. 39 www.pnas.org/cgi/doi/10.1073/pnas.1613721113

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

021

Mutations of these starch biosynthetic genes generally cause areduction in starch content and, in turn, the kernel yield (20), butrare reports show that the transgenic manipulation of starch bio-synthetic genes is able to increase them. A modified maize,Shrunken-2 (Sh2), encoding the large subunit of AGP (HS33/Rev6Sh2) was transformed to enhance yield by increasing the seednumber rather than the starch mass (21). These facts suggest thatincreasing endosperm starch content by transgenic manipulation ofstarch biosynthetic genes is difficult to achieve because of thecomplexity of starch synthesis in endosperm. On the other hand,thus far there have been only a few reports regarding transcrip-tional regulation of starch synthesis in cereals. Among them, onebarley WRKY transcription factor, SUSIBA2, and three maizetranscription factors, ZmNAC36, ZmbZIP91, and ZmEREB156,are thought to regulate the expression of starch-synthetic genesbased on transcriptional activation and/or EMSA assays of theirpromoters (22–25). Recently, a study in rice showed that OsbZIP58,the closest homologous protein of maize O2, regulates the ex-pression of multiple genes in the starch biosynthetic pathways, andits null mutants cause abnormal seed morphology with decreasedamounts of seed starch (26). O2 and PBF control kernel nutri-tional quality by regulating the synthesis of storage zein proteins.Consistent with the temporal pattern of zein gene expression, theexpression of the major starch-synthetic genes and enzymatic ac-tivities increases sharply and later decreases gradually between 10and 35 days after pollination (DAP) (8, 9, 27), suggesting that thetranscriptional regulation of storage protein and starch synthesisin endosperm may be coordinated temporally by some commonfactors. Here, we show that O2 and PBF also regulate the endo-sperm starch synthesis and, in turn, kernel weight.

ResultsReduction of Seed Weight and Starch Content in PbfRNAi and o2Mutants. The PbfRNAi and o2 mutants and their double mutantPbfRNAi;o2mutant are all in the W64A background (28). They allexhibit an opaque and soft endosperm phenotype. In addition,PbfRNAi;o2 also produces mildly shrunken kernels (Fig. 1A).Measurement of 1,000-kernel weight (KW) and test weight (TW)showed that PbfRNAi alone had less effect than o2. In thePbfRNAi mutant the KW was not overtly altered, but the TW wasreduced by 9%, whereas in the o2 mutant the KW and TW werereduced by 20% and 13%, respectively. Moreover, in the PbfRNAi;o2double mutant the two yield parameters were reduced by 43%and 23%, respectively (Fig. 1B). These results demonstrate thatPBF and O2 affect kernel-weight traits additively.To investigate the reduction in kernel weight further, we used

SEM to examine starch granules in PbfRNAi, W64Ao2, andPbfRNAi;o2 mutants, which became gradually downsized in theperipheral and central regions of endosperm as compared withnormal genotypes (NG) (Fig. 1C), similar to the effects seen insome starch biosynthetic mutants (e.g., dull1) (20). We thenquantitatively measured the starch and soluble sugars content inmature dry mutant seeds (Fig. 1D). Compared with NG, whichcontained around 65 mg starch per 100 mg dry seed flour, thestarch content was reduced by ∼5% and 11% in PbfRNAi andW64Ao2, respectively. In PbfRNAi;o2 the starch content drop-ped more severely, by 25%, significantly lower than in the singlemutants. In contrast to the reduction in starch, the levels of theirsoluble sugars increased in these mutants by 58%, 55%, and86%, respectively, compared with NG, which contained 2.8 mgsoluble sugars per 100 mg dry seed flour. This pattern of inverse

Fig. 1. Kernel phenotypes and carbohydrate determination in NG (W64A), PbfRNAi, o2, and PbfRNAi;o2 seeds. (A) Ear (Upper) and kernel (Lower) phenotypes ofW64A, PbfRNAi, W64Ao2, and PbfRNAi;o2 seeds. e, embryo; en, endosperm. (Scale bars, 1 mm.) (B) The KW and TW ofW64A, PbfRNAi, W64Ao2, and PbfRNAi;o2seeds. Error bars represent the SD of three independent replicates. (C) SEM images of starch granules in the mature kernels. (Top) Kernel transverse sections.(Middle) SEM images of the kernel peripheral regions indicated by the white-outlined boxes in the top panel; (Lower) SEM images of the kernel central regionsindicated by the black-outlined boxes in the top panel. (Scale bars, 10 μm.) (D) Carbohydrate contents in the mature kernels. Error bars represent the SD of fiveindependent replicates. (E) Dosage of O2 in W64A, W64Ao2, and their reciprocal crosses. (F) Dosage effect of O2 on KW and starch content. Data represent themean ± SD of three independent replicates. In B, D, and F, asterisks indicate a significant difference from W64A (Student’s t test, P < 0.05).

Zhang et al. PNAS | September 27, 2016 | vol. 113 | no. 39 | 10843

AGRICU

LTURA

LSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

021

accumulation of starch and soluble sugars in the three mutantswas reminiscent of the patterns seen in starch biosynthetic mu-tants (20), implying that PbfRNAi and o2 probably decreasekernel weight by affecting starch synthesis.Moreover, O2 appears to have a dosage effect on the starch

content and kernel weight. Taking advantage of triploid endo-sperm, we were able to create seeds with two dosages or onedosage of O2 by reciprocally crossing W64A and W64Ao2 (Fig.1E). Apparently, the starch contents of kernels with two dosages,one dosage, and zero dosage of O2 were decreased gradually by2%, 4%, and 12%, respectively, compared with NG with threedosages (Fig. 1F). Consequently, the KWs were accordingly de-creased by 4%, 13%, and 28%, respectively (Fig. 1F). This resultdiffers from the regulation of storage protein transcription, inwhich the opaque phenotype is not dosage dependent but is re-cessive. (The seeds used in the experiment shown in Fig. 1F werepropagated in 2014, and those used in the experiment shown inFig. 1 A–D were grown in 2015.) This observation was consistentwith a previous report that starch synthetic enzymes also have adosage effect on the starch content in maize endosperm (29).

Down-Regulation of PPDKs, SSIII, SSIIa, and SBEI in PbfRNAi and o2Mutants. To investigate the downstream gene network that PBFand O2 cooperatively regulate in the developing endosperm, en-dosperms of NG, PbfRNAi, o2, and PbfRNAi;o2 seeds were har-vested at 16 DAP, and transcriptome analysis was performed byRNA-sequencing (RNA-seq) as described in Materials and Meth-ods. About 82% of the raw reads from each sample were mappedto the annotated gene-coding regions (Table S1). Based on theglobal FPKM-expressing values, principal component and hierar-chical cluster analysis showed that the three biological replicates ofeach sample were clustered into one group (Fig. S1 A and B). Onaverage 21,788 expressed genes were detected for each sample(Fig. S1C), consistent with the previous transcriptome analysis ofdeveloping endosperms (9, 28). A total of 1,296 up-regulated and1,513 down-regulated genes were detected on average in the threemutants compared with NG (Fig. S1D and Dataset S1). Further-more, carbohydrate- and amino acid metabolism-related KyotoEncyclopedia of Genes and Genomes (KEGG) pathways werepreferentially enriched in the differentially expressed genes of o2and PbfRNAi;o2 mutants (Dataset S2). For example, lysine deg-radation, which is related to the nutritional quality and is underregulation by O2 (30), was significantly enriched in the down-reg-ulated genes of the o2 and PbfRNAi;o2 mutants. The Gene On-tology (GO) terms associated with ribosome, translation, proteinfolding, and so forth were also enriched in the differentiallyexpressed genes in the three mutants, especially in PbfRNAi;o2(Dataset S2), consistent with the recent study of the O2 regulatorynetwork (31). Among them, the nutrient reservoir activity (GO:0045735), mainly containing zein genes, was significantly enrichedin the down-regulated genes of all of the three mutants. In general,these results indicate that sugar and protein metabolism were sig-nificantly affected in the three mutants.To examine whether starch reduction in the three mutants is di-

rectly related to the changes in the expression of genes involved instarch synthesis, we analyzed the transcript levels of starch bio-synthetic pathway genes preferentially expressed in the filling stage ofendosperm and PPDK as a component of starch synthetic complexes(9, 14) and also the transcript levels of oxidative pentose phosphatepathway genes closely related with starch synthesis (Tables S2 and S3)(18, 19). The most recent study of the o2 transcriptome showed thatO2 regulated the expression of endosperm PPDK1 and PPDK2 (31).More specifically, we found that their expression was mildly down-regulated in PbfRNAi but was significantly reduced in o2 and in thedouble mutant PbfRNAi;o2 compared with NG (Fig. 2A). In addition,the expression of 11 starch biosynthetic genes, SSIII, SSI, Su1, andSSIIa was down-regulated in o2 and/or PbfRNAi;o2, whereas Su1 andZpu1 levels were up-regulated in PbfRNAi (all based on t test

analysis) (Fig. 2A). To confirm the RNA-seq data, we performedquantitative RT-PCR to analyze the expression of these genes in-dividually in the three mutants and found overall correlation betweenthe two datasets (Fig. 2B).Consistent with the RNA transcript levels, the protein levels of

PPDK, SSIII, and SSIIa in 18-DAP endosperms were reduced inthe o2 mutant and were decreased further in the double mutant(Fig. 2 B and C). Although the RNA transcript level of SBEI wasnot significantly decreased in the o2 mutant, its protein level waslower than those in NG and PbfRNAi. Like PPDK, SSIII, andSSIIa, its protein content was most severely affected in thePbfRNAi;o2 double mutant (Fig. 2C). Furthermore, immuno-blotting analysis of starch synthetic enzymes in 24-DAP and

Fig. 2. Expression of 11 starch-synthetic genes and two PPDKs in the NG,PbfRNAi, o2, and PbfRNAi;o2 seeds. (A) Relative expression of these genes in themutants relative to NG based on the RNA-seq data in Tables S2 and S3. (B) Rel-ative expressions of these genes in the mutants relative to NG based on quanti-tative RT-PCR. The data shown in A and B are from three biological replicates persample and are presented as ± SD. Asterisks indicate a significant difference fromW64A (Student’s t test, P < 0.05). (C) Protein accumulations of these genes in theNG, PbfRNAi, o2, and PbfRNAi;o2 seeds. Non-zein proteins from 18-DAP endo-sperms were separated by SDS/PAGE and subjected to immunoblot analysis. Eachlane was loaded with 12 μg total protein. ACTIN and a piece of replicated gelstained with Coomassie Brilliant Blue (CBB) were used as loading controls.

10844 | www.pnas.org/cgi/doi/10.1073/pnas.1613721113 Zhang et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

021

32-DAP endosperms demonstrated that protein levels ofPPDKs, SSIII, SSIIa, and SBEI were more reduced in the doublemutant than in NG or in the two single mutants (Fig. S2). Theseresults indicate a strong correlation between the reduction ofstarch synthesis in the o2 and PbfRNAi;o2mutants and the down-regulation of genes encoding starch biosynthetic enzymes.

Activation of the PPDK and SSIII Promoters by PBF and O2. To in-vestigate whether this correlation can be explained by directtransactivation of these starch-synthetic genes with PBF and O2,the dual luciferase transcriptional activity assay was used. In thissystem, the reporter construct contains two luciferase cassettes,one being the Renilla LUC (REN) reporter gene driven by the35S promoter (35S-REN) that is used as an internal control, andthe other being the firefly luciferase (LUC) driven by the targetgene promoter (Fig. 3A). Here, the starch synthetic gene pro-moters were fused with LUC, giving rise to 13 reporter con-structs. The effectors were 35S-O2 and 35S-PBF (Fig. 3A andFig. S3). Consistent with previous reports that O2 regulatesPPDK1 and PPDK2 (31, 32), coexpression of Pppdk1-LUC orPppdk2-LUC with 35S-O2 resulted in a 17- or 44-fold increase inLUC activity, respectively, compared with the control. Although35S-PBF alone had limited activation on the two promoters, thecombination of 35S-PBF and 35S-O2 caused 50- and 80-foldincreases in LUC signal, indicating that PBF and O2 provideadditive activation of the PPDK1 and PPDK2 promoters (Fig.3A). For the SSIII promoter, both 35S-PBF and 35S-O2 aloneexerted strong activation on its transcription, yielding 15- and 35-fold enhancement in LUC activity, respectively. When PSSIII-LUCwas coexpressed with 35S-PBF and 35S-O2, the activation wasapparently expressed as an additive action, resulting in a 64-fold

increase in LUC activity (Fig. 3A). However, promoters of SSIIaand SBEI were not activated by PBF and O2 (Fig. 3A). Similarresults were seen for the other eight starch synthetic gene pro-moters, none of which was activated by PBF and O2 (Fig. S3).

PBF and O2 Bind the P and O2 Boxes in the PPDK and SSIII Promoters.PBF and O2 belong to the DOF and bZIP transcription factorfamilies, which recognize a motif containing the AAAG (P box)and ACGT (O2 box) core element, respectively. We found thatboth the P and O2 box are present in the first 500 bp of thePPDK1, PPDK2, and SSIII promoters (Fig. S4). Interestingly,the core elements of the P and O2 boxes were side by side in thePPDK1 and PPDK2 promoters (Fig. S4). In the SSIII promoter,the two elements were separated by 25 bp, similar to the separa-tion in the 22-kDa α-zein promoters (6). EMSA was used to testwhether the two boxes are specifically recognized by PBF and O2.Because the P and O2 boxes are too close to one another in thetwo PPDK promoters, 55-bp (PPDK1) and 51-bp (PPDK2) oligo-nucleotides spanning the two boxes were used to examine thebinding affinity. Binding of His-O2 and His-PBF fusion proteins tothe probes could be visualized as retarded bands in the gel. Theresults reveal that the two probes are bound by O2 and PBF (Fig.3B). It appears that O2 and PBF have a stronger affinity for thePPDK1 probe than for the PPDK2 probe. When the P and O2boxes in the two probes were mutated, the corresponding retardedbands were abolished. The binding specificity was verified by theaddition of unlabeled intact probes in the reaction, resulting in agradual loss of all retarded bands (Fig. 3B). Similar results wereobtained for the two probes from the SSIII promoter, eachbearing only the P or the O2 box, showing that they were specif-ically bound by PBF and O2, respectively (Fig. 3C). This line of

Fig. 3. Transactivation of the PPDK1, PPDK2, SSIII, SSIIa, and SBEI promoters by PBF and O2. (A) Dual luciferase assays of the promoters of PPDK1, PPDK2, SSIII,SSIIa, and SBEI. (Upper) Schematic diagrams of the effector and reporter constructs. REN, Renilla luciferase; LUC, firefly luciferase; ter, terminator. (Lower) Theratios of LUC to REN activity in Arabidopsis protoplasts cotransformed with the reporter and effector. The numbers in the chart represent the fold-increase ofactivation compared with the negative control (empty effector). Data represent means ± SD of three independent replicates. (B and C) EMSA of PBF and O2 withprobes containing the ACGT and AAAG elements in the promoters of PPDK1 and PPDK2 (B), and SSIII (C). The positions of probes in the promoters are indicated inFig. S3. Comp, competing probes unlabeled with biotin; Mp, mutated probes with a deletion in the corresponding ACGT or AAAG core element.

Zhang et al. PNAS | September 27, 2016 | vol. 113 | no. 39 | 10845

AGRICU

LTURA

LSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

021

results demonstrates that PBF and O2 regulate PPDKs and SSIIIdirectly by recognizing their P and O2 boxes, respectively.

The Pentose Phosphate Pathway Is Affected in PbfRNAi and o2Mutants. Because the pentose phosphate pathway is also impor-tant for starch synthesis in endosperm, we investigated the ex-pression of genes in this pathway (Tables S2 and S3). Among thesegenes, the transcript levels of three glucose-6-phosphate dehy-drogenases, one 6-phosphogluconolactonase, three 6-phospho-gluconate dehydrogenases, one ribose 5-phosphate isomerase, tworibulose-phosphate 3-epimerases, one transketolase, and twotransaldolases were significantly down-regulated based on t testanalysis in PbfRNAi, o2, and/or PbfRNAi;o2 mutants comparedwith NG (Fig. S5A). We also conducted quantitative RT-PCR tovalidate 11 of the 13 down-regulated genes in the three mutantsand found overall similar patterns, as revealed by the RNA-seqdata (Fig. S5B), indicating that decrease in starch in PbfRNAi ando2 might be caused, at least in part, by the down-regulation of thepentose phosphate pathway. To investigate whether PBF and O2could regulate the genes in this pathway directly, promoters of 12down-regulated genes were amplified for the transcriptional acti-vation assay. It revealed that several promoters could be trans-activated by PBF or O2, but none of them additively (Fig. S6).

DiscussionStorage protein and starch synthesis are temporally coordinatedduring maize endosperm development through a transcriptionalregulatory network linking the two different synthetic processes (8, 9,27). This coordination is now supported by genetic and molecularevidence. PBF and O2 are two endosperm-specific transcriptionfactors that are specifically expressed from 8 to 10 DAP. The twotranscription factors control 89% of zein transcription (28). ZeinmRNAs then are translated on the polyribosomes of the roughendoplasmic reticulum (RER), and the resulting proteins aretranslocated into the lumen of the RER, where they are assembledinto protein bodies (33). In contrast to storage proteins, which haveno known enzymatic function, starch biosynthesis is far more com-plex because the respective genes encode enzymes that are orga-nized in interacting protein complexes and assemblies to producemature starch granules. In fact, starch is synthesized by a series ofcombined reactions of a fine-tuned enzyme complex and is assem-bled into semicrystalline starch granules as end products. In total, 11starch-synthetic genes in maize endosperm have been cloned (7),although some of their functions are not fully understood. Becauseof the lack of null mutants of PPDKs in maize, their functions couldnot clearly be elucidated. Based on their activity in catalyzing re-versible conversion between PYR and PEP, it has been hypothesizedthat they are involved in the synthesis and regulatory balancing ofamino acids, protein, starch, and lipids. In addition, PPDKs also havebeen shown to be components of the starch synthetic complex and tomodulate starch synthesis by interacting with other enzymes in thecomplex. In fact, RNA-seq transcriptome analysis of immature en-dosperm reveals that the expression of 13 genes increases sharplybetween 8 and 10 DAP, paralleling the temporal expression patternof storage protein zein genes (9).Previous studies showed that the level of zein proteins was re-

duced markedly in the o2 mutant and was reduced further in thedouble mutant PbfRNAi;o2. When measuring kernel weight, wefound that the KW was reduced by 20% and 43% in the o2 andPbfRNAi;o2 mutants, respectively. The TW for the two mutantsalso was reduced, by 13% and 23%, respectively. However, it isnot likely that the reduction of zein proteins contributes signifi-cantly to the loss of kernel weight, because in total zein proteinsaccount for only about 6% of the dry seed flour in NG. Moreover,the decrease in zein protein accumulation would result in acompensatory increase in non-zein protein synthesis, whichmaintains the total protein content in a relatively constant range atmaturity (28). In contrast, starch constitutes around 70% of the

dry seed mass and therefore is the most preponderant substance indetermining the kernel weight. Indeed, SEM showed that thestarch granules were progressively smaller in the PbfRNAi, o2, andPbfRNAi;o2 mutants, correlating well with the KW value. Furtherchemical quantification showed that the starch contents in thethree mutants were decreased by 5%, 11%, and 25% (Fig. 1D).We noted that the parameters of KW, TW, and starch contentwere affected far more severely in the PbfRNAi;o2 double mutantthan in the o2 mutant; however, PbfRNAi alone had limited in-fluence, indicating that O2 is the main factor and acts additivelywith PBF, affecting starch and protein composition in parallel.Still, elevated expression of PBF, when introgressed from teosinte,is associated with increased kernel weight in inbred W22 (34).O2 regulates many other genes in addition to zein genes.

Likewise, PBF could participate in a more extensive regulatorynetwork because of the pleiotropic phenotype of seeds, which isnot yet completely understood. Previous research indicated thatO2 probably is involved in the regulation of starch synthesis (35–39). Here, we found that the starch content and kernel weightwere positively correlated to the O2 dosage (Fig. 1 E and F).However, it had been unclear whether O2 directly regulatedsome, if not all, the starch-synthetic genes. Moreover, therepreviously was no genetic evidence for the regulatory role of PBFin this pathway. Taking advantage of RNA-seq transcriptomeanalysis, we found that more than 1,000 genes were down-reg-ulated or up-regulated in PbfRNAi and o2 mutants and that thedifferentially expressed genes were mainly enriched in the sugarand protein metabolism pathways. Among the down-regulatedgenes, the RNA transcript and/or protein levels of endospermPPDKs and three starch biosynthetic genes, SSIII, SSIIa, andSBEI, were reduced in developing endosperms in PbfRNAi ando2 background. Using the dual luciferase assay and EMSA, wethen verified that PBF and O2 could cooperatively activate thePPDK1, PPDK2, and SSIII promoters by recognizing the re-spective P and O2 boxes. However, the two transcription factorsfailed to activate the SSIIa and SBEI promoters (Fig. 3). Onepossible explanation for this apparent discrepancy is that o2 andPbfRNAi may affect the levels of some other transcription factor(s)or other type of mRNA regulatory factor(s) and thereby affect

Fig. 4. A proposed model for PBF and O2 controlling kernel nutritionalquality and yield in maize.

10846 | www.pnas.org/cgi/doi/10.1073/pnas.1613721113 Zhang et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

021

the transcript and protein levels of SSIIa and SBEI. This studyindicates that PbfRNAi and o2 mutations are able to cause directand indirect defects in starch synthesis.In starch synthesis, some of the enzymes are assembled into a

well-coordinated machinery complex in which the physical associ-ation of SSIII with other enzymes, including PPDK, has beenproposed to play a critical role in regulatory interactions. AlthoughPPDKs themselves are not likely to be involved directly in anycatalytic step of starch synthesis, it has been proposed that theyfunction as a modulator of the complex. Coimmunoprecipitationassays also indicated that PPDK, SSIII, SSIIa, and SBEI are as-sociated with one other (Fig. S7), as is consistent with previousreports (13, 14). Therefore we modeled the impact of PBF and O2on the regulation of kernel nutritional quality and yield, as shownin Fig. 4. The reduction or absence of both transcription factors canraise the nutritional quality of maize but at the same time reducestarch synthesis and therefore yield. Zein proteins are devoid oflysine and tryptophan. Mutations in Pbf and O2 enhance the nu-tritional value because of the reduction of zein proteins and thecompensatory increase of non-zein proteins. On the other hand,PBF and O2 directly regulate PPDKs and the starch synthetic geneSSIII and also indirectly regulate SSIIa and SBEI. In PbfRNAi ando2, the down-regulation of SSIII, SSIIa, and SBEI might directlycause decreased synthesis of starch. In addition, because PPDKsand SSIII are components of the enzyme complex and are asso-ciated with other enzymes, their reduction probably would affectthe efficiency with which the components of the complex are as-sembled into an integral structure or the stability of the complex in

o2 or PbfRNAi;o2mutants (Fig. 2C) and consequently would resultin reduced starch synthesis. Furthermore, O2 and PBF regulatemore than 1,000 genes, at least indirectly, and their mutations havea pleiotropic effect on endosperm development. It is very likelythat other affected genes and pathways related to central metab-olism [e.g., the pentose phosphate pathway (Fig. S5)] contribute tothe decrease in starch (Fig. 4) (19). Given these results, the use ofRNAi specifically against the expression of zein genes in the pro-duction of Quality Protein Maize (QPM) may be more effectivethan the classic o2 mutation, not only because α-zein RNAi isdominant but also because it does not reduce yield, as results, atleast in part, from the starch biosynthesis defect demonstrated here(40, 41).

Methods and MaterialsGenetic materials and molecular procedures, such as measurement of solu-ble sugars and starch in the mature dry seeds, RNA-seq analysis, quantitativeRT-PCR, analysis of the relative expression of the starch-synthetic genes andPPDKs, generation of antibodies, immunoblot analysis, coimmunoprecipi-tation and the dual luciferase assays, and EMSA are described in SI Materialsand Methods. Primers are listed in Table S4.

ACKNOWLEDGMENTS. We thank Dr. Baichen Wang from the Institute ofBotany, Chinese Academy of Sciences (CAS) for providing the PPDK antibodyand Dr. Peng Zhang from the Institute of Plant Physiology and Ecosystem,CAS, for providing the GBSSI antibody. This research was supported by Na-tional Natural Science Foundation of China Grants 91335109, 31371630, and31422040 (to Y.W.), CAS Grant XDA08020107 (to Y.W.), and a Chinese Thou-sand Talents Program Grant (to Y.W.). J.M. holds the Selman A. WaksmanChair in Molecular Genetics.

1. Mertz ET, Bates LS, Nelson OE (1964) Mutant gene that changes protein compositionand increases lysine content of maize endosperm. Science 145(3629):279–280.

2. Schmidt RJ, Burr FA, Burr B (1987) Transposon tagging and molecular analysis of themaize regulatory locus opaque-2. Science 238(4829):960–963.

3. Schmidt RJ, Burr FA, Aukerman MJ, Burr B (1990) Maize regulatory gene opaque-2encodes a protein with a “leucine-zipper” motif that binds to zein DNA. Proc NatlAcad Sci USA 87(1):46–50.

4. Schmidt RJ, Ketudat M, Aukerman MJ, Hoschek G (1992) Opaque-2 is a transcriptionalactivator that recognizes a specific target site in 22-kD zein genes. Plant Cell 4(6):689–700.

5. Vicente-Carbajosa J, Moose SP, Parsons RL, Schmidt RJ (1997) A maize zinc-finger proteinbinds the prolamin box in zein gene promoters and interacts with the basic leucinezipper transcriptional activator Opaque2. Proc Natl Acad Sci USA 94(14):7685–7690.

6. Wu Y, Messing J (2012) Rapid divergence of prolamin gene promoters of maize aftergene amplification and dispersal. Genetics 192(2):507–519.

7. Hannah LC (2005) Starch synthesis in the maize endosperm. Maydica 50(3-4):497–506.8. Keeling PL, Myers AM (2010) Biochemistry and genetics of starch synthesis. Annu Rev

Food Sci Technol 1:271–303.9. Chen J, et al. (2014) Dynamic transcriptome landscape of maize embryo and endo-

sperm development. Plant Physiol 166(1):252–264.10. Gao M, Wanat J, Stinard PS, James MG, Myers AM (1998) Characterization of dull1, a

maize gene coding for a novel starch synthase. Plant Cell 10(3):399–412.11. Zhang X, et al. (2004) Molecular characterization demonstrates that the Zea mays gene

sugary2 codes for the starch synthase isoform SSIIa. Plant Mol Biol 54(6):865–879.12. Tetlow IJ, et al. (2004) Protein phosphorylation in amyloplasts regulates starch

branching enzyme activity and protein-protein interactions. Plant Cell 16(3):694–708.13. Hennen-Bierwagen TA, et al. (2008) Starch biosynthetic enzymes from developing maize

endosperm associate in multisubunit complexes. Plant Physiol 146(4):1892–1908.14. Hennen-Bierwagen TA, et al. (2009) Proteins from multiple metabolic pathways asso-

ciate with starch biosynthetic enzymes in highmolecular weight complexes: A model forregulation of carbon allocation in maize amyloplasts. Plant Physiol 149(3):1541–1559.

15. Hatch MD, Slack CR (1968) A new enzyme for the interconversion of pyruvate andphosphopyruvate and its role in the C4 dicarboxylic acid pathway of photosynthesis.Biochem J 106(1):141–146.

16. Meyer AO, Kelly GJ, Latzko E (1982) Pyruvate orthophosphate dikinase from theimmature grains of cereal grasses. Plant Physiol 69(1):7–10.

17. Aoyagi K, Bassham JA (1984) Pyruvate orthophosphate dikinase mRNA organ speci-ficity in wheat and maize. Plant Physiol 76(1):278–280.

18. Hannah LC, James M (2008) The complexities of starch biosynthesis in cereal endo-sperms. Curr Opin Biotechnol 19(2):160–165.

19. Spielbauer G, et al. (2013) Chloroplast-localized 6-phosphogluconate dehydrogenaseis critical for maize endosperm starch accumulation. J Exp Bot 64(8):2231–2242.

20. Creech RG (1965) Genetic control of carbohydrate synthesis in maize endosperm.Genetics 52(6):1175–1186.

21. Hannah LC, et al. (2012) A shrunken-2 transgene increases maize yield by acting in ma-ternal tissues to increase the frequency of seed development. Plant Cell 24(6):2352–2363.

22. Sun C, et al. (2003) A novel WRKY transcription factor, SUSIBA2, participates in sugarsignaling in barley by binding to the sugar-responsive elements of the iso1 promoter.Plant Cell 15(9):2076–2092.

23. Zhang J, et al. (2014) Novel role of ZmaNAC36 in co-expression of starch syntheticgenes in maize endosperm. Plant Mol Biol 84(3):359–369.

24. Chen J, et al. (2016) ZmbZIP91 regulates expression of starch synthesis-related genesby binding to ACTCAT elements in their promoters. J Exp Bot 67(5):1327–1338.

25. Huang H, et al. (2016) Sucrose and ABA regulate starch biosynthesis in maize througha novel transcription factor, ZmEREB156. Sci Rep 6:27590.

26. Wang JC, Xu H, Zhu Y, Liu QQ, Cai XL (2013) OsbZIP58, a basic leucine zipper transcriptionfactor, regulates starch biosynthesis in rice endosperm. J Exp Bot 64(11):3453–3466.

27. Méchin V, Thévenot C, Le Guilloux M, Prioul JL, Damerval C (2007) Developmentalanalysis of maize endosperm proteome suggests a pivotal role for pyruvate ortho-phosphate dikinase. Plant Physiol 143(3):1203–1219.

28. Zhang Z, Yang J, Wu Y (2015) Transcriptional regulation of zein gene expression inmaize through the additive and synergistic action of opaque2, prolamine-box bindingfactor, and o2 heterodimerizing proteins. Plant Cell 27(4):1162–1172.

29. Singletary GW, Banisadr R, Keeling PL (1997) Influence of gene dosage on carbohy-drate synthesis and enzymatic activities in endosperm of starch-deficient mutants ofmaize. Plant Physiol 113(1):293–304.

30. Kemper EL, et al. (1999) The role of opaque2 in the control of lysine-degrading ac-tivities in developing maize endosperm. Plant Cell 11(10):1981–1994.

31. Li C, et al. (2015) Genome-wide characterization of cis-acting DNA targets reveals thetranscriptional regulatory framework of opaque2 in maize. Plant Cell 27(3):532–545.

32. Sadimantara GR, Abe T, Suzuki J, Hirano H, Sasahara T (1996) Characterization and partialamino acid sequence of a high molecular weight protein from rice seed endosperm:Homology to pyruvate orthophosphate dikinase. J Plant Physiol 149(3-4):285–289.

33. Larkins BA, Hurkman WJ (1978) Synthesis and deposition of zein in protein bodies ofmaize endosperm. Plant Physiol 62(2):256–263.

34. Lang Z, et al. (2014) Defining the Role of prolamin-box binding factor1 Gene DuringMaize Domestication. J Hered 105(4):576–582.

35. Wang X, Larkins BA (2001) Genetic analysis of amino acid accumulation in opaque-2maize endosperm. Plant Physiol 125(4):1766–1777.

36. Giroux MJ, Boyer C, Feix G, Hannah LC (1994) Coordinated transcriptional regulationof storage product genes in the maize endosperm. Plant Physiol 106(2):713–722.

37. Doehlert DC, Kuo TM (1994) Gene-expression in developing kernels of some endo-sperm mutants of maize. Plant Cell Physiol 35(3):411–418.

38. Hasjim J, Srichuwong S, Scott MP, Jane JL (2009) Kernel composition, starch structure,and enzyme digestibility of opaque-2 maize and quality protein maize. J Agric FoodChem 57(5):2049–2055.

39. Jia M, et al. (2013) Identification and characterization of lysine-rich proteins andstarch biosynthesis genes in the opaque2 mutant by transcriptional and proteomicanalysis. BMC Plant Biol 13:60.

40. Segal G, Song R, Messing J (2003) A new opaque variant of maize by a single domi-nant RNA-interference-inducing transgene. Genetics 165(1):387–397.

41. Wu Y, Messing J (2011) Novel genetic selection system for quantitative trait loci ofquality protein maize. Genetics 188(4):1019–1022.

42. Wu Y, Messing J (2012) RNA interference can rebalance the nitrogen sink of maizeseeds without losing hard endosperm. PLoS One 7(2):e32850.

43. Yoo SD, Cho YH, Sheen J (2007) Arabidopsis mesophyll protoplasts: A versatile cellsystem for transient gene expression analysis. Nat Protoc 2(7):1565–1572.

Zhang et al. PNAS | September 27, 2016 | vol. 113 | no. 39 | 10847

AGRICU

LTURA

LSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

021