A parthenogenesis gene of apomict origin elicitsembryo formation from unfertilized eggs in asexual plantJoann A. Conner, Muruganantham Mookkan1, Heqiang Huo2, Keun Chae, and Peggy Ozias-Akins3

Department of Horticulture, University of Georgia Tifton Campus, Tifton, GA 31973

Edited by Robert L. Fischer, University of California, Berkeley, CA, and approved August 5, 2015 (received for review March 24, 2015)

Apomixis is a naturally occurring mode of asexual reproduction inflowering plants that results in seed formation without theinvolvement of meiosis or fertilization of the egg. Seeds formedon an apomictic plant contain offspring genetically identical to thematernal plant. Apomixis has significant potential for preservinghybrid vigor from one generation to the next in highly productivecrop plant genotypes. Apomictic Pennisetum/Cenchrus species,members of the Poaceae (grass) family, reproduce by apospory.Apospory is characterized by apomeiosis, the formation of unre-duced embryo sacs derived from nucellar cells of the ovary and, byparthenogenesis, the development of the unreduced egg into anembryo without fertilization. In Pennisetum squamulatum (L.) R.Br.,apospory segregates as a single dominant locus, the apospory-specific genomic region (ASGR). In this study, we demonstrate thatthe PsASGR-BABY BOOM-like (PsASGR-BBML) gene is expressed inegg cells before fertilization and can induce parthenogenesis andthe production of haploid offspring in transgenic sexual pearl mil-let. A reduction of PsASGR-BBML expression in apomictic F1 RNAitransgenic plants results in fewer visible parthenogenetic embryosand a reduction of embryo cell number compared with controls.Our results endorse a key role for PsASGR-BBML in parthenogen-esis and a newly discovered role for a member of the BBM-likeclade of APETALA 2 transcription factors. Induction of partheno-genesis by PsASGR-BBML will be valuable for installing partheno-genesis to synthesize apomixis in crops and will have furtherapplication for haploid induction to rapidly obtain homozygouslines for breeding.

apomixis | parthenogenesis | AP2 transcription factor | BABY BOOM |Pennisetum

Apomixis is a naturally occurring process of asexual reproduc-tion that results in offspring that are genetically identical to the

mother plant. In nature, apomixis can be achieved through multipledevelopmental pathways, suggesting different molecular mecha-nisms have evolved to promote apomixis. Apomictic pathways havebeen categorized as adventitious embryony, a sporophytic type ofapomixis, or diplospory and apospory, two gametophytic forms ofapomixis. In gametophytic apomixis, a chromosomally unreducedembryo sac develops from the megaspore mother cell (diplospory)or from a nearby nucellar cell (apospory) in a process termedapomeiosis. Parthenogenesis, the development of the unreduced,unfertilized egg into an embryo, constitutes the second step ofthe apomictic process (1, 2).In Pennisetum squamulatum, apomixis is transmitted by a physi-

cally large, hemizygous, nonrecombining chromosomal region, theapospory-specific genomic region (ASGR) (3). Multiple copies ofthe PsASGR-BABY BOOM-like (PsASGR-BBML) gene residewithin the ASGR (4) and are postulated as strong candidate genesfor the apomictic function of parthenogenesis based on linkage tothe ASGR, strong conservation of ASGR-BBML sequences betweenapomictic Pennisetum/Cenchrus species (5, 6), a loss of the CcASGR-BBML genes in aCenchrus ciliarisASGR recombinant plant that haslost the ability to undergo parthenogenesis (7), and similarity of the

ASGR-BBML genes to BABY BOOM (BBM) genes of Arabidopsisand Brassica (8).The first BBM gene was identified as a transcript that was

induced in microspore cultures of Brassica napus (BnBBM) un-dergoing somatic embryogenesis (8). BBM genes are part of alarge gene family, the APETALA 2/ETHYLENE RESPONSEFACTOR (AP2/ERF) DNA-binding domain family. The AP2/ERFDNA-binding domain family has been identified in a wide groupof plants, including mosses, algae, gymnosperms, and angio-sperms. The AP2/ERF gene family is divided into ERF-like,which has a single AP2 domain, and AP2-like, which contains twoAP2 domains (repeat 1 and 2) that are similar to each other andseparated by a linker region. Although the AP2 domains arehighly conserved, the N-terminal and C-terminal sequences ofAP2 proteins are more distinct while still containing specificmotifs. The AP2-like clade is divided into eudicotAP2 andAINTEGUMENTA (ANT) lineages. The ANT lineage isdivided between basalANT and eudicotANT (euANT) lineagesthat contain specific motifs euANT1–4 (9). The euANT lineagecontains the PLETHORA-like (PLT), AINTEGUMENTA-like,AINTEGUMENTA-like1, AINTEGUMENTA-like5, and BBM-like subgroups. The BBM, BBM-like, and ASGR-BBM–like pro-teins share a conserved bbm-1 domain not identified in othermembers of the euANT lineage (10). Proteins within the euANTsubgroups play critical roles in meristem maintenance, cell pro-liferation, organ initiation and growth, somatic embryogenesis,

Significance

The molecular mechanisms controlling apomixis, a mode ofasexual reproduction in plants leading to clonal seed forma-tion, are largely unknown. In Pennisetum squamulatum, apo-mixis segregates as a single dominant locus, the apospory-specific genomic region (ASGR). The ASGR contains multiplecopies of the PsASGR-BABY BOOM-like (PsASGR-BBML) gene, amember of the BBM-like subgroup of APETALA 2 transcriptionfactors. Expression of a PsASGR-BBML transgene in sexualtetraploid pearl millet promoted both parthenogenesis (em-bryo formation without fertilization) and the production ofhaploid offspring. This study presents the first demonstration,to our knowledge, of function for a gene cloned from a natu-rally occurring apomictic plant that encodes a key componentcontrolling parthenogenesis.

Author contributions: J.A.C. and P.O.-A. designed research; J.A.C., M.M., H.H., and K.C.performed research; J.A.C., M.M., H.H., and K.C. analyzed data; and J.A.C. and P.O.-A.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1Present address: Division of Plant Sciences, University of Missouri, Columbia, MO 65211.2Present address: Department of Plant Sciences, University of California, Davis, CA 95616.3To whom correspondence should be addressed. Email: pozias@uga.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1505856112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1505856112 PNAS | September 8, 2015 | vol. 112 | no. 36 | 11205–11210

AGRICU

LTURA

LSC

IENCE

S

Dow

nloa

ded

by g

uest

on

May

29,

202

0

embryo differentiation, and root formation (11). Ectopic ex-pression of the BBM gene in Arabidopsis led to the formation ofsomatic embryos on seedlings (8), and deletion of the bbm-1domain eliminated the ability of transgenic plants to induce somaticembryogenesis on cotyledons in transformed Arabidopsis (10). Theinability to recover homozygous bbm/plt mutants from Arabidopsisand the early arrest of ∼25% of embryos from bbm/plt crossesindicated a redundant function in early embryogenesis for thesegenes (12).In the present study, we report ASGR-BBML promoter-GUS

activity at the cellular level in tetraploid sexual embryo sacs anddeveloping embryos of pearl millet and discover that trans-formation of tetraploid sexual pearl millet with an ASGR-BBMLtransgene, expressed under its native promoter and terminator,is able to promote parthenogenesis, embryo development with-out fertilization, in three independent transgenic lines. Offspringfrom one ASGR-BBML transgenic line was able to form diploid/dihaploid seedlings, which in turn can also produce haploidembryos based on embryo flow cytometry. Our results define apreviously unidentified role in development for an AP2-liketranscription factor, that of parthenogenesis or the initiation ofembryo development from an unfertilized egg cell. ASGR-BBMLwill find application for the synthesis of apomixis in crop plantsand for haploid induction to rapidly obtain homozygous linesfor breeding.

ResultsExpression of ASGR-BBML Genes. To date, three highly conservedgenomic duplications of PsASGR-BBML have been identifiedfrom ASGR-linked BACs p203, p207 (5), and p208. The p208PsASGR-BBML sequence is identical to the p207 PsASGR-BBML2 sequence (EU559277.1) except for the number of ATrepeats (11 vs. 17) found in intron 1. Promoter sequences of thethree PsASGR-BBML sequences are also highly conserved. Theconservation of transcribed sequence leaves in question which ofthe PsASGR-BBML genomic regions are expressed. Two ASGR-linked copies of ASGR-BBML are present in apomictic C. ciliaris(5). CcASGR-BBM-like1 (EU559278.1) is transcribed and 99.7%identical to the PsASGR-BBML across the ORF, whereasCcASGR-BBM-like2 (EU559279.1), also transcribed, containstwo nonsense mutations. Expression of ASGR-BBML has beenobserved by RT-PCR in apomictic P. squamulatum (PI 319196),C. ciliaris (B12-9), and Pennisetum glaucum backcross (BC) 7(06-63) and BC8 (06-A-58) lines (13) in unfertilized ovariesstarting 1 d before anthesis and assayed up to 2 d after anthesis,at which time ovaries without endosperm development begin tosenesce; in pollinated ovaries at day of anthesis through earlyseed development; in anthers 1 d before anthesis; and in roots.PsASGR-BBML is also expressed in embryogenic callus derivedfrom apomictic P. glaucum BC8 (06-A-58). Leaf tissue fromP. squamulatum and C. ciliaris does not show ASGR-BBML ex-pression; however, expression in leaf was seen in apomicticP. glaucum BC7 (06-63) and BC8 (06-A-58) lines. The PsASGR-BBML transcript encodes a 545-amino-acid protein derived fromthe splicing of eight exons, a 70-bp 5′UTR, and multiple 3′UTRs,with lengths ranging from 30 to 258 bp. The PsASGR-BBML genecontains two AP2 DNA-binding domains and thus is predicted tofunction as a transcription factor.

Reduction of PsASGR-BBML in Apomictic RNAi Lines. We firstattempted to ascertain the role of PsASGR-BBML in apomicticdevelopment by evaluating apomictic F1 transgenic lines carryingan RNAi construct containing the 3′ end of the PsASGR-BBMLgene. Based on Southern blot analysis, the region chosen forthe RNAi construct did not hybridize to pearl millet genomicDNA or to additional fragments other than those expected inP. squamulatum genomic DNA. As direct transformation andregeneration of apomictic P. squamulatum was not possible, an

alternative strategy and screening protocol for generating apo-mictic F1 plants with reduced expression of PsASGR-BBML us-ing the construct RNAi-BBM-3p was developed. Tetraploidsexual pearl millet lines (IA4X) were generated, via a biolisticparticle delivery system, which contained the RNAi construct.Eight independent RNAi lines were crossed with pollen fromP. squamulatum to produce segregating F1 plants. Seventy-fiveplants with genotypes ASGR/RNAi, ASGR/–, –/RNAi, and −/−

from six independent lines were kept from an initial screen of190 F1 seedlings. These plants/lines were screened for RNAitransgene expression in leaf tissue. Twenty-five plants from fiveindependent lines were chosen for detection of PsASGR-BBMLexpression in ovaries with genotypes ASGR/RNAi (14 plants),ASGR/– (five plants), and –/RNAi (six plants). Three of the 14ASGR/RNAi plants assayed, each derived from a different line,showed a significant reduction in PsASGR-BBML gene expres-sion based on semiquantitative analysis of PsASGR-BBML ex-pression at day of pollination (SI Appendix, Table S1 and Fig. S1)as compared against the control genotype ASGR/– plant. Thethree PsASGR-BBML reduced-expression F1 plants containedthe same percentage of ovules with aposporous embryo sac for-mation as the control plant (SI Appendix, Table S1). The plantswere pollinated with red tetraploid P. glaucum (Red-IA4X)pollen, and offspring were determined to be derived throughapomixis based on lack of red pigmentation of the midrib, uni-form phenotypes, and genotyping for the ASGR. Red-IA4Xplants are sexual tetraploid lines containing a dominant Rp1 al-lele, which confers a dark red pigmentation in the midrib andsheath of leaves (14). However, histological observation of ova-ries determined that the number of aposporous embryo sacsshowing parthenogenetic embryo development and the numberof parthenogenic embryos at or past the globular embryo stage2 d after anthesis without pollination were significantly reducedin the PsASGR-BBML reduced-expression lines (SI Appendix,Table S1 and Fig. S2). The identified correlation between reducedPsASGR-BBML expression and number of embryo sacs showingparthenogenesis and the diminished development of those par-thenogenic embryos suggested a potential role for PsASGR-BBMLin unfertilized embryo development.

Expression of PsASGR-BBML at the Cellular Level in Sexual OvariesBased on PsASGR-BBMLpromoter-GUS Lines. IA4X plants containingthe PsASGR-BBMLpromoter-GUS construct were generated, viaa biolistic particle delivery system, to evaluate the expression ofPsASGR-BBML at the cellular level. Sexual T1 offspring fromfive independent lines were initially evaluated. To preventunwanted fertilization, heads were bagged before stigmaemergence, and stigmas were mechanically removed from theindividual florets before pollen shed. Four sexual T1 individualswithin each line containing the PsASGR-BBMLpromoter-GUSconstruct were evaluated. Control plants were sexual, transgene-null siblings from two lines. GUS signal was detected in theegg/synergid complex in all offspring inheriting the PsASGR-BBMLpromoter-GUS construct in three of the five lines, whereasno GUS activity was identified in the control offspring. Two linesshowed very low or no GUS activity.A more stringent study using one line was undertaken to ad-

dress the possibility that pollination/fertilization triggered theGUS signal in the egg/synergid complex. Additional PsASGR-BBMLpromoter-GUS plants were germinated, genotyped for thePsASGR-BBMLpromoter-GUS insert, and isolated to a singlegreenhouse. Heads were continually removed before pollen shedto keep the greenhouse free of millet pollen. Approximately 60florets were collected from four plants 1 d before anthesis andemasculated, and the florets were placed under in vitro condi-tions. Another 15–20 ovaries were collected and placed in GUSstaining solution for the day before anthesis sampling. For thefollowing 3 d, 15–20 emasculated florets were removed from the

11206 | www.pnas.org/cgi/doi/10.1073/pnas.1505856112 Conner et al.

Dow

nloa

ded

by g

uest

on

May

29,

202

0

in vitro conditions, and the ovaries were dissected and placed inGUS staining solution. As shown in Fig. 1 for unfertilized sexualembryo sacs on the day of anthesis, the GUS signal could beobserved within the egg cell, with a weaker GUS signal observedin the synergids (Fig. 1 A–C). The synergid signal may be causedeither by a lower expression of the PsASGR-BBML gene in thesynergid cells or by leakage of the GUS signal from the egg cell.No GUS staining could be visualized in the central cell or an-tipodal cells of the sexual embryo sac (Fig. 1 A–C) or in thesurrounding ovary tissues. No embryo or endosperm develop-ment was identified in any emasculated, in vitro-cultured ovariesup to 2 d past anthesis. Florets that were pollinated and placed inthe same in vitro conditions as emasculated florets displayedboth endosperm and embryo development. GUS activity wasidentified in cells of developing embryos characterized up to 3 dafter fertilization (Fig. 1D) but not in developing endosperm.

Parthenogenesis in Transgenic Sexual Pearl Millet Lines Carrying thePsASGR-BBML Transgene Under Control of Its Endogenous Promoter.Nine independent transgenic lines (18 plants total) containingthe transgene gPsASGR-BBML were generated from IA4X via abiolistic particle delivery system. Four lines, consisting of sixplants, were not analyzed due to lack of flowering or demise ofthe plant. Approximately 50 ovaries of 12 individual plants de-rived from five independent lines were examined 2 d afteranthesis using a cleared-pistil technique and observation underdifferential interference contrast (DIC) or phase contrast opticsfor parthenogenesis (SI Appendix, Fig. S3). Fertilization wasprevented by bagging heads before stigma exsertion and the re-moval of stigmas/styles before anthesis. Structurally mature

embryo sacs containing the egg/synergid apparatus, polar nuclei,and antipodals, which indicate complete sexual embryo sac de-velopment, could be identified in all lines (Fig. 2A and SI Ap-pendix, Fig. S3). Three individuals from independent lines (g3f,g11a, and g52) exhibited parthenogenesis (Fig. 2 B and C) basedon the presence of embryo-like structures at the micropylar endof the sexual embryo sac and a lack of fertilization to createthose embryos based on the persistence of polar nuclei in thecentral cell and an absence of any endosperm development.Endosperm formation can be readily visualized in fertilizedembryo sacs in ovaries at the same developmental stage(SI Appendix, Fig. S3F). All three lines demonstrated endospermformation at day 2, when pollination was not prevented by re-moval of stigmas. A minimum of three heads and 100 ovariesfrom all heads were analyzed for g3f, g11a, and g52 plants alongwith offspring from an untransformed IA4X line. The percent-ages of structurally mature sexual embryo sacs (Fig. 2A) andembryo sacs containing parthenogenetic embryos at 2 d afteranthesis for lines g11a, g52, and g3f and untransformed offspring

Fig. 1. PsASGR-BBML expression in sexual embryo sacs. Ovaries from threesexual offspring derived from a T0 PsASGR-BBMLpromoter-GUS line areshown (A–D). The embryo sacs with antipodals have been outlined. PictureInsets are magnified regions of the egg/synergid/central cell region. TheUpper Inset in C is the next focal plane of the ovary to show the central cell.Two asterisks indicate the polar nuclei within the central cell. Arrows in-dicate synergids. GUS expression is detected in the egg cell of unfertilizedsexual embryo sacs on the day of anthesis. GUS signal is not detected in thecentral cell or antipodal cells of the mature sexual embryo sac. GUS stainingis detected in cells of the developing embryo (EM) 3 d after fertilization butnot in developing endosperm (D). No other staining in ovary tissue is iden-tified. (Scale bar, 50 μm.)

Fig. 2. Parthenogenetic embryo development in ovaries of sexual tetraploidpearl millet containing the gPsASGR-BBML transgene. Images are from unpol-linated ovaries collected and fixed 2 d after anthesis, cleared with methyl sa-licylate, and visualized using a Zeiss LSM 710 Confocal Microscope. Picture Insetsare magnified regions of the egg or embryo/synergid/polar nuclei region. Ar-rows indicate the polar nuclei within the central cell. (A) A control ovary with astructurally mature embryo sac without fertilization derived from an un-transformed tissue culture line. No embryo development is seen. Partheno-genesis in unfertilized ovaries is clearly seen in sexual transgenic line g3fcarrying the gPsASGR-BBML transgene based on the appearance of an embryo-like structure at the micropylar end of the embryo sac, polar nuclei, and an-tipodal cells (B and C). (D) Offspring from line g52 #308 carrying the gPsASGR-BBML transgene and also showing parthenogenesis. Antipodal cells have beencropped from the picture. (Scale bar, 50 μm.)

Conner et al. PNAS | September 8, 2015 | vol. 112 | no. 36 | 11207

AGRICU

LTURA

LSC

IENCE

S

Dow

nloa

ded

by g

uest

on

May

29,

202

0

are summarized in Table 1. Structurally mature sexual embryosacs ranged from 61% to 78% in the various lines. Parthenogenesisranged from 35% to 36% in lines g11a, g52, and g3f, whereas theuntransformed offspring showed no parthenogenesis. Expressionof the gPsASGR-BBML transgene was verified by RT-PCR withRNA extracted from open-pollinated ovaries 2 d after anthesisfor lines g52 and g3f (SI Appendix, Fig. S3). To rule out potentialploidy changes induced by tissue culture selection and regen-eration, the three T0 gPsASGR-BBML lines were subjectedto flow cytometry analysis (Fig. 3A). All three plants weretetraploid.Due to low germination rates and a low seed set for lines g11a

and g52, embryo rescue was used on the developing seed 10–15 dafter pollination and on the nongerminating mature seed to re-cover offspring from the three lines. Pollination with Red-IA4Xpollen over multiple heads and days was used to help compen-sate for potential pollen sterility of transgenic lines g11a and g52.Line g3f was also pollinated with Red-IA4X pollen, howeverpollen fertility of line g3f was much higher than the other twolines and self-fertilization of g3f lines also occurred. Plant g11aset a total of nine seeds, of which two offspring survived to green-house planting. Plant g52 set 97 seeds, of which 31 offspring survivedto greenhouse planting. Plant g3f set hundreds of seeds, a mix ofboth self- and cross-pollination, of which 194 were randomly selectedand 107 survived to greenhouse planting.All offspring were analyzed for inheritance of a 3,694-bp

amplicon that covers the gPsASGR-BBML ORF starting 5 bpdownstream from the start codon and amplifying into the3′ UTR (ORF amplicon). The two g11a offspring did not inheritthe transgene and were not further analyzed. All offspring fromg52 showed red pigmentation of the midrib and thus were de-rived from the fertilization of g52 sexual embryo sacs with Red-IA4X pollen. Nine offspring carried at least one copy of the ORFamplicon. The g3f offspring were a mix of both green and redpigmentation of the midrib, indicating both self- and cross-pol-lination of the ovaries and/or the production of haploid offspringfor this line. Twenty-six g3f offspring carried at least one copy ofthe ORF amplicon.All offspring from line g52 were assayed for parthenogenesis

(Table 1, averaging all offspring; SI Appendix, Table S2, data foreach individual; Fig. 2D and SI Appendix, Fig. S3 B and D). Thepercentage of structurally mature sexual embryo sacs for g52offspring with or without inheriting the ORF amplicon averaged77% and 78%, respectively. Only offspring inheriting the ORFamplicon showed parthenogenesis 2 d after anthesis. No embryoformation was identified in 791 structurally mature sexual embryosacs from 21 g52 offspring not inheriting the ORF amplicon.Ninety total sexual embryo sacs displayed embryo developmentand polar nuclei from the nine g52 offspring inheriting the ORFamplicon (Table 1). The percentage of ovaries showing parthe-nogenesis ranged from 12% to 53% in the different g52 offspring(SI Appendix, Table S2).A subset of offspring generated from line g3f were assayed for

parthenogenesis (Table 1, averaging all offspring; SI Appendix,Table S3, data for each individual; SI Appendix, Fig. S3 A and C).The percentage of structurally mature sexual embryo sacs for g3foffspring with or without inheriting the ORF amplicon averaged66% and 71%, respectively. No embryo formation was identifiedin a total of 951 structurally mature sexual embryo sacs from 28g3f offspring not inheriting the ORF amplicon. Three of the 26offspring inheriting the ORF amplicon could not be assayed forparthenogenesis, as they did not flower. Of the remaining 23offspring inheriting the ORF amplicon, 19 displayed partheno-genesis, whereas 4 did not. The percentage of structurally matureembryo sacs in which parthenogenesis was observed ranged from1% to 52% for the different g3f offspring. Additional ovariesover the initial screen of ∼50 ovaries were analyzed from the fourlines, with a total of 79, 118, 82, and 77 structurally mature sexual

embryo sacs screened for plants 123, 144, 159, and 183, re-spectively. Plants 123, 144, and 159 were assayed for transgeneexpression using RT-PCR analysis (SI Appendix, Fig. S5). In allthree offspring, gPsASGR-BBML expression was detected inunpollinated ovaries at the day of anthesis. Two transgene-spe-cific amplicons covering the gPsASGR-BBML cDNA transgenefrom the 5′ UTR through the 3′ UTR were sequenced fromplants 123, 144, and 159 along with plant 105, which displayedparthenogenesis. All sequences were identical to PsASGR-BBML cDNA sequences derived from BC7 (06-63) and BC8 (06-A-58) apomictic plants.As g3f offspring were a mix of both red and green pigmenta-

tion of the midrib, determination of the ploidy level of the greenphenotypes was required (SI Appendix, Table S4). Six offspringwere diploid/dihaploid in genome size based on flow cytometricanalysis using sorghum as the genome size reference (Fig. 3 Band C). The diploid/dihaploid offspring were further confirmedby mixing predicted diploid/dihaploid and tetraploid offspringtogether to generate three peaks (Fig. 3D). All six dihaploidoffspring carry the ORF amplicon, and of the four that flowered,all displayed parthenogenesis.

Embryo Ploidy of T1 Seed. Embryos dissected from mature seeds ofuntransformed IA4X, g3f offspring 104 (tetraploid with ORFamplicon/parthenogenesis) and 105 (diploid/dihaploid with ORFamplicon/parthenogenesis), and g3f offspring 325 (tetraploidwith ORF amplicon/parthenogenesis) were analyzed in pools offive embryos using flow cytometry. No reduction of ploidy levelwas identified in nine embryo pools (45 total embryos) whenuntransformed IA4X embryos were analyzed. Diploid/dihaploid

Fig. 3. Flow cytometry analysis to determine genome size of T0 plants andoffspring. Examples of genome size analysis using a BD-Accuri flow cytom-eter of T0 plants and g3f offspring (T1) are shown. (A) Peak analysis of sor-ghum and T0 line g11a leaf tissue. (B) Peak analysis of sorghum and g3foffspring 108. (C) Peak analysis of sorghum and g3f offspring 101. (D) Peakanalysis of g3f offspring 105 and 107. S2 and S4 designate sorghum 2n/2x/2cand 2n/2x/4c peaks, respectively. H2 and H4 designate T1 diploid/dihaploidoffspring (C and D) with 2n/2x/2c and 2n/2x/4c peaks, respectively. T2 and T4designate tetraploid T0 pearl millet (A) or tetraploid T1 offspring (B and D)with 2n/4x/2c and 2n/4x/4c peaks, respectively.

11208 | www.pnas.org/cgi/doi/10.1073/pnas.1505856112 Conner et al.

Dow

nloa

ded

by g

uest

on

May

29,

202

0

peaks were identified in three out of four embryo pools (20 totalembryos) from embryos collected from the tetraploid 104 seed.Haploid peaks were identified in three out of six embryo pools(30 embryos total) from embryos collected from the diploid/dihaploid 105 seed (SI Appendix, Fig. S6 A–C). Diploid/dihaploidpeaks were identified in three out of nine embryo pools (45 totalembryos) from embryos collected from the tetraploid 325 seed.

DiscussionWe initially attempted to analyze the cellular expression andfunction of the PsASGR-BBML gene in apomictic plants usingRNA in situ hybridization and by creating an RNAi apomicticline where expression of the PsASGR-BBML was completelyknocked down. Unfortunately, signal from RNA in situ hybrid-ization of ovaries 1 d before natural anthesis was too weak toreliably ascertain cellular localization for this low-abundancetranscript that we had shown by RT-PCR to be expressed at thistime point. We could detect in situ hybridization signals forPsASGR-BBML in developing apomictic embryos at the globularstage. The lack of an F1 aposporous transgenic line showing com-plete knockdown of the PsASGR-BBML in the RNAi experimentalso precluded us from determining the role of PsASGR-BBML inapomixis. However, expression of PsASGR-BBML in ovariesbefore embryo development based on RT-PCR analysis and thesignificant reduction in number of ovaries showing parthenoge-netic embryo development along with the degree of embryo de-velopment observed at 2 d postanthesis in the F1 PsASGR-BBMLreduced-expression lines suggested that additional work on thePsASGR-BBML gene was warranted, as some role in embryo de-velopment seemed likely.We therefore generated tetraploid sexual transgenic pearl

millet lines containing the PsASGR-BBMLpromoter-GUS con-struct to test for cellular expression of PsASGR-BBML. Thisexperiment allowed us to determine that expression from thePsASGR-BBML promoter is restricted to the egg cell and pos-sibly synergids of unfertilized sexual embryo sacs. Expression inthe egg cell before fertilization would be a requirement of aparthenogenesis gene. Expression of the PsASGR-BBML pro-moter-GUS signal continues within cells of the developing em-bryo and confirms RT-PCR results that show expression ofPsASGR-BBML in the developing seed from an apomict up to5 d after pollination.Knowing the ∼2.1-kb PsASGR-BBML promoter drove ex-

pression of the GUS gene in unfertilized egg cells, we nexttested whether the PsASGR-BBML gene regulated by this en-dogenous promoter could induce parthenogenesis in sexual pearlmillet plants expressing PsASGR-BBML. We identified three

independent IA4X lines carrying the PsASGR-BBML transgenethat clearly show parthenogenic embryo development in unfer-tilized ovaries 2 d after anthesis in sexual tetraploid pearl millet,plants that do not normally display this trait. Of the three orig-inal lines displaying parthenogenesis, two produced viable off-spring inheriting the transgene. From lines g52 and g3f, 32% and55% of rescued embryos survived to greenhouse planting andinherited the transgene ORF (29% and 24% transmission, re-spectively). The low level of embryo rescue/seedling survivalprecludes the use of Mendelian genetics to identify inheritancepatterns of the transgene(s). However, the generation of off-spring with the dominant Rp1 allele and the ORF amplicondemonstrates that the transgene can be transmitted through fe-male meiosis. Although the three original transgenic plants dis-played ∼35% parthenogenesis in unfertilized embryo sacs whenassayed by ovule clearing 2 d after natural pollen release, parthe-nogenesis in offspring with the ORF amplicon ranged from 0% to∼50%. There are many factors that could be implicated for theincomplete penetrance of parthenogenesis within the original T0lines and T1 offspring of lines g3f and g52. Variation of transgeneexpression from independent lines produced using the sametransgene construct has been widely reported and summarized(15). Factors affecting transgene expression levels includetransgene copy number, the complexity of integration patterns,and transgene integration sites. In addition, transgene segrega-tion within the sexual embryo sacs of both T0 and T1 lines;transgene modifications, such as methylation status, that couldalter the expression level and/or timing of expression of thetransgene; and/or unknown genetic factors interacting withPsASGR-BBML that enhance the protein’s ability to promoteparthenogenesis would be variable in both the T0 and the T1plants given that the starting transformation material was theheterozygous IA4X line. Transcription of the transgene wasidentified in the occasional offspring not demonstrating par-thenogenesis. If penetrance of parthenogenesis was very low(<1.0%), the number of ovaries screened would not have been ofsufficient number to detect parthenogenesis. Quantifying levelsof transgene expression to the amount of parthenogenesisidentified in the various lines is difficult due to variation in thepercentage of structurally mature sexual embryo sacs foundwithin each plant and the inability to verify that RNA is beingextracted from ovaries that are at the same developmental stage.The generation of homozygous lines containing a single copyof the gPsASGR-BBML transgene or the expression of thegPsASGR-BBML transgene using egg-specific promoters withdifferent expression levels would be advantageous to help ad-dress these questions.

Table 1. Summary of visual determination of parthenogenesis in cleared ovaries 2 d after anthesis in untransformed T0 and T1 lines

Plant designation(number of plantsor heads analyzed)

Genotypeof plantsORF +/−

Number ofovaries* withparthenogenic

embryos, distinctpolar nuclei,no endosperm

Number ofovaries* withparthenogenicembryos, nodistinct polarnuclei, noendosperm

Number ofovaries* withoutparthenogenic

embryodevelopment

Number ofovaries without

structurallymature sexualembryo sacs

Percentovaries with

structurally maturesexual embryo

sacs

Percent ovaries*displaying

parthenogenesis

Untransformed (9) – 0 0 392 94 78 0g11a (4 heads) + 29 0 53 53 61 35g3f (5 heads) + 59 0 110 81 68 35g52 (3 heads) + 27 0 49 33 70 36g52 offspring (21) – 0 0 791 229 78 0g52 offspring (9) + 90 15 268 109 77 28g3f offspring (28) – 0 0 951 392 71 0g3f offspring (23) + 122 23 903 530 66 14

*Ovaries with structurally mature sexual embryo sacs.

Conner et al. PNAS | September 8, 2015 | vol. 112 | no. 36 | 11209

AGRICU

LTURA

LSC

IENCE

S

Dow

nloa

ded

by g

uest

on

May

29,

202

0

We recovered six diploid/dihaploid plants from 170 seedlingsof line g3f. No diploid/dihaploid seedlings were recovered fromthe original g52 T0 line; however, embryo pools of g52 offspring325 were found to produce diploid/dihaploid embryos. Severalfactors could be responsible for the low recovery of diploid/dihaploid seedlings from ovaries showing parthenogenesis 2 dafter anthesis without fertilization in these plants. When theembryo develops precociously in the absence of fertilization, itlikely impedes fertilization of the central cell and endospermformation, thus leading to nonviable seed development. Mostlines with the ORF amplicon showed some level of reducedmature seed set and produced both plump and shriveled seedsfrom the various heads. Health and morphology of the g3f dip-loid/dihaploid offspring varied greatly, with two plants failing toflower. However, ploidy level does not seem to be critical forPsASGR-BBML transgene-induced parthenogenesis, as the fourchromosomally reduced diploid/dihaploid offspring derived fromthe g3f line that did flower showed a similar range of parthe-nogenesis levels as the unreduced tetraploid offspring also derivedfrom the same line. In addition, using embryo flow cytometry,we were able to detect haploid embryo formation in seed set byg3f diploid/dihaploid offspring 105. Although the majority ofnatural apomicts are polyploid, natural and experimentally de-rived apomictic diploids/dihaploids from several species havebeen identified. Polyploid apomictic plants can produce diploid/dihaploid offspring either through the genetic recombinationof the apomeiosis and parthenogenesis loci (16) or throughthe parthenogenetic development of a reduced egg carrying anapomixis locus (17).Phylogenetically, the ASGR-BBML proteins cluster within a

small group of other BBM-like proteins derived from monocotlineages including Setaria italica, Panicum virgatum, and Oryzasativa, but not with BBM-like proteins from Zea mays or Sor-ghum bicolor, which have genomes evolutionarily less divergedfrom Pennisetum than O. sativa (SI Appendix, Figs. S7 and S8).The function of the genes within this small group, except forPsASGR-BBML, is unknown. Expression of rice BBM1 is similar

to that of PsASGR-BBML, except its expression in egg cells hasnot been tested to our knowledge (SI Appendix, Fig. S9). It willbe interesting to discover if other members of this clade canpromote parthenogenesis if regulated by the PsASGR-BBMpromoter used in this experiment or if an uncharacterized do-main in the ASGR-BBML protein is required to promote par-thenogenesis. Although members of the BBM-like clade of AP2transcription factors have noted roles in somatic embryogenesisand cell proliferation, our results have uncovered a role forPsASGR-BBML in parthenogenesis. This newly discovered rolecould have a major impact on the ability to geneticallyengineer apomixis into crop species and could be used asan alternative method for haploid induction to rapidly obtainhomozygous lines for breeding.

Materials and MethodsP. squamulatum (PS26; PI 319196) and C. ciliaris (B12-9) are obligate apomictplants used in genetic studies of apomixis (3, 18). P. glaucum BC7 (06-63) (13),P. glaucum BC8 (06-A-58) (13), and P. glaucum BC8-line 63 are facultativeapomictic plants derived from double cross hybrids (19) and BCing withsexual tetraploid P. glaucum (IA4X). IA4X is a hybrid between an African andIndian line that was doubled with colchicine. Red-IA4X is derived from IA4Xcrossed with a tetraploid plant containing the Rp1 dominant marker (14).

Other methods such as DNA/RNA isolation, PCR primers and reactions(SI Appendix, Table S5), RT-PCR, RACE protocol for cDNA ends, trans-formation constructs and production of transgenic lines, sequence analysisof the PsASGR-BBML transgene, flow cytometry, chromosomal root tip counts,X-Gluc staining for β-Glucuronidase activity, semiquantitative expression andhistological analysis of PsASGR-BBML RNAi lines, and phylogenetic analysis canbe found in SI Appendix, SI Materials and Methods.

ACKNOWLEDGMENTS. We are grateful to Karen Harris-Shultz and JohnRuter for access to their flow cytometry equipment for collection of ourdata, Greg Thomas for help with ovary dissections, Muthugapatti K.Kandasamy and the Biomedical Microscopy Core at the University of Georgiafor imaging using the Zeiss LSM 710 confocal microscope, Tracey Vellidis forpreparing figures, and Evelyn Saul and Gunawati Gunawan for providingtechnical assistance. This work was supported by National Institute of Foodand Agriculture Grant 2010-65116-20449 and National Science FoundationAward DBI-0115911 (to P.O.-A. and J.A.C.).

1. Hand ML, Koltunow AM (2014) The genetic control of apomixis: Asexual seed for-mation. Genetics 197(2):441–450.

2. Barcaccia G, Albertini E (2013) Apomixis in plant reproduction: A novel perspective onan old dilemma. Plant Reprod 26(3):159–179.

3. Ozias-Akins P, Roche D, HannaWW (1998) Tight clustering and hemizygosity of apomixis-linked molecular markers in Pennisetum squamulatum implies genetic control of apos-pory by a divergent locus that may have no allelic form in sexual genotypes. Proc NatlAcad Sci USA 95(9):5127–5132.

4. Gualtieri G, et al. (2006) A segment of the apospory-specific genomic region is highlymicrosyntenic not only between the apomicts Pennisetum squamulatum and buffel-grass, but also with a rice chromosome 11 centromeric-proximal genomic region.Plant Physiol 140(3):963–971.

5. Conner JA, et al. (2008) Sequence analysis of bacterial artificial chromosome clones from theapospory-specific genomic region of Pennisetum and Cenchrus. Plant Physiol 147(3):1396–1411.

6. Akiyama Y, et al. (2011) Evolution of the apomixis transmitting chromosome inPennisetum. BMC Evol Biol 11:289.

7. Conner JA, Gunawan G, Ozias-Akins P (2013) Recombination within the aposporyspecific genomic region leads to the uncoupling of apomixis components in Cenchrusciliaris. Planta 238(1):51–63.

8. Boutilier K, et al. (2002) Ectopic expression of BABY BOOM triggers a conversion fromvegetative to embryonic growth. Plant Cell 14(8):1737–1749.

9. Kim S, Soltis PS, Wall K, Soltis DE (2006) Phylogeny and domain evolution in theAPETALA2-like gene family. Mol Biol Evol 23(1):107–120.

10. El Ouakfaoui S, et al. (2010) Control of somatic embryogenesis and embryo devel-

opment by AP2 transcription factors. Plant Mol Biol 74(4-5):313–326.11. Horstman A, Willemsen V, Boutilier K, Heidstra R (2014) AINTEGUMENTA-LIKE pro-

teins: Hubs in a plethora of networks. Trends Plant Sci 19(3):146–157.12. Galinha C, et al. (2007) PLETHORA proteins as dose-dependent master regulators of

Arabidopsis root development. Nature 449(7165):1053–1057.13. Singh M, et al. (2010) Characterization of apomictic BC7 and BC8 pearl millet: Meiotic

chromosome behavior and construction of an ASGR-carrier chromosome-specific li-

brary. Crop Sci 50(3):892–902.14. Hanna WW, Burton GW (1992) Genetics of red and purple plant color in pearl millet.

J Hered 83(5):386–388.15. Butaye K, Cammue B, Delauré SL, De Bolle M (2005) Approaches to minimize variation

of transgene expression in plants. Mol Breed 16(1):79–91.16. Noyes RD, Wagner JD (2014) Dihaploidy yields diploid apomicts and parthenogens in

Erigeron (Asteraceae). Am J Bot 101(5):865–874.17. Dujardin M, Hanna W (1986) An apomictic polyhaploid obtained from a pearl

millet x Pennisetum squamulatum apomictic interspecific hybrid. Theor Appl Genet

72(1):33–36.18. Sherwood RT, Berg CC, Young BA (1994) Inheritance of apospory in buffelgrass. Crop

Sci 34(6):1490–1494.19. Dujardin M, Hanna WW (1989) Developing apomictic pearlmillet–Characterization of

a BC3 plant. J Genet Breed 43(3):145–151.

11210 | www.pnas.org/cgi/doi/10.1073/pnas.1505856112 Conner et al.

Dow

nloa

ded

by g

uest

on

May

29,

202

0