In fl uence of glycosaminoglycans on lipid dynamics in supported phospholipid bilayers

Glycosaminoglycans (GAGs) are important constituents of extracellular matrices (ECMs). As charged polymers, they do most likely influence lipid and protein dynamics in the outer leaflet of plasma membranes. In this study, we investigated their specific effect, depending on concentration, on lipid diffusion in model membranes. In our assay, GAGs are simply attached electrostatically to supported phospholipid (DOPC) bilayers doped with small amounts of cationic lipid (DOTAP) at physiological pH. Lipid dynamics are characterized via the diffusion of fluorescent lipid analogs (DiD/DiO), determined by fluorescence correlation spectroscopy (FCS). We find that diffusion of DiD is significantly affected by the attachment of GAG. Quite surprisingly, short chains (#10 disaccharide units) of hyaluronic acid (unsulfated GAG) on the membrane surface affect the DiD diffusion coefficients stronger than medium or long chains ($100 disaccharide units). In particular, short chains of hyaluronic acids at micromolar concentrations display a 2-fold decrease of the diffusion coefficients compared to the situation without GAG. At nanomolar concentrations of hyaluronic acid of both short and long chains, DiD diffusion remains unaltered. In contrast, sulfated GAGs, such as heparan sulfate (HS) and heparin, affect the lipid diffusion already at sub-micromolar concentrations, albeit not as strongly, with a less than 1.5 fold reduction of the diffusion coefficient. Chondroitin sulfate, another class of sulfated GAGs, did not impose any effect on DiD diffusion in the supported phospholipid bilayer at the concentrations studied. We also investigated desulfated heparin, to explore the role of sulfation and to compare its effect with HA. It is observed that heparin derivatives with lower degrees of sulfation have little effect on the lipid diffusion. Altogether, our results suggest that the presence of certain carbohydrate polymers in the ECM does have a noticeable effect on lipid dynamics in biological membranes.


Introduction
Glycosaminoglycans (GAGs), an evolutionary well-conserved class of carbohydrates, are important components of the extracellular matrix (ECM), attached to the outer surface of the cell membrane. 1,2Structurally, GAG, a polymer chain of repeated disaccharide units (acidic; hexouronic acid and amino; hexosamine monosaccharides, Table 1), is covalently linked to a protein to form proteoglycans or glycoproteins, depending on the ratio between GAG and the protein. 3][6] Members of the GAG family are classied depending on their constituent sugar molecules, i.e., hexosamine (galactosamine and glucosamine), hexose (galactose), and hexuronic acid (glucoronic acid and iduronic acid).They also vary in the geometry of the glycosidic linkage and the substitution of sulfate groups on the sugar molecules.Basically, hyaluronic acid (or hyaluronan: HA) is a non-sulfated GAG, where as chondroitin sulfate (CS), keratin sulfate (KS), heparan sulfate (HS), and heparin are sulfated GAGs.GAGs are highly negatively charged, due to the carboxylic acid and substituted sulfate groups. 7The net negative charges on the GAG molecules attract cations such as Na + , and aer binding to sodium ions, they interact with the water molecules. 8,9][12] GAGs, together with other components of the extracellular matrix, are supposed to be involved in many biological functions at and across the cell membrane.In particular, signal transduction, initiated by the interaction between soluble ligands and their receptors, may be strongly inuenced by the presence of an extracellular matrix. 13Particularly the glycocalyx, an extracellular layer formed by "glycolipids" upon attachment of carbohydrates to lipids, may provide a diffusion barrier and thus, a certain protection to the cell.The specic role of this outer glycocalyx layer, as well as its counterpart in the gel-like intercellular matrix has so far not been fully elucidated.As one aspect, it is quite plausible that the presence of an ECM signicantly affects lipid and protein dynamics in the membrane, which also impacts on essential lipid-protein interactions.
In 2007, Zhang and coworkers reported that the presence of a polymer, quaternized poly(4-vinylpyridine), on the membrane surface creates heterogeneity in the lipid diffusion of the bilayer.They observed a signicant difference between the diffusion coefficients of the lipids which are in contact with the polymer (0.50 AE 0.12 mm 2 s À1 ) and those which are not (2.62 AE 0.18 mm 2 s À1 ). 14Since then, little more has been reported on carbohydrate attachment to membranes and their effects on the lipid dynamics.In a recent study, 15 the effect of mucin glycoprotein, end-functionalized by hydrophobic anchors to incorporate it into the lipid bilayer, on the lipid dynamics was characterized. 16Basically, the incorporation of derivatized mucins (extended with N-acetyl galactosamine, which is one of the monosaccharides in the GAG disaccharide unit) into the cell surface did not alter lipid mobility in the bilayer 17 Furthermore, Quemeneur and coworkers studied the adsorption of HA on a DOPC membrane system as a function of pH. 18,19Although there have been a few recent studies on carbohydrate quantication on membrane surfaces and the effects on the membrane topology, the relationship between the structure and conformation of GAGs and the dynamics of lipid and protein diffusion, and other processes in the bilayer, remains largely unexplored.
Here, we specically focus on the effect of different types and concentrations of GAGs on lipid diffusion in supported lipid bilayers (SLBs).We used four different GAGs: HA, CS, HS, and heparin (Table 1), with nano-to millimolar concentrations, depending on the solubility of the GAGs.In the case of HA, we also used three different chain lengths, i.e., LHA (low hyaluronic acid), MHA (medium hyaluronic acid), and HHA (high hyaluronic acid).The polymers are attached to the supported lipid bilayer by ionic interactions between negatively charged GAG and positively charged lipid DOTAP.As probes for lipid diffusion, we use the lipid analogues DiD and DiO, their diffusion properties are determined by uorescence correlation spectroscopy (FCS).[22][23]  a CS and HS can be obtained in different forms.From the sulfation point of view, CS can have the sulfation in glucoronic acid or N-acetyl galactose (N, 4 and 6 positions), where as HS can have the sulfations at the same position as in CS but mostly, HS is obtained as monosulfated disaccharides.

Lipid membranes
Supported lipid bilayers (SLBs) were formed by standard vesicle fusion techniques. 24The bilayers were composed of 95.0 mol% DOPC, a zwitterionic lipid, 5.0 mol% DOTAP, a positively charged lipid, and 0.01 mol% DiD, a uorescent lipid probe.To rule out a dependence of the measurements on the specic lipid probe, DiO was used as an alternative to DiD.The procedure of SLB formation started with the formation of dry thin lipid lms from the mixture of required lipids (DOPC, DOTAP and DiO/ DiD) by evaporating the solvent.Then the dry thin lipid lm was rehydrated with SLB buffer (10 mM HEPES, 150 mM NaCl and pH 7.4) to obtain MLVs (multi lamellar vesicles).In the following step, the MLVs were sonicated at room temperature to obtain SUVs (small unilamellar vesicles).In the nal step, SUVs were burst in the presence of Ca 2+ ions and spread onto the mica surface, which was previously glued to the glass surface using UV-adhesive (inert towards uorescence).The remaining SUVs (which did not burst on the mica surface) were removed by excess washing with the required buffer (Tris 20 mM, pH 7.0).The bilayer was incubated with the glycans for 45 minutes, unbound glycan was washed away, and the resulting membrane was imaged by uorescence microscopy using a LSM 510 (Zeiss) inverted uorescence microscope and a photo multiplier tube (PMT).

Fluorescence correlation spectroscopy (FCS)
FCS was employed to measure the lateral lipid mobilities and to assess the impact of glycans on lipid dynamics in supported lipid membranes. 20The experiments were performed on a commercial FCS unit, based on an LSM 510 (Confocor 3, Zeiss) inverted uorescence microscope.In brief, a 488/633 nm (as per the requirement) beam from Ar-ion/He-Ne laser was coupled into the light path of the microscope through an optical ber and focused by the water immersion objective (40Â magnication, 1.2 NA) onto the sample.A hardware correlator translates the photon arrival pulses into intensity uctuations and calculates the correlation in real time.The correlation curves were acquired and tted to analytical expressions (eqn (1)).For an averaged correlation measurement, a minimum of 10 separate correlation measurements of each 10 seconds duration were taken, and corresponding standard deviations for every point of the experimental curves were calculated from multiple experiments.
The following analytical form of the temporal auto-correlation G(s) was used to obtain the diffusion time of a uorescent molecule through the confocal observation volume in a twodimensional membrane system: Here, hNi is the average number of molecules in the observation area and s D is the lateral diffusion time of the molecules through the illuminated membrane spot.Using the reciprocal standard deviations as weights, we tted the average correlation curve for each experiment to eqn (1) to extract the s D associated with the diffusion processes.

Diffusion coefficients (D) and diffusion coefficient ratio (D ratio )
The diffusion coefficient of the lipids (particularly DiD or DiO in this case) in the supported lipid bilayer was determined from the experimentally obtained correlation time (s D ) and the known waist (r 0 ) of the confocal observation volume using the following equation: The waist of the observation volume, r 0 , was determined independently from calibration measurements using 50 nM Alexa 488 and 647 (Invitrogen Corp.) in 50 mM Tris of pH 7 and employing the known diffusion coefficient of the uorophore (435 mm 2 s À1 for Alexa 488 and 330 mm 2 s À1 for Alexa 647 in water) as a standard. 25,26The obtained r 0 value is used to calculate the unknown diffusion coefficients of lipid uorescent markers (DiD/DiO) in the lipid bilayer systems in the presence and absence of GAG on its surface.
D ratio is the ratio between the diffusion coefficients of DiD or DiO in the absence (D without GAG ) and presence (D with GAG ) of GAG.D ratio can be written as follows:

Results and discussions
The effect of different GAGs on lipid diffusion was studied by FCS in terms of D ratio (the ratio of the DiD diffusion coefficient without and with GAG), providing qualitative information about the interaction between the GAG polymer and the membrane, in particular, DOTAP.The use of the dimensionless quantity D ratio eliminates the error in determining exact diffusion coefficients, making it independent of the membrane preparation or the inuence of the support.Altogether, larger values for D ratio indicate higher impact of the GAG on lipid diffusion.The diffusion parameters obtained and reported in this article are considered as the averaged result from a heterogeneous membrane system created by the GAG aer interacting with lipid bilayer.Along with FCS, uorescence imaging is used to probe the attachment of uorescently labeled GAG polymers (Fluorescein-labeled HA, HS, CS, and heparin) with the membrane.

Effect of glycosaminoglycan attachment
GAG is a highly negatively charged constituent of the ECM, and in order to attach different types of GAG polymers, positively charged lipid (i.e., DOTAP) was doped into the articial supported phospholipid bilayers (DOPC: 95%, DOTAP: 5% and DiO or DiD: 0.01%).As a proof of GAG attachment to the membrane surface, uorescence imaging and FCS measurements were performed on the supported bilayers on a mica surface.As shown in Fig. 1A, the uorescence intensities were recorded as a function of membrane height (z-stack).The gradual decay over many mm, in spite of the small (few nm) thickness of the membrane, reects on the optical properties of the detection volume.In Fig. 1A, the blue dashed line represents the uorescence of DiO (excited with 488 nm) in the membrane.Fluorescence of DiO was followed before (blue square) and aer (blue circle) addition of LHA (low HA, M w : 3.63 kDa) in order to observe the membrane surface homogeneity.To conrm LHA attachment, a sugar binding protein, WGA: Wheat Germ Agglutinin (a class of Lectin proteins), conjugated to Alexa 594 was used as a uorescent probe, as it selectively binds to the Nacetyl galactosamine of the GAG disaccharide unit.The uorescence signal of Alexa 594 conjugated WGA (excited with 543 nm) is represented by a red dashed line in Fig. 1A.The strong variation in the uorescence signal of probe WGA as a function of distance from the membrane (z-stack) conrms the attachment of GAG to the membrane surface.Consequently, FCS measurements on DiD (excited at 633 nm) indicate a signicant decrease in diffusion coefficient in the presence of LHA (Fig. 1B and C).In a control measurement to exclude a potential role of electrostatic effects between the positively charged DiD/DiO and GAG, SLBs without DOTAP showed no difference in DiD/DiO diffusion coefficients in the presence and absence of LHA.Their inertness towards GAG in contrast to DOTAP could be due to their small sized head groups.A pioneering study by Axelrod in 1979 suggested that cyanine uorescent markers (i.e., DiI, which is similar to DiD/DiO and carries a positive charge) lie parallel to the phospholipids on the membrane surface. 27DOTAP, on the other hand, has a bulkier head group compared to DiD or DiO and thus, is more solvent accessible.
Taken together, we observe no direct interaction between GAG and DiD/DiO.In contrast, the uorescence microscopy and FCS measurements in the presence of DOTAP clearly point to signicant interactions between the positively charged View Article Online membrane surface and the negatively charged GAG molecules (Fig. 2A), which in turn affect lipid diffusion.The absolute intensities of the uorescein-GAG polymers, and thus, the absolute binding affinities, are difficult to compare, as the uorescein labeling is not specic (0.5 to 0.7 uorescein molecule per disaccharide unit depending on the GAG type, as specied by the manufacturer).

Structure-related differences, effects of sulfation
In HA, the carboxylate ion is the only interacting moiety, and in the case of sulfated GAGs, both carboxylate and sulfate groups are likely responsible for membrane binding (structural representation, Table 1).To elucidate potential differences, we compared four different uorescein labeled GAG molecules, i.e., uorescein-HHA (M w : 921 kDa), uorescein-CS (M w : 65 kDa), uorescein-HS (M w : 17 kDa), and uorescein-heparin (M w : 12 kDa).For the latter three, the uorescence signals of DiD and uorescein-GAG decrease proportionally when moving below or above the membrane (Fig. 2A), suggesting a strong interaction of GAG with DOTAP.In contrast, the signal of uorescein-HHA does not show any peak at the membrane surface, but rather stays constant above it, suggesting that it remains largely in solution.In accordance with this observation, there is no noticeable effect on DiD diffusion for HHA, in contrast to the other GAGs.DiD shows an increase in diffusion time (4.0 ms to 10 ms, Fig. 1B) upon attachment of 100 mM LHA, which can be translated to a decrease in diffusion coefficient: from 2.49 Â 10 À8 cm 2 s À1 without to 1.24 Â 10 À8 cm 2 s À1 with.Similarly, for 100 mM MHA, the diffusion coefficient drops from 2.81 Â 10 À8 cm 2 s À1 in the absence to 1.63 Â 10 À8 cm 2 s À1 in the presence (Fig. 1C).The change in DiD diffusion is indicative of the GAG interaction with the DOTAP head groups on the membrane surface.The basic differences among the three sulfated GAGs are their degree of sulfation (number of sulfate groups on one disaccharide unit), chain length, and chemical structure (occurrence of different amino-sugars in disaccharide units).HS and heparin show a noticeable effect on lipid diffusion already at high nanomolar concentrations, whereas CS does not (Fig. 2B).The observed variations in D ratio between CS, HS and heparin could be explained through their differences in chemical structure, including chain length and degree and point of sulfation.The differences in the solution structures of HS and heparin indicate that heparin has 50% trisulfate and 50% disulfate forms, and all the representative disaccharide units are sulfated.In contrast, HS has a repeat of sulfated (either mono-or di-sulfated) and non-sulfated disaccharide units. 28As a result of heavy sulfation in heparin, the concentration dependence of lipid diffusion in the presence of heparin does not seem to display a cooperative effect, compared to HS, where the curve is more sigmoidal (Fig. 2B).In contrast, HS shows a cooperative effect on DiD diffusion upon attachment to the DOTAP in the lipid bilayer.Compared to the structure of heparin, CS has also continuous repeats of sulfated disaccharide units, but it has glucoronic acid instead of iduronic acid like HS. 29,30 The conformational change in the acidic sugar of CS might result in a weaker binding with DOTAP.Consequently, there is almost no effect of CS on the DiD diffusion (Fig. 2B).
Additionally, a comparison of fully sulfated heparin with its desulfated derivative (d-heparin) has been carried out to investigate the role of sulfation.The most important aspect about these derivatives is that the chain length remains the same for all and thus, the effect of chain length can be eliminated.Besides the idea of comparing the role of sulfation, the motive of using desulfated heparin is to compare it with LHA, due to their structural similarity.Table 2 shows a decrease in D ratio from heparin to d-heparin, indicating the impact of sulfate groups on the lipid diffusion in the supported lipid bilayer at several concentrations.As a function of concentration (from 50 to 1000 nM), both derivatives show slight variations in the diffusion coefficient values, but the trend is not clear.Although d-heparin has a structural similarity with HA, the impact on lipid diffusion is different.This effect could be due to the conformational differences in the disaccharide units of HA and d-heparin.

Polymer length dependence
Different lengths of HA polymer as the only unsulfated GAG were used to gain insight into the role of chain length on the interaction between GAG and DOTAP.Besides LHA as mentioned above, the lipid diffusion was also observed in the presence of MHA (medium HA, M w : 95 kDa) and HHA (high HA, M w : $1000 kDa).LHA and MHA have signicant impact, as a function of concentration, on the DiD diffusion, as revealed from the experimental observations (shown in Fig. 3).A pronounced effect on diffusion, quantied by D ratio , could rst be observed at low micromolar concentrations of LHA and MHA (Fig. 3), further increasing up to 100 mM.In contrast, the diffusion coefficient remained unaffected (within error) in the presence of HHA.At 10 mM concentration, D ratio in the presence of LHA and MHA are 1.61 and 1.19, respectively, whereas it is 0.97 for HHA.This difference could be a result of the conformational change in the HA family as a function of chain length.It is conceivable that the binding site in HHA (i.e., carboxylic acid) is not or partially exposed to the solvent, restricting the interaction with membrane surface, compared to MHA and LHA. 31,32

Conclusions
In this study, we investigated the impact of several gylcosaminoglycans (GAGs), being essential constituents of extracellular matrixes, on the dynamics of lipid molecules in membranes.Using electrostatically attached GAGs of negative charge on supported model membranes doped with positively charged lipids (DOTAP) as anchoring sites, we demonstrated that the diffusional mobility of lipids within the GAG-decorated membrane slowed down signicantly, depending on concentration and the exact chemical nature of the GAGs.A strong interaction between GAG and membrane is observed in the case of short chains of GAG (i.e.low hyaluronic acid) and sulfated GAG (i.e.heparan sulfate), suggesting the roles of chain length and degree of sulfation.The nding that diffusing molecules within the membrane are not only directly but also indirectly affected by GAG attachment has important consequences when discussing a possible role of extracellular matrixes for lipid and protein dynamics in cell membranes, and thus, essential processes like cellular signaling.For dening appropriate future in vitro model systems of key physiological processes, constituents of the extracellular matrix will have to be taken more carefully into consideration.

Fig. 1 (
Fig. 1 (A) Fluorescence intensities of DiO in SLB without LHA (blue squares) and with LHA (blue circles) and A594-WGA (red squares) when attached to LHA on the SLB surface.(B) Diffusion curves (from FCS measurements) indicating the differences in the presence (red) and absence (green) of 100 mM LHA and (C) 100 mM MHA.The solid lines represent the fitting curves for corresponding FCS curves.

Fig. 2 (
Fig. 2 (A) Fluorescence intensities of DiD (red square) and different fluorescein-labeled GAG polymers (in blue).The dotted line at z-stack ¼ 0 indicates the membrane surface.(B) Comparison of D ratio of DiD for different sulfated GAGs (i.e., HS, heparin and CS) at nanomolar concentrations.

Fig. 3
Fig. 3 Comparison of D ratio of DiD for different unsulfated GAGs (i.e., LHA, MHA and HHA) at micromolar concentration.

Table 1
Schematic representation and brief structural information regarding the employed GAG polymers a

Table 2
Structural differences between different sulfated and desulfated heparin and their ratio in diffusion coefficients as a function of concentrations