A novel group of oleosins is present inside the pollen of Arabidopsis.

In plants, subcellular triacylglycerol granules in seeds (oil bodies) and floral tapetum (tapetosomes) are stabilized by amphipathic structural protein called oleosin. We hereby report a novel group of oleosins that is present inside the pollen of Arabidopsis thaliana. We have used the conserved sequence of oleosins to locate, via the DNA database, all 16 oleosin genes in the Arabidopsis genome. The oleosin genes can be divided into three groups according to their sequences and tissue-specific expressions, as probed by RNA blot hybridization and reverse transcriptase-PCR. The first group includes eight genes specifically expressed in the floret tapetum. The second group includes five genes specifically expressed in maturing seeds. The third, novel group includes three genes expressed in both maturing seeds and floral microspores, which will become pollen. Transgenic study using the promoter of one of these genes attached to a reporter gene has provided corroborative evidence for the specific expression of the gene in the microspores in the florets. One of the pollen oleosins can be identified by microsequencing and specific immunoblotting. Pollen oleosins synthesized by recombinant bacteria can collaborate with phospholipids in stabilizing reconstituted oil bodies. Thus, pollen has oleosins to stabilize the abundant subcellular oil bodies. Seed oil bodies and floret tapetosomes have been isolated from the miniature Arabidopsis plants, and the success indicates that the organelles can be subjected to future biochemical and genetic studies.

Eukaryotes possess intra-and extracellular granules that contain high amounts of neutral lipids, which are usually triacylglycerols (TAGs) 1 and steroid esters. The lipid granules are covered by a layer of amphipathic lipids, such as phospholipids (PLs) and glycolipids, and amphipathic proteins. These proteins, together with the amphipathic lipids, stabilize the neutral-lipid granules in an aqueous environment and may perform additional functions such as being recognition signals for the binding of intra-and extracellular lipid hydrolases and cell surface receptors. In mammals and insects, the extracellular lipoprotein granules are covered by apolipoproteins (1,2). In mammalian adipose cells and other special cells, the intracellular lipid granules are covered by proteins called perilipin and adipocyte-differentiation-related protein (3). In mammalian glands, the intracellular lipid granules are also covered by adipocyte-differentiation-related protein and are further coated with a layer of modified plasma membrane during cellular secretion (4,5). In yeast, the intracellular storage oil granules possess lipid-synthesizing enzymes, and whether they contain structural proteins is unknown (6,7). In plants, the lipid granules inside the plastids are covered by proteins termed plastid-lipid-associated proteins (8,9). In plant seeds, the intracellular storage TAG granules, called oil bodies, are covered by proteins named oleosins (10 -12). In plant floral tapetum cells, the TAGs in the tapetosomes are associated with oleosins (13,14). In prokaryotes that store polyhydroxyalkanoate, the storage granules are covered by proteins termed phasin (15).
All of the above proteins interact with the surface constituents of the lipid granules by virtue of their amphipathicity. Of these proteins, only oleosins in plants contain a long stretch of uninterrupted hydrophobic or neutral residues (10,11). The stretch has about 72 residues, and its center possesses three proline and one serine residues, which would allow the stretch to bend and form a hairpin. The hairpin would penetrate into the core of the lipid granules. Its sequence is highly conserved among oleosins of diverse plant species and represents the hallmark of all oleosins. The hairpin is flanked by less conserved, amphipathic amino and carboxylic portions, and the latter portion contains an amphipathic ␣-helix. The lengths of the amino and carboxylic portions are highly variable. As a consequence, oleosins have diverse molecular weights, usually in the range of 14,000 -45,000, depending on the isoforms and the plant species. The hydrophobic hairpin and the amphipathic amino and carboxylic portions allow the whole oleosin molecule to reside stably on the surface of the lipid granule.
In seeds, the oil bodies contain a matrix of TAGs enclosed by a layer of PLs and oleosins (10,11). Oleosins form a steric barrier on the surface of an oil body, preventing the PL layer from contacting and coalescing with the PL layers of adjacent oil bodies. Maintenance of the oil bodies as small entities provides a large surface area per unit TAG and would facilitate lipase binding and lipolysis during seed germination.
In florets, oleosins interact with the TAGs inside the tapetosomes, and the oleosins, but not the TAGs, are transferred from the tapetum cells to the surface of the adjacent maturing pollen (13, 16 -19). It has been speculated that the oleosins on the pollen surface aid attachment of the pollen to the female stigma surface and subsequently assist in the uptake of water for germination. Mutant pollen of Arabidopsis deficient in one of the pollen-surface oleosins could undergo germination and fertilization, although the water-uptake phase on the stigma is slightly prolonged (20). It is unknown whether the coat of the mutated pollen contains additional, compensatory amounts of other oleosins or whether it is also deficient in oleosin-associated coat lipids.
In addition to seeds and floret tapetum, pollen are also known to contain abundant TAGs in intracellular oil bodies (21). No oleosins have been previously found inside the pollen (18,22), and it was speculated that the pollen oil bodies were stabilized by the endoplasmic reticulum. This reported lack of oleosins in intracellular oil bodies in pollen would have strong implications on the necessity of having oleosins to stabilize the oil bodies and the mode of oil-body biogenesis.
The current literature indicates that oleosin isoforms are present in different cells within an individual plant. Many oleosin genes are known (10,18), but a comprehensive analysis of all the oleosin genes and the proteins in an individual plant has not been previously reported. The opportunity to do this analysis has become available with the completion of the sequencing of the Arabidopsis genome (23). Earlier, it was reported that in Arabidopsis, four oleosin genes, which are expressed in the floret anthers, are in tandem on a 10-kb DNA fragment (24,25). This finding has been expanded to include six oleosin genes in tandem on chromosome V, via the available Arabidopsis genome sequence, and the oleosins encoded by five of these genes have been found on the pollen surface (19).
A comprehensive analysis of all the oleosin genes in Arabidopsis will provide important scientific information. Therefore, we have used the conserved hairpin sequence of oleosins to locate, via the DNA database (23), all of the oleosin genes in the Arabidopsis genome. Analyses of these oleosin genes and their encoded proteins have provided many new insights into the restrictions and variability of the genes and proteins. Importantly, we have discovered a group of oleosin genes that are expressed in both maturing seeds and microspores, which will mature to become pollen. We further show that the pollen oleosin can function, in collaboration with PLs, in stabilizing TAG droplets in vitro. Thus, pollen has oleosins to stabilize the abundant intracellular oil bodies. This is also the first report of the isolation of the seed oil bodies and the floret tapetosomes from the miniature Arabidopsis plants, and the success indicates that the organelles can be subjected to future biochemical and genetic studies.

EXPERIMENTAL PROCEDURES
Plant Materials-Arabidopsis thaliana, ecotype Col-0, was grown from seed to flowering plants in a greenhouse maintained at 26/18°C and 14/10 h day/night cycle. Unopened flowers (florets) and siliques of different developmental stages, and leaves and stems of mature plants, were obtained. Root tissues were collected from seedlings grown for 10 days on an MS medium.
For developmental studies of the flowers, the florets were divided into five stages according to the flower bud length. For stages 1 through 5, the lengths of the buds were 0.5, 1.0, 2.0, 2.5, and longer than 2.5 mm. In stage 5, the flowers had opened. In our system, stages 1, 2, 3, 4, and 5 are roughly equivalent to the stages 7-8, 9, 10, 11-12, and 13, respectively, described earlier (26).
For developmental studies of the seeds, the siliques were divided into seven stages according to their color, length, position along the inflorescence, and firmness. For stages 1-5, the siliques were all green and their lengths were 0.7, 1.0, 1.2, 1.3, and 1.5 cm. For stage 6, they were yellow and their length was 1.5 cm. For stage 7, they were brown and opened, and their seed was dry.
For studies of the proteins inside the pollen, mature pollen was collected from flowering branches which contained freshly opened flowers at the terminal region. The flowering branches were placed in a 12-ml conical centrifuge tube, and a solution of 0.1 M K-phosphate, pH 7.0, was added. The tube was vortexed and then centrifuged at 2,000 rpm in Rotor 958 in an IEC-HN-SII microcentrifuge for 2 min. The pelleted pollen and small floret parts were resuspended in the buffer, and the resuspension was filtered through a 20 ϫ 20-m Nitex filter. The pollen on the filter was collected and air-dried in a hood. The surface materials were extracted with diethyl ether (13), and the leftover pollen was collected.
Identification of 16 Arabidopsis Oleosin Genes-The 72-amino acid sequence of the highly conserved hairpin region of the 45-kDa oleosin of Brassica tapetum (residue number from the N terminus, 39 -112; L33283; Ref. 27) and the 20-kDa oleosin of Arabidopsis seed (residue number from the N terminus, 46 -117; CAB36756; Ref. 28) were used as queries for BLAST database searches in blastp and tblastn. Each of the two searches produced the same 16 oleosin genes.
Analyses of DNA and Amino Acid Sequences-Map positions of the 16 oleosin genes were located on the Arabidopsis Genomic Initiative map using TAIR databases (www.arabidopsis.org). Analyses of amino acid sequence similarities and constructions of phylogenetic trees were performed according to the Clustal method with the use of Vector NTI Suite software (InforMax, North Bethesda, MD).
Gene-specific Primers for the 16 Oleosin Genes and Other Genes-For each of the 16 oleosin genes, a gene-specific fragment was obtained using the gene-specific forward and backward primers, described in the following paragraph. We attempted to ensure that the gene-specific nature of each probe, and thus of the RNA blot hybridization result, was valid on the basis of several measures: (a) the sequence of each gene probe bearing no or low similarity with the sequences of the other oleosin genes and their 3Ј-untranslated region was chosen, (b) the wash in the RNA blot hybridization was carried out under a high stringency condition (30 min in 0.1 ϫ SSC and 0.1% SDS at 65°C), (c) only one transcript on the RNA blot was detected and had the expected length as predicted from the gene sequence, and (d) many of the oleosin genes had different patterns of developmental expression (to be shown in Fig. 3). Despite these measures, we can only state that the possibility of crosshybridization in the RNA blot result was greatly reduced but not eliminated.
The PCR fragment included the 3Ј-terminal portion of the ORF and might also encompass a small portion of the 3Ј-untranslated region. For PCR Amplification of Gene-specific Fragments-PCR was performed using Arabidopsis genomic DNA from the leaves (9) as a template via 30 cycles of 94°C for 1 min, 55°C for 1 min, and 72°C for 1 min. The electrophoresis profile of the PCR product was a single band of the expected length. The PCR fragment was purified with the use of QIAGENE MAX Kit (Qiagen, Valencia, CA) and then 32 P-labeled with the use of Multiprimer DNA Labeling Kit (Amersham Biosciences, Piscataway, NJ).
RNA Blot Hybridization and Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR) Analyses-Total RNAs were isolated from various organs (9). Thirty to 40 g of total RNA of each sample was fractionated on a 1.3 or 1.5% formaldehyde gel, and the separated RNAs were transferred to a Hybond N membrane (Amersham Biosciences). The membrane was pre-hybridized at 65°C in 0.5 M K-phosphate, pH 7.2, 7% SDS, 1% bovine serum albumin, and 0.1 M EDTA, for 2 h, hybridized with 32 P-labeled gene-specific probe, and then washed with 2 ϫ SSC, 0.1% SDS for 20 min, 1 ϫ SSC, 0.1% SDS for 30 min, and 0.1 ϫ SSC, 0.1% SDS for 30 min, all at 65°C.
For RT-PCR analysis, 1.5 g of total RNA of each sample was used to synthesize a first-stranded cDNA using oligo(dT) 15 primer (29). This cDNA was used as a template in the presence of the pair of gene-specific primers described earlier for RT-PCR. The gene-specific fragment was analyzed with a 1.4% agarose gel.
Isolation of Seed Oil Bodies and Floret Tapetosomes-Seed oil bodies were isolated from mature seed by flotation centrifugation as described (30). Tapetosomes were isolated from stage 3 florets by step-gradient centrifugation as described earlier (13). The tapetosome fraction at the interface between 0.6 and 0.4 M sucrose was collected.
Transformation of Arabidopsis with a Chimeric SM2:GUS (␤-Glucuronidase Gene) Construct and Assay of GUS Activity-There are 367 bp between the ORF of SM2 and the ORF of the 5Ј upstream, adjacent gene. This 367-bp sequence was amplified by PCR with the use of the primers, 5Ј-AAGCTTTTTAACATACAAAAAATTGTTGTC-3Ј and 5Ј-G-GATCCTGTAGAGAGAGCGATTGTTGTTTTGT-3Ј, and Arabidopsis DNA as a template. The fragment was cloned into the HindIII/BamHI sites of pBI101, and the plasmid was transformed into Arabidopsis by the floral dip method (31) with the use of Agrobacterium tumefaciens, strain GV3101 (pMP90). Twenty individual T1 plants resistant to 50 g/ml kanamycin monosulfate were generated, and their floral tissues were examined by light microscopy for GUS activity with the use of 5-bromo-4-chloro-3-indolyl ␤-D-glucuronide (32).
Synthesis of SM3 Oleosin in Recombinant Bacteria-Two recombinant constructs, pSM3 and pSM3N, containing the full ORF of the SM3 gene and a sequence encoding the N-terminal 48 residues (MVSLL----HESSP, to be shown in Fig. 2) of SM3, respectively, were made. This N-terminal peptide is unique to SM3 among all the Arabidopsis oleosins. pSM3 was constructed by inserting a 525-bp BamHI-PstI fragment obtained by PCR using Arabidopsis DNA as template and primers P1 (5Ј-AGGATCCAATGGTAAGCTTATTAAAGCTACAAAAAC-3Ј) and P2 (5Ј-ACTGCAGCTAATGAGTAGTTGTTGTGGTTTGG-3Ј) into pRSETB vector (Invitrogen, Carlsbad, CA). pSM3N was constructed by the same procedure except that the BamHI-PstI fragment had 143 bp, made with primers P1 and P3 (5Ј-CCTGCAGATGGTGATGATTCATG-GAGACTC-3Ј). The sequences of the two fragments were confirmed by DNA sequencing. The constructs were transformed into Escherichia coli BL21(DE3) containing pLysE (Invitrogen), which were induced to produce the recombinant proteins with 1 mM isopropyl-␤-D-thiogalactopyranoside . The bacterial extracts were applied to a nickel-nitrilotriacetic acid-agarose resin column (Qiagene, Valencia, CA), and the recombinant proteins were eluted with a buffer (8 M urea, 0.1 M NaH 2 PO 4 , 0.01 M Tris-HCl, pH 4.5). The eluted preparations were dialyzed against 0.05 M K-phosphate buffer, pH 7.4. SDS-PAGE analysis showed that the recombinant proteins produced by pSM3 and pSM3N had the expected 22 and 8 kDa, respectively. The 22-kDa protein (SM3 oleosin) and the 8-kDa protein (N-terminal portion of SM3) were used for functional test and production of antibodies in chicken (13), respectively.
SDS-PAGE, Immunoblotting, and N-terminal Sequencing-All procedures followed those described earlier (13). Proteins were separated by 12.5% (w/v) SDS-PAGE, and the gel was stained with Coomassie Blue or silver stain. For immunoblotting, proteins separated on an acrylamide gel were transferred to a nitrocellulose membrane. The membrane was subjected to immunodetection with the use of chicken antibodies raised against isolated Brassica seed 20-kDa oleosins (30) or a bacteria-synthesized N-terminal fragment of SM3 oleosin (described in the preceding paragraph), or rabbit antibodies raised against a synthetic peptide of KRRMCVKPKDNP via a procedure described earlier (16). This peptide is present in Arabidopsis T3 oleosin adjacent to the hairpin stretch at the carboxylic side; its modified versions are present in other T oleosins. For sequence analyses, proteins separated on an acrylamide gel were transferred to a polyvinylidene difluoride membrane for N-terminal sequencing directly or after treatment with trypsin and high performance liquid chromatography separation of the fragments.
Test of the Function of Pollen Oleosin in Stabilizing Triacylglycerol Droplets-Triacylglycerol droplets in a suspension coalesced and floated to the top, and the floatation rate decreased if the droplets were stabilized by oleosins and phospholipids. The floatation resulted in a decrease in the absorbance of the lower portion of the mixture, and the change of this absorbance, and thus the floatation rate, was measured as follows. A 1.2-ml mixture containing 0.05 M K-phosphate buffer, pH 7.4, 2 mg of triolein, in the absence or presence of 0.02 mg of phosphatidylcholine (egg yolk, from Sigma) or phosphatidylserine (bovine brain, from Sigma), and 0.04 mg of SM3 oleosin (synthesized by bacteria as described in the preceding section) was sonicated with a 40T probe in a Braun-Sonic 2000 ultrasonic generator with a digital meter reading of 200 for 30 s. Light microscopic examination indicated that the sizes of the droplets in the various samples were the same irrespective of the presence or absence of PLs and/or oleosin. The sample was transferred to a 1.8-ml cuvette, which had a 1-cm light path. The absorbance at 600 nm of the approximately lower 0.5 ml of the mixture was measured in a Beckman DU-20 spectrophotometer. The initial absorbance was 1.5-1.7.

RESULTS AND DISCUSSIONS
The Arabidopsis Genome Has 16 Oleosin Genes-A search of the Arabidopsis genome database for genes containing the conserved hairpin sequence of oleosins has revealed 16 oleosin genes. About half of these genes have been reported previously as studied clones or expressed sequence tag sequences ( Table  I). The floret oleosin genes in Arabidopsis were initially termed Atgrp-(encoding glycine-rich proteins) because some of them are glycine-rich. The glycine-rich aspect is likely to be fortuitous due to the short repeats at the C termini in a few of the oleosins and not to reflect functionality (to be explained later under "Results and Discussions"). The original names of those genes that have been reported are listed in Table I. In the current report, highly simplified names of all the genes termed on the basis of their tissue-specific expressions are used solely for clarity in presentation. T, S, and SM genes are those expressed in the tapetum, the seed, and the seed-and-microspores, respectively (as described under "The 16 Oleosin Genes Can Be Divided into Three Groups on the Basis of Their Tissuespecific Expression"). The 16 genes are scattered throughout the five chromosomes of Arabidopsis (Fig. 1). Nevertheless, all the eight T genes are on chromosome V. Six of the T genes are arranged in tandem; T7 is separated from this cluster by three unrelated genes of ϳ10 kb; and T8 is widely separated from the gene cluster. The current result of the location of the T genes confirms and expands earlier findings of the occurrence of four oleosin genes (T2-5) in tandem via the sequence of a DNA fragment (22, 23), and six oleosin genes (T1-6) in tandem via the newly available Arabidopsis genome sequence (19). These earlier studies have shown that the oleosin genes are expressed in the florets (24,25) or that the encoded oleosins are present on the pollen surface (19). The expression of the T genes in the florets and the subsequent location of the oleosins on the pollen surface are consistent with the findings in Brassica, that the T oleosins are synthesized in the tapetum and then transferred to the pollen surface (13,14,17,18).
None of the 16 oleosin genes is a pseudogene. They all have a long ORF encoding a characteristic oleosin sequence (Table I and Fig. 2) and are expressed (next section). Three of the genes do not have an intron, whereas the other 13 genes have one intron preceding or immediately following the sequence encoding the hairpin stretch. The 50 -60 oleosin genes of diverse species that have been studied previously possess either no intron (in maize, Ref. 33) or one intron immediately following the sequence encoding the hairpin stretch (in Arabidopsis and Brassica, Refs. 27 and 34, respectively). The new information is that S1, S2, and S4 genes have an intron immediately preceding the sequence encoding the hairpin stretch. The ϳ15 oleosin genes of Arabidopsis and Brassica that have been studied previously all have one intron. SM1, SM2, and SM3 genes are unique in that they have no intron. The implication of having no intron, or one intron preceding or immediately following the sequence encoding the hairpin stretch, on the evolution of the oleosin genes remains to be seen.
The 16 Oleosin Genes Can Be Divided into Three Groups on the Basis of Their Tissue-specific Expression-We probed the expression of the 16 genes by detecting their transcripts via RNA hybridization using gene-specific probes (Fig. 3). RNAs from maturing seed, florets, leaves, stem, and roots were used. We further detected the locations of the transcripts of selected genes within the florets by the following method. We used microscopic devices to cut individual florets into the anthers, pistils, and petals plus sepals, and then separate the anthers into a microspore fraction (not contaminated by other anther cells, as observed by light microscopy) and the remaining anther-wall fraction (containing the tapetum and other anther wall cells, as well as 10 -20% of the original microspores). The extracted RNAs from these floret subfractions were of minimal quantities (about 1 g of RNA per sample from 20 florets), which were inadequate for RNA blot hybridization. They were subjected to RT-PCR for the detection of PCR fragments of individual gene transcripts with the use of gene-specific primers (same as those used to make probes for RNA blot hybridization). We believe that results from this RT-PCR test were more reliable than those obtained by in situ hybridization, which could be influenced by subjective manipulations of the hybridization conditions. All the 16 oleosin genes were expressed, as revealed in results from RNA blot hybridization (Fig. 3). Their transcripts were detected in maturing seed and/or florets but not in leaves, stems, and roots. The lengths of the transcripts are roughly proportional to those of the ORF of the respective genes ( Table  I). The results from the RNA blot hybridization, coupled with those from RT-PCR and transformation studies (as described under "The Three SM Genes Encode a Novel Group of Oleosins") have allowed the division of the 16 oleosin genes into three groups on the basis of their expressions.
The first group consists of eight T genes, which were expressed specifically in the florets (tapetum) (Fig. 3). During floret development, the transcripts of these eight genes all peaked at stage 2, when the tapetum cells were accumulating the tapetosomes. A similar pattern of the expression of T2-4 genes in the floret anthers has been reported earlier (25).
The second group consists of five S genes, which were expressed only in maturing seed (siliques) (Fig. 3). During silique maturation, some genes were expressed early (S5), intermediate (S1, S3, and S4), or late (S2). The third group consists of three SM genes, which were expressed in both maturing seeds and florets (microspores).
The Three SM Genes Encode a Novel Group of Oleosins-None of the three SM genes or their encoded oleosins has been previously reported. In the maturing seed, the transcripts were detected at mid-late maturation (Fig. 3). In the developing florets, the transcripts peaked at stage 3, in which the tapetum had begun disintegration (see "Experimental Procedures" and Ref. 26) and the T transcripts had already declined (Fig. 3). Within the subfractions of stage 2 and stage 3 florets (Fig. 4), the three SM transcripts, as detected by RT-PCR, were present largely in the microspores and much less in the anther-wall  (19); T7, BAB11438, no report; T8, BAB09005, no report; S1, AAF01542, no report; S2, BAB02690, Atol2 (39); S3, CAB36756, Atol1 (28); S4, BAB11599, A32 (40); S5, BAAQ7384, Atol3, (39); SM1, AAF69712, no report; SM2, BAB02215, no report; SM3, AAC42242, no report. The chromosome loci of the genes, from T1 to T8, S1-S5, and SM1 to SM3 are, respectively, At5g07510, At5g07520, At5g07530, At5g07540, At5g07550, At5g07560, At5g07600, At5g61610, At3g01570, At3g27660, At4g25140, At5g40420, At5g51210, At1g48990, At3g18570, and At2g25890. fraction, which contained the tapetum and other anther cells as well as about 10 -20% of the original microspores. It is likely that the SM transcripts in the anther-wall fraction were present in the leftover microspores rather than the tapetum. The SM transcripts were absent in the pistil and petal-plus-sepal fractions but present in the maturing seed, as were also revealed by RNA blot hybridization (Fig. 3). Other results from the RT-PCR analysis (Fig. 4) are also consistent with those from RNA blot hybridization (Fig. 3), in that the S3 transcript was present only in the maturing seed and the T6 transcript was restricted to the anther-wall fraction. The transcript of A3, whose Brassica and Sinapis counterparts are expressed only in the tapetum of the anthers (35,36), was present in the antherwall fraction (tapetum) and the petal-plus-sepal fraction but not in the microspore fraction. Actin-2 gene transcript was present in all the samples, as expected (37).
We examined further the tissue-specific expression of the SM genes, especially SM1 and SM2, in the anthers. The promoter sequence upstream of the ORF of SM2 was ligated in-frame to Gene-specific primers for the various genes were used. The florets were divided into pistils, petals plus sepals, and anthers, which were subdivided into a microspore and anther-wall fractions. The anther-wall fraction contained the tapetum and other anther-wall cells, plus about 10 -20% of the original microspores. Actin-2 gene, which is known to be expressed ubiquitously, was used to show that about the same amount of its transcript was present in each sample. The A3 gene, known to produce transcript in the tapetum in Brassica and Sinapis anthers, was used to show that the microspore fraction was not contaminated by tapetum RNA. the ORF of GUS, and the DNA construct was introduced into Arabidopsis. Among all the anther cells in the transformed plants, only the microspores exhibited GUS activity (Fig. 5). In stage 3 anthers, a small percentage of the microspores, whereas in stage 4 anthers, most of the microspores displayed GUS activity. This finding of GUS activity being higher in stage 4 than in stage 3 anthers reflects the peaking of the SM2 transcript in stage 3 anthers (Fig. 3). Under the same GUS assay conditions, stage 3 and stage 4 anthers of untransformed plants showed no GUS activity.
We conclude that the SM genes were expressed mostly if not completely in the microspores in the anthers. This conclusion is based on the above mentioned observations. (a) Their transcripts peaked at stage 3 when the tapetum had begun disintegration. (b) Their transcripts were present largely in the microspore fraction and much less in the anther-wall fraction isolated from stage 2 and stage 3 anthers. (c) SM-promoter-GUS was expressed only in the microspores in the anthers of transformed plants. Among the three SM genes, SM3 transcript was the most microspore-specific and present at a low level in seed, whereas SM1 and SM2 transcripts were present at a substantially higher level in seed than in microspores (Figs. 3 and 4).
The three SM genes, which form the third oleosin group, have not been reported previously in Arabidopsis or any other species. They have a novel expression pattern. The documentation of their expression in the microspores makes advances from earlier reports of no oleosins in the pollen and proposal of the pollen oil bodies being stabilized by the endoplasmic reticulum (18,22). The SM genes also share some characteristics not found in other Arabidopsis oleosin genes. They do not have an intron (Table I), and their sequences are relatively dissimilar to those of the other genes (Fig. 2).
Analyses of the Deduced Amino Acid Sequences of the 16 Oleosin Genes Generate New Insights into the Properties of the Oleosins-All 16 deduced oleosins have a complete hairpin stretch of ϳ72 hydrophobic or neutral residues (Table I and Fig. 2). In Brassica, the most abundant tapetum oleosin, of 45 kDa (7,14), possesses only 3 ⁄4 of the hairpin stretch on the amino side, and the remaining 1 ⁄4 of the stretch is highly hydrophilic. This is the only oleosin of any source known to possess an incomplete hairpin stretch. The equivalent Arabidopsis oleosin is T3 oleosin, as judged by a comparison of the Brassica DNA gene sequence and its deduced amino acid sequence with those of the Arabidopsis oleosin genes (data not shown). T3 oleosin does have a complete hairpin stretch of 72 hydrophobic residues, as do all the other Arabidopsis oleosins. The finding indicates that the lack of a complete hairpin stretch in the Brassica 45-kDa oleosin, albeit its being the predominant tapetosome oleosin in Brassica rapa, may be an aberration, rather than a characteristic of tapetosome oleosins.
The M r of the 16 Arabidopsis oleosins range from 10,000 to 53,000 (Table I), and this range is wider than the 14,000 -45,000 known previously. Because the conserved hairpin stretch of an oleosin has ϳ72 residues, the diverse M r of different oleosins reflect the length variability of the amino and carboxylic portions. The amino portions of the S and SM oleosins usually have 30 -50 residues (Table I and Fig. 2), and the sequences are amphipathic. The amino portions of the T oleosins are much shorter, and six T oleosins have only 6,9,13,13,13, and 14 residues. The function of the amino portion of an oleosin has not been defined, but is speculated to be related to guiding the nascent oleosin to the endoplasmic reticulum, to which the oleosin molecule will initially attach (11). Future elucidation of the function of the amino portion will have to take into account the very short amino portion sequences in many Arabidopsis oleosins. The carboxylic portions of the oleosins have 27-403 residues, which usually represent repeats of short sequences (Fig. 2). Some of these repeats have abundant glycine (e.g. 36, 39, and 33% glycine in the carboxylic portions of T1, T4, and T3 oleosins, respectively), and thus the oleosins were termed high-glycine proteins. This glycine abundance in the carboxylic portions is apparently fortuitous, because the short repeats can be rich in alanine (42% in T1 oleosin) or proline (22% in T5 oleosin and 21% in T6 oleosin). The function of the carboxylic portion is unknown, and the high variability in the length and residues suggests that it is not very specific.
The pI values of the 16 oleosins are generally in the alkaline range (Table I), as has been noted in the oleosins from various sources (10). Nevertheless, the current finding of pI 7.4 -7.5 in two oleosins and pI 5.8 in one oleosin indicates that the alkaline pI is not a restricted characteristic of oleosins.
The smallest oleosin in diverse sources known previously has an M r of 14,000 (10). In the current study, the smallest oleosin, the T5 oleosin, has 10,682, of which the hairpin accounts for ϳ7,000. The amino and the carboxylic portions in combination possess only ϳ3 kDa. Specifically, the amino portion (MFEIIQ-) has six residues (or five residues, if the terminal methionine had been removed) and the carboxylic portion possesses 28 residues (Table I and Fig. 2). The carboxylic portion of all oleosins known previously has an amphipathic ␣ helix (10,11), which would reside longitudinally on the oil body, interacting with the surface PLs. In T5 oleosin, the 28 residues of the carboxylic portion form an amphipathic ␣ helix in a helical wheel plot, with 11 residues in the hydrophilic phase and 7 residues in the hydrophobic phase (not shown). However, six proline residues, known to be helical breakers, are scattered along the 28-residue stretch. Thus, the carboxylic portion of T5 oleosin may not form an ␣ helix, and if so, the necessity of an oleosin to have a carboxylic amphipathic ␣ helix to perform its function is uncertain.
A Phylogenetic Tree of the Oleosins Reveals the Relationship among the Three Oleosin Groups-A phylogenetic tree constructed on the basis of the hairpin sequences of the 16 oleosins is shown in Fig. 6. Very similar trees were obtained when the construction was based on the complete oleosin sequences or the complete DNA sequences of the ORF (not shown). The eight T oleosins are separated from the other oleosins. The five S oleosins can be divided into two subgroups: S1, S2, and S4 oleosins are the H isoforms (10), whereas S3 and S5 oleosins are the L isoforms. SM3 oleosin is intermediate between the T and S oleosins, whereas SM1 and SM2 oleosins are more closely related between themselves than to the S oleosins. These relationships between the SM oleosins and the T and S oleosins reflect the duel expressions of the three SM genes in the florets and the seed, as well as the higher expression in the florets than in the seed for SM3, and vice versa for SM1 and SM2.
Seed Oil Bodies and Floret Tapetosomes Have Been Isolated-The oleosin-and TAG-containing oil bodies in seed and tapetosomes in florets of Arabidopsis have not been previously isolated for biochemical studies, partly because of the small sizes of the organs. We isolated the Arabidopsis seed oil bodies using the procedure established for the Brassica organelles (30). The total extract of mature seed contained abundant storage proteins of Ͻ10 kDa (the 2S storage proteins) and 25-40 kDa, as resolved by SDS-PAGE (Fig. 7). The oleosins of ϳ20 kDa (presumably S1, S2, and S3 oleosins) were resolved as a protein band by PAGE. Isolated oil bodies contained these 20-kDa oleosins as the predominant proteins and the storage proteins as minor contaminants. These oleosins were recognized by antibodies raised against the Brassica oleosins (30); the antibodies also recognized two minor proteins of ϳ21 kDa (presumably S4 oleosin) and 15 kDa (presumably S5 oleosin).
We isolated the Arabidopsis floret tapetosomes using the procedure established for the Brassica organelles (13). The total extract of the florets contained numerous proteins as resolved by SDS-PAGE, and a few of these proteins were enriched in the tapetosome fraction (Fig. 7). Specifically, the tapetosomes contained a major protein of 60 kDa. This protein is likely to be T3 oleosin, which is the only T oleosin possessing a deduced M r higher than 24,000 (Table I). Antibodies were raised against a synthetic peptide, KRRMCVKPKDNP, which represents a segment of the T3 oleosin adjacent to the hairpin stretch at the carboxylic side (Fig. 2); similar segments with different degrees of residue variations are present in the other T oleosins. The antibodies recognized the putative T3 oleosin FIG. 7. SDS-PAGE and immunoblotting of proteins of samples from seeds, florets, and pollen. The total seed and isolated oil body, the total floret and isolated tapetosome, and the pollen interior (after the surface materials had been removed) samples were used. In the top panels, the gels were stained with Coomassie Blue (seed and floret samples) or silver stain (pollen sample). In the bottom panels, the gels were treated for immunoblotting using antibodies against Brassica 20-kDa seed oleosins (for the seed samples), a synthetic peptide representing a segment of the Arabidopsis T3 oleosin (for the floret samples), or a bacteria-synthesized N-terminal segment of SM3 oleosin (for the pollen sample). The star indicates the major 20-kDa oleosins in the seed oil bodies; the triangle denotes the 15-kDa oleosin (T6) in the tapetosomes; and the arrow specifies the SM3 location. A band of SM3 cannot be observed because it was not separated from other proteins of similar M r . Positions of M r markers are shown on the right. The 97/1/2 (w/w/w) ratio of TAGs/PLs/oleosin was similar to that in native seed oil bodies. The suspension was placed in a 1.8-ml cuvette in a spectrophotometer, and the relative absorbance (A/A 0 ) at 600 nm of the approximately lower 0.5 ml was recorded at time intervals. The droplets coalesced and floated to the top of the cuvette, and thus the absorbance of the mixture at the lower 0.5 ml in the cuvette decreased. Oil bodies isolated from mature seed were used as a reference. and a 15-kDa oleosin, which was determined to be T6 oleosin by the finding of its N-terminal sequence being APFPLSSIIGGK.
SM3 Has Been Detected in the Pollen Interior-We attempted unsuccessfully to isolate the oil bodies from mature pollen because of the scarcity of the pollen and the difficulty of breaking the hard pollen shell without damaging the internal organelles. We then tried to detect the SM oleosins in the pollen extract. We focused our attention on SM3 because the SM3 transcript was most abundant among the three SM transcripts (Fig. 3). The surface proteins and lipids of the mature pollen were first removed by diethyl ether (13). The proteins in the leftover pollen (the pollen interior) were subjected to SDS-PAGE (Fig. 7). Numerous proteins of diverse molecular weights were observed. Because SM3 has a deduced molecular weight of 19,213, we microsequenced the separated proteins that had molecular weights of 18,000 -20,000 directly and after trypsin hydrolysis. One protein of ϳ20-kDa had an internal sequence of XH(A/G)RT(G)LNP. This sequence matches an N-terminal segment of the SM3 sequence (Fig. 2) and it, as well as its possible derivatives from the uncertain residue readings, does not match any other deduced protein of Arabidopsis. This 20-kDa protein on the SDS-PAGE gel was also recognized, via immunoblotting, by the antibodies raised against the N-terminal region of SM3 (Fig. 7). Thus, the SM3 gene was expressed inside the pollen, leading to the production of SM3 oleosin.
The Pollen Oleosin Can Function, in Collaboration with Phospholipids, to Stabilize TAG Droplets-The possible function of the SM oleosins inside the pollen in stabilizing the storage oil bodies was tested by an in vitro assay. TAG droplets in a suspension coalesced and floated to the top of the mixture. The rate of this floatation would decrease if the TAG droplets were stabilized by surface PLs and oleosins (38). The floatation rate was measured as follows. A 1.2-ml mixture containing TAGs alone or with phosphatidylcholine and/or SM3 oleosin was sonicated and then placed in a spectrophotometric cuvette, and the decrease in absorbance at 600 nm at the lower ϳ0.5-ml in the cuvette was followed at time intervals. As shown in Fig.  8, the droplets containing TAGs alone coalesced and floated to the top of the mixture, and about 90% of the absorbance value at the lower portion of the mixture disappeared after 7 h of incubation. Phosphatidylcholine or SM3 oleosin alone did not substantially reduce the floatation rate of the TAG droplets. In combination, they stabilized the droplets and reduced the floatation rate to almost that of the isolated seed oil bodies. This collaborative and synergetic effect of oleosin and PLs reflects that an oleosin molecule can assume its proper secondary structures only in the presence of PLs, such that it can insert onto, and stabilize the oil body (11). Phosphatidylserine could replace phosphatidylcholine in its collaboration with SM3 to stabilize the droplets (Fig. 8). The results indicate that pollen oleosins can function in the stabilization of storage TAG droplets inside the pollen. CONCLUSION This is the first comprehensive account of all the oleosin genes and oleosins in a species. Many new insights have been obtained. The 16 oleosin genes in Arabidopsis may contain no intron or one intron preceding or immediately following the sequence encoding the conserved hairpin stretch. There is no pseudogene. Eight of the genes are expressed in the floret tapetum; they are all on chromosome V, with 7 in tandem within a 26-kb segment. Five of the genes are expressed in the maturing seed. Three of the genes form a novel group, and are expressed in both maturing seed and microspores. Contrary to earlier suggestions, pollen do contain oleosins that can stabilize the abundant storage oil bodies. Analyses of the deduced amino acid sequences of the 16 oleosins have revealed new constraints and variability of the proteins. Although all the oleosins possess the conserved hairpin stretch of 72 hydrophobic residues, their amino portions and the carboxylic portions are highly variable in length and residues. The predominant tapetosome oleosin also has a complete 72-residue hairpin stretch; this indicates that possessing only 3 ⁄4 of the amino side of the hairpin in the counterpart Brassica oleosin is not a characteristic of tapetosome oleosins. The smallest oleosin, of 10.7 kDa, has only 6 residues in the amino portion and 28 residues in the carboxylic portion, which may not form an amphipathic ␣ helix. Oleosins from diverse species are alkaline protein, but 3 oleosins in Arabidopsis have neutral or acidic pI values. Seed oil bodies and anther tapetosomes containing TAGs and oleosins from the miniature Arabidopsis can be isolated and characterized, and thus the organelles can be subjected to future biochemical and genetic studies.