Control of cell proliferation, endoreduplication, cellsize, and cell death by the retinoblastoma-relatedpathway in maize endospermPaolo A. Sabellia,1, Yan Liua,2, Ricardo A. Dantea,3, Lucina E. Lizarragaa,4, Hong N. Nguyena, Sara W. Browna,John P. Klinglera,5, Jingjuan Yua,b, Evan LaBranta, Tracy M. Laytona, Max Feldmana,6, and Brian A. Larkinsa,1

aSchool of Plant Sciences, University of Arizona, Tucson, AZ 85721; and bState Key Laboratory for Agrobiotechnology, College of Biological Sciences,China Agricultural University, Beijing 100193, China

Contributed by Brian A. Larkins, March 18, 2013 (sent for review February 18, 2013)

The endosperm of cereal grains is one of themost valuable productsof modern agriculture. Cereal endosperm development comprisesdifferent phases characterized by mitotic cell proliferation, endor-eduplication, the accumulation of storage compounds, and pro-grammed cell death. Although manipulation of these processescould maximize grain yield, how they are regulated and integratedis poorly understood. We show that the Retinoblastoma-related(RBR) pathway controls key aspects of endosperm development inmaize. Down-regulation of RBR1 by RNAi resulted in up-regulationof RBR3-type genes, as well as the MINICHROMOSOME MAINTE-NANCE 2–7 gene family and PROLIFERATING CELL NUCLEAR ANTI-GEN, which encode essential DNA replication factors. Both themitotic and endoreduplication cell cycles were stimulated. Develop-ing transgenic endosperm contained 42–58% more cells and ∼70%more DNA than wild type, whereas there was a reduction in celland nuclear sizes. In addition, cell death was enhanced. The DNAcontent of mature endosperm increased 43% upon RBR1 down-regulation, whereas storage protein content and kernel weightwere essentially not affected. Down-regulation of both RBR1 andCYCLIN DEPENDENT KINASE A (CDKA);1 indicated that CDKA;1 isepistatic to RBR1 and controls endoreduplication through an RBR1-dependent pathway. However, the repressive activity of RBR1 ondownstream targets was independent from CDKA;1, suggestingdiversification of RBR1 activities. Furthermore, RBR1 negativelyregulated CDK activity, suggesting the presence of a feedbackloop. These results indicate that the RBR1 pathway plays a majorrole in regulation of different processes during maize endo-sperm development and suggest the presence of tissue/organ-level regulation of endosperm/seed homeostasis.

seed development | endocycle

The seed endosperm is a triploid tissue resulting from the fusionof one haploid sperm nucleus with the diploid central cell nu-

cleus within the female gametophyte. Development of the endo-sperm in flowering plants is characterized by acytokinetic mitosesof the primary endosperm nucleus, resulting in a syncytium, cel-lularization of syncytial nuclear domains, and cell proliferationthrough mitotic activity that is coupled to cell division (1, 2). Ad-ditionally, in the Poaceae (grass) family, the endosperm undergoesa rapid growth phase that coincides with accumulation of storagecompounds, such as starch and storage proteins, during a special-ized type of cell cycle known as endoreduplication. Endor-eduplication is characterized by one or more rounds of DNAsynthesis in the absence of mitosis, resulting in polyploid cells (3–5). Endoreduplication is highly correlated with cell size in manyplant and animal tissues, but its role in endosperm developmenthas not been established. Upon completion of endoreduplicationand storage metabolite synthesis, cereal endosperm cells undergoprogrammed cell death (PCD), resulting in extensive DNA deg-radation (5, 6). In maize (Zea mays L.), endosperm cells transitionfrom a mitotic to an endoreduplication cell cycle at approximately8 d after pollination (DAP), and PCD becomes obvious at ap-proximately 16 DAP.Manipulation of cell cycle regulation and cell

death during endosperm development could potentially increasegrain yield and perhaps improve its quality, yet a detailed un-derstanding of the factors underlying control and integration ofthese processes is lacking.Cylin-Dependent Kinase (CDK) and Retinoblastoma-Related

(RBR) genes play fundamental roles in cell cycle regulation in allhigher eukaryotes (7, 8). They control the onset of DNA replica-tion but also regulate other phases of the cell cycle as well as cellcycle-independent processes (9, 10). RBR genes generally restrictcell cycle progression by preventing S-phase through inhibition ofgene expression programs that depend on adenovirus E2 pro-moter binding factor (E2F) transcription factors. In turn, RBRscan be inhibited through phosphorylation by CDK activity, whichstimulates downstream gene expression and cell cycle progression.Although these mechanisms are well known, it is not clear whetherRBR genes play a role in cell cycle regulation during cereal en-dosperm development or whether they function in a pathwaythat depends on specific CDKs. The cereal RBR gene family ismore complex than that in other plants. According to phylo-genetic, gene expression, and functional studies, at least twodistinct types of RBR genes, RBR1 and RBR3, have been identi-

Significance

Cereal endosperm is a key source of dietary calories and rawmaterials for countless manufactured goods. Understandinghow the cell cycle is regulated during endosperm developmentcould lead to increased crop yield. We show that a maizeRetinoblastoma-related gene, RBR1, plays a central role inregulating gene expression, endoreduplication, and the num-ber, size, and death of endosperm cells. RBR1 is geneticallycoupled to Cyclin Dependent Kinase A;1 in controlling endor-eduplication but not gene expression. Seeds down-regulatedfor RBR1 develop normally, which suggests higher-order con-trol mechanisms regulating endosperm development that aresuperimposed on cell cycle regulation.

Author contributions: P.A.S., Y.L., R.A.D., and B.A.L. designed research; P.A.S., Y.L., R.A.D.,L.E.L., H.N.N., S.W.B., J.P.K., J.Y., E.L., T.M.L., and M.F. performed research; P.A.S., Y.L., R.A.D.,H.N.N., J.P.K., and J.Y. analyzed data; and P.A.S., R.A.D., and B.A.L. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. E-mail: psabelli@ag.arizona.edu or blarkins2@unl.edu.

2Present address: Herman B. Wells Center for Pediatric Research, Indiana University Schoolof Medicine, Indianapolis, IN 46202.

3Present address: Embrapa Agricultural Informatics, Campinas, SP 13083-886, Brazil.4Present address: Pharmacology and Toxicology Department, College of Pharmacy, Uni-versity of Arizona, Tucson, AZ 85721.

5Present address: Plant Stress Genomics Research Center, King Abdullah University ofScience and Technology, Thuwal 23955-6900, Kingdom of Saudi Arabia.

6Present address: Donald Danforth Plant Science Center, St. Louis, MO 63132.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1304903110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1304903110 PNAS | Published online April 22, 2013 | E1827–E1836

PLANTBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

fied in grasses. Additionally, the maize genome possesses almostidentical paralogs to RBR1 and RBR3, which are termed RBR2and RBR4, respectively (9, 11–13). Studies of cell cycle regulationin tissue culture showed that, in contrast to the cell cycle-repressive activity of RBR1, RBR3 plays a positive role in E2F-dependent gene expression, DNA replication, and the regenerationof transformed plant cells (13), which is uncharacteristic fora member of a family of known cell cycle inhibitors. This situationis clearly more complex than in most dicots, such as Arabidopsisthaliana, which typically possess only a single RBR gene witha clear cell cycle-inhibitory function. Both the potential in-activation by phosphorylation of the maize RBR1 gene product(14) and an increase in its expression (11, 15) during endospermdevelopment have been reported. However, whether RBRs playany role in regulating the cell cycle, endoreduplication, or otheraspects of cereal endosperm development is unknown.How cereal RBR proteins are regulated by CDKs is also un-

clear. Although there is compelling evidence that A- and D-typecyclins form complexes with CDKs that target RBR proteins forinhibitory phosphorylation (7), the identity of the kinase moiety isless certain. Biochemical and genetic evidence indicate thatA-type CDKs may be responsible for this activity (16–18), but nophysical interaction between CDKA and RBR was found in acomprehensive interactome study of cell cycle proteins in Arabi-dopsis (19). Knowledge of the role of CDKs in the cell cycle ofmaize is rudimentary. At least two CDKAs have been identified(20), but their functions during endosperm development and inrelation to the RBR pathway are not clear, beyond the observa-tion that overexpression of a dominant-negative mutant form ofCDKA;1 reduces endoreduplication (21).Here we investigated the role of RBR1 during maize endosperm

development and its relationship with CDKA;1. We show thatRBR1 controls gene expression programs, CDK activity, themitotic cell cycle, endoreduplication, cell and nuclear sizes, andPCD. An RBR1-dependent pathway mediates the function ofCDKA;1 in controlling endoreduplication, but the expression ofRBR1-repressed genes is virtually independent from CDKA;1,suggesting the existence of distinct RBR1 activities with differentsusceptibilities to inhibition by CDKA;1. Despite the abovealterations in the pathways controlling the cell cycle and PCD,mutant RBR1 RNAi kernels displayed essentially normal growthparameters, suggesting the presence of compensatory mecha-nisms governing tissue or organ homeostasis.

ResultsGeneration of RBR1DS1 Transgenic Endosperm and RBR-SpecificAntibodies. A transgenic maize line that down-regulated RBR1in developing endosperm, termed RBR1DS1, was obtained byembryo transformation with a DNA construct producing a dou-ble-stranded hairpin RNA under control of the maize 27-kDaγ-zein promoter (Fig. 1A). This well-characterized promoter ishighly active throughout the endosperm from approximately 8DAP (Fig. S1A), which typically coincides with the onset of theendoreduplication cell cycle (21, 22). Because of the high degreeof nucleotide sequence identity (∼93%) between the region ofRBR1 targeted by RNAi and the corresponding domain of theclosely related RBR2 gene, this construct could have conceivablysimultaneously down-regulated both RBR1 and RBR2. However,collateral down-regulation of the more distantly related RBR3-type genes (∼63% identity with RBR1), RBR3 and RBR4, wouldbe unlikely.Selected domains of RBR1 and RBR3 proteins were expressed

as GST fusions in Escherichia coli (Fig. 1 A and B) and the corre-sponding antisera generated. Two affinity-purified antibodies wereobtained against the N-terminal (GST-RBR1N) and large pocket(GST-RBR1P) domains of RBR1 and termed α-RBR1N andα-RBR1P, respectively. Another antibody, α-RBR3N, was raisedagainst the N-terminal domain of RBR3 (GST-RBR3N). Whentested against purified recombinant proteins, α-RBR3N detecteda 36-kDa band corresponding to GST-RBR3N but did not reactwith GST-RBR1N. The two RBR1 antibodies identified the re-spective GST-RBR1N and GST-RBR1P antigens but did not reactwith the corresponding domains of RBR3. No antibody detectedthe GST moiety (Fig. 1C). Thus, the antibodies were specific forRBR1 or RBR3 proteins. We next tested whether the two RBR1antibodies could identify RBR1 protein in endosperm extracts (Fig.1D). α-RBR1P detected a single band of ∼96 kDa, which is thepredicted molecular mass of RBR1. The α-RBR1N antibody alsodetected a band of that size, as well as several backgrounds bands,indicating that it also identified endogenous RBR1.

Up-Regulation of E2F Targets upon RBR1 Down-Regulation. Tran-script levels were measured by real-time RT-PCR in RNAextracted from wild-type and RBR1DS1 endosperms at 10, 13, 16,19, and 22 DAP isolated from ears segregating for the transgene(Fig. 2A). This period spans the phase of extensive endore-duplication. Compared with wild type, RBR1 expression in

Fig. 1. Generation of RBR1DS1 construct and RBR1/3-specific antibodies. (A) Schematic alignment ofmaizeRBR amino acid sequences relative to that of humanRB, which displays blocks of related sequences asthick bars highlighted by different shading. Theconserved “A” and “B” pocket domains are shownasthick boxes. The sense and antisense regions selectedto engineer the RBR1DS1 construct are indicatedby green and red arrows, respectively. The regionsselected for raising antibodies are indicated by bro-ken boxes. (B) SDS/PAGE of recombinant proteins.Affinity-purified polypeptides corresponding toGST-RBR1P, GST-RBR3N, GST, GST-RBR3P, and GST-RBR1N are indicated by arrowheads. (C) Affinity-purified antibodies recognize specific recombinantpolypeptides. Antibodies against GST-RBR3N (termedα-RBR3N), GST-RBR1N (α-RBR1N), and GST-RBR1P(α-RBR1P) specifically recognize the correspondingproteins. The ∼50-kDa band detected by α-RBR1Ppresumably represents an incomplete RBR1P poly-peptide. Each lane contained 10 ng of protein.Primary antibodies were used at a 1:500 dilution.(D) Identification of RBR1 protein in cellular extracts.The ∼96-kDa band corresponding to RBR1 (arrowhead) was identified in 9-DAP endosperm extracts using two independent antibodies targeted todifferent regions of the proteins: α-RBR1P and α-RBR1N. Fifty micrograms of protein extract was analyzed in each lane. Primary antibodies were used ata dilution of 1:250.

E1828 | www.pnas.org/cgi/doi/10.1073/pnas.1304903110 Sabelli et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

RBR1DS1 endosperm displayed a downward trend between 10DAP (0.8-fold) and 22 DAP (0.3-fold). RBR2 expression, whichfollows an upward trend during endosperm development similarto RBR1 (Fig. S1D) (11), was also down-regulated in the mutant(0.4- to 0.6-fold). In contrast, expression of RBR3-4 transcriptssteadily increased by as much as five- to sixfold, with a peak at 19DAP. By 22 DAP, the levels of RBR3-4 RNA declined, relativeto 19 DAP, but were still higher than in control (2.3- to 2.4-fold).The up-regulation of RBR3 upon down-regulation of RBR1confirmed previous findings, which revealed that RBR3 expressionis under negative regulation by RBR1 (11, 12). RBR4 is a recentlyidentified member of the RBR gene family (9) with unknownendosperm expression pattern. We thus determined that RBR4RNA is associated with the mitotic phase of endosperm de-velopment (i.e., before 11 DAP) (Fig. S1B) and is negativelyregulated by RBR1 (Fig. 2A), both of which are true of RBR3.Additionally, we measured transcripts of the MINICHROMO-

SOME (MCM) 2–7 gene family and the PROLIFERATINGCELL NUCLEAR ANTIGEN (PCNA) subunit of DNA poly-merase δ. Expression of these DNA replication factors is nega-tively regulated by RBR through inhibition of E2F in highereukaryotes (13, 23, 24), and they would be expected to be up-regulated upon RBR1 inactivation or loss in maize. The expressionof all these genes was indeed increased in transgenic endosperm(up to ∼15-fold for MCM7), with a significant peak at 19 DAP,similar to those of RBR3-4 (P < 0.05 for RBR3-4, MCM2-7, andPCNA, Student t test). However, PCNA was an exception, becauseits expression continued to increase (up to sevenfold) at 22 DAP(Fig. 2A). Analysis by end-point RT-PCR essentially confirmedthese results (Fig. S2). Thus, the up-regulation of E2F targetsindicated that although there were residual levels of RBR1-2

transcripts, RBR1 activity was substantially down-regulated inRBR1DS1 mutant endosperm.We next analyzed RBR1, RBR3, andMCM3 protein expression

levels by Western blotting of 16-DAP endosperm extracts. RBR1and MCM3 are normally present in endoreduplicating wild-typeendosperm, whereas RBR3 protein is not detectable (Fig. 2B andFig. S1C) (11). However, no ∼96-kDa band was detected intransgenic endospermwith either of the two anti-RBR1 antibodies,whereasRBR3was clearly expressed andMCM3was up-regulated.The α-RBR1P antibody identified RBR1 unambiguously, whereasα-RBR1N cross-reacted with several other abundant polypeptides,one of which almost comigrated with the RBR1 band in 16-DAPextracts but was not present at 9 DAP. Together, these results in-dicate the RNAi pathway triggered by RBR1DS1 not only affectedthe steady-state level of RBR1 RNA but also had a major effecton accumulation of the encoded protein, which resulted in up-regulation of E2F-dependent gene expression.

Stimulation of Endoreduplication by Down-Regulation of RBR1. Wenext investigated whether the endoreduplication cell cycle wasaffected by down-regulation of RBR1. The same segregatingendosperms that were characterized with respect to patterns ofgene expression were also analyzed by flow cytometry to measurenuclear ploidy levels. The mean ploidy of RBR1DS1 endospermnuclei increased up to 43% between 13 DAP and 16 DAP rel-ative to wild-type endosperm, and this effect was sustained forthe duration of the period of endosperm development analyzed(i.e., up to 22 DAP) (Fig. 3A). Comparisons of flow cytometricprofiles at 16 DAP revealed more high-ploidy (≥24C) nuclei inRBR1DS1 relative to wild-type endosperm (Fig. 3B and Fig.S3A). The frequency of 24C, 48C, 96C, and 192C endospermnuclei in RBR1DS1 increased 1.4-, 2-, 4.3-, and 39-fold, re-spectively, compared with wild-type endosperm. This was ac-companied by a 0.9-fold reduction in the frequency of 3C nuclei(Fig. 3B). The increased amount of DNA in 48C, 96C, and 192Cnuclei (1.5-, 3.6-, and 33-fold, respectively) in transgenic endo-sperm was accompanied by decreased DNA in 3C, 6C, and 12Cnuclei (0.6-, 0.7-, and 0.7-fold, respectively) (Fig. 3C). Theseresults are consistent with the changed patterns in gene expres-sion and indicate that down-regulation of RBR1 resulted ina sustained up-regulation of the endoreduplication cell cycle.In addition, measurements in 19-DAP and mature, segregat-

ing endosperms through a PicoGreen-based assay revealed 69%and 43% DNA increases, respectively, in RBR1DS1 endosperm(Table 1 and Table S1). Endosperm DNA content (per mg dryflour ± SEM) was also measured spectrophotometrically, whichat 19 DAP gave 1,412.4 ± 157.6 ng for RBR1DS1 and 1,111.9 ±229.3 ng for wild-type endosperm (+26% in transgenic endo-sperm). At maturity, RBR1DS1 endosperm contained 156.7 ± 5.6ng vs. 140.3 ± 7.1 ng for wild type (+11% in transgenic endo-sperm). Because RBR1DS1 endosperm contained significantlymore DNA than wild type according to two different criteria, weconcluded that stimulation of endoreduplication in RBR1DS1endosperm was sustained during endosperm development andresulted in increased DNA content in mature seeds.

Down-Regulation of RBR1 Stimulates Endosperm Cell Death. AlthoughRBR1DS1 endosperm contained more DNA than wild type at 19DAP, at maturity this difference was attenuated by 38–58%. Thissuggested that PCD was enhanced in RBR1DS1 endosperm. Todetermine whether DNA degradation was affected by down-regulation of RBR1, we measured the proportion of low-molecular-weight DNA (i.e., <1,000 bp) in mature transgenic and wild-typeendosperm (Fig. 4A and Fig. S3B). RBR1DS1 endosperm hadbetween 1.6- and 2.3-fold more low-molecular-weight DNA com-pared with wild type, indicating that DNA degradation was in-tensified by the down-regulation of RBR1.Release of cytochrome c from mitochondria is a hallmark of

PCD in both mammals and plants and generally precedes DNAdegradation and cell death (25). PCD is normally detectable inmaize endosperm by 16 DAP (6). Thus, we determined whether

A

12

4

6

8

10

RBR1 RBR2 RBR3 RBR4 MCM2 MCM3 MCM4 MCM5 MCM6 MCM7 PCNA

10 DAP

13 DAP

16 DAP

19 DAP

22 DAP

Re

la

tiv

e e

xp

re

ss

io

n

(fo

ld

d

iffe

re

nc

e)

12.3 ± 2.1 11.5 ± 1.6 15.4 ± 4.3

B + + + + - - - -

RBR1DS1

RBR1

RBR1

RBR3

MCM3

ACT

-RBR1P

-RBR1N

Fig. 2. Effects of RBR1 down-regulation on downstream gene expression.(A) Expression levels of RBR1-4, MCM2-7, and PCNA RNAs in developingRBR1DS1 endosperm relative to wild type, as measured by quantitative RT-PCR. A horizontal red line indicates the reference expression level of one unitin control samples. Error bars indicate SEMs. Each bar represents the mean ofthree to six replicates. RNA expression levels for MCM4, MCM5, and MCM7,which exceeded the scale of the histogram, are shown at the top. (B) Immu-noblot analysis showing that RBR1DS1 endosperm has no detectable RBR1protein and displays up-regulation of RBR3 and MCM3 proteins. (Upper)Presence or absence of the RBR1DS1 transgene (as scored by PCR) in 16-DAPtransgenic (+) andwild-type (-) endosperm, respectively. RBR1 protein in wild-type endosperm extracts was detected using two different antibodies(α-RBR1P and α-RBR1N). Actin proteinwas detected as a loading control. RBR1and up-regulated polypeptides are indicated by arrowheads. Both transcriptand protein analyses were carried out on half endosperms dissected fromwild-type and RBR1DS1 kernels from segregating ears.

Sabelli et al. PNAS | Published online April 22, 2013 | E1829

PLANTBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

RBR1 down-regulation altered the relative distribution of cyto-chrome c between mitochondria and cytosol in 19-DAP endo-sperm (Fig. 4B). Mitochondria- and cytosol-enriched fractionswere assayed by Western blot analysis using antibodies againstcytochrome c, the α-subunit of mitochondrial F1-ATPase (acontrol for the quality of the subcellular preparations) and actin(a gel loading control). Whereas αATPase was prevalent in themitochondrial fraction in both wild-type and RBR1DS1 extracts,cytochrome c was preferentially found in the cytosolic fraction,consistent with the notion that PCD was ongoing at this stage ofendosperm development. The cytosol/mitochondria ratio forcytochrome c was increased (by 70%) in RBR1DS1 endospermrelative to wild type, whereas virtually no change was observedfor αATPase and actin. Thus, increased mitochondrial release ofcytochrome c was in agreement with enhanced DNA degradationin RBR1DS1 endosperm, which indicated that PCD was stimu-lated by RBR1 down-regulation.

Protein Levels in Transgenic RBR1DS1 Endosperm Are Similar to Thosein Wild Type. To determine whether increased endoreduplicationassociated with down-regulation of RBR1 affected accumulationof endosperm proteins, we compared protein contents in trans-genic and wild-type endosperms. Maize endosperm proteins in-clude storage prolamins (i.e., zein) and metabolic and/or structural

proteins (i.e., nonzein) comprising albumins, globulins, and otherminor components (26). Both zein and nonzein protein fractionswere measured in 19-DAP RBR1DS1 and wild-type endosperms(Table 1 and Table S2), and no significant differences were found.A similar result was obtained for zein in mature endosperm.However, the amount of nonzein proteins in mature RBR1DS1endosperm was 19% higher than in wild type (16.5 vs. 14.7 μgprotein/mg flour). Gel electrophoresis analyses of endospermproteins were consistent with these results (Fig. S3C).The amount of protein relative to DNA was calculated for

19-DAP and mature endosperms of both genotypes (Table 1 andTable S3). At 19 DAP the zein content of wild-type endospermwas 0.3 μg/ng DNA, whereas RBR1DS1 endosperm contained0.14 μg/ng DNA, which represented a 53% decrease. Similarly, thenonzein protein/DNA ratio in RBR1DS1 endosperm (0.06 μg/ng)was nearly half that of wild type (0.12 μg/ng). A similar trend wasalso found in mature endosperm. Thus, the increase in genomecopy number in RBR1DS1 endosperm was not accompanied bya proportional increase in storage protein accumulation.

RBR1DS1 Endosperm Has Reduced Cell and Nuclear Sizes but ContainsMore Cells and Develops Similar to Wild Type. To establish whetherincreased endoreduplication in RBR1DS1 endosperm was asso-ciated with changes in cell size, the areas of transgenic and wild-type starchy endosperm cells were compared. Three relativelylarge regions, zones A–C, were selected on median longitudinalsections of kernels harvested at 10, 12, 14, 16, and 18 DAP (Fig.5A), and comparable areas of cells were measured for the twogenotypes. Cell area was reduced in RBR1DS1 endosperm forthe vast majority (90%) of the cell populations that displayedsignificant changes (P < 0.05) between the two genotypes (Fig.5B, Fig. S4 A and B, and Table S4). This reduction was generallyon the order of 10–20%, but decreases up to 36% were observed.Overall, RBR1DS1 endosperm had smaller cell size than wildtype by 12% at 12 and 14 DAP and by 15% at 16 DAP, but nosignificant differences were found at 10 or 18 DAP. Hence, in-creased E2F-dependent gene expression and endoreduplicationin RBR1DS1 endosperm were associated with reduced ratherthan increased cell size, although this effect seemed to be tran-sient over development of this tissue.Cell and nucleus areas were compared in wild-type and

RBR1DS1 cells in 16-DAP endosperms (Table 1 and Fig. S4C).The average area of transgenic RBR1DS1 cells was 82% that ofwild-type cells, and nuclear area was also similarly reduced(83%). In both genotypes, nucleus and cell areas were positivelycorrelated to virtually the same degree (Fig. S4C). Because bothcell and nucleus areas were proportionally reduced in transgenicendosperm, no significant difference relative to wild type wasfound in the cell area/nucleus area ratio (Table 1). Assumingendosperm cells and nuclei are roughly isodiametric, theseobservations indicate that increased DNA content in RBR1DS1cells is associated with similar decreases in cell and nuclear sizes,resulting in virtually unaltered karyoplasmic ratios comparedwith wild type.RBR1DS1 kernels did not seem to develop differently from

wild type with respect to structure and size (Fig. S5), nor wastheir weight significantly different at maturity (Table 1). Theseobservations, coupled with the smaller average cell size in themutant, suggested that RBR1DS1 endosperm contained a largernumber of cells than wild type, resulting from increased cell di-vision activity. The number of cells in 19-DAP wild-type andRBR1DS1 endosperms was estimated according to total DNAcontent and mean ploidy measurements (Table S5) and was alsomeasured directly in 16-DAP tissue sections (Table 1 and Fig.S4A). Mutant endosperm contained 42–58% more cells thanwild type, indicating down-regulation of RBR1 stimulated cellproliferation.

CDKA;1 Controls Endoreduplication, but Not Gene Expression, Throughan RBR1-Dependent Pathway. Earlier work showed that CDKA;1 isrequired for endoreduplication in maize endosperm (21), but

Fig. 3. Endoreduplication is stimulated in RBR1DS1 transgenic endosperm.(A) Mean ploidy levels in RBR1DS1 (red bars) and wild-type (black bars)endosperms at five developmental stages between 10 and 22 DAP. Relativetransgenic ploidy levels are expressed as fold changes in C values. (B) Dis-tribution of endosperm nuclei (expressed as a percentage of the totalnumber of nuclei) among different ploidy classes in 16-DAP RBR1DS1 (redbars) and wild-type (black bars) endosperms. (C) Distribution of nuclear DNA(expressed as a percentage of the total amount of DNA) among differentploidy classes in 16-DAP RBR1DS1 (red bars) and wild-type (black bars)endosperms. Significance levels of pair-wise differences between controland transgenic RBR1DS1 values as determined by Student t test: *P < 0.05;**P < 0.01; ***P < 0.001. Error bars indicate SEMs. All analyses were carriedout on dissected endosperms from genotyped kernels segregating on thesame ears. Each bar represents the mean of 5–12 samples.

E1830 | www.pnas.org/cgi/doi/10.1073/pnas.1304903110 Sabelli et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

whether CDKA;1 controls endoreduplication or downstream geneexpression through RBR1 is not known. The dominant-negativeCDKA1DN allele reduces CDKA;1 activity and endoreduplicationby approximately 50% (21). Hemizygous plants carrying theCDKA1DN and RBR1DS1 transgenes were crossed, resultingin four progeny endosperm genotypes: wild type, CDKA1DN,RBR1DS1, and the double mutant CDKA1DN/RBR1DS1. TheRBR1DS1 transgene did not affect expression levels of CDKA;1(Fig. 6A), indicating CDKA;1 protein expression levels are in-dependent from RBR1 regulation. RNA levels of RBR1 down-stream targets in the four genotypes showed that expression of all

transcripts was dramatically increased in endosperm carrying theRBR1DS1 transgene (Fig. 6B), consistent with previous obser-vations. Relative to wild type, CDKA1DN endosperm did notdisplay any significant changes (P > 0.05, Student t test) in levelsof RBR1-repressed transcripts. We found the MET1-homologRNA, DMT101, followed the same pattern of expression as otherRBR1 targets (Fig. 6D). Similarly, the CDKA1DN mutation didnot significantly affect gene expression when present in combi-nation with RBR1DS1, with the exception of RBR4 and PCNA.Thus, we concluded that CDKA;1 does not control expression ofRBR1 target genes, and to a large extent this is true regardless ofRBR1 activity.Based on phosphorylation of histone H1, RBR1DS1 endo-

sperm had considerably higher CDK activity than wild type (3.9-fold ± 0.9 SEM, on the basis of three experiments) (Fig. 6A),indicating that RBR1 normally plays a CDK inhibitory role. Ki-nase activity decreased in the double CDKA1DN/RBR1DS1 mu-tant relative to RBR1DS1, but it was higher than in theCDKA1DN mutant. This suggests that down-regulation of RBR1up-regulated both CDKA;1-dependent and -independentCDK activities.Nuclear DNA contents in endosperms of the four genotypes

resulting from the CDKA1DN × RBR1DS1 cross were measuredby flow cytometry (Fig. 6C). As expected from previous results,CDKA1DN endosperm had a much-reduced mean ploidy level(6.6C) than wild type (12.1C), whereas the DNA content ofRBR1DS1 endosperm nuclei was increased (15.4C). Mean ploidyof the CDKA1DN/RBR1DS1 mutant, however, was not signifi-cantly different from RBR1DS1. These results indicate that al-though CDKA;1 is required for endoreduplication, it can onlyfulfill this role in the presence of RBR1.

DiscussionWe have established RBR1 as a central regulator of maize en-dosperm development. RBR1 controls downstream gene expres-sion, CDK activity, endoreduplication, cell proliferation, cell size,nuclear size, and cell death (Fig. 7). Furthermore, genetic evi-dence indicates that CDKA;1 controls endoreduplication throughan RBR1-dependent pathway, but it has virtually no impact on

Table 1. Comparison of different parameters between wild-type and RBR1DS1 endosperms

Parameter

Wild type RBR1DS1RBR1DS1/wild

type* PMean value n Mean value n

19 DAP DNA† 227.8 ± 32.3 15 385.7 ± 33.9 15 1.69 (**) 0.002Mature DNA† 46 ± 2.9 32 66 ± 2.9 43 1.43 (***) 5.97E-0619 DAP zein‡ 50.74 ± 2.1 15 49.78 ± 1.6 15 0.98 (N.S.) 0.71919 DAP nonzein‡ 21.61 ± 2.23 15 22.63 ± 1.7 15 1.05 (N.S.) 0.718Mature zein‡ 62.66 ± 3.06 32 59.09 ± 2.25 43 0.94 (N.S.) 0.351Mature nonzein‡ 14.73 ± 0.54 32 16.5 ± 0.42 43 1.19 (*) 0.01219 DAP zein/DNA§ 0.3 ± 0.05 15 0.14 ± 0.01 15 0.47 (**) 0.00419 DAP nonzein/DNA§ 0.12 ± 0.02 15 0.06 ± 0.004 15 0.51 (*) 0.027Mature zein/DNA§ 1.59 ± 0.15 32 0.96 ± 0.05 43 0.6 (***) 0.0004Mature nonzein/DNA§ 0.38 ± 0.05 32 0.26 ± 0.08 43 0.68 (*) 0.014Kernel weight¶ 246.01 ± 31.92 370 242.73 ± 32.19 385 0.99 (N.S.) 0.1619 DAP cell numberk 81,723 — 128,972 — 1.58 —

16 DAP cell number†† 316 ± 18.1 3 447.3 ± 20.1 4 1.42 (**) 0.00516 DAP cell area‡‡ 618.54 ± 31.42 179 508.59 ± 18.35 229 0.82 (**) 0.00316 DAP nucleus area‡‡ 25.57 ± 1.24 179 21.15 ± 0.7 229 0.83 (**) 0.00216 DAP cell area/nucleus area 26.01 ± 1.12 179 25.83 ± 0.82 229 0.99 (N.S.) 0.9

*Significance levels according to Student t test: *P < 0.05; **P < 0.01; ***P < 0.001. N.S., not significant.†DNA content is expressed as ng DNA/mg dry flour ± SEM.‡Protein content is expressed as μg protein/mg dry flour ± SEM.§Amount of protein relative to DNA is expressed as μg protein/ng DNA ± SEM.¶Kernel weight (at maturity) is expressed as mg ± SEM.kEstimated number based on measurements of total DNA content and mean ploidy.††Number of cells measured on tissue sections.‡‡Area is expressed as pixels ± SEM.

Fig. 4. PCD is enhanced in RBR1DS1 endosperm. (A) Average amounts oflow-molecular-weight (<1,000 bp) DNA in transgenic RBR1DS1 endospermrelative to wild type in two mature ears. Error bars indicate SDs. Differencesbetween wild-type and transgenic endosperms for each ear are significant asdetermined by Student t test: *P < 0.05; **P < 0.01. Each bar represents themean of four to six samples. (B) Immunoblot analysis showing the relativedistribution of cytochrome c, the α-subunit of mitochondrial F1-ATPase, andactin in subcellular fractions from 19-DAP endosperm enriched for mito-chondria (M) and cytosol (C). All analyses were carried out on dissectedendosperms from genotyped kernels segregating on the same ears.

Sabelli et al. PNAS | Published online April 22, 2013 | E1831

PLANTBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

expression of RBR1-repressed genes. These results shed light onthe functions of CDKA;1 and RBR1 and the role of endoredu-plication during cereal endosperm development.

Control of Gene Expression by RBR1. Down-regulation of RBR1protein during maize endosperm development results in up-regulation of several RBR1 targets, such as RBR3-4, MCM2-7,and PCNA. RBR3 is repressed by RBR1 in embryogenic callus,where cells are undifferentiated, divide rapidly, and generallylack endoreduplication (11, 13). The present investigation dem-onstrates that RBR3 expression is similarly regulated in terminallydifferentiated, endoreduplicating endosperm cells. In addition, wefound that the recently identified RBR4 gene (9) has an overallpattern of RNA expression during endosperm development similarto RBR3, and is also negatively regulated by RBR1. In contrast,RBR2RNA level during endosperm development expression showsan upward trend, resembling that of the closely related RBR1 gene

(11). Publicly available transcriptomic datasets (http://qteller.com/qteller3/) indicate that RBR2 and RBR4 are expressed at consid-erably reduced absolute levels compared with RBR1 and RBR3,respectively. Thus, these observations collectively indicate thatRBR2 and RBR4 share similar regulation and potentially similarfunctions with their respective paralogs, RBR1 and RBR3.The up-regulation of downstream targets upon RBR1 down-

regulation can be rather dramatic, as in the case of several MCMgenes. Their increase in gene expression becomes pronounced atapproximately 13 DAP, generally reaches a peak at 19 DAP, andis sustained at least until 22 DAP. However, under closer in-spection, it is apparent that at least two major gene expressionpatterns exist. One, typical of RBR3-4 and the MCM2-7 genes,displays a steady increase in RNA levels up to 19 DAP, followedby a decline by 22 DAP. The other, represented by PCNA,involves a continuous RNA increase at least up to 22 DAP.RBR3 plays a positive role in regulating expression of theMCM2-7

Fig. 5. Reduced cell size in RBR1DS1 endosperm. (A) Cell areas were measured in three endosperm zones (indicated A–C) of segregating wild-type andRBR1DS1 kernels. (Inset) Cell contours and numbering. Scale bar = 1 mm. (B) Plots showing average cell areas in both wild-type and RBR1DS1 endosperms inthe three zones measured at 2-d intervals between 10 and 18 DAP. Error bars indicate SEMs. Each datapoint represents the mean of 600–1,503 measurements.Complete analysis is shown in Fig. S4B and Table S4.

E1832 | www.pnas.org/cgi/doi/10.1073/pnas.1304903110 Sabelli et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

gene family in tissue culture (13), and data presented here areconsistent with a similar role in the endosperm. Members of theRBR gene family, however, can have redundant roles (9, 27), andthe possibility that RBR3 could, at least in part, substitute forRBR1 in repressing endosperm gene expression (12) cannot beruled out. Down-regulation of CDKA;1 did not result in a signifi-cant decrease in expression of RBR1 targets, indicating that RBR1activity controlling gene expression is not inhibited by CDKA;1.This finding contrasts with recent results in Arabidopsis showingthat CDKA;1 is required for expression of RBR1 targets (18).

CDKA;1/RBR1 Pathway Controls Endoreduplication During EndospermDevelopment. The CDKA;1/RBR1 pathway plays an importantrole in regulation of maize endosperm endoreduplication. RBR1inhibits endoreduplication, and down-regulation of RBR1 stim-ulates it. This observation is in agreement with previous results in

tobacco (23) and Arabidopsis (28) but contrasts a recent report inwhich Arabidopsis RBR1 was down-regulated by RNAi (29).In plants, CDKA protein may have an S-phase or M-phase

regulatory function, based primarily on the type of associatedcyclin (7). However, it is not always possible to precisely de-termine the composition of CDK complexes, partly because of theredundancy in gene families encoding these proteins. Consider-able evidence implicates CDKA activity in control of endor-eduplication in plants, and in most cases CDKA is required forendoreduplication (7). However, at least in certain systems,CDKA activity has been reported to inhibit it (30). In maize, thereare at least two CDKA-related genes, and their respective roles inthe endosperm are largely unknown. It is generally believed thealternation of G- and S-phases during endoreduplication requiresactivation of pathways controlling DNA replication and/or down-regulation of those that promote mitosis and cytokinesis. Ac-cordingly, dividing cells would suppress CDK complexes involved

Fig. 6. Genetic interaction between CDKA;1 and RBR1. (A) Western blot analysis (Top and Top Middle) of CDKA;1 and CDKA1DN protein expression in 19-DAP endosperms of the four genotypes (wild type, CDKA1DN, RBR1DS1, and CDKA1DN/RBR1DS1) derived from a CDKA1DN × RBR1DS1 cross. Antibodiesagainst CDKA;1 (α-CDKA;1) recognize both the endogenous and the mutant CDKA1DN proteins. CDKA1DN protein is specifically recognized by antibodiesagainst human influenza hemagglutinin epitope tag (α-HA), which was engineered in its sequence (21). Antibodies against actin were used as gel loadingcontrol. (Bottom) Phosphorylation of histone H1 substrate by p13Suc1-adsorbed kinase activity in the four genotypes. (B) Expression levels of RBR1 targettranscripts in 19-DAP endosperms of the four genotypes described in A. Bars represent the mean of three biological replicates each made from four, pooledendosperm RNAs. (C) Mean ploidy levels in 19-DAP endosperms of the four genotypes described in A. Significance levels of pair-wise comparisons withCDKA1DN according to Student t test: *P < 0.05; **P < 0.001. (D) RNA expression levels of the DNA METHYLTRANSFERASE 1 homolog, DMT101, in the fourgenotypes described in A. Datapoint composition as described in B. All analyses were carried out on dissected endosperms from genotyped kernels segre-gating on the same ear. Error bars indicate SEMs.

Sabelli et al. PNAS | Published online April 22, 2013 | E1833

PLANTBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

in regulation ofM-phase (31), preventing cell division andmakingcells competent for endoreduplication. This switch is believed tousually be associated with initiation of cell differentiation (32, 33).Subsequently, precise fluctuations in CDK activities between G-and S-phases enable these cells to sustain an endoreduplicationcell cycle.The outcome of expressing a dominant negative CDKA;1 al-

lele indicate CDKA;1 is required for endoreduplication in maizeendosperm and suggests that CDKA;1 activity plays a necessaryrole in regulating S-phase in this type of cell cycle. Our data showthat down-regulation of CDKA;1 inhibits endoreduplication, butthis effect is essentially abolished in cells in which RBR1 is alsodown-regulated, such as in the CDKA1DN/RBR1DS1 mutant.These observations indicate epistatic regulation between CDKA;1and RBR1 and that CDKA;1 controls endoreduplication throughan RBR1-dependent pathway, most likely by targeting RBR1 forinhibitory phosphorylation.

Activity and Role of RBR1 During Endosperm Development. Our re-sults show there is substantial RBR1 activity during the endor-eduplication phase of endosperm development. RBR1 seems toantagonize and limit the extent of endoreduplication during maizeendosperm development, which is consistent with the induction ofendoreduplication by down-regulation of RBR1 in tobacco andoverexpression of E2Fa/DPa in Arabidopsis (23, 34).CDKA;1 likely functions as an S-phase kinase and targets RBR1

for inhibition during maize endosperm development. This con-clusion is supported by the phosphorylation of RBR1 by CDKcomplexes containing A-type cyclins (11). These complexes areinhibited by maize ICK/KRP-type CDK inhibitors, which typicallytarget complexes containing CDKA (35).

Down-regulation of RBR1 also results in a substantial increaseof p13Suc1-adsorbed CDK activity, which contains CDKA;1. Theinhibition of CDKA;1 activity by RBR1 suggests the presence ofa regulatory feedback loop whereby down-regulation of RBR1stimulates CDKA;1 activity, which could further inhibit RBR1.However, RBR1 also seems to inhibit the activity of CDK(s)other than CDKA;1. It is possible that B-type CDKs, which areunique to plants and thought to primarily control the G2/M-phase transition during the mitotic cell cycle, are responsible forthis RBR1-repressible activity (18). In Arabidopsis, CDKB1;1targets the CDKA1;1 inhibitor KRP2 for proteasome-dependentdegradation, thereby stimulating CDKA1;1 activity and M-phase(36). Expression of a B-type CDK may be repressed by RBR1 inmaize endosperm, and thus a similar mechanism could accountfor increased CDKA;1 activity and cell proliferation that char-acterize RBR1DS1 endosperm.An important conclusion of this investigation is that the roles

of CDKA;1 and RBR1 during maize endosperm development arepartially uncoupled. Although the RBR1 activity that repressesgene expression and endoreduplication can be severely decreasedby down-regulating RBR1, only the component inhibiting endo-reduplication can be substantially stimulated by decreased CDKA;1activity. These observations suggest there might be at least twoRBR1-dependent pathways controlling downstream effects in theendosperm: one controlling endoreduplication, which is dependenton the activity of CDKA;1 (and presumably on the phosphorylationstate of RBR1), and another inhibiting expression of E2F targets,which is CDKA;1-independent. These findings are in contrast tothe observation in Arabidopsis that both cell cycle activity andRBR1-dependent gene expression are controlled by CDKA;1 (18)and highlight the greater degree of complexity of the CDK–RBRpathway in maize and related cereal crops. This likely reflects thefunctional specialization of individual members within maizeCDKA and RBR gene families, whereas these activities areencoded by single genes in Arabidopsis (11, 20).RBR1 seems to inhibit endosperm cell proliferation. This ob-

servation implies two alternative, non-mutually exclusive sce-narios. On the one hand, endoreduplicated cells could reenter themitotic cell cycle upon down-regulation of RBR1, but this seemsunlikely owing to large numbers of intracellular starch granules.Indeed, we did not observe an increase in mitotic figures inendoreduplicating RBR1DS1 endosperms. On the other hand, thedevelopmental window in which cell division takes place beforeendoreduplication could be extended in RBR1DS1 endosperm.This scenario could have important implications about the activityof the 27-kDa γ-zein promoter, which controls the expression ofthe RBR1DS1 allele, and the role of endoreduplication in theaccumulation of storage proteins. In fact, the increased cell di-vision in RBR1DS1 endosperm suggests that the 27-kDa γ-zeinpromoter is active in both mitotic and endoreduplicating cells.This interpretation is consistent with the presence of high levels of27-kDa γ-zein transcript throughout the endosperm as early as 8DAP (22) and suggests the onset of storage protein gene ex-pression may precede, rather than accompany or follow, endor-eduplication. However, it seems at odds with the lack of an effecton cell division when the same promoter was used to express theCDKA1DN allele (21). It is also possible RBR1DS1 expression isrestricted to endoreduplicated cells and, as a result, that somediffusible factor, perhaps a small RNA molecule, feeds back tomitotic cells, thereby stimulating cell proliferation.Another function of RBR1 in the endosperm is to control cell

death. PCD starts approximately 16 DAP in maize endospermand involves extensive DNA degradation, which is especiallyprominent during endosperm maturation (5, 6). DNA degrada-tion is more pronounced in mature RBR1DS1 endosperm com-pared with wild type and is accompanied, at an earlier stage, byincreased release of cytochrome c from mitochondria, a hallmarkof PCD (25). PCD was enhanced in RBR1DS1 mutant endo-sperm, which points to an inhibitory role for RBR1 in this pro-cess. This is consistent with inhibition of PCD in animals andtobacco by RBR proteins (37–40). However, it is not clear

Fig. 7. The CDKA;1/RBR1 pathway controls different aspects of maize en-dosperm development. In this schematic model, the CDKA;1/RBR1 pathwayis shown to regulate gene expression, endoreduplication, cell proliferation,cell death, and cell size during the development of maize endosperm. CDKA;1controls endoreduplication through RBR1 but does not impact RBR1-dependent gene expression. RBR1 represses CDK activity. Broken linesindicate hypothetical relationships. Details in main text.

E1834 | www.pnas.org/cgi/doi/10.1073/pnas.1304903110 Sabelli et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

whether RBR1 has a direct PCD inhibitory function, or ratherthat increased endoreduplication stimulates cell death in RBR1DS1endosperm (6).

Role of Endoreduplication in Endosperm Development. Strong cor-relations between endoreduplication, nuclear size, and cell sizehave been observed in many systems, including the endosperm ofmaize (5). One of the proposed functions of endoreduplication isto support high rates of gene expression and metabolic output toinduce cell/tissue growth (4, 5, 41–44). Indeed, inner starchyendosperm cells are much larger and possess bigger nuclei thanthe low-ploidy cells at the periphery of the endosperm. Consis-tent with this view, a strong correlation also exists between nu-clear size, RNA transcriptional levels, and cell size (43, 45, 46).We show that nuclear size and cell size are positively and equallycorrelated regardless of RBR1 down-regulation. This suggestsa causal relationship between nuclear size and cell/cytoplasmicsize in the endosperm, supporting the karyoplasmic ratio theory(47, 48), and indicates that RBR1 does not impact it. However,down-regulation of RBR1 stimulates endoreduplication, withmarked increases in the amount of DNA in developing andmature endosperm, while resulting in smaller cells and nuclei.These observations indicate that RBR1 is involved in couplingnuclear (and cell) size to DNA content.What do these data reveal in terms of the function of endor-

eduplication in the endosperm? Inhibiting endoreduplicationthrough down-regulation of CDKA;1 activity did not affect cellsize, gene expression, and endosperm growth (21). Increasedendoreduplication in RBR1DS1 endosperm resulted in no signifi-cant changes in zein storage protein content ormature seedweight,a proxy for starch accumulation. Furthermore, whereas transgeniccells were significantly smaller between 12 and 16 DAP, theymostly appeared of normal size by 18 DAP, and the caryopsis asa whole apparently developed normally. It is thus tempting toconclude that endoreduplication does not contribute significantlyto cell growth, accumulation of storage compounds, or endospermsize. However, at least two alternative explanations caution thisinterpretation. First, evidence presented here and elsewhere sug-gests the CDKA;1/RBR1 pathway plays a pivotal role in couplingendoreduplication to cell size control, and therefore perturbingthis pathway may alter endoreduplication without a correspondingchange in cell or organ size (21, 49–51). Second, it seems thata compensatory mechanism, involving regulation of gene expres-sion, was activated in response to changes in endoreduplicationlevels as part of endosperm homeostasis. Indeed, relative to theamount of DNA, protein content decreased dramatically inRBR1DS1 endosperm compared with wild type, suggesting theadditional DNA is less transcriptionally active. This hypothesis isalso supported by the decline in RNA levels of most RBR1 targetsafter 19 DAP. It is conceivable that this response involves modi-fication of chromatin conformation. In Arabidopsis, maize, andrice, the endosperm genome is relatively hypomethylated (52–55).CG methylation in Arabidopsis, which is catalyzed by DNAMETHYLTRASFERASE (MET) 1, has a profound and primarilyrepressive effect on the expression of hundreds of genes, pseudo-genes, and transposons (56, 57). In rice, local CG hypomethylationcorrelates with the expression of endosperm specifically or pref-erentially expressed genes, including those encoding highlyexpressed storage proteins and starch biosynthetic enzymes (55). Inmaize, several imprinted genes are hypomethylated and prefer-

entially expressed in the endosperm (58). We found that expres-sion of the DMT101 transcript, a maize MET1 homolog (59), issignificantly up-regulated (P < 0.05, Student t test) in RBR1DS1endosperm (Fig. 6D), consistent with the repression of MET1 byRBR1 in Arabidopsis (60, 61). In mouse, expression of the MET1-homologDNMT1 is rate-limiting for DNAmethylation (62). Thus,down-regulation of RBR1 in maize endosperm could conceivablystimulate, at least locally, DMT101-dependent hypermethylationof endoreduplicated chromatin, contributing to gene silencing.This finding agrees with the observation that nuclear size is re-duced in RBR1DS1 endosperm despite increased DNA content,suggesting enhanced chromatin compaction. In support of thishypothesis, positive correlations both between nuclear size re-duction and chromatin condensation and between DNA hypo-methylation and nuclear size have been found in Arabidopsis seed(63) and ovarian cancer cells (64), respectively. An alternative, butnot mutually exclusive, possibility is that gene expression levels areprimarily determined by a cell size-sensing mechanism, rather thanby the intrinsic amount of nuclear DNA (65). Thus, a decrease ingene expression (on a per DNA basis) could be a key component ofendosperm homeostasis in RBR1DS1 kernels. These potentialscenarios suggest RBR1 may play a role in differentiation of en-dosperm cells, allowing expression of genes required for storagecompound accumulation through chromatin decondensation, andraise the possibility that endoreduplicated chromatin in RBR1DS1endosperm might not be functionally and/or physiologicallyequivalent to the endoreduplicated chromatin arising during thenormal development of maize endosperm.In conclusion, RBR1 seems to control endosperm develop-

ment primarily through two key mechanisms: it antagonizes cellcycle activity and stimulates nuclear and cell enlargement. Theresults presented here demonstrate that maize endosperm is anextremely plastic tissue capable of normal development despiteconsiderable changes in several key parameters governing thecell cycle, the number and size of its cells, and their viability.These findings could have important implications for the ma-nipulation of grain yield in cereal crops.

Materials and MethodsPlant Materials, Nucleic Acids, and Antibodies. All analyses refer to materialsfrom transgenic maize (Zea mays L.) plants backcrossed with the B73 inbredat BC3 or later generations. Detailed information about constructs, thegeneration of transgenic plants, antibodies, nucleic acid database accessionnumbers, and primer sequences are given in SI Materials and Methods.

Analyses of Endosperms and Kernels. To minimize artifacts due to environ-mental effects and plant-to-plant variation, all comparative analyses betweentransgenic and wild-type genotypes were carried out using individuallygenotyped endosperms or kernels harvested from the middle third of earssegregating for the transgene. Samples were either analyzed independentlyor pooled before the assay. Procedures for extraction of nucleic acids, proteinpurification, RT-PCRand SDS/PAGEanalyses, immunoblotting, assays of kinaseactivity, flow cytometry, microscopy, measurements and calculations of DNAand protein contents, cell area, nuclear area, cell number, analyses of celldeath, and mature seed weight are given in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Drs. Rudolph Jung and William J. Gordon-Kamm (Pioneer Hi-Bred Intl. Inc.) for helpful discussions and support. Thiswork was funded by US Department of Energy (Grant DE-FG02-96ER20242).R.A.D. was supported by a scholarship from the Conselho Nacional deDesenvolvimento Cientifico e Tecnologico of Brazil.

1. Sabelli PA, Larkins BA (2009) The development of endosperm in grasses. Plant Physiol

149(1):14–26.2. Sabelli PA, Larkins BA (2009) The contribution of cell cycle regulation to endosperm

development. Sex Plant Reprod 22(4):207–219.3. Kowles RV, Phillips RL (1985) DNA amplification patterns in maize endosperm nuclei

during kernel development. Proc Natl Acad Sci USA 82(20):7010–7014.4. Larkins BA, et al. (2001) Investigating the hows and whys of DNA endoreduplication. J

Exp Bot 52(355):183–192.5. Sabelli PA (2012) Replicate and die for your own good: Endoreduplication and cell

death in the cereal endosperm. J Cereal Sci 56(1):9–20.

6. Young TE, Gallie DR (2000) Programmed cell death during endosperm development.

Plant Mol Biol 44(3):283–301.7. Inzé D, De Veylder L (2006) Cell cycle regulation in plant development. Annu Rev

Genet 40:77–105.8. Cobrinik D (2005) Pocket proteins and cell cycle control. Oncogene 24(17):

2796–2809.9. Sabelli PA, Larkins BA (2009) Regulation and function of retinoblastoma-related plant

genes. Plant Sci 177(6):540–548.10. van den Heuvel S, Dyson NJ (2008) Conserved functions of the pRB and E2F families.

Nat Rev Mol Cell Biol 9(9):713–724.

Sabelli et al. PNAS | Published online April 22, 2013 | E1835

PLANTBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

11. Sabelli PA, et al. (2005) RBR3, a member of the retinoblastoma-related family frommaize, is regulated by the RBR1/E2F pathway. Proc Natl Acad Sci USA 102(37):13005–13012.

12. Sabelli PA, Larkins BA (2006) Grasses like mammals? Redundancy and compensatoryregulation within the retinoblastoma protein family. Cell Cycle 5(4):352–355.

13. Sabelli PA, et al. (2009) Positive regulation of minichromosome maintenance geneexpression, DNA replication, and cell transformation by a plant retinoblastoma gene.Proc Natl Acad Sci USA 106(10):4042–4047.

14. Grafi G, et al. (1996) A maize cDNA encoding a member of the retinoblastoma proteinfamily: Involvement in endoreduplication. Proc Natl Acad Sci USA 93(17):8962–8967.

15. Rossi V, et al. (2003) A maize histone deacetylase and retinoblastoma-related proteinphysically interact and cooperate in repressing gene transcription. Plant Mol Biol51(3):401–413.

16. Boniotti MB, Gutierrez C (2001) A cell-cycle-regulated kinase activity phosphorylatesplant retinoblastoma protein and contains, in Arabidopsis, a CDKA/cyclin D complex.Plant J 28(3):341–350.

17. Nakagami H, Kawamura K, Sugisaka K, Sekine M, Shinmyo A (2002) Phosphorylationof retinoblastoma-related protein by the cyclin D/cyclin-dependent kinase complex isactivated at the G1/S-phase transition in tobacco. Plant Cell 14(8):1847–1857.

18. Nowack MK, et al. (2012) Genetic framework of cyclin-dependent kinase function inArabidopsis. Dev Cell 22(5):1030–1040.

19. Van Leene J, et al. (2010) Targeted interactomics reveals a complex core cell cyclemachinery in Arabidopsis thaliana. Mol Syst Biol 6:397, 10.1038/msb.2010.53.

20. Colasanti J, Tyers M, Sundaresan V (1991) Isolation and characterization of cDNAclones encoding a functional p34cdc2 homologue from Zea mays. Proc Natl Acad SciUSA 88(8):3377–3381.

21. Leiva-Neto JT, et al. (2004) A dominant negative mutant of cyclin-dependent kinase Areduces endoreduplication but not cell size or gene expression in maize endosperm.Plant Cell 16(7):1854–1869.

22. Woo Y-M, Hu DW-N, Larkins BA, Jung R (2001) Genomics analysis of genes expressedin maize endosperm identifies novel seed proteins and clarifies patterns of zein geneexpression. Plant Cell 13(10):2297–2317.

23. Park JA, et al. (2005) Retinoblastoma protein regulates cell proliferation, differentiation,and endoreduplication in plants. Plant J 42(2):153–163.

24. Hirano H, Harashima H, Shinmyo A, Sekine M (2008) Arabidopsis RETINOBLASTOMA-RELATED PROTEIN 1 is involved in G1 phase cell cycle arrest caused by sucrosestarvation. Plant Mol Biol 66(3):259–275.

25. Diamond M, Mccabe PF (2011) Plant Mitochondria, Advances in Plant Biology 1, edKempken F (Springer, New York), pp 439–465.

26. Wilson CM (1983) Seed Proteins: Biochemistry,Genetics, Nutritive Value, ed GottschalkW(Niijhoff/Junk, The Hague), pp 271–307.

27. Sage J, Miller AL, Pérez-Mancera PA, Wysocki JM, Jacks T (2003) Acute mutation ofretinoblastoma gene function is sufficient for cell cycle re-entry. Nature 424(6945):223–228.

28. Desvoyes B, Ramirez-Parra E, Xie Q, Chua NH, Gutierrez C (2006) Cell type-specific roleof the retinoblastoma/E2F pathway during Arabidopsis leaf development. PlantPhysiol 140(1):67–80.

29. Borghi L, et al. (2010) Arabidopsis RETINOBLASTOMA-RELATED is required for stemcell maintenance, cell differentiation, and lateral organ production. Plant Cell 22(6):1792–1811.

30. Imai KK, et al. (2006) The A-type cyclin CYCA2;3 is a key regulator of ploidy levels inArabidopsis endoreduplication. Plant Cell 18(2):382–396.

31. Grafi G, Larkins BA (1995) Endoreduplication in maize endosperm: involvement of mphase-promoting factor inhibition and induction of s phase-related kinases. Science269(5228):1262–1264.

32. Kondorosi E, Kondorosi A (2004) Endoreduplication and activation of the anaphase-promoting complex during symbiotic cell development. FEBS Lett 567(1):152–157.

33. Ishida T, et al. (2010) Auxin modulates the transition from the mitotic cycle to theendocycle in Arabidopsis. Development 137(1):63–71.

34. De Veylder L, et al. (2002) Control of proliferation, endoreduplication anddifferentiation by the Arabidopsis E2Fa-DPa transcription factor. EMBO J 21(6):1360–1368.

35. Coelho CM, et al. (2005) Cyclin-dependent kinase inhibitors in maize endosperm andtheir potential role in endoreduplication. Plant Physiol 138(4):2323–2336.

36. Verkest A, et al. (2005) The cyclin-dependent kinase inhibitor KRP2 controls the onsetof the endoreduplication cycle during Arabidopsis leaf development throughinhibition of mitotic CDKA;1 kinase complexes. Plant Cell 17(6):1723–1736.

37. Macleod KF, Hu Y, Jacks T (1996) Loss of Rb activates both p53-dependent andindependent cell death pathways in the developing mouse nervous system. EMBO J15(22):6178–6188.

38. Tsai KY, et al. (1998) Mutation of E2f-1 suppresses apoptosis and inappropriate Sphase entry and extends survival of Rb-deficient mouse embryos. Mol Cell 2(3):293–304.

39. Guo Z, Yikang S, Yoshida H, Mak TW, Zacksenhaus E (2001) Inactivation of theretinoblastoma tumor suppressor induces apoptosis protease-activating factor-1

dependent and independent apoptotic pathways during embryogenesis. Cancer Res61(23):8395–8400.

40. Jordan CV, Shen W, Hanley-Bowdoin LK, Robertson DN (2007) Geminivirus-inducedgene silencing of the tobacco retinoblastoma-related gene results in cell death and

altered development. Plant Mol Biol 65(1-2):163–175.41. Lee HO, Davidson JM, Duronio RJ (2009) Endoreplication: Polyploidy with purpose.

Genes Dev 23(21):2461–2477.42. Edgar BA, Orr-Weaver TL (2001) Endoreplication cell cycles: More for less. Cell 105(3):

297–306.43. Sugimoto-Shirasu K, Roberts K (2003) “Big it up”: Endoreduplication and cell-size

control in plants. Curr Opin Plant Biol 6(6):544–553.44. Chevalier C, et al. (2011) Elucidating the functional role of endoreduplication in

tomato fruit development. Ann Bot (Lond) 107(7):1159–1169.45. Webster M, Witkin KL, Cohen-Fix O (2009) Sizing up the nucleus: nuclear shape, size

and nuclear-envelope assembly. J Cell Sci 122(Pt 10):1477–1486.46. Schmidt EE, Schibler U (1995) Cell size regulation, a mechanism that controls cellular

RNA accumulation: consequences on regulation of the ubiquitous transcription

factors Oct1 and NF-Y and the liver-enriched transcription factor DBP. J Cell Biol128(4):467–483.

47. Bourdon M, et al. (2012) Evidence for karyoplasmic homeostasis during endore-duplication and a ploidy-dependent increase in gene transcription during tomato

fruit growth. Development 139(20):3817–3826.48. Cavalier-Smith T (2005) Economy, speed and size matter: Evolutionary forces driving

nuclear genome miniaturization and expansion. Ann Bot (Lond) 95(1):147–175.49. Hemerly AS, et al. (1993) cdc2a expression in Arabidopsis is linked with competence

for cell division. Plant Cell 5(12):1711–1723.50. Jasinski S, et al. (2002) The CDK inhibitor NtKIS1a is involved in plant development,

endoreduplication and restores normal development of cyclin D3; 1-overexpressingplants. J Cell Sci 115(Pt 5):973–982.

51. Schnittger A, Weinl C, Bouyer D, Schöbinger U, Hülskamp M (2003) Misexpression

of the cyclin-dependent kinase inhibitor ICK1/KRP1 in single-celled Arabidopsistrichomes reduces endoreduplication and cell size and induces cell death. Plant Cell

15(2):303–315.52. Lauria M, et al. (2004) Extensive maternal DNA hypomethylation in the endosperm of

Zea mays. Plant Cell 16(2):510–522.53. Gehring M, Bubb KL, Henikoff S (2009) Extensive demethylation of repetitive

elements during seed development underlies gene imprinting. Science 324(5933):1447–1451.

54. Hsieh TF, et al. (2009) Genome-wide demethylation of Arabidopsis endosperm.Science 324(5933):1451–1454.

55. Zemach A, et al. (2010) Local DNA hypomethylation activates genes in riceendosperm. Proc Natl Acad Sci USA 107(43):18729–18734.

56. Zilberman D, Gehring M, Tran RK, Ballinger T, Henikoff S (2007) Genome-wideanalysis of Arabidopsis thaliana DNA methylation uncovers an interdependencebetween methylation and transcription. Nat Genet 39(1):61–69.

57. Lister R, et al. (2008) Highly integrated single-base resolution maps of the epigenomein Arabidopsis. Cell 133(3):523–536.

58. Waters AJ, et al. (2011) Parent-of-origin effects on gene expression and DNAmethylation in the maize endosperm. Plant Cell 23(12):4221–4233.

59. Steward N, Kusano T, Sano H (2000) Expression of ZmMET1, a gene encoding a DNAmethyltransferase from maize, is associated not only with DNA replication in actively

proliferating cells, but also with altered DNA methylation status in cold-stressedquiescent cells. Nucleic Acids Res 28(17):3250–3259.

60. Johnston AJ, et al. (2010) Dosage-sensitive function of retinoblastoma related andconvergent epigenetic control are required during the Arabidopsis life cycle. PLoSGenet 6(6):e1000988.

61. Jullien PE, et al. (2008) Retinoblastoma and its binding partner MSI1 controlimprinting in Arabidopsis. PLoS Biol 6(8):e194.

62. Biniszkiewicz D, et al. (2002) Dnmt1 overexpression causes genomic hyper-methylation, loss of imprinting, and embryonic lethality. Mol Cell Biol 22(7):

2124–2135.63. van Zanten M, et al. (2011) Seed maturation in Arabidopsis thaliana is characterized

by nuclear size reduction and increased chromatin condensation. Proc Natl Acad SciUSA 108(50):20219–20224.

64. Zeimet AG, et al. (2011) DNA ploidy, nuclear size, proliferation index and DNA-hypomethylation in ovarian cancer. Gynecol Oncol 121(1):24–31.

65. Wu CY, Rolfe PA, Gifford DK, Fink GR (2010) Control of transcription by cell size.PLoS Biol 8(11):e1000523.

E1836 | www.pnas.org/cgi/doi/10.1073/pnas.1304903110 Sabelli et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1