Objectives

Objective 1 (Yr 1, m1 to Yr 3, m12)
Fertilization-independent (FI) formation of rice endosperm and pericarp
Output 1.1 Refined FI pericarp formation based on understanding of OsAsp1 function (I)
Output 1.2 Refined FI endosperm formation based on RNAi of OsFIS genes (C)
Output 1.3 Line combining FI pericarp formation and FI endosperm formation (I + C)
Objective 2 (Yr 1, m1 to Yr 3, m12)
FI embryogenesis in rice nucellus
Output 2.1 Isolated rice orthologue of maize Mac1 gene (I)
Output 2.2 Isolated rice orthologues of Arabidopsis CLV, WUS, STM and AG genes (C + I)
Output 2.3 System to control genes of nucellus from megaspore mother cell (MMC1) (I)
Output 2.4 OsMac1-based system for inducing second MMC (MMC2) (I)
Output 2.5 WUS-induced FI embryogenesis in MMC2 (C +I)
Output 2.6 WUS-induced FI embryogenesis in nucellar cluster proximal to MMC1 (C)
Objective 3 (Yr 4, m1 to Yr 5, m12)
Apomictic hybrid rice
Output 3.1 Line displaying FI formation of embryo, endosperm and pericarp (I + C)
Output 3.1 Line in which endosperm feeds apomictic embryo during germination (I + C)
Output 3.3 One-line apomictic hybrid rice (I + C)
Objective 4 (Yr 1, m1 to Yr 5, m12)
Communication and dissemination of research results
Output 4.1 Publications and website for pre-publication releases (Yr 1, m3)
Output 4.2 Participation in Third International Symposium on Apomixis (Yr.3, m?)
Output 4.3 Mid-term review at CSIRO (Yr 3, m10)
Output 4.4 Participation in Fourth International Congress on Hybrid Rice (Yr 4, m?)
Output 4.5 Terminal workshop at IRRI (Yr 5, m10)

Note: Only those outputs scheduled for initiation in Year 1 are mentioned.
Note: (I) = IRRI activity, (C) = CSIRO activity

Objective 1 (7/'03-6/'06)
Fertilization-independent (FI) formation of rice endosperm and pericarp
Output 1.1 Refined FI pericarp formation based on understanding of OsAsp1 function (I)
OsAsp1 encodes a putative aspartate proteinase that is unusual among plant aspartate proteinases in lacking the Plant Specific Insert. When an OsAsp1:gus promoter:reporter gene fusion was transferred into rice using Agrobacterium, a form of parthenogenesis was observed (FI expansion of the pericarp and seed coat, with a clear aqueous solution instead of embryo and endosperm). This ability to by-pass the checkpoint that normally prevents ovary expansion in the absence of fertilization may be very important in relation to achieving full FI endosperm expansion (see Outputs 1.2 and 1.3). We have now shown that recombinant OsAsp1 expressed in E. coli does indeed encode an active proteinase, which is capable of not only hydrolyzing artificial substrates but also removing its own N-terminal pro-sequence.
Output 1.2 Refined FI endosperm formation based on RNA-I of OsFIS genes (C)
Mutations in Arabidopsis FIS genes produce the FI seed phenotype. We introduced an RNA-I against a rice gene (FIS3 from rice, or FIR) homologous to Arabidopsis FIS3 gene. Transgenic lines harbouring the FIR-RNA-I contained shrivelled grains, similar to fertilized fis3 mutant lines. Autonomous grain-like structures can be seen in the transgenic lines amongst empty grains at a 1:1 ratio. More work is needed in the FIR-RNA-I line to actually demonstrate autonomous growth of endosperm in these lines. The observation of empty pericarp in FIR-RNA-I lines could have implications for Output 1.3. Similar to work was reported from IRRI earlier (Output 1.1). We also observed FI pericarp structure in a tagged line isolated in CSIRO by Dr. Narayana Upadhyaya and colleagues. Cloning DNA adjacent to the insertion site identifies the gene as a small dehydrogenase/hydrolase in Chromosome 9. In Arabidopsis we showed that an antisense construct against the DNA methyltransferase (MET) gene causes FI empty seed formation. We have made RNA-I against the rice MET1 gene and the construct is about to be introduced into rice. Recent sequencing data from rice also indicate that there are 2 FIS3 like genes in rice.

Objective 2 (7/'03-6/'06)
FI embryogenesis in rice nucellus
Output 2..1 Isolated rice orthologue of maize Mac1 gene (I)
This work was accomplished by the Kurata labortatory of the National Institute of Genetics in Mishima, Japan. See Nonomura et al. (2003) Plant Cell. 15:1728-1739. As predicted in the Phase 1 grant proposal, Msp1, the rice orthologue of maize Mac1, is a leucine-rich repeat receptor-like protein kinase.
Output 2..2 Isolated rice orthologues of Arabidopsis CLV, WUS, STM and AG genes (C + I)
Putative rice orthologues of WUS and AG have been identified at IRRI. Their functional identities are now being confirmed. STM and AMP1 homologues have been identified are being isolated in Canberra.
Output 2..3 System to control genes of nucellus from megaspore mother cell (MMC1) (I)
We have identified 2 rice genes that are putative ligands of Msp1 and that may control the behavior of surrounding nucellar cells through interaction with Msp1. RNA in situ hybridization and RNA-I are being used to confirm and characterize the putative ligands. These genes will be tested for their ability to constitute a system for controlling gene expression in nucellar cells from the MMC. If successful, the system will be used to drive nucellar embryogenesis at the time of meiosis.
Output 2..4 OsMac1-based system for inducing second MMC (MMC2) (I)
The phenotype of Tos17 knock-outs of Msp1 is multiple sporocytes, including multiple MMCs. This is a good start towards generating just one additional MMC (MMC2), but we need to exert finer control than is possible with knock-outs. The system discussed under Output 2..3 will be modified to achieve this result.

Objective 4 (7/'03-6/'08)
Communication and dissemination of research results
Output 4.1 Publications and website for pre-publication releases (10/'03)
This activity has been postponed until 1/'05 because of the need to complete publication of Phase 1 data and obtain approval for release of Phase 2 data that are currently confidential.

[Note: Only outputs scheduled for initiation in Years 1-3 are mentioned. Activities: (I) = IRRI, (C) =
CSIRO.]
Our goal is to make the benefits of hybrid rice available to poor farmers through asexual seed
production. At present, farmers must buy expensive hybrid seed fresh every season, because its
heterozygosity is diminished by 50% with every cycle of sexual reproduction. We propose to develop
a synthetic form of apomixis that will fix the heterozygosity of the hybrid and allow farmers to
reproduce hybrid seeds cheaply in their own fields. Although essentially all natural apomicts are
polyploids, we expect our synthetic apomixis to be compatible with diploidy in rice.
We shall model our synthetic apomixis on the apospory of Poa pratensis, a relative of wheat and
barley in which only 5 genetic loci are required for apospory [Matzk et al. (2005) Plant Cell 17:13]24].
The key milestone will be the generation of an aposporous embryo from diploid cells in the nucellus
of the ovary. We shall convert 1-2 cells of the nucellus of rice into the equivalent of aposporous initials
(AIs) and assess their capacity to develop firstly into diploid embryo sacs and then into aposporous
embryos.
The conversion of nucellar cells into aposporous initials will take advantage of a recently
discovered gene (MSP1) that limits the number of nucellar cells able to become megaspore mother
cells (MMCs). This gene was inactivated by Tos17 insertional mutation. When the mutant was
recovered in the homozygous form (msp1 msp1), up to 15 secondary MMCs form in the ovule and
develop into haploid embryo sacs. We propose to inactivate MSP1 more subtly by RNA interference
(RNAi) to allow only 1-2 secondary MMCs to form. At the same time we shall prevent these cells from
entering meiosis, so that they form the equivalent of diploid AIs. We shall assess the conditions under
which the putative AIs form diploid embryo sacs and aposporous embryos. Objective 2 focuses on
generating an apomictic embryo by the above approach.
Apomictic embryogenesis may be pursued with or without fertilization. The fertilizationdependent
approach is in some respects more straightforward because formation of the endosperm
normally requires fertilization, even in many apomicts, and the seed coat and the pericarp, although
maternal tissues, usually fail to develop fully without fertilization. Furthermore, the fertilizationdependent
approach would produce the usual triploid endosperm rather than a diploid form. We
have opted here for the fertilization]independent (FI) approach because (1) it eliminates competition
between the sexual embryo and the apomictic embryo, and (2) it avoids controversy about the escape
of apomixis genes into the environment. However, it will be necessary to have a transient sexual
phase for the initial formation of the apomictic hybrids in breedersf fields; the FI phase can be used in
farmersf fields. Objective 1 focuses on the FI protocol.
Objective 1 (7/ 03-6/ 06)
Fertilization]independent (FI) formation of rice endosperm and pericarp
.. Output 1.1 Refined FI pericarp formation based on understanding of OsAsp1 function (I)
Both IRRI and CSIRO have engineered methods for by]passing the fertilization]dependence of
pericarp and seed coat enlargement. This by-pass is essential for achieving FI-independent
seed production, because it removes a physical constraint on the size of the endosperm that
can develop autonomously.
.. Output 1.2 Refined FI endosperm formation based on RNA]I of OsFIS genes (C)
FI endosperm formation in rice is essential for constructing a form of apomixis in which the
escape of transgenic pollen is prevented. CSIRO has a method of generating fertilisationindependent
formation of the endosperm in Arabidopsis, based on the discovery through
mutation of three Fertilization-Independent Seed (FIS) genes. This method has been introduced
into rice. Based on sequence similarity, all the FIS like genes in rice had been identified, and
transgenic lines with silencing constructs for each OsFIS gene are being evaluated.

Objective 2 (7/ 03-6/ 06)
FI embryogenesis in rice nucellus

.. Output 2.1 Isolated rice orthologue of maize MAC1 gene (I)
We are focusing on developing a synthetic form of apospory for rice. Aposporous embryos
develop from nucellar cells near the megaspore mother cell (MMC). The maize mutation
multiple archaesporial cells1 (mac1) showed that such nucellar neighbours can develop into
secondary MMCs if the MAC1 gene is inactivated. One of IRRIfs objectives was to clone the
rice orthologue, OsMAC1. However, this task was actually accomplished by Nonomura et al.
(2003) in Japan [Plant Cell. 15:1728]1739], and they named the rice gene MULTIPLE
SPOROCYTES1 (MSP1).
To identify the MSP1 promoter, Ms Xinai Zhao of IRRI examined the region 5 kb upstream
from the MSP1 transcription start site defined by Nonomura et al. (2003). She found that the
nearest upstream gene, which had been predicted by annotation software to be transcribed in
the opposite direction from MSP1, was in fact transcribed in the same direction as MSP1. As
the distance between the two genes was very small, this finding raised the possibility that the
upstream gene was really part of the MSP1 gene; either the transcriptional start site for MSP1
might be further upstream than previously thought, or there might be multiple start sites. Ms
Zhao has now assembled considerable data to support the idea that the MSP1 gene has a very
long 5f]untranslated region. She is currently conducting RNA gel blots and complementation
studies of the homozygous msp1 mutant to identify the promoter and use it in RNAi to control
MSP1 transcript levels.
A key paper on the genetics of the control of apomixis and sexuality in Poa pratensis was
published by Matzk et al. (2005) Plant Cell 17:13-24. In this paper, the authors identified a
locus APV (Apospory Prevention) and speculated that it may be the Poa ortholog of MAC1.
They identified two other loci governing the initiation of apospory (AIT, Apospory Initiation,
and mdv, Megaspore Development) controlling apomeiosis and the initiation of the sexual
pathway. IRRI is studying candidate genes as orthologs of these loci. Matzk et al. (2005)
identified two loci controlling parthenogenesis in Poa, ppv (parthenogenesis prevention) and
PPI (Parthenogenesis Initiation). In Arabidopsis, a newly identified Polycomb protein that
interacts with FIS complex, AtMSI1 may code for ppv (see section 4.2 for details). CSIRO
isolated the OsMSI1 and testing the alternative way to induce parthenogenesis in rice.

.. Output 2.2 Isolated rice orthologues of Arabidopsis CLV, WUS, STM and AG genes (C & I)
Embryogenesis in the rice nucellus will probably require re-activation of meristematic activity
through expression of orthologues of such Arabidopsis genes as CLAVATA1]3, WUSCHEL,
SHOOT MERISTEMLESS and AGAMOUS. IRRI and CSIRO identified orthologues by
sequence comparisons. In the last year, Mr. Rico Gamuyao of IRRI examined 10 members of
the rice WOX gene family to ascertain their expression patterns, using reverse transcriptionpolymerase
chain reaction (RT-PCR) and RNA in situ hybridization. He found that two genes
were highly expressed in the ovule. One of these two genes, the one most closely related in
sequence to WOX11 and WOX12 of Arabidopsis, was highly expressed also in anthers, root
tips and the shoots of 2-day-old germinated seeds. The other gene, most closely related in
sequence to WOX9 of Arabidopsis, was highly expressed only in the ovule (nucellus and
integuments) but was apparently not expressed in the MMC. In Arabidopsis, WOX9 appears
to act by maintaining cell division and preventing premature differentiation [Wu et al. (2005)
Curr Biol 15: 436-]440]. We are studying whether the rice gene has a similar function
principally in the ovule.

.. Output 2.3 System to control genes of nucellus from megaspore mother cell (MMC1) (I)
Leucine]rich receptor kinases like MSP1 are usually triggered by the binding of extracellular
ligands to the LRR domain. We are attempting to identify the ligand of MSP1 and its site of
synthesis. We are guided in this search by the finding that a LRR receptor kinase is implicated
in the control of tapetum development by pollen mother cells, PMC, in Arabidopsis. EXCESS
MALE SPOROCYTES1 (EMS1, also known as EXS) encodes a LRR receptor kinase and is
expressed in the tapetum. TAPETUM DETERMINANT1 (TPD1) encodes a small protein and is
expressed in PMCs [Yang et al. (2005) Plant Physiol. 139:186-191, Yang et al. (2003) Plant Cell
15:2792-2804]. It is possible that TPD1 interacts with the LRR domain of EMS1/EXS, possibly
disrupting a further interaction with a second LRR receptor kinase, SERK1/2 [Colcombet et al.
(2005) Plant Cell 17: 3350- 3361; Albrecht et al. (2005) Plant Cell 17: 3337-3349]. IRRI is
examining the hypothesis that rice contains a gene similar to TPD1. Two putative orthologs
(OsTPD1A and OsTPD1B) have been identified and RNA in situ hybridization indicates that
they are overlap with MSP1 in their sites of expression in the anthers, as shown already for
TPD1 and EMS1/EXS. In ovules, MSP1 and OsTPD1A are expressed. Anthers of homozygous
msp1 mutants have a layer of secondary meiocytes in stead of a tapetum. In such anthers, only
OsTPD1A is highly expressed, with transcripts of MSP1 and OsTPD1B not readily detected.
IRRI and CSIRO are collaborating to use the two]hybrid system of yeast to determine whether
OsTPD1 or OsTPD2 binds to the LRR domain of MSP1.

.. Output 2.4 OsMac1]based system for inducing secondary MMC (MMC2) (I)
In the homozygous msp1 mutant, up to 15 secondary MMCs are induced in each ovule. IRRI is
now attempting to interfere with the function of MSP1 more subtly to induce only 1-2
secondary MMCs which will then be induced to form aposporous initials through the bypassing
of meiosis. Two methods of interfering with MSP1 function are being developed: (1)
subtle down-regulation of MSP1 gene expression through RNAi under the control of the
MSP1 promoter, and (2) subtle down]regulation of the gene encoding ligand of MSP1
(putatively OsTPD1A or OsTPD1B) by RNAi. Two transgenic plants designed for RNAi of
OsTPD1A have been produced at IRRI by Ms Justina de Palma and Ms Gina Borja in the
laboratory of Dr. Philippe Herve. These plants have already been shown by reverse
transcription-PCR to down]regulate OsTPD1A expression by several fold. We are currently
examining these plants for the desired phenotype, namely, the presence of 1-2 secondary
MMCs in the ovule.

Objective 4 (7/f03]6/f08)
Communication and dissemination of research results
.. Output 4.1 Publications and website for pre]publication releases (10/f03)
Publication
Bi X, Khush GS, Bennett J. 2005. The rice nucellin gene ortholog OsAsp1 encodes an active
aspartic protease without a plant]specific insert and is strongly expressed in early embryo.
Plant Cell Physiol. 46:87-98.
Website
A website for the project is being constructed. It will contain information on Phase 1 and
Phase 2, including publications and reports.

[Note: Only outputs scheduled for initiation in Years 1-3 are mentioned. Activities: (I) = IRRI, (C) = CSIRO.]

Our goal is to make the benefits of hybrid rice available to poor farmers through asexual seed production. At present, farmers must buy expensive hybrid seed fresh every season, because its heterozygosity is diminished by 50% with every cycle of sexual reproduction. We propose to develop a synthetic form of apomixis that will fix the heterozygosity of the hybrid and allow farmers to reproduce hybrid seeds cheaply in their own fields. Although essentially all natural apomicts are polyploids, we expect our synthetic apomixis to be compatible with diploidy in rice.
We shall model our synthetic apomixis on the apospory of Poa pratensis, a relative of wheat and barley in which only 5 genetic loci are required for apospory [Matzk et al. (2005) Plant Cell 17:13-24]. The key milestone will be the generation of an aposporous embryo from diploid cells in the nucellus of the ovary. We shall convert 1-2 cells of the nucellus of rice into the equivalent of aposporous initials (AIs) and assess their capacity to develop firstly into diploid embryo sacs and then into aposporous embryos.
The conversion of nucellar cells into aposporous initials will take advantage of a recently discovered gene (MSP1) that limits the number of nucellar cells able to become megaspore mother cells (MMCs). This gene was inactivated by Tos17 insertional mutation. When the mutant was recovered in the homozygous form (msp1 msp1), up to 15 secondary MMCs form in the ovule and develop into haploid embryo sacs. We propose to inactivate MSP1 more subtly by RNA interference (RNAi) to allow only 1-2 secondary MMCs to form. At the same time we shall prevent these cells from entering meiosis, so that they form the equivalent of diploid AIs. We shall assess the conditions under which the putative AIs form diploid embryo sacs and aposporous embryos. Objective 2 focuses on generating an apomictic embryo by the above approach.
Apomictic embryogenesis may be pursued with or without fertilization. The fertilization-dependent approach is in some respects more straightforward because formation of the endosperm normally requires fertilization, even in many apomicts, and the seed coat and the pericarp, although maternal tissues, usually fail to develop fully without fertilization. Furthermore, the fertilization-dependent approach would produce the usual triploid endosperm rather than a diploid form. We have opted here for the fertilization-independent (FI) approach because (1) it eliminates competition between the sexual embryo and the apomictic embryo, and (2) it avoids controversy about the escape of apomixis genes into the environment. However, it will be necessary to have a transient sexual phase for the initial formation of the apomictic hybrids in breeders' fields; the FI phase can be used in farmers' fields. Objective 1 focuses on the FI protocol.

Objective 1 (7/'03-6/'06)
Fertilization-independent (FI) formation of rice endosperm and pericarp
Output 1.1 Refined FI pericarp formation based on understanding of OsAsp1 function (I)
Both IRRI and CSIRO have engineered methods for by-passing the fertilization-dependence of pericarp and seed coat enlargement. This by-pass is essential for achieving FI-independent seed production, because it removes a physical constraint on the size of the endosperm that can develop autonomously.
Output 1.2 Refined FI endosperm formation based on RNA-I of OsFIS genes (C)
FI endosperm formation in rice is essential for constructing a form of apomixis in which the escape of transgenic pollen is prevented. CSIRO has a method of generating fertilisation-independent formation of the endosperm in Arabidopsis, based on the discovery through mutation of three Fertilization-Independent Seed (FIS) genes. This method has been introduced into rice. Based on sequence similarity, all the FIS like genes in rice had been identified, and transgenic lines with silencing constructs for each OsFIS gene are being evaluated.

Objective 2 (7/'03-6/'06)
FI embryogenesis in rice nucellus
Output 2.1 Isolated rice orthologue of maize MAC1 gene (I)
We are focusing on developing a synthetic form of apospory for rice. Aposporous embryos develop from nucellar cells near the megaspore mother cell (MMC). The maize mutation multiple archaesporial cells1 (mac1) showed that such nucellar neighbours can develop into secondary MMCs if the MAC1 gene is inactivated. One of IRRI's objectives was to clone the rice orthologue, OsMAC1. However, this task was actually accomplished by Nonomura et al. (2003) in Japan [Plant Cell. 15:1728-1739], and they named the rice gene MULTIPLE SPOROCYTES1 (MSP1).
A key paper on the genetics of the control of apomixis and sexuality in Poa pratensis was published by Matzk et al. (2005) Plant Cell 17:13-24. In this paper, the authors identified a locus APV (Apospory Prevention) and speculated that it may be the Poa ortholog of MAC1. They identified two other loci governing the initiation of apospory (AIT, Apospory Initiation, and mdv, Megaspore Development) controlling apomeiosis and the initiation of the sexual pathway. IRRI is studying candidate genes as orthologs of these loci. Matzk et al. (2005) identified two loci controlling parthenogenesis in Poa, ppv (parthenogenesis prevention) and PPI (Parthenogenesis Initiation). In Arabidopsis, a newly identified Polycomb protein that interacts with FIS complex, AtMSI1 may code for ppv (see section 4.2 for details). CSIRO isolated the OsMSI1 and testing the alternative way to induce parthenogenesis in rice.
Output 2.2 Isolated rice orthologues of Arabidopsis CLV, WUS, STM and AG genes (C + I)
Embryogenesis in the rice nucellus will probably require re-activation of meristematic activity through expression of orthologues of such Arabidopsis genes as CLAVATA1-3, WUSCHEL, SHOOT MERISTEMLESS and AGAMOUS. IRRI and CSIRO identified orthologues by sequence comparisons. In the last year, Mr. Rico Gamuyao of IRRI examined 10 members of the rice WOX gene family to ascertain their expression patterns, using reverse transcription-polymerase chain reaction (RT-PCR) and RNA in situ hybridization. He found that two genes were highly expressed in the ovule.
Output 2.3 System to control genes of nucellus from megaspore mother cell (MMC1) (I)
Leucine-rich receptor kinases like MSP1 are usually triggered by the binding of extracellular ligands to the LRR domain. We are attempting to identify the ligand of MSP1 and its site of synthesis. We are guided in this search by the finding that a LRR receptor kinase is implicated in the control of tapetum development by pollen mother cells, PMC, in Arabidopsis. EXCESS MALE SPOROCYTES1 (EMS1, also known as EXS) encodes a LRR receptor kinase and is expressed in the tapetum. TAPETUM DETERMINANT1 (TPD1) encodes a small protein and is expressed in PMCs [Yang et al. (2005) Plant Physiol. 139:186-191, Yang et al. (2003) Plant Cell 15:2792-2804]. It is possible that TPD1 interacts with the LRR domain of EMS1/EXS, possibly disrupting a further interaction with a second LRR receptor kinase, SERK1/2 [Colcombet et al. (2005) Plant Cell 17: 3350 - 3361;
Albrecht et al. (2005) Plant Cell 17: 3337 - 3349]. IRRI is examining the hypothesis that rice contains a gene similar to TPD1.
Output 2.4 OsMac1-based system for inducing secondary MMC (MMC2) (I)
In the homozygous msp1 mutant, up to 15 secondary MMCs are induced in each ovule. IRRI is now attempting to interfere with the function of MSP1 more subtly to induce only 1-2 secondary MMCs which will then be induced to form aposporous initials through the by-passing of meiosis. Two methods of interfering with MSP1 function are being developed: (1) subtle down-regulation of MSP1 gene expression through RNAi under the control of the MSP1 promoter, and (2) subtle down-regulation of the gene encoding ligand of MSP1.

Our goal is to make the benefits of hybrid rice available to poor farmers by developing a cheaper and more flexible form of hybrid seed production that allows farmers to multiply seeds in their own fields. We aim to fix the heterozygosity of hybrids through a form of apomixis - seed production in which the genetic constitution of progeny plants is identical to that of the hybrid. To maintain a constant genetic constitution, the embryo of the progeny seed must be formed without undergoing meiosis, a form of sexual recombination. The form of apomixis that we plan to introduce into rice is known as apospory and exists in many wild grasses but not in rice or its close relatives.
In wild grasses, apospory develops in cells alongside the megaspore mother cell (MMC) of the rice ovule. These cells (aposporous initials, AI) are similar to the MMC but differ from it in failing to undergo meiosis. Multiple secondary MMCs are observed in the ovules of the mac1 mutant of maize and the msp1 mutant of rice. We hypothesized that these secondary MMCs may be a first step towards apospory and that the next step would be to by-pass meiosis in the secondary MMCs.
Unfortunately, the msp1 mutant is not a convenient platform from which to develop AIs in rice because it is male-sterile. Staff of IRRI searched for a rice gene other than MSP1 that could be inactivated to produce secondary MMCs in the ovule without causing male sterility. Based on studies in Arabidopsis thaliana, we suspected that MSP1 protein interacts with a small protein that we term TPD1-like1, or TDL1. We found two close homologues of TPD1 in the rice genome (TDL1A and TDL1B). Furthermore, these two genes are co-expressed during meiosis with MSP1. We used yeast 2-hybrid analysis and bimolecular fluorescence complementation to show that TDL1A is a ligand of MSP1. Furthermore, when we used RNA interference (RNAi) to down-regulate the OsTDL1A gene, we found that the OsTDL1A-RNAi transgenic plants produce secondary MMCs in the ovule without causing male sterility. This work has been submitted to Plant Journal for publication. We are now developing a method for by-passing meiosis in the secondary MMCs to produce AIs in rice.
Naturally occurring AIs form embryo sacs that undergo parthenogenesis to produce embryos without fertilization. Again, studies on Arabidopsis thaliana suggested that inactivation of the polycomb complex is essential not only for parthenogenesis but also for autonomous endosperm formation. Staff of CSIRO Plant Industry have, accordingly, isolated the genes encoding the rice polycomb complex. We used either RNAi or insertion mutagenesis to interfere with the expression of these genes and found that two distinct checkpoints in the sexual development of the ovary have been overcome: (i) the checkpoint that prevents ovary enlargement in the absence of fertilization and (ii) the checkpoint that prevents endosperm formation without fertilization of the cell central of the embryo sac. The rice homologue of the Arabidopsis gene CLF appears to be especially important in this context. The next steps are (i) to combine these two features to obtain large autonomous endosperms and then (ii) to manipulate the polycomb complex further to obtain parthenogenetic embryo formation. For step (ii), we are exploring the rice homologues of the Arabidopsis gene MSL1, which mediates signalling between endosperm and embryo.
In summary, our data support the emerging consensus that the sexual and aposporous pathways of plant seed production share many genes but differ at two major stages: (i) the onset of meiosis and (ii) the checkpoints preventing seed development in the absence of fertilization. This consensus raises the intriguing evolutionary question as to whether sexuality is a modification of apomixis or vice-versa.

In July 1997, CSIRO and IRRI began a three-phase project to develop apomixis for hybrid rice (Oryza sativa L.), funded by ACIAR. The goal was to make the yield advantage and stress tolerance of hybrid rice available to poor farmers in Asia and Africa by using synthetic apomixis to reduce the cost and increase the flexibility of hybrid seed production.

Phase 1 (July 1997-June 2002) involved three laboratories: CSIRO Horticulture in Adelaide, CSIRO Plant Industry in Black Mountain, and IRRI in the Philippines. It provided new insights into the similarities and differences between sexual reproduction and asexual apomixis. Highlights were (i) the identification of three FERTILIZATION-INDEPENDENT SEED (FIS) genes in the sexual plant Arabidopsis that form part of a complex to repress endosperm formation in the absence of fertilization and (ii) confirmation that FIS homologues exist in rice and in the natural apomict Hieracium piloselloides that reproduces by apospory (avoidance of meiosis).

Phase 2 (July 2003-June 2008) focused exclusively on rice and involved the Black Mountain Laboratory and IRRI. The three main objectives for Phase 2 were as follows:

Objective 1: To identify all of the known FIS orthologues in the rice genome and to functionally test each for efficacy in fertilization-independent (FI) formation of the pericarp and seed coat of rice (Objective 1a) and FI formation of the embryo and the endosperm in the rice ovule (Objective 1b).
Objective 2: To develop FI embryos in the nucellus by inducing aposporous initials (AIs).
Objective 3: To combine the above traits to generate a basic form of apospory (meiotic avoidance) with embryo and endosperm induction in rice that could be refined further in Phase 3.

Both IRRI and CSIRO contributed to Objective 1a. They found that the checkpoint that normally prevents FI formation of the pericarp and seed coat could be bypassed using transgenic approaches. However, progress with these approaches was hampered by a background level of ovary enlargement in non-transgenic control plants. A change in rice variety from Nipponbare to one in which this background is negligible would be valuable. Objective 1b was pursued by CSIRO. Based on sequence similarity with the Arabidopsis FIS gene, seven candidate rice FIS-like (OsFIS) genes were identified in the rice genome.

Transgenic rice lines containing silencing constructs to down-regulate all of these OsFIS genes were generated and evaluated, and T-DNA insertion lines in three of the FIS-like rice genes were also evaluated. Unexpectedly, autonomous endosperm formation and autonomous embryo formation were not observed in any of the hundreds of transgenic rice lines analyzed. Data on expression of these genes, combined with an evolutionary analysis of the rice and Arabidopsis genes, indicated that rice FIS-like genes do not function in an equivalent manner to that found in Arabidopsis and its relatives. Repression of rice seed formation is likely to occur by an as yet unknown mechanism that needs further investigation if synthesis of apomixis is to be attempted (Luo et al, submitted).

Under Objective 2, the first milestone was to induce the formation of multiple secondary megaspore mother cells (MeMCs) in the nucellus. This milestone was reached (Zhao et al 2008) by silencing OsTDL1A, the rice orthologue of the Arabidopsis gene TAPETUM DETERMINANT1. IRRI showed that OsTDL1A-silenced lines produce secondary MeMCs in both indica and japonica genetic backgrounds. The second milestone is to convert secondary MeMCs into AIs by eliminating meiosis. IRRI has identified the rice orthologue of a gene required for initiation of meiosis in yeast. Silencing of this rice gene may prevent entry in meiosis and allow AIformation.

Because Objectives 1a, 1b, and 2 were not attained completely, Objective 3 has not yet begun and will be postponed until Phase 3, which should focus on integrating discoveries in rice with those in Arabidopsis and natural apomicts such as Hieracium.