Abstract

   Bacterial pathogens have evolved multiple mechanisms to modulate host
   cell death, evade host immunity, and establish persistent infection.
   Here, we show that an infective endocarditis causative pathogen,
   Bergeyella cardium, is frequently detected in oral specimens from
   clinical patients. A variant strain of Bergeyella cardium (BCV) induces
   unique cytoplasmic vacuolization cell death and minor apoptosis-like
   cell death in macrophages. The cytoplasmic vacuolization cell death
   triggered by BCV is characterized by Fused LysosOme-Associated
   Termination (floatptosis) and is inhibited by the sodium channel
   inhibitor amiloride. Moreover, outer membrane vesicles (OMVs) or
   transfection of barrel-like membrane proteins, lipocalin, β-barrel, and
   PorV, dramatically induce cytoplasmic vacuolization. Endosomal solute
   carrier family 9 member A9 (SLC9A9) plays important roles in the
   process of BCV-, OMVs-, and barrel-like proteins-triggered cytoplasmic
   vacuolization cell death via promoting vacuole fusion. SLC9A9
   deficiency or amiloride administration increases host defense against
   BCV infection. These findings contribute to developing novel approaches
   to modulate cytoplasmic vacuolization cell death and treat infectious
   diseases.

   Subject terms: Cell death, Innate immunity

Introduction

   Host cell death is an intrinsic immune response that plays dual roles
   in host defense against bacterial infection^[52]1. Bacterial pathogens
   have evolved strategies to manipulate host cell death and survival
   pathways, creating an intracellular replication niche and promoting
   dissemination^[53]2. The major programmed cell death pathways,
   including apoptosis, necroptosis, and pyroptosis, have been extensively
   studied, and many virulence factors of bacterial pathogens can act
   through these pathways to increase pathogenicity^[54]3–[55]5.
   Cytoplasmic vacuolization is a morphological phenomenon characterized
   by the dilation and fusion of the endoplasmic reticulum (ER) or
   endosomal–lysosomal organelles, which can be associated with cell death
   or promote cell survival depending on the properties of chemical and
   infectious agents^[56]6. Many bacterial cytotoxins can induce
   cytoplasmic vacuolization of multiple organelles and cell death in
   different types of cells. Pathogenic E. coli subtilase cytotoxin SubAB
   induces vacuolization of the intracellular organelles through
   interaction with plasma membrane α2β1 integrin receptors^[57]7 and
   triggers Bak/Bax conformational changes, cytochrome c release and
   downstream apoptosis^[58]8. Community-acquired respiratory distress
   syndrome toxin (CARDS) from Mycoplasma pneumoniae recognizes membrane
   Annexin A2 protein as a functional receptor. It elicits extensive
   vacuolization in mammalian cells without knowing the nature of vacuoles
   and cell death activity^[59]9,[60]10. Helicobacter pylori’s vacuolating
   toxin A (VacA) is the best-studied vacuolating pore-forming toxin
   (PFT). VacA interacts with multiple host cell surface proteins for
   membrane anion-selective channel formation, internalization, and
   vacuolization of both early and late endosomes with the activation of
   v-ATPase proton pump, leading to osmotic swelling of
   endosomes^[61]11–[62]13. Meanwhile, many studies revealed that VacA not
   only induced the activation of proapoptotic proteins Bax/Bak and
   mitochondrial fragmentation for apoptotic cell death, but also
   triggered programmed necrosis via PARP activation independent of
   vacuolization, indicating vacuolization is not the direct cause for
   cell death^[63]14–[64]16. Similar to VacA, Vibrio cholerae toxin VCC
   forms an anion-selective channel in cells and causes extensive
   vacuolization of late endosomes and of the trans-Golgi network through
   autophagy, but VCC-induced vacuolization is not the cause of cell
   death^[65]17,[66]18. Instead, the vacuolated organelles, in some cases,
   can accumulate bacterial toxins to reduce stress and increase the cell
   survival potential. The relationship between vacuolization and cell
   death remains unclear, despite the extensive data available on
   bacterial toxins that induce cytoplasmic vacuolization and cell death.
   It is still unknown whether the vacuolization phenomenon is a
   programmed cell response^[67]6.

   Three types of nonapoptotic cytoplasmic vacuolization cell death have
   been defined, including paraptosis characterized by extensive
   cytoplasmic vacuolation associated with the dilation of
   ER/mitochondria^[68]19, triaptosis distinguished by early endosome
   vacuolization^[69]20, and methuosis marked by the vacuolization of late
   endosome and lysosome^[70]21. Methuosis is the best-studied type of
   cytoplasmic vacuolization death in tumor cells and was initially
   described as active Ras-triggered cell death^[71]21,[72]22. Recently,
   increasing evidence has indicated that multiple small-molecule
   compounds can induce methuosis and suppress cell growth, such as
   chalcone-related small molecules, MIPP, MOMIPP, epimedokoreanin C, and
   azaindole-based compounds, which have therapeutic potential in cancer
   treatment^[73]23–[74]25. However, the role of cytoplasmic vacuolization
   in the immune cells for the inflammatory response and immune evasion
   has not been investigated.

   We previously isolated an emerging pathogen strain with a flat dry and
   rough morphotype colony, Bergeyella cardium (BC), from a patient with
   infective endocarditis^[75]26. In the present study, we found that
   BC-unique sequences were frequently detected in oral microbiomics from
   clinical patients. We observed that a variant strain of BC (BCV) had a
   smooth morphotype and increased resistance to serum
   complement-dependent clearance. Whole-genome sequencing revealed high
   sequence identity between BC and BCV. Notably, BCV infection triggered
   robust cytoplasmic vacuolization cell death and minor apoptosis-like
   cell death in different cells. The vacuoles induced by BCV were
   colocalized with the late endosomal and lysosomal markers Rab7 and
   LAMP1. OMVs derived from BCV and transfection of lipocalin, β-barrel,
   and PorV of BCV both induced the formation of cytoplasmic vacuolization
   cell death. SLC9A9 was important for the cytoplasmic vacuolization cell
   death induced by BCV through promoting vacuole fusion. The
   intracellular propagation and in vivo pathogenicity of BCV were
   significantly more potent than those of BC, and SLC9A9 deficiency or
   amiloride administration increased host defense against BCV infection.
   Our work offers novel mechanistic insights into cytoplasmic
   vacuolization cell death mechanisms and provides potential therapeutic
   targets for treating infectious diseases.

Results

BCV is prone to high pathogenicity

   To analyze the prevalence of BC in clinical patients, which was
   previously isolated from a patient with infective endocarditis^[76]26,
   we detected BC-unique sequences using the resource data of pan-body
   pan-disease microbiomics^[77]27. Notably, BC was frequently detected in
   plaque (54.9% in 335 specimens), saliva (34.8% in 391 specimens), skin
   arm (9.3% in 183 specimens), and throat (18.7% in 327 specimens) from
   clinical patients (Fig. [78]1a, b; Supplementary Table [79]S1). The
   high proportion of BC-positive specimens in clinical patients,
   particularly oral specimens, underscores the certain noncausal
   infectious diseases potentially caused by BC. To further investigate
   the detailed bacterial characteristics and pathogenesis of BC, we
   spread the bacterial stock on Columbia blood agar plates.
   Interestingly, we found the colonies exhibited two distinct colony
   morphologies (Fig. [80]1c). The BC colonies were flat, dry, and rough.
   In contrast, the BCV colonies were raised and smooth (Fig. [81]1c). We
   performed whole-genome sequencing of BC and BCV colonies, and the
   circular genome lengths of BC and BCV were 2,036,890 bp and
   2,036,968 bp, respectively (Fig. [82]1d; GenBank [83]CP029149 and
   [84]CP114055). The genomic homology between BC and BCV was
   approximately 99.99%. We compared the BC and BCV genomes and confirmed
   that the BCV genome harbored three insertions located in the noncoding
   region, the PorV gene, and a hypothetical gene of the BC genome
   (Supplementary Fig. [85]S1a–f). The 24-bp insertion in the PorV gene
   caused an 8-amino-acid insertion in the PorV protein of BCV, resulting
   in a PorV protein longer than that of BC (Supplementary Fig. [86]S1c,
   d). The 13-bp insertion in the hypothetical gene of BC caused a
   frameshift mutation, which produced a new BatD family protein in BCV
   (Supplementary Fig. [87]S1e, f). PorV is a shuttle protein in the outer
   membrane of the type IX secretion system (T9SS) that mediates the
   secretion of major virulence factors^[88]28. To investigate whether the
   insertion of 8 amino acids in PorV of BCV (residues 344–351) led to its
   structural difference from BC, we performed structure prediction
   analysis using trRosseta^[89]29. According to the predicted structure,
   BCV PorV folded into a 14-strand β-barrel, and residues 344–351 folded
   into the β15 of the β-barrel (Supplementary Fig. [90]S1g). BC PorV was
   also predicted to be a 14-strand β-barrel (Supplementary Fig. [91]S1h).
   However, the predicted structure of BC PorV was less reliable in the
   β14–β15 region, as shown by the lower LDDT in this region
   (Supplementary Fig. [92]S1g, h). Therefore, compared with BCV PorV,
   deletion of 8 amino acids in BC PorV caused structural changes in PorV,
   which might affect the secretion of virulence factors through
   interactions with binding partners.

Fig. 1. Characterization of BC and BCV.

   [93]Fig. 1
   [94]Open in a new tab

   a The prevalence of BC in clinical patients. BC-positive is defined as
   BC-unique reads (150 nt in length) ≥ 3. The numbers below each bar
   indicate the corresponding numbers of patients. b Distribution of
   BC-unique reads within BC-positive specimens. The numbers below each
   bar indicate the corresponding numbers of BC-positive patients. c
   Bacterial colony morphology of BC and BCV. BC and BCV were cultured on
   Columbia blood agar plates for 96 h. Scale bars, 3 mm. d Circular
   genome map of BCV with its alignment with BC. The circular genome map
   includes the following, from outer to inner rings: BC CDS+, coding
   sequences (CDS) on the forward strand of the BC genome; BCV CDS+, CDS
   on the forward strand of the BCV genome; BC CDS–, CDS on the reverse
   strand of the BC genome; BCV CDS–, CDS on the reverse strand of the BCV
   genome; GC content, which helps identify genomic islands and horizontal
   gene transfer events; GC skew, indicating the over- or under-abundance
   of G or C between the leading and lagging DNA strands, which is often
   used to identify the origin and terminus of replication; and the
   locations of tRNAs and rRNAs in the genome. The BC genome is
   2,036,890 bp long, and the BCV genome is 2,036,968 bp long. e SEM and
   TEM analysis of BC and BCV. The arrows indicate puff-up and blebbing of
   BCV. Scale bars, 2 μm for SEM and 500 nm for TEM. f TEM analysis of BC
   and BCV bacteria treated with human sera for 2 h. Sera-treated BC and
   BCV were precipitated, fixed, and sectioned for TEM analysis. Scale
   bars, 5 μm for upper and 500 nm for lower. Data are representative of 3
   independent experiments with similar results (c, e, f).

   We performed electron microscopy analysis to characterize the
   phenotypes of BC and BCV in detail. Scanning electron microscopy (SEM)
   and transmission electron microscopy (TEM) revealed that BCV bacteria
   had multiple membrane puff-ups and blebbing (Fig. [95]1e). Serum
   resistance of bacteria is essential for their persistence in the
   bloodstream to cause infection and for the development of antibiotic
   tolerance, which is recognized as a stepping stone toward antibiotic
   resistance^[96]30–[97]32. Notably, the capacity for serum resistance in
   BCV was much greater than in BC (Fig. [98]1f; Supplementary Fig.
   [99]S2a), whereas antibiotic resistance was comparable between BC and
   BCV (Supplementary Fig. [100]S2b). To define the genes differentially
   expressed between BC and BCV, we performed RNA sequencing (RNA-seq)
   analysis of BC and BCV cultured on Columbia blood agar plates for 96 h.
   Overall, 89 and 487 genes were upregulated and downregulated,
   respectively, in BCV (Supplementary Fig. [101]S2c and Table [102]S2).
   Notably, the genes involved in bacterial invasion and colonization,
   stress adaptation, and secretion systems, including types I and IX,
   were highly expressed in BCV (Supplementary Fig. [103]S2d, e).
   Conversely, the genes that encode proteins involved in immunogenicity
   and pathogenicity, such as lipopolysaccharide (LPS) and lipoteichoic
   acid (LTA) biosynthesis, were dramatically downregulated in BCV
   (Supplementary Fig. [104]S2f, g). These results indicate that BC is
   highly prevalent in clinical patients, and BCV might be a more virulent
   BC variant strain.

BCV induces lysosomal fusion-mediated cytoplasmic vacuolization cell death

   To investigate the host cell death response triggered by BC and BCV
   infection, we treated bone marrow-derived macrophages (BMDMs) with BC
   or BCV for different durations. Notably, BCV infection triggered
   cytoplasmic vacuolization beginning around 6 h post-infection,
   significant cytoplasmic vacuolization at 10 h post-infection, and large
   vacuoles and single vacuole-occupied cells after 20 h post-infection
   (Fig. [105]2a; Supplementary Fig. [106]S3a). However, this phenomenon
   was not observed in BC-infected BMDMs (Fig. [107]2a). In addition to
   BMDMs, BCV infection also triggered cytoplasmic vacuolation in L929
   cells, immortalized BMDMs (iBMDMs), and the human cardiomyocyte cell
   line AC16 (Supplementary Fig. [108]S3b–d). To closely examine
   vacuolization, we performed TEM analysis of BC- and BCV-infected BMDMs.
   Notably, BCV infection caused the formation of multiple single
   membrane-bound vacuoles at 10 h post-infection (Fig. [109]2b). The size
   of the vacuoles was enlarged, and the number was reduced due to the
   fusion of vacuoles during the later phase of infection (Fig. [110]2b;
   Supplementary Fig. [111]S3e). The process of vacuole fusion and
   enlargement was also confirmed by live-cell imaging analysis, and
   plasma membrane permeabilization indicated by Propidium iodide (PI)
   staining occurred at a later time of the vacuolization process (Fig.
   [112]2c; Supplementary Video [113]S1). The vacuoles triggered by BCV
   were labeled with LAMP1 and Rab7 but not the early endosome markers
   Rab5 or EEA1 (Fig. [114]2d; Supplementary Fig. [115]S3f–h), which
   indicated that the vacuoles were derived from lysosomes and late
   endosomes. In addition, the density of BCV infection-induced
   vacuole-containing cells was lower than that of normal BMDMs, and
   vacuolated cells were separated at the middle layer of the 40% Percoll
   gradient (Supplementary Fig. [116]S3i). Cell membrane disruption was
   observed in BC-infected BMDMs at later time points (Fig. [117]2b).
   Remarkably, the cell death caused by BCV infection was much more severe
   than that caused by BC infection, as determined by real-time
   quantification of cell death and LDH release assay (Fig. [118]2e–g).

Fig. 2. BCV infection triggers cytoplasmic vacuolization cell death.

   [119]Fig. 2
   [120]Open in a new tab

   a Microscopic analysis of WT BMDMs infected with BC or BCV (400 MOI)
   for the indicated time. Scale bars, 30 μm. b TEM analysis of WT BMDMs
   infected with BC or BCV (400 MOI) for the indicated time. N indicates
   the cell nucleus; arrows indicate the intracellular BCV and enlarged
   images are shown. Scale bars, 1.2 μm. c Live-cell imaging of WT BMDMs
   infected with BCV (200 MOI) for the indicated time. Scale bar, 10 μm. d
   Confocal microscopy analysis of Rab7 and LAMP1 in BC- and BCV-infected
   (400 MOI) WT BMDMs for 20 h. Scale bars, 20 μm. e Real-time
   quantitative live-cell imaging and analysis of cell death in uninfected
   WT BMDMs and WT BMDMs infected with BC or BCV (400 MOI) (n = 16 random
   fields; 3 independent experiments). f Representative images of PI
   staining of uninfected BMDMs and BMDMs infected with BC or BCV (400
   MOI) for the indicated time, as shown in (e). Scale bars, 20 μm. g LDH
   analysis of WT BMDMs infected with BC or BCV (400 MOI) for the
   indicated time (n = 3 biologically independent samples). Data are from
   3 independent experiments (g) or representative of 3 independent
   experiments with similar results (a–f). For (g), data represent
   Mean ± SEM, and two-sided Student’s t-test without multiple-comparisons
   correction was used. For (e), two-way ANOVA was used. **P < 0.01,
   ***P < 0.001, ****P < 0.0001.

BCV-triggered cytoplasmic vacuolization cell death is distinct from methuosis

   Methuosis is recognized as nonapoptotic cell death associated with
   cytoplasmic vacuolization and is characterized by Rab7- and
   LAMP1-labeled late endosomes and lysosomes in tumor cells^[121]22. To
   define whether BCV-triggered cytoplasmic vacuolization cell death in
   BMDMs is similar to methuosis, we treated BMDMs with MOMIPP, which can
   cause methuosis in multiple tumor cells (Supplementary Fig.
   [122]S4a)^[123]22. Unexpectedly, the cytoplasmic vacuolization induced
   by MOMIPP in BMDMs but not lung cancer cells A549 was transient and
   reversible. At 6 h after treatment, MOMIPP-treated BMDMs exhibited
   remarkable cytoplasmic vacuolization (Supplementary Fig. [124]S4b–d).
   However, the cytoplasmic vacuolization in BMDMs induced by MOMIPP
   decreased after 6 h of treatment and almost disappeared after 24 h of
   continuous treatment (Supplementary Fig. [125]S4b–d). The fluid-phase
   tracer Lucifer yellow accumulated in vacuoles to form a large punctum
   in BMDMs during MOMIPP treatment, whereas the distribution of Lucifer
   yellow in BCV-infected BMDMs was restricted to the edge of vacuoles and
   failed to form big puncta inside the vacuoles (Supplementary Fig.
   [126]S4e). In line with a previous report, the vacuoles induced by
   MOMIPP in BMDMs were also labeled with Rab7 and LAMP1 but not EEA1
   (Supplementary Fig. [127]S4f). Real-time quantification of cell death
   and LDH release assay revealed that MOMIPP treatment in BMDMs did not
   cause dramatic cell death (Supplementary Fig. [128]S4g–i).
   Collectively, BCV-triggered cytoplasmic vacuolization in BMDMs shares
   common features with methuosis, but is different from the methuosis
   pathway.

BCV-triggered cytoplasmic vacuolization is a unique cell death pathway

   Type I IFN (IFN-I) signaling and inflammasome activation play essential
   roles in many bacterial pathogen infection-triggered cell death
   pathways, such as pyroptosis and PANoptosis^[129]4,[130]33,[131]34. To
   determine whether IFN-I signaling and ASC-dependent inflammasome
   activation were involved in the BCV-triggered cytoplasmic vacuolization
   pathway, WT, Ifnar^−^/^−, Asc^−^/^−, and Aim2^−^/^−Nlrp3^−^/^− BMDMs
   were infected with BC or BCV (Supplementary Fig. [132]S5a). The results
   showed that the levels of cytoplasmic vacuolization and LDH release
   induced by BCV infection were comparable among these samples
   (Supplementary Fig. [133]S5a, b). These data indicated that
   BCV-triggered cytoplasmic vacuolization cell death was not dependent on
   IFN-I signaling or ASC-dependent inflammasome activation.

   To further examine whether this type of cytoplasmic vacuolization was
   associated with apoptosis or other types of cell death, we analyzed the
   hallmarks of different kinds of cell death via protein expression and
   activation analyses of caspase-3 for apoptosis, caspase-1 for
   pyroptosis, and phosphorylation of MLKL and RIP3 for necroptosis in BC-
   and BCV-infected BMDMs. Indeed, BCV infection also triggered caspase-3
   activation at later time points but not caspase-1 activation (Fig.
   [134]3a; Supplementary Fig. [135]S5c). Notably, TEM analysis of
   BCV-infected BMDMs at later time points revealed that BCV infection
   activated two types of cell death in different cells: major cytoplasmic
   vacuolization (78.65%) and minor apoptosis-like cell death (21.35%)
   (Fig. [136]3b). Cleavage of GSDME by active caspase-3 induces membrane
   pore formation of the N-terminal domain of GSDME, which mediates the
   apoptosis switch to pyroptotic cell death^[137]35,[138]36. Indeed, both
   caspase-3 activation and downstream GSDME cleavage were observed in
   BCV-infected WT BMDMs (Fig. [139]3d). BC infection induced the
   phosphorylation of MLKL and RIP3, which might trigger the necroptosis
   and membrane rupture of BC-infected BMDMs (Figs. [140]2b, [141]3d).

Fig. 3. Amiloride inhibits BCV-triggered cytoplasmic vacuolization cell
death.

   [142]Fig. 3
   [143]Open in a new tab

   a Immunoblot analysis of caspase-3 and cleaved caspase-3 (P17) in WT
   BMDMs infected with BC or BCV (400 MOI) for the indicated time. b TEM
   analysis of WT BMDMs infected with BCV (400 MOI) for 30 h (left and
   middle) and quantification of the percentage of cytoplasmic
   vacuolization and apoptosis-like cell death within the dead cell
   population (right). N indicates the cell nucleus. A total of 89 dead
   cells were analyzed. Scale bars, 1.2 μm for the left and 1.5 μm for the
   middle. c Microscopic analysis of WT BMDMs infected with BC or BCV (400
   MOI) in the presence of amiloride hydrochloride (Amilo, 0.5 μM), z-VAD
   (25 μM), ferrostatin-1 (Fer-1, 1 mM), necrostatin-1 (Nec-1, 100 μM),
   necrosulfonamide (NSA, 500 nM), Rapamycin (Rapa, 500 nM), Wortmannin
   (Wort, 0.2 μM), 3-Methyladenine (3-MA, 5 mM), and
   pyrrolidinedithiocarbamate ammonium (PDTC, 1 μM) for 12 h. Scale bars,
   20 μm. d Immunoblot analysis of caspase-3, cleaved caspase-3 (P17),
   GSDME, cleaved GSDME (GSDME^NT), p-MLKL, MLKL, p-RIP3, RIP3, p-PI3K,
   and p-S6 in WT BMDMs infected with BC or BCV (400 MOI) together with
   various inhibitors in c for 20 h. e LDH analysis of WT BMDMs infected
   with BC or BCV (400 MOI) in the presence of the indicated inhibitors
   for 20 h (n = 3 biologically independent samples). Data are from 3
   independent experiments (e) or representative of 3 independent
   experiments with similar results (a–d). For e, data represent
   Mean ± SEM, and two-sided Student’s t-test without multiple-comparisons
   correction, **P < 0.01, ****P < 0.0001.

   The apoptosis in bystanders was also reported in necroptosis as a
   primary cell death event during SARS-CoV-2 infection^[144]37,
   suggesting that the apoptosis-like cell death in BMDMs might be a
   secondary effect of cytoplasmic vacuolization. We found that the
   supernatant without viable bacteria from BCV- but not BC-infected BMDMs
   induced caspase-3 activation (Supplementary Fig. [145]S5d). Thus, we
   performed mass spectrometry (MS) analysis of supernatants from BCV- and
   BC-infected BMDMs and found that lysosomes and lysosome-associated
   proteins, such as multiple cathepsin proteins, were enriched in the
   supernatants of BCV-treated BMDMs compared with BC-treated cells
   (Supplementary Fig. [146]S5e and Table [147]S3). We purified the
   lysosomes from BCV-infected BMDMs and treated WT BMDMs with purified
   lysosomes (Supplementary Fig. [148]S5f)^[149]38. The caspase-3
   activation, GSDME cleavage, and apoptosis-like cell death were observed
   in lysosome-treated BMDMs (Supplementary Fig. [150]S5g, h).
   Furthermore, cleaved caspase-3 was specifically detected in
   non-vacuolated BMDMs during BCV infection (Supplementary Fig.
   [151]S5i). Together, these data suggest that lysosomes and
   lysosome-associated proteinases enriched in the supernatant of
   BCV-infected BMDMs may drive apoptosis-like cell death in bystander
   cells.

   To further examine whether BCV-triggered cytoplasmic vacuolization was
   dependent on other cell death pathways and intracellular events, WT
   BMDMs were infected with BC or BCV in the presence or absence of
   different types of inhibitors, including the pan-caspase inhibitor
   z-VAD for apoptosis inhibition, the RIPK1 inhibitor necrostatin-1
   (Nec-1) and MLKL inhibitor necrosulfonamide (NSA) for necroptosis
   inhibition, ferrostatin-1 (Fer-1) for ferroptosis inhibition,
   pyrrolidinedithiocarbamate ammonium (PDTC) for NF-κb inhibition,
   3-methyladenine (3-MA) and wortmannin for PI3K and autophagy
   inhibition, rapamycin for mTORC1 inhibition and autophagy activation,
   and amiloride hydrochloride for sodium channel inhibition. Notably,
   amiloride dramatically inhibited BCV-triggered cytoplasmic
   vacuolization, caspase-3 activation, GSDME cleavage, and LDH release
   (Fig. [152]3c–e; Supplementary Fig. [153]S5j). Conversely, z-VAD
   inhibited caspase-3 activation and GSDME cleavage but not cytoplasmic
   vacuolization or cell death (Fig. [154]3c–e; Supplementary Fig.
   [155]S5j), indicating that BCV-triggered cytoplasmic vacuolization cell
   death was not dependent on apoptosis. In addition, inhibitors of
   necroptosis, ferroptosis, autophagy, mTORC1, or NF-κb did not affect
   BCV infection-triggered cytoplasmic vacuolization, caspase-3
   activation, GSDME cleavage, or LDH release (Fig. [156]3c–e;
   Supplementary Fig. [157]S5j). The necroptosis inhibitors Nec-1 and NSA
   inhibited BC infection-triggered cell death (Fig. [158]3d, e). Overall,
   these results indicated that BCV infection induced two separate cell
   death pathways: cytoplasmic vacuolization and associated apoptosis-like
   cell death. Apoptosis-like cell death may be a downstream secondary
   effect of cytoplasmic vacuolization induced by BCV, and ion channels
   are essential for BCV-triggered cytoplasmic vacuolization.

SLC9A9 is important for BCV-induced cytoplasmic vacuolization

   To explore the mechanism underlying BCV infection-induced cytoplasmic
   vacuolization, we performed a genome-wide transcriptional analysis of
   BC- and BCV-infected BMDMs. BCV infection triggered the enrichment of
   genes associated with endosomes, lysosomes, and metabolism (Fig.
   [159]4a; Supplementary Fig. [160]S6a and Table [161]S4). However, the
   expression levels of genes involved in inflammatory response pathways,
   including TNF, chemokine, NF-κB, and MAPK signaling, were dramatically
   reduced in BCV- compared with BC-infected BMDMs (Supplementary Fig.
   [162]S6a–c and Table [163]S4). Given that inhibitors targeting ion
   channels have suppressive effects on the process of BCV
   infection-induced cytoplasmic vacuolization, we found that specific
   genes encoding membrane proteins and transporters that were localized
   to intracellular compartments, especially solute carriers (SLCs), were
   highly expressed in BMDMs infected with BCV but not BC (Fig. [164]4b;
   Supplementary Fig. [165]S6d).

Fig. 4. SLC9A9 regulates BCV infection-triggered cytoplasmic vacuolization
cell death.

   [166]Fig. 4
   [167]Open in a new tab

   a, b RNA-seq analysis of the gene expression of WT BMDMs infected with
   BC or BCV (400 MOI) for 6 and 12 h. Heatmap showing the comparison of
   endosome- and lysosome-associated genes (a) and genes encoding
   receptors and transporters (b) between BC- and BCV-infected BMDMs, with
   increased expression in BCV-infected BMDMs. c Quantitative RT-PCR
   analysis of Slc9a9 expression in siCtrl- and siSlc9a9-transfected BMDMs
   for 36 h (n = 4 technical replicates; 3 independent experiments). Hprt
   stands for genes encoding hypoxanthine guanine
   phosphoribosyltransferase. d Microscopic analysis of siRNA-knockdown
   BMDMs (siCtrl and siSlc9a9) infected with BCV (400 MOI) for the
   indicated time. Scale bars, 30 μm. e LDH analysis of siCtrl- and
   siSlc9a9-transfected BMDMs in d (n = 3 biologically independent
   samples). f Microscopic analysis of WT and Slc9a9^−^/^− BMDMs infected
   with BCV (400 MOI) for the indicated time. Scale bars, 30 μm. g
   Immunoblot analysis of caspase-3, cleaved caspase-3 (P17), GSDME,
   cleaved GSDME (GSDME^NT), p-MLKL, MLKL, p-RIP3, RIP3, and SLC9A9 in WT
   and Slc9a9^−^/^− (KO) BMDMs infected with BC or BCV (400 MOI) for 20 h.
   h Immunoblot analysis of Flag-SLC9A9 co-immunoprecipitated with
   V5-Vps18, V5-Vps33a, and V5-Stx18 from lysates of HEK293T cells
   transfected with the indicated plasmids. i Confocal microscopy analysis
   of SLC9A9 and LAMP1 in BCV-infected (400 MOI) WT BMDMs at the indicated
   time. The arrows indicate SLC9A9 subcellular localization at the
   membrane of vacuoles. Scale bars, 10 μm. j A hypothetical model of the
   interaction between SLC9A9 with HOPS, SNAREs and Rab7. Data are from 3
   independent experiments (e) or representative of 3 independent
   experiments with similar results (c, d, f–i). Data represent Mean ± SEM
   for c, e, two-sided Student’s t-test without multiple-comparisons
   correction, **P < 0.01, ****P < 0.0001.

   We performed siRNA transfection for 13 candidate genes more highly
   expressed in BCV-infected than in BC-infected BMDMs (Fig. [168]4b, c;
   Supplementary Fig. [169]S6e). Notably, the cytoplasmic vacuolization
   and LDH release induced by BCV infection were markedly inhibited in
   SiSlc9a9-transfected BMDMs but not in SiSlc9a1-, SiSlc9b2-, SiSlc12a2-,
   SiSlc13a3-, SiSlc35g1-, SiAtp1a1-, SiRab7b-, SiGpr65-, SiCd5l-,
   SiOlfr56-, SiPtpn22-, or SiTlr9-transfected BMDMs (Fig. [170]4d, e;
   Supplementary Fig. [171]S6f). To provide genetic evidence for the role
   of SLC9A9 in the process of BCV-triggered cytoplasmic vacuolization, we
   generated Slc9a9^−^/^− mice (Supplementary Fig. [172]S7a–c), which were
   viable with characteristics similar to those of WT mice. SLC9A9
   deficiency in BMDMs significantly reduced the degree of BCV-triggered
   cytoplasmic vacuolization based on microscopy, quantification by 40%
   Percoll gradient separation, and LDH release assay analyses (Fig.
   [173]4f; Supplementary Fig. [174]S7d–f). In addition, the caspase-3
   activation and GSDME cleavage induced by BCV were substantially reduced
   in the absence of SLC9A9 (Fig. [175]4g).

   Membrane fusion is mediated by the zipping of soluble N-ethylmaleimide
   sensitive factor attachment protein receptors (SNAREs) and the
   tethering of homotypic fusion and protein sorting protein complex
   (HOPS) or class C core vacuole/endosome tethering (CORVET) with the aid
   of accessory proteins^[176]39–[177]41. Distinct sets of HOPS-mediated
   membrane tethering and SNARE-mediated membrane fusion regulate lysosome
   fusion with late endosomes or autophagosomes and lipid droplet
   fusion^[178]42,[179]43. To define whether SLC9A9 promotes vacuole
   fusion through interacting with membrane fusion machinery, we performed
   a co-IP analysis to identify SLC9A9-interacting proteins in HEK293T
   cells transfected with SLC9A9 and other proteins. Notably, SLC9A9
   preferentially interacted with all the HOPS subunits Vps11, Vps16,
   Vps18, Vps33, Vps39, and Vps41 (Fig. [180]4h; Supplementary Fig.
   [181]S7g, h). In addition, lysosome-resident Rab7, but not early
   endosome-associated Rab5a or Rab5b, was demonstrated to interact with
   SLC9A9 (Supplementary Fig. [182]S7g, h). Furthermore, we detected that
   SLC9A9 interacted with members within distinct sets of SNARE complexes,
   including synaptosomal-associated proteins (SNAP) and syntaxins (Stx)
   proteins. However, Vamp proteins were not detected to interact with
   SLC9A9 (Fig. [183]4h; Supplementary Fig. [184]S8a, b). In addition,
   amiloride treatment inhibited SLC9A9 interaction with Vps11, Vps16,
   Vps18, Vps33b, Vps39, Rab7, Snap47, and Esyt1 (Supplementary Fig.
   [185]S9a, b), suggesting that SLC9A9 is one of the targets of amiloride
   to inhibit cytoplasmic vacuolization. During BCV infection in BMDMs,
   SLC9A9 dramatically localized to the contact sites of LAMP1-associated
   vacuoles, revealed by confocal microscopy and live-cell imaging
   analysis (Fig. [186]4i; Supplementary Fig. [187]S10a, b). However,
   SLC9A9 deficiency did not affect the formation of cytoplasmic
   vacuolization triggered by MOMIPP (Supplementary Fig. [188]S10c).
   Collectively, these data indicate that SLC9A9 acts as an important
   regulator of vacuole fusion to mediate BCV-triggered cytoplasmic
   vacuolization cell death (Fig. [189]4j).

OMVs and barrel-like proteins of BCV induce the formation of cytoplasmic
vacuolization

   Given the difference in membrane puff-ups and blebbing produced by BC
   and BCV (Fig. [190]1e), we examined the OMVs derived from both BC and
   BCV. Notably, the production of intact OMVs in BCV dramatically
   increased compared to that in BC (Fig. [191]5a, b). Furthermore, the
   OMVs derived from BCV also triggered cytoplasmic vacuolization and
   caspase-3 activation (Fig. [192]5c–e). The cytoplasmic vacuolization
   cell death triggered by OMVs and downstream caspase-3 activation were
   substantially reduced in the absence of SLC9A9 (Fig. [193]5f–i).
   Furthermore, amiloride treatment also significantly inhibited the
   cytoplasmic vacuolization, LDH release, caspase-3 activation, and GSDME
   cleavage of WT BMDMs triggered by OMVs derived from BCV (Fig.
   [194]5j–m), suggesting that OMVs can mimic BCV to induce cytoplasmic
   vacuolization cell death.

Fig. 5. OMVs derived from BCV trigger cytoplasmic vacuolization cell death.

   [195]Fig. 5
   [196]Open in a new tab

   a TEM analysis of purified OMVs from BC and BCV. BC and BCV were
   cultured on Columbia blood agar plates for 96 h, followed by growth in
   BACTEC™ Lytic media for 16 h. The supernatants of BC and BCV were used
   for OMV purification and comparative analysis. Scale bars, 0.3 μm. b
   Quantification of the number of OMVs per field in a (n = 12 random
   fields; 3 independent experiments). c Microscopic analysis of WT BMDMs
   treated with OMVs derived from BCV (100 μg) for the indicated time.
   Scale bars, 30 μm. d Quantification of vacuole size in the BMDMs in c.
   The largest vacuole per cell was analyzed, and at least 120 cells were
   quantified for each group. e Immunoblot analysis of caspase-3 and
   cleaved caspase-3 (P17) in WT BMDMs treated with OMVs derived from BCV
   (100 μg) for the indicated time. f Microscopic analysis of WT and
   Slc9a9^−^/^− BMDMs treated with OMVs derived from BCV (100 μg) for the
   indicated time. Scale bars, 30 μm. g Quantification of vacuole size in
   the BMDMs in f. The largest vacuole per cell was analyzed, and at least
   120 cells were quantified for each group. h LDH analysis of WT and
   Slc9a9^−^/^− BMDMs treated with OMVs derived from BCV (100 μg) for the
   indicated time (n = 4 biologically independent samples). i Immunoblot
   analysis of caspase-3, cleaved caspase-3 (P17), and SLC9A9 in WT and
   Slc9a9^−^/^− BMDMs treated with OMVs derived from BCV (100 μg) for the
   indicated time. j Microscopic analysis of WT BMDMs treated with OMVs
   derived from BCV (100 μg) in the presence and absence of amiloride
   hydrochloride (Amiloride, 0.5 μM) treatment for the indicated time.
   Scale bars, 30 μm. k Quantification of vacuole size in the BMDMs in j.
   The largest vacuole per cell was analyzed, and at least 120 cells were
   quantified for each group. l LDH analysis of WT BMDMs treated with OMVs
   derived from BCV (100 μg) in the presence and absence of amiloride
   hydrochloride (Amiloride, 0.5 μM) treatment for the indicated time. m
   Immunoblot analysis of caspase-3, cleaved caspase-3 (P17), GSDME, and
   cleaved GSDME (GSDME^NT) in WT BMDMs treated with OMVs derived from BCV
   (100 μg) in the presence and absence of amiloride hydrochloride
   (Amiloride, 0.5 μM) treatment for the indicated time. Data are from 3
   independent experiments (h, l) or representative of 3 independent
   experiments with similar results (a–g, i–k, m). Data represent
   Mean ± SEM for (b, d, g, h, k, l), ***P < 0.001, ****P < 0.0001, by
   two-sided Student’s t-test without multiple-comparisons correction.

   Next, we performed MS analysis of bacterial proteins in the
   supernatants from BCV- and BC-infected BMDMs, and detected high levels
   of PorV, lipocalin, β-barrel protein, and adenylosuccinate lyase (ADSL)
   in the supernatants of BCV-infected BMDMs but not in those of
   BC-infected cells (Supplementary Fig. [197]S11a and Table [198]S3).
   PorV, lipocalin, β-barrel protein, and adenylosuccinate lyase (ADSL)
   were also detected in the OMVs derived from BCV (Supplementary Fig.
   [199]S11b and Table [200]S5). However, the expression of PorV,
   lipocalin, β-barrel, and ADSL was comparable between BCV and BC
   (Supplementary Table [201]S2), indicating that OMVs derived from BCV
   might contribute to the enrichment of these proteins in the
   supernatants of BCV-infected BMDMs. Thus, we cloned PorV, lipocalin,
   β-barrel, and ADSL from the BCV genome into the expression vector and
   purified the recombinant proteins (Supplementary Fig. [202]S11c). The
   recombinant proteins of GST-fused PorV, GST-fused lipocalin, GST-fused
   β-barrel, GST-fused ADSL, or GST alone did not cause cytoplasmic
   vacuolization following their direct addition to the BMDMs
   (Supplementary Fig. [203]S11d). However, transfection of GST-fused
   lipocalin or GST-fused β-barrel proteins into the cells resulted in
   profound cytoplasmic vacuolization between 1 and 12 h after
   transfection (Fig. [204]6a; Supplementary Fig. [205]S12a). The
   cytoplasmic vacuolization caused by the transfection of the GST-fused
   PorV protein gradually decreased after 2 h and substantially reduced up
   to 12 h after treatment (Fig. [206]6a; Supplementary Fig. [207]S12a).
   In contrast, transfection of GST-fused ADSL or GST protein alone did
   not cause cytoplasmic vacuolization in BMDMs (Fig. [208]6a;
   Supplementary Fig. [209]S12a).

Fig. 6. Transfection of the lipocalin, β-barrel, and PorV proteins of BCV
induces the formation of cytoplasmic vacuolization cell death.

   [210]Fig. 6
   [211]Open in a new tab

   a Microscopic analysis of WT BMDMs transfected with GST (10 μg),
   GST-PorV (10 μg), GST-Lipocalin (10 μg), GST-β-barrel (10 μg), and
   GST-ADSL (10 μg) proteins for 1, 2, and 12 h as indicated. The arrows
   indicate large vacuole-containing cells. Scale bars, 30 μm. b Confocal
   microscopy analysis of GST and LAMP1 in BMDMs transfected with GST (10
   μg), GST-PorV (10 μg), GST-Lipocalin (10 μg), GST-β-barrel (10 μg), and
   GST-ADSL (10 μg) proteins for 2 h. The arrows indicate the subcellular
   localization of the GST-fused proteins. Scale bars, 10 μm. c
   Microscopic analysis of WT and Slc9a9^−^/^− BMDMs transfected with the
   combination of GST-Lipocalin (10 μg), GST-β-barrel (10 μg), and
   GST-PorV (10 μg) proteins for the indicated time. The arrows indicate
   vacuole-occupied cells. Scale bars, 30 μm. d Quantification of vacuole
   size in WT and Slc9a9^−^/^− BMDMs in c. The largest vacuole per cell
   was analyzed, and at least 120 cells were quantified for each group. e
   LDH analysis of WT and Slc9a9^−^/^− BMDMs in c (n = 4 biologically
   independent samples). f Immunoblot analysis of caspase-3, cleaved
   caspase-3 (P17), and SLC9A9 in WT and Slc9a9^−^/^− BMDMs transfected
   with GST-fused proteins in c for 20 h. g Confocal microscopy analysis
   of GST and SLC9A9 in WT BMDMs transfected with the GST-fused proteins
   in c for 2 and 10 h. Scale bars, 10 μm. Data are from 3 independent
   experiments (e) or representative of 3 independent experiments with
   similar results (a–d, f, g). Data represent Mean ± SEM for d, e,
   *P < 0.05, ***P < 0.001, ****P < 0.0001, by two-sided Student’s t-test
   without multiple-comparisons correction.

   The intracellular lipocalin, β-barrel, and PorV proteins were
   preferentially accumulated on the membrane of LAMP1-associated vacuoles
   (Fig. [212]6b). Moreover, transfection of the GST-fused lipocalin
   protein together with the GST-fused β-barrel protein or the combination
   of the GST-fused lipocalin, GST-fused β-barrel, and GST-fused PorV
   proteins induced the formation of large vacuoles at 12 h and the
   formation of single vacuole-occupied cells at 20 h after transfection
   (Supplementary Fig. [213]S12b, c). In line with BCV infection and OMVs
   treatment, SLC9A9 deficiency significantly inhibited the cytoplasmic
   vacuolization cell death and downstream caspase-3 activation triggered
   by the combined transfection of lipocalin, β-barrel, and PorV proteins
   (Fig. [214]6c–f). Moreover, amiloride treatment also significantly
   inhibited the cytoplasmic vacuolization, LDH release, caspase-3
   activation, and GSDME cleavage of WT BMDMs transfected with combined
   lipocalin, β-barrel, and PorV proteins (Supplementary Fig.
   [215]S12d–g). Protein structure prediction revealed that lipocalin,
   β-barrel, and PorV had multiple β-strands forming barrel-like
   structures (Supplementary Fig. [216]S12h). However, the majority of
   intracellular SLC9A9 distribution did not colocalize with the
   transfected lipocalin, β-barrel, or PorV proteins (Fig. [217]6g),
   indicating that the transfected proteins did not directly interact with
   SLC9A9 to trigger cytoplasmic vacuolization cell death.

BCV increases in vivo pathogenicity

   Confocal microscopy and CFU analysis revealed that the amount of
   intracellular BCV in BMDMs was significantly higher than that of BC
   (Supplementary Fig. [218]S13a–c). The intracellular BCV was primarily
   distributed in the vacuoles labeled with LAMP1 at 8 h post-infection
   (Supplementary Fig. [219]S13a). After infection in BMDMs, 276 genes
   were upregulated in BCV compared with BC, including the genes
   contributing to stress adaptation, secretion system, invasion, and
   colonization of bacteria (Supplementary Fig. [220]S13d, e and Table
   [221]S6). We further performed bacterial killing assay to investigate
   the bacterial entry and intracellular survival of BCV and BC, and the
   results showed that the bacterial entry of BC into BMDMs was
   significantly lower than that of BCV, and the bacterial burden of BCV
   at 22 h post-infection was dramatically greater than that of BC (Fig.
   [222]7a). However, the bacterial growth of BC and BCV was comparable in
   culture media (Supplementary Fig. [223]S14a), suggesting that BCV
   evolved the capacity to undergo intracellular propagation.

Fig. 7. BCV-triggered cytoplasmic vacuolization cell death increases
pathogenicity in vivo.

   [224]Fig. 7
   [225]Open in a new tab

   a Bacterial killing ability of BC and BCV in BMDMs. WT BMDMs were
   infected with BC (20 MOI) and BCV (20 MOI) for 2 h, followed by
   treatment with gentamicin (50 μg/mL) for 1 h. Infected BMDMs were
   washed, lysed, and cultured on Columbia blood agar plates for 96 h to
   enumerate intracellular (2 h) bacteria. Washed BMDMs were further
   cultured in fresh media for 22 h, and the total number of intracellular
   and extracellular bacteria (22 h) was enumerated after culturing on
   Columbia blood agar plates for 96 h (n = 3 biologically independent
   samples). b, c WT female mice were intranasally infected with 4.0 ×
   10^8 CFU BC (n = 12) or BCV (n = 12), and the body weight change (b)
   and bacterial burden in the lungs on Day 1 after infection were
   measured (c). d TEM analysis of OMVs in bronchoalveolar lavage fluid
   (500 μL volume per mouse) from BC- and BCV-infected mice in c. Scale
   bars, 0.3 μm. e Microscopic analysis of cytoplasmic vacuolization in
   isolated AMs from BC- and BCV-infected mice in c. Scale bars, 20 μm. f
   H&E staining of lung sections from uninfected and BC- and BCV-infected
   mice in c. Scale bars, 100 μm. g Disease scores based on inflammation
   in lung sections in f from uninfected (Uninf, n = 2), BC-infected
   (n = 4), and BCV-infected (n = 4) mice. h Expression of genes encoding
   IL-1α and IL-6 was analyzed in lung tissues from uninfected (Uninf,
   n = 3) and BC-infected (n = 10), and BCV-infected (n = 10) mice in c. i
   Bacterial killing ability of BCV in WT and Slc9a9^−^/^− BMDMs. WT and
   Slc9a9^−^/^−BMDMs were infected with BCV (20 MOI) for 2 h, followed by
   treatment with gentamicin (50 μg/mL) for 1 h. Infected BMDMs were
   washed, lysed, and cultured on Columbia blood agar plates for 96 h to
   enumerate intracellular (2 h) bacteria. Washed BMDMs were further
   cultured in fresh media for 22 h, and the total number of intracellular
   and extracellular bacteria (22 h) was enumerated after culturing on
   Columbia blood agar plates for 96 h (n = 3 biologically independent
   samples). j, k WT and Slc9a9^−^/^− female mice were intranasally
   infected with 4.0 × 10^8 CFU BCV (n = 8 mice for each group), and the
   body weight change (j) and bacterial burden in the lungs on Day 1 after
   infection were measured (k). l H&E staining of lung sections from
   BCV-infected mice in k. Scale bars, 100 μm. m Disease scores based on
   inflammation in lung sections in l from BCV-infected WT and
   Slc9a9^−^/^−mice (n = 5 mice for each group). n ELISA analysis of IL-1α
   and IL-6 in sera from uninfected and BCV-infected WT and
   Slc9a9^−^/^−mice (n = 5 mice for each group) in k. o WT female mice
   were intranasally infected with BCV (4.0 × 10^8 CFU), and the bacterial
   burden in the lungs was measured 24 h after infection. Amiloride
   indicates that the mice were intravenously injected with amiloride
   hydrochloride (10 mg/kg) twice, at 0 and 12 h after BCV infection
   (n = 7 mice for each group). p H&E staining of lung sections from DMSO-
   or amiloride hydrochloride-treated mice infected with BCV in o. Scale
   bars, 100 μm. q Disease scores based on inflammation in the lung
   sections in p (n = 3 mice for each group). Data are from 2 independent
   experiments (j, k) or representative of 3 independent experiments with
   similar results (a–i, l–q). Data represent Mean ± SEM for (a–c, g–k,
   m–o, q), two-sided Student’s t-test without multiple-comparisons
   correction, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

   To examine the pathogenicity of BCV and BC in vivo, we first infected
   alveolar macrophages (AMs) isolated from pulmonary lavage fluid ex vivo
   with BC or BCV. Consistent with the results obtained using BMDMs,
   BCV-infected AMs demonstrated remarkable cytoplasmic vacuolization
   (Supplementary Fig. [226]S14b). Next, WT mice were intranasally
   infected with BCV or BC (4.0 × 10^8 CFU per mouse), and the bacterial
   burden and host response were assessed. Both the body weight loss and
   bacterial burden in the lungs of the BCV-infected mice were
   significantly greater than those of the BC-infected mice 24 h
   post-infection (Fig. [227]7b, c). TEM analysis revealed that the OMVs
   in bronchoalveolar lavage fluid were detected in BCV-infected mice but
   not BC-infected mice (Fig. [228]7d), and cytoplasmic vacuolization in
   AMs was also observed in BCV-infected mice (Fig. [229]7e). Furthermore,
   inflammatory responses and lung pathology, including vascular muscle
   hypertrophy, infiltration of immune cells, and inflammatory cytokine
   expression, were more severe in the BCV-infected than in the
   BC-infected mice (Fig. [230]7f–h).

   To determine the roles of SLC9A9 in BCV pathogenesis and host defense,
   we performed BCV infection in WT and Slc9a9^−^/^− BMDMs and mice. The
   intracellular replication of BCV in BMDMs was reduced in the absence of
   SLC9A9 (Fig. [231]7i). During intranasal infection with BCV in WT and
   Slc9a9^−^/^− mice, the body weight loss and bacterial burden in the
   lungs of the BCV-infected Slc9a9^−^/^− mice were significantly less
   than those in WT mice at 24 h post-infection (Fig. [232]7j, k), Lung
   pathology and inflammation, and systemic inflammatory cytokines IL-1α
   and IL-6 in sera were attenuated in the BCV-infected Slc9a9^−^/^− mice
   compared with WT mice (Fig. [233]7l–n). To examine whether immune cells
   or nonimmune cells contribute to the increased host defense in
   Slc9a9^−^/^− mice during BCV infection, we performed bone marrow
   transplantation between WT (CD45.1^+) and Slc9a9^−^/^− (CD45.2^+) mice.
   Flow cytometry analysis revealed that donor-derived bone marrow cells
   were predominantly detected in the bone marrow, blood, and spleen of
   recipient mice six weeks after transplantation (Supplementary Fig.
   [234]S14c), and anti-SLC9A9 western blot analysis showed that SLC9A9
   was notably detected in the spleen and lung from recipient
   Slc9a9^−^/^−mice that received bone marrow from WT mice, but not
   detected in the spleen or lung from recipient WT mice that received
   bone marrow from Slc9a9^−^/^− mice (Supplementary Fig. [235]S14d),
   indicating that the transplantation system is successful in the
   chimeric mice. Next, chimeric mice were intranasally infected with BCV,
   and the bacterial loads in the lung were determined. Interestingly, WT
   and Slc9a9^−^/^− mice that received bone marrow from Slc9a9^−^/^− mice
   exhibited significantly lower bacterial loads in the lung and less body
   weight loss than WT and Slc9a9^−^/^− mice that received bone marrow
   from WT mice, respectively (Supplementary Fig. [236]S14e, f), whereas
   recipient WT and Slc9a9^−^/^−mice that received same donor cells had a
   comparable capacity to regulate the bacterial burden and body weight
   loss (Supplementary Fig. [237]S14e, f). H&E staining revealed that the
   lung pathology was decreased in WT and Slc9a9^−^/^− mice that received
   bone marrow from Slc9a9^−^/^− mice (Supplementary Fig. [238]S14g).
   These data indicate that SLC9A9 in immune cells plays a dominant role
   in the host defense against BCV infection.

   Given that amiloride can inhibit cytoplasmic vacuolization cell death
   induced by BCV in BMDMs, to further determine whether cytoplasmic
   vacuolization contributed to the high pathogenicity of the BCV
   pathogen, we combined BCV administration with amiloride treatment in
   vivo, which might inhibit the cytoplasmic vacuolization induced by BCV
   infection. Interestingly, the bacterial burden of BCV in the lung was
   significantly reduced in the WT mice treated with amiloride (Fig.
   [239]7o). In addition, H&E staining revealed that the lung pathology
   was decreased in the presence of amiloride (Fig. [240]7p, q). The
   amiloride administration in Slc9a9^−^/^− mice further increased host
   defense against BCV infection, characterized by less body weight loss,
   reduced bacterial burden, and decreased lung pathology (Supplementary
   Fig. [241]S14h–j). Overall, these results indicated that BCV was a more
   virulent variant of the Bergeyella cardium bacterial strain, OMVs and
   the membrane barrel-like proteins lipocalin, β-barrel, and PorV were
   important for BCV-induced cytoplasmic vacuolization cell death via
   SLC9A9-mediated vacuole fusion (Supplementary Fig. [242]S15).

Discussion

   Previous studies on Bergeyella cardium strains are limited to a few
   clinical case reports due to the difficulty of growth in multiple sets
   of blood cultures^[243]26,[244]44. We isolated both BC and BCV from a
   patient with infective endocarditis who had used antibiotics long
   before the clinical examination^[245]26. Our work demonstrated that BCV
   exhibited several characteristics contributing to immune evasion,
   including acquired cell entry and intracellular replication capacity,
   robust OMV release, increased serum killing resistance, downregulated
   inflammatory cytokine expression, induced cytoplasmic vacuolization
   cell death, and increased in vivo pathogenicity. Whether the long-term
   use of antibiotics in patients induced genetic changes or epigenetic
   switching of BCV to promote immune evasion capacity remains unclear.
   Our study shows Bergeyella cardium as an emerging pathogen frequently
   detected in oral specimens from clinical patients and presents a
   variant of oral microbiomes has high virulence and is potentially
   associated with noncausal infectious diseases.

   Genomic analysis revealed gene differences between BC and BCV,
   including PorV and BatD. The Bat proteins have been shown to partially
   compensate for the oxidative stress response in bacteria and provide
   defense against oxidative damage^[246]45. In Francisella tularensis,
   the BatD homolog mutant reduced the intracellular replication capacity
   in macrophages and virulence in a mouse model^[247]46. Thus, adding the
   BatD protein to BCV might partially contribute to stress adaptation,
   intracellular replication, and in vivo pathogenicity mechanisms. PorV
   is a vital machinery component of T9SS and functions as a shuttle
   protein to deliver T9SS substrates to the attachment complex on the
   cell surface^[248]47. Structural prediction revealed that the deletion
   corresponding to residues 344–351 of BCV PorV led to structural changes
   in PorV. It is possible that the ‘BCV β15’ strand promoted β-barrel
   formation altogether, which might be associated with the robust
   production of OMVs in BCV through mediating the secretion and
   attachment of cargo proteins to the cell surface and
   OMVs^[249]48,[250]49. We speculate that the increased virulence of BCV
   could be partially due to the robust production of OMVs and secretion
   of barrel-like proteins, causing cytoplasmic vacuolization and cell
   death. A limitation of this study is the lack of genetic knockout and
   complementation experiments in Bergeyella cardium strains to confirm
   whether these genomic mutations contributed to robust OMV biogenesis
   and the secretion of virulence factors in BCV. Challenges remain in
   establishing genetic manipulation techniques for novel organisms due to
   the lack of genetic manipulation tools^[251]50, which should be a focus
   for future development. In this study, BCV and BC actually represent
   two mutual mutants of each other for understanding the association
   between phenotype and genetic differences.

   Bacterial OMVs play a critical role in host-microbial interactions that
   influence pathogenesis by delivering virulence factors and eliciting
   inflammatory responses^[252]51. The large number of intact OMVs shed by
   BCV might result in an increasing delivery of virulence factors such as
   lipocalin, β-barrel, and PorV into BMDMs to cause cytoplasmic
   vacuolization, whereas the engagement with membrane receptors and
   internalization were essential for vacuolization triggered by the
   vacuolating toxins SubAB, CARDS, and VacA^[253]7,[254]10,[255]11.
   Lipocalin family members are ancestral proteins found in all kingdoms
   of life, and secreted lipocalin polypeptides have been reported to
   induce apoptosis of leukocytes through an autocrine pathway^[256]52.
   β-barrel proteins contribute to forming β-barrel outer membrane
   proteins and are conserved across species in terms of folding and
   insertion into the outer membrane^[257]53. Transmembrane β-barrel
   proteins can fold spontaneously and assemble into lipid membranes to
   form stable pores^[258]54. Vacuolating toxin VacA also contains five
   β-helix subdomains and a C-terminal β-barrel domain^[259]55.
   Barrel-like proteins can be delivered into the cytoplasm either through
   the secretion system of the intracellular BCV itself or via BCV-derived
   OMVs. The barrel-like structures of lipocalin, β-barrel, and PorV of
   BCV might contribute to their assembly into the membrane, leading to
   the initiation and formation of cytoplasmic vacuolization cell death.
   Defining the receptor’s engagement with the transfected barrel-like
   proteins to trigger cytoplasmic vacuolization is inspiring.

   The interplay between lipids and proteins plays a crucial role in
   dynamic membrane remodeling, and phosphatidylinositol bisphosphate
   (PIP[2]) controls the formation and spatiotemporal organization of
   protein complexes involved in vesicle budding, trafficking, and
   membrane curvature and fusion^[260]56. Recently, PIKfyve, a class III
   phosphoinositide (PI) kinase, was identified as an inhibition target of
   MOMIPP for inducing methuosis^[261]57. Interestingly, deficiency of the
   PIKfyve lipid kinase complex components, including the phosphoinositide
   (PI) kinase PIKfyve, the scaffolding protein Vac14, and the lipid
   phosphatase Fig4, inhibited the conversion of
   phosphatidylinositol-3-phosphate (PI3P) to
   phosphatidylinositol-3,5-bisphosphate PI(3,5)P[2] and caused loss of
   PI(3,5)P[2] and remarkable lysosome vacuolation in multiple cells in
   both mice and humans^[262]58–[263]64. Likewise, PI3P depletion causes
   the accumulation of deidentified early endosomes, triggering
   triaptosis^[264]20. Thus, deidentified early endosome and lysosome
   might be the direct cause for the formation of vacuolization, and the
   barrel-like proteins of BCV might ultimately influence the homeostasis
   and distribution of PI(3,5)P[2] to modulate lysosome membrane dynamics
   and cytoplasmic vacuolization, which requires further investigation.

   The equilibration of osmotic pressure within organelles by water
   diffusion, rather than programmed gene expression and protein
   regulation, is involved in transient vacuolization formation induced by
   natural or synthetic chemical compounds and virulence factors from
   bacterial and viral pathogens^[265]65–[266]70. Previously, membrane
   fusion SNARE proteins Syntaxin 7 and VAMP7 and late endosome docking
   protein Rab7 were also reported to be essential for clustering late
   endocytic compartments and cell vacuolization induced by Helicobacter
   pylori VacA treatment^[267]71–[268]73. This study found that
   intracellular membrane fusion mediated by SLC9A9 (NHE9) interaction
   with HOPS components and specific SNARE proteins was critical for the
   cytoplasmic vacuolization cell death during BCV infection. SLC9A9 also
   colocalizes with intracellular late endosomes and contributes to the
   transport of sodium and hydrogen ions across the membrane to maintain
   the pH balance of intracellular organelles^[269]74,[270]75.
   Overexpression of SLC9A9 has been reported to lead to endosomal
   alkalization, SLC9A9 knockdown has been shown to induce endosomal
   acidification, and the organellar pH is essential for enzyme activity,
   membrane fusion, and cell volume^[271]74. In contrast, VacA induced
   vacuolization mainly through its anion-selective channel for the influx
   of anions into endosomes and stimulation of the v-ATPase proton pump,
   which caused the acidification of endosomes^[272]13. Recently,
   endosomal PI(3,5)P[2] was corroborated to directly bind with SLC9A9 for
   mediating its dimerization and sodium/proton exchange
   activity^[273]76,[274]77. Beyond this canonical function, our data
   revealed that SLC9A9 might principally interact with membrane fusion
   proteins localized in Rab7-resident vacuoles to initiate assembly of
   HOPS complex and promote vacuole fusion. However, the interplay between
   SLC9A9 and the intracellular PI(3,5)P[2] composition and the balanced
   regulation between vacuole sodium/proton exchange and the membrane
   fusion via SLC9A9 remains elusive. SLC9A9 represents one of the
   significant targets modulated by BCV to hijack for immune evasion.
   Thus, BCV-induced vacuolization cell death shares common
   characteristics with methuosis and VacA-triggered vacuolization of late
   endosome and lysosome hybrid, but also has distinct features of
   initiation specificity and involvement of the membrane fusion process.
   BCV-triggered cytoplasmic vacuolization in BMDMs represents a unique
   cell death, and we term it Fused LysosOme-Associated Termination
   (Floatptosis).

   Solute carriers are a large group of membrane transport proteins with
   more than 300 members in humans, and the SLC9 family (also called
   sodium/hydrogen exchanger, NHE) is divided into three subgroups based
   on membrane localization and cell type specificity^[275]74. Solute
   carrier transporters are essential metabolic regulators of immune cells
   and play critical roles in cancer immunotherapy^[276]78,[277]79.
   Dysfunction of SLC9 family members is associated with multiple
   diseases, such as cancer, neurological disorders, gastrointestinal
   tract disorders, and kidney disease^[278]80–[279]82. Thus, the finding
   of SLC9A9-mediated vacuole fusion and cytoplasmic vacuolization cell
   death upon BCV infection provides novel insights into the mechanisms of
   both SLC9A9- and vacuolization-associated diseases.

Experimental procedures

Mice

   Ifnar^−^/^−, Aim2^−^/^−Nlrp3^−^/^−, and Asc^−^/^− mice were previously
   described^[280]83. Slc9a9^−^/^− mice were generated by GemPharmatech
   Co., Ltd (Suzhou, Jiangsu). Exon 2 of the Slc9a9 gene was knocked out
   by CRISPR-Cas9 system. The strategy for constructing the targeting
   vector is illustrated in Supplementary Fig. [281]S7a. The knockout of
   the Slc9a9 gene was validated by genotyping, qRT-PCR, and western blot
   (Fig. [282]4g; Supplementary Fig. [283]S7b, c). WT and knockout mice
   were kept under specific pathogen-free conditions in the Animal
   Resource Center at Shandong University, Jinan, Shandong Province,
   China. All animal experiments were conducted in accordance with
   guidelines approved by the Ethics Committee of Scientific Research of
   Shandong University.

Bacteria and OMVs preparation

   BC and BCV were cultured on Columbia blood agar plates
   (Autobio,0004763) for 96 h, followed by growing in commercial BACTE^TM
   Lytic media from BD BACTEC^TM Lytic/10 Anaerobic/F Culture Vials (BD,
   442021) with shaking (220 rpm) for 16 h at 37 °C as previously
   described^[284]26. The bacteria were precipitated by centrifugation at
   3000 rpm for 30 min at 4 °C, and resuspended in DMEM/F-12 media for
   infection experiments. For OMVs preparation, 200 mL BCV culture
   (OD[600]is around 1.0) were precipitated by centrifugation at 3000 rpm
   for 30 min at 4 °C, and the supernatants were passed through a 0.45 µm
   filter to remove debris, and the filtered supernatants were
   precipitated by ultracentrifugation at 40,000× g for 2 h at 4 °C, and
   resuspended in 1 mL BACTEC™ Lytic media or PBS for TEM analysis and
   treatment experiments, respectively. The concentration of OMVs was
   quantified using Pierce BCA protein assay kit (23227).

Preparation of BMDMs, treatment, bacterial infection, and siRNA transfection

   To generate BMDMs, bone marrow cells were cultured in L929
   cell-conditioned DMEM/F-12 supplemented with 10% FBS, 1% nonessential
   amino acids, and 1% penicillin-streptomycin for 5 days. The siRNAs
   targeting Slc9a9, Slc9a1, Slc9b2, Slc12a2, Slc13a3, Slc35g1, Atp1a1,
   Rab7, Gpr65, Cd5l, Olfr56, Ptpn22, and Tlr9 were ordered from RiboBio
   Company. siRNAs were electroporated into BMDMs using the Neon^TM
   Transfection System following the manufacturer’s instructions. Briefly,
   0.25 nmol siRNAs were electroporated into 5.5 M BMDMs in 110 μL Neon™
   Electroporation T Buffer, Electroporation was performed using the Neon™
   Transfection System at 2050 V with a single 25 ms pulse. After 36 h,
   the transfected BMDMs were infected with BCV for vacuolization. The
   siRNA sequences are listed in Supplementary Table [285]S7.

   Inhibitors z-VAD (Calbiochem, 627610), necrostatin-1 (Nec-1;
   Calbiochem, 480065), necrosulfonamide (NSA; MCE, HY-100573),
   ferrostatin-1 (Fer-1; MCE, HY-100579), pyrrolidinedithiocarbamate
   ammonium (PDTC; TargetMoI, T3147), 3-Methyladenine (3-MA; APExBIO
   Technology, A8353), wortmannin (CST, 9951S), rapamycin (MCE, HY-10219),
   and amiloride hydrochloride (Alomone labs, A-140) were used to treat
   BMDMs for 2 h with indicated concentration ahead of bacterial
   infection.

   To induce cytoplasmic vacuolization, we infected WT BMDMs in a 12-well
   plate (1 M per well) with different MOIs of BCV (100–400 MOIs) up to
   40 h. BCV-infected BMDMs were examined by microscopy, and cytoplasmic
   vacuolization was evaluated according to the number and size of
   vacuoles using ImageJ software. BCV at 400 MOI can induce more
   cytoplasmic vacuolization and large vacuoles at the early
   post-infection time period compared with lower MOIs. For comparing
   experiments, WT, inhibitor-treated, and siRNA-transfected BMDMs were
   infected with BC and BCV (400 MOI) as indicated times. The treated and
   control cells were analyzed for cytoplasmic vacuolization under a
   microscope and lysed for RNA and protein analysis.

Bacterial killing assay

   BMDMs were infected with BC and BCV with MOIs of 10–20 for 2 h and
   treated with antibiotics; cells were washed twice and cultured in fresh
   media (DMEM/F-12 and 10% FBS). After 22 h, the supernatant
   (extracellular bacteria) and BMDMs lysed in PBS (intracellular
   bacteria) were serially diluted, plated onto Columbia Blood agar
   plates, and incubated for 96 h for CFU enumeration.

Bacterial infection of mice

   BC and BCV bacterial strains were cultured on Columbia blood agar
   plates for 96 h, then growing in BACTEC™ Lytic media (BD, 442021) with
   shaking (220 rpm) for 16 h at 37 °C. Eight- to ten-week-old and
   gender-matched wild-type mice were infected intranasally with BC and
   BCV (4.0 × 10^8 CFUs per mouse). Mice were weighed and monitored on day
   0 and day 1. Mice were euthanized on day 1 after infection, and lungs
   were harvested to determine the bacterial burden and cytokine
   expression.

Immunoblot analysis and antibodies

   Samples were separated by 12% SDS-PAGE, followed by electrophoretic
   transfer to polyvinylidene fluoride membranes, and membranes were
   blocked and then incubated with primary antibodies. The following
   primary antibodies were used: anti-caspase-3 (CST, 9662); anti-cleaved
   caspase-3 (CST, 9661); anti-SLC9A9 (Proteintech, 66577-1-Ig);
   anti-caspase-1 (AdipoGen, AG-20B-0042); anti-MLKL (Abcepta, AP14272B);
   anti-p-MLKL (CST, 37333); anti-RIP3 (CST, 95702S); anti-p-RIP3 (CST,
   91702S); anti-p-PI3K (CST, 4228); anti-p-S6 (CST, 4856); anti-Cas9
   (abcam, ab191468); anti-Flag (Sigma, F3165); anti-V5 (CST, 13202) and
   anti-GAPDH (CST, 5174). HRP-labeled anti-rabbit (CST, 7074), anti-mouse
   (CST, 7076), or anti-goat (Santa Cruz, sc-2006) was used as the
   secondary antibody.

Immunofluorescence staining and microscopy

   For LAMP1, RAB7, SLC9A9, GST, EEA1, RAB5, and cleaved caspase-3
   immunostaining, treated and untreated BMDMs were fixed in 4%
   paraformaldehyde for 15 min at room temperature. Cells were washed with
   PBS and blocked in 1× ELISA buffer with 0.1% saponin for 1 h. Cells
   were stained with anti-LAMP1 (Invitrogen, 14-1071-85); anti-RAB7 (CST,
   9367); anti-SLC9A9 (Proteintech, 66577-1-Ig), anti-GST (Proteintech,
   10000-0-AP), anti-EEA1 (CST,3288), anti-RAB5 (CST, 3574) or
   anti-cleaved caspase-3 (CST, 9661) — all at 1:300 to 1:500 dilution,
   overnight at 4 °C. Cells were washed, stained with a
   fluorescence-conjugated secondary antibody (Invitrogen, A-11008, Alexa
   Fluor™ 488, Goat anti-Rabbit; Invitrogen, A-21422, Alexa Fluor™ 555,
   Goat anti-Mouse; Invitrogen, A-11077, Alexa Fluor™568, Goat anti-Rat)
   at 1:300 dilution for 40 min at 37 °C, and mounted using a mounting
   medium (Vector Laboratories, H-1200). Lucifer Yellow was ordered from
   Aladdin (L131282). Cells were observed on the ZEISS-LSM880 and ANDOR
   High-speed confocal microscope, and image acquisition and data analysis
   were performed using ZEN black_2-3SP1, ZEN blue 2.6, and Imaris
   software.

Live-cell imaging for cell death

   BMDMs (0.5 × 10^6 cells/well) were seeded in 24-well plates. BMDMs were
   infected with BC and BCV and stained with propidium iodide (PI; Life
   Technologies, P3566) and Hoechst (Beyotime Biotechnology, C1029)
   according to the manufacturer’s instructions. The plate was scanned,
   and images were acquired in real-time every 2 h from 0 to 24 h
   post-treatment by Opera Phenix High Content Screening System by
   PerkinElmer, Inc. Hoechst staining indicates the total number of cells;
   PI-positive dead cells are marked with a red mask for visualization.
   The image analysis, masking, and quantification of dead cells were done
   using the Harmony software package supplied with the system.

RNA-seq and data analysis

   Total RNA was extracted from BC- and BCV-infected and uninfected WT
   BMDMs cultured in L929 cell-conditioned DMEM/F-12 media and subjected
   to commercial RNA-seq analysis (Novogene). Each transcript’s RPKM
   values (reads per kilobase of transcript per million reads mapped) were
   calculated. Differential expression analysis between comparative groups
   was conducted using DESeq2 software (version 1.42.1). Differentially
   Expressed Genes (DEGs) were selected based on the criteria of
   |log2(Fold change)| ≥ 1 and a P-value less than 0.05. KEGG pathway
   enrichment analysis of the DEGs was performed using clusterProfiler
   software (version 4.10.1).

   BC and BCV were cultured on Columbia blood agar plates
   (Autobio,0004763) for 96 h, and the bacterial RNA was extracted for
   RNA-seq analysis (LC-BIO Biotech). Differential expression analysis was
   performed with R package edgeR (version 4.2.0). DEGs were selected
   based on a logarithmic fold change greater than two and a false
   discovery rate (FDR) below 0.05. KEGG pathway analyses were conducted
   using KOBAS to understand the functions of the DEGs. DEGs were
   considered significantly enriched if their Bonferroni-corrected P-value
   was less than 0.05. The heatmap was plotted using the R package
   pheatmap version 1.0.12.

Library preparation, genome sequencing, and assembly

   The whole genome of BC and BCV were sequenced using PacBio Sequel
   platform and Illumina NovaSeq PE150 at the Beijing Novogene
   Bioinformatics Technology Co., Ltd. Briefly, the DNA sample was
   fragmented by sonication to a size of 350 bp, then DNA fragments were
   end-polished, A-tailed, and ligated with the full-length adapter for
   Illumina sequencing with further PCR amplification. At last, PCR
   products were purified (AMPure XP system) and libraries were analyzed
   for size distribution by Agilent2100 Bioanalyzer and quantified using
   real-time PCR. The low-quality reads were filtered (less than 500 bp)
   to improve analysis accuracy and obtain clean data. The long reads were
   selected by SMRT portal (more than 6000 bp) as the seed sequence, and
   the other shorter reads were aligned to the seed sequence by Blasr. The
   SMRT Link software with Illumina data corrected the results of the
   preliminary assembly. Cyclization was confirmed according to the
   overlap between the head and the tail, and the initiation site was
   corrected by the blast with the DNAa database. The National Center for
   Biotechnology Information (NCBI) did the genome component prediction
   and gene annotation. The genome sequence dataset has been deposited in
   GenBank under accession no. [286]CP029149 and [287]CP114055. The
   genomic alignment between two BC and BCV was conducted using SnapGene,
   and scalable circular genome maps were generated using the Java package
   CGView to display the alignment of the BCV genome with the BC genome.

BC prevalence analysis in clinical patients

   The metagenomic sequencing data of clinical patients used in this study
   after removing ambient human DNA (150 nt pair-end reads) was obtained
   from the NCBI public database under accession code PRJNA1057503. The
   downloaded dataset was subjected to quality control, and low-quality
   and adapter sequences were removed using TrimGalore (version 0.6.6),
   and the high-quality sequences were mapped to the Bergeyella cardium
   genome sequence (GenBank [288]CP029149) using BLAT. Sequences that
   mapped to the reference genome with an alignment rate of ≥ 90% were
   used for further analysis. Sequences specially mapped to Bergeyella
   cardium genome and not any other species were recognized as potential
   unique reads of Bergeyella cardium by using the kraken2_nt_20230502 and
   krakenuniq_Standard (20220616) databases^[289]84,[290]85. Finally, the
   potential Bergeyella cardium-unique sequences were verified using
   BLASTn against the NCBI nt Database, and the specimens with Bergeyella
   cardium-unique reads above three were classified as Bergeyella
   cardium-positive specimens.

MS analysis

   The supernatants from BCV- and BC-infected BMDMs, and OMVs derived from
   BCV were performed using mass spectrometry (MS) analysis by Applied
   Protein Technology (Shanghai). Briefly, protein digestion was performed
   by trypsin. The digest peptides of each sample were desalted on C18
   Cartridges, concentrated by vacuum centrifugation and reconstituted in
   40 µL of 0.1% (v/v) formic acid. MS analysis was performed on a timsTOF
   Pro mass spectrometer (Bruker) coupled to Nanoelute (Bruker Daltonics)
   for 45 min. The peptides were loaded on a C18-reversed phase analytical
   column in buffer A (0.1% formic acid) and separated with a linear
   gradient of buffer B (99.9% acetonitrile and 0.1% formic acid) at a
   flow rate of 300 nL/min. The mass spectrometer was operated in positive
   ion mode. The MS raw data for each sample were combined and searched
   using the MaxQuant software for identification and quantification
   analysis. Proteins identified in the supernatants from BCV- and
   BC-infected BMDMs are listed in Supplementary Table [291]S3. Proteins
   identified in the OMVs derived from BCV are listed in Supplementary
   Table [292]S5.

Plasmid construction, transfection and co-IP experiments

   The full length of Slc9a9 was amplified from a mouse cDNA library and
   subcloned into the pCDH vector. GFP was amplified and subcloned into
   pCDH-SLC9A9 vector to make GFP-SLC9A9 fusion plasmid. Lipocalin,
   β-barrel, ADSL, and PorV genes were amplified from the DNA of BCV and
   subcloned into the pGEX-6P-2 vector. The full length Vps11, Vps16,
   Vps18, Vps33a, Vps33b, Vps39, Vps41, Vps35, Rab5a, Rab5b, Rab7, Stx2,
   Stx11, Stx12, Stx18, Stxbp1, Stxbp2, Stxbp3, Snap29, Snap47, Snapin,
   Vamp2, Vamp3, Vamp4, Vamp8, Sec22b, Bnip1, Bet1, Bet1l, Esyt1, and Ehd1
   were amplified from a mouse cDNA library and subcloned into pCDNA3.1
   vector. All plasmids were confirmed by DNA sequencing. The primer
   sequences for vector construction are listed in Supplementary Table
   [293]S7. Lipofectamine 3000 reagents were used for the transient
   transfection of plasmids into HEK293T cells.

   For IP, whole HEK293T cells collected 36 h after transfection were
   lysed in IP buffer composed of 50 mM Tris-HCl (pH 7.4), 1 mM EDTA,
   150 mM NaCl, 1% Triton, and protease/phosphatase inhibitor cocktails
   (BioTools). After centrifugation, supernatants were collected and
   incubated with protein A/G Plus–Agrose (Santa Cruz Biotechnology,
   sc-2003) and 3 μg of the corresponding antibodies for 12 h at 4 °C,
   followed by 5 rounds of washing with IP buffer. Immunoprecipitated
   components were eluted in the SDS loading buffer. For immunoblot
   analysis, immunoprecipitates and input lysates were separated by
   SDS-PAGE, followed by transfer onto PVDF membranes, and detected by
   specific antibodies.

Lentivirus production and infection

   The viral particles were prepared by transfecting HEK293T cells with
   GFP and fusion GFP-SLC9A9 plasmids in combination with packaging
   vectors. Twelve hours later, the media were replaced with fresh
   complete DMEM. Viral supernatant was harvested and passed through 0.45
   μm syringe filter 48 h and 72 h after transfection. To establish stably
   transduced cells, WT and Slc9a9^−^/^− bone marrows were infected twice
   with filtered lentiviral supernatant at an MOI of 1.0 in day 3 and day
   4 in the presence of polybrene (8 μg/mL) as previously
   described^[294]86. The transduced cells were cultured in fresh media
   for further treatments and analysis.

Lysosome isolation and treatment

   Lysosomes from BCV-infected BMDMs (10 M, 18 h) were isolated using a
   lysosome enrichment kit with cultured cell sonication and density
   gradient centrifugation, followed by lysosome precipitation according
   to the manufacturer’s manual (PI89839; Thermo Fisher Scientific). The
   purity of isolated lysosomes was determined by detecting the lysosomal
   membrane protein LAMP1, cathepsin B, nuclear protein lamin B, and GAPDH
   using immunoblotting analysis as previously described^[295]38. The
   concentration of lysosome was quantified using Pierce BCA protein assay
   kit (23227).

Bone marrow transplantation

   The mouse model of bone marrow transplantation was established using WT
   (CD45.1^+) and Slc9a9^−^/^− (CD45.2^+) mice as the donors and the
   recipients, which were similar in age and within 8–12 weeks old as
   previously described^[296]87. Six to eight hours prior to
   transplantation, the recipient mice were irradiated with a dosage of
   10 Gy. For transplantation, 10 million bone marrow cells were injected
   into the tail veins of the recipient mice. Since the second day of
   transplantation, recipients were fed antibiotics for two weeks. Six
   weeks after transplantation, bone marrow chimeric mice were used for
   infection experiments.

Flow cytometry analysis

   For flow cytometric analysis of CD45.1^+ cells and CD45.2^+ cells,
   cells prepared from spleens, bone marrow and peripheral blood were
   stained using a subset of antibodies (Biolegend, QA18A43, QA18A15,
   1:300). Cell preparation and staining with cell surface CD45.1 and
   CD45.2 were carried out as described previously^[297]87, and cells were
   analyzed on a BD LSR Fortessa Cell Analyzer (BD Biosciences).

Expression, purification, and transfection of recombinant proteins

   The control PGEX-6P-2 vector and GST-fused protein constructs
   PGEX-6P-2-PorV, PGEX-6P-2-Lipocalin, PGEX-6P-2-β-barrel, and
   PGEX-6P-2-ADSL were transformed into the Escherichia coli strain
   Rosetta (DE3) pLysS for expression. The expressed proteins were loaded
   onto Glutathione Agarose (MCE) and Mono S^TM 5/50 GL (Cytiva, USA)
   pre-equilibrated with the lysis buffer for purification. Endotoxin was
   removed using Pierce High-Capacity Endotoxin Removal Resin (Thermo,
   88274) according to the manufacturer’s protocol. GST and GST-fused
   proteins were transfected into the cells using Xfect™ Protein
   Transfection Reagent (TaKaRa, 631324) following the manufacturer’s
   instructions.

Transmission electron microscopy (TEM) and scanning electron microscopy (SEM)

   BC- and BCV-infected BMDMs were fixed in 2% paraformaldehyde and 2.5%
   glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4) for 1 h at 37 °C.
   Cells were embedded and sectioned for TEM by Jindi Medical Technology
   (Jinan, Shandong). BC and BCV were cultured on Columbia blood agar
   plates for 96 h, then grown in BACTEC™ Lytic media (BD, 442021) with
   shaking (220 rpm) for 16 h at 37 °C. An equal number of bacteria were
   precipitated and fixed in 2.5% glutaraldehyde and phosphate buffer, and
   TEM and SEM analyzed sectioned samples at Servicebio (Wuhan, Hubei).

Preparation of tissue sample for HE staining

   The superior lobes of the right lungs were fixed in 10% formalin, and
   5-μm sections were stained with hematoxylin and eosin (H&E) and
   examined with a microscope. The severity of lung disease was scored on
   the basis of the presence of inflammation by a pathologist blinded to
   the experimental groups according to the grade standard (0 = absent; 1
   = rare, minimal; 2 = scattered mild; 3 = multifocal, moderate; 4 =
   extensive, marked; 5 = severe).

Real-time qRT-PCR

   Total RNA was isolated from cells and tissues using TRIzol reagent
   (Invitrogen, Thermo Fisher Scientific). cDNA was reverse transcribed
   using M-MLV reverse transcriptase (Promega). Real-time qRT-PCR was
   performed on the Roche LightCycler 96 Real-Time Detection System. Hprt
   gene (hypoxanthine guanine phosphoribosyltransferase) is an internal
   control for data analysis. The primer sequences are listed in
   Supplementary Table [298]S7.

Lactate dehydrogenase (LDH) release assay

   Cell culture supernatants were collected at the indicated times, and
   lactate dehydrogenase activity was measured using the Promega
   cytotoxicity kit (G1781) according to the manufacturer’s protocols.

ELISA analysis

   Sera from uninfected and infected mice were analyzed for cytokine
   release using ELISA MAX Standard (BioLegend) following the
   manufacturer’s instructions.

Protein structure prediction and analysis

   The protein structures of Lipocalin, β-barrel, PorV, and ADSL were
   predicted by the online server trRosetta
   ([299]https://yanglab.qd.sdu.edu.cn/trRosetta/)^[300]29,[301]88,[302]89
   , and visualized by UCSF ChimeraX^[303]90. Briefly, the protein
   sequence was used as the input sequence for structure prediction,
   performed with default parameter settings. trRosetta provided five
   models as results for each prediction. The highest-ranked model of each
   prediction, model 1, was used for further analysis. The confidence of
   the overall structure prediction is reflected by the TM-score. A
   TM-score (0–1) above 0.5 usually indicates a model with correct
   topology. The confidence of the prediction at each residue is indicated
   by per-residue LDDT scores, ranging from 0 to 100, located at the
   B-factor column of the PDB file of the structure. For structure
   analysis and comparison of BC and BCV PorV, structural figures were
   made using Open-Source PyMOL (Schrödinger). BCV and BC PorV models are
   colored by the ‘b-factor’ ‘spectrum’, reflecting the per-residue LDDT
   scores. The structure model of BC PorV was superposed onto BCV PorV
   using the ‘super’ command in PyMOL, resulting in an RMSD of 0.489 over
   1961 atoms.

Statistics

   Data are presented as the mean ± SEM. Statistical analyses were
   performed using two-way ANOVA, two-tailed Student’s t- and log-rank
   tests. P-values of 0.05 or less were considered significant.

Study approval

   The study was approved by the Ethics Committee of Scientific Research
   of Shandong University (ECSBMSSDU2021-2-171 and ECSBMSSDU2021-1-086),
   Jinan, Shandong Province, China.

Supplementary information

   [304]Supplementary figures^ (13.7MB, pdf)
   [305]Supplementary Video S1^ (9.4MB, mp4)
   [306]Supplementary Table S1^ (36.9KB, xlsx)
   [307]Supplementary Table S2^ (313.4KB, xlsx)
   [308]Supplementary Table S3^ (71.3KB, xlsx)
   [309]Supplementary Table S4^ (3.8MB, xlsx)
   [310]Supplementary Table S5^ (149.6KB, xlsx)
   [311]Supplementary Table S6^ (263.4KB, xlsx)
   [312]Supplementary Table S7^ (16.5KB, xlsx)

Acknowledgements