The new SH3_T domain increases the structural and functional variability among SH3b-like CBDs from staphylococcal phage endolysins
Department of Biotechnology, Ghent University, Ghent, Belgium
Centro de Investigación Biomédica en Red de Enfermedades Respiratorias (CIBERES), Madrid, Spain
Instituto de Productos Lácteos de Asturias (IPLA-CSIC), Villaviciosa, Asturias, Spain
DairySafe Group. Instituto de Investigación Sanitaria del Principado de Asturias (ISPA), Oviedo, Spain
#Author to whom correspondence should be addressed (yves.briers@ugent.be)ABSTRACT
Endolysins, proteins encoded by phages to lyse their hosts and release their progeny, have evolved to adapt to the structural features of each host. The endolysins from Staphylococcus-infecting phages typically feature complex architectures with two enzymatically active domains (EADs) and one cell wall-binding domain (CBD) belonging to the bacterial SH3 (SH3b) superfamily. This study focuses on three SH3b-like CBDs from exemplary staphylococcal phage endolysins (LysRODI, LysC1C, and LysIPLA5) that were structurally and functionally characterized. While RODI_CBD and C1C_CBD were assigned to the well-known SH3_5 family, a new family, SH3b_T, was identified using the CBD from LysIPLA5 as a model. GFP-fused CBDs were created to assess their differential binding to a collection of staphylococcal strains. IPLA5_CBD showed enhanced binding to Staphylococcus epidermidis, while RODI_CBD and C1C_CBD exhibited distinct binding profiles, with RODI_CBD targeting Staphylococcus aureus specifically and C1C_CBD displaying broad binding. Sequence comparisons suggested that a few differences in key amino acids could be responsible for the latter binding difference. The CBDs modulated the activity spectrum of synthetic EAD-CBD combinations in accordance with the previous binding profiles, but in a manner that was also dependent on the EAD present in the fusion protein. These results serve as a context for the diversity and versatility of SH3b domains in staphylococcal endolysins, providing insights on how (i) the CBDs from this superfamily have diverged to adapt to diverse bacterial ligands in spite of sharing a common fold; and (ii) the evolution of specificity relies on the EAD-CBD combination rather than solely the CBD.
IMPORTANCE
Clinical management of bacterial infections is nowadays compromised by the rise in antimicrobial resistance. The development of new antimicrobial therapies with diverse modes of action is therefore of pivotal importance to complement the current standard of care. Phage endolysins are a new class of antibacterial agents based on rapid peptidoglycan degradation. The natural reservoir of phage endolysins offers a practically infinite diversity. This works reveals a broadly spread but still unknown phage endolysin domain targeting staphylococci while providing structural-functional insights that are paramount to understand the evolution of endolysins and how they can be applied as an antimicrobial.
INTRODUCTION
Endolysins are one of the gene products that dsDNA (bacterio)phages use to release the viral progeny from their host bacterial cells. They contain a catalytic, peptidoglycan-degrading activity, and thus when released to the periplasm via different tightly regulated mechanisms, they provoke bacterial cell lysis by osmotic shock due to the disruption of the peptidoglycan (1). Besides their natural key role in the phage infection cycle, endolysins have sparked interest since they can be purposed as alternative antimicrobial agents (2–4). Due to the escalating burden of antibiotic resistance among clinically relevant bacteria (5), the discovery and development of novel antimicrobials is one of the current main scientific priorities set up by many healthcare authorities and international organizations (6, 7). Recombinantly produced endolysins have been extensively shown to be effective as exogenous antibacterial agents in vitro and in vivo, and different clinical trials have been conducted (8–10). Besides their lesser probability to cause resistance in bacteria, the most interesting features of lysins both from an applied and a fundamental perspective are probably their great natural variability and modularity (Criel et al, 2021; Vázquez et al, 2021). Both characteristics are intertwined, as phages rapidly and dynamically evolve in a modular manner, exchanging functionally autonomous modules of genetic information between each other and with their bacterial hosts (11). At the endolysin level, their modularity means that they may comprise different functional domains: one or more enzymatically active domains (EADs) and a cell wall-binding domain (CBD). Usually, phages that infect Gram-negative hosts bear lysins with a single EAD, whereas those from Gram-positive hosts typically have several domains, at least one of each kind (12, 13). The preferential presence of CBDs in endolysins from a Gram-positive background is hypothetically explained either by (i) the need for a tropism of the enzyme towards its insoluble, non-diffusible substrate (as is the case for many enzymes acting on polymeric substrates (14); (ii) for the endolysin to remain tightly bound to the cell debris of the lysed host thus preventing the killing of neighboring cells that are potential new hosts for the phage progeny; or by a combination of both reasons (15). Importantly, due to their typically high affinity, CBDs are thought to be the main determinant for the observed endolysin specificity, as proven, for example, by domain swapping experiments in endolysins derived from phages infecting Streptococcus or Listeria (Diez-Martinez et al, 2015; Schmelcher et al, 2011; Vázquez et al, 2017).
The case of endolysins from staphylococcal phages has been extensively studied due to the prominent role of many staphylococcal species in human or animal microbiota and disease (16). For example, Staphylococcus aureus is one of the most burdensome human bacterial pathogens globally (17), and Staphylococcus epidermidis plus some other so-called coagulase-negative staphylococci are widespread components of the human skin microbiota that are also responsible for nosocomial infections (18). Endolysins from staphylococcal phages have a typical bicatalytic structure (Vázquez et al., 2021), with evolutionarily conserved CBDs belonging to the bacterial SH3 (SH3b) superfamily (19). SH3b (bacterial Src Homology 3) is a superfamily of widespread ligand-binding domains that appear in many bacterial and phage proteins and are also related to homologous ligand-binding domains in other kingdoms (20). The SH3 fold is one of the simplest and oldest ones (21) and its main structural feature is a β-barrel layout usually devoted to a ligand-binding function. The SH3b domains in particular are mainly known to bind cell wall motifs, thus playing a prominent role in cell wall-remodeling enzymes, autolysins and phage endolysins. In the particular case of staphylococcal lysins, the studied CBDs from SH3b have been classified into the SH3_5 (PF08460) family, and are assumed to specifically bind the peptide moiety of staphylococcal peptidoglycan, including the peptide stem and the peptide cross-link, as recently shown for the SH3_5 CBD of lysostaphin (22, 23). However, SH3b domains comprise representatives that, while sharing the characteristic SH3-like β-barrel topology, have evolved to recognize a variety of cell wall ligands. For example, the SH3_5 from Lactiplantibacillus plantarum major autolysin Acm2 is a broad-range CBD that recognizes many different peptidoglycan chemotypes (24), the SH3b-like, PSA_CBD (PF18341) domain from Listeria phage endolysins recognizes serovar-specific motifs at the cell wall teichoic acids (25), and the SH3b CBD from the endolysin of Bacillus phage PBC5 binds to the glycan chain (26).
In this work, we aimed at providing insights on the specificity range of SH3b-like CBDs from staphylococcal endolysins and how they impact the antibacterial spectrum of the lysins in which they are inserted. To this end, we focused on three Staphylococcus phage endolysins: LysRODI, LysC1C and LysIPLA5 (27, 28). The binding profiles of the selected CBDs were experimentally characterized both as a standalone and in connection with their ability to modulate the activity range of EADs derived from LysRODI and LysC1C. In this way, we expect this work contributes to understand how the structural diversity of staphylococcal CBDs connects to their peptidoglycan-binding function, and how this ability cooperates with intrinsic features of EADs to produce the experimentally observed activity spectra in lysins purposed for exogenous lysis.
MATERIALS AND METHODS
Bacterial strains and culture conditions
The staphylococcal strains used in this work (Table 1) were grown in tryptic soy broth (TSB) at 37 °C with shaking (200 rpm) or on TSB plates containing 2% (w/v) bacteriological agar. Escherichia coli TOP10 was used for cloning and E. coli BL21(DE3) for protein expression. Acinetobacter baumannii RUH 134 (29) was used as a control strain. All the former Gram-negative bacteria were grown in LB medium at 37 °C with shaking (200 rpm). For the positive selection of pVTEIII or pVTD3 E. coli transformants, 100 μg/ml ampicillin or 50 μg/ml kanamycin were used, respectively, together with 5% (w/v) sucrose to negatively select against plasmids lacking insertion (as explained in (30)). 100 μg/ml ampicillin was used to select transformants of vectors based on pET21(a). Bacterial stocks were made by adding 20% v/v glycerol to grown bacterial cultures and were kept at −80°C.
Plasmid construction and DNA manipulation
The sequences encoding LysRODI, LysC1C and LysIPLA5 were codon optimized (GenSmart Codon optimization), synthetized and cloned into a pET21(a) vector (between NdeI and XhoI restriction sites) by GenScript (Rijswijk, Netherlands). For all other proteins used in this work, the expression vectors were constructed through the VersaTile workflow as described in (30). In brief, each individual domain (EAD, CBD or eGFP) was PCR-amplified from its source plasmid with specific primers including BpiI and BsaI recognition sites at both the 5’ and 3’ end, according to the VersaTile method. A restriction/ligation reaction with BpiI was carried out with these amplicons to insert them into the entry vector pVTEIII (AmpR, SucS). The ligation products were subsequently use for transformation of E. coli TOP 10 by electroporation and transformants bearing pVTEIII plasmids with the inserted tile (AmpR, SucR) were selected on LB plates with ampicillin and sucrose. The TOP10 cells were used as a source for tiles, which were all confirmed by Sanger sequencing (LGC Genomics) and stored at the VersaTile repository of Ghent University. Tile ligation into the destination vector pVTD3 (KanR, SucS) was conducted by setting up restriction/ligation reactions with BsaI and the appropriate tiles from the repository (e.g., eGFP plus RODI_CBD and a 6×His tag). All chimeric coding sequences were designed with a C-terminal 6×His tag for purification unless otherwise stated. Final constructs were used for transformation of E. coli BL21(DE3), selecting the transformants with kanamycin and sucrose, and their sequence was verified by Sanger sequencing. A list of the tiles used in this work, including their source NCBI entry and the delineation coordinates can be found in Table 1.
Protein expression and purification
The fusion proteins used throughout this work were expressed in E. coli BL21(DE3) strains bearing the corresponding pVTD3 or pET21(a) vectors prepared as described in the previous section. Protein expression and purification was performed as previously described (31). After purification, the buffer was exchanged to 50 mM sodium phosphate buffer pH 7.4 using Zeba™ Spin Desalting Columns, 7K MWCO, 5 ml (Thermo Fisher Scientific) following the supplier’s recommendations. Finally, proteins were sterilized by filtration (0.45 μm PES membrane filters, VWR).
Protein concentration was quantified using the Quick Start Bradford Protein assay (BioRad). Relevant information on the proteins used in this work is in Table 2.
Quantification of bacterial binding
Binding of CBDs to bacterial substrates was measured by recording the fluorescence of eGFP fusions of the different domains. Such fusions were prepared using VersaTile and comprised eGFP at the N-terminal, the CBD at the central and a 6×His tag at C-terminal position. To perform the binding assay, exponential (OD600 ≈ 0.5 – 0.6) or stationary phase (OD600 ≈ 1– 2) cultures of the strains to be tested were centrifuged (10, 000 × g, 1 min) and the pellets were washed with PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, pH 7.4). The bacterial suspensions were adjusted to OD600 ≈ 1.0 and dispensed on dark flat bottom 96-well plates (180 µl per well). 20 µl of 20 µM solutions of the eGFP-CBD fusion proteins (or just buffer for bacterial autofluorescence controls) were added to each well and then the plates were incubated for 10 min in the dark at room temperature. Then the plates were centrifuged (1000 × g, 5 min), the supernatants were removed, and the pellets washed once with PBS and finally suspended in 200 µl of PBS. 200 µl of 10 µM fluorescein were added to the plate as internal control to automatically optimize gain, as well as positive fluorescence controls for each eGFP-CBD fusion protein (180 µl PBS plus 20 µl of the 20 µM protein stock solution). Fluorescence was then measured in a TECAN Infinite 200 PRO plate reader (TECAN, Männedorf, Switzerland) with excitation/emission wavelengths of 485 nm and 530 nm, respectively. Fluorescence measurements were then corrected for comparability between proteins by applying a correction factor Fmax/Fprot in which Fmax is the maximum fluorescence recorded and Fprot is the fluorescence of each eGFP-CBD at 2 µM. Fluorescence measurements were acquired for three biological replicates.
Minimum Inhibitory Concentration
The minimum inhibitory concentrations (MICs) of the antimicrobial proteins in this work were determined by the broth microdilution assay as described before (31). The MIC values reported correspond to the mode of three independent biological replicates.
Bioinformatic Analyses
Two complementary approaches were taken to build a protein sequence dataset of SH3b-like domains related to staphylococcal lysins (Supplementary Figure S1). To find representative sequences of the IPLA5_CBD family, termed SH3b_T, a phmmer search (32) was conducted against Reference Proteomes restricted to viral taxa (taxid: 10239) and using the first IPLA5_CBD as query (UniProt I6T7G5, from coordinate 380 to 458). The 38 significant hits (i.e., sequences comprising only the SH3b_T-like domains) were clustered with CD-HIT and an exclusion cutoff of 97% identity was applied to decrease redundancy (33). A second phmmer iteration was conducted, now against the full UniProt database, using two queries: the first repeat of the LysIPLA5 CBD and the SH3b_T domain from A0A499SIE6 (positions 169 to 250), displaying the lowest % identity, in amino acid sequence, with the former (39%). This yielded 366 significant hits. A length cutoff was applied to the sequences (source full proteins > 100 amino acids; query coverage > 60 amino acids) and a second CD-HIT redundancy reduction with 97% identity cutoff was applied, reducing the final set to 59 representatives for which additional metadata (domain predictions for the full-length sequences, bacterial hosts) were mined from UniProt or predicted using hmmscan against the Pfam database. Alternatively, representative domain sequences from the SH3_5 family were retrieved from PhaLP database of phage lysins (34), restricting the search to endolysins from host genus Staphylococcus and domain name SH3_5 (236 entries total). The CBD sequences were extracted from these entries using the delineating coordinates from the Pfam SH3_5 predictions stored in PhaLP, and then the same length and % identity cutoffs were imposed to obtain 53 final representative SH3_5 domain sequences. The same PhaLP-based pipeline was used to retrieve PSA_CBD examples (type: “endolysin”, domain name: “PSA_CBD”), while the phmmer pipeline was used to retrieve PBC5_CBD examples. The first SH3b-like repeat of LysPBC5 (A0A218KCJ1, residues 230 to 279) was used as query for the phmmer searches. The dataset built this way can be accessed as Supplementary Dataset S1.
Multiple sequence alignments (MSAs) and phylogenetic analyses were performed and represented in R. The packages ‘msa’ and ‘ggmsa’ were used respectively to build the MSAs (with algorithm ClustalW) and to represent them (35, 36). The phylogenetic trees were built based on the MSAs for which a sequence similarity-based pairwise distance matrix was computed, which was then used as input for building an UPGMA phylogenetic tree and representing it (37–39).
Protein three-dimensional structure predictions were performed using AlphaFold2 algorithm as implemented in ColabFold (40) with default parameters plus Amber relaxation, and models were analyzed and rendered with DeepView (41), which was also used for structural alignments.
Statistical analyses and experimental data representation
Representation of experimental data was performed using R package ‘ggplot2’ (42). Error bars in bar plots represent the standard deviation of three independent replicates. Additional statistical analyses (Kruskal-Wallis test) were performed in R.
RESULTS
LysIPLA5 has an atypical bicatalytic architecture with an SH3b-like C-terminal end
The endolysins from phages phiIPLA-RODI and phiIPLA-C1C (i.e., LysRODI and LysC1C) are examples of the canonical CHAP (PF05257):Amidase_2 (PF01510):SH3_5 architecture described for a majority of staphylococcal endolysins. The S. epidermidis phage vB_SepiS-phiIPLA5, however, bears an endolysin (LysIPLA5) with two predicted EADs, Amidase_2 and CHAP, in the reverse order. In addition, LysIPLA5 has a C-terminus where usually a CBD would be present, but no CBD could be predicted (Figure 1a). In a preliminary functional study of LysIPLA5 domains, neither the full protein nor any of its individual domains showed in vitro antimicrobial activity against S. epidermidis F12, probably due to very low expression yields and subsequently low working concentrations, at least for the full protein (Supplementary Figure S2). The C-terminal domain, henceforth referred to as IPLA5_CBD, was nevertheless efficiently expressed and a concentration as high as 34.16 µM was also unable to exert any growth inhibition. On the other hand, whereas ∼6 µM IPLA5_CHAP was also inactive, a MIC (0.75 µM) was achieved when IPLA5_CHAP was fused to the native C-terminal domain IPLA5_CBD. These results initially supported the possibility of IPLA5_CBD being a CBD.
According to a three-dimensional structure prediction of LysIPLA5, this C-terminal stretch contains two repeats with a fold akin to the typical SH3 β-barrels (Figure 1b). The sequences of these repeats substantially differ from the SH3_5 CBDs of LysRODI and LysC1C (percent identities between 16-21%), while the repeats are 44% identical to each other (Figure 1c). Thus, it was concluded that LysIPLA5 bears a C-terminal CBD that belongs to the SH3b superfamily but that constitutes a different family than the usual SH3_5, which was then termed SH3b_T.
SH3b_T family shows differences in their bacterial distribution with respect to SH3_5
The tree shown in Figure 2 indicates that the 59 SH3b_T-like sequences retrieved by an iterative phmmer search substantially differ in sequence from the 53 established SH3_5 examples obtained from PhaLP, as they are allocated in different clades and are clearly distinguished in the similarity-based distance matrix heatmap. They also display a marked difference in the bacterial species to which they are associated. In the case of SH3_5, they mostly appear in phages whose host is annotated as S. aureus, whereas the preferred host for IPLA5_CBD is S. epidermidis or other coagulase-negative staphylococci, with only a few exceptions to this rule. In addition, a trend can also be established with respect to the preferred full endolysin architecture: while the preferential architecture for endolysins with an SH3_5 CBD is the canonical CHAP:Amidase_2:CBD, the order of the EADs is mostly reversed for lysins with a IPLA5_CBD, as in LysIPLA5 itself.
RODI_CBD, C1C_CBD and IPLA5_CBD present different binding profiles
To confirm the nature of IPLA5_CBD as a true CBD and explore the specificity differences to which the differential taxonomical distributions in Figure 2 point, eGFP fusions of IPLA5_CBD, RODI_CBD and C1C_CBD were obtained. These eGFP-tagged domains were then assayed for binding capacity against a set of staphylococcal strains (Figure 3a). Well-marked trends could be observed in the binding specificity of the three different domains. RODI_CBD bound almost exclusively to S. aureus and a few other coagulase-negative staphylococci (CoNS), C1C_CBD was able to bind generally to all the tested staphylococci, and IPLA5_CBD bound preferentially to S. epidermidis plus other CoNS, although not in such a specific manner as RODI_CBD. A summary and the statistical significance supporting these trends are available in Figure 3b. Although generalizations should be made with care, these results provide an experimental explanation to the preferred association of IPLA5_CBD family to S. epidermidis and CoNS, while SH3_5 is more commonly associated to S. aureus and only in some cases to CoNS or staphylococci in general (and LysC1C is an example of the latter). An additional conclusion of the results in Figure 3 is that binding (i.e., the magnitude of the recorded fluorescence value) seems to be generally lower when cells in stationary phase are used versus using exponential phase ones. An explanation for this may be provided by the nature of the ligand that has been described before for SH3b domains in anti-staphylococcal lysins, which has been usually identified as the peptidoglycan peptide moiety (22, 23, 43, 44). Stationary phase S. aureus cells are known to have fewer cross-links in the peptidoglycan (45), which would then mean a lower number of potential binding ligands for SH3b-like CBDs, thus explaining the results in Figure 3.
The specificity profile of SH3b-like CBDs modulates the antibacterial spectrum of accompanying EADs
To better understand the contribution of the sole-CBD specificity profile to the full lysins’ activity spectrum, fusions of the CHAP EADs of LysRODI and LysC1C with either their wild-type CBD or IPLA5_CBD were obtained. Then, the MIC for each of the constructs, including only the EAD, was calculated against the full set of staphylococcal strains (Figure 4). A general conclusion from this experiment is that fusing a CBD normally improves the antimicrobial activity since all mean MIC log2 fold-change values in Figure 4b are negative, reflecting a decrease in the MIC as a result of fusing CBDs to the EADs. This is in accordance with the common notion that CBDs are necessary for the efficient action of endolysins. There are, however, a few cases in which the EAD-CBD fusion underperforms when compared with the sole EAD, namely RODI_CHAP is less active against some CoNS and a S. epidermidis strain when fused to RODI_CBD. In fact, the MIC improvement is not too impressive when fusing RODI_CBD to RODI_CHAP against any strain (with mean log2 fold-change values of about –1, in contrast with the 5 log2 fold decrease achieved by the IPLA5_CBD fusion against S. epidermidis, for example). This may be explained by the fact that RODI_CHAP already seems a highly optimized EAD against S. aureus, with clearly lower MIC values when compared with S. epidermidis or even the other coagulase negative staphylococci. The RODI_CHAP-IPLA5_CBD fusion does exhibit a remarkably lower MIC against S. epidermidis strains, but not against S. aureus, a behavior correlating to the specificity spectrum shown by IPLA5_CBD in Figure 3. In contrast with the S. aureus-specialized RODI_CHAP, C1C_CHAP is a broad-range EAD, as much as its native CBD is also broad range, although with a slight preference towards S. epidermidis (however non-significant). This intrinsic optimization towards S. epidermidis is more apparent when the effect of fusing C1C_CHAP to the different CBDs is considered: while, as expected, adding IPLA5_CBD improves the performance against S. epidermidis, the fusion with the broad-range C1C_CBD does not decrease the MIC value equally against S. aureus and S. epidermidis; in fact, it shows a similar effect to the IPLA5_CBD fusion (Figure 4b). This confirms a concomitant specialization of C1C_CHAP towards S. epidermidis peptidoglycan rather than that of S. aureus.
The difference in specificity between RODI_CBD and C1C_CBD could be explained by variability in key residues
The structures of the CBDs under investigation in this work were analyzed to find determinants for the perceived functional differences shown in Figure 3 and Figure 4. To this end, MSAs were obtained using the representative sequence sets from Figure 2. A direct MSA-based comparison between the SH3_5 and the SH3b_T domains was not possible due to their low reciprocal identity (Figure 1c); thus, the analysis was split to focus first on SH3_5 domains (Supplementary Figure S3) and, therefore, on the structural differences that explain the different specificities of RODI_CBD and C1C_CBD. A set of key residues for peptidoglycan binding in SH3_5 domains was determined from the thorough, previously published works on the SH3_5 CBD of lysostaphin (22, 23). Then, the corresponding residues in RODI_CBD and C1C_CBD (Table 3) were identified assisted by MSA in Supplementary Figure S3 and the comparison of the predicted 3D models of RODI_CBD and C1C_CBD with the experimentally determined structure of lysostaphin CBD (Figure 5a).
Table 3 shows little or no change in the chemical nature of the residues known to bind the peptide stem of the peptidoglycan across the three CBDs. Conversely, C1C_CBD displays radical shifts in amino acid properties at positions 439, 440 and 461, within the peptidoglycan cross-bridge binding site with respect to lysostaphin and RODI_CBD. Whereas C1C_CBD contains two polar, non-charged amino acids (N439, Q440) and a negatively charged one (E461), RODI_CBD and lysostaphin bear two negatively charged residues (D450, E451) and a positively charged one (R470) at the corresponding sites. Zooming into the analogous structures of these CBDs, it can be concluded that these changes imply a substantial rearrangement of the chemical environment at the groove that is assumed to bind the peptidoglycan cross-bridge. Particularly, D450 and R470 in RODI_CBD are at a distance between 2.98 Å and 4.13 Å (depending on the atoms considered), which makes it possible for them to form a salt bridge (46). This possibility is disrupted in C1C_CBD by the presence of N439 and E461 (Figure 5b).
The facts that (i) the key residues in RODI_CBD and lysostaphin CBD are relatively unchanged but (ii) there are obvious differences in C1C_CBD only at the cross-bridge-binding region is interpretable in the light of the experimental results presented in Figure 3 and Figure 4. While RODI_CBD binds S. aureus specifically, as lysostaphin CBD does, C1C_CBD seems to have no clear preference between binding S. aureus or S. epidermidis. Assuming that the three SH3_5 CBDs compared bind to the same ligand, the peptide moiety of peptidoglycan, which seems plausible given their sequence similarity (Supplementary Figure S2) and the conservation of the residues putatively devoted to binding the peptide stem (Table 3), then the differences found at the cross-bridge binding pocket of C1C_CBD must explain its promiscuous binding profile. In fact, the major difference between the peptidoglycans of S. aureus and S. epidermidis is the structure of their cross-bridges. While the cross-bridge of S. aureus is the well-known pentaglycine bridge, the cross-bridging peptide in S. epidermidis is either GGSGG or AGGGG (47). Since the introduction of a central serine residue in the cross-bridge is a known resistance mechanism to lysostaphin binding (48), it is reasonable that the RODI_CBD, structurally equivalent to lysostaphin CBD, is unable to bind the serine-containing S. epidermidis peptidoglycan cross-bridge. The specificity mechanism in lysostaphin CBD is thought to be one of steric constraint (the cross-bridge binding pocket can only accommodate a pentaglycine peptide (22)). Therefore, in C1C_CBD, the variants N394, Q440 and E461 should provide a greater flexibility for ligand placing at the cross-bridge binding site. This increased flexibility may be achieved by the disruption of the D450-R470 salt bridge in C1C_CBD (Figure 5b).
The IPLA5_CBD fold is closer to PSA_CBD or PCD5_CBD and may bind a different ligand than SH3_5 CBDs
Given the low sequence similarity between the SH3_5 and the SH3b_T sequences, a structure-based comparison was attempted using a few bona fide experimentally determined examples from the wider SH3b CBD superfamily (i.e., lysostaphin CBD, PSA_CBD and PCD5_CBD), which were aligned to the predicted structure of a single IPLA5_CBD repeat (Figure 6a). This comparison initially suggested that IPLA5_CBD repeats were predicted in an SH3b fold more similar to that of the PSA_CBD family (PF18341) or the still poorly described PCD5_CBD family. When sets of representative sequences from PSA_CBD and PCD5_CBD families were added to a phylogenetic tree together with SH3b_T and SH3_5, this was made apparent by the clustering of the former three apart from the latter (Figure 6b).
A preliminary conclusion from this observation is that the ligand for IPLA5_CBD family may be different from that of SH3_5 (i.e., not the peptidoglycan cross-bridge or peptide moiety) given that PSA_CBD domains are known to bind sugar moieties in the teichoic acids of Listeria cells (25), while the ligand for PCD5_CBD has been described as the peptidoglycan glycan strands (26). Since IPLA5_CBD family is closer to PSA_CBD and PCD5_CBD than to SH3_5, it may bind a glycan moiety rather that a peptidic one. Given that the cell wall teichoic acids are also, in general, a differential feature between S. aureus and S. epidermidis (49), teichoic acids might be the ligand of IPLA5_CBD explaining its preference for S. epidermidis in detriment of S. aureus.
DISCUSSION
In this work, we have identified a new family of SH3b-like domains that binds elements of the staphylococcal cell wall, with a preference towards S. epidermidis, and we have set a context for part of the underexplored diversity of the versatile SH3b-like folds among lysins. Here, we have compared four SH3b families present in phage endolysins: the SH3_5 family commonly found in staphylococcal lysins, the newly described SH3b_T present in LysIPLA5, the listerial PSA_CBD and the PBC5_CBD domains found in Bacillus and their phages. All of them share the common β-barrel structure, typical of SH3 folds (Figure 6a), although with differing topologies (Figure 7). These differences may correlate with the type of cell wall ligands they recognize, namely (i) the peptidoglycan peptide moiety (or more specifically the cross-bridge) for SH3_5 CBDs or (ii) different glycan moieties for PSA_CBD, PBC5_CBD and, perhaps, IPLA5_CBD. Regarding their topology, all of them conserve the general structure of the SH3 fold, and share the common SH3b trait of an extended RT β-hairpin (equivalent of the RT loop in SH3e, the topology of eukaryotic SH3 domains) plus the conserved central antiparallel β-sheets 2 to 4 (“β-core”, Figure 7). Their main topological differences are located (i) at the N- and C-proximal β-sheets, such as the presence/absence of an additional N-terminal β-sheet (β0, present in SH3_5 and SH3b_T) or an additional β-sheet connecting β4 and β5 (β4-5, in SH3_5 and PSA_CBD); and also (ii) at the RT β-hairpin, which is clearly more elongated in PSA_CBD, PBC5_CBD and SH3b_T, with the latter having the longest version. These differences should somehow account for the different ligands of the families, although this work does not provide concrete insights about the possible ligand.
On a different note, the results hereby presented show the versatility of the SH3b domains for evolving slightly different structures that bind different ligands, which is particularly evident in the comparison between RODI_CBD and C1C_CBD. While both belong to the same SH3_5 family, their divergence only in a small set of key residues sets them rather radically apart in their specificity, determined by the nature of the peptidoglycan cross-bridge on the ligand side. This suggests that not only the SH3 fold can evolve towards topologically diverse families, each of them recognizing a different kind of ligand, but also within each family, SH3b domains can fine-tune the residues at the binding pockets to select the specific ligands they bind. Thus, the wide-spread presence of SH3b-like domains in endolysins from phages that infect very diverse groups of bacteria can be explained on the basis of this versatility of SH3 folds as “raw material” to evolve CBDs targeted at very specific ligands. However, the binding specificity dictated by CBDs, as we have shown, does not fully determine the activity spectrum of the derived lysins in which they are allocated (Figure 4). For example, the increased activity against S. epidermidis provided by IPLA5_CBD was more prominent when fused to an EAD with poor activity against this bacterium (RODI_CHAP) than when accompanying an EAD already optimized against S. epidermidis (C1C_CHAP). Thus, our set of results imply that, while the final antibacterial outcome correlates in terms of specificity to the observed specificity profile of the CBDs, it is also true that such outcome is influenced by the intrinsic activity range displayed by the EAD. Therefore, while the CBD can be a specificity determinant, it must be considered that it does not fully determine endolysin specificity, at least in the case of anti-staphylococcal lysins, and this may be of greater importance when engineering new tailor-made lysins.
Acknowledgements
RV was supported by a postdoctoral fellowship of the ‘Bijzonder Onderzoeksfonds’ (BOF) Ghent University (01P10022). DGu was supported by the Research Foundation– Flanders (FWO) under grant G066919N. This work was also funded by MCIN/AEI/10.13039/501100011033/FEDER, UE, grant number PID2022-140988OB-I00 awarded to PG.
Statements and declarations
Competing Interests
DGu is currently employed by Telum Therapeutics SL. DGr and YB are co-founders and (part-time) employee of Obulytix.
Data availability
All data supporting the findings of this study are available within the paper and its Supplementary Information.