Origin of β-carotene-rich plastoglobuli in Dunaliella bardawil.

The halotolerant microalgae Dunaliella bardawil accumulates under nitrogen deprivation two types of lipid droplets: plastoglobuli rich in β-carotene (βC-plastoglobuli) and cytoplasmatic lipid droplets (CLDs). We describe the isolation, composition, and origin of these lipid droplets. Plastoglobuli contain β-carotene, phytoene, and galactolipids missing in CLDs. The two preparations contain different lipid-associated proteins: major lipid droplet protein in CLD and the Prorich carotene globule protein in βC-plastoglobuli. The compositions of triglyceride (TAG) molecular species, total fatty acids, and sn-1+3 and sn-2 positions in the two lipid pools are similar, except for a small increase in palmitic acid in plastoglobuli, suggesting a common origin. The formation of CLD TAG precedes that of βC-plastoglobuli, reaching a maximum after 48 h of nitrogen deprivation and then decreasing. Palmitic acid incorporation kinetics indicated that, at early stages of nitrogen deprivation, CLD TAG is synthesized mostly from newly formed fatty acids, whereas in βC-plastoglobuli, a large part of TAG is produced from fatty acids of preformed membrane lipids. Electron microscopic analyses revealed that CLDs adhere to chloroplast envelope membranes concomitant with appearance of small βC-plastoglobuli within the chloroplast. Based on these results, we propose that CLDs in D. bardawil are produced in the endoplasmatic reticulum, whereas βC-plastoglobuli are made, in part, from hydrolysis of chloroplast membrane lipids and in part, by a continual transfer of TAG or fatty acids derived from CLD.


INTRODUCTION
Eukaryotic cells accumulate neutral lipids in different tissues mainly in the form of lipid droplets (Murphy 2012). Most lipid droplets consist of a core of triglycerides (TAG) and/or sterol esters coated by a phospholipids monolayer and embedded with proteins (Zweytick et al. 2000). Plants accumulate triglycerides (TAG) in different tissues, primarily in seeds, but also in fruit, such as in palm oil, in flowers and in leaves. The best characterized system for TAG metabolism are oil seeds, in which TAG serves as the major carbon and energy reservoir to be utilized during germination (Huang 1992, Huang 1996. Recent studies show that lipid droplets are not just static pools of lipids, but have diverse metabolic functions (Farese Jr and Walther 2009). In addition, plants also contain plastoglobuli, small chloroplastic lipid droplets consisting primarily of storage lipids and pigments. Proteome analyses of plastoglobuli suggest that they are involved in synthesis and degradation of lipids, pigments and coenzymes (Ytterberg et al. 2006, Lundquist et al. 2012. It has been shown that plant plastoglobuli are associated with thylakoid membranes (Austin et al. 2006, Ytterberg et al. 2006).
It is not entirely clear where are TAG synthesized in the plant cell. Until recently it has been assumed that most TAG is made in the ER from fatty acids which are mostly synthesized in the chloroplast and imported to the cytoplasm (Joyard et al. 2010).
However, the recent identification of the enzyme diacylglycerol acyl transferase (DGAT) in plant plastoglobuli (Lundquist et al. 2012), suggests that TAG may be synthetized directly in chloroplasts, although direct evidence for this is missing. TAG may be synthesized also from galactolipid fatty acids during stress or senescence by phytyl ester synthases, which catalyze acyl trans-esterification from galactolipids to TAG (Lippold et al. 2012). Phosphatidyl choline (PC) plays a major role in acyl transfer of newly synthesized fatty acids from the chloroplast into TAG at the ER in plants ( Bates et al. 2009). An indication for the origin of glycerolipids in plants is the identity of the fatty acids at the sn-2 position: if it originates in the chloroplast, it is mostly C16:0 whereas if it was made in the ER it is mostly C:18 (Heinz and Roughan 1983).
Many species of unicellular microalgae can accumulate large amounts of TAG under growth-limiting conditions such as nitrogen deprivation (Shifrin and Chisholm 1981, Roessler 1990, Avron and Ben-Amotz 1992, Thompson 1996. In green microalgae, TAG are usually synthesized and accumulated in CLD (Murphy 2012), although in some cases, such as in C. reinhardtii starch-less mutants, they accumulate also in chloroplasts (Fan et al. 2011, Goodson et al. 2011. Recent studies indicate that the CLD are closely associated with ER membranes and possibly also with chloroplast envelope membranes (Goodson et al. 2011, Peled et al. 2012. Green microalgae also contain two distinct types of chloroplastic lipid droplets. The first are plastoglobuli, similar in morphology to higher plants plastoglobuli (Bréhélin et al. 2007, Kessler andVidi 2007). The second is the eyespot (stigma), part of the visual system in microalgae. The eyespot is composed of a cluster of β -carotene containing lipid droplets, organized in several layers, between grana membranes in the chloroplast (Häder andLebert 2009, Kreimer 2009). Recent proteomic analysis of algal eyespot proteins revealed that they contain diverse structural proteins, lipid and carotenoid metabolizing enzymes, transporters and signal transduction components (Schmidt et al. 2006).
The origin of TAG in microalgae is still not clear. In C. reinhardtii it was found that the Plants and algae lipid droplets contain structural major proteins, localized at the lipid droplet periphery, whose major function appears to be stabilization and prevention of fusion (Huang 1992, Katz et al. 1995, Huang 1996, Frandsen et al. 2001, Liu et al. 2009).
In plant seed oils the major classes of lipid droplet proteins are oleosins and caleosins which have a characteristic hydrophobic loop with a conserved three proline domain (Hsieh and Huang 2004, Capuano et al. 2007, Purkrtova et al. 2008, Tzen 2012. Oleosin and caleosin analogous were recently identified also in some green microalgal species (Lin et al. 2012, Vieler et al. 2012, Huang et al. 2013). However, the most abundant lipid droplets proteins in Volvocales order green algae are a new family of major lipid droplet proteins (MLDP), structurally distinct from plant oleosins and caleosins (Moellering and Benning 2010, Peled et al. 2011, Davidi et al. 2012. Plastoglobules have different major lipid-associated proteins termed PAP-fibrillins, which form a distinct protein family with no sequence or structural similarities to oleosins (Kim and Huang 2003). We have previously identified in the β C-plastoglobuli a lipid-associated protein, termed carotene globule protein (CGP), whose degradation destabilized the lipid droplets (Katz et al. 1995). The proteome of Chlamydomonas lipid droplet indicates that algal CLD also contain several enzymes suggesting that they are involved in lipid metabolism (Nguyen et al. 2011).
The halotolerant green algae Dunaliella bardawil and D. salina Teodoresco, are unique in that they accumulate under high light stress or nitrogen deprivation large amounts of plastidic lipid droplets (βC-plastoglobuli), which consist of TAG and of two isomers of β -carotene, all-trans and 9-cis (Ben-Amotz et al. 1982, Ben-Amotz et al. 1988. D. bardawil also accumulates under the same stress conditions CLD, similar to other green algae (Davidi et al. 2012). It has been demonstrated that the function of β C-plastoglobuli is to protect the photosynthetic system against photoinhibition (Ben-Amotz et al. 1989).
The enzymatic pathway for β -carotene synthesis in D. bardawil and D. salina has been partly identified, but the sub-cellular localization of β -carotene biosynthesis is not known (Jin and Polle 2009). The synthesis of β -carotene depends on TAG biosynthesis (Rabbani et al. 1998), however the origin of β C-plastoglobuli is not known. Are they formed within the chloroplast or are they made in the cytoplasm? Is the TAG in β C-plastoglobuli and in CLD identical or different and where is it formed?
D. bardawil is an excellent model organism for isolation of lipid droplet for several reasons: First, D. bardawil contains large amount of both CLD and β C-plastoglobuli (Ben-Amotz et al. 1982, Fried et al. 1982, making it possible to obtain sufficient amounts of proteins and lipids from the two types of lipid pools for detailed analyses. Second, Dunaliella does not have a rigid cell wall and can be lysed by a gentle osmotic shock, which does not rupture the chloroplast. Therefore it is possible to sequentially release pure CLD and β C-plastoglobuli by a two-step lysis (Katz et al. 1995). Third, D.
bardawil seems to lack the eyespot structure, which can be clearly observed in other Dunaliella species even in a light microscope or by electron microscopy, but has never been observed in D. bardawil by us. This avoids the risk of cross-contamination of β Cplastoglobuli with eyespot proteins. Finally, the availability of protein markers for the major lipid droplet associated proteins, CGP and MLDP, enabled both good immunolocalization and careful monitoring of the purity of the preparations by Western analysis.
In this work we describe the purification, lipid compositions and protein profile of two lipid pools from D. bardawil: CLD and plastidic β C-plastoglobuli. A detailed proteomic analysis of these lipid droplets will be described in another manuscript. Combined with detailed EM studies, these results led to surprising conclusions regarding the origin of the plastidic β C-plastoglobuli.

Isolation and lipid composition of CLD and plastidic
β C-plastoglobuli from D.

bardawil
CLD and β C-plastoglobuli were isolated from D. bardawil essentially as described earlier (Katz et al. 1995), with one modification. In brief, cells incubated for 48 h in N-depleted medium at high light (DB-N), were washed and lyzed by a gentle osmotic shock. We noticed that the osmotic shock does not release most of the CLD from the lyzed cells, and therefore introduced a syringe treatment after osmotic cell disruption, which released the majority of the CLD as shown by the protein SDS-PAGE profile and by the level of TAG in the preparation (Fig. S1). Chloroplasts were washed and lyzed by sonication. CLD and β C-plastoglobuli were isolated by flotation centrifugation through sucrose density layers.
Control cells grown in N-sufficient medium (DB+N) were also analyzed.
The two populations of purified lipid droplets, CLD and β C-plastoglobuli, were analyzed by Nile red and TLC. In the Nile red fluorescence staining (Fig. 1A), both CLD and β Cplastoglobuli show the characteristic staining of neutral lipids at 580-590 nm, which is apparent also in intact N-deprived cells. No chlorophyll fluorescence emission was observed at 685 nm, indicating that the two lipid pools are practically clean of contamination by chloroplast membranes. TLC analysis of neutral lipid composition ( Fig.   1B) reveals that CLD contain mostly TAG, whereas β C-plastoglobuli also contain high amount of β -carotene and another minor carotenoid. Both preparations also contain small amounts of polar lipids and no detectable chlorophyll. In order to further analyze the trace polar lipids, larger amounts of total lipid extracts were resolved on TLC, and the samples were compared to thylakoid lipid extracts (Fig. 1C). The polar lipid compositions differ between the two preparations: whereas β C-plastoglobuli polar lipids are almost identical to thylakoid in lipid composition, predominated by monogalactosyldiacylglycerol (MGDG) and digalactosyldiacylglycerol (DGDG), as confirmed by galactololipid staining (Fig. 1C, right), the CLD contain two different polar lipid components which are not galactolipids.
In order to compare the TAG molecular species of the two droplet populations, we analyzed lipid extracts by reverse-phase HPLC using a Halo-C8 column. The chromatogram of CLD and of β C-plastoglobuli lipid extracts is shown in Fig. 2. The two chromatograms are very similar, exept for three peaks which appear only in the latter.
Two major peaks (at 5 min and 8 min), were identified as β -carotene and phytoene (a β carotene precurssor), respectively, based on their absorption spectra, as determined by a diode-array detector. The third peak (MS-9, at 33.3 min, Fig. 2), is a unique TAG species that does not appear in other Dunaliella species. Of the major TAG peaks (between 9 to 28 min.), each composed of 2-3 individual components, the minor smaller peaks (between 9-17 min), show slight differences in relative peaks distributions, whereas the 3 major peaks (between 18-28 min) are almost identical in both fractions. TAG peaks were identified by reference to TAG standards.
Fatty acid composition analysis of D. bardawil cells and the two types of lipid droplets was conducted by GC analysis of the fatty acid methyl esters (FAMEs) (Fig. 3A). The two population of lipid droplets were highly enriched in the fatty acid 18:1ω9 in comparison to whole cell extracts, reaching about 40% of the fatty acid content of isolated lipid droplets. In contrast, both populations were relatively deprived of fatty acids 18:3ω3 and 16:4ω3, which are components of the chloroplast thylakoid membrane (Evans et al. 1982). These results are in agreement with our previous study in D. salina (Davidi et al. 2012) and studies in C. reinhardtii which showed elevated levels of 18:1ω9, under N-deprivation (Wang et al. 2009, James et al. 2011, Siaut et al. 2011).
As noted above, the identity of the fatty acids at the sn-2 position in higher plants is considered as an indicator for its origin: 18C when made in the ER or 16C when made in the chloroplast. In order to test the applicability of this criterion to Dunaliella lipids, we first analyzed the fatty acid compositions at the sn-2 and sn-1 positions of cytoplasmatic membrane polar lipids (microsomal fraction) and of thylakoid membrane galactolipids, as controls for ER-made and for chloroplast-made lipids. Since we found that microsomal lipid extracts contain significant amounts of galactolipids, indicating chloroplast membrane contamination, we performed the positional analysis on isolated PC, which is a component in microsomal membranes, but is excluded from chloroplast membranes (Table S1). The analysis was made by exposure of the former to phospholipase A2, which specifically cleaves fatty acids at the sn-2 position, and of the latter to Rhizopus lipase, which specifically cleaves fatty acids at sn-1+3 positions. As expected, galactolipids sn-2 fatty acids are mostly 16C in length predominated by 16:4, consistent with a chloroplast origin (see also (Cho andThompson 1987a, Cho andThompson 1987b)). Conversely, the microsomal sn-2 fatty acids of PC are slightly enriched in 18C fatty acids, consistent with previous a report in another Dunaliella species (Ha and Thompson 1991). Conversely, the chloroplast membranes sn-1 fatty acids are composed of a mixture of 16C and 18C fatty acids, whereas PC sn-1 shows a clear 16C bias, consistent with previous reports (Ha and Thompson 1991). Next, we analyzed the sn-1+3

Time-course of formation of cytoplasmic and chloroplastic lipid droplets
In order to learn if the formation of CLD and of chloroplastic β C-plastoglobuli occour simultaneously or if one preceeds the other, we performed a physical seperation and quantitation of the TAG contents of the two lipid populations at different times after exposure to N-deprivation.
As shown in Fig. 4A, the rate of TAG accumulation in the two lipid populations is very different: whereas CLD increase in level until 48 h and than decrease, chloroplastic plastoglobuli increase mostly after 3-6 days. These results suggest that the formation of CLD precedes that of chloroplastic β C-plastoglobuli.
Pulse-labeling experiments with 14 C-palmitic acid ( 14 C-PA) were designed to answer 3 questions: Are TAG in the two lipid pools synthesized de novo, from newly-incorporated fatty acids? Are TAG produced from degradation of membrane polar lipids in the chloroplast or in microsomal cytoplasmic membranes? Are TAG in β C-plastoglobuli made from pre-formed TAG in CLD? 14 C-PA was chosen because we found that this fatty acid, which is a major component in all polar and neutral glycerolipid fractions in D.
bardawil, is rapidly incorporaed into both cytoplasmic and chloroplastic membane lipids, as shown below. To answer the 1 st and 3 rd questions, N-deprived cells were labeled for 4h with 14 C-PA, than all free 14 C-PA was removed by washing and supplemetation of 250-fold excess unlabelled PA. After different periods of incubation in N-deprived medium, cells were disrupted and 14 C contents in CLD, βCplastoglobuli, cytoplasmatic microsomal membranes and in chloroplast membranes were analyzed. If β C-plastoglobuli TAG are produced from CLD TAG, then we expected to find an increased labeling in β C-plastoglobuli correlated with a decreased labeling in CLD.
The 14 C distribution pattern (Fig. 4B, Table S1) showed fast incorporation into both microsomal and chloroplast membranes, followed by decrease with time, indicating that there is no permeability barrier for PA into the chloroplast. The decrease with time in 14 C content in the two membrane fractions differ: whereas in microsomal membranes the decrease continues for 72h, reaching about 15% from the point of PA dilution, in the chloroplast membranes 14 C decrease levels off after 24h, and it is roughly correlated with the decrease in chlorophyll (Fig. 4C), suggesting that in the microsomal membranes, the decrease is mostly due to degradation and to resynthesis of new lipids (turnover), whereas in chloroplast membranes the decrease results from net degradation of about 60% of the membranes and no de novo synthesis.
The time-courses of 14 C incorporation into the two polar lipid droplets greatly differ: whereas in CLD there is a 7-fold increase in labeling in the first 24h, followed by a subsequent small decrease, in β C-plastoglobuli there is a small but progressive increase in labeling during 72h (Fig. 4B). The large increase in incorporation into CLD in the first 24h, is correlated with the major decrease in labeling of both chloroplast and microsomal membrane lipids, which could indicate acyl editing, namely, that PA is first incorporated into specific polar membrane lipids and next transacylated into TAG (Bates et al. 2009, Li et al. 2012). The subsequent decrease in labeling in CLD between 40h and 88h of N 1 0 deprivation is roughly correlated with the increase in labeling in β C-plastoglobuli (53,000cpm and 41,000cpm, respectively).
In order to test if β C-plastoglobuli TAG may be synthsized from degradation of preformed chloroplast (or other membrane) polar lipids, we designed another 14 C-PA pulselabeling experiment, in which D. bardawil cells were incubated with 14 C-PA in Nsufficient (complete) growth medium before entering N-deprivation, and then residual 14 C-PA was removed by washing and dilution with access unlabeled PA. Samples of cells taken at 0h or 48h after entering N-deprivation, were disrupred and fractionated into membrane and lipid droplets and their 14 C contents were analyzed.
As shown in Fig. 5 and in Table 2, the 14 C distribution in CLD and in β C-plastoglobuli is very different from the de novo synthesis experiment (Fig. 4B). Over a third of the total 14 C in all lipid fractions was recovered in the β C-plastoglobuli fraction, compared to 5% in the de novo synthesis after 40h, whereas the CLD fraction contained about 20% of the total 14 C, compared to over 80% in the de novo synthesis experiment. The number of counts recoverd from the β C-plastoglobuli fraction was similar to the decrease in counts in the chloroplast membrane fraction (285,000cpm compared to 260,000cpm, respectively), constituting about 50% of the total 16C fatty acids in polar chloroplalast membrane lipids. Also the recovery of 14 C labeled TAG in CLD was close to 40% of the microsomal 14 C contents at the onset of N deprivation, suggesting significant utilization of 16C fatty acids for production of TAG from degradation of polar membrane lipids in both fractions, particularly of β C-plastoglobuli. The larger decrease in microsomal 14 C contents during 48h of N-deprivation, is consistent with the faster lipid turnover in microsomal membranes.

Protein composition
Protein analysis of the purified droplets was conducted by SDS-PAGE ( of at least two components, which was previously characterized in our group (Katz et al. 1995) and termed carotene globule protein (CGP).
Western blot analysis of the purified CLD and β C-plastoglobuli with anti-MLDP and anti-CGP antibodies was conducted (Fig. 6B). For this purpose anti-MLDP and anti-CGP specific rabbits polyclonal antibodies were utilized (Katz et al. 1995, Davidi et al. 2012).
The results demonstrate the specificity of anti-MLDP antibodies to the CLD and of anti-CGP antibodies to the chloroplast β C-plastoglobuli. No cross-reaction with anti-Rubisco antibodies, a common chloroplast contaminant, could be detected in the purified lipid droplets proteins (Fig. 6B, upper lane) strengthening the conclusion that the two lipid pools are pure.

CGP structure
CGP has been isolated and cloned in our lab (Shoham 1995) and found to be a nuclear analysis (Guermeur et al. 1999) (Fig. S3B) shows that MLDP is highly structured protein, consisting mostly of α -helices, whereas CGP is mostly unstructured protein, probably due to its high content of proline residues.
Using software for prediction of post-translational modifications revealed no putative conserved sites for palmitoylation, prenylation, myristoylation or GPI-anchor for the D. S5B-H). Notably, many CLD seem to be closely associated with the outer chloroplast surface and in some cases appear almost engulfed by the chloroplast (Figs. S5E,F).
Within the chloroplast appear arrays of smaller droplets (of about 100 µm, marked by red arrows), bordered by the chloroplast envelope membranes and by the outermost thylakoid membranes. In some cases, these small droplets appear tightly sqeezed to the large CLD, separated by a nonuniformly-stained chloroplast envelope membrane (Fig. 12D). A movie of the tomography reconstruction that best represent our findings is avaliable on Movie S1. This morphology was obsereved in dozens of cells and was also obsreved in cells after 2 and 3 days of N-deprivation, although at a lower frequency (Fig. S6). After 7 days of deprivation, the cells appear swollen, the cytoplasm contains only a few lipid droplets whereas the chloroplast contains large amounts of starch granules and large number of β C-plastoglobuli, ranging in size from 200-500 µm, and localized mostly at the outer periphery ( Fig. S7A,B). Some blank spaces and tears to chloroplast membranes in the vicinity of starch granules can be seen, possibly resulting from damage by starch granules during centrifugation.

DISCUSSION
In this work we tried to clarify the interrelations between two neutral lipid droplets in Dunaliella bardawil, CLD and β C-plastoglobuli, the latter being unique to this species.
This was achieved by improved isolation of the two lipid pools combined with lipid composition analyses, biochemical analyses of lipid biosynthesis and electron microscopy.
The isolated β C-plastoglobuli and CLD have similar TAG molecular species and fatty acid compositions (in both total fatty acid, fatty acid at sn-1+3 and sn-2 positions), but differ in the presence or absence of β -carotene, in their polar lipid compositions and in their major lipid-associated proteins, as will be detailed below.
A major question that concerned us in this study is the origin of TAG in CLD and in β Cplastoglobuli in D. bardawil. The finding that TAG sn-2 fatty acids in both CLD and in β C-plastoglobuli are a mixture of 16C and 18C, give no clear indication for a cytoplasmatic origin at the ER, which shows a clear 18C bias both in higher plants (Heinz and Roughan 1983) and in algae including Dunaliella (Ha and Thompson 1991), nor for a chloroplastic origin, which has a clear 16C bias ((Cho and Thompson 1987a), Table S1), although the cytoplasmatic TAG fatty acid composition is much closer to that of PC than to chloroplast membrane lipids (Table 1). A similar sn-2 fatty acid composition was reported previously for both cytoplasmatic and chloroplast associated TAG in D. salina (Ha and Thompson 1991 finding that PA incorporation into polar lipids precedes the incorporation into TAG (Fig.   4B) suggests that acyl editing is probably part of TAG biosynthesis also in D. bardawil.
The origin of TAG in D. bardawil lipid droplets is not clear. The very similar fatty acid composition (Fig. 3) and the similar TAG molecular species (Fig. 2), strongly suggest a common origin. In contrast, according to the 14 C-PA pulse-labeling experimets, the two TAG pools seem to have different origins: In the first 40-48h of N deprivation, most CLD TAG seem to be produced by direct incorporation of newly-synthetized fatty acids (Fig.   4B) and a smaller part from pre-formed polar lipids (Fig. 5), whereas for βCplastoglobuli TAG, significant part of fatty acids, amounting to 50% of 16C fatty acids in chloroplast membranes, seem to be released from degraded chloroplast membrane polar lipids (Figs. 5, Table 2). However, chloroplast membrane lipids degradation cannot provide most TAG fatty acids simply because the amount of total membrane fatty acid contents in D. bardawil is only around 20% of that of the fatty acid TAG content (see for example Fig. 1B, lane 2).
The only difference in fatty acid composition between the two TAG pools is the higher contents of 16:0 FA in β C-plastoglobules, particularly at the sn1+3 position (Fig. 3).
Also the only significant differences in the TAG molecular species compositions in the two lipid pools are in the low molecular weight species, which are enriched in 16C fatty acids (peaks between 9-17min, Fig. 2). The reason for these differences may be the higher contribution of 16C fatty acids derived from degradation of chloroplast membrane lipids and incorporated into TAG in plastoglobules, consistent with the 14 C-palmitic acid pulse-labeling experiments (Fig. 5, Table 2) during the first 48h of N-deprivation.
The time-course of TAG accumulation in the two lipid pools, and the 14 C-PA pulselabeling experimets (Fig. 4A,B) suggest that the major biosynthesis of CLD takes place in the first 48h of N deprivation/high light stress, whereas the accumulation of consistent with the idea of lipid transfer from the cytoplasm into the chloroplast for the formation of β C-plastoglobuli.
A possible mechanism that may lead to a similar TAG composition of the cytoplasmatic and chloroplastic lipid droplets is a dynamic transfer of fatty acids or of TAG molecules from the CLD to the β C-plastoglobules through the chloroplast envelope membranes.
This possibility is consistent with the close proximity of CLD and of β C-plastoglobules to chloroplast envelope membranes observed by electron microscopy (Figs S5-S7, Fig.   12). At present we do not have any biochemical evidence for such a fatty acid or TAG transfer, but we have identified in a proteomic analysis lipases and acyl transferases in the CLD proteome, and of three phytyl ester synthases (PES) homologs in the β Cplastoglobuli proteome, which could be involved in such a mechanism (Davidi et al, unpublished results). Such a dynamic lipid transfer may explain also why at later stages of N deprivation/high light stress, as β -carotene biosynthesis proceeds within the chloroplast, increasing part of TAG is mobilized from the cytoplasm into the chloroplast to generate more plastoglobules for incorporation of the pigment.
In summary, we propose that at the early stages of stress, CLD TAG is produced at the ER, mostly by incorporation of new-synthetized fatty acids at the ER and the process may involve fatty acid shuttling through PC or another polar membrane lipid.
Subsequently, plastoglobuli start to be created within the chloroplast from TAG made in part from fatty acids hydrolyzed from chloroplast membrane lipids and in part from fatty acids or TAG molecules derived from cytoplasmic droplets.
It is interesting to compare the similarities and differences between D. bardawil and between C. reinhardtii: starchless mutants of C. reinhardtii also accumulate lipid droplets both in the cytoplasm and in the chloroplast, but the chloroplastic droplets resemble in size those in the cytoplasm, do not contain β-carotene and they have not been characterized. Interestingly, the cytoplasmatic lipid droplets in these mutants were observed to adhere to the outer chloroplast envelope (Goodson et al. 2011) similar to the observation in D. bardawil. Moreover, as already discussed, the fatty acids at the sn-2 position in C. reinhardtii TAG is primarily C16, suggesting that it is made in the chloroplast by the same enzyme that produces chloroplast membrane lipids whereas in D. Third, Chlamydomonas does not synthesize massive amounts of β -carotene, which has to be deposited in TAG droplets as is the case in D. bardawil under N deprivation. It is possible, therefore, that D. bardawil evolved a unique mechanism for TAG accumulation in the chloroplast to enable the deposition of β-carotene in this lipid droplets.
The finding that the polar lipid compositions in CLD and in β C-plastoglobuli differ (Fig.   1C), suggests that the origins of the polar lipid monolayers of these lipid droplets are different. CLD polar lipids most probably originate in the ER. In contrast, the similarity of β C-plastoglobuli polar lipids to thylakoid/stroma membrane galactolipids (Fig. 1C The β C-plastoglobuli and the CLD are characterized by different major lipid-associated proteins (Figs. 6,9), CGP (Katz et al. 1995) and MLDP (Davidi et al. 2012) in β Cplastoglobuli and in CLD, respectively. The CGP protein, which has been previously characterized at the protein level (Katz et al. 1995), differs in sequence from green algae MLDPs suggesting that it has a different origin.
We have shown in this work that formation of two lipid bodies in Dunaliella bardawil is a complex event that involves both hydrolysis of membrane lipids and de-novo synthesis and probably also of trans-membrane lipid transfer into the chloroplast. Even though the molecular details of this mechanism are still not clear, we believe that the present study and our following proteomic analysis of these lipid droplets, will contribute to the presently poor understanding of the biogenesis of lipid droplet in plants and in microalgae.

Strain and growth condition
Dunaliella bardawil is an isolated species (

Lipid droplets isolation
Isolation of lipid droplets was performed essentially as previously described with some modifications (Jiménez andPick 1994, Katz et al. 1995). In brief, algae after 2 days of Ndeprivation were washed, osmotically-lyzed and centrifuged at low speed for separation of the cytoplasmatic and chloroplastic fractions. Chloroplasts were passed twice through a 25 ml syringe (1.5 inches, gauge 21) and centrifuged for 15 min at 5000 g. This treatment was found to release the majority of CLD. Chloroplasts were pelleted by centrifugation, washed twice and lyzed by sonication. Purification of CLD and of β Cplastoglobuli was performed by floatation on a discontinuous sucrose gradient consisting of three layers (30% sucrose containing the droplet fraction; 15% sucrose and 5% sucrose, all containing 10 mM Tris-HC1, pH 8) and centrifugation at 75,000 g for 2 h.
The crude lipid droplets, recovered from the top fraction, were collected and re-purified by flotation on a second sucrose gradient. The purified lipid droplets were collected from the top and kept frozen in liquid nitrogen.

Cloning the CGP gene
This part of the work was performed during 1993-95 by T. Shoham and I. Gokhman (Shoham 1995 of high light-induced D. bardawil cells, with the aid of anti-CGP polyclonal antibodies. In brief, D. bardawil culture in complete growth medium (Fisher et al. 1994) was exposed for 36 h to high light intensity of 1,600 μE m -2 s -1 in a Warburg temperaturecontrolled shaker. RNA was isolated using Tri Reagent. cDNA to poly(A+) mRNA was synthesized and cloned into the gamma Uni-ZAP XR expression vector (Promega, Madison, WI) as described (Fisher et al. 1996). A phage clone was isolated and a screen with anti-CGP antibodies (Katz et al. 1995) was subsequently shown to include a partial 3' end of the CGP gene. The full-length cDNA was isolated by the 5'-RACE procedure (CLONTECH Laboratories, Palo Alto, CA). To obtain the corresponding genomic DNA sequences, primers based on cDNA sequences were used with templates of genomic DNA digested by Sau3A, HaeII, TaqI or MspI, followed by fragment circularization by ligase, in several consecutive steps of inverted PCR amplification and subsequent cloning.

Generation of anti-MLDP and anti-CGP antibodies
Polyclonal antibodies against CGP and against MLDP were raised in rabbits as described before (Katz et al. 1995, Davidi et al. 2012.

Gel electrophoresis and Western analysis
Protein extracts from total cells were generated from pellets of culture samples containing Cells pre-cultured for 48h in complete growth medium were collected by centrifugation, washed once and cultured in N-deficient medium. After 0, 6,12,24,32,48,72,96,168 h, samples of 10ml containing 1-2x10 7 cells, were taken for RNA isolation. The cells were collected by centrifugation and immediately flash-frozen in liquid nitrogen and stored at -80°C for further use. Total RNA was isolated using tri-reagent procedure according to manufacture protocol (Molecular Research Center, Cincinnati, OH, USA).
Independent RNA isolations were conducted for each growth period. Template cDNA was synthesized using 0.1 µg total RNA in a total volume of 20 µl, using Superscript kit

Bioinformatics analysis
Sequences were routinely searched using BLAST (Altschul et al. 1997). Sequences were aligned using the CLUSTALW multiple sequence alignment program (Thompson et al. 1994  The sn-1+3 and sn-2 distributions of fatty acids in TAG from purified lipid droplets were determined using Rhizopus lipase (Sigma) as previously described (Fischer et al. 1973) with minor modifications. The TAG were extracted and separated on TLC as described above. The band containing the TAG was scrapped off, dissolved in n-hexane and dried.
Tris-HCl buffer (40 mM, pH 7.2) (0.1 mL) containing 50 mM sodium borate (to reduce positional migration of fatty acids) was added to the dried lipid sample and the mixture was sonicated for 10 min. One hundred µL of lipase (100 units The sn-1 and sn-2 of microsomal and chloroplast membrane lipids were determined by exposure to Rhizopus lipase or to phospholipase A2 (Sigma) and analysis of release compared to total fatty acids. Microsomal and chloroplast membranes were isolated as previously described (Jiménez andPick 1994, Bates et al. 2009). Tris-HCl buffer (40 mM, pH 7.2) (900µl) and diethylether (1.5 ml) were added to the samples together with twenty µl of enzyme for 30 min incubation with shaking in room temperature. The reaction ended by evaporation of the diethylether. Lipids were extracted, separated and fatty acids were analyzed as described above. fraction. Both fractions were subjected to lipid extraction and the lipids were analyzed by TLC as described above. Three repeats were made for each sample.

Incorporation of 14 C-palmitic acid into D. bardawil lipids
To estimate the distribution of 14 C in the different fractions two protocols were used: In one, 5 µCi of 14 C-PA (1 µM final concentration) was added to 300ml D. bardawil culture that has been induced for 12 h in high light at N-deficient medium. After 4h, the cells were washed, diluted into fresh N-deprived medium supplemented with 250 µM unlabeled palmitic acid. Samples were collected after 0, 24 h, 48 h or 72h (total 16, 40h, 64h or 88h of N deprivation, respectively). Cytoplasmatic and chloroplastic TAG was separated as described above. In the second, cells were labeled for 5h in complete (N sufficient) medium before transfer to N deprivation. The incubation with 14 C-PA was terminated by washing and dilution with 250-fold unlabeled PA and the cell were transferred for 48h to N deprivation at high light. Cell samples were taken before and 48h after N deprivation for lysis and fractionation into cytoplasmic membranes, chloroplast membranes, cytoplasmic droplets and β C-plastoglobuli as described above. 14 C in each fraction was counted. The extracted lipids from each fraction were also separated on TLC (as described above) and the TAG bands were scraped off and dissolved in 1 ml n-hexane and 10 ml scintillation solution. The 14 C radiolabelling was measured using a Tri-carb liquid scintillation counter (PerkinElmer, U.S.A.).

Lipid analysis by High-Performance Liquid Chromatography (HPLC)
The lipid profile of D. bardawil purified lipid droplets was determined by Reverse phase β -carotene and phytoene were detected using same system with Photodiode-array detector (Waters Corp., Milford, MA, USA) and were identified by their absorbance spectra at 488 nm and 286 nm, respectively.
Electron Microscopic Techniques

Gold immunolabeling
Samples were prepared according to the Tokuyasu method (Tokuyasu 1973). In brief: Algae cultured for 0,1 or 6 days in N-deprived medium, were fixed in 2% glutaraldehyde The Netherlands). SEM (Zeiss, Germany) using a VCT 100 and were observed using a secondary electrons in-lens detector at an acceleration voltage of 2 kV at a temperature of -120°C .

TEM observation and STEM tomography
Algae cultured in complete or in N-deprived media for 1, 2, 3 or 7 days were fixed in 2%    -plastoglobuli (black) and was compared to TAG standards. β C: β -Carotene, MS-9: TAG peak which was collected and send to MS and FAME analysis.    (C) Phylogenic tree of CGP with SOUL-domain proteins from algae and Arabidopsis. The alignment was generated by the CLUSTAL W program and the phylogram was constructed by the neighbor-joining method using MEGA5 software (Tamura et al. 2011 Algae were cultured in complete (DB+N) or N-deficient media (DB-N) for 6 days. Cryosections of fixed cells were treated with anti-MLDP or with anti-CGP polyclonal rabbit antibodies followed by incubation with 10 nm gold-conjugated goat anti-rabbit antibodies.   The values shown represent means and SD of three replicates.    Table S1 and in Figs 3B, 3C. Table 2. Comparison of 14 C distribution between lipid fractions labeled during N deprivation or from preformed polar lipids. A comparison of 14 C distribution in lipid fractions in cells labeled during N-deprivation and extracted 24h after termination of labeling (de-novo synthesis, total time at N deprivation 40h, see Fig. 4B), or before N deprivation (synthesis from preformed polar lipids, Fig. 5). The values in the table represent the 14 C distribution in each lipid fraction expressed as % of total. Figure S1. Improvement of CLD purification. CLD were isolated with or without syringe treatment following the osmotic shock. (A) SDS-PAGE of CLD protein before (1) and after (2) syringe treatment, together with protein marker (3). (B) TLC of lipid extracted from CLD before (1) and after (2) syringe treatment, together with triolein standard (3). Figure S2. Multiple sequence alignment of CGP and SOUL heme-binding protein orthologs. The alignment was constructed by CLUSTALW algorithm. CGP and orthologs (followed by NCBI accession numbers in parentheses): Chlamydomonas reinhardtii (XP_006691398.1) Volvox carteri (XP_002947474.1), Chlorella variabilis (EFN56543.1), Micromonas sp. RCC299 (XP_002502391.1). Figure S3. Hydropathy plot and secondary structure prediction of MLDP and CGP.

SUPPLAMENTARY MATERIAL
(A) Hydropathy plot of MLDP and CGP were generated employing the Kyte-Doolittle algorithm (Kyte and Doolittle 1982), using the Prot Scale program at http://www.expasy.ch/tools/protscale.html . The G value in each graph is the GRAVY for each protein calculated using the GRAVY calculator program at http://www.gravy-calculator.de/ . (B) Secondary structure prediction of MLDP and CGP was obtained by consensus secondary prediction program http://npsa-pbil.ibcp.fr/cgibin/npsa_automat.pl?page=/NPSA/npsa_server.html. Blue lines represent α -helices, red lines represent extended strand, and purple lines represent unstructured domains.    Movie S1. Tomographic reconstruction movie of D. bardawil cell after 1 day of Ndeprivation. The movie shows sequence tomographic slices throughout a tomographic reconstructed volume. It can be seen that the two CLD are surrounded by a distorted chloroplast envelope membrane and that several small β C-plastoglobuli are formed and accumulate in the adjacent chloroplast. Each slice in the movie is an average of 20 tomographic slices, 2 nm in thickness each. The full frame width is 4 µm. Table S1. Fatty acid sn-2 and sn-1 positional analysis of membrane lipids. Chloroplastic and cytoplasmatic membranes were isolated, lipids were extracted, PC from cytoplasmatic membranes was isolated from two-dimensional TLC plates, and then lipids were treated with Rhizopus lipase or with phodpholipase A2. Released and total fatty acid compositions were determined and expressed as % of total. Data represent averages of 3 repeats.