Phospholipid Dependence and Liposome Reconstitution of Purified Hyaluronan Synthase*

Previous radiation inactivation and enzyme characterization studies demonstrated that the Streptococcus equisimilis hyaluronan synthase (seHAS) is phospholipid-dependent and that cardiolipin (CL) is the best phospholipid for enzyme activation. Here we investigated the ability of seHAS, purified in the absence of added lipid, to be activated by synthetic phosphatidic acid (PA), phosphatidylserine, or CL lipids containing fatty acyl chains of different length or different numbers of double bonds. The most effective lipid was tetraoleoyl CL (TO-CL), whereas tetramyristoyl CL (TM-CL) was ineffective. None of the phosphatidylserine species tested gave significant activation. PAs containing C10 to C18 saturated acyl chains were not effective activators, and neither were oleoyl lyso PA, dilinoleoyl PA, or PA containing one oleoyl chain and either a palmitoyl or stearoyl chain. In contrast, dioleoyl PA stimulated seHAS ∼10-fold, to ∼20% of the activity observed with TO-CL. The tested acidic lipids such as PA and CL activated the enzyme most efficiently if they contained only oleic acid. Mixing experiments showed that the enzyme interacts preferentially with TO-CL in the presence of TM-CL. Similarly, seHAS incorporated into phosphotidylcholine-based liposomes showed increasing activity with increasing TO-CL, but not TM-CL, content. Inactivation of membrane-bound seHAS by solubilization with Nonidet P-40 was prevented by TO-CL, but not TM-CL. The pH dependence of seHAS in the presence of synthetic or naturally occurring CLs showed the same pattern of lipid preference between pH 6 and 10.5. Unexpectedly, HAS showed lipid-independent activity at pH 11.5. The results suggest that Class I HAS enzymes are lipid-dependent and that assembly of active seHAS-lipid complexes has high specificity for the phospholipid head group and the nature of the fatty acyl chains.

Expression and Purification of seHAS-E. coli SURE TM cells containing pKK223-3 plasmid encoding seHAS, with a C-terminal His 6 fusion, were grown at 30°C in Luria broth, HAS expression was induced, and membranes containing seHAS were prepared as described earlier (12). The membrane pellets were washed once with phosphate-buffered saline containing 1.3 M glycerol and protease inhibitors, sonicated briefly, aliquoted, recentrifuged at 100,000 ϫ g for 1 h, and stored at Ϫ80°C. The extraction buffer, membrane solubilization procedure, and affinity chromatography over Ni 2ϩ -nitrilotriacetic acid resin (Qiagen Inc.) were described previously (12). After loading DDM extract, the column was washed with 10 volumes of wash buffer, and seHAS was released using 3 volumes of elution buffer. No added phospholipids were present during any stage of the purification.
HAS Activity Assays-Activity of seHAS was determined in 100 l (final volume) of assay buffer with or without phospholipids at the concentrations indicated in the figures, as described previously (12,24,25). Purified seHAS (ϳ0.3 g) was added to initiate the enzyme reaction, and the mixtures were gently agitated in a Taitec MicroMixer, model E-36, at 30°C for 60 min, or for the indicated time. The reactions were terminated by the addition of 20 l of 12% (w/v) SDS and analyzed by descending paper chromatography using Whatman 3MM paper and elution with 1 M ammonium acetate, pH 5.5 (adjusted with glacial acetic acid), and 95% ethanol (7:13). The values are presented as the means Ϯ S.E. (n ϭ 3-4).
Phospholipid Preparation-Portions of stock chloroform or ethanol solutions of each phospholipid were dried under nitrogen, resuspended in 0.05% (w/v) DDM in distilled water, and added to assay buffer to give the desired final concentration. In experiments with combinations of TM-CL and TO-CL, the two lipids were reconstituted in DDM either separately and then mixed when added to assay buffer, or the chloroform solutions were mixed prior to evaporation and resuspension in DDM.
pH Dependence of seHAS Activity in the Presence of Different CLs-A modified assay buffer without sodium and potassium phosphate, UDP-sugars, or lipid was prepared at room temperature. Phosphate buffers ranging in pH from 4.5 to 12 were prepared using 500 mM solutions of phosphoric acid or mono-, di-, or tribasic potassium phosphate, which were mixed to achieve the desired final pH when diluted to 25 mM phosphate in modified assay buffer. The final pH values were measured using an Accumet model 15 pH meter. One of the four different CL preparations was then added to a final concentration of 2 mM. The purified seHAS was then added to start the reaction.
Preparation of seHAS Proteoliposomes-Recombinant seHAS purified in the absence of added lipid was reconstituted into proteoliposomes containing various proportions of PC and TO (or TM as indicated) using OG and Bio-Beads SM 2, as described by Rigaud and Levy (26). Phospholipids (50 mg) in chloroform were dried under vacuum and resuspended in 3 ml of diethyl ether. One ml of phosphate buffer was added, and the two-phase system was then sonicated on ice for 3 min at 3 W using a Sonicator Misonix 3000 with a micro tip. The organic solvent was evaporated at room temperature under reduced pressure (300 -400 mm Hg). During evaporation the suspension converts into a viscous gel, which on further evaporation again forms a suspension. At this point 2 ml of phosphate buffer was added, evaporation (ϳ700 mm Hg) continued for ϳ30 min to remove traces of solvent, and the liposome suspension was sequentially extruded through 0.4-and 0.2-m syringe filters to produce uniformly sized liposomes. Preformed liposomes were diluted in phosphate buffer to a final concentration of 2-3 mg/ml lipid, and a 2.2-fold weight excess of detergent was added from a 10% (w/v) solution of OG in phosphate buffer. The OG-phospholipid mixture was incubated for 2 h at room temperature, the desired amount of purified seHAS was added, and the seHAS-liposome mixture was incubated for 90 min at room temperature. OG was then removed by two treatments with wet Bio-Beads, at a 30:1 (w/w) ratio of beads:OG, for 1 h at room temperature with gentle mixing on a rotating wheel. The beads were allowed to settle at 1 ϫ g, the proteoliposome suspension was transferred and centrifuged at 100,000 ϫ g for 1 h at 4°C. The pellet was resuspended by pipetting up and down in phosphate buffer, and unincorporated protein was removed by recentrifugation. Final purified seHAS-proteoliposomes were resuspended in phosphate buffer, stored at 4°C, and used within 1 week.
General-Protein concentrations were determined by the method of Bradford (27) with the Coomassie protein assay reagent (Pierce), using bovine serum albumin as the standard. SDS-PAGE and Western analysis procedures were performed using the methods of Laemmli (28) and Burnette (29), respectively.

RESULTS
seHAS Activation by Synthetic Phospholipids-When seHAS is affinity-purified from DDM extracts of E. coli membranes in the absence of exogenous lipids, the enzyme displays very low and variable activity. The purified enzyme was greatly stimulated by mixtures of some, but not all, naturally occurring phospholipids ( Fig. 5 in Ref. 12). As noted previously, bovine CL activated seHAS, purified in the absence of added lipids, to a much greater extent than phosphatidic acid (PA), PC, phos-phatidyl ethanolamine, phosphatidyl glycerol, phosphatidylinositol, or phosphatidylserine (PS). PS was the only other lipid that stimulated seHAS activity significantly, to a level severalfold above background and ϳ25% of the activity with bovine CL. These results indicated a significant lipid dependence for the streptococcal enzyme and confirmed earlier radiation inactivation studies, in which CL was identified as a component of the active synthase (11).
To define better the fatty acyl requirements for HAS activity, we tested the abilities of well defined synthetic lipids to stimulate the activity of lipid-depleted, purified seHAS. The reference lipid in these comparisons was TO-CL, which was previously found to be a better activating lipid than crude bovine or E. coli CL preparations (24). We first tested a series of synthetic PA species (Fig. 1). The activity of purified seHAS was not stimulated by 2 mM of PA variants containing two identical fatty acyl groups ranging from C10 to C18 in length (Fig. 1A). Although the enzyme was several-fold more active in the presence of PA (16:0) than the other members of this series, this level of activity was only ϳ5% of that in the presence of TO-CL. Based on the activation of seHAS by CL-containing oleoyl groups, we expected PA (18:1) to be a good activating lipid, and it was by far the most effective PA variant tested (ϳ18% of the TO-CL-stimulated activity). PA species containing a single oleoyl chain (lyso) or with a second palmitoyl or stearyl fatty acyl chain were unable to activate seHAS (Fig. 1A). Fig. 1B shows the concentration dependence for activation of seHAS by dioleoyl PA and TO-CL and the lack of effect with dilinoleoyl PA or distearyl PA.
We then tested a series of PS variants with two identical saturated fatty acyl chains ranging from C12 to C18 (Fig. 2). Although in this experiment the mean activity of seHAS in the presence of dilauroyl PS was ϳ3-fold greater than in the presence of TM-CL, the results with this PS lipid were variable and not statistically different from any of the other variants tested except TO-CL. Dioleoyl PS was unable to stimulate seHAS activity (Fig. 2), in contrast to dioleoyl PA (Fig. 2). Lyso-oleoyl PS was also unable to activate the enzyme (Fig. 2). Although a naturally occurring PS mixture stimulated activity (12), none of the synthetic PS variants tested showed the ability to activate the enzyme. Presumably, other PS species that were not tested in this study are capable of activating HAS, or an activating contaminant (e.g. CL) may be present in some commercial PS preparations.
Among the commercially available synthetic PA and CL lipids tested, the two containing only oleic acid were consistently capable of activating purified seHAS (TO-CL Ͼ Ͼ dioleoyl PA). Dilinoleoyl PA and distearoyl PA showed inconsistent and low levels of stimulation, and some of this variability, as noted above, is likely due to the presence of variable amounts of endogenous tightly bound annular lipids still associated with seHAS (i.e. not removed by the purification procedure).
seHAS Preferentially Interacts with TO-CL Compared with TM-CL-If seHAS and multiple CL molecules form discrete, specific complexes, then the way in which lipid is presented to the lipid-depleted enzyme might affect the kinetics of forming active HAS-lipid complexes and affect the experimental results. To assess this possible complication, the abilities of the above lipids (as well as TM-CL, bovine CL, and E. coli CL) to activate seHAS were tested using two different protocols. In the standard assay, the enzyme is added last to assay buffer containing substrates and lipid. If enzyme and lipid require substantial time to interact to achieve seHAS activation, then a lag time would occur under these conditions; in this case, a preincubation of enzyme and lipid, prior to the addition of substrate, should decrease such a lag. When this was tested ( Fig. 3), in all but one case (with E. coli CL), seHAS activity was actually greater under the standard conditions, without a 30-min pretreatment. This result indicates that seHAS and phospholipids interact quickly, with no apparent kinetic lag, to form enzyme complexes.
An important feature of a lipid requirement for any enzyme is whether the enzyme interaction with preferred lipid is much more favored, and therefore more specific, than for other lipids that are not capable of activating the enzyme. Is the enzyme capable of binding many different types of lipids with relatively little difference in the energetics of binding each lipid, or is there great specificity (presumably indicating a greater K eq ) for the activating lipid? Since two synthetic CL species were available with dramatically different abilities to activate seHAS, we performed mixing experiments to test whether the purified enzyme was more readily able to interact with the activating TO-CL compared with the inactive TM-CL.
In the first experiment ( Fig. 4), seHAS was mixed with either 1 mM TM-CL or 1 mM TO-CL and increasing amounts of the second CL species. If TM-CL was fixed at 1 mM, then seHAS activity was initially very low as expected but increased sharply as the TO-CL concentration increased from 0.25 to 1 mM. TO-CL was able to activate seHAS in the presence of a greater or equal concentration of TM-CL. In contrast, when the fixed lipid was TO-CL, the addition of TM-CL, even up to 2 mM, did not decrease the ability of TO-CL to activate seHAS. These results indicate that seHAS interacts preferentially with TO-CL compared with TM-CL.    DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48

JOURNAL OF BIOLOGICAL CHEMISTRY 36545
In a second kinetic experiment (Fig. 5), we assessed the affects on the rate of HA synthesis of 1 mM TO-CL or TM-CL alone or as mixtures of the two CL species. The behavior of the enzyme was different depending on whether the two lipids were premixed and dried from organic solvent before aqueous suspension (Fig. 5A) or were mixed separately from aqueous suspensions (Fig. 5B). The activity of seHAS in the presence of either TM-CL or TO-CL alone was not affected by the method of lipid preparation (Fig. 5, solid circles and squares). When the lipids were premixed prior to suspension, the addition of 1 mM TM-CL had no effect on the rate of seHAS activity in the presence of 1 mM TO-CL (Fig. 5A, open circles), consistent with the results in Fig. 4. However, the enzyme was less active in this same lipid mixture when the lipids were presented to the pro-tein as separate aqueous suspensions (Fig. 5B, open circles). Interestingly, after a lag time of ϳ20 -25 min, the steady-state rate of HA synthesis in the presence of the two CL species was identical to that with TO-CL alone. These results indicate that when the two lipids are presented as a mixture from separate aqueous suspensions, seHAS interacts with each of the CL species (e.g. as micelles or liposomes) in a kinetically different way. The protein-lipid interactions are dynamic (i.e. kinetically reversible), and the interaction with TO-CL is probably more favorable thermodynamically than with TM-CL.
seHAS Shows the Same Order of CL Preference between pH 6 and 10.5-In a previous characterization of the purified seHAS, we noted an unusually broad pH range for enzyme activity, from pH 6 to pH 10.5-11.0 (24). Because this result was obtained in the presence of BCL, we wanted to determine whether the enzyme behaved any differently with the more effectively activating TO-CL or whether the inability of TM-CL FIGURE 5. The effect of mixtures of activating and inactivating synthetic CL on the kinetics of HA synthesis by purified seHAS. seHAS purified in the absence of lipid was incubated with 1 mM TM-CL (f) or TO-CL (F) alone or a combination of the two, each at 1 mM (E). Reactions at 30°C were initiated by the addition of seHAS to assay buffer containing the lipids and stopped at the indicated times as described under "Experimental Procedures." A, the lipids were premixed in chloroform, evaporated, reconstituted in DDM, and added to the assay buffer. B, the individual lipids were evaporated from chloroform separately and reconstituted in DDM, and then each was added to the assay buffer. to activate seHAS was pH-dependent (Fig. 6A). As expected, lipid-depleted purified seHAS had very low activity with TM-CL from pH 4.5 to 10.5, although a very small peak of activity was observed between pH 7.5 and 8.5. The bovine, E. coli, and TO-CL preparations showed the same overall pattern of pH dependence, and their order of effectiveness did not change with pH. In each case, an interesting multi-component profile was observed between pH 6 and 10.5 showing two distinct peaks with maximum activities at pH ϳ7.5-8 and pH 9 -10. This result was observed in three independent experiments, although the decreased activity observed in the region at pH 8.5 was variable. In all cases, the enzyme was essentially inactive at pH 10.5, but another small peak of activity was found between pH 11 and 12, an unusually high pH for retention of enzyme activity. The results show that the lipid requirement for seHAS activity and the preference for TO-CL are still present and consistent over a broad range between pH 6 and pH ϳ10.5.
seHAS Becomes Lipid-independent above pH 11-The additional peak of enzyme activity seen in Fig. 6A above pH 10.5-11 was unexpected and seemed unusual. This activity was reproducible, but on further examination, we found that phospholipids were not required at this very high pH for seHAS to make HA (Fig. 6B). Treatment with Streptomyces hyaluronidase verified that the enzyme synthesized normal HA products (not shown). In multiple experiments, the level of HAS activity at pH 11.5 without lipid ranged up to ϳ10% of that at pH 7.6 with BCL.
Lipid Dependence of seHAS Incorporated into Proteoliposomes-We next wanted to characterize the lipid dependence of seHAS incorporated into liposomes of defined size and composition. Proteoliposomes were prepared by disrupting preformed PCbased liposomes with OG, adding purified seHAS, and then removing the detergent. Transmission electron microscopy ( Fig. 7A) showed single intact liposomes of relatively uniform size (ϳ50 nm). To determine the efficiency of seHAS incorporation into liposomes, the protein remaining in the supernatant after ultracentrifugation was precipitated using trichloroacetic acid and analyzed by SDS-PAGE and Western blotting (Fig.  7B). Only trace amounts of seHAS were detected in the supernatant after centrifugation, demonstrating that essentially complete incorporation of purified recombinant seHAS protein into liposomes was achieved. Consistent with the above and previous (12) results, seHAS had little or no activity in proteoliposomes containing only PC (not shown).
We then tested the effect of different total lipid-to-protein molar ratios on seHAS activity in proteoliposomes containing TO-CL. When liposomes were prepared with a constant PC:TO-CL molar ratio (4:1) and different amounts of purified seHAS, enzyme activity increased with increasing protein-to-lipid ratios (Fig. 7C).
To assess whether seHAS showed the same specific activation by TO-CL in proteoliposomes as in the above experiments, FIGURE 7. Incorporation of seHAS into proteoliposomes. A, seHAS proteoliposomes were prepared with PC and TO-CL (1:1 molar ratio) and purified recombinant seHAS (a protein to lipid molar ratio of 1:1000) as described under "Experimental Procedures." Aliquots (10 l) of the final liposome suspension were attached to EM grids in a humidified chamber and stained with 2% phosphotungstic acid, pH 6.24 for 10 s and imaged using a Hitachi H7600 electron microscope and digital camera (AMT 2K ϫ 2K). The bar represents 100 nm. B, seHAS proteoliposomes were prepared as above except that the molar ratio of purified seHAS to lipid was either 1:550 (lanes 1 and 2) or 1:1100 (lanes 3 and 4). After centrifugation at 100,000 ϫ g, the liposome pellets ( lanes  2 and 4) and supernatants (lanes 1 and 3) were subjected to precipitation with 10 % (w/v) trichloroacetic acid (in the presence of 0.2 mg/ml bovine serum albumin as carrier), followed by SDS-PAGE and electro-transfer to nitrocellulose membranes. The membranes were stained for 20 min with fluorescent His 6 tag stain, and the image was taken under UV light on an Alpha Innotech imager. C, purified recombinant seHAS was reconstituted into proteoliposomes with a PC:TO molar ratio of 4:1 and the indicated protein to total lipid ratio expressed as mol % of enzyme. seHAS activity was determined in triplicate and is presented as the mean dpm Ϯ S.E. ( 14 C-GlcUA incorporated into HA)/20 l of liposomes/h. DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 we varied the TO-CL:PC molar ratio, while keeping the lipidto-protein molar ratio constant. As a control to assess the potential activity of the enzyme, we assayed the activity of seHAS in proteoliposomes in the presence or absence of 2 mM BCL. seHAS showed much higher activity in liposomes produced using a 1:1 molar ratio rather than a 1:2 ratio of TO-CL:PC (Fig. 8A). Inclusion of exogenous BCL had a strong stimulatory effect on seHAS activity if the liposome TO-CL content was 33 mol % but did not further increase the enzyme activity when the TO-CL content was 50 mol %. This result indicates that proteoliposomes with a PC-to-TO-CL molar ratio of 1:1 provide sufficient opportunity for seHAS to be activated optimally by TO-CL (or conversely, to be not inhibited by interaction with PC). This finding was valid for proteoliposome preparations with a range of seHAS-to-lipid molar ratios (Fig. 8B). If purified seHAS was reconstituted in a similar way using proteoliposomes containing TM-CL, rather than TO-CL, the enzyme showed little or no activity (not shown).

Phospholipid Dependence and Liposome Reconstitution of HAS
TO-CL Rescues the Activity of Detergent-solubilized seHAS-Finally, we examined the ability of TM-CL or TO-CL to prevent the apparently irreversible inactivation of HAS that typically occurs during extraction with nonionic detergents (Fig. 9).
Based on a titration experiment (not shown) to assess the sensitivity of membrane-bound HAS to detergent, we found that activity was completely inhibited by Ն0.01% Nonidet P-40. We then assessed the effect of increasing concentrations of TO-CL or TM-CL on inactivation of seHAS with 0.01% Nonidet P-40. HAS inactivation was completely prevented by including 0.25 mM TO-CL during treatment with detergent, whereas 0.25 mM TM-CL had only a minor effect. As the concentration of TO-CL increased to 2 mM, there was actually a further activation of seHAS to a level that was ϳ151% of the original membrane activity. In contrast, in the presence of 2 mM TM-CL, the enzyme showed only ϳ15% of its initial membrane activity. Thus, membrane-bound seHAS was solubilized with excellent retention of activity, only when the best activating phospholipid was present during extraction.

DISCUSSION
This is the first report in which a HAS has been purified and reconstituted in an active form into liposomes. The characteristics of purified seHAS reconstituted in proteoliposomes were similar to and very consistent with the phospholipid dependence and preference for TO-CL, rather than TM-CL, demonstrated when the enzyme was simply mixed with the lipids in suspension. Another major finding is that membrane-bound seHAS could be solubilized without inactivating the enzyme, if the best activating phospholipid was present during extraction.
Streptococcal HASs have been amenable to detailed study because these enzymes can be solubilized by the mild detergent DDM with retention of activity and purified (12,24). However, purification and study of the eukaryotic enzymes has been more difficult because these HASs are generally inactivated in the presence of any detergent, including DDM (3). This lability of  HAS to detergent solubilization has hindered progress in the field, because no one has been able to study an active purified mammalian HAS. The only exception is a study by Yoshida et al. (30) in which mouse FLAG-HAS1 was extracted from membranes with 3-[(3-cholamidopropyl) dimethyl ammonio]-1-propanesulfonic acid, purified by affinity chromatography, and reconstituted in an active form by layering onto buffer containing a high concentration of the same detergent. This twophase system apparently mimics features of the lipid bilayer required by HAS.
Kimata and co-workers 3 recently purified recombinant mouse HAS2 expressed in insect cells using a baculovirus expression vector. Significantly, they found that the best detergent was DDM, and DDM-solubilized HAS2 was activated by subsequent addition of one particular phospholipid. Thus, it seems likely that all of the HAS family members are lipid-dependent, as reported originally for the streptococcal enzymes (12). The single exception, the Pasteuella enzyme, is structurally and functionally different from all other HAS proteins (2,31).
We note that the rescue of HAS activity by TO-CL after Nonidet P-40 treatment (Fig. 9) is the first successful solubilization of a Class I HAS using a nonionic detergent (other than DDM) under conditions that retain or rescue activity of HAS suitable for enzyme purification. These results support the conclusion that detergent inactivation of HAS is due to removal of activating lipids associated with the enzyme. If these lipids are replaced, as in the case of the lipid-depleted purified seHAS, or not removed completely, then HAS can remain active. Taking advantage of this strategy should allow other mammalian HASs to be purified and characterized.
Lipids associated with membrane proteins are found in three general types of interactions: tightly bound in an annular shell, nonannular in cavities of the protein surface, and as integral protein lipids that can be within and parallel to the bilayer (32). In preliminary mass spectrometry experiments (not shown), we detected CL m/z signature profiles in seHAS that had been DDM-solubilized and purified without added lipid. These residual CL molecules likely represent tightly bound annular shell lipids, some of which may be required for correct folding of the protein. Experiments are in progress to determine the fatty acyl composition of the endogenous annular CL associated with the purified lipid-depleted seHAS.
Some of the proteins and enzymes known to require CL for activity or stability include the sodium/proton antiporter, cytochrome bc 1 complex, ATP synthase, bacteriorhodopsin, and the EmrE multidrug transporter. In general, the membrane proteins known to be CL-dependent are channels or transporters that use multiple membrane domains to constitute a pore within the bilayer. An interesting exception is MurG, an E. coli glycosyltransferase involved in peptidoglycan biosynthesis (14). Overexpression of this peripheral membrane protein results in elevated CL content in cell membranes and association of the protein with unusual intracellular vesicles.
The distinct bimodal pH optima for seHAS are unusual for a single enzyme. HAS, however, is actually a multifunctional enzyme with four distinct donor/acceptor-binding sites, a hyaluronyl-UDP translocation activity, and two separate glycosyltransferase activities (31). Unlike the vast majority of known glycosyltransferases, HASs that elongate at the reducing end use UDP-GlcUA and UDP-GlcNAc as acceptors rather than donors. The donors in these two reactions are the hyaluronyl-UDP products with either GlcNAc or GlcUA attached to UDP at the reducing end. Further studies are needed to determine whether the bimodal pH optima reflect distinct partial reactions catalyzed by HAS. Despite the phospholipid requirement for activity over a broad pH range, HAS showed unexpected activity at pH 11-12 in the absence of any added lipid. This high pH is well above the pK a values of Glu, Asp, Lys, Arg, and His residues in HAS and also likely near or above the pK a values for Ser, Thr, Cys, and Tyr. Interestingly, this pH range is also likely to deprotonate many of the -OH groups in GlcNAc and GlcUA residues in HA and, in HA-UDP, GlcNAc-UDP and GlcUA-UDP. Thus, in this lipid-independent mode at high pH, the enzyme and substrates are all expected to be highly negatively charged. The enzyme-activating phospholipids (e.g. CL) are also negatively charged, which is consistent with the possibility that negative charge is important for HAS activity.
The finding that HAS activity is lipid-independent at high pH levels is not only surprising, but it could also be very significant for new approaches to characterize the structure, function, and mechanism of HAS. The ability to be active in the absence of lipid means that it might be feasible to study purified streptoccal HASs using biophysical techniques typically employed to study protein structure (e.g. circular dichroism, fluorescence spectroscopy, ultracentrifugation, or calorimetry). Such techniques could not be easily used before, because of the presence of large amounts of lipid needed for activity.
Our results indicate there is something special about 18:1(⌬9) oleoyl chains for seHAS activation, because the only lipids able to activate the enzyme contained this fatty acid. A possible explanation is that the fluidity of this acyl chain provides advantages for sealing the mobile bilayer-protein interface during domain movement of HAS as it polymerizes HA. It will be interesting, although difficult, to determine whether the fatty acyl groups are any different for the annular, nonannular, or integral lipids associated with native protein. Mixing experiment results (Fig. 5) indicate that seHAS interacts with either TM-CL or TO-CL in a dynamic, reversible way. If the enzyme first associates with TM-CL, then it takes time for it to dissociate from TM-CL and bind sufficient TO-CL molecules to become active. This result indicates that seHAS interacts with TO-CL in a thermodynamically more favorable way than with TM-CL.
None of the known CL-dependent proteins is reported to require as many CL molecules associated with the protein as HAS, although few investigations have been able to ascertain the molar ratios within protein-lipid complexes. Radiation inactivation studies of Streptococcus pyogenes and Streptococcus equisimilis membranes containing HAS as well as E. coli membranes containing each of the two recombinant streptococcal HASs revealed that the active unit is a HAS monomer associated with an additional mass of ϳ23 kDa (11). This addi-3 A. Murakawa and K. Kimata, personal communication. DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 tional mass was identified as CL based on the restoration of a target size equivalent to a HAS monomer when CL was added to DDM-solubilized membranes that had been irradiated. Independent studies also demonstrated the lipid requirement of the purified streptococcal HASs and their preference for CL (12). A later study (33), also using radiation inactivation, found that the target mass of the active Xenopus laevis HAS1 is a protein monomer with an additional ϳ23 kDa of unknown origin. Although this result is strikingly similar to the streptococcal HAS results, this group was not able to verify a lipid requirement for the enzyme. If the molar mass of sodium TO-CL is similar to the mass of endogenous CL associated with HAS, then 23 kDa corresponds to 15.3 CL molecules/HAS. We assume a slightly lower CL mass and use a value of 16 required CL molecules/ HAS monomer. Of course this number is an estimate and not certain, because the masses and distribution of endogenous CLs or other lipids associated with HAS are unknown.

Phospholipid Dependence and Liposome Reconstitution of HAS
Regardless of the exact number, HAS requires interactions with many phospholipid molecules, as illustrated schematically in Fig. 10, which might be explained by a proposed mechanism in which HA synthesis and HA translocation are coupled. Genetic and biochemical evidence indicate that only the streptococcal HAS protein is needed for HA synthesis using the two UDP-sugar precursors (34 -36). These cytoplasmic precursors are polymerized into a UDP-HA chain by the enzyme, which continuously translocates the growing HA to the cell exterior (1,10). The substrates are intracellular, and the HA product is extracellular. Because the HAS proteins have 6 -8 transmembrane or membrane-associated domains, we suggested (25) that the protein is complexed with many CL molecules to create an intra-protein pore through which the chain is moved after the transfer of HA from UDP-HA to a new UDP-sugar (or disac-charide unit) at the reducing end. The very nature of such a translocation function for HAS likely requires a mechanism involving the movement of multiple protein domains.
In support of this intra-protein pore model, we recently found that two polar residues within MD2 and MD4, which are conserved within the HAS family, interact directly with each other or with HA during synthesis (37). Fig. 10 illustrates the proximity of MD2 and MD4 in the context of the proposed HAS-lipid complex. In the original model (25), we referred to membrane-associated domains 1 and 2 and transmembrane domains 1-4 differently, with a different numbering system and assumed that their order of interaction in the protein was sequential. Now we know that MD2 and MD4 are partners (37). In fact, all the HAS MDs are likely close together, and specific MDs interact with each other. The organization of HAS domains might be similar to that of many transporter or channel proteins whose folding and topology within the membrane create an intra-protein space important for moving the ligand or ion across the membrane bilayer. The molecular mechanism that appears most consistent with the topological and structural features of these HAS enzymes is the formation of a pore through which the HA product moves.
Reconstitution of purified HAS into defined liposomes will allow us to address some specific questions about the above pore model of HAS and the vectoral nature of HA synthesis. Because proteoliposomes have an internal lumen, transport studies can be performed, and inside and external components can be separated. The method used to prepare proteoliposomes (26) is designed to give an orientation of intracellular domains out (i.e. facing the medium rather than the lumen), although this has yet to be confirmed for seHAS. If this is the orientation of most incorporated seHAS molecules, then HA synthesis might proceed by use of external substrates and translocation of HA product to the proteoliposome interior. Lumenal overaccumulation of HA might result in either disruption of proteoliposomes (and continued HA synthesis) or a time-dependent inhibition of HA synthesis if the proteoliposomes remained intact. Studies are in progress to answer these and other similar questions.
Based on many of these above considerations, we proposed the Pendulum Hypothesis model (38) for polysaccharide synthesis. This model has several conceptually similar variants. In one variant of the Pendulum Hypothesis, the HASs have two functional domains that move, and each domain contains one of the two UDP-sugar-binding sites, a binding site for one of the two donor HA-UDPs, and performs one of the two hyaluronyltransferase functions. Each functional domain can move into different positions in which its transferase function is either inactive or active, and only one domain can be active at a time. These activities are reciprocal, and when one domain is able to bind a UDP-sugar acceptor, the other is active as a transferase. As the UDP-HA chain is elongated, the alternating domain movements create the translocation mechanism that extrudes the HA chain through the protein and across the membrane. Large scale movement of protein domains required for such a translocation process would likely require many associated phospolipid molecules to fill in gaps on the protein surface and keep a seal at the lipid bilayer and the enzyme interface. HA and MD4-MD5 are likely to be associated, because they are connected by very short extracellular loops (10). We recently showed that a pair of polar residues within MD2 (Lys 48 ) and MD4 (Glu 327 ) mediate the interaction of these domains within the membrane (37). Based on radiation inactivation studies (11), an active synthase is a complex of one HAS monomer and ϳ16 CL molecules. The large number of CLs are schematically shown (solid ovals) associated with multiple MDs to create an intramolecular pore through which the growing HA chain (hexagon) can be translocated across the membrane. synthesis would not be well tolerated if water, ions, or small molecules were able to enter or leave the cell because of leakiness. Studies are in progress to test this model and the role of CL in activating HAS.