The Baseplate of Lactobacillus delbrueckii Bacteriophage Ld17 Harbors a Glycerophosphodiesterase*

Glycerophosphodiester phosphodiesterases (GDPDs; EC 3.1.4.46) typically hydrolyze glycerophosphodiesters to sn-glycerol 3-phosphate (Gro3P) and their corresponding alcohol during patho/physiological processes in bacteria and eukaryotes. GDPD(-like) domains were identified in the structural particle of bacterial viruses (bacteriophages) specifically infecting Gram-positive bacteria. The GDPD of phage 17 (Ld17; GDPDLd17), representative of the group b Lactobacillus delbrueckii subsp. bulgaricus (Ldb)-infecting bacteriophages, was shown to hydrolyze, besides the simple glycerophosphodiester, two complex surface-associated carbohydrates of the Ldb17 cell envelope: the Gro3P decoration of the major surface polysaccharide d-galactan and the oligo(glycerol phosphate) backbone of the partially glycosylated cell wall teichoic acid, a minor Ldb17 cell envelope component. Degradation of cell wall teichoic acid occurs according to an exolytic mechanism, and Gro3P substitution is presumed to be inhibitory for GDPDLd17 activity. The presence of the GDPDLd17 homotrimer in the viral baseplate structure involved in phage-host interaction together with the dependence of native GDPD activity, adsorption, and efficiency of plating of Ca2+ ions supports a role for GDPDLd17 activity during phage adsorption and/or phage genome injection. In contrast to GDPDLd17, we could not identify any enzymatic activity for the GDPD-like domain in the neck passage structure of phage 340, a 936-type Lactococcus lactis subsp. lactis bacteriophage.

all domains of life (from bacteria to humans) (1). During degradation of the bacterial and eukaryotic cell membrane, the glycerophospholipid building blocks are first deacylated by phospholipases A 1 and A 2 , resulting in the formation of glycerophosphodiesters. These glycerophosphodiesters, which differ based on the alcohol moiety present (e.g. choline, inositol, or glycerol), are further hydrolyzed by GDPDs, producing the corresponding alcohol and sn-glycerol 3-phosphate (Gro3P). GDPDs vary in their substrate specificity, biological function, and localization inside the cell. For bacteria, GDPDs, e.g. the well characterized Escherichia coli periplasmic GlpQ and cytosolic UgpQ, play an important role in glycerophospholipid metabolism where the released alcohol moiety may act as an essential bacterial growth factor and where the produced Gro3P represents a major carbon and phosphate source (2,3). Some GDPDs are known to contribute to bacterial pathogenesis. For example, the Haemophilus influenzae GDPD, GlpQ, is a lipoprotein located in the outer membrane and contributes to bacterial pathogenesis through choline generation from the abundant pools of degradation products of the eukaryotic cell membrane. Consequent decoration of the H. influenzae cell wall with phosphorylcholine allows evasion from the host immune system through mimicry of the eukaryotic cell membrane (4). The canonical plant GDPD-encoding genes are upregulated by inorganic phosphate deprivation and have been shown to contribute to seedling growth (5) or root hair development and density (6). The seven different GDPD isoforms identified in humans display a high degree of substrate specificity (not necessarily a glycerophosphodiester) and tissue functionality. For example, human kidney GDE2 is an osmoregulated enzyme that controls the levels of the osmoprotector glycerophosphocholine (7), whereas it triggers motor neuron differentiation as part of the nervous system (8).
The present study investigates the GDPD derived from the L. delbrueckii subsp. bulgaricus group b bacteriophage 17 (phage Ld17) and the L. lactis subsp. lactis bacteriophage 340 (phage 340) and represents the first characterization of bacteriophage-encoded GDPD. Besides an in-depth biochemical and molecular analysis of the GDPD of phage Ld17 (GDPD Ld17 ), we studied substrate specificity and highlight the importance of the native GDPD Ld17 during phage infection.

Results
The Presence of a GDPD Domain in Structural Phage Proteins with Distinct Topologies-The structural module of the genomes of all known group b L. delbrueckii-infecting phages and of the L. lactis subsp. lactis-infecting phages 340 and 645 contains a gene that encodes a protein containing a GDPD(-like) domain. Because of the high protein identity of these putative GDPD-containing proteins within the group b Ldb-infecting phages (93-95% identity) and between the two L. lactis ssp. lactis-infecting phages (phages 340 and 645; 99% identity), the GDPDs of phage Ld17 (designated here as GDPD Ld17 ) and 340 (named GDPD 340 ) were selected as representatives of the respective phage groups.
Bioinformatics analysis revealed an apparent tripartite domain structure in both GDPDs (Fig. 1A): (i) an N-terminal domain that is presumably required for physical association with the phage particle, (ii) a putative carbohydrate-binding domain, and (iii) the GDPD(-like) domain with the latter two domains displaying an inverse arrangement in GDPD Ld17 and GDPD 340 . NCBI Conserved Domain Detection analysis (20) identified a DUF2479 domain (residues 23-167; 9.76E Ϫ03 ) within the N terminus of GDPD 340 ; this domain is a typical feature of the neck passage structure (NPS) of lactococcal 936-  (20). The predicted catalytic domains, GDPD (cd08556; 5.11E Ϫ38 ) and GDPD_like_2 (cd08582; 7.40E Ϫ15 ) are indicated as dark gray boxes. Sequence similarity (GDPD Ld17 , 9.86E Ϫ27 ; GDPD340, 2.65E Ϫ09 )) to the GDPD UgpQ of E. coli is depicted. The predicted carbohydrate-binding (CB) domain (CBM_4_9; 9.76E Ϫ03 ) of GDPD Ld17 and the putative carbohydrate-binding domain (residues 169 -618) of GDPD 340 consisting of three Ϯ124amino acid repeating units and one partial repeating unit, as predicted by the Trust server (22), are indicated by black boxes. The extreme N-terminal domain of GDPD 340 , probably involved in physical association with the virion, contains a DUF2479 domain (light gray box; 9.76E Ϫ03 ), which is typically found in the neck passage structure of lactococcal 936-type phages. The double arrows delineate the "GDPD" and the "GDPD ϩ CBM modules" of GDPD Ld17 cloned and designated further as the "catalytic" and "catalytic ϩ carbohydrate-binding" modules. B, ClustalW comparison of the UgpQ-like GDPD domains (amino acids in light gray were excluded from the NCBI Conserved Domain Detection delineation). The putative catalytic histidine residues are marked in black, and the glutamic acid and aspartic acid residues involved in binding of the divalent cations are marked in dark gray. Identical residues are marked by an asterisk, and residues marked with a point and colon display chemical and physical similarity, which is higher for the latter. Gaps, indicated by horizontal lines, were introduced into the sequences to maximize the alignment. type phages to which phage 340 belongs (21). Transmission electron micrographs of immunogold-labeled GDPD 340 confirm the presence of GDPD 340 as part of the phage 340 NPS (Fig.  2B). No conserved domains were detected in the N terminus of GDPD Ld17 , whereas immuno-EM using anti-GDPD Ld17 antibodies revealed that the GDPD protein of Ld17 is associated with the baseplate of phage Ld17 ( Fig. 2A).
Although a conserved carbohydrate-binding domain (CBM_4_9; residues 664 -788; 9.76E Ϫ03 ) is present in the C terminus of GDPD Ld17 , we hypothesize that a carbohydratebinding function is present in the central domain of GDPD 340 based on the identification of a Ϯ124-amino acid repeating unit using the Trust server (22). The full repeat region (residues 169 -618) consisting of three full repeating units of Ϯ124 amino acids and one partial unit suggests interaction with a (repeated) component of the bacterial cell envelope and is therefore assigned as the putative carbohydrate-binding domain of GDPD 340 .
Sequence similarity between GDPD Ld17 and GDPD 340 is restricted to their GDPD(-like) domains, which share 28 and 20% amino acid identity, respectively, with the well characterized cytoplasmic E. coli protein UgpQ. The GDPD domain (residues 378 -613) of GDPD Ld17 belongs to the large family of the canonical prokaryotic and eukaryotic GDPDs (EC 3.1.4.46) represented by UgpQ. ClustalW alignment (Fig. 1B) highlights in the GDPD-domain of GDPD Ld17 the presence of five conserved amino acid residues, two catalytic histidine residues (His-381 and His-421 in GDPD Ld17 ) and three residues (Glu-408, Asp-410, and Glu-491) for binding a divalent cation, which are believed to be involved in the catalytic metal ion-dependent acid-base reaction mechanism to hydrolyze glycerophosphodiester linkages (23). The GDPD-like domain (residues 635-795) of GDPD 340 belongs to a subfamily of uncharacterized bacterial GDPDs (cd08582; 7.40E Ϫ15 ); curiously, ClustalW alignment identified just two (His-669 and Glu-737) of the five conserved residues that are associated with the typical GDPD-mediated reaction mechanism.
Structural Characteristics-Consistent with the immunogold electron microscopy findings (Fig. 2), mass spectrometric analyses of Ld17 (17) and 340 (supplemental Table S2) virions indicates the presence of the full-length GDPDs as part of the structural phage particle. Full-length GDPD Ld17 and GDPD 340 were recombinantly expressed and purified using the L. lactis NZ9000-pTX8048 expression system (24). Mass calculations using SEC-MALS-RI analyses of the purified recombinant GDPD Ld17 and GDPD 340 proteins revealed the presence of complexes with estimated masses of 325 and 326 kDa, respectively, which correspond to the homotrimeric native conformations (theoretical molecular masses of 111.9 and 110.7 kDa for GDPD Ld17 and GDPD 340 monomers, respectively) (Fig. 3A). Moreover, GDPD Ld17 appears to form a sodium dodecyl sulfate (SDS)-resistant homotrimer at room temperature (Fig. 3B). An estimated molecular mass of ϳ335 kDa, consistent with a homotrimer, was observed when unheated samples of GDPD Ld17 were loaded on a 7% (w/v) SDS-polyacrylamide gel, whereas a bandconsistentwiththetheoreticalmolecularmassofthemonomer (111.9 kDa) was observed when the protein sample was boiled for 5 min prior to loading on the SDS-polyacrylamide gel. The GDPD 340 protein does not display this SDS resistance characteristic at room temperature and appears to disassemble Phages were incubated with polyclonal anti-GDPD rabbit antibodies raised against the GDPD specific for that phage and labeled with anti-rabbit secondary mouse antibodies conjugated to 5-nm gold particles. into its monomers with or without a heating step prior to or upon electrophoresis.
Enzymatic Activity and Substrate Preference of GDPD Ld17 -During metabolism or pathogenesis, bacterial GDPDs are known to hydrolyze deacylated phospholipids, called glycerophosphodiesters, into the corresponding alcohol and Gro3P. Therefore, we analyzed in the first instance the ability of the bacteriophage-encoded GDPDs to degrade glycerophosphorylcholine (GPC), a glycerophosphodiester, with HPAEC-PAD analysis in parallel with an enzyme-coupled spectrophotometric assay quantifying the formation of Gro3P. The specific GDPD activity of GDPD Ld17 , expressed as ⌬A 340 nm /(mol of GDPD ϫ min) using 2.5 mM GPC substrate after 30-min incubation at 30°C was 7466.67 Ϯ 108.44 (supplemental Fig. S1A). In parallel, the HPAEC-PAD chromatogram showed one additional peak at 42 min after GDPD Ld17 treatment of the GPC substrate that we attribute to the formation of Gro3P (data not shown).
As the bacterial GDPDs perform a second step in cell membrane degradation subsequent to deacetylation of the glycerophospholipids by phospholipase A, we presumed that these glycerophosphodiesters are not the natural substrates of the bacteriophage-encoded GDPD but that this activity is serendipitous. Because bacteriophage-encoded GDPDs, which are a structural component of the virion, are most likely to interact with phosphodiester-containing substrates (e.g. teichoic acids (25) or other surface-associated carbohydrates (26)) in the host cell envelope, we implemented an extraction procedure to concentrate cell wall-associated carbohydrates (CWaCs) excluding peptidoglycan.
HPAEC-PAD analysis of the CWaC extract of Ldb17, the bacterial host of phage Ld17, displayed an extra peak at 42 min following GDPD Ld17 treatment compared with the negative buffer control (data not shown). This peak corresponds to the Gro3P peak liberated in the presence of the GPC substrate. Using an enzyme-coupled spectrophotometric assay specific for the detection of Gro3P, the GDPD Ld17 activity on the Ldb17 CWaC extract was quantified as 2071.11 Ϯ 37.91 ⌬A 340 nm /(mol of GDPD ϫ min). These results imply the presence of carbohydrate substrates in the Ldb17 cell envelope that, following GDPD Ld17 hydrolysis, generate a reaction product identical to that generated during GPC hydrolysis, namely Gro3P.
To gain further insight into the substrate specificity of GDPD Ld17 , a spectrum of purified carbohydrates after treatment with GDPD Ld17 (in parallel with a buffer control) was analyzed with HPAEC-PAD. According to HPAEC-PAD anal- Enzymatic activity of GDPD Ld17 on purified carbohydrates GDPD activity expressed as ⌬〈 340 nm /(mol of GDPD ϫ min) was evaluated using an enzyme-coupled spectrophotometric assay on 1.5 g/liter purified carbohydrates. Average and S.D. of three repeats are given. HPAEC-PAD chromatograms of purified carbohydrates (2.5 g/liter) exposed to GDPD Ld17 at 30°C for 2 h detected only Gro3P as the end product of GDPD Ld17 activity.

Chemical structure
Ref. yses (Table 1), the sole end product of GDPD Ld17 hydrolysis is Gro3P, which is removed from the poly(glycerol phosphate) CWTA backbone, the Gro3P-decorated linear D-galactan, and the bacterial GPC. The single Gro3P peak observed in the HPAEC-PAD chromatograms of all poly(glycerol phosphate) CWTA exposed to GDPD Ld17 at any time point in the incubation when measuring GDPD Ld17 activity in the linear activity range or when excess GDPD Ld17 protein (10-fold the amount compared with the highest amount used in the linear range of activity) is applied during longer incubation periods (up to 2 h) suggests exolytic activity for GDPD Ld17 whereby Gro3P units are released one by one from the CWTA ends. Furthermore, the absence of peaks other than that corresponding to Gro3P and the correlation between the observed Gro3P peak volume in the HPAEC-PAD chromatograms and the Gro3P quantification by an enzyme-coupled spectrophotometric assay suggests that decorated, typically by glycosylation or D-alanylation for CWTA, Gro3P units in the CWTA backbone are not removed. This indicates that GDPD Ld17 -mediated degradation of CWTA is terminated at a substituted Gro3P unit. This notion is consistent with the low GDPD Ld17 activity (11.11 Ϯ 0.77) observed for Staphylococcus epidermidis RP62A CWTA (28), which has a higher degree of substitution compared with S. aureus NM8m  (31), are insensitive to GDPD Ld17 hydrolysis, thus indicating a requirement for glycerol adjacent to the phosphodiester bond (Table 1). However, lipoteichoic acid (LTA) is not degraded by GDPD Ld17 despite the fact that CWTA and LTA share a similar 133 poly(glycerol phosphate) backbone. It should be noted that CWTA possesses a Gro3P repeating unit, whereas the repeating unit (and thus the unit available for GDPD Ld17 at the end) of the LTA backbone is sn-glycerol 1-phosphate, which has an opposite stereochemistry on C2 of the glycerol unit compared with Gro3P. The absence of any degradation in the HPAEC-PAD chromatogram of purified LTAs following GDPD Ld17 treatment indicates a strict preference for the Gro3P configuration.

GDPD activity HPAEC-PAD
Biochemical and Structural Requirements for GDPD Ld17 Activity-The optimal biochemical characteristics for GDPD Ld17 activity were assayed on Ld17 CWaC extract (3 g/l) in the linear range of enzymatic activity using an enzyme-coupled FIGURE 4. NMR analysis of the GDPD Ld17 -treated Gro3P-substituted linear D-galactan (sPS2) of Ldb17. Chemical structure (27) and part of the heteronuclear single quantum coherence spectrum of the Ldb17 sPS2 (green-black) and GDPD Ld17 -treated Ldb17 sPS2 (red-violet). Note the absence of Gro and AЈ signals in the GDPD Ld17 -treated sPS2 spectrum. spectrophotometric assay (supplemental Fig. S3). The highest GDPD Ld17 activity is observed in the temperature range of 30 -42°C at pH 8 in the presence of 10 mM Ca 2ϩ . Although Mg 2ϩ gradually inhibits GDPD Ld17 activity upon increasing concentration with 25% relative activity at 100 mM MgCl 2 , other divalent cations (Mn 2ϩ , Co 2ϩ , Ni 2ϩ , Cd 2ϩ , and Zn 2ϩ ) display a stronger inhibitory effect with complete absence of enzymatic activity at 5 mM.
Because most bacterial GDPDs in the NCBI database are of similar size (30 -40 kDa) as the GDPD domain of GDPD Ld17 , the individual central GDPD domain alone (48.8 kDa) or this domain in combination with the putative carbohydrate-binding domain (73.8 kDa) was recombinantly expressed using the pTX8048 lactococcal expression vector, and their enzymatic activity was evaluated using an enzyme-coupled spectrophotometric assay on the Ldb17 CWaC extract and GPC. Both truncated derivatives had lost more than 98% activity in comparison to the full-length GDPD Ld17 protein (supplemental Fig. S1A) as well as the ability to form homotrimers resistant to SDS at room temperature (supplemental Fig. S1B). As trimerization of the truncated proteins was not analyzed by size exclusion chromatography, one may argue that the ability to form homotrimers was already lost upon truncation. Nevertheless, loss of (SDS-resistant) trimerization upon truncation and near complete loss of enzymatic activity of the truncated domains indicate the involvement of the deleted regions in both processes. Individual site-directed mutagenesis of each of the five conserved residues (His-381, Glu-408, Asp-410, His-423, and Glu-491; Fig. 1B) of GDPD Ld17 putatively involved in a typical GDPD-reaction mechanism reduced the total activity to below 3%, both on the GPC substrate and on the (assumed) native substrate (Ld17 CWaC-extract) for each of the five generated mutant proteins. SDS-PAGE without prior boiling of the protein sample indicates that these five mutagenized proteins are still able to form SDS-resistant homotrimers (supplemental Fig. S1B), implying no major conformational changes of the mutagenized proteins and affirming the involvement of the five conserved residues within GDPD Ld17 in the hydrolysis of the typical glycerophosphodiester and of the native Ld17 carbohydrates.
Absence of Detectable Enzymatic Activity for GDPD 340 -The HPAEC-PAD chromatograms did not show any activity of GDPD 340 on GPC and on CWaC-extracts of the phage 340 host, IL1403, or of other lactococcal and Lactobacillus strains (data not shown). Also an enzyme-coupled spectrophotometric assay with higher sensitivity than HPAEC-PAD for Gro3P detection could not detect any GDPD activity. This absence of enzymatic activity for GDPD 340 suggests that this protein does not possess GDPD activity, which may have been caused by mutations that altered three of the five conserved residues implicated in the GDPD catalytic reaction mechanism (Fig. 1B).
GDPD Activity and Phage Ld17 Characteristics-The cation Ca 2ϩ is essential for enzymatic activity of the recombinantly expressed GDPD Ld17 , whereas other divalent cations inhibit enzymatic activity (see above). To evaluate the importance of the native GDPD activity in the baseplate of the Ld17 virion for the phage infection process, different phage characteristics (native GDPD activity, adsorption, and efficiency of plating) were assessed in the presence of a diverse set of 20 mM divalent cations ( Table 2). Similar to the recombinantly expressed GDPD Ld17 , the native GDPD activity of the Ld17 virion is stimulated by the addition of Ca 2ϩ ions (relative activity of 127%), whereas the other divalent cations have an inhibitory effect. Mg 2ϩ is again the less potent inhibitor with 33% activity remaining compared with the complete absence of GDPD activity after addition of any other inhibiting divalent cation. In parallel with the GDPD activity, the adsorption and the relative efficiency of plating of phage Ld17 on the host Ldb17 are optimal with values of 91% and 1.00, respectively, in the presence of 20 mM CaCl 2 . Adsorption and plaque formation are less efficient in the presence of the other inhibiting divalent cations with only 71% adsorption and a relative efficiency of plating of 0.57 for Mg 2ϩ .
Further evaluation of the enzymatic activity of the recombinantly expressed GDPD Ld17 on CWaC extracts of a diverse set of 12 L. delbrueckii subsp. bulgaricus strains (Table 3)  The native GDPD activity present in the Ld17 virion was assayed in 50 mM MOPS (pH 7.5), 300 mM NaCl on 2.5 mM GPC substrate after incubation at 30°C for 2 h with the enzyme-coupled spectrophotometric assay. 100% activity is set equal to the condition without additional divalent cations. Adsorption of phage Ld17 on its host Ldb17 is expressed as the ratio of adsorbed phage after 10 min to the number of free phages at time point 0. The relative efficiency of plating (EOP) of phage Ld17 on its host Ldb17 was compared with the optimal condition containing 20 mM CaCl 2 . Relative efficiency of plating was not determined (ND) for Cd 2ϩ and Zn 2ϩ due to the absence of bacterial growth of the Ldb17 host under these conditions. For each condition, the average and S.D. of three independent experiments is given.

Discussion
The GDPD of bacteriophage Ld17 displays the characteristics of the canonical GDPD activity as identified in the literature (1): enzymatic activity requires (i) a divalent cation(s) and (ii) a conserved set of five amino acids for a metal ion-dependent, acid-base reaction mechanism to hydrolyze a (iii) glycerophosphodiester substrate to Gro3P and an alcohol. GDPD Ld17 solely requires Ca 2ϩ for activity, whereas other divalent cations are inhibitory for activity. This is probably a reflection of an adaptation to the high Ca 2ϩ levels present in the dairy environment from which the Ld17 phage was isolated (17). Also the optimal temperature range (30 -42°C) for activity reflects the preferred growth conditions of the Lactobacillus host. Besides degradation of the simple glycerophosphodiester GPC, GDPD Ld17 also elicits activity toward more complex substrates, the Gro3P decoration of the Ld17 D-galactan polysaccharide and the poly(glycerol phosphate) backbone of CWTA, in all cases releasing Gro3P. Degradation of CWTA occurs according to an exolytic mechanism, and substitution of the CWTA backbone is presumed to be inhibitory for GDPD Ld17 activity. Although GDPD hydrolysis of more complex substrates has not been identified before for the canonical GDPDs, the GDPD Ld17 activity is still in congruence with the recognition of the glycerophospho moiety by the canonical GDPDs and the formation of Gro3P end product (23,32): the three conserved residues (Glu, Asp, and Glu) in the catalytic groove coordinate a divalent cation, which stabilizes the cyclic phosphate intermediate formed by glycerol and phosphate during a two-step reaction mechanism with the two catalytic histidine residues acting as a general acid and base. The involvement of the glycerol unit in the catalytic reaction mechanism explains the absence of any hydrolyzing activity by GDPD Ld17 on carbohydrates with phosphodiester bonds not adjacent to a glycerol unit, such as the L. lactis MG1363 pellicle and poly(ribitol phosphate) TA. Moreover, the reaction mechanism is specific for sn-glycerol 3-phosphate, the repeat unit of the CWTA backbone, whereas the stereoisomer sn-glycerol 1-phosphate, the repeat unit of the poly(glycerol phosphate) backbone of lipoteichoic acid, is not recognized by GDPD Ld17 .
CWTA degradation by bacterial enzymes has so far only been described in, now vintage, research articles by . During sporulation, B. subtilis strain Marburg produces a TA-degrading enzyme with absolute substrate specificity for the endogenous ␣-glucosylated glycerol TA of the cell, producing the ␣-D-glucose 132 (sn)glycerol 3-phosphate according to an exolytic mechanism (33,34). Phosphatestarved Bacillus pumilus DSM27 cells produce a TA-degrading enzyme that produces Gro3P using unsubstituted TA of the poly(glycerol phosphate) type as substrate, whereas the enzyme is unable to remove substituted glycerol phosphate (46). Although both bacterial TA-degrading enzymes display GDPD-like activity, the question remains whether the canonical bacterial GDPDs also display this teichoic acid-degrading activity. Further research of the bacterial GDPD substrate specificity is, as such, required.
Immuno-EM analysis showed that GDPD Ld17 is part of the baseplate of phage Ld17. GDPD Ld17 activity removes the Gro3P decoration of the D-galactan polysaccharide, the major charged carbohydrate associated with the Ldb17 cell wall, and can further degrade the short chain partially glycosylated oligo(glycerol phosphate) CWTA, which is a minor cell envelope component (27). Within bacteriophages, a very diverse set of enzymatic activities within baseplate-associated proteins, tail spikes, or fibers that interact with a component of or are closely associated with the bacterial cell envelope have been identified. These virion-associated enzymes (partially and locally) degrade the (corresponding) bacterial cell envelope component to accomplish accessibility to the underlying primary bacterial receptor to which the bacteriophage will irreversibly adsorb. To this group belong structural enzymes degrading lipopolysaccharide (36 -38), exopolysaccharide (39 -41), and capsule (42)(43)(44) structures and a small set of capsule deacetylases (45,46). Furthermore, peptidoglycan-degrading enzymes associated with the tail tip of the virion ensure injection of the phage genomic DNA into the bacterial cell (47)(48)(49)(50)(51). Within the baseplate-encoding region of phage Ld17, no peptidoglycan-degrading activity was predicted, and because the carbohydrate substrates of GDPD Ld17 are newly identified, it is difficult to infer whether GDPD activity of phage Ldb17 is involved in rendering accessibility to the bacterial receptor and/or injection of the phage genomic DNA into the cytosol. Moreover, the Ca 2ϩ requirement for optimal GDPD Ld17 activity and for efficient phage Ld17 adsorption and plaque formation suggests a general adaptation of the enzymatic activities of phage Ldb17 and the different steps of its infection cycle to a Ca 2ϩ -rich environment or a dependence of adsorption and/or genomic DNA injection on GDPD Ldb17 activity, resulting in a more efficient infection cycle.
L. delbrueckii subsp. bulgaricus is a commonly used in starter cultures in the commercial production of fermented milk products (yogurts and cheeses) where it contributes to the flavor profile of the final product (52). Infection of starter cultures during dairy fermentation by bacteriophages may disable (part of) the starter culture, thereby negatively influencing the fermentation process and the organoleptic characteristics of the dairy product. Despite its industrial significance, only one specific bacteriophage-host interaction between a L. delbrueckii strain, ATCC 15808, and its infecting phage, the group a L. delbrueckii phage LL-H, has been studied in depth (53)(54)(55)(56). However, group b L. delbrueckii-infecting phages represent the most important genotype infecting L. delbrueckii (17). All group b L. delbrueckii-infecting phages possess a structural GDPD that shows 93-95% amino acid identity to GDPD Ld17 , and all have a full set of the conserved five amino acids involved in the canonical GDPD activity. Also all tested CWaC extracts of L. delbrueckii subsp. bulgaricus strains were shown to be highly sensitive for GDPD activity ( Table 3). The absence of any detectable LTA and only limited amounts of CWTA (27), which is susceptible to GDPD hydrolysis, further suggest bacteriophage adsorption to a cell wall component different from TA, whereas most bacteriophages infecting Gram-positive bacteria (e.g. Staphylococcus species (57,58), B. subtilis (59 -61), Listeria monocytogenes (62,63), and the group a L. delbrueckii phage LL-H (64) adsorb to TA structures. As such, further research unraveling the precise role of the GDPD during the infection process of group b L. delbrueckii phages will provide us with new insights for strategies to prevent lysis of starter cultures during dairy fermentation.
We were not able to identify any enzymatic activity for GDPD 340 . In comparison with the canonical GDPDs, GDPD 340 seems to have lost three of the five amino acid residues essential for canonical GDPD activity. GDPD 340 is part of the NPS of phage 340, a 936-type phage. The lactococcal 936-type phages are a well studied phage group because of their importance for the dairy industry, often (negatively) influencing dairy fermentation processes. The "older" 936-type phage isolates have no NPS (e.g. sk1 and p2) or a truncated NPS without any enzymatic domain (e.g. bIL140 and P008). More recent phage isolates from the dairy environment frequently possess a large NPS with a potential enzymatic domain, a GDPD-like domain (phages 340 and 645; Refs. 18 and 19), or an SNGH hydrolase domain (21,65,66). These "novel" enzymatic activities within the 936-type NPS are probably beneficial, but not essential, for bacteriophage infection in the dairy environment. It therefore appears that such enzymatic activity merely constitutes an accessory function that can be lost without loss of phage viability as can be hypothesized for GDPD 340 . Nonetheless, GDPD 340 may conversely represent a non-canonical GDPD with activity different from the typical hydrolysis of glycerophosphodiesters, and the reaction conditions used in this study may have been unsuitable for detection of this enzymatic activity. For example, the mammalian GDE4 and GDE7 degrade glycerolysolipids (1), the GDPD of Mycoplasma hyorhinis displays, besides a typical GDPD activity, also nonspecific phospholipase A activity (68). Furthermore, Simocková et al. (69) identified phospholipase C activity that controls the phosphatidylglycerol content of the cell membranes in the bioinformatically annotated GDPD-like protein Pgc1p of Saccharomyces cerevisiae.
Adsorption Assays-A 1-ml sample was taken of a bacterial suspension grown to midexponential phase (A 600 nm of 0.550; Ϯ5 ϫ 10 7 CFU/ml) and centrifuged (15, 000 ϫ g, 10 min). The bacterial pellet was resuspended in 900 l of MRS broth supplemented with 1% (w/v) lactose, 0.5% (w/v) tryptone, and a 20 mM concentration of the relevant divalent cations prewarmed at 42°C. 100 l of 5 ϫ 10 5 pfu phages were added to generate a final multiplicity of infection of 0.02. Following incubation at 42°C for 10 min, the bacterial host cells (including irreversibly adsorbed phages) were removed by centrifugation (15,000 ϫ g, 7 min). The supernatant was titrated using the double agar overlay method with MRS soft and solid agar supplemented with 20 mM CaCl 2 to determine the number of non-adsorbed or reversibly adsorbed phages.
Analysis of Structural Phage 340 Proteins-The structural proteins of the L. lactis-infecting phage 340 were extracted from a CsCl-purified phage suspension (10 10 pfu) using the methanol/chloroform (1:1:0.75, v/v/v) extraction protocol described by Moak and Molineux (48). The resulting protein pellet was resuspended in an SDS-PAGE loading buffer, boiled for 10 min, and loaded onto a standard 12% polyacrylamide gel (71). The structural proteome lane on the polyacrylamide gel was cut into slices that were individually subjected to trypsin digestion and peptide extraction according to Lavigne et al. (72). The collected peptides were analyzed by electrospray ionization-MS/MS.
Cloning of GDPD and Derivatives-Purified bacteriophage genomic DNA served as template for amplification of the coding regions (of the truncated derivatives) of the (putative) glycerophosphodiester phosphodiesterase-encoding genes using the KOD high fidelity DNA polymerase (Novagen, UK) and specific primer pairs (Eurofins MWG, Germany) (supplemental Table S1). Phage genomic template DNA was prepared by infecting L. lactis subsp. lactis IL1403 and L. delbrueckii subsp. bulgaricus 17 at an A 600 nm of 0.15 with 340 and Ld17, respectively, followed by incubation until lysis was observed. The resulting lysate was passed through a 0.45-m filter (Sarstedt, Germany) followed by total genomic DNA extraction (73). PCR products were digested with BamHI (Roche Applied Science) and XbaI (Roche Applied Science), cloned into the similarly digested expression vector pTX8048 (24), and finally ligated using T4 DNA ligase (Promega). Ligation mixtures of PCR fragments and pTX8048 were introduced into L. lactis subsp. cremoris NZ9000 by electroporation, and resulting transformants were selected on GM17 agar plates supplemented with 5 g/ml chloramphenicol. The generated constructs were verified by DNA sequencing using vector-specific and internal sequencing primers (supplemental Table S1).
Site-directed Mutagenesis of Amino Acid Residues Involved in the Catalytic Reaction Mechanism-Site-directed mutagenesis was performed by PCR using the template plasmid pTX8048gdpd and the respective primer couples Ld17_XxxxAf and Ld17_XxxxAr (the catalytic residue, X, on position xxx of the protein sequence was replaced by an alanine residue, Ala) (supplemental Table S1). The "overhang" primers consist of 11 nucleotides 5Ј of the mutated triplet and 27 nucleotides 3Ј. The template DNA was extracted using a GeneJET Plasmid Miniprep kit (Thermo Scientific) from dam ϩ E. coli MC1000 to ensure its methylation. PCR amplification was performed in 50-l volumes with ϳ10 ng of template, 1 unit of KOD Hot Start DNA polymerase (EMD Chemicals, Gibbstown, NJ), 1ϫ KOD Hot Start DNA polymerase buffer, 200 M dNTPs, 1.5 mM MgSO 4 , and a 0.5 M concentration of the respective primers. Cycling conditions consisted of an initial polymerase activation (95°C, 2 min) followed by 95°C for 20 s, 60°C for 10 s, and 70°C for 3 min for 35 cycles followed by a final extension for 7 min. PCR-generated amplicons were purified using the GeneJET PCR Purification kit (Thermo Scientific) followed by DpnI treatment (Stratagene) at 37°C for 3 h to restrict all methylated template plasmid. After a final purification (GeneJET PCR Purification kit), the mutated pTX8048-gdpd was transformed into electrocompetent L. lactis subsp. cremoris NZ9000 and selected for on GM17-Cm 5 agar plates. After plasmid extraction, the mutations were verified by DNA sequencing using vector-specific and internal sequencing primers (supplemental Table S1).
Recombinant Protein Expression and Purification-Protein expression using the nisin-inducible expression system of pTX8048 provides an N-terminal protein fusion to thioredoxin to improve protein solubility and to a His 6 tag for protein purification purposes (25). Expression occurred at 30°C for 3 h in L. lactis subsp. cremoris NZ9000 in GM17 broth supplemented with 5 g/ml chloramphenicol after induction at an A 600 nm of 0.20 with 40 ng/ml nisin using Nisaplin powder (2.5% nisin; Danisco, UK) dissolved in sterile water. Cells were harvested from suspension by centrifugation (8,700 ϫ g, 10 min, 4°C). The bacterial pellet was washed once in lysis buffer (10 mM Tris-HCl (pH 7.5), 0.3 M NaCl, 10 mM CaCl 2 ) and finally stored overnight at Ϫ20°C.
The bacterial pellet was thawed at room temperature, resuspended in lysis buffer (10 mM Tris-HCl (pH 7.5), 0.3 M NaCl, 10 mM CaCl 2 ) supplemented with 30 mg/ml hen egg white lysozyme (Sigma-Aldrich), and incubated at 20°C for 30 min. The bacterial suspension was subsequently placed on ice and disrupted with glass beads in a Mini Bead Beater (Biospec Products, Bartlesville, OK) for 90 s followed by 5 min on ice for three cycles. Cell debris and insoluble components were discarded by centrifugation (25,000 ϫ g, 30 min, 4°C). Purification of the recombinant N-terminal His 6 -tagged proteins was performed using the nickel-nitrilotriacetic acid-agarose (Qiagen, UK) according to the manufacturer's protocol with the wash and elution buffers composed of 10 mM Tris-HCl (pH 7.5), 0.3 M NaCl, 10 mM CaCl 2 supplemented with 20 and 150 mM imidazole, respectively.
Protein purity was at least 95% as estimated by SDS-PAGE. Protein concentration was estimated by the Bradford assay (74). Purified protein was dialyzed overnight against a volume of buffer at least 1000ϫ the volume of the protein sample using Slide-A-Lyzer MINI dialysis units (Pierce).
Immunoelectron Microscopy-Purified phage preparation was dialyzed for 15 min against TGB buffer (200 mM Tris-HCl (pH 7.5), 500 mM glycine, 2% (v/v) butanol) using a Novagen D-Tube TM dialyzer (molecular mass cutoff, 3.5 kDa). Dialyzed phages were put over a nickel glow-discharged grid for 30 min and incubated overnight (20°C) with primary antibody solution diluted 1:90 in TGB buffer. The grid was washed then in TGB buffer and incubated for 1 h in a 1:40 dilution of the secondary antibody solution (goat anti-rabbit immunoglobulin G 5-nm gold conjugate solution). Fixation was achieved using 0.25% (v/v) glutaraldehyde for 20 min in phosphate-based saline buffer at room temperature. The grid was washed and blotted five times in filtered dialyzed water. Finally, the samples were stained with 2% phosphotungstic acid (pH 7.0) for about 30 s and observed using a Tecnai Spirit electron microscope operated at 120 kV and a 2,000 ϫ 2,000-pixel charge-coupled device camera. Polyclonal antibodies against the recombinantly purified GDPD Ld17 and GDPD 340 were raised in rabbits by Davids Biotechnologie GmbH (Regensburg, Germany).
Extraction of Cell Wall-associated Carbohydrates-A culture flask of 800 ml of fresh medium was inoculated with a 1% bacterial overnight culture and grown overnight (Ϯ20 h) at the appropriate temperature. Cells were harvested by centrifugation (8,700 ϫ g, 15 min) and washed three times in 20 mM NH 4 C 2 H 3 O 2 ⅐AcOH (pH 4.7). The bacterial pellet was resuspended in 20 mM NH 4 C 2 H 3 O 2 ⅐AcOH (pH 4.7) supplemented with 4% SDS and boiled in a water bath with intensive stirring for 1.5 h. After cooling the bacterial suspension to room temperature, the bacterial cells were collected by centrifugation (8,700 ϫ g, 15 min), and the cell pellet was washed three to five times with 20 mM NH 4 C 2 H 3 O 2 ⅐AcOH (pH 4.7) until no SDSmediated foaming was observed. The pellet was resuspended in 50 ml of 5% trichloroacetic acid and stirred for 48 h at 4°C in the presence of acid-washed glass beads (Sigma-Aldrich). Insoluble material was removed by centrifugation (8,700 ϫ g, 15 min), and the supernatant was dialyzed against water at 4°C. Subsequently, the dialysate was lyophilized, resuspended in 5 ml of H 2 O, and centrifuged once more at high speed (25,700 ϫ g, 20 min) to remove all residual insoluble material. The supernatant was collected, lyophilized, and stored at room temperature.
Purified Carbohydrates-LTA material from S. aureus and B. subtilis was purchased from Sigma-Aldrich. The LTA of L. lactis subsp. cremoris MG1363 was prepared by extraction with aqueous butanol according to Morath et al. (76). The extracellular teichoic acid of L. monocytogenes ScottA was purified as described by Brauge et al. (77). Extracellular teichoic acid is defined as the TA material embedded in the biofilm matrix, thereby differentiating it from the TA material present in the culture supernatant or in the cell envelope (77). Purification of the CWTAs of the S. epidermidis strains 1457 and RP62A and of the S. aureus strain MN8m was achieved as described previously by Sadovskaya et al. (28) and Vinogradov et al. (29), respectively. The cell wall-associated polysaccharide (polysaccharide pellicle), consisting of hexasaccharide repeating units linked via phosphodiester bonds, of MG1363 was purified according to Chapot-Chartier et al. (26). Extraction and purification of the individual cell wall-associated carbohydrates (sPS1, sPS2, and CWTA) of Ldb17 was achieved according to Vinogradov et al. (27). CWTA was additionally repurified on a BioGel P-2 column.
Monosaccharide Analysis of S. epidermidis 1457 and 5 CWTAs-The CWTA was treated with 48% hydrofluoric acid (4°C, 24 h), hydrolyzed with 4 M trifluoroacetic acid (120°C, 2 h), and converted to alditol acetates by conventional methods, thereby avoiding the evaporation to dryness. GC-MS was performed on a Trace GC ULTRA system (Thermo Scientific) equipped with a capillary column (NMTR-5MS; 30 ϫ 0.25 mm) using a temperature gradient of 140 (3 min) to 250°C at 15°C/ min with a DSQ II MS detector. Control mixtures with equimolar amounts of glycerol, glucose, and N-acetylglucosamine were treated in parallel.
Enzyme-coupled Spectrophotometric Assay-The enzymatic activity of GDPD was quantified by measuring the production of Gro3P in a two-step coupled spectrophotometric assay. The GDPD reaction mixture (100 l) consisted of 50 mM MOPS (pH 7.5); 20 mM CaCl 2 ; recombinantly purified GDPD Ld17 , GDPD 340 , or derivatives; and a substrate of interest. The final concentration of the substrates used was 2.5 g/liter for the purified carbohydrates and the cell wall-associated carbohydrate extracts and 1 mM for L-␣-glycerophosphorylcholine (Sigma-Aldrich). After 30-min incubation at 30°C, the reaction was stopped by boiling (95°C, 10 min), and the reaction mixture was cooled to room temperature. Buffer conditions (pH, ionic concentration, and divalent cations) and temperature were adjusted if required to study the relevant GDPD reaction parameters. To evaluate the optimal pH for GDPD Ld17 activity, a universal buffer (150 mM KCl, 10 mM KH 2 PO 4 , 10 mM sodium citrate, and 10 mM boric acid) adjusted to pH values from 4 to 13 was tested. After 30-min incubation with GDPD Ld17 , the pH of the boiled samples was adjusted to 7 to prevent any influence on the coupled enzymatic reaction.
The amount of Gro3P produced by GDPD was separately measured in a second reaction using Gro3P dehydrogenase, an oxidoreductase that generates NADH in the presence of NAD ϩ and Gro3P. A volume of 80 l of the GDPD reaction was added to the 900 l of the Gro3P dehydrogenase reaction mixture consisting of 0.8 M hydrazine hydrate buffer (pH 9) supplemented with 0.2 M glycine, 0.5 mM NAD ϩ , and 10 units/ml Gro3P dehydrogenase. This assay mixture was incubated at 37°C for 1 h until oxidation of Gro3P was complete. The reduction of NAD ϩ to NADH was measured by monitoring the absor-bance increase at 340 nm with a spectrophotometer. The GDPD activity is expressed as ⌬A 340 nm /(mol of GDPD ϫ min) with ⌬A 340 nm ϭ A 340 nm (GDPD) Ϫ A 340 nm (control without GDPD). For each substrate, the GDPD concentration was chosen in the linear range of activity.
NMR Analysis of Ldb17 sPS2 after GDPD Ld17 Treatment-Solutions of purified Ldb17 sPS2 (2 g/liter; solution A) and GDPD Ld17 (10 M; solution B) were prepared in a 50 mM MOPS (pH 7.5), 20 mM CaCl 2 buffer system. 0.5 ml of solution A and 0.1 ml of solution B were mixed, and the final volume was adjusted to 1 ml with MOPS-CaCl 2 buffer (final concentrations, 1 mg/ml purified Ldb17 sPS2 and 1 M GDPD Ld17 ). In the control experiment, buffer was used instead of GDPD Ld17 solution. Reaction mixtures were incubated for 2 h at 30°C. A 100-l TCA solution (50%) was added to precipitate the proteins, and the mixtures were fractionated on a Sephadex G-50 column. Elution profiles of two samples were identical, indicating the absence of depolymerization. Fractions corresponding to polysaccharides were collected, lyophilized, repurified on a Sephadex G-50 to remove traces of MOPS, and analyzed by NMR spectroscopy as described in Vinogradov et al. (27). In a separate experiment, the phosphate content in the GDPD Ld17treated and untreated sPS2 sample was assessed according to Chen et al. (78).
Exohydrolytic Activity of GDPD Ld17 on CWTA-The reaction mixture contained 1 ml of purified CWTA (1 g/liter) of S. epidermidis strains 1457 and 5 (67), 1 M GDPD Ld17 , and 50 mM MOPS (pH 7.5), 20 mM CaCl 2 buffer. After incubation for 2 h at 30°C, the chromatographic profiles of the reaction mixture and the untreated purified CWTAs were analyzed on a Sephadex G-50 gel filtration column as described in Vinogradov et al. (27).
Author Contributions-A. C., D. v. S., and J. M. conceived the idea for the project. A. C. conducted most of the experiments, analyzed the results, and wrote the paper. A. C., D. v. S., J. M., I. S., and E. V. were involved in interpretation of the results. E. C. and F. D. B. set up the initial scientific framework for the project. I. S. and E. V. purified the carbohydrates and conducted NMR analysis and Sephadex G-50 gel filtration. J.-P. N. performed mass spectrometry. C. C. and S. B. performed SEC-MALS-RI analysis. C. C. and S. S. created the immuno-EM images. All authors reviewed the results and approved the final version of the manuscript.