Synthesis of Gb3 Glycosphingolipids with Labeled Head Groups: Distribution in Phase‐Separated Giant Unilamellar Vesicles

Abstract The receptor lipid Gb3 is responsible for the specific internalization of Shiga toxin (STx) into cells. The head group of Gb3 defines the specificity of STx binding, and the backbone with different fatty acids is expected to influence its localization within membranes impacting membrane organization and protein internalization. To investigate this influence, a set of Gb3 glycosphingolipids labeled with a BODIPY fluorophore attached to the head group was synthesized. C24 fatty acids, saturated, unsaturated, α‐hydroxylated derivatives, and a combination thereof, were attached to the sphingosine backbone. The synthetic Gb3 glycosphingolipids were reconstituted into coexisting liquid‐ordered (l o)/liquid‐disordered (l d) giant unilamellar vesicles (GUVs), and STx binding was verified by fluorescence microscopy. Gb3 with the C24:0 fatty acid partitioned mostly in the l o phase, while the unsaturated C24:1 fatty acid distributes more into the l d phase. The α‐hydroxylation does not influence its partitioning.


Introduction
Theeukaryotic plasma membrane of animals is aheterogeneous structure with aplethora of different lipids.The main lipid components are glycerophospholipids,s terols,a nd sphingolipids. [1] Among them, glycosphingolipids serve aparticular role.T hey are found in the outer leaflet of the plasma membrane and are discussed to reside preferentially in so-called raft domains,w hich are enriched in sphingomyelin (SM) and cholesterol (Chol). [2][3][4] Their size,chemical composition, and physical characteristics are tightly associated with their signal processing capabilities. [5] Raft domains are supposed to have diameters of 10-200 nm and are highly dynamic structures. [3,6] This combination of smallness and dynamics bears the major challenge in visualizing raft domains in cellular membranes. [4] Hence,t wo approaches have been pursued within the last decades to shed some light on the structure and function of these domains.O nt he one hand, detergent-resistant membranes were extracted from cells and their composition analyzed, however they turned out to be prone to artefacts. [7] On the other hand, artificial membranes with lipid compositions resembling the outer leaflet of the plasma membrane were reconstituted, which separate into aliquid-disordered (l d )and aliquid-ordered (l o ) phase. [8] Typical lipid compositions comprise al ow-melting glycerophospholipid, ah igh-melting glycerophospholipid or SM, and Chol. [9] The l d phase has loose lateral lipid packing, acyl chains with gtg kinks,a nd fast lateral diffusion. In contrast, the l o phase is characterized by at ighter lipid packing and ah igher degree of order,b ut still rather fast lateral diffusion. [10] However,the size and physical properties of l o domains formed in artificial membranes are very different from those found in the plasma membrane.T his difference becomes obvious if comparing,f or example,t he physicochemical properties of coexisting l o /l d phase-separated GUVs with those of phase-separated cell-derived membranes termed giant plasma membrane vesicles (GPMVs). [11] Despite this difference between the natural and artificial membrane systems,a rtificial coexisting l o /l d membranes have been frequently used to analyze the partitioning of receptor lipids and proteins, [12] such as bacterial toxins,i nt he different phases. [13] Bacterial toxins are known to bind to specific glycosphingolipids embedded in the outer leaflet of the plasma membrane.Cholera toxin (CTx) produced by Vibrio cholerae and Shiga toxin (STx) produced by Shigella dysenteriae and by enterohemorrhagic strains of Escherichia coli,both belonging to the class of AB 5 toxins, [14] bind specifically to monosialotetrahexosylganglioside (G M1 ) [15] and globotriaosyl ceramide (Gb 3 ), [16,17] respectively.W hile the head groups of the glycosphingolipids indeed define the specificity of protein binding,not much attention has been drawn to the variability of the ceramide backbone harboring different fatty acids.I n various cell types (human colon Caco-2, HCT-8 epithelial cells,h uman endothelial cell lines,p rimary human umbilical vein endothelial cells,primary human endothelial cells of the brain and the kidney, [18] and references therein), aconserved repertoire of Gb 3 species was found carrying saturated C 16:0 , C 22:0 ,orC 24:0 fatty acids as well as the unsaturated C 24:1 fatty acid. Results of Lingwood and co-workers [19] suggest that the pathogenic outcome of Shiga toxin producing E. coli (STEC) infections is related to the different Gb 3 species.T og ather more molecular information, artificial membranes doped with Gb 3 were employed. In coexisting l o /l d supported lipid membranes,Gb 3 species differing in their fatty acid gave rise to adifferent phase behavior before and after binding of the B subunits of STx (STxB) as well as differences in the protein organization on the membrane surface. [20,21] In giant unilamellar vesicles (GUVs), Gb 3 species with an unsaturated acyl chain caused the formation of tubular invaginations upon STxB binding,i nc ontrast to Gb 3 with as aturated acyl chain. [22] In all these studies,i tb ecame evident that STxB binds exclusively to the l o phase,w hich also implies that the receptor Gb 3 is localized in the l o phase after protein binding. However,i tr emains unclear how Gb 3 is distributed in coexisting l o /l d membranes prior protein binding.
To get access to this information, an approach based on fluorescently labeled Gb 3 molecules can be pursued. However,itturned out that, if afluorescent label is attached to the fatty acid position to ensure that the STxB interaction with the head group is not influenced by the fluorophore,binding of STxB is greatly altered. [23] If af atty acid labeled Gb 3 is reconstituted into l o /l d phase-separated GUVs,t he protein binds to the l d phase and not to the l o phase as known from membranes containing naturally occurring Gb 3 .
To date,o nly af ew examples are found in the literature where synthetic routes towards glycosphingolipids with labeled head groups have been described. [24] Here,wedecided on anew strategy in line with approaches pursued for G M1 and G M3 [25] and focused on head group labeled Gb 3 .The idea is to develop fluorescently labeled Gb 3 glycosphingolipids without altering its binding properties to STxB.W ea ttached af luorophore via an oligoethylene glycol spacer to the 2'-OH group of the middle galactose of the Gb 3 head group,w hich is not involved in STxB binding as deduced from crystal structure analysis [17] and binding studies of different trisaccarides. [26] This approach in turn allows us to alter the fatty acid of the Gb 3 molecules.

Results and Discussion
We synthesized as et of Gb 3 sphingolipids as depicted in Scheme 1. Altogether eight different glycosphingolipids were synthesized and they consist of the globotriaose head group with two different oligoethylene glycol (PEG) linkers,t o which aB ODIPY fluorophore was attached and the sphingosine.S aturated, unsaturated, a-hydroxylated derivatives, and acombination thereof were prepared, all based on aC 24 fatty acid. C 24 fatty acids were chosen as they are the major constituent (> 50 %) found in natural Gb 3 mixtures such as toxin insensitive erythrocytes, [27] HeLa-cells, [28] and HEp-2 cells. [29] To access the head group labeled Gb 3 derivatives with different fatty acids and PEG linker lengths we designed am odular convergent synthesis in which av ariation of the fatty acid and the fluorophore is possible with minimal synthetic effort (Scheme 1). In contrast to semisynthetic approaches,ac onvergent total synthesis ensures the highly defined nature of the obtained material, which was crucial for our biophysical experiments.T he retrosynthetic analysis of the desired structures 1-8 led to four different components. Thec ommercially available BODIPY dye 9 should be attached to the carbohydrate head group in the last step of the synthesis by aH uisgen cycloaddition (click chemistry). Thes phingosine core should be introduced as the azido sphingosine 10.T he azide serves as am asked amine which undergoes amide coupling with the four selected fatty acids (11)(12)(13)(14)w ith aC 24 backbone.A ssembling the globotriose building blocks 15 and 16,inwhich the 2-hydroxy group of the middle galactose was modified with the PEG linker and the reducing end was activated for the glycosylation reaction with 10,w ould be the most challenging endeavor during this synthesis.M onosaccharide building blocks with carefully chosen patterns of temporary and permanent protecting groups had to be synthesized starting from the simple monosaccharides d-glucose and d-galactose.
Naturally occurring Gb 3 molecules carry 24 carbon long fatty acids,e ither saturated or monounsaturated. [30] The galactosyl trichloroacetimidate 17,g alactosyl phosphate 18, and glucoside 19 were identified as suitable precursors to build up the trisaccharide (Scheme 2). They were prepared according to literature procedures. [31] Theunion of 18 and 19 under Lewis-acidic conditions utilizing TMSOTf as apromoter afforded the respective (1!4)-linked disaccharide.Perfect b-selectivity was observed because of the neighboring-group participation of the Fmoc group of 18.D uring the course of the reaction the para-methoxybenzyl group of the galactose was cleaved, [32] yielding the lactose acceptor for the second glycosylation step with 17 under Lewis-acidic conditions without any additional deprotection step.T he desired aconfigured product was isolated as the main product when diethyl ether was used as acosolvent. Subsequent removal of the Fmoc protecting group with piperidine led to the trisaccharide 20.
Thet risaccharide 20 was then equipped with ap entenyl chain in the position where the fluorophore needs to be attached (Scheme 3). In the next step the substrate was subjected to Birch conditions to remove all benzyl protecting groups.Despite the strongly reducing conditions,the anomeric CH 2 CH 2 TMS group and the pentenyl handle stayed intact. Deprotection was followed by DMAP-mediated benzoylation. In contrast to benzyl groups,b enzoyl esters have the advantage that they can be easily removed at the end of the synthetic route without affecting the double bond in the lipid part of the glycosphingolipid. [21] To attach the PEG linker,the double bond of the pentenyl handle was first transformed into at hioester with tioacetic acid under radical conditions.T his species was hydrolyzed under basic conditions and the emerging highly nucleophilic thiol was subsequently reacted with the PEG bromides 21 (13 ethylene glycol units) and 22 (3 ethylene glycol units). To ensure af ull protection of all hydroxy groups,t he benzoylation step was repeated. Finally, the anomeric protecting group was removed with trifluoroacetic acid and the reducing end was converted into the corresponding trichloroacetimidates 15 and 16.
To build up the glycolipid, the trichloroacetimidates were reacted with the protected azidosphingosine 10, [33] which was synthesized starting from the chiral pool compound l-serine (for detailed information see the Supporting Information), in ag lycosylation reaction utilizing TMSOTf as the Lewis acid to afford 23 and 24 in moderate yields (Scheme 4). Comparing experiments with globotriaosyl trichloroacetimidates devoid of the PEG modification indicated that the Lewis-basic linker might hamper this very sensitive glycosylation step.S taudinger reduction of the azides and direct coupling with the fatty acids 11-14, [21,34] without isolating the intermediary amines,afforded the PEG-modified glycosphingolipids 25-32.
Global deprotection under ZemplØnc onditions set the stage Scheme 2. Assembly of the Gb 3 trisaccharide. for the final step of the synthesis.The commercially available BODIPY dye 9 was introduced into the glycosphingolipids by coupling its azide unit with the alkyne moiety of the PEG linker under mild copper(I)-catalyzed conditions (Scheme 5).
In total, eight different fluorescently labeled glycosphingolipids (1)(2)(3)(4)(5)(6)(7)(8), varying in the PEG linker length and the acyl chain of the fatty acid, were obtained. Thelinker length (n)iseither 3o r1 3o ligoethylene glycol groups.T he fatty acid (C m:D )i s either saturated (C 24:0 )o ru nsaturated (C 24:1 ). Hydroxylation at the a-position is indicated by OH, and non-hydroxylation is indicated by H. Starting with the saturated C 24 fatty acid and aP EG spacer composed of 13 oligoethylene glycol units,w ep repared GUVs composed of the well-known raft mixture 1,2dioleoyl-sn-glycero-3-phosphocholine( DOPC)/SM-porc/ Chol labeled with 5mol % 1 and 0.25 mol %T exas Red-DHPE (39.75/35/20/5/0.25) to address the question of whether STxB is indeed still capable of binding to the head group modified Gb 3 and whether it binds to the l o phase as expected. Figure 1shows representative confocal images of aGUV in a5 00 nm STxB-Cy5 (monomer) solution. Texas Red-DHPE partitions preferentially in the l d phase,visualizing the coexisting l o /l d membrane ( Figure 1A). Thef luorescence image of STxB-Cy5 shows that STxB binds to the GUV and that it binds to the l o phase ( Figure 1B). This result confirms our hypothesis that the fluorescent label at the 2'-OH position does not greatly interfere with the binding properties of STxB and is suited to investigate the partition of different Gb 3 species as af unction of the fatty acid in coexisting l o /l d membranes.
To quantitatively compare the phase partitioning among the different Gb 3 species,w eu sed the fluorophore Dy731-DOPE as l d marker [23] to guarantee that the fluorescence of the BODIPY labeled Gb 3 does not spectrally overlap with the absorption of the l d marker.M oreover,t he concentration of Gb 3 was reduced from 5t o1mol %t oe nsure that selfquenching of the BODIPY fluorophore is minimized (see Figure S1 in the Supporting Information). GUVs composed of DOPC/SM-porc/Chol/Gb 3 /Dy731 (39/39/20/1/1) were prepared. As it is known that the composition of GUVs obtained by electroformation is rather heterogeneous, [35] at least two independent GUV preparations with about 30 individual GUVs each were analyzed. Confocal z-stack images were measured for each GUV and line profiles were taken from each slice,where phase separation was visible.Anexample of fluorescence images of a l o /l d coexisting GUV together with the line profile is shown in Figure 2. Thefluorophore Dy731-DOPE indicates the l d phase (Figure 2A). [23] From the BODIPY fluorescence intensity ( Figure 2B), the preferential localization of 1 is visible.T oq uantify the partition of 1,t he BODIPY intensity of the l d phase (I(l d )) and of the l o phase (I(l o )) as obtained from the corresponding line profile was determined and the l o distribution (%l o )w as calculated [Eq. (1)]: [23] Several tens of line profiles were taken from each GUV. All %l o values were cast into ah istogram ( Figure 2C). Data obtained in this manner are presented as violin plots throughout the manuscript.
There is increasing evidence that the size of the linker attached to the head group of alipid alters the phase behavior  of the fluorescently labeled lipid. [36,37] To investigate whether the linker length, that is,the number of ethylene glycol units, influences the partition of the Gb 3 sphingolipids in phaseseparated GUVs,w esynthesized Gb 3 molecules differing in their fatty acid with either 13 ethylene glycol units (PEG 13 )or 3( PEG 3 ). Independent of the fatty acid, the same trend is observed ( Figure 3). All Gb 3 sphingolipids with PEG 13 partition more in the l o phase than the corresponding Gb 3 species with PEG 3 .
Them ean values are summarized in Table 1. Thed ifference between the l o distribution of PEG 13 Gb 3 species and PEG 3 Gb 3 species lies between 0.15 and 0.33 (Table 2, DPEG). Such altered partitioning of al ipid as af unction of linker length, to which af luorophore has been attached, was also observed by Honigmann et al. [36] They reported on af luorophore that was either directly connected to the lipid 1,2distearoyl-sn-glycero-3-phosphoethanolamine (DSPE) or connected by aP EG-linker with 45 ethylene glycol units, and was reconstituted into supported lipid membranes composed of 1,2-diphytanoyl-sn-glycero-3-phosphocholine (DPhPC)/1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC)/Chol. Afluorescence analysis of the partition clearly showed that the fluorescent lipid lacking the PEG-linker was preferentially localized in the l d phase,w hile that with the PEG-linker partitioned into the l o phase.S imilarly,M omin et al. [38] and Bordovsky et al. [37] found that an increase in length of the hydrophilic PEG linker at the head group of lipids that are expected to be localized in the l o phase of coexisting l o /l d membranes is required to favor their partitioning in the l o phase.T his observation is explained by the notion that the fluorophore itself is partially hydrophobic and might be also bulky.Itchanges the packing parameter of the lipid. If the fluorophore is directly connected to the lipid or   (Table 1). The errors are the standard deviation of the mean. N = number of line profiles.

Angewandte Chemie
Research Articles attached by ashort linker, the size of the lipidshead group is expanded and the lipid is more conically shaped, favoring the l d phase. [39] Fort he slightly hydrophobic but small BODIPY fluorophore used in our study,ahydrophilic PEG spacer of suitable length is required to mitigate interactions with the membrane.M omin et al. [38] found al inker with 10 ethylene glycol units to be sufficient to decouple the fluorophore from the membrane. [39] In our study,a13-unit long linker decoupled the fluorophore from the membrane interface with the result that 1,w hich is expected to at least preferentially partition into the l o phase,i ndeed has a l o distribution of almost 0.75. From these results,w ec onclude that the Gb 3 species with PEG 13 are better suited to report on the natural partition of Gb 3 than those with PEG 3 .Thus,the experiments in which we compare the influence of unsaturation and hydroxylation of the fatty acid of Gb 3 are all performed with the PEG 13 species.T he corresponding results with the PEG 3 linker can be found in the Supporting Information (see Figures S2 and S3).
We investigated the influence of the fatty acid saturation on the partition behavior of Gb 3 (Figure 4). Theresults show that introducing af atty acid with a cis-double bond redistributes the Gb 3 sphingolipid in the l d phase,a nd can be rationalized by the increased space requirement of the Gb 3 species with the C 24:1 fatty acid. Thedifferences between the l o distribution of 1/5 and 3/7 harboring the PEG 13 linker are significant and range between 0.21 and 0.27 (Table 2, DC24).
Bjçrkqvist et al. [40] investigated different glycosphingolipids as well as sphingomyelins and found by differential scanning calorimetry (DSC) that the phase-transition temperature was decreased by about 20 Kf or all sphingolipids harboring the C 24:1 fatty acid compared to the corresponding C 24:0 sphingolipids,d emonstrating their different packing behavior. Theability to pack tightly with ordered acyl chains in case of as aturated fatty acid [41] is ar equirement for membrane lipids to partition into l o domains and they concluded that the C 24:1 sphingolipids are less likely to partition into the l o phase.F luorescence quenching experiments revealed that sphingolipids with aC 24:0 fatty acid form l o domains in multicomponent membranes composed of either the sphingolipid or mixed with palmitoyl sphingomyelin. [40] This behavior was also found by Mate et al., [42] who reported that sphingomyelin with the C 24:0 fatty acid reconstituted into aD OPC/Chol membrane leads to visible phase separation into an l o and l d phase,w hile the sphingomyelin with the C 24:1 fatty results only in one lipid phase.
These results support our notion that the packing of the unsaturated Gb 3 species disfavors its partition in the l o phase. Similar to our in vitro results,Legros et al. [43] found in primary human blood brain barrier endothelial cells that Gb 3 with C 24:1 fatty acids resides more strongly in non-detergent-resistant membranes compared to Gb 3 with C 24:0 fatty acids.
In nature,a bout 50 %o ft he Gb 3 sphingolipids are decorated with an OH group in the a-position of the fatty acid, raising the question, whether this OH group alters the Gb 3 partition. Ther esults ( Figure 5) clearly indicate that the OH group in the a-position does not influence its distribution. Monolayer experiments on galactosyl ceramide (GalCer), harboring either an a-hydroxylated or nonhydroxylated C 24:0 fatty acid on aL angmuir trough, suggest that the ahydroxylation does not change the area per lipid at 30 mN m À1 , [44] as urface pressure that reflects the packing density of bilayers. [45] Using 2 HNMR spectroscopy,M orrow and co-workers [41,46] also demonstrated that the order parameter of the fatty acids of GalCer embedded in aP OPC/Chol membrane and the orientation of the head group does not change considerably.
This report is in line with our observation that the OH group does not significantly alter the partitioning of the Gb 3 species in phase-separated GUVs.H owever,i nap revious study,wefound that the 2-OH group influences the fraction of l o phase in phase-separated supported lipid bilayers. [21] In the case of the hydroxylated C 24:0 fatty acid, the l o fraction was smaller than that of the nonhydroxylated species.S lotte and co-workers [47] showed that the 2-OH group increases the hydration in the membrane interface and decreases the affinity of as phingolipid for sterols.T he same was found by Lingwood et al. [48] and Yahi et al. [49] and implies that the   Table 1). recruitment of Chol into the l o phase by hydroxylated Gb 3 is reduced compared to the nonhydroxylated species,leading to asmaller l o fraction, while the amount of Gb 3 in the l o fraction is the same.
Our results clearly demonstrate that the fatty acid of Gb 3 influences its partitioning into the l o phase.One reason might be found in the interaction of the Gb 3 fatty acid with the fatty acid of SM, which we next analyzed. To investigate this aspect in more detail, we replaced the SM mixture isolated from pigs with synthetic pure SM. Exchanging asphingomyelin mixture with sphingomyelins with ad efined fatty acid is known to alter the phase separation behavior of ternary mixtures. [50] Five different SM species with as aturated fatty acid of varying length were chosen, namely palmitoyl SM (C 16:0 ), stearoyl SM (C 18:0 ), arachidoyl SM (C 20:0 ), behenoyl SM (C 22:0 ), and lignoceroyl SM (C 24:0 ), and the l o distribution of each Gb 3 species in these membranes was determined (Figure 6, Table 3).
Thef atty-acid chain length also determines the length difference between the two hydrophobic chains,w hich increases with an increase in fatty-acid chain length. This mismatch results in interdigitation of both leaflets, [51] which was-for fatty acids with al ength of more than 20 carbon atoms-not only observed in the gel phase but also in the liquid-crystalline phase. [52,53] Interdigitation was also reported for glycosphingolipids carrying aC 24 fatty acid. [53,54] Hence,it is likely that the Gb 3 species under investigation preferentially partition into the l o phase if SM interdigitates.I nterdigitation of SM in the liquid-crystalline phase occurs for C 20 fatty acids and longer, in agreement with our observation that the partition in the l o phase is increased for SM species with C 20 fatty acids or longer.
However,the l o phase consists not only of SM but also of Chol owing to its better solubility in SM membranes than in PC membranes. [55][56][57] Chol is best soluble in SM C 16:0 . [56,57] If the solubility of Chol in the l o phase greatly influenced the Gb 3 distribution in the l o phase,the opposite trend would have been observed. This trend was not found and agrees with the idea that the interaction of Gb 3 with Chol is less important than the one with SM.  (Table 3).    Conclusion G M1 and Gb 3 detection by fluorescently labeled Cholera toxin Bsubunits (CTxB) and Shiga toxin Bsubunits (STxB), respectively is aw ell-established tool for monitoring l o membrane domains [58] and implies that these glycosphingolipids are localized in the l o phase.However,aseach CTxB and STxB pentamer can recruit am aximum of 5( CTx) or 15 (STx) receptor lipids,t he glycosphingolipid partitioning in coexisting l o /l d membranes after protein binding does not necessarily reflect the situation prior protein binding.Hence, to be able to quantify the partitioning of Gb 3 in phase coexisting l o /l d membranes by means of fluorescence readout, chemical access to fluorescently labeled pure Gb 3 molecules is required. Thea pproach of synthesizing head group labeled glycosphingolipids enables one to address the question how the fatty acid of aglycosphingolipid influences its distribution in l o /l d phase-separated membranes,aquestion that has been hardly addressed because most of the glycosphingolipids are not available in chemically pure form. Our results clearly demonstrate that the fatty acid (un)saturation significantly shifts the Gb 3 molecules from the l o phase (C 24:0 )t ot he l d phase (C 24:1 ). As STxB exclusively binds to Gb 3 in the l o phase, the amount of redistributed Gb 3 and probably also other l o phase lipids thus depends on the fatty acid of Gb 3 .However, the a-hydroxylation does not alter the partition of Gb 3 ,even though it has been shown that the OH group of Chol can form ah ydrogen bond only to the nonhydroxylated fatty acid. Instead, the length match of the fatty acids of SM and Gb 3 appear to play amore decisive role in determining where the Gb 3 glycosphingolipids are preferentially localized. As the combination of the attached fatty acids of SM and Gb 3 considerably impacts the distribution of the Gb 3 glycosphingolipids,itisconceivable that the overall recruitment of lipids and thus the Shiga toxin induced membrane reorganization that eventually leads to the invagination of the protein into the host cell, is strongly influenced by the fatty acid composition of Gb 3 .