Structural analysis and immunostimulatory potency of lipoteichoic acids isolated from three Streptococcus suis serotype 2 strains

Abstract

Streptococcus suis serotype 2 is an important porcine and human pathogen. Lipoteichoic acid (LTA) from S. suis has been suggested to contribute to its virulence, and absence of d-alanylation from the S. suis LTA is associated with increased susceptibility to cationic antimicrobial peptides. Here, using high-resolution NMR spectroscopy and MS analyses, we characterized the LTA structures from three S. suis serotype 2 strains differing in virulence, sequence type (ST), and geographical origin. Our analyses revealed that these strains possess–in addition to the typical type I LTA present in other streptococci–a second, mixed-type series of LTA molecules of high complexity. We observed a ST-specific difference in the incorporation of glycosyl residues into these mixed-type LTAs. We found that strains P1/7 (ST1, high virulence) and SC84 (ST7, very high virulence) can attach a 1,2-linked α-d-Glcp residue as branching substituent to an α-d-Glcp that is 1,3-linked to glycerol phosphate moieties and that is not present in strain 89-1591 (ST25, intermediate virulence). In contrast, the latter strain could glycosylate its LTA at the glycerol O-2 position, which was not observed in the other two strains. Using LTA preparations from WT strains and from mutants with an inactivated prolipoprotein diacylglyceryl transferase, resulting in deficient lipoprotein acylation, we show that S. suis LTAs alone do not induce Toll-like receptor 2–dependent pro-inflammatory mediator production from dendritic cells. In summary, our study reveals an unexpected complexity of LTAs present in three S. suis serotype 2 strains differing in genetic background and virulence.

Introduction

Streptococcus suis serotype 2 is a Gram-positive bacterial pathogen of pigs responsible for important economic losses to the porcine industry worldwide, causing sudden death, meningitis, and a variety of other pathologies (1). Moreover, S. suis is an emerging zoonotic agent responsible for septic shock and meningitis in humans and has become a public health issue, particularly in South-East Asia (2). However, S. suis serotype 2 strains are genotypically and phenotypically heterogeneous, both in terms of geographical distribution and virulence, with different sequence types (STs),⁴ based on multilocus sequence typing (MLST), as described previously (3). This molecular typing technique is based on differences in the genetic sequences of conserved housekeeping genes, in which mutations are assumed to be largely neutral (4). MLST has classified the different S. suis strains into a large number of STs, based on differences in these genes, which has facilitated the understanding of its evolutionary divergence and epidemiology (3, 4). Data provided by the S. suis MLST database revealed that 701 different STs have been identified for S. suis (all serotypes combined), from a total of 1528 isolates. Serotype 2 strains (which is by far the most important serotype for both pigs and humans) belong to 111 of these 701 STs. Regardless of this elevated number, these STs can be regrouped into only a few clusters called clonal complexes (CC). As depicted in Fig. S1, ST1 is the most prevalent ST in terms of the number of isolates among serotype 2, and it is the founder of CC1. An important CC1 subgroup, founded by the ST7 (Fig. S1), has substantially grown in importance in recent years and was responsible for the 1998 and 2005 human S. suis outbreaks in China (5, 6). As such, ST1 strains, which predominate throughout Eurasia, South America, and Australia, and ST7 strains that are present in China, are genetically similar and generally virulent/highly virulent, respectively (3, 7, 8). Meanwhile, the CC25 and CC28, founded by the ST25 and ST28 respectively, rank as the second and third most important clusters (Fig. S1). In fact, the ST25 and ST28 account for 95% of North American strains isolated from clinical cases of porcine and human disease (3, 9). Although ST25 strains are of intermediate virulence, ST28 strains are considered low virulent and have been mostly associated with secondary infections and immunocompromised individuals (8, 9). Consequently, we chose ST1, ST7, and ST25 strains for our analyses because of their importance as predominant virulent S. suis serotype 2 strains, which are responsible for most porcine and human infections worldwide (3).

Besides peptidoglycan and lipoproteins (LPs), wall teichoic acids and lipoteichoic acids (LTA) are the major constituents of the Gram-positive cell wall. LTA has been suggested to contribute to the virulence of S. suis, whereas the absence of d-alanine residues from its LTA has been associated with an increased susceptibility to the action of cationic antimicrobial peptides (10). In another study (11), LPs, which are often co-purified with LTA (12, 13), were determined to be major activators of the porcine innate immune system. As of now, the detailed structure of the S. suis LTA remained elusive. In general, LTA contains a lipophilic anchor formed by diacylglycerol (DAG), which anchors these molecules to the cell membrane. The DAG is substituted with a glycosyl moiety at the O-3 position. These glycosyl moieties, as well as the attached complex backbone structures consisting of repetitive units (RUs), are highly variable between different species of Gram-positive bacteria. So far, five different types of LTA have been described, which are mainly characterized by the architecture of their RUs (polyglycerol phosphate (type I), complex glycosyl-glycerol-phosphate (type II + III), glycosyl-ribitol-phosphate (type IV), and glycosyl-phosphate (type V)-containing polymers) (14).

We describe herein the structural analysis of LTA isolated from three S. suis serotype 2 strains of different background as representatives of the most clinically and epidemiologically important STs using chemical degradations, high-resolution MS analysis, as well as one- and two-dimensional, homo- and heteronuclear NMR spectroscopy. Finally, the immunostimulatory properties of these well-characterized LTA preparations were evaluated to understand their role in the activation and modulation of the host innate immune response by S. suis.

Results

For the structural analyses, LTA preparations from the three S. suis serotype 2 strains P1/7 (ST1), SC84 (ST7), and 89-1591 (ST25), as well as from their respective Δlgt mutants, were prepared according to our previously published workflow (15). The latter strains lack the gene encoding for the lipoprotein diacylglyceryl transferase (Lgt) and are therefore deficient in lipidation of prolipoproteins (16). In a first step, we compared the ¹H NMR spectra of native LTA from all three strains (Fig. 1). Spectra have been recorded in deuterated 25 mm sodium phosphate buffer, pH 5.5, at 300 K to suppress fast de-alanylation. In all three ¹H NMR spectra, the typical NMR signals for polyglycerol phosphate chains of type I LTA molecules (14, 17) can be observed. Furthermore, all three strains are capable of modifying the glycerol O-2 position with alanine as indicated by the broad signal at δ_H 5.43–5.35 for proton H-2 of the alanine-substituted glycerol moieties, the signal at δ_H 4.33–4.27 for the CH group, and the doublet at δ_H 1.63 for the CH₃ group of alanine (17). Whereas the spectra of LTA from strains P1/7 and SC84 are almost identical, the spectrum of LTA from strain 89-1591 indicates a different binding position for some of the present glycosyl residues (Fig. 1). Therefore, the detailed analysis of the latter LTA will be described separately below. Notably, the ¹H NMR spectra of LTA from strains P1/7 and SC84 are virtually identical with the published ¹H NMR spectrum for LTA isolated from S. suis serotype 2 strain 31533 (10).

Figure 1.

¹H NMR analysis of native LTA of the investigated S. suis serotype 2 strains indicates differences in the binding position of glycosyl residues. Shown are ¹H NMR spectra (δ_H 6.0–0.0) isolated from S. suis strains P1/7 (ST1) (A), SC84 (ST7) (B), and 89-1591 (ST25) (C) recorded in deuterated 25 mm sodium phosphate buffer, pH 5.5, at 300 K. Black arrows in C indicate deviating major anomeric signals in LTA of strain 89-1591 compared with the other two spectra.

For a better characterization of the LTA structures, we generated defined part structures by hydrofluoric acid (HF) treatment, from which we obtained the respective lipid anchor and de-phosphorylated RUs. Afterward, we analyzed the interconnection of the RUs using de-O-acylated LTA, which was obtained by hydrazine treatment, using NMR and MS.

Preparation and structural analysis of the lipid anchor and LTA part structures from LTA of strains P1/7 and SC84

To elucidate the nature of the lipid anchor, we treated the isolated LTA of strains P1/7 and SC84 with 48% HF for 2 days at 4 °C according to our previously published procedure (18). This treatment cleaves all phosphodiester bonds and leads to the formation of dephosphorylated RUs as well as to the release of the glycolipid anchor from the LTA. The HF-treated LTA was applied to hydrophobic–interaction chromatography as described for the purification of intact LTA (shown for strain P1/7 in Fig. S2). Early in the gradient, the dephosphorylated RUs and other fatty acid–free part structures were eluted in fractions 6–9. To collect the lipid anchor, the later eluting UV-active fractions 26–34 were combined. The exact structure of the carbohydrate part of the glycolipid anchor of S. suis P1/7 and SC84 LTA was analyzed by NMR (Fig. S3 and Table S1) and was determined as α-d-Glcp-(1→2)-α-d-Glcp-(1→ (kojibiose). Consequently, the glycolipid anchor of S. suis P1/7 and SC84 LTA is α-d-Glcp-(1→2)-α-d-Glcp-(1→3)-DAG (kojibiose-diacylglycerol; 1a; bold numbers always refer to the structures shown in Fig. 2). Early fractions were combined, lyophilized, and subsequently submitted to gel-permeation chromatography (Bio-Gel P-10; shown for strain P1/7 in Fig. S4). As major molecule, we identified α-d-Glcp-(1→2)-α-d-Glcp-(1→1)-glycerol (2; Fig. 2) by NMR analysis (Fig. S5 and Table S2). In minor amounts, α-d-Glcp-(1→1)-glycerol (3; Fig. 2) was present as well (NMR data in Table S3). Furthermore, no further substituted glycerol (4; Fig. 2) was observed (signals marked with # in Fig. S5).

Figure 2.

LTA part structures of the investigated S. suis strains. Shown is a compilation of structures of the observed glycolipid anchor 1a (LA; R = FA (fatty acids)), its de-O-acylated version 1b (A; R = H), and monomerized LTA repeats of different nature (2–7) isolated after HF treatment. Assignment of the different glucose moieties (Glc I, Glc II, Glc III, Glc I^LA, and Glc II^LA) is used throughout.

Structural analysis of hydrazine-treated LTA of strains P1/7 and SC84

Hydrazine treatment cleaves off all ester-bound residues like fatty acids and alanine and thus reduces the structural heterogeneity of LTA molecules. Therefore, the basic structural features of the LTA chain is accessible, because phosphodiester bonds remain intact (18–20). In the following, the structural investigation of hydrazine-treated LTA (LTA_N2H4) of S. suis strains P1/7 and SC84 by NMR is discussed. In Fig. 3, the ¹H,¹³C-heteronuclear single quantum correlation (HSQC) NMR of LTA_N2H4 of strain P1/7 is depicted as an example for both strains, because almost identical spectra were obtained. The complete NMR chemical shift data are summarized in Table 1. LTA isolated from WT strains and their respective Δlgt mutants are structurally identical, and the respective ¹H NMR spectra of native LTA preparations from Δlgt mutants are depicted in Fig. S6 (for spectra of LTA preparations from WT strains see Fig. 1).

Figure 3.

NMR analysis of de-O-acylated LTA of S. suis strain P1/7 (ST1). Shown is a section (δ_H 5.5–3.0; δ_C 105–50) of the ¹H,¹³C HSQC NMR spectrum (recorded in D₂O at 300 K) obtained from hydrazine-treated LTA of S. suis strain P1/7 (ST1), including assignment of signals. The corresponding NMR chemical shift data are listed in Table 1. (# indicates the anomeric signal of a tiny amount of Glc I lacking Glc II.)

TABLE 1.

¹H (700.4 MHz), ¹³C NMR (176.1 MHz), and ³¹P NMR (283.5 MHz) chemical shift data (δ, ppm) (J, Hz) for S. suis strain P1/7 (ST1) LTA after hydrazine treatment (de-O-acyl LTA)

* indicates nonresolved multiplet.

The MS analysis of LTA_N2H4 of S. suis strains P1/7 and SC84 revealed a remarkably high diversity of LTA molecules. In total, we observed >60 different LTA moieties, which could be grouped into two major structural types. One series of molecules belonged to type I LTA with the observed de-O-acylated lipid anchor 1b (Fig. 2) with different numbers (X = 3–14) of glycerol phosphate (GroP)-repeating units. The second type of observed LTA contained, in addition, more complex RUs consisting of α-d-Glcp-(1→2)-α-d-Glcp-(1→3)-GroP (= Y). The number of GroP repeats (X) varied from 3 to 10, and 1 to 10 RUs of structure Y were present in these LTA_N2H4 molecules. However, we only observed LTA molecules with 4–9 GroP moieties, which had multiple repeats (more than one repeat) of structure Y attached. In Fig. 4A, a representative mass spectrum of the LTA_N2H4 of strain P1/7 is depicted. In Fig. 4B, the region of 800–2700 Da is enlarged, and peaks for observed LTA_N2H4 molecules are assigned. The mass region containing molecules with higher molecular weight is shown in Fig. S7. The full list of identified LTA_N2H4 molecules for S. suis strains P1/7 and SC84 is given in Table 2.

Figure 4.

Molecular species distribution of de-O-acylated LTA of S. suis strain P1/7 (ST1). A, charge-deconvoluted spectrum of a representative MS analysis performed in the negative ion mode (the mass region of 2700–5300 Da has been magnified by a factor of 5). B, zoom into the region between 800 and 2700 Da, including detailed assignment for molecules consisting of the de-O-acyl glycolipid anchor A (1b; Fig. 2) and different combinations of RUs X (= GroP) and Y (= α-d-Glcp-(1→2)-α-d-Glcp-(1→3)-GroP). Details of all assigned molecular species are summarized in Table 2. Relative abundance for a spectral region was always normalized to the respective base peak. Additional information about molecular assignments for higher molecular species are depicted in Fig. S7.

Table 2.

Mass spectrometric analysis of de-O-acyl LTA of S. suis strains P1/7 (ST1) and SC84 (ST7)

Summary of calculated and observed molecular masses (Da) for LTA preparations after hydrazine treatment. For each preparation, two independent MS analyses have been performed, and the identified molecules are listed as a combined list (A = de-O-acyl glycolipid anchor (1b); RU X = GroP; RU Y = α-Glcp-(1→2)-α-d-Glcp-(1→3)-GroP). Masses observed only in one of the two replicates are written in italic type. Accuracy of the measurement is stated as Δppm; ND= not detected.

Molecule	Calculated mass	Strain P1/7		Strain SC84
		Observed mass	Accuracy	Observed mass	Accuracy
	Da	Da	Δppm	Da	Δppm
X3A	878.162	878.163	0.5	ND
X4A	1032.165	1032.166	0.4	1032.167	1.6
X5A	1186.169	1186.169	0.0	1186.168	−0.4
X6A	1340.172	1340.171	−0.6	1340.170	−0.9
X7A	1494.175	1494.174	−0.4	1494.173	−1.0
X8A	1648.178	1648.177	−0.8	1648.176	−1.1
X9A	1802.181	1802.181	0.2	1802.181	−0.2
X10A	1956.184	1956.186	1.1	1956.186	0.8
X11A	2110.187	2110.188	0.1	2110.187	−0.1
X12A	2264.190	2264.192	0.8	2264.191	0.4
X13A	2418.193	2418.192	−0.5	2418.192	−0.7
X14A	2572.197	2572.204	2.9	ND
Y1X3A	1356.271	1356.270	−0.7	1356.270	−1.0
Y1X4A	1510.274	1510.274	−0.2	1510.273	−0.5
Y2X4A	1988.383	1988.385	0.9	1988.384	0.6
Y3X4A	2466.492	2466.494	1.1	2466.493	0.7
Y4X4A	2944.601	2944.602	0.3	2944.601	0.0
Y5X4A	3422.709	3422.707	−0.6	3422.707	−0.6
Y6X4A	3900.818	3900.818	−0.1	3900.808	−2.7
Y7X4A	4378.927	4378.929	0.4	4378.925	−0.4
Y8X4A	4857.036	4857.033	−0.6	4857.031	−1.0
Y9X4A	5335.144	ND		5335.140	−0.8
Y1X5A	1664.277	1664.276	−0.6	1664.276	−0.6
Y2X5A	2142.386	2142.389	1.5	2142.389	1.4
Y3X5A	2620.495	2620.494	−0.2	2620.494	−0.4
Y4X5A	3098.604	3098.601	−0.8	3098.600	−1.0
Y5X5A	3576.712	3576.707	−1.6	3576.708	−1.2
Y6X5A	4054.821	4054.810	−2.8	4054.808	−3.2
Y7X5A	4532.930	4532.929	−0.2	4532.928	−0.4
Y8X5A	5011.039	5011.038	−0.2	5011.035	−0.7
Y9X5A	5489.147	5489.160	2.2	5489.142	−1.1
Y1X6A	1818.280	1818.281	0.5	1818.282	1.1
Y2X6A	2296.389	2296.391	0.6	2296.390	0.3
Y3X6A	2774.498	2774.493	−1.7	2774.494	−1.6
Y4X6A	3252.607	3252.599	−2.5	3252.600	−2.2
Y5X6A	3730.716	3730.703	−3.5	3730.708	−2.1
Y6X6A	4208.824	4208.816	−1.9	4208.810	−3.4
Y7X6A	4686.933	4686.933	−0.1	4686.923	−2.0
Y8X6A	5165.042	5165.056	2.8	5165.038	−0.7
Y9X6A	5643.151	ND		5643.149	−0.3
Y10X6A	6121.259	6121.272	2.1	ND
Y1X7A	1972.284	1972.285	0.9	1972.285	0.6
Y2X7A	2450.392	2450.395	1.2	2450.394	0.7
Y3X7A	2928.501	2928.503	0.7	2928.502	0.4
Y4X7A	3406.610	3406.608	−0.4	3406.608	−0.4
Y5X7A	3884.719	3884.707	−3.0	3884.709	-.5
Y6X7A	4362.827	4362.827	−0.1	4362.829	0.3
Y7X7A	4840.936	4840.928	−1.6	4840.929	−1.6
Y8X7A	5319.045	5319.046	0.1	5319.040	−0.9
Y9X7A	5797.154	ND		5797.149	−0.7
Y10X7A	6275.262	ND		6275.257	−0.8
Y1X8A	2126.287	2126.287	0.0	2126.286	−0.3
Y2X8A	2604.395	2604.396	0.0	2604.394	−0.5
Y3X8A	3082.504	3082.502	−0.8	3082.501	−1.1
Y4X8A	3560.613	3560.604	−2.6	3560.605	−2.4
Y5X8A	4038.722	4038.717	−1.1	4038.719	−0.7
Y6X8A	4516.831	4516.829	−0.4	4516.828	−0.7
Y7X8A	4994.939	4994.947	1.6	4994.937	−0.4
Y8X8A	5473.048	ND		5473.045	−0.6
Y1X9A	2280.290	2280.291	0.7	2280.291	0.4
Y2X9A	2758.399	2758.406	2.7	ND
Y3X9A	3236.507	3236.515	2.4	3236.506	−0.5
Y4X9A	3714.616	3714.614	−0.5	3714.614	−0.6
Y5X9A	4192.725	4192.717	−2.0	4192.714	−2.6
Y6X9A	4670.834	4670.842	1.9	4670.828	−1.2
Y7X9A	5148.942	ND		5148.939	−0.6
Y1X10A	2434.293	2434.292	−0.4	2434.291	−0.9

To further investigate the order of the different RUs, we selected different molecules for MS/MS experiments. As an example, the MS/MS spectrum for LTA_N2H4 with X = 5 and Y = 3 (mass = 2620.495 Da) is shown in Fig. 5. In this way, we could verify the consecutive order of the two different RU types X and Y. In Fig. 5A the complete overview of the MS/MS spectrum obtained in the negative ion mode is shown. Masses unequivocally presenting fragments occurring from a fragmentation starting from the de-O-acyl linker are labeled in blue in Fig. 5A, and the ones occurring from a fragmentation starting at the terminus are shown in red. Fragments that can exist from both cleavage directions are labeled black in Fig. 5A. In Fig. 5B, the observed fragment ions are assigned to the respective cleavage position in the molecule. A complete list of the observed fragments and their assignment is given in Table S4. Based on integration values from ¹H NMR spectra of LTA_N2H4, the chains of S. suis P1/7 LTA consist on average of 62% GroP (RU X), 33% RU Y, and 5% RU Y lacking Glc II. The values for LTA of S. suis SC84 are very similar: 57% GroP (RU X), 39% RU Y, and 4% RU Y lacking Glc II. As reference, the integral of H-1 of Glc I^LA (δ_H 5.17) was set to 1.0. As measures for the different repeats, the following signals were used for integration: H-2 of Gro (δ_H 4.08–4.04) for RU X, H-1 of Glc I (δ_H 5.21–5.18) for RU Y, as well as H-1 of Glc I lacking Glc II (δ_H 4.97–4.94; marked with # in Fig. 3). Integration of signals was done in ¹H NMR spectra obtained from LTA_N2H4 of the WT and their respective Δlgt strains and have been averaged for the evaluation of RU ratios.

Figure 5.

MS/MS analysis of de-O-acylated LTA molecule Y₃X₅A of S. suis strain P1/7 (ST1). As a representative molecule for this investigation, LTA_N2H4 Y₃X₅A, with a calculated monoisotopic mass of 2620.495 Da has been chosen. A, complete MS/MS spectrum of the doubly charged precursor m/z 1309.243 is shown (the mass region of 400–1275 Da has been magnified by a factor of 20 and the mass region of 1315 to 2200 Da by a factor of 50). Fragment ions representing part structures starting from the de-O-acyl linker are labeled in blue, and fragment ions originating from the terminus are colored in red. Fragment ions that can be produced from both cleavage directions are labeled in black. B, all observed fragment ions are assigned to the linearized structure model of the LTA. The complete list of observed fragments and their assignment is given in Table S4.

Preparation and structural analysis of the lipid anchor and LTA part structures from LTA of strain 89-1591

The analysis of the lipid anchor and LTA part structures isolated from LTA of strain 89-1591 after HF treatment was done as described above. The glycolipid anchor of this strain was identified, as for the other two strains, as kojibiose-diacylglycerol (1a in Fig. 2). However, the observed molecules representing the dephosphorylated RUs of the LTA of strain 89-1591 differed significantly from the previously observed molecules. The major present molecule was identified as α-d-Glcp-(1→1),α-d-Glcp-(1→2)-glycerol (5); 3 and α-d-Glcp-(1→2)-glycerol (6) were observed as well. Besides that, small amounts of 2 (resulting from the completely de-O-acylated glycolipid anchor), 1-O-Ala-glycerol (7), as well as unbound glycerol (4; structures for all molecules are depicted in Fig. 2) and alanine are detectable in this preparation (Fig. S8). Molecule 7 is a result of the migration of the alanine moiety from O-2 to O-1 of the glycerol after phosphodiester bond cleavage (19), indicating the substitution of some Gro-P repeats with alanine on position O-2. The unbound alanine results from the decomposition of 7 during long-term NMR measurement into 4 and alanine. All NMR chemical shift data of 5, 6, and 7 (Fig. 2) are listed in Tables S5–S7.

Structural analysis of hydrazine-treated LTA of strain 89-1591

Hydrazine treatment of LTA isolated from S. suis strain 89-1591 was performed as for the other two strains. The ¹H,¹³C HSQC NMR of LTA_N2H4 of strain 89-1591 is depicted in Fig. 6; the complete NMR chemical shift data are summarized in Table 3.

Figure 6.

NMR analysis of de-O-acylated LTA of S. suis strain 89-1591 (ST25). Shown is a section (δ_H 5.5–3.0; δ_C 105–50) of the ¹H,¹³C HSQC NMR spectrum (recorded in D₂O at 300 K) obtained from hydrazine-treated LTA of S. suis strain 89-1591 (ST25), including assignment of signals. The corresponding NMR chemical shift data are listed in Table 3.

TABLE 3.

¹H (700.4 MHz), ¹³C NMR (176.1 MHz), and ³¹P NMR (283.5 MHz) chemical shift data (δ, ppm) (J, Hz) for S. suis strain 89-1591 (ST25) LTA after hydrazine treatment (de-O-acyl LTA)

* indicates nonresolved multiplet.

The MS analysis of LTA_N2H4 of S. suis strain 89-1591 (Fig. 7A) revealed an even higher diversity of LTA molecules than observed for the other two investigated strains. As before, we observed one series of molecules that belonged to type I LTA with the observed de-O-acylated lipid anchor 1b carrying different numbers (X = 4–20) of GroP repeats. In addition, more complex versions of these LTA molecules with additionally bound hexoses have also been observed. On the one hand, these can be α-d-Glcp residues 1,2-linked to GroP, which would lead to the de-phosphorylated RU molecule 6 (Fig. 2). On the other hand, these residues can be α-d-Glcp 1,3-linked to the GroP, leading to de-phosphorylated RU molecule 3 (Fig. 2). Finally, both glycosyl attachments can be present in the same RU, leading to de-phosphorylated RU molecule 5 (Fig. 2), which was the molecule with the highest abundance observed. Because of the multitude of combinatorial possibilities, especially with regard to the nonstoichiometric α-d-Glcp substituents of the glycerol O-2 position, which has the same additional mass as one glucose moiety within the LTA chain, the present LTA_N2H4 molecules cannot be determined in such detail as for strains P1/7 and SC84. The mass spectrometric analysis depicted in Fig. 7A as well as two magnified sections of the spectrum (Fig. 7, B and C) show this increased complexity. The full list of identified LTA_N2H4 molecules for S. suis strain 89-1591 can be found in Table S8. In total, we identified more than 165 different LTA moieties in the LTA_N2H4 preparation of S. suis strain 89-1591, all of them measured with a mass deviation of ≤3.5 ppm. An evaluation of RU ratios as described above for LTA_N2H4 from ST1/ST7 strains is not possible for LTA_N2H4 of this strain, because the signal for H-2 of Gro has too much overlap with other signals in the ¹H NMR spectra.

Figure 7.

Molecular species distribution of de-O-acylated LTA of S. suis strain 89-1591 (ST25). A, charge-deconvoluted spectrum of a representative MS analysis performed in the negative ion mode (The mass region of 3000 to 6150 Da has been magnified by a factor of 5.). B and C, zoom into the regions between 1750 and 2050 Da (B) and 5135 and 5185 Da (C), including assignment of mass peaks for molecules consisting of the de-O-acyl glycolipid anchor A (1b; Fig. 2) and different numbers of GroP and Glc residues. The difference Δm = +8.05 Da corresponds to one GroP moiety less but one Glc moiety more in the overall composition. All identified molecules are listed in Table S8. Relative abundance for a spectral region was always normalized to the respective base peak.

Analysis of the fatty acid composition of S. suis LTA preparations

For the analysis of the fatty acid composition, the LTA preparations isolated from the Δlgt mutants have been used. These should give the most reliable values because most likely no other fatty acid-containing molecules are co-purified. As the most prominent fatty acid, hexadecanoic acid (16:0) was observed in all strains. Furthermore, 12:0, 14:0, 16:1, 18:0, and 18:1 have been detected. The molar ratios were determined to be 1.0/1.9/3.2/15.1/2.9/4.9 (12:0/14:0/16:1/16:0/18:1/18:0) for P1/7Δlgt LTA, 1.0/2.7/4.6/24.0/4.8/8.1 for SC84Δlgt LTA, and 1.0/1.4/3.6/15.2/6.2/6.2 for 89-1591Δlgt LTA.

Evaluation of the immunostimulatory properties of S. suis LTA preparations

The immunostimulatory properties of the different S. suis LTA preparations were characterized using murine bone marrow-derived DCs. DCs are innate immune cells known for their central role in the S. suis infection, including the production of pro-inflammatory mediators (21). The pro-inflammatory cytokines interleukin (IL)-6 and tumor necrosis factor (TNF), as well as the chemokines CXC motif chemokine ligand (CXCL) 1 and CC motif chemokine ligand (CCL) 3, were selected based on the fact that they are produced in important concentrations by DCs following S. suis infection (21). Significant levels of all four mediators were observed following activation of DCs with concentrations of 1, 3, 10, and 30 μg/ml native LTA preparations from the three S. suis WT strains (Fig. S9). The observed levels of these mediators were very similar for LTA preparations from the three S. suis strains (P1/7, SC84, and 89-1591). Only for CCL3 was an increased level at 30 μg/ml induced by LTA of strain 89-1591 observed. Moreover, levels of IL-6, CXCL1, TNF, and CCL3 were induced by all LTA preparations in a dose-dependent manner (Fig. S9).

LPs are often co-purified alongside LTA, and bacterial LPs of other Gram-positive pathogens are important activators of the innate immune response (22). Given the elevated levels of pro-inflammatory mediators produced by DCs following activation with the S. suis LTA preparations, the immunostimulatory properties of LTA preparations following treatment with H₂O₂ was evaluated. H₂O₂ oxidizes the thioether bond of LPs, which abolishes immunostimulatory activity of potentially co-purified LPs (13, 23). H₂O₂-treated LTA preparations only induced little IL-6, CXCL1, TNF, or CCL3 production following activation with 3 or 30 μg/ml, whereas nontreated LTA preparations induced significantly increased activity (Fig. 8), suggesting that co-purified LPs, but not the LTA, are important inducers of pro-inflammatory mediators by DCs, and this for all three strains of S. suis was evaluated.

Figure 8.

Pro-inflammatory mediator production by DCs following activation with the different LTA preparations from S. suis strains P1/7 (ST1), SC84 (ST7), and 89-1591 (ST25). A–D, production of IL-6 (A), CXCL1 (B), TNF (C), and CCL3 (D) 24 h following activation of DCs with 3 or 30 μg/ml of the nontreated LTA preparations from WT strains, the LTA preparations from WT strains following treatment with H₂O₂ for 24 h (H₂O₂), or the LTA preparations from Lgt-deficient mutants (Δlgt). Secreted mediators were quantified by sandwich ELISA. Data represent the mean ± S.E. (n = 3). C denotes the culture medium alone; n.d., not detected. ***, p < 0.001, indicates a significant difference between the WT and H₂O₂ or Δlgt LTA preparations.

To specifically determine the immunostimulatory potential of the LTA molecules themselves, without additive or synergistic effects of the pro-inflammatory LPs, LTA was prepared from Lgt-deficient mutants. Lgt is required for LPs to be biologically active and recognized by TLR2 (16, 20). Accordingly, activation of DCs with LTA preparations from Lgt-deficient mutants of the three S. suis strains led to a complete abrogation of pro-inflammatory mediator production, regardless of the concentration of LTA used (p < 0.001) (Fig. 8).

Taken together, these results suggested that the co-purified LPs, but not the LTA, are the main activators of DCs when using LTA preparations from the S. suis strains P1/7, SC84, and 89-1591. Because LPs are recognized by TLR2 following dimerization with either TLR1 or TLR6, which allows us to discriminate between triacyl and diacyl motifs of LPs (24), DCs derived from WT and TLR2^−/− mice were used. In accordance with the above-mentioned results, TLR2 deficiency resulted in a complete abrogation of pro-inflammatory mediator production by DCs, and this is regardless of the LTA concentration used (3, 10, or 30 μg/ml) (p < 0.001) (Fig. 9).

Figure 9.

Pro-inflammatory mediator production by WT and TLR2^−/− DCs following activation with the LTA preparations from the S. suis strains P1/7 (ST1), SC84 (ST7), and 89-1591 (ST25). A–D, production of IL-6 (A), CXCL1 (B), TNF (C), and CCL3 (D) 24 h following activation of DCs derived from WT or TLR2^−/− mice with 3, 10, or 30 μg/ml of the LTA preparations. Secreted mediators were quantified by sandwich ELISA. Data represent the mean ± S.E. (n = 3). C, denotes the culture medium alone. ***, p < 0.001, indicates a significant difference between WT and TLR2^−/− DCs.

Discussion

This study provides the first detailed structural characterization of LTA isolated from the important porcine and opportunistic human pathogen S. suis. We investigated three different serotype 2 strains: P1/7 (ST1), SC84 (ST7), and 89-1591 (ST25). In preparation for further immunological studies, we constructed the respective Δlgt mutants, which are deficient in Lgt-mediated prolipoprotein acylation (16, 22), and we analyzed their LTA as well. For all strains, the LTA isolated from Δlgt mutants had the same chemical structure as the one isolated from the respective parental WT strain.

As the glycolipid anchor, we identified in LTA of all three strains kojibiose-diacylglycerol, a glycolipid anchor that has also been found in other streptococci and closely related species like Lactococcus lactis (19, 25). When combining NMR and MS-based data, we were able to show that S. suis strains P1/7 and SC84 produce an almost identical LTA, which is most likely the same as present in S. suis serotype 2 strain 31533 (which is also an ST1 strain) as judged from a published ¹H NMR (10). This is in line with the close relationship of ST1 and ST7 strains, because they both belong to the CC1 (7). Interestingly, these strains contained two different types of LTA. One series of LTA molecules represents a type I LTA carrying only polymeric GroP chains connected to the kojibiose-diacylglycerol lipid anchor, which are identical to those LTA molecules identified in L. lactis G121 (25), as well as Streptococcus uberis 233, Streptococcus dysgalactiae 2023, and Streptococcus agalactiae 0250 (19). The second series of observed LTA molecules comprises, in addition, more complex glycosyl residue-containing RUs consisting of α-d-Glcp-(1→2)-α-d-Glcp-(1→3)-GroP. Similar RU constitutions are known from type II or type III LTA molecules (14). By MS/MS experiments, we could verify the consecutive order of the two different repeating unit types in these LTA molecules. In S. suis strains P1/7 and SC84, a defined subset of LTA molecules with 3–10 GroP repeats was found to be elongated with 1–10 repeats of the more complex glycosyl residue-containing units. To the best of our knowledge this is the first example of such a regulated synthesis of a mixed-type LTA.

In S. suis strain 89-1591, LTA molecules of two different types were also observed. The type I LTA is of the same structure as the one determined in the other two strains. The RUs of the more complex LTA type consist either of α-d-Glcp residues 1,2-linked to GroP, α-d-Glcp residues 1,3-linked to the GroP, or most prominently both of these glycosyl attachments to GroP are present in the same repeat. A multitude of combinatory possibilities, especially with regard to the nonstoichiometric α-d-Glcp substituents at the glycerol O-2 positions in these LTA molecules, makes it impossible at this stage to determine whether the order of GroP repeats and more complex RUs is as regular as in strains P1/7 and SC84. The summary of the identified LTA structures considering also identified fatty acids and the presence of the possible, nonstoichiometric alanine substitution of the O-2 position of the glycerol moieties is depicted in Fig. 10. It is important to note that strain 89-1591 is the ST25 of the CC25, which is genetically distinct from strains of CC1. This strain also possesses lower virulence than ST1 and ST7 strains (9).

Figure 10.

Chemical structures of LTA isolated from S. suis serotype 2 strains P1/7 (ST1), SC84 (ST7), and 89-1591 (ST25). From all three investigated strains, a type I LTA composed of kojibiose-diacylglycerol and polyglycerol phosphate chains was isolated. All strains are capable of modifying the glycerol O-2 position with alanine. Exclusively in strain 89-1591, a potential glycosylation at the glycerol O-2 position was additionally observed. In all strains, a second kind of LTA molecule has been identified, whereas ST-specific difference with regard to the incorporation of glycosyl residues into the complex mixed-type LTA has been observed. Strains P1/7 and SC84 are able to attach an 1,2-linked α-d-Glcp residue as branching substituent to the α-d-Glcp 1,3-linked to the GroP, whereas in strain 89-1591, this branching substituent is absent.

In our investigation, we showed that the pro-inflammatory potency of S. suis LTA molecules themselves is quite low when tested for pro-inflammatory mediator production from DCs. H₂O₂ treatment of LTA preparations as well as the use of LTA isolated from Δlgt strains lead to a complete abrogation of inflammatory activity independently of the strain, which can only be observed if LTA preparations of WT strains are used. This activation of DCs is totally TLR2-dependent and can therefore be ascribed to the LPs co-purified with the LTA. This is consistent with a study describing the LPs as the important activators of the swine innate immune system present in S. suis (11).

In summary, our study revealed an unexpected complexity of LTA molecules present in S. suis serotype 2 strains from different genetic and virulence backgrounds. In all investigated strains, two different kinds of LTA molecules have been identified, whereas an ST-specific difference with regard to the incorporation of glycosyl residues into the complex mixed-type LTA has been observed. Strains P1/7 and SC84 are able to attach an 1,2-linked α-d-Glcp residue as branching substituent to the α-d-Glcp 1,3-linked to the GroP. In strain 89-1591, an exclusive glycosylation at the glycerol O-2 position was observed. Just recently, the first enzyme required in the glycosylation process of this position in LTA of Listeria monocytogenes has been identified (26). The identification and analysis of respective homologous glycosyltransferases involved in such reactions in S. suis as well as the analysis of the impact on bacterial physiology and virulence are currently under investigation. This will further foster the recent achievements in the understanding of the biological role of teichoic acid glycosylation in Gram-positive bacteria (27).

Experimental procedures

Bacterial strains and growth

The well-characterized representative S. suis strains P1/7 (ST1; United Kingdom; pig meningitis (28)), SC84 (ST7; China; human streptococcal toxic shock-like syndrome (6)), and 89-1591 (ST25; Canada; pig septicemia/meningitis (29)) were used in this study. The different S. suis WT and mutant strains were grown as described previously in 6-liter batches (2 × 3 liters) (30) in Todd-Hewitt broth (BD Biosciences) to late logarithmic phase (A₆₀₀ ≈1) and harvested by centrifugation (10,000 × g, 40 min, 4 °C). Bacteria were resuspended in citric buffer (50 mm, pH 4.7) and heat-killed (10 min, 100 °C). Cells were then stored at −80 °C and subsequently lyophilized.

Construction of the S. suis prolipoprotein diacylglyceryl transferase (Lgt) isogenic mutants

Lgt-deficient mutants for strains P1/7, SC84, and 89-1591 were constructed as described previously (31), including DNA manipulations. Deletion of genes was confirmed by PCR and sequencing. The oligonucleotide primers used for the constructions are listed in Table S9. The growth of the different Lgt-deficient mutants was determined to be similar to that of their respective WT strains (data not shown).

Extraction and isolation of LTA

LTA purification was performed as described elsewhere (15). Yields of LTA preparations from 6 liters of bacterial culture were as follows: strain P1/7, 21.2 mg; strain P1/7Δlgt, 20.9 mg; strain SC84, 20.2 mg; strain SC84Δlgt, 30.9 mg; strain 89-1591, 18.7/15.6 mg, and strain 89-1591Δlgt, 20.6 mg.

Chemical treatments of LTA

Hydrazine treatment (to yield de-O-acyl LTA) or HF treatment (to isolate the LTA glycolipid anchor and the dephosphorylated repeats) was performed following previously described procedures (18). Notably, de-O-acylated S. suis LTA has to be desalted by dialysis against water (MWCO: 500–1000 Da) instead of performing a size-exclusion chromatography. To destroy the TLR2 activity caused by potentially co-purified LPs, LTA preparations were treated with 1% H₂O₂ for 24 h at 37 °C followed by dialysis as described previously (23).

Quantification of fatty acids

Fatty acids were extracted and quantified from LTA preparations of the Δlgt mutant strains following our earlier described procedure (18), but with n-pentadecanoic acid (15:0; Sigma) used as an internal standard. Reported data for fatty acid ratios are the mean of two independent hydrolyses of the same LTA batch, both measured as two technical replicates. Different isoforms for unsaturated fatty acids (16:1; 18:1) are reported as one sum value.

NMR spectroscopy

Deuterated solvents were purchased from Deutero GmbH (Kastellaun, Germany). NMR spectroscopic measurements were performed in D₂O or deuterated 25 mm sodium phosphate buffer (pH 5.5; to suppress fast de-alanylation) at 300 K on a Bruker Avance^III 700 MHz (equipped with an inverse 5-mm quadruple-resonance Z-grad cryoprobe). Acetone was used as an external standard for calibration of ¹H (δ_H = 2.225) and ¹³C (δ_C = 30.89) NMR spectra (32), and 85% of phosphoric acid was used as an external standard for calibration of ³¹P NMR spectra (δ_P = 0.00). Analysis of glycolipid 1 was performed in CD₃OD, and spectra were calibrated using the residual solvent peak (δ_H = 3.31, δ_C = 49.0) (32). All data were acquired and processed by using Bruker TOPSPIN V 3.0 or higher. ¹H NMR assignments were confirmed by 2D ¹H,¹H COSY and total correlation spectroscopy (TOCSY) experiments. ¹³C NMR assignments were indicated by 2D ¹H,¹³C HSQC, based on the ¹H NMR assignments. Inter-residue connectivity and further evidence for ¹³C assignment were obtained from 2D ¹H,¹³C heteronuclear multiple bond correlation and ¹H,¹³C HSQC-TOCSY. Connectivity of phosphate groups were assigned by 2D ¹H,³¹P HMQC and ¹H,³¹P HMQC-TOCSY.

Mass spectrometry

All mass spectrometric analyses were performed on a Q Exactive Plus (ThermoFisher Scientific, Bremen, Germany) using negative ion mode. LTA fractions were diluted to a final concentration of 0.03 mg/ml in propan-2-ol, water, 30 mm ammonium acetate (50:50:4, v/v/v), which was adjusted with acetic acid to pH 4.5. The HESI source was operated at −3 kV with a flow rate of 5 μl/ml using nitrogen as sheath gas at 5 atomic units. Survey MS¹ and MS² spectra were acquired with a resolution of 288,000 full width at half-maximum at m/z 200. MS² analyses were performed using the Triversa Nanomate (Advion, Ithaca, NY) as ion source applying a spray voltage of −1.1 kV and back pressure of 1.0 p.s.i. Precursor ions were selected with isolation window width of 1.5 Da, and collision-induced dissociation was performed with 20 normalized collision energies. Deconvoluted spectra were computed using Xtract module of Xcalibur 3.1 software (ThermoFisher Scientific, Bremen, Germany).

Generation of bone marrow-derived DCs and cell activation

Murine bone marrow-derived DCs were generated as described previously (21) from the femur and tibia of WT (C57BL/6J) or TLR2^−/− (B6.129-Tlr2tmKir/J) mice. Prior to activation, cells were seeded at 1 × 10⁶ cells/ml, and different concentrations of the LTA preparations or cell culture medium alone (negative control) were added. Cell supernatants were collected 24 h later for quantification of secreted IL-6, TNF, CXCL1, and CCL3 by sandwich ELISA using pair-matched antibodies (R&D Systems).

Statistical analyses

Normality of data were verified using the Shapiro-Wilk test. Accordingly, parametric (unpaired t test) or nonparametric tests (Mann-Whitney rank sum test), where appropriate, were performed to evaluate statistical differences between groups. p < 0.05 was considered statistically significant. As software, GraphPad Prism 6 was used.

Author contributions

N. G., J.-P. A., D. S., and M. G. conceptualization; N. G., D. R., J. X., D. S., and M. G. resources; N. G., D. S., and M. G. supervision; N. G., J.-P. A., S. T., D. R., and D. S. investigation; N. G. and J.-P. A. writing-original draft; N. G. and M. G. project administration; N. G., J.-P. A., D. S., and M. G. writing-review and editing.