Lipoteichoic acid polymer length is determined by competition between free starter units

Significance

Carbohydrate polymers are found in all domains of life, and their biological activities can change depending on their length. Gram-positive bacteria decorate their cell membrane with lipoteichoic acids (LTAs), which are important for essential processes such as cell division, osmotic stability, and antimicrobial resistance. The length of LTA polymers affects fitness, pathogenicity, and susceptibility to beta-lactams. Using in vitro reconstitution, we have identified a unique mechanism that controls polymer length. In addition to providing fundamental mechanistic insight into carbohydrate polymerases, our results provide a platform for studying this conserved cell envelope enzyme and should facilitate the development of antibiotics targeting it.

Abstract

Carbohydrate polymers exhibit incredible chemical and structural diversity, yet are produced by polymerases without a template to guide length and composition. As the length of carbohydrate polymers is critical for their biological functions, understanding the mechanisms that determine polymer length is an important area of investigation. Most Gram-positive bacteria produce anionic glycopolymers called lipoteichoic acids (LTA) that are synthesized by lipoteichoic acid synthase (LtaS) on a diglucosyl-diacylglycerol (Glc₂DAG) starter unit embedded in the extracellular leaflet of the cell membrane. LtaS can use phosphatidylglycerol (PG) as an alternative starter unit, but PG-anchored LTA polymers are significantly longer, and cells that make these abnormally long polymers exhibit major defects in cell growth and division. To determine how LTA polymer length is controlled, we reconstituted Staphylococcus aureus LtaS in vitro. We show that polymer length is an intrinsic property of LtaS that is directly regulated by the identity and concentration of lipid starter units. Polymerization is processive, and the overall reaction rate is substantially faster for the preferred Glc₂DAG starter unit, yet the use of Glc₂DAG leads to shorter polymers. We propose a simple mechanism to explain this surprising result: free starter units terminate polymerization by displacing the lipid anchor of the growing polymer from its binding site on the enzyme. Because LtaS is conserved across most Gram-positive bacteria and is important for survival, this reconstituted system should be useful for characterizing inhibitors of this key cell envelope enzyme.

All cell surfaces are rich in carbohydrate polymers that act as structural components, scaffolds for other molecules, and participants in signaling processes (1). The biological functions of a carbohydrate polymer are often greatly affected by its length. For example, depending on molecular weight, hyaluronic acid polymers can promote cell migration, differentiation, and inflammation or can inhibit these processes (2, 3). Similarly, the number of repeat units in bacterial O-antigen has a profound effect on complement activation and host cell uptake (4, 5). Unlike protein and nucleic acid polymers, which are assembled on a template that determines both length and composition, carbohydrate polymers are assembled without the use of a template. Template-independent length regulation is not as precise as template-directed polymerization, but physiological lengths of carbohydrate polymers typically fall into a defined range that is important for function (6). How different polymerases achieve length control is a fundamental question in the field.

Several mechanisms for carbohydrate polymer length determination have been described. Some polymerases include a “molecular ruler” domain that measures the polymer against a portion of the enzyme (7), some use a dedicated “termination enzyme” to control length (8), and others rely on repeat unit concentration to control polymerization (9). These mechanisms are not mutually exclusive and can act together to control length (10, 11). The degree to which a polymerase is processive also influences product length. Processivity, a fundamental property of polymerases, refers to the number of elongation steps that occur without release of the growing polymer (12). A polymerase may be partially processive, in that more than one monomer addition occurs while the polymer is bound to the enzyme, but the polymer can be released and then rebind to continue elongation. A polymerase may also act in a distributive manner, where the growing polymer is released after each round of monomer addition. While some general mechanisms and aspects of length control for carbohydrate polymerases are known, here we describe a previously unknown mechanism for length regulation of a common type of lipoteichoic acid (LTA), a cell surface polymer that is crucially important to the physiology of most Gram-positive bacteria (13, 14).

In the Gram-positive pathogen Staphylococcus aureus (Sa), LTA is a membrane-anchored poly(glycerol-phosphate) polymer involved in virulence (15–19), regulation of cell size and division (20–23), and osmotic stability (24, 25) (Fig. 1A). Sa LTA is assembled by the conserved lipoteichoic acid synthase (LtaS) on the cell surface using glucose(β1,6)-glucose(β1,3)-diacylglycerol (Glc₂DAG) as the membrane-anchored “starter unit” (20, 26). The polymer elongates in a process that involves the repeated transfer of phosphoglycerol units from phosphatidylglycerol (PG) to a catalytic threonine in LtaS (T300) and then to the tip of the growing polymer (Fig. 1B) (27–29). Repeat units may be modified by D-alanyl esters or, less commonly, GlcNAc moieties (24, 30). Because LTA is so important for Sa survival (13, 14, 21, 22), LtaS is a proposed target for antibiotics, and understanding its behavior may facilitate inhibitor development.

Fig. 1.

LTA is a lipid-anchored polymer assembled from Glc₂DAG and PG on the bacterial cell surface. (A) Chemical structure of LTA from Sa. Phosphoglycerol repeat units may be modified with D-alanine esters or GlcNAc moieties. (B) Mechanism of LTA synthesis by LtaS. Phosphoglycerol units are transferred from PG to residue T300 to form a covalent intermediate, releasing DAG. Phosphoglycerol is then transferred to a Glc₂DAG starter unit to form GroP-Glc₂DAG. Additional repeat units are added to the glycerol tip of the polymer. (C) In Sa, PgcA and GtaB synthesize UDP-glucose from glucose-6-phosphate. UgtP uses UDP-glucose and DAG to make Glc₂DAG. LtaA exports Glc₂DAG to the cell surface. LtaS transfers phosphoglycerol units derived from PG to T300, releasing DAG for recycling. (D) Anti-LTA Western blot of Sa RN4220 wild-type (wt) or ΔugtP lysates. ΔugtP mutants lack Glc₂DAG, and LTA is instead polymerized directly on PG (20).

Glc₂DAG, the starter unit for LTA polymerization, is biosynthesized on the cytoplasmic leaflet of the membrane by the sequential action of three enzymes: the phosphoglucose mutase PgcA, the UTP-glucose-1-phosphate uridylyltransferase GtaB, and the diacylglycerol β-glucosyltransferase UgtP (also called YpfP) (15, 20). Glc₂DAG is exported to the cell surface by the flippase LtaA (Fig. 1C) (15). An interesting feature of LtaS is that it can use PG as an alternative starter unit if Glc₂DAG synthesis or export is blocked (20). However, polymers formed on this alternative starter unit (PG-LTA) are significantly longer than polymers formed on Glc₂DAG (Glc₂DAG-LTA, Fig. 1D) (15, 23), and cells that make these longer polymers have cell division defects (20, 23), are much less virulent (15, 16), and are more sensitive to beta-lactam antibiotics and other cell envelope stresses (23). Whether the shorter polymers assembled on Glc₂DAG reflect the intrinsic behavior of LtaS or the action of other cellular factors is an important question that cannot be definitively answered with genetic approaches.

Here we used in vitro reconstitution to test whether the identity of the LTA membrane anchor determines the length of the polymers that LtaS synthesizes. We show that the length differences observed between wild-type and mutant cells lacking Glc₂DAG are recapitulated in a proteoliposome system that contains only purified LtaS, PG, and either Glc₂DAG or an alternative anchor. Based on our studies, we propose a model for how polymer length can be controlled in polymerases that operate without a template.

Results

Purified UgtP Produces Glc₁DAG and Glc₂DAG in a Reconstituted System.

In Sa, the glycolipid anchor Glc₂DAG is synthesized by UgtP, which attaches both glucose units (20, 26). We purified UgtP and incubated it with diacylglycerol (DAG) and excess UDP-glucose. Lipids were extracted from the reaction and separated by thin-layer chromatography (TLC) (Fig. 2 A and B). Glc₂DAG was the major product, but a small amount of glucose(β1,3)-diacylglycerol (Glc₁DAG) was also produced, consistent with observations from previous experiments in which membrane fractions were used to test UgtP activity (20, 26). Glycolipids extracted from the UgtP reaction were purified by preparative-TLC, and their identities were confirmed by electrospray ionization mass spectrometry (SI Appendix, Fig. S1). Quantities were estimated by staining and comparing intensities with known galactose(β1,6)-galactose(β1,3)-diacylglycerol (Gal₂DAG) and glucose(α1,3)-diacylglycerol (Glc₁αDAG) standards (Fig. 2A).

Fig. 2.

Glc₂DAG is the only glycolipid that can serve as a starter unit for LTA polymerization, and its use reduces LTA polymer length. (A) Chemical structures of glycolipid anchors tested. (B) UgtP was reconstituted in a purified system and produces Glc₁DAG and Glc₂DAG. Lipids extracted from reactions were separated by TLC. S, standard lane with Gal₂DAG and Glc₁αDAG standards. (C) LtaS proteoliposome schematic. Proteoliposomes were formulated with PE, PG, CL, and an optional glycolipid (Glc₂DAG, Glc₁DAG, Gal₂DAG, or Glc₁αDAG), and reactions were initiated with Mn²⁺. (D) Polyacrylamide gel of LTA extracted from proteoliposome reactions and detected with an Alcian blue/silver stain. The amount of material loaded was adjusted to provide similar staining intensities and is shown at the bottom of the gel. (E) Zoomed-in region of D at the red bracket. Blue dots highlight the offset banding pattern of Glc₂DAG-anchored LTA, and orange dots highlight the pattern of PG-anchored LTA. (F) [¹⁴C]-labeled Glc₁DAG and Glc₂DAG were synthesized with UgtP and incorporated into proteoliposomes. Radiolabeled starting material was separated from reaction products by cellulose paper strip chromatography (SI Appendix, Fig. S4). Averages and SDs from three independent reactions are shown. (G) Completed reactions with radiolabeled glycolipids were separated by TLC and detected by autoradiography.

The Lipid Anchor Determines LTA Length.

To assess if and how the lipid anchor directly affects LTA length, we assembled LtaS proteoliposomes containing 69 mol% phosphatidylethanolamine (PE), 20 mol% PG, 10 mol% cardiolipin (CL), and 1 mol% of PE, Glc₂DAG, Glc₁DAG, Gal₂DAG, or Glc₁αDAG (Fig. 2C). Reactions were initiated by adding Mn²⁺, an essential cofactor of LtaS (31, 32), and quenched after 5 h. To visualize polymer length, LTA was extracted from the reactions and resolved on an Alcian blue/silver-stained polyacrylamide gel (Fig. 2D and SI Appendix, Fig. S2A). Reactions without Glc₂DAG formed substantially less product than those with Glc₂DAG (see below); therefore, it was necessary to load LTA from higher reaction volumes to achieve comparable staining intensities. We found that proteoliposome reactions containing Glc₂DAG resulted in polymers >20 repeat units shorter than proteoliposomes lacking this glycolipid (SI Appendix, Fig. S2B). All other glycolipids tested, including Glc₁DAG, resulted in LTA with a length distribution indistinguishable from LTA in samples containing only PG as a starter unit.

We more closely examined the banding pattern of the polymer for clues to possible structural differences that correlated with the length differences. Direct structural analysis was not possible owing to the limited amounts of material produced in the reconstituted system; however, a densitometry scan revealed altered periodicity of individual polymer bands from reactions with Glc₂DAG compared with the bands from the other proteoliposome reactions, all of which had the same mobility throughout the gel as bands from reactions containing only PG (SI Appendix, Fig. S2B). The periodic mobility differences between polymers formed from Glc₂DAG proteoliposomes and other proteoliposomes are shown in the expanded region of the gel in Fig. 2E and SI Appendix, Fig. S2C. We found that products from in vitro reactions with Glc₂DAG had the same mobility as LTA isolated from wild-type Sa, while bands from proteoliposome reactions that produced PG-anchored LTA had spacing that matched LTA isolated from ΔugtP mutants lacking Glc₂DAG (SI Appendix, Fig. S3 A and B). These results suggest that Glc₂DAG was incorporated into the LTA polymers, but the other glycolipids were not.

To confirm the incorporation of Glc₂DAG into LTA, we synthesized radiolabeled Glc₁DAG and Glc₂DAG using UgtP and UDP-[¹⁴C]glucose. Radiolabeled glycolipids were incorporated into proteoliposomes, and reactions were monitored by TLC and paper strip chromatography, in which glycolipid starting material migrates up the strip while products remain on the baseline (SI Appendix, Fig. S4). The addition of Mn²⁺ to the proteoliposomes to initiate polymerization resulted in a decreased signal for [¹⁴C]-Glc₂DAG and a concomitant increase in baseline material over time, showing that the radiolabeled starting material was incorporated into polymer (Fig. 2 F and G). In contrast, [¹⁴C]-Glc₁DAG was not consumed, consistent with cellular evidence indicating that Glc₁DAG is not a substrate of LtaS (15). Taken together, our findings show that only PG and Glc₂DAG are competent starter units for LTA. Because the relative lengths of the polymers formed using Glc₂DAG or PG as starter units recapitulate those in wild-type and ΔugtP cells, we have concluded that LtaS has an intrinsic mechanism for length regulation that depends on the identity of the lipid anchor.

Initiation Is the Rate-Limiting Step of LtaS Polymerization of LTA.

To investigate the mechanism of LTA polymerization, we used an anti-LTA antibody to monitor polymer formation as a function of time in reactions containing Glc₂DAG as the lipid anchor and PG to supply phosphoglycerol units, or only PG. Reactions containing Glc₂DAG had a significantly faster overall rate and produced more than five times as much product (Fig. 3A). The faster rate was not due to glycolipid-mediated alterations in the physical properties of the proteoliposomes, as the time courses of reactions with Gal₂DAG or Glc₁αDAG closely matched those of reactions without glycolipid (SI Appendix, Fig. S5). Reactions with Glc₁DAG proceeded more slowly than other reactions, suggesting that this substrate is modestly inhibitory (SI Appendix, Fig. S5). Notably, the progress curve for the Glc₂DAG reactions exhibited a lag phase of approximately 2 h before appreciable amounts of polymer were detected using the anti-LTA antibody (Fig. 3A).

Fig. 3.

LTA polymerization on Glc₂DAG exhibits a lag phase in product formation that is shortened by using a primed substrate, GroP-Glc₂DAG. (A) Proteoliposome reactions contained no glycolipid or 1 mol% of UgtP-synthesized Glc₂DAG, Sa Glc₂DAG, or Sa GroP-Glc₂DAG. LTA formation was monitored by dot blot with an anti-LTA antibody and normalized to the final amount of product in the Glc₂DAG sample (green triangles; set to 100). The averages and SDs of four independent replicates from at least two separate proteoliposome preparations are shown. (B) Model for priming by LtaS. Phosphoglycerol units are transferred from an LtaS covalent intermediate to the starter unit. Short oligomers freely dissociate until a primer of sufficient length is formed, after which more rapid polymerization occurs. The key is as in Fig. 1.

A lag phase in a polymerization reaction is often attributed to a “priming” step in which short oligomers are slowly formed before a rapid extension phase (33–37). We did not observe a lag phase in product formation in the progress curves obtained using the paper strip assay (Fig. 2F); however, this discrepancy is easily reconciled if products retained at the baseline of the paper strip include short primers that would not be detected by the anti-LTA antibody. To assess whether the observed lag phase could be due to a priming step, we sought to determine whether adding primers to the proteoliposomes accelerated polymerization. Although Sa contains only a single LTA synthase, some other Gram-positive organisms encode both an LTA synthase and a homologous enzyme that acts as a dedicated LTA primase to add a single phosphoglycerol unit to the glycolipid anchor (38, 39). We made use of a previously engineered Sa strain (38) containing an inducible copy of the Bacillus subtilis LTA primase yvgJ to isolate both Glc₂DAG and glycerol-phosphate-diglucosyl-diacylglycerol (GroP-Glc₂DAG) (SI Appendix, Figs. S6 and S7). The purified Sa Glc₂DAG and Sa GroP-Glc₂DAG substrates were tested separately in proteoliposome reactions. The reactions with Sa Glc₂DAG were first compared with reactions with Glc₂DAG synthesized in vitro to control for differences in acyl chain composition because Sa makes only saturated lipids, of which a substantial fraction are branched (40), whereas the synthetic material contains unbranched, unsaturated acyl chains. The progress curves for proteoliposome reactions containing Sa Glc₂DAG closely resembled those with synthetic Glc₂DAG, showing that acyl chain differences do not affect the reaction (Fig. 3A). However, the lag phase for reactions containing Sa GroP-Glc₂DAG, the primed substrate, was substantially shorter than that for reactions with either Glc₂DAG substrate, implying that GroP-Glc₂DAG is a competent substrate for LtaS (Fig. 3A). We also noted that the amount of product formed in the primed reactions was greater in reactions with this substrate (Fig. 3A). Because we observed increased optical density of the reactions after several hours, irrespective of the proteoliposome lipid composition, we attributed the differences in overall product formation for reactions with different substrates to destabilization/aggregation of the proteoliposomes over long reaction times. Nevertheless, the progress curves over the first few hours for unprimed and primed substrates support a model in which a slow priming phase precedes a more rapid elongation phase. Because the exogenous addition of primed substrate accelerates the reaction, we infer that primed substrates that form slowly in the reaction with Glc₂DAG can reversibly dissociate and rebind (Fig. 3B). We also suspect that priming requires addition of more than one phosphoglycerol unit, because a short lag phase is still observed for reactions with the singly primed substrate.

LtaS Is Processive and Does Not Rebind Released Polymers.

Some polymerases are fully processive and, once elongation begins, catalyze multiple rounds of extension before releasing polymeric product; other polymerases release and rebind the growing polymer between rounds of elongation (12). One method to assess whether an enzyme is processive is to track the distribution of polymer lengths over time (41). We set up time courses using either Glc₂DAG or PG lipid anchors and tracked the product lengths by polyacrylamide gel electrophoresis (PAGE) to assess the processivity of LtaS. With either starter unit, the polymer lengths remained constant over time even as the abundance of individual lengths increased (SI Appendix, Figs. S8 and S9). This constant distribution of product lengths over time implies that once a polymer is released, it cannot rebind for further extension. Because product lengths increase over time for distributive or partially processive enzymes where products are able to rebind, we infer that LTA polymerization is fully processive once the extension phase of the reaction begins.

Starter Unit Concentration Determines Polymer Length and Identity of the Lipid Anchor.

We next asked whether the concentrations of Glc₂DAG and PG would affect polymer product lengths. Therefore, we formulated LtaS proteoliposomes with a fixed concentration of PG (20 mol%) and varied amounts of Glc₂DAG (0.25 to 4 mol%), keeping the total amount of lipid constant by adjusting the amount of the major lipid component, PE, and monitored product length. We found that LTA length decreased as the concentration of Glc₂DAG increased (Fig. 4A and SI Appendix, Fig. S10A). At very low concentrations of Glc₂DAG, product lengths were similar to those observed in the absence of Glc₂DAG and spacing of polymer bands was consistent with predominant use of PG as a starter unit (SI Appendix, Fig. S11 A and B). At intermediate concentrations, individual product bands could not be resolved in a region of the gel in which product band differences are most apparent, and at high concentrations they resolved again but with mobilities differing from those at low concentrations (Fig. 4B and SI Appendix, Figs. S10B and S11B). We inferred that these changes in product bands were due to a mixture of PG-LTA and Glc₂DAG-LTA at intermediate Glc₂DAG concentrations and of primarily Glc₂DAG-LTA at higher concentrations. Therefore, polymer length is determined not only by starter unit identity, but also by the concentration of potential starter units. We observed qualitatively similar results when proteoliposomes containing 40 mol% PG and increasing concentrations of Glc₂DAG were used, but in this case the average length of polymers in the reactions was shorter than for proteoliposomes containing 20 mol% PG; moreover, the concentration of Glc₂DAG required for a clear shift in product mobility increased (Fig. 4 A and B and SI Appendix, Fig. S10 A and B). Taken together, these two sets of experiments show that Glc₂DAG and PG are both used as membrane anchors in reactions that contain a 20:1 to 40:1 PG/Glc₂DAG ratio, while lower ratios result in the exclusive use of Glc₂DAG. Therefore, Glc₂DAG is the strongly preferred starter unit. In wild-type Sa, the ratio of PG to Glc₂DAG is <10:1 (42, 43), well within the range that would result in near-exclusive utilization of Glc₂DAG.

Fig. 4.

Glc₂DAG is the preferred LTA lipid anchor and LTA length is determined by the concentration of Glc₂DAG and PG starter units. (A) Alcian blue/silver stained polyacrylamide gel of LTA extracted from proteoliposome reactions containing varying amounts of PG and Glc₂DAG. The asterisk indicates the bromophenol blue loading dye. (B) Zoomed-in region of Fig. 4A at the blue bracket where the periodicities of the banding patterns for different anchors are most out of phase. Blue dots highlight the banding pattern produced by Glc₂DAG-anchored LTA, and orange dots highlight the offset banding pattern from PG-anchored LTA. The banding pattern is blurred at intermediate concentrations of Glc₂DAG, reflecting a heterogenous mixture of Glc₂DAG- and PG-anchored LTA.

A Model for Polymer Length Control by Starter Unit Concentration.

Given that Glc₂DAG reactions are substantially faster than reactions using PG as a membrane anchor, we were initially surprised that Glc₂DAG polymers are shorter. Faster reaction rates typically correlate with longer polymer length for processive reactions. We could not find a description in the literature of any other processive carbohydrate polymerase for which the use of a preferred substrate that reacted quickly led to shorter polymers than the use of a less preferred substrate that reacted more slowly. To help understand what features of LtaS are important for its intrinsic control of LTA length, we used Phyre2 (44) to build a homology model of Sa LtaS from the structure of a distant homolog, the lipid A-phosphoethanolamine transferase EptA from Neisseria meningitidis (45), and then replaced the modeled LtaS extracellular domain with the previously solved crystal structure of this domain (Fig. 5A and SI Appendix, Fig. S12A) (29). We also performed an evolutionary covariance analysis to identify residue contacts between the extracellular and membrane domains (SI Appendix, Table S1) (46). The residue contacts are consistent with the model in showing that the extracellular domain is positioned above the third and fourth transmembrane helices (Fig. 5A).

Fig. 5.

Working model: LTA length is regulated by starter unit competition. (A) Homology model of LtaS constructed from the structure of the extracellular domain of Sa LtaS (cyan; PDB ID code 2W5T) and Phyre2 (tan) (29, 44). A complete description of model generation is provided in Materials and Methods. The nucleophile, T300, is shown in purple along with a manganese ion (gray sphere) and molecule of glycerol-phosphate (green) bound in the active site. Red lines indicate pairs of residues across the extracellular and membrane domains that coevolved. Mutations in the residues in magenta were found to decrease LTA length and/or abundance (complete list in SI Appendix, Fig. S12D) (23). (B) Schematic of length determination by competition for the acceptor binding site. Short primers are extended by processive polymerization. Length is determined on termination, which is enhanced by competition for the acceptor binding site between free Glc₂DAG and PG starter units. The key is as in Fig. 1.

The LtaS extracellular domain contains the catalytic residues necessary to hydrolyze PG, yet hydrolytic activity is very low without the membrane domain, and LTA polymer does not form (47). Carbohydrate polymerases typically contain at least two substrate binding sites: an acceptor site that holds the growing polymer and a donor site in which the substrate that donates the extender unit binds. The LtaS membrane domain likely contains binding sites for both the lipid anchor and for PG, which serves as the phosphoglycerol donor. Because LTA elongates from the growing end (27, 28), and (as our data show) the released polymers cannot productively rebind, there must also be a region in the extracellular domain that holds the end of the growing chain. To identify potential lipid-binding sites in the membrane domain, we analyzed sequence conservation (48) and identified two clusters of well-conserved residues, one directly adjacent to the active site of the extracellular domain in our model and the other more distal from the active site and extracellular domain (SI Appendix, Fig. S12B). A likely path for the growing polymer between these sites has been identified based on the location of conserved residues and a cocrystallized glycerol-phosphate molecule (SI Appendix, Fig. S12C) (49). We propose that the proximal site is where PG binds and transfers phosphoglycerol to T300. We speculate that the distal site binds the lipid anchor of the growing polymer chain. Supporting the importance of these conserved regions, and consistent with a role for each in substrate binding, we previously identified clusters of mutations in both conserved regions that reduce LTA polymer length and abundance (Fig. 5A and SI Appendix, Fig. S12D) (23).

We can propose a model for LTA polymer length control by combining the structural model with our results on how product lengths vary with substrate identity and abundance (Fig. 5B). First, a primer forms slowly, and once it reaches an appropriate length, it binds in the distal acceptor site and is processively extended. The tip of the growing polymer must remain associated with the extracellular domain for elongation to occur, but the lipid end may reversibly dissociate. Binding of a new starter unit (Glc₂DAG or PG) or primer in its place would increase the probability of full dissociation of the entire polymer as the length increases. This model explains the observation that increasing the concentration of either the preferred starter unit Glc₂DAG or the alternative starter unit PG results in shorter polymers.

Discussion

Previous work has shown that control of Sa LTA length is critical for virulence (15, 16), control of cell size and division (20, 23), and cell envelope integrity (23). Genetic approaches identified a role for the LTA glycolipid anchor, Glc₂DAG, in regulating LTA length (15), but the mechanism underlying anchor-dependent polymer length control was a mystery. Here we show that length control is an intrinsic property of LtaS that depends only on the identity and concentration of the membrane-anchored starter units in the reaction. First, we reconstituted LTA synthesis from pure components in a proteoliposome system containing only LtaS, PG, Glc₂DAG (optionally), and additional phospholipids required for liposome assembly. We found that the differential observed in cells between long LTA polymers assembled on PG and short LTA polymers assembled on Glc₂DAG was recapitulated in this system, demonstrating that no other factors from Sa are required to control polymer length. Second, we showed that polymer lengths vary inversely with the concentrations of PG and Glc₂DAG. Finally, we confirmed a processive elongation mechanism for LtaS, but found that shorter polymers are produced with the preferred starter unit, Glc₂DAG, even though the overall reaction is faster. Taken together, these results are consistent with a model of LTA length regulation in which free starter units promote dissociation of the growing LTA polymer from LtaS to irreversibly terminate processive polymerization.

Our proposed mechanism implies that LtaS has three binding sites: a donor site that briefly holds PG as phosphoglycerol is transferred to the catalytic threonine, an acceptor site that binds the lipid anchor of the growing polymer, and a region that holds the elongating end of the growing polymer near the active site as phosphoglycerol units are added. With respect to two-site binding of the growing polymer chain, our mechanism for LtaS polymerization resembles the tethering model previously proposed for Mycobacterium tuberculosis galactofuranosyltransferase GlfT2 (6). As formulated, however, the tethering model predicts product length outcomes that are the opposite of what we observe for LtaS. In the tethering model, both ends of the growing polymer are bound to the enzyme during processive elongation, and the increasing conformational entropy of the growing polymer increasingly favors dissociation (6). Binding interactions are susceptible to competition from unbound substrates, and increasing concentrations of free substrates can result in decreased polymer length (6). However, in the tethering model, acceptor substrates that bind with higher affinity have a lower probability of dissociation and thus produce longer polymers. LtaS is functional for polymer synthesis only in proteoliposomes, and it is challenging to obtain reliable binding affinities for substrates. Nevertheless, the observation that LtaS preferentially uses Glc₂DAG as a starter unit in the presence of a 20-fold excess of PG suggests that it binds more tightly than PG to the lipid anchor site. The overall reaction rate is also faster in vitro, and yet Glc₂DAG-LTA is shorter than PG-LTA. To account for these differences, we propose a starter unit competition model that incorporates the two-point interactions of the tethering model while also accounting for the capacity of LtaS to initiate polymer synthesis on two distinct substrates. Although we have suggested that the anchor substrates compete by directly displacing the anchor end of the growing polymer from its binding site, it is also possible that displacement is caused by binding to an allosteric site or to the donor lipid-binding site. Whereas binding of PG in the donor site would result in further elongation, binding of Glc₂DAG in that same site would result in a pause in elongation during which dissociation could occur. However, an argument against this particular model for displacement is that including Glc₁DAG in the proteoliposomes does not result in shorter polymers, even though it would be accommodated in the binding site.

Our study also shows that LtaS exhibits a lag phase in polymer formation that is shortened by introducing a short primer. If the lipid-binding sites identified in the membrane domain are indeed correct (Fig. 5A), then substrate priming would require conformational changes to bring the acceptor close enough to the active site nucleophile (T300) to react, which may explain the slow initial reaction rates. Alternatively, the priming steps may use a ping-pong mechanism at the donor site in which PG first transfers phosphoglycerol to T300 and then, after DAG release, a lipid anchor substrate binds at the same site and reacts with the enzyme-phosphoglycerol intermediate. Short primers formed by this mechanism could dissociate and migrate to the acceptor site for processive elongation. Finally, the lag phase could also be caused by a slow conformational change (50) in LtaS on the addition of Mn²⁺ that occurs before the processive elongation phase. Structures of full length LtaS bound to Glc₂DAG or primed substrates may shed light on the initial stages of LTA polymerization, and together with our complete reconstitution, could enable development of inhibitors of this important Sa enzyme.

LTA length may be controlled in a similar manner in other bacteria that contain LtaS. Although LTA is generally important for the fitness of Gram-positive organisms, its structure can vary, and five different classes of LTA have been characterized (14, 51). Sa produces type I LTA, the most-studied type, but two other LTA structural types are produced by organisms that contain LtaS homologs. Work on the biosynthesis of these other LTA structural types is sparse, but the presence of LtaS in the organisms that produce them suggests that they may use a related polymerization mechanism. Two additional classes of LTA are produced by organisms that lack LtaS homologs, such as Streptococcus pneumoniae. The S. pneumoniae LTA precursor is fully assembled inside the cell and then transferred to a glycolipid after export to the cell surface (52). It is not known how length control is achieved for S. pneumoniae LTA, or whether this is important.

In closing, we note that our reconstitution of LtaS also explains a formerly perplexing observation about LTA product lengths. Whereas wild-type cells make short LTA and ΔugtP cells make long LTA, ΔltaA cells produce LTA of intermediate lengths. LtaA is a polytopic membrane protein that acts as a flippase to move Glc₂DAG from the inner leaflet of the membrane to the outer leaflet, where it can be used for LTA assembly. It was previously shown that ΔltaA cells produce heterogeneous LTA, with some polymer assembled on PG and some assembled on Glc₂DAG that is evidently exported by another unknown mechanism (15). If LTA length were determined solely by the identity of the anchor lipid, then one might expect ΔltaA cells to produce LTA with a bimodal length distribution comprising long, PG-anchored polymers and short, Glc₂DAG-anchored polymers. Instead, there is a gradient of LTA lengths from wild-type to ΔltaA to ΔugtP cells, which is explained by our results showing that the ratio of Glc₂DAG to PG affects polymer length. Unlike ΔugtP cells, ΔltaA cells still have some Glc₂DAG in the outer leaflet of the membrane to serve as a competing starter unit. Our in vitro studies suggest that there is considerable potential for Sa to control its physiology by altering membrane composition to affect LTA length, a property that profoundly impacts numerous aspects of the cell (23).

Materials and Methods

Expanded descriptions of the experiments are provided in SI Appendix, Materials and Methods.

UgtP Expression and Purification.

Sa UgtP with a C-terminal His₆-tag was expressed in Escherichia coli and purified by immobilized metal affinity chromatography (IMAC) followed by size exclusion chromatography (SEC) in a buffer consisting of 20 mM Hepes pH 7.4, 400 mM NaCl, 500 μM DTT, and 5% glycerol and stored at −80 °C.

LtaS Expression and Purification.

Sa LtaS with a C-terminal His₆-tag was expressed and purified from E. coli in a similar manner as described previously (32). Membrane fractions were solubilized in detergent, and full-length LtaS was sequentially purified by IMAC, DEAE-Sepharose, and SEC in a buffer consisting of 25 mM Hepes pH 7.4, 200 mM NaCl, and 0.05% DDM and stored at −80 °C.

UgtP Reactions and Glycolipid Purification.

UgtP reactions generally consisted of 1 μM enzyme, 190 μM DAG, and 1 mM UDP-glucose in a solution consisting of 20 mM MES pH 6.2, 200 mM NaCl, 0.6% CHAPS, and 10% DMSO at room temperature. Reactions were extracted twice with CHCl₃/MeOH/H₂O. TLC was run on silica gel 60 plates (Millipore Sigma) and stained with n-(1-napthyl)ethylenediamine/H₂SO₄. Glycolipids were purified by preparatory TLC and quantified with n-(1-napthyl)ethylenediamine/H₂SO₄ against known Glc₁αDAG and Gal₂DAG standards. [¹⁴C]-labeled Glc₁DAG and Glc₂DAG were prepared under the same reaction conditions but with 5:1 cold UDP-Glc/UDP-Glc[¹⁴C(U)]. Radiolabeled glycolipids were purified on a Bakerbond spe C18 column with an MeOH gradient.

Purification of Glycolipids from Sa.

An engineered Sa strain with inducible expression of Sa ltaS and B. subtilis yvgJ was grown similarly as described previously to accumulate GroP-Glc₂DAG (38). Cells were lysed with a cell disruptor, insoluble material was removed, and lipids were extracted with CHCl₃/MeOH/H₂O. Glycolipids were purified from the crude lipid extract by preparatory TLC with silica gel 60 plates and quantified with n-(1-napthyl)ethylenediamine/H₂SO₄ against a Gal₂DAG standard.

Mass Spectrometry.

Dried glycolipid samples were resuspended in 5 mM ammonium acetate pH 6.45 (or H₂O for Sa GroP-Glc₂DAG) at 40 to 80 μM. Samples were directly injected into an Agilent 1200 series high-pressure liquid chromatograph in line with an Agilent 6520 Q-TOF mass spectrometer using electrospray ionization with acetonitrile plus 0.1% formic acid at 1 mL/min. Glycolipids were detected in positive mode, except for Sa GroP-Glc₂DAG, which was detected in negative mode. The peak in the total ion chromatogram was integrated in MassHunter quantitative analysis software (Agilent).

Detection of LTA in Sa Lysates.

Cells were grown at 37 °C, lysed with lysostaphin and sodium dodecyl sulfate (SDS), treated with proteinase K, and detected by Western blot as described previously (23).

Extraction and Purification of LTA from Sa.

LTA was extracted and purified in a similar manner as described previously (15, 30). In brief, cells were lysed with a cell disruptor, and LTA was extracted in the aqueous layer of a 1-butanol/H₂O mixture. The aqueous extract was suspended in 5% isopropanol and purified on a Bakerbond spe C18 column with an isopropanol gradient. Fractions with LTA, as determined by Western blot analysis, were pooled and stored in water at −20 °C.

Preparation of LtaS Proteoliposomes.

Appropriate lipids were mixed in CHCl₃ and dried with N₂, followed by lyophilization. Liposomes were prepared by extrusion through a 0.1-μm polycarbonate membrane (Avanti Polar Lipids) in a buffer consisting of 25 mM Hepes pH 7.4, 200 mM NaCl, and 100 μM EDTA. Liposomes were saturated with DDM, and purified LtaS was added at a final concentration of 1 μM. Detergent was removed with Bio-Beads SM-2 (Bio-Rad), and proteoliposomes were stored at 4 °C. The final concentration of lipids was 8.33 mM.

LtaS Proteoliposome Reactions.

Proteoliposome reactions were initiated by dilution of proteoliposomes into four volumes of a buffer consisting of 20 mM MES pH 6.5, 50 mM NaCl, 5% DMSO, and 1 mM MnCl₂ and run at room temperature. For Western blot analysis, samples were quenched with an SDS/PAGE loading buffer. To purify LTA from reactions, LTA was extracted into the aqueous layer of a 1-butanol/H₂O mixture. Proteoliposome reactions with radiolabeled glycolipids were prepared in the same manner but were quenched with MeOH. These reactions were then separated on cellulose-based chromatography paper strips and quantified by liquid scintillation counting, or separated by TLC and detected by exposure to a general-purpose phosphor screen.

PAGE of LTA Polymers.

LTA samples from proteoliposomes were treated with a lipase (resinase HT; Novozymes) (30). LTA samples from cells were adjusted to pH 8 to 9 and treated with resinase HT at 50 °C overnight to cleave D-alanyl and fatty acid esters, followed by treatment with proteinase K. The 20%, 20 × 20 cm polyacrylamide gels were prepared as described previously (53). Gels were run at 28 mA/gel at 4 °C for 16 to 17 h and stained with Alcian blue, followed by a silver stain (Bio-Rad).

LtaS Homology Model Generation.

A model of Sa LtaS was generated using Phyre2 (44) in intensive mode. The final model was created by separately aligning the membrane domain (residues 1 to 229) from the Phyre2 model and the known structure of the Sa LtaS extracellular domain (residues 230 to 641; Protein Data Bank [PDB] ID code 2W5T) (29) to the known structure of EptA (PDB ID code 5FGN) (45). ConSurf (48) was used to analyze sequence conservation among 500 LtaS-related sequences.

Data Availability.

All study data are included in the main text and SI Appendix.