Attachment of phosphorylcholine residues to pneumococcal teichoic acids and modification of substitution patterns by the phosphorylcholine esterase

The bacterial lung pathogen Streptococcus pneumoniae has a unique nutritional requirement for exogenous choline and attaches phosphorylcholine (P-Cho) residues to the GalpNAc moieties of its teichoic acids (TAs) in its cell wall. Two phosphorylcholine transferases, LicD1 and LicD2, mediate the attachment of P-Cho to the O-6 positions of the two GalpNAc residues present in each repeating unit of pneumococcal TAs (pnTAs), of which only LicD1 has been determined to be essential. At the molecular level, the specificity of the P-Cho attachment to pnTAs by LicD1 and LicD2 remains still elusive. Here, using detailed structural analyses of pnTAs from a LicD2-deficient strain, we confirmed the specificity in the attachment of P-Cho residues to pnTA. LicD1 solely transfers P-Cho to α-d-GalpNAc moieties, whereas LicD2 attaches P-Cho to β-d-GalpNAc. Further, we investigated the role of the pneumococcal phosphorylcholine esterase (Pce) in the modification of the P-Cho substitution pattern of pnTAs. To clarify the specificity of Pce-mediated P-Cho hydrolysis, we evaluated different concentrations and pH conditions for the treatment of pneumococcal lipoteichoic acid with purified Pce. We show that Pce can hydrolyze both P-Cho residues of the terminal repeat of the pnTA chain and almost all P-Cho residues bound to β-d-GalpNAc in vitro. However, hydrolysis in vivo was restricted to the terminal repeat. In summary, our findings indicate that LicD1 and LicD2 specifically transfer P-Cho to α-d-GalpNAc and β-d-GalpNAc moieties, respectively, and that Pce removes distinct P-Cho substituents from pnTAs.

Crucial host interactions of the human lung pathogen Streptococcus pneumoniae are mediated by its bacterial cell wall components. It is well known that lipoproteins (LPs) 2 are the predominant TLR2 stimuli (1,2). Diacylated LPs induce signaling via a TLR2/TLR6 heterodimer, whereas triacylated LPs induce signaling via a TLR2/TLR1 heterodimer (3)(4)(5). Furthermore, (pneumococcal) peptidoglycan (PGN) is digested by lysozyme, and cytosolic sensing of the digestion products by NOD2 (Nod-like receptor 2) takes place (6,7). Lipoteichoic acid (LTA) and wall teichoic acid (WTA) of S. pneumoniae are decorated with phosphorylcholine (P-Cho) residues, a modification that is assumed to be of great significance in pneumococcal host-pathogen interactions (8). These P-Cho moieties serve as anchor for surface-located choline-binding proteins (CBPs), which are involved in various pathophysiological functions of this bacterium (8,9). Furthermore, P-Cho is recognized by components of the host immune response, such as the human C-reactive protein and the platelet-activating factor receptor (10 -12). It was shown that human L-ficolin directly interacts with P-Cho residues of the pneumococcal teichoic acids (pnTAs), inducing thereby activation of the lectin complement pathway (13). S. pneumoniae depends on the nutritional uptake of choline, which is further metabolized by a cascade of three enzymes (LicA, LicB, and LicC). Subsequently, two additional enzymes, LicD1 and LicD2, mediate the attachment of P-Cho to pnTA precursor chains. P-Cho is attached specifically to the O-6 positions of the two GalpNAc residues, which are present in each repeating unit (RU) of pnTAs (Fig. 1). LicD1 is assumed to incorporate one of these P-Cho residues and has been shown to be essential (14). LicD2 probably transfers the second P-Cho residue but is a nonessential enzyme (15). However, the specificity of the LicD1-and LicD2-mediated P-Cho transfer to pnTAs has remained elusive so far. We have shown earlier that the terminal RU of pnTAs can occur in different variants with regard to its P-Cho content. In pnLTA isolated from strain D39⌬cps⌬lgt, both GalpNAc residues of the terminal repeat were almost completely substituted with P-Cho independently of the pH value of the culture medium. In contrast, the pnLTA isolated from strain TIGR4⌬cps possessed a terminus in which one or both of these P-Cho residues were significantly reduced or even absent, but only when bacteria had been cultured under mild acidic conditions (1). However, the enzyme(s) involved in this specific modification have not been definitely identified so far. The most likely candidate is the pneumococcal phosphorylcholine esterase (Pce). Pce is a member of the CBP family, and its enzymatic activity was first described in 1974 (16). It has been shown that Pce is able to hydrolyze about 30% of the total P-Cho residues attached to the GalpNAc moieties of WTA and LTA in vitro (17). The crystal structure of Pce revealed the presence of two structural modules, the catalytic module (residues 1-300) and the choline-binding module (residues 313-540). Both are joined by a small linker, which comprises residues 301-312. Analysis of the crystal structure suggested that the removal of P-Cho residues is limited by the configuration of the active site of Pce in such a way that only residues on the end of the TA chains are accessible to the catalytic center (18). Until now, the focus of previous investigations has mainly centered on choline metabolism and its significance for host interactions (19 -22). The specificity of the P-Cho hydrolysis by Pce has not been elucidated to date.
In this study, we investigated the structural specificity of Pcemediated removal of P-Cho residues from pnLTA. Therefore, LTA from a Pce-deficient S. pneumoniae strain was isolated, incubated with enzymatic active, recombinant Pce under various conditions, and analyzed by high-resolution MS and NMR. Furthermore, using an LicD2-deficient strain, the specific attachment of P-Cho residues by LicD1 and LicD2 was clarified.

Assessing the P-Cho substitution pattern of pneumococcal LTA in a Pce-deficient strain
The prerequisite for a reliable analysis of the Pce-mediated P-Cho hydrolysis was the availability of completely P-Cho sub-stituted pnLTA molecules. Therefore, we isolated the LTA of a Pce-deficient strain in the nonencapsulated TIGR4 background (TIGR4⌬cps⌬pce). The LTA was de-O-acylated by hydrazine treatment and purified by gel permeation chromatography as described previously (23). The high-mass region of the deconvoluted spectrum is shown in Fig. 2B (top), and the complete spectrum is depicted in Fig. S1. It shows the typical proportional distribution of chain lengths with the predominant presence of molecules with 6 and 7 RUs for LTA isolated from TIGR4 strains (1,23,24). Furthermore, the spectrum indicates that all de-O-acylated pnLTA molecules of this preparation are completely P-Cho-substituted. The respective 31 P NMR spectrum is depicted in the top panel of Fig. 2A, and the corresponding chemical structure for de-O-acyl pnLTA (1) is shown in Fig.  2C (and in Fig. 2D as a schematic representation), with X ϭ P-Cho at residues H and G for this preparation. In 31 P NMR, LTA of the Pce-deficient strain displays P-Cho signals at ␦ P 0.33 ppm for P-Cho at ␤-D-GalpNAc moieties (residues D and G, Fig. 2C), at ␦ P 0.12 ppm for P-Cho at the terminal ␣-D-GalpNAc (residue H) and at ␦ P Ϫ0.15 ppm for P-Cho at all other ␣-D-GalpNAc moieties (residue E). Signals for ribitol-P (residues CЈ/C) occur at ␦ P 1.89/1.80 ppm.

The specificity and efficiency of the Pce-mediated P-Cho hydrolysis are influenced by the pH value
To determine the specificity and efficiency of P-Cho hydrolysis mediated by Pce, we used the above described, completely P-Cho-substituted LTA of pneumococcal strain TIGR4⌬cps⌬pce in its native and therefore acylated form. An earlier study showed that Pce has its highest activity against p-nitrophenylphosphorylcholine and cell wall components at  pH 8.0 (17). To further evaluate this finding, we treated purified pnLTA with different concentrations of heterologously expressed Pce and compared the enzyme activity at physiological pH value (pH 7.4) and a more basic pH of 8.0. Subsequently, Pce-treated pnLTAs were treated with anhydrous hydrazine and purified by gel permeation chromatography, to avoid aggregates or micelle formation and thus to obtain reliable 1 H and 31 P NMR integral values.

P-Cho attachment and hydrolysis in pneumococcal TAs
Analysis of the Pce-treated LTA from strain TIGR4⌬cps⌬pce (pH 7.4) by 31 P and 1 H NMR revealed changes in the P-Cho substitution pattern depending on the Pce concentrations used (80, 160, or 240 g of Pce/mg of isolated LTA; Fig. 2A and Table  1). At 80 g of Pce/mg of isolated LTA, partial hydrolysis of the P-Cho DϩG and P-Cho H moieties could be observed. An increase of the Pce concentration to 160 g/mg showed only an effect on the hydrolysis of these P-Cho moieties, whereas the amount of Rib-ol-P C/CЈ and P-Cho E remained unaltered. At this concentration, almost all P-Cho H moieties have been hydrolyzed, and the amount of P-Cho DϩG was lowered by more than half. An increase of the Pce concentration from 160 to 240 g/mg had only a marginal effect. In Fig. 2B, the respective MS spectra recorded from these preparations are depicted, focusing on de-O-acylated LTA molecules with 6 RUs (the full versions of these MS spectra are shown in Figs. S2-S4). Molecular mass differences of 165 Da between two mass peaks are indicated, which corresponds to the loss of one P-Cho moiety each. In total, we observed the loss of up to seven P-Cho residues (7 ϫ Ϫ165 Da). In combination with the 31 P NMR results for P-Cho H and P-Cho DϩG , this indicates that both P-Cho residues from the terminal RU and all P-Cho substituents at ␤-D-Galp-NAc moieties within the RU (residue D) can be hydrolyzed. However, mass peak VI (7054.44 Da) is of highest abundance in the preparations after treatment with 160 or 240 g of Pce/mg of LTA, whereas mass peak VII (6889.39 Da) is only marginally present. Whether this is a specific P-Cho moiety that is not efficiently hydrolyzed (e.g. P-Cho D of the first RU) can only be speculated. The observed masses for all LTA molecules are in agreement with their respective calculated masses and are listed in Table 2. A change in the pH value from pH 7.4 to pH 8.0 improved the efficiency of hydrolysis on P-Cho H for the concentration of 80 g/mg LTA ( Fig. 3 and Table 3). Nonetheless, pH 8.0 shows a hydrolysis behavior with a lower overall effi-ciency for P-Cho attached to residues D and G. An increase of the Pce concentration from 80 g/mg LTA to 160 g/mg LTA at pH 8.0 resulted in no further significant change of the hydrolysis of moieties P-Cho DϩG . In summary, our in vitro analysis clearly demonstrates that Pce hydrolyzes only the moieties P-Cho DϩG and terminal P-Cho H . P-Cho E and Rib-ol-P C/CЈ moieties are not hydrolyzed by Pce.

LicD2 transfers P-Cho substituents to ␤-D-GalpNAc moieties in pnTAs
To study the specificity of the attachment of P-Cho residues by LicD2, we isolated and analyzed LTAs of LicD2-deficient strains. The native LTAs of S. pneumoniae TIGR4⌬cps⌬licD2 and TIGR4⌬cps⌬pce⌬licD2 were isolated, purified, and subsequently analyzed by NMR and MS (Fig. 4). The 31 P NMR spectra from isolated LTA revealed only a P-Cho substitution at the ␣-D-GalpNAc residues within the RUs (P-Cho E ) and the ␣-D-GalpNAc at the terminus (P-Cho H ). The respective signal for P-Cho substituents at the ␤-D-GalpNAc at 0.33 ppm (P-Cho DϩG ), which is present in pnLTA isolated from the respective parental strains TIGR4⌬cps⌬pce ( Fig. 2A, top) and TIGR4⌬cps (1), is absent (Fig. 4A). All NMR data for the hydrazine-treated LTA of TIGR4⌬cps⌬pce⌬licD2 are listed in Table  4, and a section of the respective 1 H, 13 C HSQC NMR is shown in Fig. S5, including assignment of signals. In the MS analysis of this de-O-acyl LTA, only the expected masses for pnLTA with one P-Cho residue per RU were observed (Fig. 4B, top) Table 2). This is equivalent to 6 P-Cho residues less in these LTA molecules of strain TIGR4⌬cps⌬pce⌬licD2. In the MS analysis of de-O-acyl LTA of strain TIGR4⌬cps⌬licD2, the above described Pce-mediated hydrolysis of P-Cho residues at the ␣-D-GalpNAc of the terminus (P-Cho H ) is clearly visible (Fig. 4B (middle panel; the complete MS spectrum is shown in Fig. S7) and Table 5). Notably, LTA chains of strain TIGR4⌬cps⌬licD2 tended to be slightly longer, as observed for TIGR4⌬cps⌬pce⌬licD2. 31 P NMR analysis of the correspond-  Table 2. C and D, current structural model for de-O-acyl LTA of S. pneumoniae strain TIGR4 (and mainly all other pneumococcal strains (24)), depicted as a detailed chemical drawing (C) as well as a schematic cartoon (D). Table 1 Integration values from 31 P NMR spectra from native LTA of TIGR4⌬cps⌬pce before and after treatment with the indicated amounts of Pce shown in Fig. 2A To reveal comparable integrals, the signal of P-Cho H was set to 1.0 in the spectrum of de-O-acylated LTA of non-Pce-treated TIGR4⌬cps⌬pce ( Fig. 2A, top). The resulting integral value for Rib-P C´/ Rib-P C residues served then as reference for signal integration in 31 P NMR spectra recorded from the Pce-treated pnLTA preparations. For each Pce-treatment condition, two independent experiments were performed. Stated integral values always reflect the mean of two such independent experiments Ϯ S.D.

P-Cho attachment and hydrolysis in pneumococcal TAs
ing PGN-WTA complexes after LytA treatment indicated the same P-Cho substitution pattern for both LicD2-deficient strains (Fig. 5), further strengthening our findings that solely LicD2 is mediating the attachment of P-Cho substituents to ␤-D-GalpNAc moieties in pnTAs.
To investigate the specificity and efficiency of Pce-mediated P-Cho hydrolysis for LTA comprising only one P-Cho residue per RU, LTA isolated from strain TIGR4⌬cps⌬licD2 was treated with heterologously expressed Pce as described above. These experiments indicated that Pce is capable of hydrolyzing P-Cho at the ␣-D-GalpNAc of the terminus, whereas P-Cho E residues were not hydrolyzed (Fig. 4 (bottom) and Tables 5 and  6). The observed masses for pnLTA molecules isolated from the two different LicD2-deficient strains as well as from the men-tioned Pce-treated preparation are all in accordance with the respective calculated monoisotopic masses (Table 5).

Discussion
Previous studies showed that the pneumococcal Pce is capable of removing in vitro about 15-30% of the total P-Cho residues attached to pneumococcal TAs (16,17). Our study demonstrates now, by combining NMR and MS analyses, the specificity and efficiency of the P-Cho hydrolysis at GalpNAc residues mediated by Pce. With one-dimensional 31 P and 1 H NMR spectra, a direct assignment of specific P-Cho residues to the respective ␣or ␤-configured GalpNAc residues of pnTA RUs is possible. With this direct assessment, it is further possible to judge and compare the Pce-mediated P-Cho hydrolysis at different pH values. Compared with LTA of S. pneumoniae strain TIGR4⌬cps, LTA of the Pce-deficient strain exhibited a complete P-Cho substitution pattern. Isolated pnLTA of the Pce-deficient strain was treated at different concentrations and different pH values with heterologously expressed Pce. P-Cho substituents at the terminal ␣-D-GalpNAc residue (P-Cho H ) and at the ␤-D-GalpNAc moieties (P-Cho DϩG ) were efficiently hydrolyzed at a concentration of 80 g of Pce/mg of LTA at both tested pH values. An increased hydrolysis rate was observed at a pH of 7.4, when the concentration was increased from 80 to 160 g/mg LTA, which was not the case at pH 8.0. In contrast, at pH 8.0 and a Pce concentration of 160 g/mg LTA, the P-Cho H residue was completely hydrolyzed, which was not observed at pH 7.4. However, P-Cho substituents at ␤-D-Galp-NAc residues were less efficiently hydrolyzed by the Pce at pH 8.0 compared with pH 7.4. A possible reason for the altered hydrolysis efficiency at the studied pH values could be conformational changes of the catalytic domain of Pce. In summary, a removal of about 20 -30% of P-Cho residues by hydrolysis was observed, which is in good agreement with previously described findings for in vitro studies (17).
The observed specificity for the Pce-mediated hydrolysis indicates that P-Cho residues bound to either ␣-D-GalpNAc or ␤-D-GalpNAc have a different biological importance and makes it most likely that different enzymes are responsible for their specific attachment. Previous studies using the monoclonal antibody TEPC-15, which recognizes a P-Cho epitope in S. pneumoniae, showed that LicD2 mutant strains differ in their P-Cho substitution compared with WT strains (14). By applying detailed chemical structural analyses, we showed here that LicD2-deficient S. pneumoniae strains exclusively exhibit a P-Cho decoration on the ␣-D-GalpNAc of their LTA as well as Table 2 Calculated and observed masses as well as the mass accuracy of de-O-acylated pnLTA molecules with 6 RU, considering different numbers of missing P-Cho residues (signals I-VII in Fig. 2B

P-Cho attachment and hydrolysis in pneumococcal TAs
of their WTA. This provides clear proof that the P-Cho attachment to the ␤-D-GalpNAc residues is mediated by LicD2. By implication, LicD1 catalyzes the attachment of P-Cho to ␣-D-GalpNAc (Fig. 6), because it is the only P-Cho transferase besides LicD2 that has been identified in the pneumococcal genome (15). A direct validation of the LicD1-mediated P-Cho transfer is not possible due to its essentiality for the pneumococcus (14). Our results further confirm that LTA and WTA in S. pneumoniae are synthesized via a shared biosynthesis route (15,23). Using LTA isolated from TIGR4⌬cps⌬licD2, Pce still hydrolyzes the P-Cho substituents on the terminal ␣-D-Galp-NAc (P-Cho H ) residues, whereas P-Cho moieties at ␣-D-Galp-NAc located within the RU were not affected. This finding is an indication that P-Cho residues on the ␤-D-GalpNAc are not required for binding of Pce. Because the P-Cho residues on the ␣-D-GalpNAc moieties are essential for binding of CBPs, whereas P-Cho residues on ␤-D-GalpNAc are not, a structural explanation for the essential nature of LicD1 is given (14,15). This is also in line with the results of analyzed pneumococcal strains only possessing one P-Cho per RU in their TAs, which all lack only the P-Cho at the ␤-D-GalpNAc moieties (21,24,25). In summary, our experiments indicate that Pce is the only enzyme responsible for P-Cho removal from pnTAs. Pce only hydrolyzes the P-Cho residues on ␤-D-GalpNAc residues as well as on the terminal ␣-D-GalpNAc. This in vitro removal corresponds to up to 30% of the total P-Cho content. In vivo, Pce is only able to hydrolyze P-Cho residues present in the terminal repeating unit. This finding is in line with the previous observation that the activity of Pce might be limited by more complex constraints related to the topography of the pneumococcal surface (17). Only residues that are located at the end of the TA chains may be accessible to the active site of the catalytic center of Pce in vivo. Moreover, P-Cho residues on the bacterial Table 3 Integration values from 31 P NMR spectra from native LTA of TIGR4⌬cps⌬pce before and after treatment with the indicated amounts of Pce at pH 8.0 shown in Fig. 3 To reveal comparable integrals, the signal of P-Cho H was set to 1.0 in the spectrum of de-O-acylated LTA of non Pce-treated TIGR4⌬cps⌬pce (a different batch of LTA has been used here compared with Table 2). The resulting integral value for Rib-P CЈ /Rib-P C residues served then as reference for signal integration in 31 P NMR spectra recorded from the Pce-treated pnLTA preparations. For each Pce treatment condition, two independent experiments were performed. Stated integral values always reflect the mean of two such independent experiments Ϯ S.D. ND, not detectable.

P-Cho attachment and hydrolysis in pneumococcal TAs
surface are a target for components of the host immune response, such as the human C-reactive protein, the plateletactivating factor receptor, or L-ficolin, which leads to different reactions in immune defense, such as activation of the lectin complement pathway (24). A selective modification of the P-Cho pattern on the surface, mediated by Pce activity, could impair targeting of the pneumococcus by these host components and may thus favor infection and colonization by S. pneu-

P-Cho attachment and hydrolysis in pneumococcal TAs
moniae. Furthermore, we could show on the molecular level the specificity of P-Cho attachment to pnTAs mediated by LicD1 and LicD2. Our study revealed that LicD2 solely promotes the attachment of P-Cho residues to the ␤-D-GalpNAc residues, and LicD1 catalyzes the attachment of P-Cho residues to ␣-D-GalpNAc (Fig. 6).

Bacterial strains and growth
Pneumococcal strains used in this study are listed in Table  S1. Bacteria were grown on Columbia blood agar plates (Oxoid) or in Todd-Hewitt broth supplemented with 0.5% yeast extract (THY; Roth) containing appropriate antibiotics (kanamycin (150 g/ml), erythromycin (5 g/ml), and chloramphenicol (5 g/ml)). Cultivation on plates or in liquid cultures was performed at 37°C and 5% CO 2 .

Mutant construction
All primers used are listed in Table S2. For the construction of the pneumococcal licD2 mutant in S. pneumoniae TIGR4⌬cps, a DNA fragment consisting of the S. pneumoniae TIGR4 licD2 gene and ϳ500-base pair up-and downstream flanking regions were amplified by PCR using primers LicD2SphIfor and LicD2SacIrev. The resulting PCR product was cloned into plasmid pUC18 (Thermo Fisher Scientific). This plasmid was used as template for an inverse PCR with primer Invrev1130BamHI and Invfor1130SmaI. Afterward, an ermB gene, amplified by PCR from vector pTP1 using primer InvrevBamHIErm and InforSmaIErm was inserted (26). The final recombinant plasmid was used to transform and mutagenize S. pneumoniae TIGR4⌬cps.
To delete the pce gene in S. pneumoniae TIGR4, primer CBPE1 and CBPE2 were used to amplify the pce gene from chromosomal DNA of S. pneumoniae R6 (without cholinebinding repeats) by PCR. The resulting PCR fragment was cloned into plasmid pQE-30 (Qiagen) via BamHI and HindIII restriction sites. Primer CpgE6fXma and CbpE4rXma were used for an inverse PCR, and the constructed plasmid was used as template DNA. After digestion with XmaI, an erythromycin resistance cassette (ermB) was cloned into the plasmid. Afterward, S. pneumoniae TIGR4 was transformed, and resulting clones were selected on blood agar plates containing erythromycin.
S. pneumoniae TIGR4⌬cps⌬pce was constructed by transformation of S. pneumoniae TIGR4⌬pce with chromosomal DNA of S. pneumoniae TIGR4⌬cps and selection on blood agar plates containing kanamycin and erythromycin (27).
The triple deletion mutant S. pneumoniae TIGR4⌬cps⌬pce⌬licD2 was generated by replacement of the ermB gene in the licD2 deletion plasmid by a chloramphenicol resistance gene (cat) and transformation of S. pneumoniae TIGR4⌬cps⌬pce with the resulting plasmid.

Extraction, isolation, and chemical treatment of pnTAs
Isolation and purification of pnTAs (LTA and WTA) as well as treatment of purified LTA with anhydrous hydrazine to generate de-O-acyl LTA have been performed as described earlier (23).

Expression and purification of Pce
E. coli BL21 pRGR12 strain was used for heterologous expression and purification of pneumococcal Pce as described (28). Briefly, bacteria were grown in Luria broth medium until exponential phase (A 600 of 0.7) on an environmental shaker at 30°C. At this time point, isopropyl-␤-D-thio-galactopyranoside (1 mM) was added, and incubation proceeded for 3 h. After centrifugation, bacterial pellet was solubilized by ultrasonic treatment. The heterologously expressed Pce was purified using DEAE-cellulose, as described (29).

Phosphorylcholine esterase treatment of pnLTA
Treatment of pnLTA with Pce was performed basically as described elsewhere but with some modifications in detail (17). Purified pnLTAs were dissolved in 50 mM K 2 HPO 4 /KH 2 PO 4 to obtain a concentration of 2.62 mg/ml at pH 7.4 or pH 8.0, respectively. Heterologously expressed Pce was added in different concentrations (80, 160, and 240 g/mg of pnLTA). The solution was incubated for 24 h at 37°C. To inactivate the enzyme, the sample was incubated for 5 min at 100°C and afterward centrifuged at 10,000 ϫ g for 20 min at 4°C. Finally, the solution was lyophilized. For each Pce treatment condition, two independent experiments were performed.

NMR spectroscopy
NMR spectroscopic measurements were performed in D 2 O at 300 K on a Bruker Avance III 700-MHz spectrometer (equipped with an inverse 5-mm quadruple-resonance Z-grad cryoprobe). Deuterated solvents were purchased from Deutero GmbH (Kastellaun, Germany). For calibration of 1 H (␦ H ϭ 2.225 ppm) and 13 C (␦ C ϭ 30.89 ppm) NMR spectra, acetone was used as an external standard. The 31 P NMR spectra (␦ P ϭ 0.0 ppm) were calibrated with 85% phosphoric acid in D 2 O as an external standard. All data were acquired and processed using Bruker TOPSPIN version 3.0 or higher. 1 H NMR assignments were confirmed by two-dimensional 1 H, 1 H COSY and TOCSY experiments, and 13 C NMR assignments were indicated by two-dimensional 1 H, 13 C HSQC, based on the 1 H NMR assignments. From two-dimensional 1 H, 13 C HMBC and 1 H, 13 C HSQC-TOCSY experiments, interresidue connectivity was obtained. Connectivity of phosphate groups was assigned by two-dimensional 1 H, 31 P HMQC and 1 H, 31 P HMQC-TOCSY.

Mass spectrometry
All samples were measured on a Q Exactive Plus mass spectrometer (Thermo Scientific, Bremen, Germany) using a Triversa Nanomate (Advion, Ithaca, NY) as ion source. All measurements were performed in negative-ion mode using a spray voltage of Ϫ1.1 kV. Samples were dissolved in a water/propan-2-ol/trimethylamine/acetic acid mixture (50:50:0.06:0.02, v/v/ v/v). The mass spectrometer was externally calibrated with glycolipids of known structure. All mass spectra were chargedeconvoluted and given mass values refer to the monoisotopic mass of the neutral molecules, if not indicated otherwise.  Integration values from 31 P NMR spectra from native LTA of TIGR4⌬cps⌬pce⌬licD2 as well as TIGR4⌬cps⌬licD2 before and after treatment with the indicated amount of Pce shown in Fig. 4 To reveal comparable integrals, the signal of P-Cho H was set to 1.0 in spectra of de-O-acylated LTA of non-Pce-treated TIGR4⌬cps⌬pce⌬licD2. The resulting integral value for Rib-P CЈ /Rib-P C residues served then as reference for signal integration in the other 31 P NMR spectra. Isolations of pnLTAs were done twice, and for the Pce treatment, two independent experiments were performed. Stated integral values always reflect the mean of two such independent experiments Ϯ S.D. ND, not detectable.