Ubiquitous and Endoplasmic Reticulum–Located Lysophosphatidyl Acyltransferase, LPAT2, Is Essential for Female but Not Male Gametophyte Development in Arabidopsis

Arabidopsis Has Four Putative Cytoplasmic AtLPATs That Can Be Divided into Two Groups

Five Arabidopsis genes (AtLPAT1-5) could encode LPATs on the basis of their predicted amino acid sequences being similar to those of demonstrated LPATs in other plant and nonplant systems (Kim and Huang, 2004; Yu et al., 2004). Several additional Arabidopsis genes that encode proteins whose amino acid sequences are similar to those encoded by the above putative AtLPAT genes have recently been shown to encode GPAT (Zheng et al., 2003).

An unrooted phylogenetic tree of the amino acid sequences of proteins derived from AtLPAT1-5 and closely related genes of other plant species, yeast, and Escherichia coli reveals several groups of putative LPATs (Figure 1). One group includes the known plastid (AtLPAT1) and E. coli LPAT. Another group comprises the studied microsomal LPATs of several other plant species and AtLPAT2 and 3. A third group consists of AtLPAT4 and 5 and an unstudied rice (Oryza sativa) LPAT derived from genomic sequence. A fourth group encompasses the studied coconut and meadowfoam microsomal LPATs. This last group might represent seed-specific LPATs that can acylate a unique acyl group to the sn-2 position of a TAG molecule (Kim and Huang, 2004), such as C-12 in coconut (Oo and Huang, 1989; Knutzon et al., 1995) and C-22:1 in meadowfoam (Cao et al., 1990; Hanke et al., 1995). Nevertheless, this last group also includes a rice LPAT derived from a genomic sequence, and rice seed TAG is not known to possess a unique acyl moiety at the sn-2 position. Finally, the yeast LPAT does not have a closely related LPAT in plants.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

An Unrooted Phylogenetic Tree of Five Putative LPATs of Arabidopsis and Related LPATs of Other Plant Species Constructed on the Basis of Their Predicted Amino Acid Sequences.

The LPATs of E. coli and yeast are included for comparison. Genes and proteins were obtained after a BLAST search of the databases of The Arabidopsis Information Resource and National Center for Biotechnology Information with use of the amino acid sequences of maize LPAT (Z29518) and B. napus LPAT1 (AF111161) as queries. They were of Arabidopsis (LPAT1, At4g30560; LPAT2, At3g57650; LPAT3, At1g51260; LPAT4, At1g75020; LPAT5, At3g18850), Brassica (LPAT2, Z95637), rice (LPAT1, AC068923; LPAT2, NM191504; LPAT3, AL606645), meadowfoam (LPAT1, S60478; LPAT2, S60477), coconut (U29657), almond (AF213937), E. coli (plsC, M63491), and yeast (slc1, L13282).

AtLPAT2-5 (hereafter, the prefix At is omitted for simplicity) can be divided into two groups on the basis of their amino acid sequences. The identity of residues in amino acid sequences is 57% between LPAT2 and 3; 54% between LPAT4 and 5; and 23% or less between the two pairs. LPAT2 and 3 share a similar location of the putative transmembrane segments along the sequences, and this location differs from that shared by LPAT4 and 5 (Kim and Huang, 2004). LPAT2 and 3 have a similar pattern of introns, which differs strikingly from that shared by LPAT4 and 5 (Figure 2). Yet, the four LPATs all contain the four conserved domains (I to IV) of acyltransferases in diverse organisms, and their genes are all expressed in the plant.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Structures of Four Putative Arabidopsis LPAT Genes Encoding Cytoplasmic LPATs and Their Predicted Amino Acid Sequences.

Top panel: the exons (boxes) and introns (lines) of the four genes. Bottom panel: an alignment of their predicted amino acid sequences with the AlignX program from Vector NTI Suite. Residues identical to those in LPAT2 in some of the LPAT3-5 are lightly shaded, and those identical in all four LPATs are heavily shaded. The I [NH(X₄)D], II, III (EGT), and IV motifs conserved in related acyltransferases in various species are boxed. The C-terminal sequence of LPAT2, whose corresponding synthetic polypeptide was used to prepare antibodies, is boxed. The insertion site of T-DNA in LPAT2 and the corresponding location on LPAT2 are indicated by inverted arrows.

A pileup of the amino acid sequences of LPAT2-5 shows that LPAT2 has a C-terminal sequence absent in the other three LPATs (Figure 2). A synthetic peptide of this unique sequence, absent also in all other Arabidopsis proteins deduced from the complete genome sequence, was used to prepare antibodies against LPAT2.

LPAT2 Functionally Complemented an E. coli LPAT Mutant, and Its Encoded Protein Synthesized in the Bacteria Had in Vitro LPAT Enzyme Activity

We tested whether LPAT2-5 encode proteins that possess LPAT enzymatic activity. E. coli JC201, a temperature-sensitive mutant of LPAT (Coleman, 1990), was used. The mutant grows at 30°C but not at 42°C. The full-length open reading frame (ORF) of each of the four LPATs was inserted into the expression vector pQE, which contains the T5 promoter transcription-translation system. The resulting pQE-LPAT(2-5) and the control pQE were transformed into JC201, which had been transformed with pREP4 to prevent potential leakiness of the T5 promoter.

Cells containing pQE-LPAT(2-5) grew as well as the wild type containing pQE in the absence of isopropyl-β-thiogalactoside (IPTG) but poorly when IPTG at the low concentration of 0.2 mM was added to the growth medium. Apparently, LPAT2-5 strongly inhibited bacterial growth, a phenomenon that is quite common when eukaryotic membrane proteins are synthesized in transformed bacteria (Laage and Langosch, 2001). We used cells grown in 0.2 mM IPTG overnight to test functional complementation on an agar plate and for 1 h (when the cells started to express the Arabidopsis LPAT and accumulated its LPAT) in a liquid medium to prepare membrane fractions for biochemical analyses.

Figure 3A shows that JC201 harboring either pQE or pQE-LPAT2 grew at 30°C, but only that harboring pQE-LPAT2 grew at 42°C. At 42°C, the colonies were irregularly shaped and fewer than those at 30°C. Thus, LPAT2 was active in vivo and complemented the defective E. coli LPAT. Attempts to observe similar functional complementation in JC201 with LPAT3, 4, and 5 were unsuccessful, presumably because LPAT3, 4, and 5 were more potent in inhibiting bacterial growth.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Expression of LPAT2 and LPAT3 in the E. coli Strain JC201, a Temperature-Sensitive Mutant Defective in LPAT.

(A) Bacteria transformed with pQE, pQE-LPAT2, or pQE-LPAT3 were grown at 30^° C or 42^° C for 18 h and photographed. pQE-LPAT2 (shown) but not pQE-LPAT3 (data not shown) complemented the mutation.

(B) Membrane fractions of bacteria transformed with pQE or pQE-LPAT2 grown at 30^°C were analyzed for their protein constituents by SDS-PAGE (left) followed by immunoblotting with antibodies against LPAT2 (right). The antibodies recognized a single recombinant LPAT2 protein of 45 kD. Molecular mass markers are on the left lane.

(C) Membrane fractions from the three types of transformed bacteria were assayed for LPAT activities with the use of equal amounts of protein (70 μg), LPA-18:1, and either 18:1- or 16:0-CoA.

Proteins in E. coli JC201 harboring pQE and pQE-LPAT2 were analyzed by SDS-PAGE (Figure 3B). Proteins in the membrane fractions from the two bacterial samples separated on the gel had an identical pattern. After immunoblotting, a protein of 45 kD (45,352 D being the expected molecular mass of the 6-His–tagged polypeptide derived from the ORF) in the sample of bacteria harboring pQE-LPAT2 but not that harboring pQE was recognized by anti-LPAT2. The membrane fractions from these two bacterial samples were assayed for LPAT activity with use of LPA-18:1 as the acyl acceptor and either 16:0- or 18:1-CoA as the acyl donor (Figure 3C). The membrane fraction from bacteria harboring pQE had minimal enzymatic activity, which was higher with 16:0-CoA than with 18:1-CoA, a characteristic of the prokaryotic enzyme (Ohlrogge and Browse, 1995); the activity should represent that of the native bacterial enzyme. The membrane fraction from bacteria harboring pQE-LPAT2 had several hundred times higher LPAT activity, and the activity was higher with 18:1-CoA than 16:0-CoA, a characteristic of the eukaryotic (cytoplasmic) enzyme (Ohlrogge and Browse, 1995); the activity should represent mostly LPAT2 activity. Attempts to observe similar in vitro enzymatic activities in LPAT3, 4, and 5 in the transformed bacteria were largely unsuccessful. The membrane fraction from bacteria harboring pQE-LPAT3 had LPAT activity, which was two to three times higher than that in the membrane fraction from bacteria harboring pQE and preferred 18:1-CoA over 16:0-CoA (Figure 3C). This acyl-CoA preference should reflect largely that of LPAT3 and not the native bacterial enzyme. The membrane fraction from bacteria harboring pQE-LPAT4 and 5 had only background (bacterial) activity that was higher with 16:0-CoA than with 18:1-CoA (data not shown).

LPAT2 and 3, but Not LPAT4 and 5, Synthesized in Yeast Had in Vitro LPAT Enzyme Activities

LPAT2-5 synthesized in transformed E. coli apparently inhibited bacterial cell growth. Inhibition of fungal cell growth was not observed when LPAT2-5 were transformed into and expressed in yeast. LPAT was placed in the yeast-transforming plasmid pYES under the control of the GAL1 promoter, which can be activated with galactose and to a lesser extent glucose (Giniger et al., 1985). Grown in glucose, yeast harboring pYES-LPAT contained the LPAT transcript at a low level (Figure 4A). Four hours after the addition of galactose to the growth medium, the yeast possessed several times more of the LPAT transcript. The total lysate of the yeast harboring pYES-LPAT resolved on a SDS-PAGE gel possessed a protein band of ∼43 kD (Figure 4B), which is the deduced molecular mass of LPAT2-5. The ∼43-kD protein band was also present in the lysate of yeast transformed with pYES. The yeast genome has genes that could encode 132 unmodified proteins of 42 to 44 kD (http://db.yeastgenome.org), and the lone yeast LPAT has 34 kD. Therefore, the ∼43-kD protein band in the lysate of the yeast harboring pYES-LPAT should represent non-LPAT proteins of the yeast and LPAT2-5. LPAT2 in the 43-kD protein band from the lysate of yeast transformed with LPAT2 was positively identified with antibodies against an LPAT2-specific polypeptide. The amount of LPAT2 was higher in the transformed yeast grown in galactose than in glucose (Figure 4C), consistent with the level of the LPAT2 transcript (Figure 4A). Yeast lysate containing LPAT2 or 3 showed in vitro LPAT activities substantially higher than that in the control lysate of yeast transformed with pYES, whereas those containing LPAT4 and 5 had activities similar to those in the control yeast lysate (Figure 4D). Both LPAT2 and LPAT3, as well as the yeast native LPAT, had activities higher with 18:1-CoA than with 16:0-CoA, a characteristic of the eukaryotic enzyme. These findings on LPAT2-5 synthesized in yeast are consistent with those in E. coli.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Expression of LPAT2-5 in Saccharomyces cerevisiae.

(A) Yeast transformed with pYES or pYES-LPAT2-5 were analyzed for the presence of LPAT2-5 in DNA (cells grown in glucose) and LPAT2-5 transcripts in RNA (cells grown in glucose [Glu] and then transferred to a galactose medium and grown for 4 h [Gal]) by PCR. Samples containing approximately equal amounts of total RNA (shown) were used in the RT-PCR.

(B) Lysates of the transformed yeast were analyzed for protein constituents by SDS-PAGE. Asterisks indicate the protein band containing LPAT2-5; the band also contained other yeast proteins because it was also present in the lysate of yeast transformed with pYES.

(C) SDS-PAGE immunoblotting of membrane fractions of transformed yeast grown in glucose and then galactose for 4 h (top); the antibodies were raised against a synthetic peptide unique to LPAT2. Sensitivity of the immunoblotting was shown with decreasing amounts of the sample from galactose-grown yeast transformed with LPAT2 (bottom). Positions of molecular mass markers are shown at the right in (B) and (C).

(D) Lysates of the transformed yeast grown in glucose and then galactose for 4 h were analyzed for LPAT activities with the use of equal amounts of protein (50 μg) and LPA-18:1 and either 18:1- or 16:0-CoA.

The LPAT2 Transcript Was the Most Abundant and Ubiquitous, the LPAT3 Transcript Was Abundant in Pollen, and the LPAT4 and 5 Transcripts Were Ubiquitous but Scarce

Three different methods were used to study the prevalence of the LPAT2-5 transcripts. Each method has its advantages and uncertainties, but together they would produce complementary results revealing a general theme.

RT-PCR results show that the LPAT2, 4, and 5 transcripts were present in all of the five organs examined: siliques, inflorescences, rosette leaves, stems, and roots (Figure 5). The LPAT3 transcript was detected in only inflorescences and roots. In addition, the LPAT2, 4, and 5 transcripts but not the LPAT3 transcript were detected in developing embryos, whereas all were present in seedlings. The LPAT3 transcript was apparently more abundant than the LPAT2 transcript in pollen (Figure 5). Although this RT-PCR result with pollen is at best semiquantitative, it is consistent with the findings of the β-glucuronidase gene (GUS) expression (to be described in Figure 7). It is also consistent with the reported transcriptomes of the Arabidopsis pollen, where LPAT3 and not LPAT2, 4, and 5 was present in abundance (Becker et al., 2003; Honys and Twell, 2003).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

RT-PCR Analyses of the Transcripts of LPAT2-5 in Various Organs.

The organs included siliques (Si), unopened flowers (F), rosette leaves (RL), stems (St), roots (R), and early (E) and late (L) maturing embryos and seedlings. The bottom panel shows a comparison of the levels of LPAT2 and 3 transcripts in mature pollen. Approximately equal amounts of transcripts of an Arabidopsis actin gene (ACTIN) were present in the various samples. The lengths of the RT-PCR fragments are those expected from the primers used. The RT-PCR products were obtained with primers and cycle numbers at a linear range, as revealed in the results of DNA-PCR with use of increasing amounts of full-length LPAT cDNA and the same primers and cycle numbers.

The recently available massively parallel signature sequencing (MPSS) database of Arabidopsis (Meyers et al., 2004) was used to assess the prevalence of the LPAT2-5 transcripts. This 17-mer MPSS database of a reasonable size (2 to 4 million signature sequences from each cDNA library) was constructed with Sau3A restriction fragmentation, and the ORFs of LPAT2-5 have 3, 5, 4, and 3 Sau3A sites, respectively. Thus, a comparison of the prevalence of the four transcripts in the MPSS database should provide an approximation. Figure 6 shows that in all of the five examined organs, the LPAT2 transcript was substantially more abundant than all the other three transcripts combined.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Transcripts of the Four Putative Genes in Different Organs with Use of the MPSS Database.

Information was obtained from http://mpss.udel.edu/at (Meyers et al., 2004). Signature sequences of 17 nucleotides of each gene in five different libraries are shown. The signatures of 17-mer tags generated via cutting the cDNA in the libraries at the Sau3A restriction sites were normalized to transcripts per million to facilitate comparisons among libraries.

The promoter regions of the four LPATs were each attached to the reporter gene GUS, and the constructs were transformed into Arabidopsis plants. GUS activity in the various organs of the transformed plants was examined (Figure 7). It was highest in the veins of leaves, inflorescence stalks, and roots when GUS was driven by the LPAT2, 4, and 5 promoters. GUS activity was detected mostly in pollen when GUS was driven by the LPAT3 promoter. In pollen, the LPAT3 promoter appeared to be more active than the LPAT2 promoter, whereas LPAT4 and 5 promoters were totally inactive in driving the expression of GUS (Figure 7). This observation of the relative strength of the LPAT2 and 3 promoters in driving GUS in pollen is consistent with the RT-PCR results (Figure 5). In the pistil, GUS activity in the female gametophyte was detected only when GUS was driven by the LPAT2 promoter; it was clearly detected at the four-nucleate stage and increased to a maximum at the eight-nucleate stage and then declined after fertilization (Figure 7). This decline could be due to dilution of the GUS transcript or protein as a result of cell multiplication.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

GUS Activities in Different Parts of Arabidopsis Plants Transformed with a Construct Containing the Promoter of LPAT2-5 Ligated to the ORF of a GUS Gene.

The bottom panel details GUS activities from GUS driven by the LPAT2 promoter in the ovule during development before and after fertilization. Bars represent 50 μm in the root, 20 μm in the anther and pistil, and 10 μm in the ovule samples. ch, chalazal pole; co, cotyledon; hy, hypocotyl; mp, micropylar pole; r, root; rt, root tip; sm, shoot meristem.

Overall, results obtained from the above three different methods reveal a general theme of the expression of the four LPATs. The LPAT2, 4, and 5 transcripts are ubiquitous, but the LPAT2 transcript dominated. LPAT3 is expressed mostly in pollen, where its transcript is more abundant than the LPAT2 transcript and where the LPAT4 and 5 transcripts are absent. The LPAT2 transcript is the only one present in the female gametophyte.

LPAT2 Was Ubiquitous and Located in the ER

Antibodies against a synthetic peptide corresponding to the C terminus of LPAT2 were raised. This C-terminal sequence in LPAT2, or a similar version, is absent in LPAT3-5 (Figure 2) and all other Arabidopsis proteins deduced from the genome sequence. The antibodies reacted with LPAT2 but not LPAT3-5 synthesized in transformed yeast (Figure 4C). They detected LPAT2 as a single protein band by SDS-PAGE immunoblotting in the total extracts of all examined Arabidopsis organs, including siliques, floral buds, stems, leaves, and roots (Figure 8B). Thus, LPAT2 was ubiquitous, as was the LPAT2 transcript (Figure 5). Anti-LPAT2 antibodies also recognized BnLPAT2 in Brassica napus anthers (Figure 8B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Detection of LPAT2 in Extracts of Different Organs and Its Localization in the ER of the Tapetum Cells of Arabidopsis and Brassica.

(A) An SDS-PAGE gel of the total leaf extracts from the wild type and the heterozygous mutant (LPAT2/lpat2) and an identical gel after immunoblotting with anti-LPAT2.

(B) LPAT2 in the total extracts of different organs analyzed by SDS-PAGE and then immunoblotting. In (A) and (B), the asterisk denotes the recognized LPAT2.

(C) Immunofluorescence microscopic images of double-labeled Arabidopsis and Brassica tapetum cells. The image of each cell was observed by bright-field (BF) microscopy and immunofluorescent microscopy labeled with antibodies against LPAT2, calreticulin (ER marker), and plastid lipid associated protein (PAP; plastid marker). All images are of the same magnification. Bar = 5 μm.

(D) An SDS-PAGE gel of the total sporophytic anther extract from Brassica and its subfractions by centrifugations and identical gels that had been subjected to immunoblotting with antibodies against the indicated proteins. Two protein bands of 50 to 60 kD were recognized by antibodies against castor bean (Ricinus communis) calreticulin. The amounts of the various subfractions loaded to the gel corresponded to those from the same amount of the total extract. In all gels, positions of molecular mass markers are shown at the left.

We tested whether LPAT2 was present in the ER or other subcellular compartments. We chose the tapetum cells of Arabidopsis and Brassica, in which the ER is abundant and the interfering autofluorescent chloroplasts are absent (Platt et al., 1998), to study localization. Double labeling with antibodies against LPAT2 and the ER marker calreticulin shows that in Arabidopsis tapetum cells, LPAT2 was largely colocated with calreticulin (Figure 8C). Small regions in the ER were not occupied by either LPAT2 or calreticulin, and the existence of these nonoverlapping regions reaffirms the presence of specialized subdomains in the ER (Papp et al., 2003). Similar colocation of BrLPAT2 and calreticulin was observed in Brassica tapetum cells (Figure 8C). Results of control experiments with antibodies against other organelle markers reaffirm that BrLPAT2 was colocated with calreticulin in the ER in Brassica tapetum cells. For example, the plastid protein, plastid lipid–associated protein (Kim et al., 2001), was localized in the plastids independent of the calreticulin in the ER (Figure 8C).

Complementary data supporting the presence of LPAT2 in the ER were obtained by subcellular fractionation of Brassica sporophytic anther extract followed by SDS-PAGE and immunoblotting (Figure 8D). LPAT2 and calreticulin were colocalized in the 100-kg (the microsomes) and the 10-kg pellets, whereas the plastid lipid-associated proteins were present mostly in the 10- and 1-kg pellets.

A Heterozygous (LPAT2/lpat2) Mutant Possessed Half of the LPAT2 Transcript and LPAT and No Greatly Altered Leaf Acyl Composition or Vegetative Morphology

Seeds containing T-DNA–inserted LPAT2 (lpat2) were studied (Salk_058130, T-DNA inserted into At3g57650). Of the plants grown from the 13 seeds we obtained, eight were wild type (LPAT2/LPAT2) and five heterozygous mutants (LPAT2/lpat2). All these heterozygous mutants shared the same vegetative and reproductive morphology, and we will describe one in detail. The heterozygous mutant had only one copy of T-DNA inserted into its genome, as evidenced by DNA gel blot results (Figure 9). Leaf DNA was cut with PstI, whose sites are present in the T-DNA and at the opposite ends of LPAT2. Fragments on the blot were hybridized with a 1.6-kb DNA probe, which represents the 5′-terminus of LPAT2 (1.2 kb) and the 5′-terminus of the T-DNA (0.4 kb) (to be shown as fragment a in Figure 10A). The wild-type DNA had only one PstI fragment of the expected 2.8 kb. The heterozygous mutant DNA had an additional ∼5.4-kb PstI fragment, which represents the 1.2-kb 5′-terminus of LPAT2 and ∼4-kb 5′ terminus of the T-DNA.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Comparisons of the Heterozygous Mutant (LPAT2/lpat2) with the Wild Type (LPAT2/LPAT2).

DNA hybridization shows that both plants contained the PstI fragment of LPAT2 (2.8 kb), whereas the mutant also had the PstI fragment of lpat2 (5.4 kb). RNA hybridization reveals that the mutant had approximately half the LPAT2 transcript (1.4 kb) of the wild type and a transcript of 1.3 kb that could represent the truncated lpat2 transcript (marked with an asterisk). Approximately equal amounts of the 25S and 18S rRNA (stained with ethidium bromide) were present in the mutant and wild-type samples. The mutant and wild type had similar vegetative morphology. The siliques produced by the mutant were shorter than those of the wild type. During maturation at 4, 10, and 12 DAF, the siliques contained half maturing seeds and half apparently aborted ovules (arrows). Sections of the pistils observed under a light microscope (stained with tuluidine blue, panel second to the bottom) show that the lpat2 female gametophyte at the four-cell stage in the mutant ovule differed from that of the wild type in having numerous, large, purple starch granules in the central cell. The sporophytic cells enclosing the gametophyte in both the mutant and wild type appeared to be similar. Sections of the central cells observed under an electron microscope (bottom panel) show that the cytoplasm of the lpat2 female gametophyte contained numerous large starch granules (st) in amyloplasts instead of abundant ER as the wild type did. ch, chalazal pole; cv, central cell vacuole; cw, cell wall; en, egg nucleus; ev, egg vacuole; mp, micropylar pole; sen, secondary endosperm nucleus; sn, synergid cell nucleus.

• Download figure
• Open in new tab
• Download powerpoint

Figure 10.

Analyses of the Offspring of LPAT2/lpat2 Transformed with LPAT2 or LPAT3.

(A) The structures of LPAT2, lpat2, LPAT2P:LPAT2-c, and LPAT2P:LPAT3-c and results of PCR genotyping. The primers used and the expected PCR fragments from the DNA are shown. LPAT2 contains introns (lines) and exons (boxes). The gene lpat2 contains a T-DNA (inverted triangle with the left border, LBa1) inserted into LPAT2. Results of PCR with use of the indicated primers are shown; X indicates the location of the missing fragment on the gel.

(B) RT-PCR analysis of LPAT2 and LPAT3 transcripts in 10-d-old shoots (a) and roots (b) of the wild type and LPAT2/lpat2 mutants (T3-1, 2, and 3) transformed with LPAT2P:LPAT3-c. Approximately equal amounts of transcripts of an Arabidopsis gene (ACTIN) were present in the various samples. The RT-PCR products were obtained with primers and cycle numbers at a linear range, as revealed in the results of DNA-PCR with use of increasing amounts of full-length LPAT cDNA and the same primers and cycle numbers. M represents RNA length markers.

(C) Images of the siliques from the various genotypes at the indicated DAF. The siliques were treated with ethanol (thus turned brown) for easy observation of the seeds. Green siliques produced by transformed plants (LPAT2/lpat2 + LPAT2P:LPAT3-c) at 8, 10, and 12 DAF are also shown; underdeveloped seeds are indicated with an asterisk, whereas the aborted ovules cannot be seen. Magnified, autofluorescent images of 6- and 10-DAF siliques taken under a fluorescent microscope reveal three types of ovules or seeds: shrunken, unfertilized ovules (arrows), fertilized embryo whose development stopped at the globular stage (asterisks), and fertilized embryo on a normal developmental track (arrowheads).

RNA gel blot analysis of the leaf RNA from the mutant and the wild type was performed with the probe used for the above-mentioned DNA gel blot analysis (Figure 9). The heterozygous mutant contained the wild-type LPAT2 transcript of 1.4 kb of approximately half the amount in the wild type and another fragment of 1.3 kb, which could represent a truncated LPAT2 transcript. The leaves of the heterozygous mutant also had approximately half of the LPAT2 present in the wild type (Figure 8A).

The heterozygous mutant was similar to the wild type in vegetative morphology (Figure 9), except that its rosette leaves were slightly shorter. Its leaf acyl composition was also similar to that of the wild type, possessing 18:3, 18:2, 16:0, 16:3, and 16:1 as the abundant acyl moieties (data not shown).

The Heterozygous (LPAT2/lpat2) Mutant Produced Shorter Siliques That Contained Half Normal Seeds and Half Remnants of Aborted Ovules

Although the heterozygous mutant was similar to the wild type in vegetative growth, it was not so in its reproductive growth (Figure 9). Its mature siliques were approximately two-thirds the length of those in the wild type. They contained approximately half normal-sized seeds and half empty slots filled with apparently aborted ovules that had not been fertilized. No seed or ovule of an intermediate size indicative of abortion after fertilization was observed. The aborted ovules in the siliques could be recognized soon after fertilization of the flowers 4 d after flowering (DAF) because they did not increase in size, whereas the fertilized ovules continued to do so. Sections of the ovules before fertilization reveal that the lpat2 female gametophyte began to differ from that of the wild type at the four-cell stage (Drews and Yadegari, 2002). At this stage, the central cell of the wild type contained abundant ER, whereas that of the lpat2 mutant had numerous large starch grains (Figure 9). The presence of abundant ER in the central cell of the wild type indicates the need of abundant membranes and ER activities; thus, a loss of LPAT2 in the lpat2 female gametophyte is deadly. In the lpat2 central cell, the reduced carbon designated for the ER membrane activities might have been channeled to form the starch granules. The wild-type egg cell did not contain abundant ER but was packed with different organelles and a large vacuole. The lpat2 egg cell also had similar subcellular structures, although its membrane structures in general were less defined (data not shown). Thus, the abortion of the lpat2 female gametophyte before or immediately after fertilization could have been caused by the absence of LPAT2 in the egg cell (which would become the embryo after fertilization), the central cell (which would become the endosperm), or both cells. It is known that the endosperm is required for successful embryogenesis, and aborted endosperm leads to death of the embryo regardless of whether the embryo is initially viable (Costa et al., 2004). The sporophytic cells enclosing the gametophyte in both the mutant and wild type at the four-cell stage appeared to be similar.

Results of Segregating lpat2 after Selfing the Heterozygous Mutant or Crossing It with the Wild Type Reveal Deleterious Effect of lpat2 in the Female Gametophyte

Results of reciprocal crossing between the heterozygous LPAT2/lpat2 mutant and the wild type support the notion that lpat2 causes ovule rather than embryo or pollen lethality (Table 1). Selfing of the heterozygous mutant produced siliques containing half of the number of seed per silique (compared with that in the wild type) and empty slots (approximately the same number as the slots filled with seed) possessing aborted ovules (Figure 9, Table 2). This 1:1 ratio is indicative of inheritance of the female or male gametophyte (Drews and Yadegari, 2002). Results of genetic crossing reveal that the deleterious effect of lpat2 was exerted only at the female gametophyte. Crossing the heterozygous mutant as the female with wild-type pollen also produced siliques that contained half of the number of seeds per silique as that in wild-type siliques. Plants grown from 20 of these seeds were found to be all of the LPAT2/LPAT2 genotype. Crossing the wild type as the female with pollen from the heterozygous mutant produced siliques that contained the same number of seeds as that in wild-type siliques. Plants grown from 20 of these seeds were examined; 10 had the LPAT2/LPAT2 genotype and 10 the LPAT2/lpat2 genotype.

In summary, we conclude that lpat2 causes lethality in the female gametophyte on the basis of the following findings. First, LPAT2 is the only LPAT in the female gametophyte because the promoter of LPAT2 but not LPAT3-5 drove the expression of GUS in the female gametophytes of transformed plants (Figure 7). Second, the central cell of the lpat2 female gametophyte at the four-cell stage began to differ from that of the wild type in possessing numerous starch grains instead of abundant ER, consistent with the loss of LPAT that is essential for phospholipid synthesis (Figure 9); less obvious but noticeable changes in the membranes in the lpat2 egg cell also occurred. Third, results of reciprocal crossing of the heterozygous mutant with the wild type provide genetic evidence for female and not male gametophytic lethality (Table 1). We speculate that lpat2 did not cause pollen lethality because the more abundant LPAT3 there compensated for the loss of LPAT2.

LPAT2-cDNA Functionally Complemented lpat2 in Heterozygous (LPAT2/lpat2) and Homozygous (lpat2/lpat2) Mutants

The above-mentioned phenotypes of the lpat2 mutant were caused solely by a T-DNA insertion into LPAT2, as evidenced by the successful rescue of the mutation after the heterozygous mutant was transformed with the LPAT2-promoter:LPAT2-cDNA (abbreviated as LPAT2P:LPAT2-c). Five of the T2 plants were heterozygous (LPAT2/lpat2) and contained LPAT2, lpat2, and LPAT2-c (Figure 10A). They produced 81% of the seeds per silique as that of the wild type; this proportion is substantially higher than the 46% produced by the untransformed heterozygous plant (Table 2). Four of the T2 plants were homozygous (lpat2/lpat2) and contained lpat2 and LPAT2-c (Figure 10A). They produced 85% of the number of seeds per silique as that of the wild type; again, this proportion is substantially higher than the 46% produced by the untransformed, heterozygous plant (Table 2). We never found the genotype lpat2/lpat2 in the offspring from selfing the untransformed heterozygous mutant.

LPAT3-cDNA Functionally Complemented lpat2 in Part in Heterozygous (LPAT2/lpat2) and Homozygous (lpat2/lpat2) Mutants

Pollen of the genotype lpat2 were viable and could fertilize eggs (Table 1). We speculate that LPAT3, whose transcript was abundant in pollen, compensated for the loss of LPAT2. Recombinant LPAT3 produced by E. coli (Figure 3) and yeast (Figure 4) had LPAT activity in vitro. We tested whether LPAT3-c could functionally complement lpat2 in the ovules by transforming the heterozygous mutant (LPAT2/lpat2) with LPAT2P:LPAT3-c. The shoots and roots of several transformed LPAT2/lpat2 mutants contained LPAT2 in amounts similar among themselves and to the wild type (Figure 10B). They contained LPAT3 in amounts substantially higher than those in the shoots (not detectable) and roots (a low amount) of the wild type. Thus, LPAT3 driven by the LPAT2 promoter was expressed properly in the shoots and roots of the LPAT2/lpat2 mutants. The T3 plants (LPAT2/lpat2 + LPAT2P:LPAT3-c) were selfed, and the seeds produced from five selfed T3 lines were analyzed (Table 3). The mature siliques produced by these lines were intermediate in length between those of the wild type and the heterozygous mutant (Figure 10C). They contained three distinguishable types of seeds, aborted seeds, or ovules (Figure 10C). The first type appeared to be normal, maturing seed, having a heart-shaped embryo at 6 DAF and a fully formed embryo at 10 DAF. These seeds should have the genotypes of LPAT2/LPAT2 or LPAT2/lpat2 plus or minus LPAT2P:LPAT3-c. The second type was immature seeds containing an aborted embryo, which had reduced development compared with the maturing, normal embryos at 6 and 10 DAF. These aborted seeds did not germinate. They might have derived from lpat2 + LPAT2P:LPAT3-c female gametophytes fertilized with pollen of the genotype LPAT2 or lpat2 plus or minus LPAT2P:LPAT3-c. The third type was white ovules that apparently had not been fertilized. The female gametophytes of these aborted ovules could have the genotype of lpat2 without LPAT2P:LPAT3-c. The normal seeds, aborted seeds, and aborted ovules were in the ratio of 54:34:12. Although it is difficult to pinpoint the genotypes of these seeds and ovules, it is apparent that LPAT3-c allowed ovules of the genotype lpat2 to survive and be fertilized but did not allow the fertilized ovule to undergo full maturation into seeds.

Plant Materials and Growth Conditions

Seeds potentially containing a T-DNA inserted in the At3g57650 locus (LPAT2, Salk_058130) were obtained from the Salk Institute (http://signal.salk.eud/cgi-bin/tdnaexpress) via the Arabidopsis Biological Resource Center (Ohio State University, Columbus). These seeds and seeds of the wild type (ecotype Columbia), as well as various transgenic plants, were grown to flowering in a growth chamber maintained at 100 μE m⁻² s⁻¹ and 20°C under a 16-h-light/8-h-dark photoperiod.

For the studies of LPAT transcripts by RT-PCR, the following organs were collected: siliques of mixed development stages, unopened flowers, rosette leaves, and stems were obtained from mature plants; roots were collected from seedlings grown for 10 d on MS medium (Gibco, Grand Island, NY); developing embryos were dissected from seeds in siliques 10 (termed early maturation) and 18 (late maturation) DAF; and 2-d-old seedlings were obtained from seeds grown on MS medium. Mature pollens were harvested by the following procedure: 100 just-opened flowers were suspended in water in a tube, and the tube was vortexed. The released pollen suspended in water was collected on a Nitex filter (Tetko, Elmsford, NY) with 20 × 20-μm openings.

Seeds of the wild type and three lines of LPAT2/lpat2 + LPAT2P:LPAT3-c were grown for 10 d on agar plates containing MS medium. The plantlets were divided into shoots and roots for the analyses of the presence of LPAT2-3 transcripts.

Subcellular Fractionation of the Sporophytic Anther Extract

Brassica rapa anthers were chopped in 0.1 M Hepes-NaOH, pH 7.5, and 0.4 M sucrose with a razor blade and ground gently with a mortar and pestle. The homogenate was filtered through a Nitex cloth (20 × 20-μm pore size), which retained most of the microspores (∼25-μm diameter) and the unlysed outer anther cells that were more difficult to be broken than the cell wall–depleted tapetum cells. The filtrate was regarded as the crude extract, which represented materials mostly from the tapetum cells. It was centrifuged successively at 1000g for 10 min, 10,000g for 30 min, and 100,000g for 90 min to produce the 1-kg, 10-kg, and 100-kg pellets (resuspended in the same grinding buffer) and a 100-kg supernatant. The subfractions were subjected to SDS-PAGE and immunoblotting analyses.

DNA Database Search and Sequence Analyses

One plastid (LPAT1) and four putative cytoplasm (LPAT2-5) LPATs of Arabidopsis thaliana and related LPATs from other sources were identified from The Arabidopsis Information Resource (http://www.Arabidopsis.org) and GenBank via amino acid sequence similarities with use of a maize (Zea mays) LPAT (Z29518) and a Brassica LPAT (AF111161) as queries (Kim and Huang, 2004).

Protein sequence alignments among the four Arabidopsis putative cytoplasmic LPATs were conducted by the ClustalW algorithm (Thompson et al., 1994) with use of the residue substitution matrix (blosum62mt2) of the AlignX application of Vector NTI Suite (InforMax, North Bethesda, MD). A phylogenetic tree of the aligned sequences was built with use of the neighbor-joining method (Saitou and Nei, 1987).

The Arabidopsis MPSS database (http://mpss.udel.edu/at/java.html) was used to estimate the prevalence of transcripts of LPAT1-4 in cDNA libraries of five organs. The signatures of 17-mer tags generated via cutting the transcript cDNA in the libraries at the Sau3A restriction site were normalized to transcripts per million to facilitate comparisons among libraries. The transcripts per million were based on 1,963,474; 1,791,360; 2,885,229; 3,645,414; and 2,018,785 signatures in callus, inflorescence, leaf, root, and silique libraries, respectively.

RT-PCR and cDNA Cloning of LPAT Genes

RT-PCR analyses were performed with gene-specific primers for LPAT2-5 as described (Kim and Huang, 2004). Gene-specific primers were designed to correspond to the sequence diversity regions between the 3′ terminal coding region and the 3′ untranslated region. The respective forward and reverse primers for LPAT2-5 were 5′-GCGTACTAACTCTTGGAGCAA-3′ and 5′-CAAAACTGACACGCGCTTCTT-3′ for a 294-bp fragment, 5′-TTTCCTTGTATTCGGTGGTTTC-3′ and 5′-TGGTTGCAATTGTTAGACAACA-3′ for a 341-bp fragment, 5′-GAAACAACAACAGGTCACTAAAC-3′and 5′-TTACAATTCTACATGAGCTTGTA-3′ for a 461-bp fragment, and 5′-GATTGCCTTCACCACCATCTG-3′ and 5′-CCCAAGCTATGATCTAACATGTG-3′ for a 325-bp fragment.

Full-length cDNAs of LPAT2-3 were amplified by RT-PCR from RNA of flowers and cloned into a pGEM-T vector. The primers represented sequences at the 5′ terminal end of the ORF and the untranslated region. For LPAT2-3, the primers were (p1) 5′-TTGAGGATGGTGATTGCTGCA-3′ and (p2) 5′-GTTTTTACTTCTCCTTCTCCG-3′, and 5′-TCAGACATGAAGATCCCTGCG-3′ and (p5) 5′-GTCCTAAGCAGAAATAAGCTG-3′.

The primers, p1, p2, p5, (p3) 5′-ACTGAAACGTGGAAATACCAGTACATCCATA-3′, and (p4) 5′-ATTGCATAGTTCAAAACTTCAAAGTTGGCCA-3′ were used for genotyping LPAT2-3 and lpat2 in the wild type, mutants, and their transgenic plants. The uses of these primers are shown in Figure 10A.

DNA-PCR was performed to evaluate whether the above RT-PCR was performed at conditions that would reveal the quantities of the LPAT transcripts. It was performed with the same gene-specific primers used for the RT-PCR and LPAT cDNA from the respective clones as templates as well as the same PCR cycle numbers.

LPAT Promoter:GUS Construction and Histochemical Localization of GUS Expression

The promoter regions of LPAT2-5 were each chosen from the sequence 5′ upstream of the predicted translation initiation site and the adjacent 5′ upstream gene. They were amplified by PCR with specific primers and leaf DNA as a template and then inserted into a pMOSBlue vector (Amersham Pharmacia Biotech, Piscataway, NJ) to produce unique recognition sites by restriction endonucleases. The specific primers for LPAT2-5 were 5′-GAGGATAAAATGCCAAAGATTGAATGA-3′ and 5′-CCTCAAAAAATCCCACCAACAAAAG-3′, 5′-TCTACGAGGCCAACGGAAGATCCCT-3′ and 5′-GTCTGAGAGAGATAAGGAACAACAC-3′, 5′-TGCTCCGCTTCTACGGAACCACTTC-3′ and 5′-TTCCAGATAAACCTCAAGACAAAACTGC-3′, and 5′-GACTGAGAAGGAGAAGCAGCAGATT-3′ and 5′-ACACCTTAAGATAATCTAATTAACCTCCC-3′, respectively. Each of the inserts was confirmed by DNA sequencing.

For each of LPAT2-5, an LPAT promoter:GUS was constructed by inserting the above-mentioned promoter sequence, cut from pMOSBlue-LPAT, into pBI101 in frame with GUS. The products of LPAT2P:GUS, LPAT3P:GUS, LPAT4P:GUS, and LPAT5P:GUS had a 3.0-kb XbaI-Smal fragment, a 2.0-kb HindIII-SmaI fragment, a 1.8-kb HindIII-SmaI fragment, and a 0.9-kb XbaI-SmaI fragment, respectively, representing the promoters. T1 or T2 plants showing kanamycin resistance were investigated for GUS expression patterns.

GUS staining of organs was performed with use of the substrate 5-bromo-4-chloro-3-indolyl β-d-glucuronide according to the method described by Jefferson (1987). Potassium ferricyanide and ferrocyanide (3 mM) were added to limit diffusion of GUS reaction products. Samples were incubated in a β-d-glucuronide solution at 37°C for 20 h. After incubation, samples were cleared in ethanol and photographed.

DNA and RNA Hybridization of Leaf Samples from Plants of Various Genotypes

DNA from plants of various genotypes was isolated from leaves with use of a CTAB procedure described by Dellaporta et al. (1983) and digested with PstI. One PstI site is at the opposite ends of LPAT2 and 1 is in the T-DNA. The inserted T-DNA, LPAT2, and lpat2 were detected by hybridization of genomic DNA with a ³²P-labeled DNA probe. The probe was obtained from DNA of the heterozygous mutant with use of primers p1 and LBa1 and labeled with ³²P-dCTP with use of the Multiprimer DNA labeling kit (Amersham Phamacia Biotech). The probe was ∼1.6 kb and contained 1.2 kb of the 5′ terminal portion of the LPAT2 coding region and 0.4 kb of the 5′ terminal portion of the T-DNA. The procedure for hybridization was as described previously (Kim et al., 2001).

Total RNAs were isolated from leaves as described (Verwoerd et al., 1989). RNA hybridization probe and the procedure were as described for DNA hybridization.

Expression of LPATs in Escherichia coli JC201 and Assay for in Vitro LPAT Enzyme Activities

Full-length fragments of cDNAs of LPAT2-5 were amplified by RT-PCR from RNA of flowers. Unique restriction sites were generated by PCR: BamHI-HindIII with primers 5′-GGATCCGATGGTGATTGCTGCAGCTGTC-3′ and 5′-AAGCTTTTACTTCTCCTTCTCCGTTTC-3′ for LPAT2; BamHI-SalI with 5′-GGATCCCATGAAGATCCCTGCGGCTCTT-3′ and 5′-GTCGACCTAAGCAGAAATAAGCTGTTC-3′ for LPAT3; SphI-HindIII with primers 5′-GCATGCGATGGAAGTTTGCGGGGATCTG-3′ and 5′-AAGCTTTTAATGGTTTTTAACTTTACAAGAATC-3′ for LPAT4; and SphI-HindIII with primers 5′-GCATGCAATGGAAAAAAAGAGTGTACCA-3′ and 5′-AAGCTTTTATTTGTTTACTAATTTGAGGGAATT-3′ for LPAT5. The fragments were ligated in frame into the pQE31 expression vector. The constructs were transformed into the E. coli strain, JC201, which had been previously transformed with pREP4 (Qiagen, Valencia, CA) that confers kanamycin resistance and expresses the lac repressor protein LacI. The bacteria were grown on Luria-Bertani agar at 30°C and 42°C. Successful functional complement would be achieved if the transformed bacteria grew at 42°C after IPTG induction.

The following procedure was used to prepare bacteria extracts for LPAT enzyme assay. A 1-mL overnight culture in a Luria-Bertani liquid medium was used to inoculate a 200-mL medium containing 100 μg/mL of streptomycin, 50 μg/mL of ampicillin, and 25 μg/mL of kanamycin. The culture was allowed to grow at 30°C until the A₆₀₀ reached ∼0.5. IPTG was added to a final concentration of 0.2 mM, and induction was allowed to continue for 1 h. The culture did not grow well beyond this 1-h period, which indicates the lethality of LPAT2-5 in the bacteria cells. After the 1-h induction, the cells were pelleted by low-speed centrifugation, and the pellet was resuspended in 4 mL of 50 mM Tris-HCl, pH 8.0, 2 mM MgCl₂, and 2 mM DTT. The resuspended cells were disrupted by sonication with a 40T probe in a Braun-Sonic 2000 ultrasonic generator (Freeport, IL) with a digital meter reading of 200. The total extract was centrifuged at 10,000g for 15 min at 4°C to remove unbroken bacteria and debris. The supernatant was centrifuged at 100,000g for 1.5 h at 4°C. The pellet containing membranes was resuspended in 1 mL of the above buffer and stored at −80°C. It was used to assay for LPAT activity with LPA-18:1 and ¹⁴C-CoA-18:1 or CoA-16:0 as described previously (Kim and Huang, 2004) and for SDS-PAGE and immunoblotting analyses.

Expression of LPATs in Saccharomyces cerevisiae and Assay for in Vitro LPAT Enzyme Activities

Full-length ORF fragments of cDNAs of LPAT2-5 were reamplified by PCR from LPAT cDNA clones with the following primers: 5′-GCGATGGTGATTGCTGCAGCTGTC-3′ and 5′-CTTTTACTTCTCCTTCTCCGTTTC-3′ for LPAT2, 5′-GCCATGAAGATCCCTGCGGCTCTT-3′ and 5′-GACCTAAGCAGAAATAAGCTGTTC-3′ for LPAT3, 5′-GCGATGGAAGTTTGCGGGGATCTG-3′ and 5′-CTTTTAATGGTTTTTAACTTTACAAGA-3′ for LPAT4, and 5′-GCAATGGAAAAAAAGAGTGTACCA-3′ and 5′-CTTTTATTTGTTTACTAATTTGAGGGA-3′ for LPAT5.

The cDNAs were cloned into pYES2.1 TOPO TA cloning vector with use of topoisomerase (Invitrogen, Carlsbad, CA) to produce pYES-LPAT2-5. The plasmids with the LPAT inserts were first transformed into competent E. coli, TOP10F′. Plasmid DNAs from the resulting bacterial colonies were isolated and confirmed of their inserted LPAT sequences by sequencing with use of the GAL1 forward primer (5′-AATATACCTCTATACTTTAACGTC-3′) and V5 C-term reverse DNA sequencing primer (5′-ACCGAGGAGAGGGTTAGGGAT-3′). Then, the plasmids containing LPATs were transformed into S. cerevisiae, strain INVSc1, and the transformed cells were selected on plates containing the SC-U minimal medium (0.67% yeast nitrogen base, 2% glucose, complete supplement mixture lacking uracil, and 20 g/L agar).

DNA was extracted from the transformed yeast with use of a method described by M. Kaeberlein (http://web.mit.edu/guarente/protocols/quickprep.html), and the presence of LPATs of the expected sequences was confirmed by obtaining PCR fragments with use of the above-mentioned gene-specific primers. Each transformed yeast was grown to the late-log phase in the SC-U medium containing 2% glucose. The cells were harvested by centrifugation and inoculated at a concentration of OD₆₀₀ = 0.4 in an induction medium (SC-U medium containing 2% galactose). The induction was allowed to proceed for 4 h. The galactose would induce high-level expression of the transgene controlled by the GAL1 promoter in the plasmid vector.

The yeast cells were harvested by centrifugation, washed once with water, and harvested again. The washed cells were resuspended in 50 mM Na phosphate, pH 7.4, 1 mM EDTA, 5% glycerol, and 1 mM PMSF and broken with glass beads and a mortar and pestle. The samples were centrifuged at 8000g for 15 min at 4°C to remove cell debris. The resulting supernatant was the lysate, which was used for protein analyses by SDS-PAGE and LPAT activity assays (preceding section). The lysate was also centrifuged at 100,000g for 90 min to yield a membrane pellet for use in immunoblotting.

Preparation of Antibodies against LPAT2 and SDS-PAGE and Immunoblotting

Among the four putative LPAT2-5, LPAT2 has a unique peptide, KPKDNHHPESSSQTETEKEK, at the C terminus (Figure 2). All other Arabidopsis proteins do not have a sequence similar to this unique peptide. The Brassica LPAT2 related to Arabidopsis LPAT2 has a similar sequence, KPKDNHQSGPSSQTEVEEK, at its C terminus. A synthetic peptide corresponding to the Arabidopsis LPAT2 peptide was made, and rabbit antibodies against it were prepared via a procedure described earlier (Wang et al., 1997). Antibodies against castor bean calreticulin were kindly provided by Sean Coughlan and Tony Kinney.

SDS-PAGE and immunoblotting of the gel were performed by procedures described earlier (Kim and Huang, 2004). Protein quantities in the bacterial and yeast extracts were determined with the Bradford method (Bradford, 1976) and BSA as the standard.

Immunofluorescence Microscopy

Parafilm sections 7 μm thick of Arabidopsis flower buds and B. rapa anthers were used. Double immunolabeling with two polyclonal antibodies was performed as previously described, with modifications (Li et al., 2002). Primary antibodies included rabbit anticalreticulin (1:1000) against the protein from castor bean, chicken anti-PAP (1:500; Brassica plastid lipid-associated protein) (Kim et al., 2001), and rabbit anti-LPAT2 (1:50). Organ sections on slides were incubated with the primary antibodies (anticalreticulin) at 4°C overnight. After the antibody solution was removed, the sections were washed with PBST (0.05% Tween 20 in PBS) three times for 5 min each. The sections were incubated with Cy3-conjugated Fab fragment goat anti-rabbit IgG (H+L) (1:50 dilution) at room temperature for 3 to 5 h in the dark. The sections were rinsed briefly and post-fixed with 4% paraformaldehyde in 50 mM Na-phosphate buffer, pH 7.4, at room temperature for 20 min. After being washed with PBST three times for 5 min each, the sections were incubated with primary antibodies (rabbit anti-LPAT2 or chicken anti-PAP) at 4°C overnight in the dark and then Cy5-conjugated goat anti-rabbit IgG (H+L) (1:200 dilution) at room temperature for 1 h. A drop of antifade solution (Molecular Probes, Eugene, OR) was added to the section, which was then covered with a cover slip and sealed with nail polish. All confocal images were collected with the use of a Leica TCS SP2 confocal microscopy system (Leica Microsystems, Wetzlar, Germany), and three-dimensional images were created with the combination of multiple serial optical sections. Images were processed with the use of Adobe Photoshop software (San Jose, CA).

Construction of LPAT2P:LPAT2-cDNA and LPAT2P:LPAT3-cDNA and Their Transformation into a Heterozygous (LPAT2/lpat2) Mutant

LPAT2P:LPAT2-c and LPAT2P:LPAT3-c, containing 3.0 kb of the LPAT2 promoter and the full-length cDNA of LPAT2 and LPAT3, respectively, were constructed. GUS in pBI101 harboring LPAT2P:GUS was replaced with either LPAT2-c or LPAT3-c, which were obtained from pGEM-T containing these cDNA. pBI101 harboring LPAT2P:LPAT2-c or LPAT2P:LPAT3-c was transformed into heterozygous (LPAT2/lpat2) mutants by the floral dip method with use of the Agrobacterium tumefaciens strain GV3101. Selection for kanamycin resistance was made via nptII in pBI101 because the T-DNA in the LPAT2/lpat2 heterozygous mutant did not confer kanamycin resistance. Kanamycin-resistant seedlings (T1) on a Petri dish were transferred to soil. The T2 and T3 generation plants were examined for lpat2, LPAT2, and LPAT3 by PCR and for seed appearance by stereomicroscopy (LEICAMZ125; Leica Microsystems).

Phenotypes and Genetic Analyses of LPAT2/lpat2 Mutants and Transgenic Plants

Thirteen seeds potentially containing lpat2 (a T-DNA inserted into LPAT2) were obtained from the Salk Institute. They were grown to mature plants, and leaves were analyzed for LPAT2 and lpat2 by PCR. No lpat2/lpat2 individual was found. Five LPAT2/lpat2 heterozygous individuals identified all had vegetative parts quite similar to those of wild types, with a minor exception that their leaves were slightly elongated. They all produced siliques, each of which was half filled with seeds and half with aborted ovules or embryos. The progeny of each of these five heterozygous individuals were kanamycin sensitive, even though the T-DNA inserted into LPAT2 was supposed to contain nptII. One of the five heterozygous individuals was found to contain only one copy of T-DNA per diploid genome, as revealed by DNA hybridization analysis. From this individual, a 1.6-kb DNA fragment was generated by PCR with p1 and LBa1 primers and sequenced. The sequence confirmed that LPAT2 was inserted by a T-DNA in the fifth intron after nucleotide 21,362,009 in chromosome 3 (Figure 2). This individual was used for all subsequent studies.

The embryo sacs of the wild type and the heterozygous mutant were observed by microscopy. The pistils of different developmental stages were cut into 1-mm segments and fixed, dehydrated, infiltrated, and embedded in Embed 812 (Electron Microscopy Sciences, Hatfield, PA) as described previously (Christensen et al., 1997). Thin sections (10 μm) were stained with tuluidine blue for light microscopy with a Nikon MICROPHOT-FXA (Tokyo, Japan) attached to a Spot digital camera or lead citrate and uranyl acetate for electron microscopy (Platt et al., 1998).

Lipid Analysis

Rosette leaves were immediately frozen in liquid nitrogen after being harvested, and the lipids were extracted according to the protocol of the Lipodomics Center at Kansas State University (http://www.ksu.edu/lipid/lipidomics/leaf-extraction.html). The acyl moieties of the lipids were saponified with alkaline in ethanol, and the acidified samples of free fatty acids were derivatized to methyl esters with boron trifluoride in methanol, which were subjected to gas–liquid chromatography analysis.