Insights into the molecular determinants of host specificity in Pseudomonas aeruginosa-infecting phages: a structural and functional analysis of tail fibre proteins

Abstract

Background

Bacteriophages and bacteria frequently occupy the same ecological niches, driving complex and dynamic host–virus interactions. In Pseudomonas aeruginosa, phages from the Migulavirinae subfamily, tail fibre proteins (TFPs) are crucial to host recognition. These proteins, located within the phage tail structure, are subject to frequent recombination and may play a key role in shaping host range. This study investigates the molecular basis of host specificity in Litunavirus and Luzseptimavirus phages, focusing on the structure and variation of their TFPs.

Results

Host spectrum analysis divided phages into three categories; however, contrary to expectations, no direct correlation was found between TFP recombination history and host range, most likely because subsequent single amino acid changes in the pyocin knob regions, critical for adsorption, altered the host spectrum after the recombination event. Notably, phages sharing highly similar pyocin knob 2 domain architectures displayed identical host spectra, suggesting a strong link between this region and host specificity. Despite high sequence variability, all TFPs adopted a conserved trimeric fold with five regions: N-terminal, GrpE-like, GDSL-like with a carbohydrate-binding module, pyocin knob, and C-terminal. Structural similarities to bacterial PilA and pyocins were noted. Variation in the pyocin knob region, especially substitutions involving polar residues, was partially correlated with host range, likely via hydrogen bonding with the O-antigen. The GrpE-like domain resembled type IV pili, suggesting a role in reversible attachment, while the GDSL-like domain may support enzymatic processing of the O-antigen.

Conclusions

Our findings support a multi-step adsorption mechanism of Migulavirinae phages, initiated by random encounters with the bacterial surface, followed by specific, stable interactions between the pyocin knob region and the bacterial lipopolysaccharide (LPS) O-antigen. Final stabilization involves additional interactions with the LPS core region. While the GrpE-like domain may contribute to transient stabilization near the surface, its structural similarity to PilA suggests a possible evolutionary convergence rather than a direct pilus-binding function. Despite high sequence variability, TFPs maintain conserved structural features, allowing for modular adaptations that precisely adjust host specificity. Importantly, the lack of a direct link between TFP recombination and host range suggests that factors beyond recombination influence phage host specificity.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12915-025-02445-y.

Background

Pseudomonas aeruginosa is a highly adaptable opportunistic pathogen responsible for a wide range of infections, including chronic lung infections in cystic fibrosis patients, ventilator-associated pneumonia, urinary tract infections, and infections of burn wounds. These infections are often difficult to treat due to the bacterium’s intrinsic and acquired resistance to antibiotics, as reported by WHO [1]. In light of growing antimicrobial resistance, bacterial viruses (bacteriophages, phages) are gaining renewed attention as potential therapeutic tools [2–4]. However, beyond phage therapy, bacteriophages have been increasingly recognized for their diverse applications in medicine, biotechnology, agriculture, food safety, and environmental protection [5, 6]. This growing diversity of applications highlights the need for a deeper understanding of phage biology, particularly the molecular mechanisms of host recognition and infection.

Phages offer advantages such as short replication cycles, large burst sizes, and high mutation rates, which enable the rapid adaptation necessary to overcome bacterial defences [7]. One of the most important features of a therapeutic phage is its lytic host range, defined as the spectrum of bacterial strains it can infect and lyse productively [4]. While some phages are highly specific and infect only a narrow range of strains [2, 8], others exhibit polyvalency, targeting bacteria from different species or even genera [9]. Both strategies can be valuable depending on the clinical application [10].

Host range is shaped by specific interactions between phage structures and bacterial receptors. In P. aeruginosa, known phage receptors include lipopolysaccharide (LPS) and type IV pili [11–13]. Recent studies on phage DEV (Migulavirinae), a Schitoviridae member, showed that long tail fibre proteins (TFPs) bind the LPS O-antigen likely through a carbohydrate esterase-like domain, aiding adsorption but necessarily being essential for infection. TFPs anchor the phage to smooth LPS, while the short tail fibre protein gp56, acting as a tail plug, engages a secondary receptor exposed on rough LPS, triggering plug release and genome ejection [13, 14].

The dynamics of phage-host interactions are often described using evolutionary models such as the arms race, where both phages and bacteria undergo counter-adaptations to gain the upper hand [15, 16]. On the phage side, this often involves rapid genetic changes in genes encoding receptor-binding structures such as tail fibres, spikes, and other appendages [17, 18]. Indeed, tail fibre genes are frequently recombined in many phage genomes, including those infecting P. aeruginosa [19–21].

Thanks to recent developments in structural bioinformatics, these interactions can now be explored in silico with increasing accuracy and biological relevance. Protein structure prediction tools—supported by crystallographic data and machine-learning-driven de novo approaches—allow us to investigate receptor-binding mechanisms quickly and reproducibly [22, 23]. Although traditionally studied through empirical methods, phage structural analysis is becoming a vital aspect of understanding host specificity [24].

In this study, we examined a set of sewage-derived Migulavirinae phages infecting P. aeruginosa. Building on our previous work on recombination in this subfamily [2], we investigated how amino acid variation in long TFPs translates into structural differences and functional outcomes in host range. Our aim was to identify specific elements involved in LPS receptor binding, with potential implications for the development of customized therapeutic phages targeting resistant P. aeruginosa strains [25, 26].

Results

Molecular identification of newly isolated phages

A total of 18 phages were isolated from municipal sewage. Two of them were selected for further analysis due to their highest titres expressed in plaque-forming units per millilitre (PFU/mL). Based on a 759-nucleotide (nt) fragment of the portal protein gene, the first phage showed high identity and coverage (99.87% and 100%, respectively) with the phage Ka4 (NCBI accession number PP054054.1) [27], belonging to the genus Luzseptimavirus; hence, it was named Ka4-like throughout text. The second virus was named PASB7-like after Litunavirus PASB7 (acc. OR509539.1) [28], with the highest identity and sequence coverage for a 787-nt fragment of the portal protein gene (100% and 100%, respectively; NCBI accession WNV46176.1). However, the gene encoding TFP in these phages does not correspond to the molecular identification based on the portal protein. Unlike the reference phages, their TFPs lack recombinant fragments. Obtained sequences are available in Figshare [29].

Host spectrum and adsorption specificity of phages

Using 25 strains of P. aeruginosa, we have determined and compared the host spectra for Litunaviruses isolated and characterized earlier: vB_Pae575P-3, vB_Pae1369P-5, PA26 and vB_PaeP_MAG4 [8, 21, 25, 30] as well as PASB7-like Litunavirus and Ka4-like Luzseptimavirus, isolated and preliminarily identified in this study. Out of 25 tested P. aeruginosa strains, 11 were susceptible to the phages studied (Table 1). We observed differences in phages’ specificity to the strains studied, and the comparison of detailed host spectra is presented in Fig. 1 and Table 1. Interestingly, among phage-sensitive strains, we observed diverse types of lytic activity in spot assays, corresponding to different plaque morphologies and efficiencies of plating. In several cases, phages formed distinct, countable plaques even after dilution, indicating productive infection (designated as P; e.g. vB_Pae575P-3, vB_Pae1369P-5, vB_PaeP_MAG4 on P. aeruginosa 705/1996). In other instances, plaques were either barely visible (P^b) or appeared only at high phage concentrations (P^a), suggesting low-level productivity or reduced adsorption/replication efficiency. Moreover, some strains exhibited clearing zones without any visible plaques, even after lysate dilution (e.g. PA26 on strain 705/1996), which we classified as N*, indicative of lysis from without rather than true infection. To verify this, we performed adsorption assays for all N* cases, which did not confirm phage adsorption to the host cells, supporting the hypothesis of non-productive interaction. Finally, no visible effect (no clearing or plaques) was interpreted as complete resistance (N) (Fig. 1; Table 1). Based on the patterns of lytic activity and plaque morphology, the phages appeared to fall into three general categories. The first included broad-spectrum, high-efficiency phages such as MAG4 and PASB7-like, which were capable of infecting and lysing all the tested strains with clearly visible plaque formation. The second group comprised phages with a moderately broad host range, including vB_Pae575P-3, vB_Pae1369P-5, and PA26, which exhibited productive infection in many strains but with lower efficiency in some cases, reflected by reduced plaque size or number (P^a, P^b). The third group included narrow-spectrum or non-productive phages, exemplified by Ka4-like, which showed limited infectivity and failed to produce plaques in several strains, suggesting either an inability to adsorb or induction of lysis from without (N*) only at high phage concentrations. Of particular interest, strain 961/1996 was susceptible only to MAG4 and PASB7-like phages, which caused productive infection, whereas all other phages induced only weak lysis from without. In contrast, strains such as 1369/2003 and 886/2003 were susceptible to all tested phages, highlighting the variability of surface receptor availability and phage susceptibility among P. aeruginosa clinical isolates.

Table 1.

Lytic and adsorptive bacterial host spectrum of phages

Pseudomonas aeruginosa strain	Phages and their spectrum type
	Ka4-like	vB_Pae575P-3	vB_Pae1369P-5	PA26	MAG4	PASB7-like
436/1996	P	P	P	P	P	P
705/1996	P	P	P	N*	P	P
961/1996	N*	N*	N*	N*	P	P
1947/2003	N	Pa	Pa	P	P	P
1864/2003	P	P	P	Pb	P	P
886/2003	P	P	P	P	P	P
1369/2003	P	P	P	P	P	P
575/2003	P	P	P	Pb	P	P
1702/2009	N	N*	N*	P	P	P
27853	P	P	P	P	P	P
10145	P	N	N	P	P	P

P – infective, productive phage, exhibiting adsorption followed by productive cycle and host lysis, plaques visible (white cells); P^a – P displaying low efficiency of plating, producing few plaques only at high concentrations (light grey cells); P^b – P displaying high efficiency of plating, producing small, barely visible plaques (light grey cells); N – non-infective phage, displaying no adsorption and no productive cycle, no plaques and no lysis observed (dark grey cells); N* – N displaying bacterial lysis from without, observed at high lysate concentrations, no plaques observed (medium grey cells).

Fig. 1.

Various types of host susceptibility to studied phages observed in spot test. Left side: phage PA26 forms clearings without visible plaques on host strain 705/1996 in contrast to phages vB_Pae575P-3, vB_Pae1369P-5, vB_PaeMAG4, Ka4-like and PASB7-like showing clearings with plaques. Right side: host strain 1947/2003 resistant to phage Ka4-like, effectively lysed by PA26, vB_PaeMAG4 and PASB7-like (clearings with plaques) and showing low efficiency of plating for phages vB_Pae575P-3, vB_Pae1369P-5 (single plaques only at high concentrations)

It is worth noting that phage lysates used for host spectrum studies were purified by caesium chloride (CsCl) density gradient centrifugation to avoid any interference from bacterial components with the results. However, lysates extracted with chloroform-extracted or precipitated with polyethylene glycol (PEG) without CsCl density gradient purification produced equivalent results (Additional file 1: Table S1).

Structural model of tail fibre protein

Based on our previous identification of the highly variable region of the TFP in representatives of the Schitoviridae family (Migulavirinae subfamily) representatives, specifically two genera: Litunavirus and Luzseptimavirus [21], we aimed to find the connection between the structure of the TFP and the viral host spectrum pattern. At first, we confirmed the presence of tail fibre structures in the representative of the Litunavirus genus, vB_Pae1369P-5, by transmission electron microscopy (Fig. 2). In the microphotograph, two phage particles are shown with their icosahedral heads of approximately 70 nm in diameter, portal structures and short tails clearly indicating their podoviral morphology with tangled fibre-like structures of the two phage particles (Fig. 2). While these images provide general confirmation of the presence of fibre-like appendages, they do not allow an unambiguous assignment of the TFP to either side or central tail fibres due to limited resolution and image contrast. Furthermore, we performed a comparative analysis of published microphotographs of additional Litunavirus and Luzseptimavirus phages. Notably, none of these studies provided direct evidence linking the observed fibre-like structures to receptor recognition (Additional file 1: Table S2) [2, 8, 25, 28, 30–34]. Nonetheless, structural and functional analyses of the TFP, including the presence of a GrpE-like coiled-coil domain (named after the bacterial heat-shock nucleotide exchange factor GrpE) and homology to characterized long tail fibre proteins (e.g. of phage DEV) [13], support its role in host recognition rather than in DNA injection.

Fig. 2.

Transmission electron microphotograph of phage vB_Pae1369P-5 particles. Portal structures (A) and its tangled fibres (B) pointed with arrows

Subsequently, a preliminary functional model of the TFP was constructed based on the full-length structural analysis in accordance with the NCBI Conserved Domain Database. The TFP is a trimeric protein composed of five distinct regions (Fig. 3). The boundaries of these regions are clearly identifiable through the protein's secondary and tertiary structures, including β-sheet barrels and α-helices (Fig. 3). Based on the residue positions in phage vB_Pae1369P-5, the first 112 residues were assigned to the N-terminal domain. The subsequent residues (113–246) were designated to the coiled-coil region, corresponding to the GrpE-like domain. Residues 247–686 were classified as the GDSL-like (Gly-Asp-Ser-Leu motif-containing) / CMB-like (carbohydrate-binding module-like) region. Residues 702–940 were assigned to the pyocin knob region. Finally, residues 941–1081 were attributed to the C-terminal lectin-like domain. These boundaries correspond to those proposed by Buth et al. [35] for the receptor-binding fibres of bacteriocins produced by Pseudomonas aeruginosa, (R1 pyocin) and TFPs of other Litunavirus phages, such as phage DEV [13, 35]. The amino acid sequences and model structures of TFP trimers were compared across the phages included in this study, in accordance with previously described recombination events between the two genera within the Migulavirinae subfamily: Litunavirus and Luzseptimavirus [21]. Notably, despite variations in the amino acid sequences, as depicted in the overview of alignments in Additional file 1: Fig. S1, the models exhibited a conserved structure across the phages. The origin of the functional domains within the structures is delineated using distinct colours. Phages that are non-recombinant representatives of the Litunavirus and Luzseptimavirus genera, namely LIT1, PASB7-like, and LUZ7, Ka4_like (representing pure ancestral lineages), are monochromatic, whereas the remaining phages exhibit recombinant characteristics expressed in various distinct forms. The Litunaviruses vB_PaeP_MAG4 contain a recombined GrpE-like fragment, whereas the remainder of their structure derives from the Litunavirus lineage. Phages vB_Pae1369P-5 and PA26 share similar structural origins, with their conserved N-terminal domains originating from Litunavirus, while the GrpE-like domains, carbohydrate esterase-like regions, and pyocin knobs exhibit greater similarity to the Luzseptimavirus amino acid sequences. Furthermore, phages vB_Pae1369P-5 and PA26 possess highly similar C-terminal domains. It is noteworthy that the TFP domains of phage origin exhibit significant similarity to the pyocin R1 and R2 fibres from Pseudomonas aeruginosa (Fig. 3). Both pyocin knobs, the knob connector, and the C-terminal domain precisely mimic their bacterial counterparts, facilitating a comparison of their biochemical properties.

Fig. 3.

Structural models and comparison of long tail fibre protein (TFP) trimers from various phages. Corresponding functional regions and domains as well as Pseudomonas aeruginosa R1 and R2 pyocins are identified and marked with brackets on the left. The recombined fragments within each TFP trimer are colour-coded to indicate their origin: blue for Litunavirus origin and purple for Luzseptimavirus origin. This figure was generated using AlphaFold [36] and visualized with Chimera [37]. The input sequences and coordinates of the predicted models are available in Figshare [29]

Variation of TFP domains in relation to the host spectrum of phages

The comparison of TFP models among Migulavirinae representatives showed high structural similarity, as visible in Fig. 3, despite the differences in host spectrum patterns. This finding led us to a more detailed analysis of the amino acid composition and polarity in the studied phages. The most prevalent nonpolar amino acids included alanine, glycine, valine, and leucine, whereas threonine, asparagine, serine, and aspartic acid were the most abundant among the polar residues (Additional file 1: Table S3). Particular attention was given to the distribution of polar amino acids with exposed functional side chains, especially within the pyocin knob region, as these structural elements are implicated in interactions with bacterial surface saccharides. Detailed analyses of amino acid composition and surface accessibility within individual domains are presented in the subsequent sections.

The N-terminal domain, in the form of a coiled barrel, is highly conserved, with high sequence and structural model similarity across phages (Additional file 1: Fig. S1A). These findings support the conserved and non-recombined nature of the domain, which was not subject to hybridization occurring in Migulavirinae ancestral lineages (Fig. 3). No correlation was observed between the amino acid composition of this domain and the phage host.

The GrpE-like domain exhibits a conserved, long coiled-coil structure (134 residues in phage vB_Pae1369P-5). However, the residue composition varies among phages, indicating a recombination hotspot within the TFP (Additional file 1: Fig. S1B). Interestingly, a comparison between the GrpE-like domain of the phage TFP and the P. aeruginosa pilus protein PilA (PDB: 1OQW) [38] revealed structural similarities (Fig. 4). Structural comparisons of all analysed GrpE-like domain models with the PilA protein model demonstrated a high degree of structural similarity—α-helices of both these structures align in reverse-complement conformation (the N-terminus of the phage protein domain corresponds to the C-terminus of the bacterial protein). The highest alignment score was observed for phage vB_PaeP_MAG4 (Additional file 1: Table S4) [39–41].

Fig. 4.

GrpE-like domain structural comparison with the bacterial PilA protein. A Structural model of the N-terminal and GrpE-like domains of phage vB_Pae1369P-5. B Structural model of the PilA monomer. This figure was generated using Robetta [42] and visualized with Chimera [37]. The input sequences and coordinates of the predicted models are available in Figshare [29]

The GDSL-like esterase domain and the CBM module of the vB_Pae1369P-5 tail fibre protein exhibit homology to the corresponding region in coliphage G7C [43]. This region is relatively large (440 residues) and highly conserved. It is composed of α-helices and β-sheets intricately folded, forming the broadest portion of the fibre (Fig. 3). The GDSL-like domain, located between residues 452 and 686, contains conserved amino acids such as glycine, aspartate, serine, and leucine, which are characteristic of the active sites of enzymes such as esterases, lipases, and deacetylases [44]. Although these residues are dispersed in the primary amino acid sequence, they are spatially co-localized within the three-dimensional structure, forming a functional active site (Fig. 5B, C). The structure exhibits a high degree of similarity to domain 4 of gp63.1 from coliphage G7C, adopting the fold of an SGNH (Ser-Gly-Asn-His) esterase [43]. The carbohydrate-binding module, located between residues 257 and 451, displays a high degree of similarity to domains 5 and 6 of gp63.1 from coliphage G7C [43].

Fig. 5.

Detailed view of the GDSL/CMB-like region. A An overview of TFP protein structure of phage vB_Pae1369P-5 aligned with the O-antigen structure: core oligosaccharide in navy blue and saccharide polymer in light red. B Identification of the catalytic site (yellow) and residues forming it within the GDSL-like domain (light purple). C Structural comparison of the SGNH domain in phage G7C (beige, with residues marked in yellow) with the corresponding region in phage vB_Pae1369P-5 (blue, with residues marked in red), including annotated critical residues. This figure was generated using AlphaFold [36] and visualized with Chimera [37]. The input sequences and coordinates of the predicted models are available in Figshare [29]

Alignment analysis of the complete GDSL-like/CBM region reveals a consistent level of variability, with an average amino acid similarity of 84.66% across all phages (including both Litunavirus and Luzseptimavirus). The observed variability does not correlate with the host spectrum of the studied phages.

The region following the GDSL-like/CBM domain is separated by a short inter-domain α-helical linker (~15 residues) and comprises the pyocin knob architecture, as defined by Buth et al. [35]. This region includes the knob1 domain (residues 702–785), the inter-knob connector (residues 786–855), and the knob2 domain (residues 856–940), forming a modular assembly characteristic of bacteriocin tail fibre structures [35]. The knob1 domain is characterized by higher sequence conservation compared to the knob2 domain (Additional file 1: Fig. S1D). Despite differences in amino acid composition, the structures remain preserved among all phages and are analogous to those of P. aeruginosa R1 pyocin fibre (Fig. 3). Both domains display phylogenetic variability in terms of amino acid sequence origin. Phage vB_PaeP_MAG4 shows a high degree of similarity to Litunavirus strains LIT1 and PASB7-like, while the remaining phages (vB_Pae575P-3, vB_Pae1369P-5, and PA26) are more similar to the Luzseptimavirus strain LUZ7. Regarding the variation of these domains in relation to the host spectrum, knob1 does not follow a specific pattern, remaining identical between phages vB_Pae1369P-5 and PA26, despite their distinct adsorption spectra. In contrast, both the inter-knob connector and knob2 are more variable and reveal a few mutations between phages expressing distinct host ranges, and thus might be crucial for effective host recognition (Table 2). In the case of phages sharing an identical adsorption spectrum (vB_PaeP_MAG4 and PASB7-like), no amino acid variability was observed within pyocin knob1 and knob2; however, in the inter-knob connector, they differed at two amino acid positions (Additional file 1: Fig. S1).

Table 2.

Comparative analysis of amino acid variability within the tail fibre protein domains (inter-knob connector and knob2) differentiating vB_Pae1369P-5 and PA26. The one-letter codes correspond to specific amino acids according to the standard amino acid nomenclature: S—serine, R—arginine, N—asparagine, G—glycine, A—alanine, H—histidine, D—aspartic acid, L—leucine. The chemical properties of amino acids are indicated as follows: no ‘d’ sign for nonpolar residues, ‘d’ for polar uncharged residues, ‘d + ’ for polar positively charged residues, and ‘d − ’ for polar negatively charged residues

Phage	Mutation position within inter-knob connector region			Mutation position within pyocin knob2 domain
	793	838	850	862	910
vB_Pae575P-3	S d	G	N d	N d	D d −
vB_Pae1369P-5	S d	G	N d	N d	H d +
PA26	R d +	D d −	H d +	D d −	N d
Ka4-like	S d	G	N d	L	S d
vB_PaeP_MAG4	N d	A	N d	N d	A
PASB7-like	N d	A	N d	N d	A

Within the inter-knob connector region of bacteriophage PA26, three amino acid substitutions were identified, along with two additional substitutions within the pyocin knob2 domain. Notably, PA26 exhibited a broader host range compared to phages vB_Pae575P-3 and vB_Pae1369P-5 (Table 1). In most instances, the observed substitutions in PA26 involved a transition from polar uncharged or nonpolar residues, as present in vB_Pae575P-3 and vB_Pae1369P-5, to polar charged residues. Representative examples include S793R, G838D, and N862D. An exception to this trend was the D/H910N substitution, in which the negatively charged aspartic acid (D) in vB_Pae575P-3 and the positively charged histidine (H) in vB_Pae1369P-5 were both replaced by the polar uncharged asparagine (N) in PA26 (Table 2). A reciprocal pattern was observed at position 862, where PA26 retained aspartic acid (D), while both vB_Pae575P-3 and vB_Pae1369P-5 exhibited asparagine (N). These alterations at positions 862 and 910 may influence the host specificity of the phages. Interestingly, bacteriophage vB_PaeP_MAG4, despite displaying a more similar adsorption profile to PA26 than to vB_Pae1369P-5, showed substantial divergence in the corresponding amino acid sequences (Table 2; Additional file 1: Fig. S1E). It is also noteworthy that vB_PaeP_MAG4 possesses a higher number of asparagine residues, particularly within the knob2 domain (10/86) and C-terminal domain (12/138), which may also contribute to its adsorption properties and host range. The asparagine (N) residues are highlighted in Fig. 6.

Fig. 6.

Comparison of pyocin knob regions and C-terminal domains across phages. From left to right: vB_Pae575P-3, vB_Pae1369P-5, PA26, and vB_PaeP_MAG4. Asparagine Asn, N) residues are marked in blue. This figure was generated using AlphaFold [36] and visualized with Chimera [37]. The input sequences and coordinates of the predicted models are available in Figshare [29]

The C-terminal domain forms a conserved three-dimensional structure located at the distal tip of the TFP (Fig. 3), and is implicated in receptor recognition, particularly in binding to LPS structures, as previously demonstrated for phage T4 [45]. Distinct ligand-binding cavities of vB_PaeP_MAG4 are evident at the base of this domain (Fig. 7), suggesting a potential role in host interaction. However, these cavities differ structurally from those of phage T4, which targets LPS core or OmpC-like β-barrel proteins. Based on domain localization, surface topology, and the physicochemical profile of this site, we hypothesize that it is optimized to recognize the core oligosaccharide region of LPS, rather than the outer membrane protein C (OmpC) porin or the variable O-antigen. Importantly, the observed structural differences in cavity architecture may reflect adaptation to the distinct composition of the LPS core in Pseudomonas aeruginosa, the host of our phage, which differs significantly from the LPS core of E. coli, the host of phage T4 [46]. The amino acid sequence of this domain in phage vB_PaeP_MAG4 differs substantially from those of the other phages analysed, resembling the divergence observed in the pyocin knob domains. This divergence supports the hypothesis of a recombination-driven origin of the gene fragment encoding the C-terminal domain (Additional file 1: Fig. S1F). Nevertheless, sequence similarity among the phages does not align with their respective host range profiles, indicating that host specificity may be determined by additional structural or physicochemical factors beyond simple sequence conservation.

Fig. 7.

Detailed view of the C-terminal binding surfaces. Left: C-terminal TFP domain (blue) and its comparison with the analogous structure from phage T4 (right, beige), showing  visible crevices. This figure was generated using AlphaFold [36] and visualized with Chimera [37]. The input sequences and coordinates of the predicted models are available in Figshare [29]

Discussion

In our previous studies, we demonstrated that recombination events within the gene encoding the tail fibre protein in Litunavirus phages, members of the Schitoviridae family (Migulavirinae subfamily), significantly impact their phylogenetic positioning and classification within the viral taxonomy [21]. A key finding from these studies was the occurrence of intergeneric recombination between phages from two distinct genera, Litunavirus and Luzseptimavirus, resulting in the emergence of genetic variants potentially influencing host range and adsorption mechanisms, including shifts in host specificity [21]. Phages vB_Pae1369P-5 and PA26 share a recombination history of TFP; hence, we would expect them to exhibit similar host spectrum patterns. Our observations from performed experiments state that these assumptions were not true. Therefore, this study aims to build upon previous findings by investigating the impact of recombination within the tail fibre on the host spectrum of phages infecting Pseudomonas aeruginosa. Additionally, a detailed structural analysis of the TFP model provides insights into the potential adsorption mechanisms employed by Litunavirus phages for host recognition and attachment.

Trans-genera recombination

Genes involved in host recognition, encoding tail fibres, tail spikes, or other receptor-binding proteins, are typically modular and highly susceptible to recombination, making them hotspots for genetic exchange [7, 21]. This study provides further evidence for genetic influx from the Luzseptimavirus genus into Litunavirus, resulting from a trans-genera recombination event in an ancestral lineage, as previously suggested by our research team [21]. Additionally, amino acid sequence analysis identified repeated motifs, that may represent residual traces of homologous recombination events, consistent with earlier reports [47]. The high abundance and variable positioning of these motifs suggest a frequent occurrence of recombination, likely driven by adaptive evolutionary pressure to modify or expand the host range of these phages [18].

Correlation between host spectrum and the structure of the tail fibre protein domains

This study presents a structural–functional model of TFP, essential for phage adsorption and host specificity. Structural modelling revealed a conserved trimeric architecture with five domains: (i) N-terminal, (ii) coiled-coil GrpE-like, (iii) GDSL-like/CBM, (iv) pyocin knob (knob1, connector, knob2) and (v) C-terminal lectin-like. These assignments align with prior studies [13, 35], supporting the model’s validity.

While plaque assays defined host range, some strains exhibited diffuse lysis without detectable adsorption, likely resulting from lysis from without [48, 49]. However, other mechanisms, such as cell wall weakening by phage-derived enzymes or abortive infection, cannot be excluded. As fluorescence-based assays for DNA injection were not employed, such non-productive interactions should be interpreted with caution.

Although a consistent correlation between global TFP structure and host range was not observed across all analysed phages, our structural analysis revealed that specific polar amino acid substitutions in the pyocin knob2 region were associated with broader host tropism. For example, a substitution at position D862N in phage vB_Pae1369P-5 and H/D910N in phage PA26 replaces charged residues with polar asparagine, enabling multivalent hydrogen bonding. Given that each virion displays 15 TFP subunits, these mutations amplify the binding potential to O-antigen sugars [13, 50]. Similar enrichment in asparagine, a polar residue, was observed in vB_PaeP_MAG4 and PASB7-like, which also has broad host range, further supporting this association. Asparagine’s amide (–CONH₂) side chain enables hydrogen bonding with hydroxyl (-OH) and carboxyl (-COOH) groups typically present on sugars of LPS O-antigen chains [50]. Therefore, it is the specific positional occurrence of such polar residues, rather than the overall sequence similarity, that appears crucial for host recognition [50].

Importantly, the observed distribution pattern of polar and charged residues across receptor-binding domains, as summarized in Table S3, provides mechanistic insight into phage host specificity. Rather than reflecting global sequence conservation, it is the precise positional occurrence and chemical nature of such residues, particularly within the knob2 region, that appear to modulate local binding affinity through electrostatic and hydrogen bonding interactions. These findings support a model in which host recognition is mediated by such localized physicochemical forces, with the knob2 domain playing a central role. This interpretation is further supported by our observation of specific amino acid substitutions, including D → N and H → N changes, that introduce polar side chains such as asparagine capable of multivalent hydrogen bonding. These substitutions are correlated with a broader host range, suggesting functional consequences at the adsorption level.

This hypothesis aligns with previous studies showing that even single amino acid changes can drastically alter host range by modifying interaction surfaces or charge distribution [51, 52]. As previously reported, point mutations in structural genes of tail fibre proteins may profoundly influence phage–host dynamics and facilitate adaptation to new bacterial targets [18]. Similar principles are observed in other viruses: for example, a single glutamic acid to valine substitution in the SARS-CoV-2 spike protein altered both its structural stability and receptor binding affinity [53, 54]. These examples underscore how even subtle shifts in polarity or electrostatic character of surface residues can drive adaptive changes in viral tropism. Therefore, amino acid polarity analysis complements structural modelling and adsorption data, highlighting the critical role of localized physicochemical properties, rather than overall sequence similarity, in determining receptor recognition and host range.

Within PA26, the pyocin knob connector harbours two arginine residues (positions 793, 804), forming a positively charged motif potentially interacting with negatively charged LPS components. Similarly, mutation hotspots, such as S793 and D910, remain targets for future validation via mutagenesis.

The C-terminal lectin-like domain, despite structural similarity to R2 pyocins and T4 phage [45], showed no correlation with host range. However, its role in stabilizing adsorption via conserved LPS core structures may support broader specificity [55].

The GDSL-like domain with a carbohydrate-binding module, also reported in phage G7C [43], is part of the SGNH hydrolase superfamily, whose members hydrolyse ester bonds in diverse substrates [44]. This domain may enhance infectivity by degrading bacterial surface components, including biofilms [56]. In our study, it was enriched in polar residues (Asn, Asp), which could facilitate hydrogen bonding with O-antigen sugars. While no direct link to host specificity was found, its physicochemical features may modulate adsorption efficiency.

Lastly, we observed a tertiary fold similarity between the GrpE-like domain of the phage tail fibre protein and the PilA monomer, the major subunit of type IV pili (T4P) in Pseudomonas aeruginosa. Although no direct interaction or terminal alignment was detected, this structural resemblance may reflect convergent evolution toward surface-associated functionality. Given that T4P plays a key role in host adhesion, motility, and biofilm formation [38], the similarity between phage TFP and host PilA could suggest an evolutionary adaptation enabling the phage to mimic or exploit host surface structures during adsorption.

Collectively, our findings suggest that TFP-mediated host specificity arises from localized physicochemical optimization, especially in knob2, magnified by the trimeric, multivalent structure. Even single substitutions may dramatically affect binding affinity, providing a mechanistic basis for tropism shifts. Functional studies (e.g. D910N, S793R mutations) are now warranted to validate these hypotheses.

Putative receptor binding mechanism

The phage initially encounters P. aeruginosa through Brownian motion. The GrpE domain may transiently stabilize this interaction, preventing premature detachment. As the virion approaches the surface, the remaining TFP domains recognize LPS as the primary receptor. The CBM specifically binds the O-antigen, facilitating access of the catalytic residues to substrate bonds. The distal GDSL-like domain, containing an SGNH-like catalytic tetrad (Ser, Gly, Asn, His), likely initiates hydrolysis of the O-antigen, e.g., via deacetylation of sugar residues. This enzymatic activity, positioned upstream of the receptor-binding site, may reduce steric hindrance and increase polysaccharide flexibility, aiding access to deeper LPS regions [57, 58]. Analogous mechanisms have been described in phage G7C, where tail fibre-associated deacetylases enhance epitope exposure [43]. We propose that this dual function—enzymatic remodelling followed by receptor binding—is essential for efficient adsorption (see Fig. 5A). Subsequently, the pyocin knob region, enriched in asparagine, binds hydrophilic O-antigen components such as ManNAc3NAcA. Structural analysis suggests that knob1 and knob2 are spatially arranged to engage specific repeating motifs within the O-antigen, enabling multivalent, cooperative binding (Fig. 5A). This configuration enhances specificity and avidity. Finally, the C-terminal domain binds core LPS sugars, such as 3-deoxy-D-manno-octulosonic acid (Kdo) or phosphate-rich regions, enabling electrostatic stabilization via positively charged residues. Additionally, short tail fibres, as secondary receptors, contribute to LPS core binding, as reported for phage DEV [13], facilitating genome injection.

Conclusions

This study demonstrates that phage host specificity in Pseudomonas aeruginosa-infecting viruses from the Migulavirinae subfamily is shaped not only by overall structural rearrangements but also by subtle physicochemical variations in receptor-binding domains. Our integrative approach combining host range phenotyping, molecular identification, and domain-resolved structural modelling reveals that key residues, particularly polar and charged amino acids, within the pyocin knob domain may drive precise adaptations towards diverse bacterial targets. Interestingly, despite recombination across genera, the tail fibre protein maintains a conserved modular organization, enabling evolutionary flexibility without structural compromise. These insights offer valuable clues for the rational design of phage-based antimicrobials and broaden our understanding of molecular host recognition mechanisms within environmental phage populations.

Methods

Bacterial strains and phages

For laboratory analysis, 22 clinical and 1 laboratory strains of Pseudomonas aeruginosa were utilized as described earlier [8] together with 2 standard P. aeruginosa strains (ATCC 27853 and ATCC 10145; Microbiologics). For host range studies, we used previously characterized Litunaviruses: vB_Pae575P-3, vB_Pae1369P-5 [8], PA26 [30] and vB_PaeP_MAG4 [25], as well as phages newly isolated in this study.

Bacteria and previously isolated bacteriophage propagation

For P. aeruginosa strain propagation, a standard Luria–Bertani (LB, Graso, Poland) medium was used, either solid, containing 1.5% agar (w/v) (BTL, Poland), or liquid. Bacteria were cultivated aerobically with shaking (liquid cultures) for 20–24 h at 37 °C. For bacteriophage propagation, the double-layer agar (DLA) method was used, as described before [8]. Briefly, 0.2 mL of overnight bacterial culture was added to 3.8 mL of LB, containing molten 0.7% agar (w/v) (BTL, Poland) at 45 °C and poured onto a Petri dish, containing solid LB medium to prepare a bacterial lawn. Serial tenfold dilutions of phage lysates were prepared in TM buffer (10 mM Tris–HCl, 10 mM MgSO₄; pH 7.2) and spotted onto or added into the bacterial lawn prepared as described above to obtain single phage plaques. When dry, plates were inverted and incubated at 37 °C for 20–24 h. Single plaques were then cut out of the top agar layer, added to host bacteria growing exponentially in 3 mL of LB liquid medium and cultivated at 37 °C with vigorous shaking until lysis occurred. Then, 300 µL of chloroform (Merck, Germany) was added to each lysate, vortexed and centrifuged at 3000 × g. The resulting supernatants were kept in the dark at 4 °C until further use.

New phage isolation from raw municipal wastewater

Since the collection lacked the representative of Luzseptimavirus, potential candidates were isolated from municipal wastewater, as it was described earlier [8, 31, 59]. Raw municipal sewage samples were obtained from “Dębogórze” Group Wastewater Treatment Plant (Dębogórze-Wybudowanie, Poland) and used for new bacteriophage isolation. For this aim, overnight liquid LB cultures of P. aeruginosa strains were mixed into 3 cocktails (S—sensitive, SR—semi-resistant, R—resistant) according to their preliminary sensitivity for infection (Additional file 1: Table S5) [8, 25, 28, 30]. Each cocktail contained 2 mL of each bacterial strain. Next, mixtures were enriched with sewage, constituting 10% (v/v) of their final volume, and set for overnight incubation with shaking at 37 °C. After this time, each mixture was subjected to chloroform extraction as described above; the resulting supernatants were screened against bacterial hosts from cocktails in search of clearing zones, using a spot test [60] and DLA method, as described above. Those supernatants that revealed clearing zones were further tenfold, serially diluted, and added into the lawns of sensitive hosts. Resulting phage plaques were propagated in their host, extracted with chloroform and kept in the dark at 4 °C until further use. To ensure the uniformity of bacteriophage lysates, the single plaque isolation step was repeated 3 times for each lysate.

Phage lysates purification

All phage lysates were propagated using the same sensitive host strain, followed by their precipitation using 10% polyethylene glycol (PEG; Roth, Germany) and 1 M NaCl (Merck, Germany) with stirring at 4 °C overnight. The precipitate was centrifuged (30 min, 10,000 × g) and resuspended in TM buffer with 1 M NaCl. PEG residues were removed by repeated chloroform extraction, and the solution underwent ultracentrifugation in caesium chloride (CsCl; Roth, Germany) density gradient (1.45, 1.50, 1.70 g/ml) for 2 h in 67,500 × g. The resulting phage bands were collected and dialysed against Spectra/Por 4 membrane (Roth, Germany) in TM buffer for a week [10]. All lysates were kept in the dark at 4 °C until host spectrum determination and electron microscope analysis.

Host spectrum determination

A total of 22 clinical isolates of P. aeruginosa were selected based on their genetic diversity, antimicrobial resistance profiles, biofilm-forming ability and clinical origin (wounds, suppuration, decubitus ulcers, bronchial mucus and patients with cystic fibrosis) to represent a broad spectrum of phenotypes commonly encountered in hospital settings (Additional file 1: Table S5) [8, 25, 28, 30]. Additionally, two reference strains (ATCC 27853 and ATCC 10145) were included as standard controls for comparative purposes. All P. aeruginosa strains were subjected to host spectrum determination using the DLA method and the spot test. LB agar plates with all bacterial strain lawns were spotted [60] with 3 µL of tenfold dilutions of phages’ CsCl density gradient-purified phage lysates. Additionally, lysates after chloroform extraction only and PEG precipitation, followed by chloroform extraction, were tested along with CsCl-purified lysates for comparison. When dry, plates were inverted and incubated overnight at 37 °C. All phage–host interaction experiments (DLA and spot tests) were performed in three independent biological replicates, each starting from a separate overnight culture of the tested bacterial strain. For each replicate, phage dilutions were freshly prepared and applied separately. The observed lysis profiles (e.g. presence/absence and clarity of plaques or lysis zones) were recorded independently for each replicate. A phage was considered capable of infecting a host strain only if lysis was observed in all three replicates under the same preparation method. Inconsistent results were excluded from further interpretation.

Adsorption assay

The phage adsorption assay was performed as previously described by Topka-Bielecka et al. (2020), with minor modifications. To evaluate the adsorption efficiency of phages during the infection cycle, overnight cultures of P. aeruginosa strains were inoculated into LB medium and incubated at 37 °C with shaking until reaching a cell density of 10⁸ CFU/mL (after approximately 180 min). Next, 1 mL of each culture was centrifuged at 2000 × g at 4 °C for 10 min to separate bacterial cells from the medium. The supernatant was discarded, and the resulting bacterial pellet was resuspended in 1 mL of 0.85% NaCl solution and centrifuged again. After discarding the supernatant, the washed bacterial cells were resuspended in 1 mL of LB medium and incubated at 37 °C with shaking for 15 min to allow acclimation. Following this, phage lysates were added to reach a multiplicity of infection (MOI) of 0.1 and 0.01 for two parallel replicates of the experiment. Samples were then incubated with shaking at 37 °C for a total of 40 min. During this period, 50 µL of the cultures were collected at 11 time points: 0 s (immediately after phage addition), 30 s, 1 min, 3 min, 5 min, 10 min, 15 min, 20 min, 25 min, 30 min, and 40 min (end of the experiment). At each time point, the samples were centrifuged at 10,000 × g for 2 min to separate bacterial cells with adsorbed phages from unbound phages. The supernatant containing unbound phages was then serially tenfold diluted in TM buffer. Dilutions were spotted onto DLA bacterial lawns as described above, allowed to dry, inverted, and incubated at 37 °C for 20–24 h. After incubation, plates showing individual plaque formation were selected, and plaques were used for PFU/mL determination by spotting onto fresh DLA lawns.

Transmission electron microscopy

For single phage particle observation, the previously described procedure was applied [61]. Briefly, CsCl density gradient-purified phage lysate was adsorbed onto carbon-coated copper mesh grids, stained with uranyl acetate (1.5%) and visualized using a transmission electron microscope (Tecnai Spirit BioTWIN, FEI) with a magnification of 340,000 ×. Microphotography was performed by the Laboratory of Electron Microscopy (Faculty of Biology, University of Gdansk, Poland).

Isolation and genomic characterization of phage DNA

The viral DNA was liberated from virions using the method described previously [8]. To degrade the bacterial nucleic acids, all samples were treated with DNase and RNase (EURx, Gdańsk, Poland) and incubated for 1 h at 37 °C. In the next step, the enzymes were inactivated for 10 min at 75 °C. The encapsulated phage DNA was then extracted with NucleoSpin Virus Kit (Macharay-Nagel GmbH & Co., Düren, Germany) according to the manufacturer’s recommendations. Purified DNA was used in PCR amplification of the portal protein gene (843 base pairs [bp]; forward primer 5′AAGGCTGATGGTCGGTACAAG ′3 and reverse primer 5′CAGATCGAAGTTGCCACGCAA ′3—this study; Merck) and GrpE-like domain of the tail fibre protein (TFP) gene (1164 bp; forward primer 5′AAGGTGATCTTCACCTTCCA ′3 and reverse primer 5′CAAACCAGTGAACATGGT ′3—this study; Merck). Final concentrations of reagents within the 25 μL reaction volume were: 0.1 μM of each primer, 0.6 U of Taq polymerase (EURx, Gdańsk, Poland), 0.2 mM dNTPs, and Taq Polymerase buffer (15 mM MgCl₂; EURx, Gdańsk, Poland). The reactions were conducted with a protocol consisting of preliminary denaturation (5 min, 95 °C), 30 cycles (denaturation − 30 s, 95 °C; annealing − 30 s, 56 °C; extension − 90 s, 72 °C), and final extension (5 min, 72 °C). PCR products were purified using NucleoSpin Gel and PCR Clean-Up Kit (Macharay-Nagel GmbH & Co., Düren, Germany) and sequenced (Macrogen Europe, Amsterdam, Netherlands). Sequences were trimmed and compared to records available in the NCBI GenBank using BLASTn (NCBI, Bethesda, USA; accessed on March 31, 2024).

To verify the identification and recover complete TFP gene sequences, DNA extracts from the bacteriophages were subjected to whole-genome sequencing (Genomed, Warsaw, Poland). Following DNA fragmentation, sequencing libraries were prepared using the NEBNext Ultra II DNA Library Prep Kit (New England Biolabs) according to the manufacturer’s recommendations for Illumina platforms. Sequencing was carried out on the Aviti system, generating 300-nt paired-end reads. Raw reads were quality-filtered using Fastp v0.23.4 [62], and de novo genome assemblies were constructed with metaviralSPAdes v3.15.5 [63].

Tail fibre protein modelling, domain determination and structure similarity comparison

Full TFP amino acid sequences of phages were submitted to AlphaFold [36] for structure prediction. A detailed list of Litunavirus and Luzseptimavirus phages, TFP gene and amino acid sequences with their accession numbers is listed in Table S6 in Additional file 1. The structure was polymerized into a trimer, as mentioned previously [13]. Results were analysed using Chimera X 1.7.1 [37], and exact domain boundaries were determined based on both sequence alignment and structural features of models, with names of domains assigned by NCBI Conserved Domain Data Base [64]. Next, spatial models of all TFP domains were calculated using the Robetta [42] online server. Models with the best resolution score (lowest Å values) [65, 66] were chosen for structural alignment and comparison with P. aeruginosa R1 pyocin fibre (accession number in Protein Data Base PDB: 6CL5), fimbria protein model of PilA monomer of PAK pilin (PDB: 1OQW) [67], and T4 phage long tail fibre (PDB: 2XGF) [45], using the Chimera X program and Matchmaker tool provided by the software [37].

To investigate the potential interaction between the domains of the tail fibre protein (TFP) of analysed phages and the O-antigen of Pseudomonas aeruginosa, 3D structural models of lipopolysaccharide (LPS) components were generated. The LPS core with a single O-antigen repeating unit was obtained from the Carbohydrate Structure Database (CSDB ID: 23,236) [68] and exported in PDB format using the CSDB/SNFG structure editor available on the same platform. The O-antigen fragment consisting of 30 repeating units was built using GLYCAM-Web [69] via the “Build via Text” option with the following input string: [4DManpNAcb1-4DManpNAcb1-3DFucpa1-] < 30 > OH. The structure was then generated and exported in PDB format [69]. Both PDB files were imported into Chimera X, where the LPS components were spatially aligned with the domains of the TFP for visualization of possible binding interfaces [37]. This step aimed to illustrate the structural feasibility of the TFP-O-antigen interaction.

Abbreviations

bp

Base pairs

CBM

Carbohydrate-binding module

DLA

Double-layer agar

GDSL

Gly-Asp-Ser-Leu

Kdo

3-Deoxy-D-manno-octulosonic acid

LPS

Lipopolysaccharide

nt

Nucleotide

OmpC

Outer membrane protein C

PEG

Polyethylene glycol

PFU

Plaque-forming units

SGNH

Ser-Gly-Asn-His

T4P

Type IV pili

TFP

Tail fibre protein

Authors’ contributions

Conceptualization, A.Z., M.G., A.J-K., M.K. and H.M.; methodology, A.Z., M.G and A.J-K.; validation, M.G. and A.J-K.; formal analysis, A.Z., A.J-K., M.G; investigation, A.Z., A.J-K., M.G.; writing—original draft preparation, A.Z., A.J-K.; writing—review and editing, A.Z., M.G., A.J-K., M.K. and H.M.; visualization, A.Z., M.G.; supervision, M.G. and A.J-K.; funding acquisition, A.Z. All authors read and approved the final manuscript.

Funding

Project has been funded by the internal grant program for Young Scientists 2023, provided by the Dean of the Faculty of Biology at the University of Gdańsk. Project number: 539-D200-B086-23.

Faculty of Biology,University of Gdansk,539-D200-B086-23

Data availability

All data generated or analysed during this study are included in this published article, its supplementary information files and publicly available repositories (FigShare DOI: 10.6084/m9.figshare.30231001).

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.