Binding of Bacillus subtilis dynamin‐like protein DynA to the bacterial membrane is essential for effective phage defense

TL;DR

This study shows that the bacterial protein DynA protects Bacillus subtilis from phage infection by binding to the membrane, which helps prevent cell rupture.

Contribution

The study identifies the specific membrane-binding site in DynA and demonstrates its essential role in phage resistance.

Findings

Membrane binding by DynA is mediated by lysine and phenylalanine residues in the D1 subunit.

Mutations disrupting membrane binding increase phage susceptibility in B. subtilis.

DynA stabilizes the bacterial membrane to prevent explosive lysis during phage infection.

Abstract

Bacterial dynamin‐like proteins are large GTPases that play crucial roles in membrane dynamics. Bacillus subtilis dynamin‐like protein A (DynA), a two‐headed bacterial dynamin‐like protein, possesses membrane‐binding and membrane‐tethering functions in trans. The formation of large DynA clusters on bacterial membranes in response to pore‐forming antibiotics and phages demonstrates its potential role in maintaining bacterial membrane integrity under various environmental stresses. In this study, we identified the membrane‐binding site of B. subtilis DynA within the D1 subunit of the protein that includes positively charged lysine residues K360 and K367, as well as hydrophobic phenylalanine residues F363, F364, and F365. For experimental validation, recombinant proteins with amino acid substitutions in the lysine and phenylalanine residues were produced and used in liposome binding assays. Nonconservative substitutions led to a complete loss of DynA's membrane‐binding capability. In vivo data showed strains with DynA variants lacking membrane‐binding capability exhibit significantly increased susceptibility to phage infection compared with wild‐type cells, further emphasizing the importance of DynA's membrane interaction in conferring phage resistance. Our findings bridge the gap between the structural characteristics of DynA and its functional implications in maintaining bacterial membrane integrity and mediating phage resistance.

The bacterial dynamin‐like protein DynA from Bacillus subtilis plays a key role in phage defense. It binds the plasma membrane through a unique paddle domain in its D1 subunit. This membrane association is essential for its protective function, likely preventing explosive membrane rupture following phage infection. By stabilizing the membrane, DynA helps protect the bacterial population against explosive lysis, minimizing dispersal of bacteriophages.

Figures

AlphaFold v2.0 model of B. subtilis bacterial dynamin‐like protein DynA. (A) DynA possesses two dynamin subunits D1 (purple) and D2 (cyan) with their GTP‐binding domains. The D1 subunit possesses a putative membrane‐binding site (pink) at the end of the extended stalk. (B) A close‐up view of the putative membrane‐binding site highlighting the arrangement of key lysine residues at Position 360 and 367 and three phenylalanine residues at Positions 363, 364, and 365.

GTPase activity of 5 μm DynA and DynA variants. Basal GTPase activity was retained in DynA variants (F363A/F364A/F365A, K360A/F367A, and F363W/F364W/F365W), confirming preservation of their structural and functional integrity. The GTPase‐deficient variant DynA K56A/K625A served as the negative control. Error bars indicate the standard deviation of three independent experiments.

Liposome sedimentation assays of wild‐type (WT) DynA and DynA variants. (A) Graphical illustration of the liposome sedimentation assay. Large, unilamellar vesicles (LUVs) were incubated with the protein at room temperature, then centrifuged to separate the supernatant and liposome containing pellet. Protein in both fractions was analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS/PAGE). (B) SDS/PAGE gel image of liposome sedimentation assays of WT DynA and DynA variants where ‘s’ indicates proteins in supernatant and ‘p’ indicates the proteins in pellet. (C) Supernatant to pellet ratios of DynA, DynA (F363A, F364A, F365A), DynA (K360A, K367A), DynA (K56A, K625A) and DynA (F363W, F364W, F365W) variants, respectively. WT DynA exhibits a low supernatant to pellet ratio indicating effective membrane binding. While both DynA F363A, F364A, F365A and DynA K360A, K367A variants remained in the supernatants, the conservative variant DynA F363W, F364W, F365W retained the membrane binding. Statistical analysis is based on One‐way ANOVA (top bracket) and post hoc Tukey's Honestly Significant Difference (HSD) test for pairwise significance (, P value less than 0.01; *, P value less than 0.001). Error bars depict standard deviation of three independent experiments.

Liposomes incubated with 25 μm DynA in the absence of the nucleotide were examined by transmission electron microscopy. Wild‐type DynA binds the liposome membrane, leading to tethering and clumping of liposomes. Arrowheads point to the electron‐dense layer representing contrasted DynA protein on the surface of cup‐shaped liposomes. In contrast, in the presence of the DynA F363A, F364A, F365A variant, liposomes remain dispersed. In the presence of the DynA K360A and K367A variant, liposomes were destabilized (double arrowheads) and insoluble lipids accumulated on the film (arrows). Liposomes were embedded in 0.2% visualized by TEM after embedding in 1.8% methyl cellulose with 0.2% uranyl acetate. Scale bar: 1 μm (upper panel), 200 nm (lower panel).

Membrane binding of DynA is essential for bacterial innate immunity. The lytic effect of Ф29 infection (MOI = 1) on different B. subtilis strains was tested (i) by measuring the optical densities of liquid growth cultures in a plate reader, and (ii) by ɸ29 plaque formation: photographs of agar plates are displayed in the middle column; quantitative analysis is presented in the right column. Statistical analysis is based on One‐way ANOVA (top bracket) and post hoc Tukey's Honestly Significant Difference (HSD) test for pairwise significance (*, P value less than 0.05; , P value less than 0.01; *, P value less than 0.001). Error bars indicate the standard deviation of three independent experiments. The following bacterial strains were used for comparison: wild‐type (WT) B. subtilis 168 (WT), DynA‐deficient strain (ΔdynA), DynA‐overexpressing strain (DynA++). (A) The DynA F363A, F364A, F365A ++ variant strain shows a similar pattern of cell lysis as the DynA‐deficient strain (i), and larger plaques compared with WT (ii). (B) The DynA K360A, K367A ++ variant strain shows a similar pattern of cell lysis as that of the DynA‐deficient strain (i), and larger plaques compared with WT (ii). (C) Cells expressing only the D1 subunit of DynA show resistance to the phage infection similar to the overexpressing strain, while the D2 subunit alone is ineffective in preventing host cell lysis. Amino acid substitutions in the D1 subunit (F363A, F364A, F365A) render the community sensitive to phage infection (i, ii). (D) DynA conservative substitutions show a protective effect against ɸ29 infection (i, ii).

Tables

Table 1. Bacterial strains used in this study.

Strain	Genotype	Source/reference
Bacillus subtilis strains
168	trpC2	Laboratory collection
25 152	trpC2, metBl O, xin‐1, SPβ‐(erm)	Laboratory collection
FBB002	trpC2, dynA::tet	[32]
FBB018	trpC2, amyE::Pxyl‐dynA‐gfp spc	[32]
SSB004	trpC2, amyE::Pxyl‐dynA‐F363A‐F364A‐F365A‐gfp spc dynA::tet	This study
SSB005	trpC2, amyE::Pxyl‐dynA‐K360A‐K367A‐gfp spc dynA::tet	This study
LJ‐B05	trpC2, amyE::Pxyl‐D1‐yfp‐spec dynA::tet	This study
LJ‐B06	trpC2, amyE::Pxyl‐D2‐gfp‐spec dynA::tet	This study
SSB006	trpC2, amyE::Pxyl‐D1 F363A‐F364A‐F365A‐yfp‐spec dynA::tet	This study
SSB007	trpC2, amyE::Pxyl‐dynA‐F363W‐F364W‐F365W‐gfp spc dynA::tet	This study
Escherichia coli strains
BL21 (DE3)	F− omp T hsd SB (rB −, mB −) gal dcm (DE3)	Novagen
SSE001	BL21 (DE3) pET16b_DynA‐His6	This study
SSE002	BL21 (DE3) pET16b_DynAF363A,F364A,F365A‐His6	This study
SSE003	BL21 (DE3) pET16b_DynAK360A,K367A‐His6	This study
SSE004	BL21 (DE3) pET16b_DynAF363W,F364W,F365W‐His6	This study
SSE005	BL21 (DE3) pET16b_DynAK56A,K625A‐His6	This study

Full text

Introduction

The dynamin superfamily of proteins is a large group that plays crucial roles in several essential cellular processes [1], including fission and fusion of biological membranes [2, 3], endocytosis and exocytosis [4], membrane remodeling [5], and immune responses to pathogens [6]. This large family is further divided into classical dynamins and dynamin‐like proteins (DLPs), based mainly on sequence homology and notable structural features, including the highly conserved N‐terminal GTPase domain, a central hinge domain, and an extended alpha‐helical stalk domain [7]. The proteins are mechanochemical GTPases, which utilize the energy of GTP hydrolysis to remodel cellular membranes. They are over 60 kDa and can self‐assemble on biological membranes, which activates their GTPase activity [8]. GTP binding and hydrolysis are mediated by four well‐conserved motifs within the GTPase domain: G1, G2, G3, and G4 [1]. The G1 motif primarily binds to phosphates, the G2 motif, with a conserved threonine, interacts with Mg^2+^ ions to carry out GTP hydrolysis, the G3 motif also interacts with Mg^2+^ ions through aspartate residues, and the conserved glycine interacts with the γ‐phosphate, and the G4 motif interacts with the nucleotide base [9].

After the discovery of the first dynamin from bovine tissue and the shibire (ts) mutant in fly, it became quickly clear that dynamin‐like proteins are ubiquitous in eukaryotic cells [10, 11]. The eukaryotic dynamin‐1 was the first dynamin protein studied, and it is involved in endocytosis by assembling into large helical structures around the neck of budding vesicles. This assembly is followed by GTPase activity, which releases the vesicles from the plasma membrane [12, 13, 14, 15]. The other two isoforms, dynamin‐2 and dynamin‐3, which share 80% sequence similarity, mainly differ in their proline‐rich domains and expression profiles [16]. Eukaryotic DLPs are classified as either membrane fission or membrane fusion DLPs based on their biological functions [17]. Mitochondrial division by Drp1 and peroxisomal division by DRP3A are examples of fission DLPs [18, 19], while membrane‐fusing DLPs, such as Fzo1 mitofusins and Mgm1/OPA1, fuse the outer and inner mitochondrial membranes, respectively, and Atlastins fuse membranes of the endoplasmic reticulum [3, 20, 21]. A fascinating class of DLPs, called Mx proteins, has been shown to have a protective effect against a wide range of DNA and RNA viruses, making them part of the eukaryotic innate immune response [22]. However, their mode of action in antiviral activity is still largely unclear [23].

Bacterial dynamin‐like proteins (BDLPs), a subclass of DLPs, were first predicted in 1999, and since then, several hundred bacterial species have been identified to possess DLPs [24]. BDLPs gained attention when the full‐length crystal structure of BDLP1 from Nostoc punctiforme was reported [25]. While the exact function of BDLP1 in N. punctiforme remained unclear, it demonstrated ordered self‐assembly on liposome surfaces, reshaping them into highly curved tubules while disrupting the outer lipid layer [7]. Similarly, the cyanobacterial DLP forms oligomers in solution, suggesting membrane‐independent self‐assembly [26]. BDLPs have been found in both Gram‐positive and Gram‐negative bacteria. Some BDLPs, such as DynA and DynB from Streptomyces venezuelae, a filamentous bacterium, have been shown to play a role in cell division during sporulation [27] Other BDLPs, such as IniA and IniC in Mycobacterium tuberculosis, and LeoA in enterotoxigenic Escherichia coli, are involved in extracellular vesicle release [28, 29], while the BDLP from Campylobacter jejuni is speculated to fuse biological membranes [30, 31].

DynA, a two‐headed BDLP in Bacillus subtilis, was first described in 2011 and found to have two connected dynamin‐like subunits, D1 and D2, suggesting it resulted from gene duplication and fusion events [32, 33]. In vitro studies have revealed that DynA can tether and fuse membranes in trans. The D1 subunit of DynA is crucial for membrane tethering and fusion, while the D2 subunit facilitates the process [32, 34]. GTP binding facilitates full membrane fusion and likely DynA disassembly, but is not required for membrane tethering and lipid mixing [32, 34]. Further research has shown that DynA is involved in a complex stress response pathway that protects the plasma membrane's integrity. Deletion of dynA from B. subtilis renders the bacteria more sensitive to environmental stresses, such as pore‐forming antibiotics and phages [35]. The protein exhibits high lateral mobility in the plasma membrane but forms large, confined clusters in response to the pore‐forming antibiotic nisin, suggesting its recruitment to sites of membrane damage [35, 36]. This protective effect of DynA was further studied using the smallest B. subtilis phage, Ф29. DynA‐deficient strains exhibited increased susceptibility to phage infection, as shown by larger plaque formation and extensive lysis in liquid cultures. Interestingly, DynA‐overexpressing strains showed significantly less and delayed host cell lysis at the population level. During phage infection, DynA formed large immobile clusters in the later stages, indicating its role in blocking phage progeny release by preventing membrane rupture [36]. DynA is the first reported BDLP with a critical role in bacterial stress response and innate immunity.

This study focuses on characterizing the membrane‐binding site within the D1 subunit of DynA. Using a DynA AlphaFold model as a template, we identified a candidate membrane‐binding site comprising key positively charged lysine residues (K360 and K367) and hydrophobic phenylalanine residues (F363, F364, and F365). These residues play a significant role in facilitating interaction with biological membranes. To investigate the functional relevance of this membrane‐binding site, we engineered a series of point mutations in the dynA gene to produce DynA protein variants. Our findings reveal that the DynA variant, DynA F363A, F364A, and F365A, as well as DynA K360A and K367A, exhibit a complete loss of membrane‐binding capability, as demonstrated by liposome sedimentation assays. In contrast, a variant engineered with conservative substitutions, DynA F363W, F364W, and F365W, maintained its membrane‐binding ability, suggesting that structural integrity around these residues is crucial for DynA's membrane interaction. Furthermore, we employed electron microscopy to visualize DynA's binding to lipid membranes. Wild‐type (WT) DynA was observed to envelop liposomes, inducing liposome clustering indicative of successful membrane interaction, while the nonconservative variants displayed a stark absence of binding, further confirming our biochemical findings. The in vivo relevance of DynA's membrane‐binding capability was assessed through phage infection assays. We found that the strains possessing DynA variants lacking membrane‐binding exhibited significantly increased susceptibility to lysis compared with WT strains, highlighting the importance of this interaction in conferring phage resistance. Moreover, our experiments demonstrated that the D1 subunit alone could effectively mediate this protective function when expressed in B. subtilis, emphasizing its pivotal role.

Results

Membrane‐binding site of DynA

DynA consists of two structurally similar subunits, D1 and D2, both containing GTP‐binding domains essential for GTP hydrolysis [32]. However, previous studies indicated that the D1 subunit is primarily responsible for membrane tethering and fusion, as demonstrated by in vitro assays [32, 34]. In our initial screening of potential membrane‐binding regions within the D1 subunit, we identified a candidate site using an AlphaFold v2.0 generated model [37]. The model shows two globular GTPase domains within each subunit located in close proximity to each other. Each of these GTPase domains is connected to an alpha‐helical stalk region. The stalk of the D1 subunit exhibits an extended conformation while the stalk of the D2 subunit adopts a more folded confirmation forming an overall compact structure of DynA. The extended tip of the D1 stalk has two positively charged lysine residues at Positions 360 and 367, along with three consecutive hydrophobic phenylalanine residues at Positions 363, 364, and 365, making this region electrostatically and hydrophobically ideal to interact with lipid membranes (Fig. 1). Moreover, this specific arrangement of residues is only present within the stalk of the D1 subunit while being absent in the D2 subunit.

AlphaFold v2.0 model of B. subtilis bacterial dynamin‐like protein DynA. (A) DynA possesses two dynamin subunits D1 (purple) and D2 (cyan) with their GTP‐binding domains. The D1 subunit possesses a putative membrane‐binding site (pink) at the end of the extended stalk. (B) A close‐up view of the putative membrane‐binding site highlighting the arrangement of key lysine residues at Position 360 and 367 and three phenylalanine residues at Positions 363, 364, and 365.

DynA F363A, F364A, F365A and DynA K360A, F367A variants lack membrane binding

To further examine the region between residues K360 and K367 as a potential membrane‐binding site, we introduced three amino acid substitutions to create the DynA F363A, F364A, F365A variant and two amino acid substitutions to generate the DynA K360A, K367A variant. Additionally, we engineered a third variant, DynA F363W, F364W, F365W, with conservative substitutions to assess whether the membrane‐binding capability of DynA is retained under these conditions. We also included the previously described DynA K56A, K625A variant, which lacks GTPase activity, as a control in our experiments [32].

The WT DynA protein and its four variants were heterologously expressed with a poly‐His tag in E. coli and purified in a two‐step chromatography (see Materials and Methods section and Fig. S1). All proteins were purified to homogeneity (Fig. S1) and subsequently used for in vitro assays. DynA, DynA F363A, F364A, F365A variant, DynA K360A, F367A variant, and DynA F363W, F364W, F365W variant exhibited basal GTPase activity confirming structural and functional integrity, while DynA K56A, K625A variant lacks GTPase activity and was used as a control (Fig. 2). We interpreted the presence of GTPase activity as an indication that the generated variants of DynA contained their overall fold.

GTPase activity of 5 μm DynA and DynA variants. Basal GTPase activity was retained in DynA variants (F363A/F364A/F365A, K360A/F367A, and F363W/F364W/F365W), confirming preservation of their structural and functional integrity. The GTPase‐deficient variant DynA K56A/K625A served as the negative control. Error bars indicate the standard deviation of three independent experiments.

Next, we evaluated the membrane‐binding ability of DynA and its variant proteins using liposome sedimentation assays. Large, unilamellar vesicles (LUV), approximately 400 nm in diameter, were prepared and incubated at 24 °C with the protein of interest. The protein–liposome mixture was then subjected to ultracentrifugation, and the resulting pellets and supernatants were separated and analyzed for protein presence using SDS/PAGE (Fig. 3A).

*Liposome sedimentation assays of wild‐type (WT) DynA and DynA variants. (A) Graphical illustration of the liposome sedimentation assay. Large, unilamellar vesicles (LUVs) were incubated with the protein at room temperature, then centrifuged to separate the supernatant and liposome containing pellet. Protein in both fractions was analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS/PAGE). (B) SDS/PAGE gel image of liposome sedimentation assays of WT DynA and DynA variants where ‘s’ indicates proteins in supernatant and ‘p’ indicates the proteins in pellet. (C) Supernatant to pellet ratios of DynA, DynA (F363A, F364A, F365A), DynA (K360A, K367A), DynA (K56A, K625A) and DynA (F363W, F364W, F365W) variants, respectively. WT DynA exhibits a low supernatant to pellet ratio indicating effective membrane binding. While both DynA F363A, F364A, F365A and DynA K360A, K367A variants remained in the supernatants, the conservative variant DynA F363W, F364W, F365W retained the membrane binding. Statistical analysis is based on One‐way ANOVA (top bracket) and post hoc Tukey's Honestly Significant Difference (HSD) test for pairwise significance (**, P value less than 0.01; **, P value less than 0.001). Error bars depict standard deviation of three independent experiments.

The results from the liposome sedimentation assay demonstrated a complete loss of membrane‐binding ability in both the DynA F363A, F364A, F365A and DynA K360A, K367A variants compared with the WT DynA. In contrast, the DynA K56A, K625A variant, while lacking GTPase activity, retained its membrane‐binding capability, as previously described [32]. Moreover, the variant with conservative substitutions, DynA F363W, F364W, F365W, was able to bind liposomes (Fig. 3B,C). These data provide compelling evidence that the region between residues K360 and K367 constitutes the membrane‐binding site (paddle domain) of DynA.

Electron microscopy reveals loss of membrane binding in DynA F363A, F364A, F365A and DynA K360A, K367A

To gain further information on the interaction of DynA with liposome membranes, we employed transmission electron microscopy to gain visual insights into the DynA–membrane interaction. When 200 nm liposomes were incubated with 25 μm WT DynA, the protein was observed to completely envelop the liposomes, and several liposomes aggregated to form clusters held together by DynA. Arrowheads indicate the electron‐dense layer representing the contrasted DynA protein on the surface of the cup‐shaped liposomes. On the contrary, the DynA F363A, F364A, F365A variant was unable to attach to or aggregate any liposome vesicles. Similarly, the DynA K360A, K367A variant showed no binding affinity for the liposomes; however, liposome destabilization and degradation were evident in the presence of this protein (Fig. 4).

Liposomes incubated with 25 μm DynA in the absence of the nucleotide were examined by transmission electron microscopy. Wild‐type DynA binds the liposome membrane, leading to tethering and clumping of liposomes. Arrowheads point to the electron‐dense layer representing contrasted DynA protein on the surface of cup‐shaped liposomes. In contrast, in the presence of the DynA F363A, F364A, F365A variant, liposomes remain dispersed. In the presence of the DynA K360A and K367A variant, liposomes were destabilized (double arrowheads) and insoluble lipids accumulated on the film (arrows). Liposomes were embedded in 0.2% visualized by TEM after embedding in 1.8% methyl cellulose with 0.2% uranyl acetate. Scale bar: 1 μm (upper panel), 200 nm (lower panel).

DynA's membrane binding is essential for phage protection

Phage Ф29 produces small plaques on a lawn of B. subtilis cells [38]. Previous studies from our laboratory have demonstrated a unique protective effect of DynA during Ф29 phage infection. Cells lacking DynA exhibited significant host cell lysis in the presence of phages [35], as shown by plaque assays and infection assays in liquid cultures, suggesting a protective effect of DynA. This protective effect of DynA becomes more pronounced with DynA overexpression, resulting in a substantial reduction in host cell lysis [36]. Our objective was to determine whether the membrane‐binding capability of DynA is essential for its in vivo protective effect against phage infection. To investigate this, B. subtilis strains were generated that expressed various DynA variant proteins as a sole DynA copy ectopically from the amyE locus under a xylose‐inducible promoter. In vivo assays in liquid medium and on agar plates were performed to test the resistance of these strains to Ф29 phage lysis.

First, we analyzed the lysis behavior of a B. subtilis strain expressing the DynA F363A, F364A, F365A variant (strain SSB004). The DynA variant was expressed as a GFP fusion to allow microscopic inspection. Exponentially growing cultures of WT B. subtilis, the DynA deletion strain (B. subtilis, ΔdynA), the B. subtilis (DynA F363A, F364A, and F365A++) strain, and the DynA overexpression strain (DynA++) were infected with a multiplicity of infection (MOI) of 1, and optical density was measured continuously in a plate reader over a period of 14 h. The strain expressing DynA F363A, F364A, F365A as a sole copy exhibited significantly increased cell lysis, similar to the deletion strain, indicating that the absence of DynA–membrane binding renders cells more susceptible to lysis upon phage infection (Fig. 5Ai). A similar effect was observed in plaque assays, where the lawn of the B. subtilis (DynA F363A, F364A, F365A++) infected with Ф29 displayed much larger plaques and twice the number of plaques compared with WT B. subtilis 168 (Fig. 5Aii).

Membrane binding of DynA is essential for bacterial innate immunity. The lytic effect of Ф29 infection (MOI = 1) on different B. subtilis strains was tested (i) by measuring the optical densities of liquid growth cultures in a plate reader, and (ii) by ɸ29 plaque formation: photographs of agar plates are displayed in the middle column; quantitative analysis is presented in the right column. Statistical analysis is based on One‐way ANOVA (top bracket) and post hoc Tukey's Honestly Significant Difference (HSD) test for pairwise significance (, P value less than 0.05; **, P value less than 0.01; **, P value less than 0.001). Error bars indicate the standard deviation of three independent experiments. The following bacterial strains were used for comparison: wild‐type (WT) B. subtilis 168 (WT), DynA‐deficient strain (ΔdynA), DynA‐overexpressing strain (DynA++). (A) The DynA F363A, F364A, F365A ++ variant strain shows a similar pattern of cell lysis as the DynA‐deficient strain (i), and larger plaques compared with WT (ii). (B) The DynA K360A, K367A ++ variant strain shows a similar pattern of cell lysis as that of the DynA‐deficient strain (i), and larger plaques compared with WT (ii). (C) Cells expressing only the D1 subunit of DynA show resistance to the phage infection similar to the overexpressing strain, while the D2 subunit alone is ineffective in preventing host cell lysis. Amino acid substitutions in the D1 subunit (F363A, F364A, F365A) render the community sensitive to phage infection (i, ii). (D) DynA conservative substitutions show a protective effect against ɸ29 infection (i, ii).

Next, we analyzed the B. subtilis strain synthesizing the DynA K360A, K367A++ variant (strain SSB005), again expressed as a GFP fusion. The infection trend for this strain mirrored that of the phenylalanine triple substitution variant described before, showing a higher degree of cell death when exponentially growing cells were infected with Ф29 at a MOI of 1 compared with the WT strain. In contrast, the DynA overexpression strain (DynA++) demonstrated resistance to cell lysis, consistent with prior observations (Fig. 5Bi). The plaque assay for the B. subtilis (DynA K360A, K367A++) strain revealed larger and more numerous plaques, similar to those seen in the DynA deletion strain, underscoring the importance of positively charged residues for membrane binding (Fig. 5Bii).

Previous work based on liposome sedimentation assays has shown that the D1 subunit of DynA exhibits in vitro membrane binding, while the D2 subunit does not possess any binding ability and only assists in maintaining membrane cluster stability [32]. We aimed to determine whether the D1 subunit alone could fulfill the functional role of DynA in vivo. To this end, we constructed a B. subtilis strain expressing only the D1 subunit of DynA, fused to YFP (strain: LJ‐B05), and a B. subtilis strain expressing only the D2 subunit of DynA, fused to GFP (strain: LJ‐B06). We compared the D1++ expressing strain to a D2++‐expressing strain by simultaneously infecting both with Ф29 phages during their exponential growth phase at a MOI of 1. The D1++‐expressing strain showed resistance to cell lysis, while the D2++ expressing strain exhibited no resistance phenotype. Additionally, we created a B. subtilis D1 (F363A, F364A, F365A)++ strain (SSB006), expressed as a GFP fusion. This D1 (F363A, F364A, F365A)++ expressing strain was sensitive to cell lysis upon infection with Ф29 phages (Fig. 5Ci). The plaque assay results were consistent with the infection assays in liquid cultures, further confirming that membrane binding is integral to the D1 subunit of DynA and that this binding is essential for its protective effect against phages (Fig. 5Cii).

A B. subtilis (DynA F363W, F364W, F365W++) strain (strain SSB007) with hydrophobic side chains was developed to assess whether we can conserve the membrane binding of DynA in vivo. This DynA variant was again expressed as a GFP fusion. Exponentially growing cultures of WT B. subtilis, DynA deletion strain (B. subtilis, ΔdynA), B. subtilis (DynA F363W, F364W, F365W++) strain, and DynA overexpression strain (DynA++) were infected with Ф29 phages at a MOI of 1, and optical density was measured. The strain expressing DynA F363W, F364W, F365W++ variant showed significantly less cell lysis, similar to the DynA overexpression strain (DynA++), further confirming that phenylalanine residues at Positions 363, 364, and 365 are integral parts of the membrane‐binding site of DynA (Fig. 5Di). Similar results were obtained by plaque assays (Fig. 5Dii). In summary, these data confirm that membrane binding of the D1 subunit of DynA plays a crucial role in phage defense.

Discussion

The bacterial dynamin‐like protein DynA in B. subtilis is part of a novel phage protection system that acts as a critical last line of defense against phage infection. By delaying host cell lysis and preventing the immediate release of phage progeny after replication and assembly, DynA helps to mitigate the rapid spread of phages through the bacterial population [36]. To fully understand the molecular mechanisms underlying this protective effect, it is essential to explore the complex structure of DynA and the roles of its individual subunits and domains. DynA is a membrane‐associated protein with an approximate molecular weight of 137 kDa. It consists of two subunits, D1 and D2, and the holoenzyme dimerizes in vitro. Both subunits contain independent GTP‐binding domains. Each GTP‐binding domain features a central P‐loop motif, which includes an essential lysine residue crucial for nucleotide binding. Mutation of both P‐loops (K65A and K625A) in DynA abolishes GTP binding and hydrolysis, as shown in previous studies [32]. Notably, sequence analysis of DynA reveals the absence of traditional transmembrane domains [32]. However, earlier studies also revealed that DynA and the DynA D1 subunit co‐sediment with liposomes, while the DynA D2 subunit does not exhibit membrane binding. This suggests that membrane tethering is governed exclusively by the D1 subunit, whereas the D2 subunit plays a role in stabilizing the DynA complex [32, 34]. Interestingly, DynA and a distantly related antiviral dynamin Mx1 share a unique functional similarity. Neither the DynA–membrane tethering nor the Mx1 interaction with the viral ribonucleoprotein requires GTP hydrolysis [32, 39]. However, both proteins require GTP hydrolysis for their antiviral protective effect [36, 40].

In our study, we used an AlphaFold model to investigate the potential membrane‐binding site within the D1 subunit of DynA. This analysis identified two positively charged lysine residues at Positions 360 and 367, alongside three consecutive hydrophobic phenylalanine residues at Positions 363, 364, and 365 in the extended stalk region of the D1 subunit as a potential membrane‐binding site. Previous studies on human MxA proteins have documented the importance of lysine residues at the tip of the stalk region in facilitating electrostatic interactions with cellular membranes [41, 42].

To test the potential membrane‐binding role of this region in DynA, we introduced substitutions in the D1 subunit. Specifically, we generated the DynA F363A, F364A, F365A variant and the DynA K360A, K367A variant. Both variants demonstrated a complete loss of membrane binding in liposome sedimentation assays. Interestingly, the DynA F363W, F364W, F365W variant, which incorporated conservative substitutions, retained the ability to co‐sediment with liposomes. These findings suggest that the arrangement of positively charged lysine residues surrounding hydrophobic phenylalanine residues in the D1 stalk region is essential for DynA's membrane‐binding capability. Similarly, hydrophobic residues in the paddle domain of BDLP from N. punctiforme have been shown to be involved in membrane binding, where mutations in phenylalanine and leucine residues disrupted lipid association [7]. Therefore, the paddle domain of BDLP and DynA shares structural and functional similarities.

Next, we aimed to investigate the importance of DynA's membrane binding in vivo. Previous studies have demonstrated that DynA plays a unique protective role against phage infection by delaying host cell lysis [36]. However, the molecular mechanism of DynA in bacterial innate immunity remained unclear. To determine with this protective effect is mediated through direct interactions with the plasma membrane, we compared the lysis behavior of B. subtilis strains expressing either the DynA F363A, F364A, F365A or DynA K360A, K367A variant, as the sole DynA copy, to that of the WT strain during phage infection. Both variant strains exhibited significantly increased cell lysis during Ф29 phage infection compared with the WT strain, confirming that DynA's protective effect against phage infection is lost when its membrane‐binding ability is compromised. Furthermore, a strain expressing the sole D1 subunit of DynA exhibited reasonable protection against phage infection, while substitutions in the D1 subunit (F363A, F364A, F365A) failed to protect against phage infection. These results suggest that the DynA D1 subunit directly interacts with the membrane via its membrane‐binding site, forming a physical barrier to prevent rupture during phage‐induced host cell lysis, whereas the D2 subunit has no direct role during phage infection and may only facilitate the D1 subunit in stabilizing DynA–membrane complexes [34].

DynA protective phenotype against phages is similar to Mx dynamins that are a part of the innate immune response in eukaryotes by possessing antiviral activity. Both dynamins share remarkable structural and functional similarities [35, 42]. However, unlike MxA, DynA shows no direct interaction with the phage particle so far, but the possibility persists. Two interesting bacterial dynamins, LeoA of enterotoxigenic Escherichia coli (ETEC) and IniA from Mycobacterium tuberculosis, have been implicated in membrane fission and release of membrane vesicles [43, 44]. IniA has been speculated to assemble on the periplasmic side of the plasma membrane to bud membrane vesicles encapsulating antituberculosis drugs, thus contributing to resistance against antituberculosis drugs [45]. In contrast, DynA binds the membrane from the cytoplasmic site, and hence production of vesicles that could act as phage decoys is unlikely. However, there is a possibility that DynA protects against phages by forming large assemblies on the plasma membrane where it captures phage progeny that remain stuck at the lysing cell debris, thus preventing disruptive bursts and phage progeny dispersal.

In summary, this study provides a detailed understanding of the structural and functional characteristics of DynA, linking its membrane‐binding ability to its protective role against phage infection. By elucidating the specific interactions at the membrane‐binding site within the D1 subunit of DynA, we provide the first evidence that DynA's function in bacterial innate immunity requires its membrane‐binding activity.

Materials and methods

Bacterial strains

All bacterial strains used in this study are listed in Table 1.

Growth and isolation of phages

Ф29 phages were obtained from the German Collection of Microorganisms (DSMZ GmbH). The host strain B. subtilis 25 152 was cultured to mid‐log phase in LB medium at 37 °C with continuous shaking. High‐titer phages (approximately 10^8^–10^10^ PFU) were then added to the bacterial culture. The mixture was incubated at 37 °C for 6 h with shaking. After incubation, the phage–host mixture was centrifuged at 3800 ** g ** for 20 min to separate cell debris. The phage supernatant was filtered through 0.2 μm pore filters and stored at 4 °C.

Phage purification by NaCl‐PEG precipitation

One liter of the bacteria–phage mixture was centrifuged at 3800 ** g ** for 20 min to separate the phages from the cell debris. The supernatant was then filtered, and solid NaCl was gradually added to a final concentration of 1 m, stirring until fully dissolved. Next, 1% solid PEG 8000 (Millipore Sigma, Burlington, MA, USA) was slowly added to the supernatant, and the mixture was stirred continuously for 30 min. The phage mixture was pelleted by centrifugation at 15 000 ** g ** for 10 min at 4 °C. The supernatants were gently decanted without disturbing the pellet. The pellets were then completely resuspended in 20 mL of SM Buffer (100 mm NaCl, 8 mm MgSO_4_·7H_2_O, 50 mm Tris‐Cl, pH 7.5). The phages in SM buffer were centrifuged again at 15 000 ** g ** for 10 min at 4 °C. The purified phages were separated from the pellets and stored at 4 °C.

Phage infection assay in liquid medium

Infection assays using Φ29 phage were conducted following the method described previously [36]. Overnight cultures of B. subtilis and strains that expressed DynA variant proteins were diluted to an OD_600_ of 0.1 and grown until optical density reached approximately 0.3. The cultures were then infected with Φ29 phage at a multiplicity of infection (MOI) of 1, and optical density was measured every 10 min using Infinite200 PRO (Tecan, Grödig, Austria) over a period of 14 h at 37 °C with continuous shaking.

Quantitative plaque assay

Quantitative plaque assays using Φ29 phage were conducted following the method described previously [36]. Overnight cultures of B. subtilis were diluted to an OD_600_ of 0.1 in fresh LB medium and grown until they reached an OD_600_ of 0.5–1.0. The purified phages were serially diluted tenfold (from 10^1^ to 10^10^) in gelatin‐free SM buffer. For each dilution, 100 μL of the phage solution was mixed with 1 mL of bacterial culture and incubated at room temperature for 10 min to facilitate phage attachment. Each bacterial‐phage mixture was then combined with 4 mL of overlay agar (containing 0.4% agar in LB medium) and poured onto LB agar underlay plates (1% agar). The plates were incubated at 24 °C for 18 h, after which plaques were observed.

Protein purification

Purification of DynA and its variants was performed according to the previously described method, with some modifications [32]. Heterologous expression of DynA‐His₆ and its variants was conducted in Escherichia coli BL21 (DE3) strains in LB (lysogeny broth) medium supplemented with the appropriate antibiotics (carbenicillin at 100 μg·mL^−1^). Protein expression was induced by adding IPTG to a final concentration of 0.5 mm when the OD_600_ reached approximately 0.7. The cells were incubated overnight at 18 °C. Cultures were harvested by centrifugation at 6500 ** g ** for 15 min at 4 °C, washed with T2 buffer (50 mm Tris, 200 mm NaCl, 20 mm imidazole, 10% glycerol, pH 8.0), and then centrifuged again at the same speed and temperature. The cell pellets were frozen and stored at −80 °C.

The frozen pellets were resuspended in cold T2 buffer and supplemented with lyophilized DNase I and Protease Inhibitor Cocktail (cOmplete™, EDTA‐free, Roche 04693132001), along with 0.7% Triton X‐100. The suspension was passed through a French Pressure Cell (SLM Instruments, Rochester, NY USA) to disrupt the cells, with the process repeated three times at an inner pressure of 20 000 psi. The resulting suspension was centrifuged at 15 000 ** g ** at 4 °C using a Beckman Coulter Avanti J‐25 centrifuge with a JA‐10 rotor and Falcon adapters to remove cell debris. The supernatant was transferred to a fresh Falcon tube and mixed with T2‐buffered Ni‐NTA agarose (Qiagen), followed by incubation at 4 °C overnight with continuous mixing.

The following day, the bound protein was extensively washed with T2 buffer and eluted using T5 buffer (50 mm Tris, 500 mm NaCl, 1 m imidazole, 10% glycerol, pH 8.0) for 1 h at 4 °C with continuous shaking. The proteins were concentrated as needed using Amicon filter devices with the appropriate molecular weight cutoff (Millipore; Merck, Darmstadt, Germany). The eluted protein was reduced with 1 mm DTT and subjected to size‐exclusion chromatography using a Superose 6 Increase column (Cytiva, Marlborough, MA, USA) in 50 mm Tris, 500 mm NaCl, 10% glycerol, pH 8.0. Finally, the protein concentration was determined using the Bradford assay and stored at −80 °C until further use.

GTP hydrolysis assays

GTP hydrolysis was performed based on a previously described method [46]. GTPase activity of DynA and its variants was analyzed in continuous reactions using the EnzChek Phosphate Assay Kit (E‐6646, Molecular Probes Inc., now Thermo Fisher, Waltham, MA, USA), following the manufacturer's instructions. Reaction volumes were reduced to 100 μL and assayed in flat‐bottom 96‐well plates (Greiner‐UV‐Star 96‐well plates). Reactions were measured at 37 °C, continuously, every minute, over a 3‐h time course in a Tecan Infinite 200 Pro with the Tecan i‐control v.2.0 software. Each reaction contained 5 μm DynA or the respective DynA variants, buffered in 50 mm Tris, 200 mm NaCl, pH 8.0. Before starting the reactions with 2 mm Mg‐GTP, the reactions were preincubated for 10 min to eliminate phosphate contamination. Data were analyzed with Excel (normalized, with subtraction of GTP auto‐hydrolysis and subtraction of the no‐substrate control, and linear regression was used to determine the hydrolysis rate per minute). Data visualization and further analysis were performed with GraphPad Prism version 5.03 for Windows (GraphPad Software).

Liposome sedimentation assay

A thin film of Avanti total Escherichia coli polar lipids was dissolved in chloroform and dried under a nitrogen stream for 3 h. The dried film was then rehydrated in 50 mm Tris, 10% glycerol, pH 7.1, at 37 °C for 1 hour. The resulting suspension was diluted to 2 mg·mL^−1^ in assay buffer (50 mm Tris, 200 mm NaCl, 10% glycerol, pH 7.4) and extruded 40 times through a 400 nm pore size filter using a mini extruder. Two micromolar DynA and its variant proteins were incubated with 1 mg·mL^−1^ liposome solution at 25 °C. The mixture was then ultra‐centrifuged at 90 000 ** g ** for 20 min. Nucleotides were used at a 1 mm concentration with 5 mm MgSO_4_. The supernatants were carefully separated from the pellets, and protein presence was analyzed using SDS/PAGE.

Liposome preparation for electron microscopy

A thin film of Avanti total Escherichia coli polar lipids was dissolved in chloroform and dried under a nitrogen stream for at least 1 h. The lipids were further dehydrated in a vacuum dryer for another 4 h. The lipids were then rehydrated in 50 mm Tris, 10% glycerol, pH 7.1, at 37 °C for 3 h. The resulting suspension was diluted to 2 mg·mL^−1^ in assay buffer (50 mm Tris, 200 mm NaCl, 10% glycerol, pH 7.4) and extruded 40 times through a 200 nm pore size filter using a mini extruder.

Electron microscopy

For the liposome membrane‐binding assay, 25 μm DynA/variant protein was incubated with 1 mg·mL^−1^ liposomes in the absence of the GTP nucleotide at 24 °C for 10 min. The samples were immediately analyzed by negative staining transmission electron microscopy. To this end, copper, 400 square mesh, formvar‐coated TEM grids were glow discharged for 60 s at 0.6 mbar air pressure and 10 mA glow current using a Safematic CCU‐010 unit, and then incubated with 5 μL of the liposome suspension for 4 min. Grids were washed shortly on 7 drops of water and embedded in a thin layer of 1.8% methyl cellulose with 0.2% uranyl acetate. Samples were imaged with a Tecnai G2 Spirit BioTwin transmission electron microscope (FEI/now Thermo Fisher Scientific, Waltham, MA, USA) operated at 80 kV using TEM User interface (v. 4.2) and equipped with a LaB6 filament. Images were recorded with a MegaView III G2 CCD camera, using the iTEM v.5 software (both Olympus Soft Imaging solutions/now EMSIS).

Statistical analysis

We used the SPSS software (SPSS Inc., Chicago, Ill., USA) to calculate statistical significances.

Conflict of interest

The authors declare no conflict of interest.

Author contributions

SS performed the majority of the experiments and data analysis and wrote the original manuscript. UR performed the electron microscopy. MB contributed to the conceptualization, supervision, data analysis, wrote original manuscript with input from all authors, and secured funding.

Supporting information

Fig. S1. SDS/PAGE gel stained with Coomassie blue showing purified DynA proteins.