Báo cáo khoa học: Identification of a glycosphingolipid transfer protein GLTP1 in Arabidopsis thaliana ppt

Structural models showed that the residues that are most critical for glycosphingolipid binding in human GLTP are conserved in AtGLTP1, but some of the sugar-binding residues are unique,

protein GLTP1 in Arabidopsis thaliana

Gun West1,*, Lenita Viitanen1,*, Christina Alm2, Peter Mattjus1, Tiina A Salminen1

and Johan Edqvist3

1 Department of Biochemistry and Pharmacy, A ˚ bo Akademi University, Turku, Finland

2 Department of Plant Biology and Forest Genetics, Swedish University of Agricultural Sciences, Uppsala, Sweden

3 IFM-Biology, Linko¨ping University, Sweden

Glycosphingolipids (GSLs) carry one or more sugar

units on a ceramide backbone [1] These lipids are

major constituents of eukaryotic plasma membranes,

and function in several cellular processes such as cell

death [2,3], adhesion [4] and cell–cell recognition [5,6]

In plants, two types of GSLs are found: the neutral

cerebroside glucosylceramide (GlcCer), which has a

glucosyl residue at the primary hydroxyl group of sphinganine, and the so-called phytoglycosphingolipids

or inositol phosphorylceramide, with a ceramide-1-phosphate base, to which glycosylated inositol residues are bound via a phosphodiester bond [7–10] Typical mammalian sphingolipids, e.g galactosylcera-mide (GalCer), lactosylceragalactosylcera-mide (LacCer), neuraminic

Keywords

ceramide; GLTP; glycolipids; lipid transfer;

sphingolipids

Correspondence

J Edqvist, IFM-Biology, Linko¨ping

University, SE-581 83 Linko¨ping, Sweden

Fax: +46 13 281399

Tel: +46 13 281288

E-mail: Johed@ifm.liu.se

*These authors contributed equally to this

study

(Received 24 January 2008, revised 12

March 2008, accepted 30 April 2008)

doi:10.1111/j.1742-4658.2008.06498.x

Arabidopsis thaliana At2g33470 encodes a glycolipid transfer protein (GLTP) that enhances the intervesicular trafficking of glycosphingolipids

in vitro GLTPs have previously been identified in animals and fungi but not in plants Thus, At2g33470 is the first identified plant GLTP and we have designated it AtGTLP1 AtGLTP1 transferred BODIPY-glucosyl-ceramide at a rate of 0.7 pmolÆs)1, but BODIPY-galactosylceramide and BODIPY-lactosylceramide were transferred slowly, with rates below 0.1 pmolÆs)1 AtGLTP1 did not transfer BODIPY-sphingomyelin, monoga-lactosyldiacylglycerol or digamonoga-lactosyldiacylglycerol The human GLTP transfers BODIPY-glucosylceramide, BODIPY-galactosylceramide and BO-DIPY-lactosylceramide with rates greater than 0.8 pmolÆs)1 Structural models showed that the residues that are most critical for glycosphingolipid binding in human GLTP are conserved in AtGLTP1, but some of the sugar-binding residues are unique, and this provides an explanation for the distinctly different transfer preferences of AtGLTP1 and human GLTP The AtGLTP1 variant Arg59Lys⁄ Asn95Leu showed low BODIPY-gluco-sylceramide transfer activity, indicating that Arg59 and⁄ or Asn95 are important for the specific binding of glucosylceramide to AtGLTP1 We also show that, in A thaliana, AtGLTP1 together with At1g21360 and At3g21260 constitute a small gene family orthologous to the mammalian GLTPs However, At1g21360 and At3g21260 did not transfer any of the tested lipids in vitro

Abbreviations

DGDG, digalactosyldiacylglycerol; GalCer, galactosylceramide; GlcCer, glucosylceramide; GLTP, glycolipid transfer protein; GSL,

glycosphingolipid; GST, glutathione S-transferase; LacCer, lactosylceramide; MGDG, monogalactosyldiacylglycerol; POPC, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine; SM, sphingomyelin.

(sialic) acid-containing ceramides (gangliosides) and

sphingomyelin (SM), have not been found in higher

plants, such as Arabidopsis thaliana [7–9]

A remarkable property of the GSLs is that they

have a high melting temperature due to the high

sat-uration of the hydrocarbon chains, and, furthermore,

the region between the polar head group and the

hydrophobic backbone contains chemical groups that

can function both as hydrogen bond donors and

hydrogen bond acceptors [11] These properties allow

GSLs to self-associate and bring local order to

otherwise disordered and fluid membranes [12] The

ordered membrane microdomains or lipid rafts,

which are enriched in GSLs and sterols, are believed

to play important roles in protein sorting, signal

transduction and infection by pathogens, as the rafts

appear to mediate a lipid-based sorting mechanism

that could facilitate protein–protein interactions by

selectively including or excluding proteins [13,14]

Membrane microdomains have mostly been studied

in animal and yeast cells; however, it was recently

suggested that similar lipid rafts enriched in GlcCer

and sterols also exist in plant plasma membranes

from tobacco leaves and BY2 cells as well as in

callus membranes [15,16]

Serine palmitoyltransferase catalyses the first step

in sphingolipid biosynthesis, which is the formation

of 3-ketosphinganine from the condensation of serine

and palmitoyl CoA [17,18] The 3-ketosphinganine is

reduced to sphinganine, which is subsequently

acyl-ated to produce ceramide In mammalian cells,

cera-mide is synthesized in the endoplasmic reticulum

(ER) and translocated to the Golgi compartment for

further conversions into more complex sphingolipids

The ceramide transport protein mediates intracellular

trafficking of ceramide between ER and the Golgi in

a non-vesicular manner [19] The biosynthesis of

GlcCer is catalyzed by a UDP-glucose:ceramide

glucosyltransferase (GlcCer synthase), which transfers

glucose to the ceramide backbone [20] In mammalian

cells, GlcCer is synthesized at the cytosolic surface of

the Golgi membrane In Drosophila melanogaster, the

GlcCer synthase GlcT-1 has also been identified in

pre-Golgi compartments including the ER, indicating

that ER to Golgi ceramide transport may not always

be necessary for GlcCer synthesis [2] GlcCer synthase

has also been identified in plants, but the intracellular

location of the plant enzyme has not been determined

[21] GlcCer is enriched in the plasma membrane and

endosomes, suggesting that there is a need for

trans-port of GlcCer from the Golgi to the plasma

mem-brane Transport of GlcCer probably occurs via

transport vesicles and non-vesicular monomeric

trans-port through the cytosol [22] Non-vesicular transtrans-port may be mediated by glycolipid transfer proteins (GLTPs), which accelerate the transfer of GSLs between membranes in vitro GLTPs are specific for GSLs, such as GlcCer and GalCer for example, which have sugar residues attached via b-linkages to the lipid hydrocarbon backbone [23]

Glycolipid transfer protein was discovered initially

in membrane-free cytosolic extracts of bovine spleen [24], and later in a wide variety of tissues [23,25] It

is a ubiquitous, basic (pI 9), soluble protein of

24 kDa [26] The crystal structures of apo-GLTP and lactosylceramide-bound GLTP show a topology dominated by a-helices with a single binding site for the GSL [27] So far, no phenotypes have been asso-ciated with a lack of functional GLTP in metazoans The HET-C2 protein from the filamentous fungi Podospora anserina shows sequence similarity to the mammalian GLTPs, GSL transfer activity [28] and a functional GSL binding site similar to that of mam-malian GLTPs [29] Inactivation of the het-c2 gene leads to abnormal ascospore formation, and it has been suggested that HET-C2 is involved in regulating cell-compatibility interactions during the hetero-karyon formation that occurs during hyphal fusion between different strains [30,31] The mammalian four-phosphate adaptor protein 2 (FAPP2) protein contains a domain with similarity to GLTPs, con-nected to a pleckstrin homology domain Two recent studies have indicated that GlcCer synthesis in early Golgi compartments, as well as its transport by FAPP2 to distal Golgi compartments, is required for protein transport out of the distal compartments [32,33]

The lethal recessive knockout of the A thaliana gene ACCELERATED CELL DEATH 11 (ACD11) shows activation of programmed cell death ACD11 shares 30% similarity to mammalian GLTPs, and has been suggested to be orthologous to mammalian GLTPs However, ACD11 does not translocate GSLs in vitro; instead, it facilitates the intermembrane transfer of single-chain sphingosine [34] Our aim was to deter-mine whether plants encode and express GLTPs with specificity for GSLs We have identified three genes in the A thaliana genome, At1g21360, At2g33470 and At3g21260, that, based on sequence analysis, encode GLTP-like sequences At2g33470 and At3g21260 were also recently identified as putative GLTPs by Jouhet

et al [35] According to our structural models, At1g21360 and At2g33470 have the necessary features for binding GSL In vitro lipid transfer assays con-firmed that At2g33470 (designated as AtGLTP1) is in fact a GLTP with specificity for GlcCer

Identification of A thaliana glycolipid transfer

proteins

We used the amino acid sequences of bovine and

human GLTPs to search databases for GLTP-like

pro-teins from other eukaryotic organisms Putative

GLTPs were detected in vertebrates, insects and

nema-todes, but also in the cnidarians Hydra magnipapillata

and Nematostella vectensis, in the choanoflagelate

Mo-nosiga ovata, in fungi classified as zygomycota,

basid-omycota and ascbasid-omycota, in green algae and land

plants, in species of the phylum Apicomplexa, such as

Cryptosporidum and Plasmodium, and in the diplomo-nad Giardia lamblia We could not identify any GLTP-like proteins in the yeasts Schizosaccharomyces pombe and Saccharomyces cerevisae, in the slime mold Dictyo-stelium discoideum, in ciliates, or in Trypanosoma and Leishmania(Kinetoplastida)

There have been no reports of any plant GLTP with specificity for GSLs, and therefore it was particularly interesting to discover that A thaliana proteins from five genes (ACD11, At1g21360, At2g33470, At3g21260 and At4g39670; Fig 1A), gave blast e-values below 5e-05 when the amino acid sequence of human GLTP was used as bait These five genes encode proteins with amino acid sequences that show 18–25% identity and

A

B

Fig 1 Analysis of the amino acid

sequences of putative Arabidopsis thaliana

GLTPs (A) Percentage of amino acid

sequence similarity and identity from

pair-wise comparisons of the identified putative

A thaliana GLTPs and human (Hs) GLTP In

each case, the value before the solidus

cates identity, and that after the solidus

indi-cates similarity (B) Multiple amino acid

sequence alignment of AtGLTP1,

At1g21360, At3g21260, ACD11 and

At4g39670 The amino acid sequences of

human GLTP, the fungus Podospora

anseri-na HET-C2 and the red alga Galdieria

sulphu-raria GLTP are also included Black boxes

indicate that identical amino acids are

pres-ent in at least four of the sequences, while

shaded boxes indicate that amino acids with

similar physicochemical properties are

pres-ent in at least four of the sequences.

32–39% similarity to human GLTP (Fig 1B) The

highest similarity scores between the A thaliana

puta-tive sphingolipid transporters are found between

At1g21360 and At3g21260, which share 54% identity

and 70% similarity, and between ACD11 and

At4g39670, with 45% similarity and 61% identity The

putative molecular masses of these A thaliana proteins

range from 17 kDa for At3g21260 to 27 kDa for

At4g39670 ACD11 has previously been shown to

facilitate the intermembrane transfer of single-chain

sphingosine in vitro, but does not transfer GSLs

in vitro [34] The lethal recessive knockout of ACD11

shows activation of programmed cell death There are

no reports on the physiological function, biochemical activity or regulation of the other proteins identified

To investigate the relationship between these A tha-liana proteins and known and putative eukaryotic GLTPs, a phylogenetic tree (Fig 2) was constructed from the amino acid sequences using the maximum-likelihood method [36] The phylogenetic analysis sug-gests that At1g21360, At2g33470 and At3g21260 share

a common origin with metazoan and fungal GLTPs The tree indicates a close relationship between At1g21360 and At3g21260, and suggests that this gene pair evolved from a duplication event that occurred after the split of monocotyledons and dicotyledons

Fig 2 Phylogenetic tree of glycolipid trans-fer protein amino acid sequences recon-structed by the maximum-likelihood method Numbers indicate the percentage

of 100 bootstrap re-samplings that support the inferred topology Only bootstrap values over 50% are shown Sequences are identi-fied by gene names or by National Center for Biotechnology Information GI numbers Arrows indicate amino acid sequences of putative GLTPs from A thaliana.

The tree further indicates that a gene duplication that

occurred before the split of gymnosperms and

angio-sperms is responsible for the formation of the

At2g33470 and At1g21360⁄ At3g21260 gene families

ACD11 and At4g39670 group close together in a

sepa-rate plant-specific branch, containing sequences from

other land plants and green algae The evolutionary

relationship of this plant-specific branch to the GLTP

branch of the tree is unclear On the basis of the

results of the sequence analyses, we concluded that

At1g21360, At2g33470 (here on referred to as

AtGLTP1) and At3g21260 are possible candidates for

A thalianaGLTPs We therefore focused our attention

on gaining insight into the biological role, activity and

ligand specificity of these proteins

Structural modeling of putative A thaliana

GLTPs

Structural models of AtGLTP1 (At2g33470),

At1g21360 and At3g21260 in apo form were

con-structed (supplementary Fig S1) in order to examine

whether they have similar structural features to human

GLTP, supporting the theory that the proteins are

GLTPs Based on the sequence alignments used for

modeling, AtGLTP1 shares a sequence identity of

23% with human GLTP and 27% with Galdieria

sulphuraria GLTP (GsGLTP) The overall folding of

the human apo-GLTP and apo-GsGLTP X-ray

struc-tures is very similar, but they are clearly different at

the N- and C-termini The longer N-terminal part of

GsGLTP forms an a-helix The C-terminal part of

GsGLTP is a long unstructured loop stretching away

from the sugar-binding site, whereas the C-terminus in

human GLTP is much shorter and participates in

ligand binding in the complex structures [27,37]

The AtGLTP1 (supplementary Fig S1A) and

At1g21360 models have a two-layer all-a-helical

topol-ogy, with a binding pocket for a sugar moiety lined by

polar amino acids and a hydrophobic tunnel suitable

for binding the hydrocarbon chains of lipids The

hydrophobic nature of the tunnel is highly conserved,

although only a few of the amino acids are totally

con-served At3g21260 is considerably shorter than

AtGLTP1 and At1g21360, missing residues 1–57 and

1–74, respectively (Fig 1B) This means that the model

of At3g21260 lacks the first layer of a-helices and

consequently half of the hydrophobic tunnel

In human GLTP, the residues Asp48, Asn52, Trp96

and His140 have been shown by point mutations to be

the most important residues for the recognition of

sugar-amide moieties [27,38] (Fig 3B) In AtGLTP1,

these residues are conserved and correspond to Asp52,

Asn56, Trp99 and His138 (Figs 1B and 3C–E), and are also totally conserved in At1g21360 (Fig 1B and supplementary Fig S1B) At3g21260 lacks the aspar-tate and the asparagine, as it is much shorter at the N-terminus, but has the conserved tryptophan and his-tidine (Fig 1B and supplementary Fig S1D) When

we compared the other residues (Lys55, Leu92, Tyr207 and Val209) that interact with GSLs in human GLTP– GSL complex structures, some interesting differences were identified between human GLTP and the putative

A thaliana GLTPs Firstly, Lys55 in human GLTP is replaced by Arg59 in the sugar recognition center of AtGLTP1, and there is also an arginine in this posi-tion in At1g21360 and At3g21260 Secondly, the resi-due equivalent to Leu92 in human GLTP is Asn95 in AtGLTP1 (Figs 1B and 3C–E and supplementary Fig S1B) This residue is replaced by an arginine in both At1g21360 and At3g21260 (Fig 1B) Thirdly, the residue corresponding to Tyr207 in the human GLTP

is a lysine in both AtGLTP1 (Lys200) and At3g21260, but an arginine in At1g21360 This makes the sugar-binding pocket of At1g21360 very rich in arginines Fourthly, the residue corresponding to Val209 in human GLTP corresponds to Ser202 in AtGLTP1, a methionine in At1g21360 and a proline in At3g21260 (Fig 1B) In summary, the modeling shows that the AtGLTP1 and At1g21360 proteins are probably GLTPs, as they share extensive structural similarities with human GLTP Amino acid replacements in the sugar recognition center suggest that AtGLTP1 and At1g21360 may have different sugar-binding properties compared to human GLTP At3g21260 has an incom-plete hydrophobic binding cavity, which indicates that

it is not able to bind GSLs

Lipid transfer capability of AtGLTP1

In order to examine whether AtGLTP1, At1g21360 and At3g21260 show a glycolipid transfer activity simi-lar to that found for mammalian GLTPs [39], we expressed the Arabidopsis proteins in Escherichia coli

To test the lipid transfer capacity of the proteins, we used a previously established transfer assay, which relies on resonance energy transfer between a transfer-able (energy donor) fluorescent lipid and a non-trans-ferable (energy acceptor) fluorescent lipid from a donor vesicle population to an acceptor population [40–43] The intervesicular trafficking of three BODIPY-labeled glycolipids, GlcCer, GalCer and LacCer, and a BODIPY-labeled SM was monitored as

a function of time between donor 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) vesicles and POPC acceptor vesicles (in a tenfold excess) using 4 lg

of protein AtGLTP1 is able to efficiently transfer

BODIPY-GlcCer (Fig 4A, black trace), whereas

BODIPY-GalCer and BODIPY-LacCer are transferred

only to a limited extent (Fig 4A, red and green

traces) BODIPY-SM was not transferred at all (yellow

trace, Fig 4A) On the basis of its capacity to enhance

the translocation of BODIPY-GlcCer, we decided to

designate At2g33470 as AtGLTP1 At1g21360 and

At3g21260 are not able to transfer any of the

BODIPY-labeled lipids under the conditions of the

resonance energy transfer assay (Fig 4B) Human

GLTP is able to efficiently move all three labeled

gly-colipids, but not BODIPY-SM (Fig 4D) Numerical

values for the transfer rates are given in Table 1

Using a competition assay, we also analyzed the

substrate specificity of AtGLTP1 for

monogalactosyl-diacylglycerol (MGDG) and digalactosylmonogalactosyl-diacylglycerol

(DGDG; supplementary Fig S2) There was no change

in the transfer of BODIPY-GlcCer after addition of

MGDG and DGDG, which indicates that neither

MGDG nor DGDG are substrates for AtGLTP1 Addition of POPC vesicles was used as a reference Human GLTP appears to be able to transfer DGDG and MGDG to some extent (supplementary Fig S2),

A

Fig 3 The human GLTP crystal structure and the sugar-binding pocket of AtGLTP1 (At2g33470) (A) The fold of the human GLTP crystal structure with bound LacCer in yellow (B) Close-up of the sugar-binding pocket in the crystal structure of human GLTP with bound GalCer

in yellow Binding residues are shown in grey (C–E) Sugar-binding pocket in the structural models of AtGLTP1 (At2g33470) in complex with GlcCer (C), LacCer (D) and GalCer (E) The difference between the glucosyl and galactosyl units [in (B), (C) and (E)] is the orientation of the OH4 hydroxyl (marked with arrow) (F) Sugar-binding pocket in the structural model of the AtGLTP1 Arg59Lys ⁄ Asn95Leu mutant Conserved residues are shown in grey and non-conserved residues are shown in green (C–F) GSLs are in yellow.

Table 1 AtGLTP1 and human GLTP in vitro lipid transfer activity The GLTP-mediated (4 lg) BODIPY-labeled lipid transfer was exam-ined using a fluorescence assay, and the values given are means ± SD from at least four analyses Rates are given as pmol transferred per second.

Transfer rate (pmolÆs)1)

BODIPY-GalCer 0.08 ± 0.04 BODIPY-LacCer 0.02 ± 0.01

BODIPY-GalCer 0.93 ± 0.06 BODIPY-LacCer 0.83 ± 0.04 Arg59Lys ⁄ Asn95Leu mutant BODIPY-GlcCer 0.006 ± 0.01

BODIPY-GalCer 0.012 ± 0.01

which is in agreement with previous findings using porcine GLTP purified from brain [44] Control experiments (supplementary Fig S3) with addition of unlabeled GlcCer to the AtGLTP1 transfer assay indicate that GlcCer competes for the labeled BODIPY-GlcCer substrate, as the transfer trace tapers off significantly after addition of the GlcCer vesicles The BODIPY-GlcCer transfer continues at the same rate after addition of POPC, MGDG and DGDG

Analysis of the GSL transfer specificity of AtGLTP1

In order to obtain an understanding of the differences in the ligand transfer activities between the plant and mammalian GLTPs, we attempted to change the GSL binding specificity of AtGLTP1 through mutagenesis Models of AtGLTP1 in complex with GSLs were con-structed in order to identify amino acids suitable for mutagenesis In the human GLTP–GSL complex struc-tures [27,37], the first sugar unit stacks with Trp96 and forms a network of hydrogen bonds with Asp48, Asn52, Lys55 and Tyr207 (Fig 3B) In AtGLTP1, Trp99, Asp52 and Asn56 are conserved, while the lysine and tyrosine are replaced by Arg59 and Lys200 (Fig 3C–E)

In the AtGLTP1–GlcCer and AtGLTP1–GalCer models, Arg59 could hydrogen bond with the sugar unit similarly to the corresponding residue Lys55 in human GLTP (Fig 3C,E) Interestingly, however, in the AtGLTP1–LacCer model, Arg59 appears to sterically hinder binding of the lactosyl group (Fig 3D) In the human GLTP–LacCer structure [27], Leu92 forms a hydrophobic interaction with the Gal ring, while the corresponding Asn95 residue in AtGLTP1 cannot form similar hydrophobic contacts (Fig 3D) On the other hand, Asn95 appears to play an important role in the specific binding of GlcCer, as its amine group binds the OH4 hydroxyl of Glc in our AtGLTP1–GlcCer model (Fig 3C) In agreement with the low GalCer transfer activity of AtGLTP1 (Table 1), Asn95 cannot bind the OH4 hydroxyl of Gal in our AtGLTP1–GalCer model,

as it points away from the amine group of Asn95 (Fig 3E)

A

B

C

D

Fig 4 Representative time-course traces for BODIPY-labeled Glc-Cer, GalGlc-Cer, LacCer and SM transfer by (A) AtGLTP1 (At2g33470), (B) At1g21360 and At3g21260 (no activity), (C) AtGLTP1 Arg59Lys ⁄ Asn95Leu (no activity), and (D) human GLTP, from donor to accep-tor vesicles The donors contained 0.5 mol% of BODIPY-GlcCer, BODIPY-GalCer, BODIPY-LacCer or BODIPY-SM and 3 mol% DiI-C18 in a POPC matrix, and the acceptor vesicles contained 100% POPC The assay was run at 37 C in sodium phosphate buffer containing 140 m M NaCl.

As the structural models indicated that Arg59 and

Asn95 could be responsible for the altered GSL

trans-fer specificity of AtGLTP1, these residues were chosen

for site-directed mutagenesis to the corresponding

human GLTP residues, Lys and Leu, respectively

According to the ligand docking results, the AtGLTP1

Arg59Lys⁄ Asn95Leu mutant generated has a similar

GlcCer binding mode as the wild-type AtGLTP1 (data

not shown) The activity of the AtGLTP1 mutant was

tested in the lipid transfer assay, which showed that it

had lost the specific GlcCer transfer capability of

AtGLTP1, without gaining any increased capacity for

GalCer transfer (Fig 4C and Table 1) Seemingly,

Arg59 and⁄ or Asn95 are responsible for the specific

binding of GlcCer to AtGLTP1 However, substituting

Arg59 and Asn95 with the corresponding residues of

human GLTP did not increase the overall GSL trans-fer activity of AtGLTP1, and thus this diftrans-ference can-not explain why AtGLTP1 shows much lower GSL transfer activity compared to human GLTP (Table 1)

Expression of AtGLTP1, At1g21360 and At3g21260 during development

To assess the expression pattern of AtGLTP1, At1g21360 and At3g21260, we retrieved relevant data from microarray analyses of gene expression during

A thaliana development, accessible in public databases (http://www.arabidopsis.org, http://www.genevestigator ethz.ch, http://www.weigelworld.org) Figure 5 shows the expression of AtGLTP1, At1g21360 and At3g21260 in 63 samples from various tissues or

Fig 5 Expression of AtGLTP1, At1g21360 and At3g21260 in A thaliana tissues The data are from the microarray experiment AtGenExpress expression atlas of A thaliana [45] (http://www.weigelworld.org) The investigated tissue samples are from roots (RO; sam-ples 1–7), stems (ST; samsam-ples 8–10), leaves (LE; samsam-ples 11–25), whole plants (WP; samsam-ples 26–36), shoot apex (SA; samsam-ples 37–40), floral organs (FL; samples 41–55) and seeds (samples 56–63) of A thaliana Col-0 Plants were grown on soil, unless an alternative growth sub-strate is indicated (1) root, 7 days; (2) root, 17 days; (3) root, 1· MS agar, 1% sucrose, 15 days; (4) root, 8 days, 1· MS agar; (5) root,

8 days, 1· MS agar, 1% sucrose; (6) root, 1· MS agar, 21 days; (7) root, 1· MS agar, 1% sucrose, 21 days; (8) hypocotyl, 7 days; (9) 1st node, ‡ 21 days; (10) 2nd internode, ‡ 21 days; (11) cotyledons, 7 days; (12) leaf numbers 1 + 2, 7 days; (13) rosette leaf number 4,

10 days; (14) rosette leaf number 2, 17 days; (15) rosette leaf number 4, 17 days; (16) rosette leaf number 6, 17 days; (17) rosette leaf num-ber 8, 17 days; (18) rosette leaf numnum-ber 10, 17 days; (19) rosette leaf numnum-ber 12, 17 days; (20) petiole leaf numnum-ber 7, 17 days; (21) proximal half of leaf number 7, 17 days; (22) distal half of leaf number 7, 17 days; (23) leaf, 1· MS agar, 1% sucrose, 15 days; (24) senescing leaves,

35 days; (25) cauline leaves, ‡ 21 days; (26) seedling, green parts, 7 days; (27) seedling, green parts, 1· MS agar, 8 days; (28) seedling, green parts, 1· MS agar, 1% sucrose, 8 days; (29) seedling, green parts, 1· MS agar, 21 days; (30) seedling, green parts, 1· MS agar, 1% sucrose, 21 days,; (31) rosette after transition to flowering, but before bolting, 21 days; (32) rosette after transition to flowering, but before bolting, 22 days; (33) rosette after transition to flowering, but before bolting 23 days; (34) vegetative rosette, 7 days; (35) vegetative rosette,

14 days; (36) vegetative rosette, 21 days; (37) shoot apex, vegetative + young leaves, 7 days; (38) shoot apex, vegetative, 7 days; (39) shoot apex, transition (before bolting), 14 days; (40) shoot apex, inflorescence (after bolting), 21 days; (41) flower, stage 9; (42) flower, stage 10 ⁄ 11; (43) flower, stage 12; (44) flower, stage 15; (45) flower, 28 days; (46) pedicel, stage 15; (47) sepal, stage 12; (48) sepal, stage 15; (49) petal, stage 12; (50) petal, stage 15; (51) stamen, stage 12; (52) stamen, stage 15; (53) pollen, 6 weeks; (54) carpel, stage 12; (55) carpel, stage 15; (56) siliques, with seeds stage 3; mid-globular to early heart embryos (57) siliques, with seeds stage 4; early to late heart embryos (58) siliques, with seeds stage 5; late heart to mid-torpedo embryos (59) seeds, stage 6, without siliques; mid to late-torpedo embryos (60) seeds, stage 7, without siliques; late-torpedo to early walking-stick embryos (61) seeds, stage 8, without siliques; walking-stick

to early curled cotyledons embryos; (62) seeds, stage 9, without siliques; curled cotyledons to early green cotyledons embryos; (63) seeds, stage 10, without siliques; green cotyledons embryos.

stages of development The data are from the

AtGenExpress expression atlas

(http://www.weigel-world.org/resources/microarray/AtGenExpress) [45]

AtGLTP1 mRNA is ubiquitous in all tissues and at

all stages of life of the plant The highest levels of

AtGLTP1 mRNA were found in floral tissues

(Fig 5, samples 41–55), such as petals (sample 50),

stamens (sample 51) and sepals (sample 48), and in

stems (Fig 5, samples 8–10) The At3g21260

tran-script was most abundant in roots (sample 6), but

was also detectable in most other tissues and

devel-opmental stages The levels of At3g21260 transcripts

were lower in all analyzed tissues compared to

AtGLTP1 mRNA The transcription of At1g21360 is

more restricted, as the transcript was only detected

in roots (samples 3–7) The expression of AtGLTP1

and At1g21360 was also analyzed using RT-PCR

(supplementary Fig S4), and AtGLTP1 mRNA was

found to be ubiquitous in all tissues and at all

developmental stages tested At1g21360 mRNA was

also detectable in all tested tissue samples

(supple-mentary Fig S4), suggesting that At1g21360 mRNA

is expressed at a low level in the whole plant

We fused the AtGLTP1 and At1g21360 promoters

to the GUS reporter gene The constructs were

transformed into A thaliana, and the temporal and

spatial patterns of expression from these gene fusions

were assessed during plant growth and development

(Fig 6) In young seedlings carrying AtGLTP1::GUS,

staining in roots was restricted to the tips

(Fig 6A,B) In the roots of more mature plants,

staining was still found in the tips, but also in the

stelar tissue of the elongation zone (Fig 6D) The

root cap did not show any GUS expression In

young seedlings, GUS activity was also present in

the tips of the cotyledons and in the first leaf

pri-mordia (Fig 6A,C) Additionally, staining was seen

in hydathodes and epidermis of cotyledons (Fig 6E)

and rosette leaves (Fig 6F) In leaf epidermis, GUS

staining appeared to be more intense in stomatal

cells (Fig 6F) GUS staining was also seen in floral

tissues, such as the receptacle (Fig 6G,I), petals

(Fig 6G), floral buds (Fig 6G,H), styles (Fig 6H)

and anther filaments (Fig 6I) Staining was most

evi-dent in distal regions of the floral tissues Expression

from At1g21360::GUS was only detected in roots

(Fig 6J–O) In young At1g21360::GUS seedlings,

GUS staining was restricted to cells in the region of

the root, from which root hairs develop, and to root

hairs (Fig 6J–L) In older At1g21360::GUS plants,

GUS activity could also be detected in the basal

regions of lateral roots

Discussion

In this report, we identified three A thaliana paralogs, At1g21360, At2g33470 and At3g21260, as orthologs to mammalian GLTPs At1g21360, At2g33470 and At3g21260form a small gene family that has its origin

in a gene duplication event before the split of gymno-sperms and angiogymno-sperms, and another duplication that occurred after the split of monocotyledons and dicoty-ledons We designated At2g33470 as AtGTLP1 because we had shown that it was a true GLTP with capacity to stimulate the in vitro transfer of GSLs from donor to acceptor vesicles AtGLTP1 could efficiently transfer GlcCer, but the transfer of GalCer and Lac-Cer was negligible Human GLTP efficiently moved all three tested glycolipids It appears that amino acid replacements that narrow the GSL transfer repertoire have been tolerated in AtGLTP1 due to the lack of GalCer and LacCer in Arabidopsis tissues

Based on modeling of the AtGLTP1 structure, we concluded that the Lys55⁄ Arg59 and Leu92 ⁄ Asn95 replacements most likely mediate the differences in GSL transfer specificity between human GLTP and AtGLTP1 We therefore constructed an AtGLTP1 Arg59Lys⁄ Asn95Leu mutant (Fig 3F), which had a very low transfer activity for both GlcCer and GalCer (Table 1), confirming that Arg59 and⁄ or Asn95 in AtGLTP1 are extremely important for specific GlcCer binding (Fig 3C) Lys200 and Ser202 (Tyr207 and Val209 in human GLTP, Fig 3B) are the only differ-ences with respect to human GLTP in the sugar-bind-ing site of the AtGLTP1 Arg59Lys⁄ Asn95Leu mutant (Fig 3F) In the human GLTP–LacCer crystal struc-ture [27], the ceramide amide group is oriented by hydrogen bonds, which are aligned by the hydrophobic contacts between Val209 and the initial three-carbon ceramide segment of LacCer, but Ser202 in AtGLTP1 cannot form similar hydrophobic contacts Lys200 in AtGLTP1 is equivalent to Tyr207 in human GLTP, which forms a hydrogen bond with the glucose ring of LacCer [27] The role of Tyr207 in the GSL transport

of human GLTP has been studied by point mutation

to a leucine, which had a slight effect on transfer activ-ity [27], but there is no documentation regarding the importance of Val209 on GSL transfer activity In con-clusion, we have shown that Asn95 and⁄ or Arg59 are involved in GlcCer binding Further mutational studies will be conducted in order to pinpoint the residues responsible for the specific binding of GlcCer to AtGLTP1 and to determine the reason for the lower overall GSL transfer activity of AtGLTP1 compared

to human GLTP

B C

G

H

O

L A