A phospholipase B effector exploits membrane lipid divergence to drive inter-Bacteroides competition and enables universal genome editing

ABSTRACT

The Type VI secretion system (T6SS) mediates interbacterial competition, yet the mechanisms underlying its effector diversity and cell-targeting specificity remain poorly understood. Here, we identify Btle, a phospholipase B (PLB) effector defining a unique Tle subfamily (Tle7), encoded in approximately 18.0% of Bacteroides fragilis strains. Lipidomic profiling reveals that Btle preferentially targets Bacteroides glycerophospholipids, disrupting membrane integrity and driving strict inter-Bacteroides competition, with self-intoxication prevented by the cognate immunity protein Btli1. Guided by structural modeling and rational mutagenesis, we developed a Btle variant (Btle^M) that evades Btli1 neutralization, establishing a broadly applicable counterselection marker for homologous recombination across Bacteroides. Secretion assays combined with AlphaFold structural prediction demonstrate that Btle delivery depends on a dual-effector complex, uncovering a conserved secretion mechanism within this genus. Collectively, this work expands the Tle superfamily with a calcium-dependent PLB effector, clarifies how glycerophospholipid compositional divergence shapes interbacterial targeting specificity, and establishes a universal counterselection strategy for genetic manipulation of gut Bacteroides.

Introduction

The human gastrointestinal tract harbors one of the most densely populated and diverse microbial ecosystems on earth, whose balance profoundly shapes host health and disease. ^1-5 Bacteroidota, the dominant Gram-negative gut phylum, achieve stable gut establishment through intense interbacterial antagonism during early colonization, ⁶ and broadly affect other microbes and host physiology through secreted metabolites. ^7-9

The Type VI Secretion System (T6SS) is a key contact-dependent weapon deployed by Gram-negative bacteria, including Escherichia coli, Vibrio cholerae, and Bacteroides fragilis, for interbacterial competition in the gut. ^10-13 Operating as a molecular syringe, the T6SS translocates effectors directly from the attacker into target cells, where they strike peptidoglycan, ¹⁴ cell membrane, ^{15 , 16} nucleic acids, ¹⁷ small molecules ^18-20 and proteins. ^21-23 In Bacteroides, T6SSs are classified into three genetic architecture (GA) types: GA1, GA2, and GA3. GA1 and GA2 are typically encoded on integrative conjugative elements (ICEs), whereas GA3 is unique to B. fragilis. ^{24 , 25} The GA3 cluster contains two variable regions, V1 and V2, which encode distinct effector-immunity pairs conferring competitive advantages. ^{26 , 27} The V1-encoded effectors Bte1 and Bfe1 could be delivered via an Hcp2-Hcp3 heterohexamer, ²⁸ whereas the V2-encoded effectors BtpeA and BtaeB employ a coordinated paired-delivery mechanism, ²⁹ underscoring the breadth of translocation strategies within this system.

Among the cellular structures targeted by T6SS effectors, the bacterial membrane represents a conserved yet chemically diverse and particularly vulnerable interface. ^{30 , 31} Type VI lipase effectors (Tle) from Proteobacteria, which target bacterial membrane, fall into six superfamilies. ^{15 , 32} Tle1–Tle4 and Tle6 carry the conserved G-X-S-X-G motif and typically display phospholipase A (PLA) or B (PLB) activity, ^33-37 while Tle5 harbors the H-X-K-XXXX-D motif and acts as a phospholipase D (PLD). ¹⁵

The enzymatic activities and delivery mechanisms of Tle1–Tle6 have been extensively characterized, with structural studies illuminating their catalytic domains, membrane-binding regions, and inactivation mechanisms. ^{38 , 39} Most characterized Tle proteins are delivered by VgrG via specific protein–protein interactions. ³⁵ Beyond membrane disruption, they serve diverse functions: biofilm regulation, the "tit-for-tat" retaliatory response in Pseudomonas, modulation of motility, and activation of the PI3K/Akt pathway in eukaryotic cells to promote bacterial internalization. ^{33 , 40} Collectively, these findings highlight the remarkable functional diversity of the Tle family. Nevertheless, the bactericidal spectra and species specificity of Tle proteins remain poorly understood, and despite the T6SS being a recognized virulence determinant in Bacteroides, Tle proteins in this genus remain virtually unstudied.

Here, we report Btle (Bacteroides T6SS lipase effector), a lipase effector encoded within the V2 locus of B. fragilis GS075 among eight hypothetical proteins. Btle harbors the canonical G-H-S-H-G motif and a Ser-His-Asp catalytic triad, displays phospholipase B activity, and occupies a distinct phylogenetic branch within the Tle family. Functional assays demonstrate that Btle targets and permeabilizes recipient bacterial membranes, with lipidomic profiling implicating divergent phospholipid compositions as the basis of its target selectivity. We further demonstrate that self-protection is mediated by the cognate immunity protein Btli1, whereas the adjacent paralogue Btli2 is non-functional. Rational mutagenesis yielded Btle^M, a Btle variant that escapes Btli1 neutralization and serves as a novel and universal counterselection marker for Bacteroides genome editing. Additionally, GS075_1252 (a PAAR-like protein), GS075_1253 (a BtapC homolog), and GS075_1256 (a candidate effector) are identified as key contributors to T6SS assembly and effector delivery. Together, our findings expand the Tle superfamily, uncover membrane phospholipid composition as a determinant of T6SS target specificity, and establish an effector-derived genetic tool for manipulating gut Bacteroides.

Results

Btle is a Bacteroides T6SS effector

Sequencing of B. fragilis isolates from our human microbiome culture collection revealed a GA3-type T6SS cluster in B. fragilis strain GS075 with a notably gene-rich V2 region, containing eight open reading frames encoding proteins of unknown function (Figure 1a). Phylogenomic analysis revealed that the GS075-like GA3 T6SS, which harbors vgrG and V2 regions highly similar to those of the GS075 configuration, occurs in approximately 18.0% of B. fragilis strains. This finding suggests that it represents a common but underexplored V2 architecture.

Figure 1.

A T6SS effector of B. fragilis GS075 is required for interbacterial antagonism. (a) Gene organization of the V1 and V2 regions of the B. fragilis GS075 T6SS locus. Scale bar, 1 kb. (b) Signal peptide prediction and domain architecture of the GS075_1251–GS075_1249 locus. (c) Western blot detection of secreted Btle^HA and Hcp1 in culture supernatants (Sup) and whole-cell lysates (WCL) of B. fragilis GS075-btle ^HA and the GS075 -btle ^HA-Δ tssC mutant. (d) Co-culture competition assays between the indicated donor strains (GS075::CmR, GS075Δbtle::CmR, or GS075ΔtssC::CmR) and the recipient GS075ΔbtleΔbtli1-2::ErmR. Data are mean ± SD; significance assessed by t-test (ns, non-significant; ****, P < 0.0001). (e) In vivo competitive colonization of antibiotic-treated mice inoculated with donor (GS075::CmR) and recipient (GS075ΔbtleΔbtli1-2::ErmR) strains (n = 5). Data are mean ± SEM.

JackHMMER analysis of these eight proteins identified a putative T6SS effector, ⁴¹ which we designated Btle (Bacteroides T6SS lipase effector; GS075_1251). Btle carries a lipase (class 3) domain (PF01764.28) featuring the G-X-S-X-G motif, a hallmark of esterases and lipases, including Tle1–Tle4. ¹⁵ Unusually, rather than a single immunity gene, two homologous genes, namely btli1 (GS075_1250) and btli2 (GS075_1249), lie immediately downstream of btle, each encoding a putative immunity protein bearing an N-terminal lipoprotein signal peptide (Figure 1b). ⁴²

To determine whether Btle secretion relies on the T6SS machinery, we generated a GS075 strain expressing C-terminal HA-tagged Btle (GS075-btle ^HA) and a T6SS-deficient mutant in this background (GS075-btle ^HA-ΔtssC). Secretion assays verified extracellular accumulation of Btle in the GS075-btle ^HA strain, whereas no secreted Btle was detected in the GS075-btle ^HA-ΔtssC mutant (Figure 1c). To assess competitive function, we generated a Δbtle single mutant and a ΔbtleΔbtli1-2 triple mutant. In interbacterial competition assays using GS075ΔbtleΔbtli1-2::ErmR as the recipient, the GS075Δbtle::CmR donor showed complete loss of antagonistic activity, comparable to GS075ΔtssC::CmR, whereas wild-type GS075::CmR exhibited strong killing (Figure 1d). We further performed monocolonization and co-colonization experiments in SPF mice. Under monocolonization conditions, wild-type (GS075::CmR) and mutant strain (GS075ΔbtleΔbtli1-2::ErmR) exhibited comparable colonization capacity (Figure S1a–b). By contrast, the wild-type strain displayed a marked colonization advantage during co-colonization (Figure 1e). Together, these data establish Btle as a T6SS-secreted effector with a G-X-S-X-G motif and lipase domain that mediates interbacterial antagonism.

Btle is a calcium-dependent Tle with phospholipase B activity

Having established Btle as a T6SS-dependent competitive factor, we next investigated its mechanism of toxicity. We performed compartment-specific expression of Btle in B. fragilis NCTC9343 using cytoplasmic or periplasmic inducible systems. Periplasmic expression (ssbtle) caused severe growth inhibition, whereas cytoplasmic expression exhibited no detectable growth defect (Figure 2a). Similar periplasmic toxicity was observed in B. thetaiotaomicron VPI-5482 and B. ovatus ATCC 8483 (Figure 2b). We further performed in vitro co-culture competition assays using GS075Δbte1::ErmR and GS075Δbte1Δbtle::ErmR as donor strains, and Bo::CmR and Bt::CmR as recipient strains. Compared with the GS075Δbte1Δbtle::ErmR control strain, the GS075Δbte1::ErmR strain, which expresses and secretes functional Btle, exhibited significant antagonistic activity against both Bt::CmR and Bo::CmR recipients (Figure 2c). This observation suggests broad-spectrum activity of Btle, which likely reflects the conserved phospholipid composition of Bacteroides membranes. ^{43 , 44}

Figure 2.

Btle is a calcium-dependent PLB effector. (a–b) Spotting assays of serially diluted (10-fold) B. fragilis NCTC9343, B. thetaiotaomicron VPI-5482, and B. ovatus ATCC 8483 strains carrying the indicated constructs on BHI agar with or without aTc induction. ss, Bacteroides-derived signal peptide; CT, uninduced control. (c) Co-culture competition assays between the indicated donor strains (GS075Δbte1::ErmR and GS075Δbte1Δbtle::ErmR) and the recipients (Bo::CmR and Bt::CmR). (d) Esterase activity of purified Btle incubated with Tween-20 and 3 mM CaCl₂ at 37 °C overnight. Data are mean ± SD of three biological replicates. (e) Predicted catalytic triad residues of Btle shown as sticks. (f) Esterase activity of wild-type Btle and catalytic triad mutants. (g) Spotting assay of serially diluted B. fragilis NCTC9343 strains carrying the indicated constructs on BHI agar with or without aTc. (h) Representative GC chromatogram showing free FA15:0 released from PE(15:0_15:0) after incubation with Btle in buffer containing 3 mM CaCl₂, in the absence or presence of 5 mM EDTA. (i) Schematic illustrating the strategy to distinguish PLA or PLB activity of Btle. (j) Phylogenetic tree of Tle family members; bootstrap support is proportional to circle size.

To characterize enzymatic activity, we first assessed Btle’s hydrolysis of Tween-20. ⁴⁵ Under standard conditions (3 mM CaCl₂), visible calcium precipitate of the fatty acid (FA) product formed (Figure 2d). Since lipases and phospholipases commonly employ a Ser-His-Asp catalytic triad, with the catalytic Ser embedded in the G-X-S-X-G motif, we used AlphaFold3 to identify candidate His and Asp residues near Ser129. ^{46 , 47} The predicted α/β hydrolase fold domain was modeled with high confidence (Figure S2a), revealing His457 at 3.0 Å from Ser129 and Asp239 at 2.6 Å from His457 (Figure 2e), implicating Ser129-His457-Asp239 as the catalytic triad. Purified single-point mutants Btle^S129A, Btle^D239A, and Btle^H457A all lacked esterase activity in vitro (Figure 2f), and periplasmic expression of each mutant in B. fragilis NCTC9343 markedly reduced antimicrobial activity relative to wild-type (Figure 2g), confirming that this triad is essential for both catalysis and bactericidal function.

During Tween-20 assays, increasing Ca²⁺ concentrations (10–50 mM) markedly enhanced reaction (Figure S2b), indicating Ca²⁺ contributed to Btle catalysis rather than merely precipitating fatty acid products. Given that B. thetaiotaomicron is enriched in PE(15:0_15:0), ⁴³ we performed GC-based assays to assess the calcium dependence of Btle phospholipase activity. Btle produced abundant FA15:0 from PE(15:0_15:0) in the presence of calcium, whereas EDTA-mediated calcium chelation strongly reduced product formation (Figure 2h). AlphaFold3 modeling of the Btle-Ca²⁺ complex identified a contiguous electronegative cluster predicted to coordinate Ca²⁺ via Asp242 and Asp243 (Figure S2c–d). Alanine substitution at either residue abolished enzymatic activity in reaction systems even supplemented with 30 mM CaCl₂ (Figure S2e), demonstrating that Btle is a calcium-dependent lipase.

To examine the PLA or PLB activity of Btle, we incubated Btle with 500  μM PE(15:0_15:0). For this substrate, PLA activity would produce no more than 500 μM FA15:0, while PLB hydrolysis yields up to 1000 μM FA15:0 (Figure 2i). Released FA15:0 was quantified by GC against a FA15:0 calibration curve (Figure S2f). The measured peak area corresponded to 745.6 μM FA15:0, with the entire confidence interval (gray shaded area) above 500 μM, confirming that Btle acts as a PLB.

Phylogenetic analysis of Btle alongside Tle1–Tle6 revealed that Btle and its homologs form a novel clade, phylogenetically distant from all six established Tle branches (Figure 2j). Although overall sequence identity within each family is low, the canonical G‑X‑S‑X‑G motif is strongly conserved across family members (Tle1: GFSRG; Tle2/Tle6: GHSLG; Tle3: XHSQG; Tle4: THSMG). Notably, Btle and its homologs carry a conserved GHSHG motif that differs from the corresponding motifs in Tle1–Tle6. Together with the phylogenetic clustering, this distinct motif feature distinguishes Btle and its homologs from previously defined Tle families, supporting their designation as a new family, Tle7.

Btle-mediated phospholipid hydrolysis disrupts membrane integrity in target cells

To verify Btle’s in vivo enzymatic activity and mechanism, we performed 4 h induction starting at either 0 h or 1 h to express Btle in the periplasm of the sensitive strain B. thetaiotaomicron VPI-5482 (Bt). Unlike the catalytically inactive Btle^S129A mutant or the empty-vector control, periplasmic Btle expression significantly increased propidium iodide (PI) uptake, indicating compromised membrane integrity (Figure 3a). ⁴⁸ These results confirm that Btle is an antibacterial phospholipase that disrupts the target cell membrane.

Figure 3.

Periplasmic Btle expression disrupts membrane integrity. (a) Confocal micrographs of B. thetaiotaomicron VPI-5482 expressing periplasmic Btle, stained with propidium iodide (PI) for membrane integrity and FM 1-43 for membrane visualization. The ssbtle 1 h group was supplemented with aTc after subculture 1 hour, while the remaining three groups were treated with aTc at the time of subculture. Scale bar, 5 μm. (b–c) Growth curves (OD₆₀₀) of B. thetaiotaomicron VPI-5482 and E. coli C43(DE3) expressing periplasmic Btle or Btle^S129A following induction with aTc (b) or IPTG (c) at the indicated time points. Blue, green, and orange arrows correspond to inductions at 0 h, 1 h, and 2 h, respectively. (d) Confocal micrographs of E. coli C43(DE3) expressing periplasmic Btle or Btle^S129A, stained with PI and FM 1-43. The pelB-btle 1 h group was supplemented with IPTG after subculture 1 hour, while the remaining three groups were treated with IPTG at the time of subculture. Scale bar, 5 μm. (e–f) GC-MS analysis of products from Btle-mediated hydrolysis of B. thetaiotaomicron (e) and E. coli (f) membrane fractions in vitro.

Despite targeting the conserved bacterial cell membrane, Btle could be expressed and purified in large quantities from the periplasm of E. coli C43(DE3), suggesting that E. coli is relatively insensitive to its activity. Consistent with this, induction of Btle expression starting at 0 h, 1 h, and 2 h post-inoculation revealed that Btle strongly inhibited the growth of Bt (Figure 3b). In contrast, the growth of E. coli was only weakly inhibited, and this mainly occurred when Btle was induced at 0 h (Figure 3c). Correspondingly, PI uptake in E. coli peaked upon induction at 0 h but remained much lower than in Bt (Figure 3d). We further confirmed that Btle was low expressed in B. thetaiotaomicron but abundantly expressed in E. coli and efficiently translocated into the periplasm (Figure S3a–b).

To probe the biochemical basis of this differential activity, we incubated membrane fractions from B. thetaiotaomicron and E. coli with Btle or lysis buffer, and measured fatty acid release by GC-MS. Btle released substantially more fatty acids (e.g., FA15:0, FA14:0, and FA13:0) from B. thetaiotaomicron membranes than that from E. coli membranes (Figure 3e–f), confirming greater enzymatic activity against Bt phospholipids and suggesting that differential activity against membrane phospholipids contributes to target selectivity.

Divergent glycerophospholipid composition underlies Btle’s reduced activity against E. coli

To understand why Btle preferentially targets Bt membranes, we performed lipidomic profiling of B. thetaiotaomicron and E. coli expressing periplasmic Btle. PCA score plots clearly separated Btle-expressing strains (Bt-Btle and Ec-Btle) from empty-vector controls (Bt and Ec), with tight intra-group clustering confirming strong biological reproducibility (Figure 4a). It also resolved Bt from Ec groups, indicating distinct lipid compositions between the two strains. Glycerophospholipids constitute the major lipid components in both Bt and Ec, accounting for ~84.7% of total lipids in Bt and ~94.2% in Ec (Figure 4b). Within the glycerophospholipid pool, phosphatidylethanolamine (PE), the dominant class in Bt (~82.2%), declined to ~6.9% in Bt-Btle (Figure 4c). Volcano plot analysis confirmed substantially greater reductions in PE species abundance in Bt-Btle versus Bt (Figure 4d). Notably, lysophosphatidylethanolamine (LPE) did not accumulate in Bt-Btle; rather, multiple LPE species were significantly depleted (Figure 4e), which is consistent with phospholipase B activity.

Figure 4.

Btle remodels the phospholipid composition of B. thetaiotaomicron and exhibits substrate preference. (a) PCA of lipid profiles from Btle-expressing strains (Bt-Btle and Ec-Btle) and empty-vector control (Bt and Ec). (b) Relative abundances of glycerophospholipid across Bt and Ec. (c) Relative abundance of individual glycerophospholipid species in Bt and Bt-Btle. (d–e) Volcano plots depicting Btle-induced changes in PE (d) and LPE (e) species abundance in Bt. (f) Relative abundance of PE and LPE in Ec and Ec-Btle. (g) The relative abundance of glycerophospholipids with different total acyl carbon counts in Bt and Ec. (h) Relative abundance of glycerophospholipids with different total acyl carbon counts from Bt (black bars) and Bt-Btle (red bars). (i) Relative abundance of diverse major glycerophospholipid species in Bt and Bt-Btle. (j) GC chromatograms detecting free fatty acids released from different PE substrates after incubation with Btle.

In E. coli, Btle expression caused a moderate reduction in PE abundance (from ~85.7% to ~41.8%), whereas LPE species were even moderately elevated in Ec-Btle (Figure 4f). Consistent with the weaker Btle-mediated E. coli killing and lower free fatty acid release (Figure 3c–d and Figure 3f), these results confirm poor membrane cleavage activity of Btle against E. coli, further demonstrating the pronounced substrate preference of Btle.

Although glycerophospholipid dominated in both bacteria (Figure 4b), individual glycerophospholipid species differed substantially (Figure S3c). Notably, Bt membranes were dominated by total acyl carbon-30 (36.2%), followed by carbon-31 and carbon-29, whereas Ec glycerophospholipids were dominated by total acyl carbon-33, followed by carbon-34 (Figure 4g), indicating that Ec glycerophospholipids generally possess longer acyl chains. Consistent with this difference, short-acyl-chain glycerophospholipid species were markedly depleted in Bt upon Btle expression, whereas longer-chain glycerophospholipids (total acyl carbon ≥ 32) were much less affected (Figure 4h). Specifically, the abundances of multiple highly abundant short-chain glycerophospholipids, such as PE(15:0_15:0), PE(15:0_16:0) and PE(14:0_15:0), in Bt were significantly downregulated, whereas those of PE(16:0_18:1), PE(33:1), and PE(16:0_16:1) remained unchanged (Figure 4i).

To validate this observation, we performed in vitro enzymatic activity assays of Btle using PE(14:0_14:0), PE(15:0_15:0), PE(16:0_16:0), and PE(17:0_17:0) as substrates and detected the released fatty acids via GC. Btle showed substantially higher hydrolytic activity toward PE(14:0_14:0) and PE(15:0_15:0), whereas PE(16:0_16:0) and PE(17:0_17:0) were poorly hydrolyzed (Figure 4j and Figure S3d). These in vivo and in vitro data suggest that acyl-chain length may determine Btle substrate specificity, with Btle preferentially hydrolyzing glycerophospholipids with shorter acyl chains.

Btli1 is the functional immunity protein for Btle

Despite Btle being followed by two putative immunity genes, bioinformatic and structural analyzes revealed that Btli1 and Btli2 only share 50% sequence identity and both carry predicted tetratricopeptide repeat (TPR) motifs. AlphaFold3 modeling showed that both adopt similar α-helical architectures (Figure 5a and Figure S4a–b).

Figure 5.

Btli1 neutralizes Btle toxicity, and the engineered Btle^M variant escapes Btli1 to serve as a counterselection marker. (a) AlphaFold3-predicted structures of Btli1 (pink) and Btli2 (teal), with α-helices annotated; the helix absent in Btli2 is highlighted in red. (b) Co-culture assays of btli1 and btli2 mutant and complemented strains. ****, P < 0.0001; ns, non-significant. (c) In vivo His pull-down analysis of interactions between Btli1^His, Btli2^His, and Btle^HA. Whole-cell lysates (WCL) served as input, and co-precipitated proteins were detected by western blot. (d) AlphaBridge diagram of the predicted Btli1-Btle complex; the inner ring shows per-residue pLDDT confidence (blue, very high; cyan, high; yellow, low; orange, very low). (e) His pull-down analysis of ^HisBtle, ^HisBtle^M, and Btli1 interactions. (f) Spotting assay of serially diluted B. fragilis GS075 strains carrying the indicated constructs on BHI agar ± aTc. (g) Schematic of the modified plasmid pGX1 for GS075_2836 knockout using ssBtle^M as the counterselection marker; oriT, RP4 origin of transfer; ErmR, erythromycin resistance gene. (h) PCR verification of GS075_2836 knockout. (i) Validation of counterselection efficiency of Bfe1, Btle and Btle^M for homologous recombination in Bt::btli1 ^HA and Bo::btli1 ^HA. (j) Growth differences between wild-type and gene-deleted strains in minimal medium supplemented with fructose. Data are mean ± SD of three independent experiments; significance assessed by Student’s t-test (****, P < 0.0001).

To examine the potential immune function of Btli1 and Btli2, we performed in vitro co-culture assays. Competitive indices were determined by normalizing the abundance of the recipient co-cultured with GS075::CmR to that of the recipient co-cultured with GS075Δbtle::CmR. The GS075ΔbtleΔbtli1::ErmR recipient, but not the GS075ΔbtleΔbtli2::ErmR recipient, displayed substantially reduced competitiveness against GS075::CmR. The deletion mutant GS075ΔbtleΔbtli1-2::ErmR similarly exhibited severe competitive defects. Complementation with btli1 alone restored immunity, whereas co-complementation with btli1 and btli2 conferred no additional benefit (Figure 5b). Consistently, periplasmic expression in B. fragilis NCTC9343 showed that Btli1 alone or co-expression with Btli1 and Btli2 fully suppressed Btle toxicity, whereas Btli2 alone did not (Figure S4c). In vivo pull-down assays confirmed that only Btli1 interacts with Btle (Figure 5c). Together, these data establish Btli1 as the sole functional immunity protein for Btle.

Since Btli2 resembles Btli1 yet fails to neutralize Btle, we reasoned it might confer immunity against a Btle homolog. Foldseek searches of T6SS gene clusters in NCBI identified Btle homologs from B. ovatus (Btle^Bo) and an unclassified Bacteroides sp. (Btle^Bs) (Figure S4d–f), ⁴⁹ each followed by a single immunity protein. Despite sequence identities of 59–69% among the four immunity proteins (Figure S4g–h), Btli^Bo and Btli^Bs were structurally more similar to Btli1 than to Btli2 (Figure S4i). Pull-down assays revealed that heterologously expressed and purified Btle^Bo partially bound Btli1, whereas no detectable interaction was observed with Btli2 (Figure S4j), arguing against Btli2 being the cognate immunity protein for either homolog.

Sequence and structure alignments demonstrated that Btli2 lacks the α5 and α6 helices present in Btli1 (Figure S4h and Figure 5a). To test whether the missing α5–α6 helices account for Btli2’s non-functionality, we engineered a reconstructed variant, Btli3, by restoring these helices in Btli2. AlphaFold3 predicted that Btli3 closely resembles Btli1 (RMSD = 1.18 Å vs. Btli2-Btli1 RMSD = 3.85 Å) (Figure S4k–l). Yet, periplasmic Btli3 still failed to confer immunity against Btle (Figure S4m). Heterologous expression and purification verified that Btli3 was properly folded (Figure S4n). Pull‑down assays further revealed no obvious interaction between Btle and Btli3, while Btli1 showed stable binding to Btle (Figure S4o), confirming that Btli3 is unable to neutralize Btle. Given the high sequence identity between the upstream regions of btli1 and btli2 (Figure S4p), we propose that btli2 arose by duplication of btli1 and subsequently lost function through accumulated mutations.

Btle^M enables universal Bacteroides genome editing

To dissect the immunity mechanism, we modeled the Btle-Btli1 complex using AlphaFold3 (Figure S5a) and mapped interfacial contacts with AlphaBridge (Figure 5d) ⁵⁰ , revealing that Btli1 engages primarily the C-terminus of Btle. Guided by this interaction interface, we generated a Btle variant: the E468K mutant with residues 482–489 deleted (hereafter Btle^M), which maintains a consistent overall protein folding as wild‑type Btle (Figure S5b). While Btle^M completely abolished Btli1 binding (Figure 5e), periplasmic expression of Btle^M retained strong antimicrobial activity against B. fragilis GS075 (Figure 5f), confirming that Btle^M combines full toxicity with complete Btli1 evasion.

Selection by inducible T6SS effectors is a versatile and modular allelic exchange method for Bacteroides. ^51-53 However, given the prevalence of T6SS effector-immunity pairs and orphan immunity genes across human gut Bacteroides, ⁵⁴ it is essential to confirm that the target strain lacks immunity proteins against the chosen counterselectable marker prior to genetic manipulation. The immunity-evasive property of Btle^M prompted us to exploit it as a universal counterselection marker for homologous recombination. We cloned Btle^M into the pSIE1 backbone ⁵¹ to generate pGX1, a plasmid for universal genome editing in Bacteroides (Figure 5g). As GS075 endogenously encodes Btli1, it served as the primary test strain, with the HTCS-encoding gene GS075_2836, implicated in fructose utilization, selected as the knockout target (Figure S5c). PCR screening with primers Seq-F and Seq-R confirmed deletion by the expected 775 bp product at reasonable efficiency (Figure 5h).

To assess universality, we introduced a Btli1^HA expression plasmid into B. thetaiotaomicron (Bt) and B. ovatus (Bo) (Figure S5d), then targeted the GS075_2836 homologs Bt_1754 and Bo_RS19685 for deletion. During counterselection, wild-type Btle failed in strains expressing Btli1, whereas Btle^M supported robust counterselection efficiency comparable to Bfe1 (Figure 5i), with knockouts confirmed by PCR (Figure S5e–f). All deletion strains showed impaired growth in minimal medium with 0.3% fructose as the sole carbon source (Figure S5g), ^{55 , 56} validating both the knockout and the functional relevance of the targeted gene (Figure 5j).

Multiple V2-encoded hypothetical proteins are required for Btle delivery

Unlike the V2 regions of GA3 T6SS clusters in B. fragilis NCTC9343 (Bf9343) and 638R (Bf638R), each encoding a single effector-immunity pair, the GS075 V2 region contains btle, btli1, and btli2 alongside several additional hypothetical proteins. We performed tblastn searches against the NCBI whole-genome shotgun contigs database using amino acid sequences of the eight ORFs encoded within V2 region as queries. This analysis retrieved T6SS loci exhibiting partial sequence similarity across the entire gene cluster. In addition to the V2 region sharing the identical genetic organization as GS075, we identified two other GS075-type T6SS V2 regions, from strains DCMOUH0017B (100% coverage, 94.02% identity to the GS075 T6SS V2 region) and BFG-89 (79% coverage, 47.92% identity), although BFG-89 carries a distinct putative effector at the equivalent locus (Figure 6a). Together, these findings establish the GS075-type GA3 T6SS as a widespread cluster variant.

Figure 6.

Functional characterization of hypothetical proteins in the GS075 GA3 T6SS V2 region. (a) Gene cluster alignment of GS075-type GA3 T6SS V2 regions. (b) Western blot detection of GS075_1252^strep in culture supernatants and whole-cell lysates of B. fragilis GS075-btle^HA::1252^strep and B. fragilis GS075-btle^HA-ΔtssC::1252^strep. (c) Structural alignment of GS075_1252 (green) with GA3-type T6SS PAAR structures: PAAR^GS086 (pink), PAAR^Bf9343 (light blue), and PAAR^638R (tan). (d) Phylogenetic tree of PAAR proteins, including GS075_1252 and its homologs. (e) Structural alignment of GS075_1253 (blue) and BtapC (green). (f) Signal peptide prediction and domain architecture of GS075_1256 and GS075_1255. (g) Co-culture competition assays between donor strains (GS075::CmR and GS075ΔtssC::CmR) and recipient GS075Δ1256Δ1255::ErmR. (h) In vivo competitive colonization of antibiotic-treated mice inoculated with donor (GS075::CmR) and recipient (GS075Δ1255Δ1256::ErmR) strains (n = 5). Data are mean ± SEM. (i) Co-culture assay verifying the immune activity of GS075_1255. (j) Secretion assay assessing the effects of deletion and complementation of V2 region genes on Btle^HA and Hcp1 delivery. (k) Proposed model for Btle delivery and antagonistic activity.

Structural and functional analysis of four conserved genes in this region (GS075_1252, _1253, _1255, and _1256) revealed distinct roles. AlphaFold3 modeling showed that the N-terminus of GS075_1252 adopts a canonical PAAR spike fold (Figure S6a), and secretion assays confirmed its T6SS-dependent extracellular release, which was abolished in the ΔtssC mutant (Figure 6b). Structural alignment of GS075_1252 (PAAR^GS075) against PAAR^GS086 (PDB: 9JG8) yielded an RMSD of 1.427 Å, whereas PAAR^GS075 exhibited low structural similarity to the PAAR from Bf9343 and Bf638R (Figure 6c). Phylogenetic analysis grouped PAAR proteins into two clades, one containing Bf638R and Bf9343 and the other containing GS075, GS086, BFG-89, and DCMOUH0017B (Figure 6d). GS075_1253 was predicted to form a β-sheet-rich barrel closely resembling BtapC from GS086 (Figure 6e and Figure S6b), an adapter that mediates co-secretion of T6SS effectors, ²⁹ suggesting shared assembly mechanisms between the GS075- and GS086-type clusters.

We next investigated whether GS075_1256 and GS075_1255 constitute an additional effector-immunity pair. GS075_1256 lacks a signal peptide and harbors four predicted structural domains (Figure 6f and Figure S6c), while GS075_1255 carries a lipoprotein signal peptide (Figure 6f), a hallmark of Bacteroides T6SS immunity proteins. In vitro co-culture assays confirmed that GS075_1256 and GS075_1255 confer a fitness advantage to GS075 (Figure 6g). In antibiotic-treated mice, wild-type (GS075::CmR) and mutant strain (GS075Δ1255Δ1256::ErmR) showed comparable colonization capacity under monocolonization (Figure S6d), whereas the GS075::CmR strain exhibited a striking competitive advantage upon co-colonization (Figure 6h). In addition, the GS075Δ1255Δ1256 double-deletion recipient acquired protective immunity upon complementation with GS075_1255 (Figure 6i).

Deletion of GS075_1256 or GS075_1253 impaired T6SS secretion, which was restored by complementation (Figure 6j). This indicates that GS075_1253, like BtapC, facilitates the co-secretion of multi-effectors, whereas GS075_1256 functions both as a candidate effector and as a structural component required for delivery. Although GS075_1254 is predicted to comprise two α/β domains (Figure S6e), genomic analysis showed it is not conserved across GS075-type clusters (Figure 6a) and is dispensable for secretion (Figure 6j).

Collectively, these findings establish that multiple V2-encoded hypothetical proteins are integral to T6SS assembly, effector delivery, and interbacterial competition in GS075 (Figure 6k).

Discussion

This study identifies Btle, a GA3-type T6SS effector from B. fragilis GS075, which likely forms a delivery complex with VgrG and multiple V2-encoded proteins and is translocated into recipient cells upon T6SS contraction. In susceptible targets, Btle cleaves specific membrane glycerophospholipids, releasing fatty acids and disrupting membrane integrity; GS075_1256 may contribute additional bactericidal activity via an unknown mechanism. In resistant strains, periplasmic Btli1 binds Btle and blocks its activity (Figure 6k).

Unlike previously characterized Proteobacterial Tle proteins, Btle is potently active against Bacteroides yet shows limited activity against E. coli, despite the cell membrane being a conserved structure. To assess whether membrane repair or stress response mechanisms may contribute to the reduced susceptibility of E. coli to Btle, we quantified transcript levels of representative stress-responsive and lipid-metabolic genes (osmB, bdm, wza, and lplT/aas) in E. coli by RT-qPCR (Figure S7a). Compared with the catalytically inactive Btle^S129A mutant, wild-type Btle expression caused no significant changes in osmB, bdm and wza, but led to minor upregulation (~2-fold) of lplT/aas. Consistent with this, lipidomic analysis revealed increased LPE levels in Ec-Btle (Figure S7b), suggesting that the upregulated transcription of lplT/aas may represent an adaptive response to LPL accumulation. Moreover, Btle showed low hydrolytic activity against E. coli membrane fractions (Figure 3e–f). These results collectively indicate that the lack of a robust bactericidal phenotype of Btle against E. coli is attributable to limited membrane damage rather than to efficient membrane repair.

Lipidomic profiling revealed substantial divergence in phospholipid composition between B. thetaiotaomicron and E. coli. Bacteroides species are enriched in shorter-chain phospholipids, whereas gut bacteria such as E. coli predominantly harbor longer-chain phospholipids. Although lipidomic analysis detected altered abundances of various glycerophospholipids in the presence of Btle, inducible Btle expression yields higher levels than T6SS-mediated delivery under physiological conditions, potentially altering substrate utilization profiles or disrupting phospholipid metabolic homeostasis. To validate the lipidomic findings, we performed in vitro enzymatic activity assays, which showed that Btle hydrolyzed PE(14:0_14:0), PE(15:0_15:0) more efficiently than PE(16:0_16:0) and PE(17:0_17:0). Together, our in vivo and in vitro data suggest that acyl-chain length may determine Btle substrate specificity, with Btle preferentially hydrolyzing shorter-chain glycerophospholipids. This finding provides a molecular basis for this differential sensitivity and illustrates how compositional variation within an otherwise conserved cellular compartment enables T6SS effectors to discriminate between targets.

Two putative immunity genes, btli1 and btli2, lie downstream of btle, yet only Btli1 neutralizes Btle toxicity. Two immunity genes are not unusual at effector loci, though their functionality varies—both may be active, or one may represent a non-functional pseudogene. Sequence analysis suggests btli2 arose by duplication of btli1 and subsequently accumulated inactivating mutations. Whether Btli2 retains any physiological role, such as immunity against an unidentified effector, remains open, and expanding genome databases may yet provide an answer.

AlphaFold3 modeling of the Btle-Btli1 complex revealed an unexpected interaction mode: rather than occluding the catalytic site or substrate-binding pocket, Btli1 engages the C-terminus of Btle, adding to the growing evidence for mechanistic diversity in immunity protein function. Based on the predicted structure of Btle-Btli1, we further generated the Btle^M mutant, which can evade binding to Btli1 while retaining toxicity. This mutant was subsequently applied as a counterselection marker in genome editing. This strategy avoids the failure of conventional counterselection markers caused by endogenous homologous immunity proteins in target strains, and greatly simplifies the genome editing procedure.

Characterization of additional V2 hypothetical proteins revealed further parallels to the T6SS of strain GS086. GS075 encodes a BtapC homolog (GS075_1253), suggesting a shared co-secretion mechanism. A key distinction, however, is that GS086 encodes a PAAR-effector fusion, whereas in GS075 the PAAR (GS075_1252) and Btle are expressed as independent proteins; in CP103088, PAAR is followed by yet another putative effector. This decoupling of PAAR from effector likely reflects an adaptive evolutionary strategy that has contributed to T6SS effector diversification.

GS075_1256 functions as an additional candidate effector conferring a competitive advantage both in vitro and in vivo, yet its mechanism remains unclear owing to the absence of recognizable enzymatic domains and poor heterologous expression. By analogy to the GS086 multi-effector system, GS075_1256 may perturb membrane integrity or participate in membrane repair. Resolving its activity will further clarify how Bacteroides T6SS clusters combine multiple cargo proteins to achieve efficient interbacterial antagonism.

In summary, Btle broadens our understanding of T6SS-mediated antagonism among Bacteroides and expands the Tle superfamily. Notably, our study identifies that cell membrane lipid composition functions as a key determinant governing the target specificity of Btle, which substantially advances our understanding of the target selectivity underlying interbacterial antagonism mediated by T6SS. ⁵⁷ On this basis, systematic phospholipid profiling across gut microbes will inform the rational engineering of Btle as a precision tool for targeted gut microbiome regulation.

Materials and methods

Bacterial strains and growth conditions

Bacterial strains are listed in Supplementary Table 1. Bacteroides strains used in this study were activated and cultured anaerobically at 37 °C on solid or in liquid Brain Heart Infusion medium (BHI) supplemented with L-cysteine (1 g/L) and hemin (5 mg/L). E. coli strains were cultured at 37 °C in Luria-Bertani (LB) broth with shaking at 220 rpm. Where required, the following antibiotics were added at the indicated concentrations: ampicillin 100 μg/mL, kanamycin 50 μg/mL, gentamicin 200 μg/mL, erythromycin 50 μg/mL, chloramphenicol 20 μg/mL. To assess the growth of HTCS mutant strains, minimal medium (MM) supplemented with 0.3% (w/v) fructose as the sole carbon source was used. ⁵⁵

Plasmid construction

All plasmids used in this study (Supplementary Table 2) were generated through Gibson assembly and verified by Sanger sequencing at Sangon Biotech. For recombinant protein expression in E. coli C43(DE3), the gene encoding wild-type Btle and inactive mutants of Btle were cloned into the pET22b vector, introducing an N-terminal signal peptide pelB for stable expression in the periplasm and a C-terminal 6×His tag. For interaction validation, ^HisBtle, ^HisBtle^Bo, and ^HisBtle^M were generated by cloning the protein-coding sequences into the pET15b vector to introduce an N-terminal 6×His tag. Btli1 lacking its signal peptide was cloned into pET26b to fuse with an N-terminal pelB signal peptide (no affinity tag), whereas pET22b was used for the expression of His-tagged Btli1. The genes encoding Btli1^strep, Btli2^strep, and Btli3^strep were cloned into pET15b-His-SUMO vectors.

For inducible expression by anhydrotetracycline (aTc) in Bacteroides species, the target genes were cloned into the pNBU2-P1T_DP ^GH023 vector. ⁵³ The N-terminal signal peptide (ss), derived from BT_4676 with the amino acid sequence MKKILSLLVMAIVAIQFSF, was used to achieve stable expression in the periplasm of Bacteroides species. For co-expression of two or more genes (e.g., btle and btli1-2), the genomic fragment spanning btle to the immunity protein genes was PCR-amplified and inserted into the vector via Gibson assembly. Both genes were controlled by the same inducible promoters, with expression triggered by aTc treatment. For constitutive overexpression, these genes were cloned into pNBU2-P1E6 vector. ⁵⁸

For genetic knockouts or epitope insertion (e.g., HA-tag) at the native locus in Bacteroides, ~1000 bp regions flanking the gene or epitope insertion site were cloned into the pSIE-ssBfe1-ErmR vector, which utilizes Bfe1 as a counterselection marker. To verify the feasibility of Btle^M as a counterselection marker, we constructed pGX1-ssBtle^M-ErmR, and designed upstream and downstream homologous arms based on the target gene.

Deletion mutant and complementation

Knockout or complementation plasmids were transferred into specific Bacteroides strains via conjugation with E. coli S17-1 λpir. Overnight cultures of donor strain S17-1 λpir were diluted 1:250 in fresh LB medium supplemented with 100 μg/mL ampicillin, while recipient Bacteroides strains were diluted 1:100 in fresh BHI broth. All cultures were incubated at 37 °C. When donor OD₆₀₀ (optical density at 600 nm) reached ~0.3 and recipient OD₆₀₀ reached ~0.2, cells were mixed at a 1:10 donor-to-recipient ratio. The cell mixture was pelleted by centrifugation at 9,000× g for 10 min, resuspended in 100 μL LB broth, and spotted onto 6-cm non-selective BHI agar plates. After 20–24 h of aerobic incubation at 37 °C, cells were scraped and resuspended in 1 mL non-selective LB broth. Suspensions (undiluted or 10⁻¹ diluted) were plated on BHI agar containing 200 μg/mL gentamicin plus either erythromycin or chloramphenicol.

Transconjugants were PCR-screened to confirm plasmid integration. For knockout or epitope insertion verification, PCR-confirmed integrants were cultured overnight in 1 mL BHI broth, and 100 μL of 10-fold or 100-fold diluted suspensions were plated on BHI agar with 200 ng/mL aTc. Plates were incubated anaerobically at 37 °C in the dark for 24–36 h. Single colonies were streaked in parallel onto BHI agar and BHI agar with gentamicin/erythromycin or gentamicin/chloramphenicol as applicable. Colonies growing only on BHI (antibiotic-sensitive) were selected for PCR and DNA sequencing to confirm gene deletion or epitope insertion.

Secretion assay and western immunoblot analysis

For whole-cell lysate (WCL) preparation, the tested Bacteroides strains were cultured anaerobically in 10 mL fresh BHI broth to an OD₆₀₀ of 0.3. Cells were harvested by centrifugation, and the pellet was resuspended in 100 μL PBS mixed with an equal volume of 2× SDS loading buffer. Meanwhile, for supernatant samples, the culture supernatant was sterilized via filtration through a 0.22 μm membrane filter, concentrated to approximately 100 μL, and supplemented with 100 μL 2× SDS loading buffer. For western blot analysis, protein samples were separated by 15% glycine SDS-PAGE and electrotransferred onto PVDF membranes (0.45 μm pore size, Millipore). After blocking, membranes were incubated with primary antibodies (Mouse α-HA, MBL #M180-3, 1:5000; Mouse α-His, MBL #M136-3, 1:5000; Rabbit α-DnaK^Ec, Cusabio #CSB-PA633459HA01EGW, 1:5000; Rabbit α-Bte1, α-Hcp1 and α-DnaK^Bf, this study, 1:5000), followed by incubation with horseradish peroxidase HRP-conjugated secondary antibodies (goat anti-rabbit, MBL #458, 1:5000; goat anti-mouse, MBL #330, 1:5000). Immunoreactive protein bands were developed using M5 HiPer ECL Western HRP Substrate (Mei5 Biotechnology, #MF074-05), and chemiluminescence signals were documented with a Tanon-5200 chemiluminescence imaging system (Tanon).

Co-culture

To enable post-co-culture isolation, the erythromycin-resistance marker ErmR or chloramphenicol-resistance marker CmR was integrated into the genomes of attacker or recipient strains via conjugative transfer of empty pNBU2 vector harboring the corresponding resistance cassette. Overnight cultures of donor and recipient strains were diluted 1:100 in 10 mL fresh BHI broth. When cultures reached OD₆₀₀ ≈ 0.3, cells were mixed at a 10:1 donor-to-recipient ratio. The mixed culture was pelleted by centrifugation at 7,000× g for 10 min and resuspended in 100 μL BHI broth. Then, 10 μL mixture was spotted onto a 6-cm BHI agar plate. After 24 h incubation, cells were harvested, resuspended in 1 mL BHI broth, and serially diluted 10-fold. Dilutions were spotted onto erythromycin-containing BHI agar plates to count colony forming units (CFUs). To calculate the competitive index, the abundance of the recipient strain co-cultured with the experimental strain was normalized to the abundance of the recipient strain co-cultured with the control strain.

Animal experiments

All animal experiments were performed under the supervision and approval of the Animal Research Ethics Review Form (SYDWLL-2023-058) of the School of Life Sciences, Shandong University.

6–8-week-old female C57BL/6J specific pathogen-free (SPF) mice were purchased from GemPharmatech Co., Ltd (Jiangsu, China). These mice were randomly allocated and housed in groups under controlled SPF conditions, with the temperature maintained at 24 ± 2 °C, relative humidity at 45% ± 5%, and a 12-hour light–dark cycle. Cage positions were rotated regularly and sample measurement order was randomized. Prior to the experiments described below, all mice were housed for a one-week acclimatization period. Then the mice were treated with an antibiotic cocktail following a previously established protocol. ²⁹ Specifically, 100 μL of the antibiotic mixture (containing 10 mg/mL metronidazole, 100 mg/mL neomycin, and 100 mg/mL ampicillin) and 100 μL of vancomycin (100 mg/mL) were administered via oral gavage to each mouse once per day for 5 d. Subsequently, all antibiotic-treated mice were supplied with autoclaved water and supplemented with the antibiotic cocktail (0.5 g/L vancomycin, 1.0 g/L metronidazole, 1.0 g/L neomycin, and 1.0 g/L ampicillin) for 5 d. After this period, all antibiotic-treated mice were switched to autoclaved water for an additional 3 d. Before gavage, fecal samples were plated onto selective agar plates to confirm clearance of endogenous antibiotic-resistant gut bacteria. All mice received an equal volume (100 μL) of bacterial suspension via a single oral gavage on Day 0. For monocolonization, the gavage suspension contained ~5 × 10⁹ viable bacterial cells. For co-colonization experiments, mice were gavaged with a mixture containing ~5 × 10⁹ viable cells of attacker and an equal number of recipients.

Fresh fecal samples were collected at predetermined time points to monitor bacterial colonization in individual mice. The collected samples were weighed and vortexed in 1 mL of phosphate-buffered saline (PBS). The diluted fecal samples were then plated individually on selective BHI agar plates containing 200 μg/mL gentamicin and 20 μg/mL chloramphenicol or 50 μg/mL erythromycin. Colony morphology was monitored, and representative single colonies were picked for colony PCR to validate strain identity and exclude contaminating indigenous gut microbes.

Agar spot assay

Overnight cultures of Bacteroides were adjusted to a uniform OD₆₀₀, serially diluted 10-fold, and then 3 μL samples of each dilution were spotted onto BHI agar plates with or without 50 ng/mL aTc. Plates were incubated at 37 °C under anaerobic conditions for 12–24 h and subsequently photographed for documentation and analysis.

Protein expression and purification

E. coli C43(DE3) cells harboring the recombinant plasmid for target protein expression were grown in LB broth supplemented with 100 μg/ml ampicillin or 50 μg/ml kanamycin at 37 °C with shaking at 220 rpm. When the culture reached an OD₆₀₀ of ~0.8, protein expression was induced by adding isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 200 μM. Induction proceeded overnight at 20 °C with shaking at 220 rpm. Cells were harvested by centrifugation at 4,500× g for 15 min. The cell pellet was resuspended in 30 ml of lysis buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl) and lysed using a high-pressure cell crusher (Union-Biotech). The lysate was clarified by centrifugation at 17,000× g for 50 min. His-tagged protein was purified from the supernatant using Ni-NTA agarose resin (Qiagen). For the Btli1^strep–Btli3^strep, the 6×His-SUMO tag was removed via ULP1 protease cleavage during protein purification. Further purification was achieved by anion-exchange chromatography (HiTrap Q, GE Healthcare) and size-exclusion chromatography (Superdex 200 Increase 10/300 GL column, GE Healthcare), yielding highly pure protein.

Enzymatic activity assays

The Tween-20 hydrolysis assay was performed as previously described. ⁴⁵ These experiments were performed at 37 °C at a final enzyme concentration of 1.5 μM in a buffer consisting of 20 mM Tris-Cl pH 8.0, 150 mM NaCl, 2% (v/v) Tween-20 and different concentrations of CaCl₂. Lipases hydrolyze the ester bonds of Tween-20, releasing free fatty acids. These liberated fatty acids combine with calcium ions present in the reaction mixture, forming insoluble calcium fatty acid salts. The resultant precipitation increases light scattering, enabling measurement of enzymatic activity by monitoring optical density at 500 nm (OD₅₀₀) over time.

For the validation of PLA/PLB activity of Btle, the substrate PE(15:0_15:0) (500 μM), together with Btle or Btle^S129A mutant protein (final concentration 1.5 μM) and 3 mM CaCl₂ was added to the lysis buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl). The reaction mixture (100 μL) was incubated at 37 °C for 12 h, followed by the addition of 2 μL HCl. After vortexing thoroughly, 100 μL ethyl acetate was added and the mixture was vortexed again. Subsequent centrifugation was performed at 13,000 rpm for 5 min, and 80 μL of the supernatant was collected for gas chromatography (GC) analysis (7890B, Agilent). The fatty acid products were analyzed by methods modified from Jiang et al. ⁵⁹ GC analysis was carried out using an SH-Rtx-5 capillary column (221-75701-30, SHIMADZU, 30 m × 0.25 mm × 0.25 μm). Nitrogen was used as the carrier gas at a constant column flow rate of 2.0 mL/min. The oven program was set initially at 40 °C and held for 5 min, then ramped up to 280 °C at 10 °C/min and maintained at 280 °C for 5 min. The injection port temperature was set at 300 °C, with an injection volume of 1 μL.

GC-MS analysis of membrane fraction products from in vitro Btle digestion

To examine the enzymatic activity of Btle against bacterial membrane fractions, we first harvested bacterial cells from 1 L of Bacteroides culture and 1 L of E. coli culture via centrifugation (4,500× g for 15 min) and resuspended cell pellet in 25 mL lysis buffer (20 mM Tris-HCl, pH 8.0, and 150 mM NaCl). Cells were lysed using a high-pressure homogenizer, and the supernatant was separated from the pellet by centrifugation (17,000× g, 30 min, 4 °C). Membrane fractions were isolated from the supernatant via ultracentrifugation (150,000× g, 60 min, 4 °C) and resuspended in 2 mL lysis buffer. Bacterial input biomass and recovered membrane amounts were maintained consistent for all preparations. Btle protein (final concentration, 1.5 μM) and 2% (v/v) membrane fractions were added to lysis buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl). The reaction mixture (100 μL) was incubated at 37 °C for 12 h, followed by the addition of 2 μL HCl. After thorough vortexing, 100 μL ethyl acetate was added, and the mixture was vortexed again. Subsequent centrifugation was performed at 13,000 rpm for 5 min, and 80 μL of the supernatant was collected for gas chromatography-mass spectrometry (GC-MS) analysis.

The GC2030-GCMS-QP2020 NX system equipped with a DB-5ms column (30 m, 0.25 mm, 0.25 μm; Agilent Technologies, Inc., USA) was utilized for GC-MS analysis. The electron ionization (EI) source was employed, with the source set at 280 °C and interface temperature set at 280 °C. The full scan mode was utilized for data acquisition with the scan range of 25–350 m/z. The carrier gas flow rate of helium was 1.0 mL/min, and the injection temperature was 300 °C with splitless using liquid injection. The temperature program for the GC analysis was as follows: started at 40 °C for 5 min, ramped to 280 °C at 10 °C/min and held for 10 min, resulting in a total runtime of 34 min. Data were analyzed by GCMSsolution 4.45 (Shimadzu, Japan) together with searching against the NIST23 database.

Protein interaction analysis experiment

For in vivo interaction verification: GS075 strains expressing Btle^HA and Btli1^His/Btli2^His in the periplasm were collected by centrifugation and resuspended in lysis buffer containing 20 mM Tris-HCl, pH 8.0 and 150 mM NaCl. Bacterial cells were lysed using a high-pressure cell disruptor (Union-Biotech). The lysate was clarified by centrifugation, and the supernatant was loaded onto a Ni-NTA agarose resin (Qiagen). Following washing with lysis buffer plus 20 mM imidazole to remove non-specifically bound proteins, the target proteins were finally eluted with elution buffer (lysis buffer with 250 mM imidazole). Elution fractions were subjected to SDS-PAGE and analyzed by western blot.

Strains expressing His-tagged Btle^M (or Btle) and untagged Btli1 were combined before cell lysis via a high-pressure cell disruptor. The proteins were further purified, and the eluted fractions were examined for complex formation.

Sequence alignment and phylogenetic analysis

Protein sequences representing Tle1–Tle6 families and Btle homologs were collected. ^{15 , 32} Each family was aligned using the MUSCLE algorithm and phylogenetic trees were generated using the Molecular Evolutionary Genetic Analysis (MEGA) tool version 11 to construct maximum-likelihood tree with bootstrap analysis of 1000 replicates. ⁶⁰ Phylogenetic topologies were visualized and annotated in iTOL v6.7.2 (https://itol.embl.de/) ⁶¹ . The method for constructing the phylogenetic tree of PAAR homologous proteins is as described above.

Fluorescence microscopy

Overnight cultures of strains were subcultured at a 1:100 ratio into fresh BHI medium (for Bacteroides strains) or LB medium with ampicillin (for E. coli strains). Inducer was supplemented at the time of subcultivation (0 h) and 1 h post-subcultivation (1 h), followed by 4 h of induction for Btle expression, bacterial cells were harvested by low-speed centrifugation, washed twice with sterile PBS, and resuspended in PBS containing 5 μg/mL FM 1-43 and 20 μg/mL propidium iodide (PI). Following 30 min of dark incubation at room temperature, samples were examined using a Zeiss LSM 900 laser scanning confocal microscope.

Subcellular fractionation

To perform subcellular fractionation, bacterial culture and induction were performed as described for lipidomic sample preparation. Then, bacterial cells were washed with PBS, and resuspended in 285 μL sucrose buffer (20 mM PBS, 20% sucrose, 2.5 mM EDTA) for 20 min at room temperature. After that, 285 μL ice-cold 0.5 mM MgCl₂ was added and incubated for 5 min on ice. The suspension was centrifuged at 7000× g for 20 min at 4 °C to collect the supernatant containing periplasmic proteins (Peri). The pellet was resuspended in PBS buffer and defined as cytoplasmic (Cyto). ⁶² All samples were resuspended in equal volumes and analyzed by SDS-PAGE and western blotting.

Lipidomic analysis

Overnight cultures of Bt5482::pNBU2-P1T_DP ^GH023 (Bt), Bt5482::pNBU2-P1T_DP ^GH023 -ssbtle (Bt-Btle), E. coli C43(DE3)/pET22b (Ec) and E. coli/pET22b-pelB-btle (Ec-Btle) were subcultured at a 1:100 ratio into fresh BHI medium (for Bacteroides strains) or LB medium with ampicillin (for E. coli strains) and grown to OD₆₀₀ ≈ 0.17. Subsequently, aTc or IPTG was supplemented to final concentrations of 50 ng/mL and 200 μM. After 4 hours, cells were harvested, rapidly flash-frozen in liquid nitrogen, and stored for subsequent lipidomic analysis.

Lipid extraction and profiling were performed by BGI Group (Shenzhen, China) using a UPLC-MS/MS platform. Weigh 25 mg of cells, place them in a 2 mL thickened centrifuge tube. Add two small steel beads, 800 μL of pre-cooled dichloromethane/methanol (3:1, v/v), and 10 μL of configured internal standard 2 for each sample. Grind in a Tissuelyser for 5 minutes, sonicate for 10 minutes in ice bath, and let stand in a refrigerator overnight at −20 °C. Centrifuge at 25,000× g, 4 °C for 15 minutes, take 600 μL of supernatant, and drain into a freezer drainer. Redissolve with 120 μL of lipid resolving solution (isopropanol: acetonitrile: water = 2:1:1) and shake for 10 min. Sonicate in ice bath for 10 min. Centrifuge at 25,000× g, 4 °C for 15 min. Take 20 μL of the mixed QC again for each sample.

Chromatographic conditions: The column used was a CSH C18 chromatography column (1.7 μm, 2.1 × 100 mm, Waters, USA). The mobile phase for positive ion mode mobile phase contains 60% acetonitrile water + 10 mM ammonium formate + 0.1% formic acid (Liquid A) and 90% isopropanol + 10% acetonitrile + 10 mM ammonium formate + 0.1% formic acid (Liquid B). The negative ion mode mobile phase contains 60% acetonitrile water + 10 mM ammonium formate (Liquid A) and 90% isopropanol + 10% acetonitrile + 10 mM ammonium formate (Liquid B). The following gradients are used for elution: 0 ~ 2 min, 40% ~ 43% liquid B; 2 ~ 2.1 min, 43% ~ 50% liquid B; 2.1 ~ 7 min, 50% ~ 54% liquid B; 7 ~ 7.1 min, 54% ~ 70% liquid B; 7.1 ~ 13 min, 70% ~ 99% liquid B. 13 ~ 13.1 min, 99% ~ 40% B liquid. 13.1 ~ 15 min, 40% B liquid. The flow rate was 0.4 mL/min, the column temperature was 55 °C, and the injection volume was 5 μL.

Mass spectrometry conditions: First- and second-level mass spectrometry data acquisition was performed using a Q Exactive mass spectrometer (Thermo Fisher Scientific, USA). The mass spectrometry scan covered an m/z range of 200–2000. The MS1 resolution was set to 70,000, the automatic gain control (AGC) target was 3 × 10⁶, and the maximum injection time (IT) was 100 ms. According to the strength of the parent ion, Top3 is selected for fragmentation, and the secondary information is collected. For MS2 acquisition, the MS2 resolution was set to 17,500, the AGC target was 1 × 10⁵, the maximum injection time was 50 ms, and the stepped normalized collision energy (NCE) was set to 15, 30, and 45 eV. Electrospray ionization (ESI) ion source parameters were set as follows: sheath gas flow at 40 arb, auxiliary gas flow at 10 arb; spray voltage was 3.80 kV in positive ion mode and 3.20 kV in negative ion mode; the ion transfer tube temperature was 320 °C, and the auxiliary gas heater temperature was 350 °C.

Circular dichroism (CD) spectroscopy

Far-UV CD spectra of ^HisBtle and ^HisBtle^M were recorded using a Circular Dichroism Spectrometer (JASCO J-1500, JASCO, Tokyo, Japan). Protein was diluted to 0.2 mg/mL, and a 200 μL aliquot was transferred to a 1 mm path-length quartz cuvette. Spectra were acquired at 25 °C over a wavelength range of 250–190 nm at a data pitch of 0.1 nm, with a bandwidth of 1 nm and a scan speed of 200 nm/min. Five replicate scans (5 cycles) were averaged to obtain the final spectrum.

Analysis of GS075-like T6SS V2 region

An in-house nucleotide database of 2611 Bacteroides fragilis genome sequences was constructed (dataset updated to September 2026). After removing redundant strains, 2157 non-redundant genomes remained. The tblastn screening identified 388 strains that simultaneously carry homologs of VgrG^GS075, GS075_1253, and Btle, representing a prevalence of approximately 18.0%. These clusters encode proteins with high similarity to VgrG^GS075, GS075_1256/1255/1254/1253/1252 and Btle, while a minority contain only one immunity protein.

RNA extraction, cDNA synthesis, and RT-qPCR

Total RNA was extracted from bacteria using the RNAsimple Total RNA Kit (TIANGEN Biotech) according to the manufacturer’s protocol. Reverse transcription was performed using the PrimeScript™ RT reagent Kit (Perfect Real Time) (TAKARA). Quantitative RT-PCR was carried out using TB Green® Premix Ex Taq™ (Tli RNaseH Plus) (TAKARA). Relative changes were calculated with the $2^{-\text{ΔΔ}{\text{C}}_{\text{t}}}$ method, normalized to E. coli 16S rRNA transcript levels.

Quantification and statistical analysis

Data are presented as the arithmetic mean ± standard deviation (SD) unless otherwise indicated. All statistical analyses were performed using GraphPad Prism v8. Unpaired two-tailed Student’s t-tests were used for comparing means between two groups, unless otherwise indicated. The significance of mean comparisons was denoted as follows: NS, non-significant; ^∗, P < 0.05; ^∗∗, P < 0.01; ^∗∗∗, P < 0.001; ^∗∗∗∗, P < 0.0001. A significance level of P < 0.05 was considered statistically significant.

Supplementary Material

Supplementary Table 1_List of bacterial strains used in this study.docx

Supplementary Table 1_List of bacterial strains used in this study.docx

Supplementary Table 3_Most prominent bacterial lipid peaks detected with the UPLC MS method.xlsx

Supplementary Table 3_Most prominent bacterial lipid peaks detected with the UPLC MS method.xlsx

Supplementary Table 2_List of plasmids used in this study.docx

Supplementary Table 2_List of plasmids used in this study.docx

Supplementary Material

Supplementary_Data_S1_Sequences of all genes and proteins.docx

Supplementary Material

Attachments.zip

Funding Statement

The present study was supported by the Shandong Provincial Natural Science Foundation (Grant nos. ZR2024ZD47 and ZR2024YQ047), the Taishan Young Scholars Program (no. tsqn202408046), the SKLMT Frontiers and Challenges Project (Grant no. SKLMTFCP-2023-01), the Intramural Joint Program Fund of SKLMT (Grant no. SKLMTIJP-2025-01), and the Young Talent Development Program of SKLMT (Grant no. M2025YA01).

Disclosure statement

No potential conflicts of interest were disclosed.

Data availability statement

The data reported in this paper have been deposited in the OMIX, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (https://ngdc.cncb.ac.cn/omix: accession no. OMIX018322). Raw Illumina sequencing reads of strain GS075 have been deposited in the NCBI Sequence Read Archive (SRA) under accession SRR40408787, associated with BioProject PRJNA1507262 and BioSample SAMN62158796. All materials created in this study are available from the corresponding author. Constructs, strains, plasmids, and mutants generated in this study are available from the corresponding author under a Materials Transfer Agreement with the Shandong University.

Supplementary material

Supplemental data for this article can be accessed at https://doi.org/10.1080/19490976.2026.2750098.