Abstract

   Neutrophils are highly abundant in the oral mucosal tissues, and their
   balanced activation and clearance are essential for immune homeostasis.
   Here, we demonstrate that neutrophils infected with the bacterial
   pathogen Porphyromonas gingivalis (Pg) are captured alive by
   macrophages in a manner that bypasses all known receptor-ligand
   interactions involved in the phagocytosis of either live or dead cells.
   Mechanistically, upon interaction with Pg, or its protease RgpB
   (gingipains), live neutrophils undergo rapid remodeling of their
   proteomes, generating neoepitopes. N-terminomics-based proteomic
   profiling identified multiple RgpB cleavage sites on several
   azurophilic granule proteins that are translocated to the surface of
   live neutrophils via low-level degranulation and activate macrophage
   α[M]β[2] integrin receptors, thus mediating internalization of
   non-apoptotic neutrophils within macrophage phagosomes. Macrophages
   with entrapped live neutrophils exhibit phenotypic and transcriptional
   reprogramming, consistent with inflammatory outcomes in vitro and in
   vivo. In contrast to the immunosuppressive outcomes associated with
   efferocytosis of apoptotic neutrophils, live neutrophil entrapment
   failed to fully activate several catabolic and metabolic processes and
   exhibited a defective activation of PPAR-γ mediated pro-resolution
   pathways, thereby promoting bacterial persistence and hindering the
   resolution of inflammation. Thus, our data demonstrate a novel immune
   subversion strategy unique to Pg and reveal a previously unknown mode
   of live neutrophil sequestration into macrophages during an infection.

   graphic file with name 41419_2025_7808_Figa_HTML.jpg

   Subject terms: Inflammation, Immune evasion

Introduction

   Neutrophils are the most abundant leukocytes in the body and play
   important roles in immune surveillance, antimicrobial immunity, and
   tissue homeostasis. Upon recruitment to sites of infection, they
   rapidly phagocytose and kill invading bacteria by degranulating and
   releasing antimicrobial mediators, thus playing a critical role in
   sterilizing immunity. Most recruited neutrophils eventually undergo
   apoptosis, a process characterized by distinct molecular alterations to
   the cell surface proteins, sugars, receptor repertoire, and plasma
   membrane lipid composition that collectively serve as “eat me” signals
   and facilitate their clearance via efferocytosis within tissues. For
   example, the exofacial localization of phosphatidylserine (PS) is a
   well-conserved, apoptosis-associated signal recognized directly or
   indirectly by many efferocytic receptors that regulate the phagocytic
   uptake of dying neutrophils within macrophages, termed efferocytosis
   [[44]1, [45]2]. While most infected, aged, or apoptotic neutrophils
   (ANs) are cleared via efferocytosis [[46]3] or phagoptosis [[47]4],
   several studies now show that the expression of apoptotic markers does
   not always predicate neutrophil engulfment by another cell, and under
   certain circumstances, live neutrophils might also be internalized by
   other live cells [[48]5–[49]8].

   Viable neutrophils appear to invade and transit through endothelial
   cells during transcellular migration by forming a transient
   transcellular channel between the apical and basal membranes of
   endothelial cells within blood vessels to exit circulation [[50]6].
   Emperipolesis is another example in which live neutrophils invade
   megakaryocytes and either remain within vacuolar structures called
   ‘emperisomes’ or enter the megakaryocyte cytosol, eventually egressing
   and carrying the megakaryocyte membrane [[51]5, [52]7]. Transcellular
   migration and emperipolesis is mediated in part via ICAM-1 and β[2]
   integrin interactions, transcellular cup formation, and actin
   remodeling [[53]5, [54]6]. Recently, using a model of allergic
   inflammation, Mihlan et al. showed that live neutrophils can also be
   entrapped by degranulating mast cells (MCs) in a leukotriene B4 (LTB4)
   dependent manner, resulting in a cell-in-cell structure [[55]8]. Here,
   we show that live neutrophil ingestion can also occur under pathogenic
   conditions, such as bacterial infections, creating an environment
   conducive for bacterial survival.

   Porphyromonas gingivalis (Pg) is a human-adapted bacterial colonizer of
   the oral mucosal tissues and is etiologically associated with
   periodontal inflammation and other chronic inflammatory or degenerative
   diseases [[56]9]. Its pathogenic potential is primarily linked to the
   production of arginine-specific (HRgpA and RgpB) and lysine-specific
   (Kgp) proteases called gingipains that cleave proteins at the arginine
   (Arg-X-aa) and lysine (Lys-Xaa) peptide bonds, respectively [[57]10].
   Gingipains cleave a large number of host proteins and are particularly
   adept at manipulating neutrophil antimicrobial and inflammatory
   responses [[58]1, [59]11, [60]12]. Interestingly, despite nearly
   identical catalytic activity, HRgpA and RgpB differentirely impact
   neutrophil responses and viability. While HRgpA induces robust NETosis
   and apoptosis, RgpB minimally impacts neutrophil viability or
   lifecycle, but instead dampens their antimicrobial capacity and induces
   the cleavage of non-phagocytic signals (CD31) of neutrophil surface
   macrophages [[61]11, [62]13, [63]14]. However, the full extent of
   RgpB-mediated modifications to the neutrophil proteome and its impact
   on tissue inflammatory responses or Pg fitness with the host have not
   been explored. Here, we show that through the activity of RgpB, Pg
   remodels the proteomes of live neutrophils, generating non-canonical
   uptake signals that engage α[M]β[2] integrin receptors on macrophages
   and their subsequent sequestration within macrophages. The entrapment
   of live neutrophils within macrophages dysregulates pro-resolution
   pathways and favors bacterial persistence in vivo.

RESULTS

RgpB protease is necessary and sufficient for the entrapment of live
neutrophils within macrophages

   We previously showed that purified RgpB can cause a modest cleavage of
   antiphagocytic signals on the surface of live neutrophils, causing
   their ingestion by a subset of human monocyte-derived macrophages
   [[64]14]. However, since Pg produces several virulence factors, we
   first determined whether gingipains alone were necessary and sufficient
   for the entrapment of live neutrophils upon infection. Human
   neutrophils infected with Pg mutants lacking either the major fimbriae
   (ΔfimA), peptidylarginine deaminase (Δppad), or gingipains (ΔrgpA,
   Δrgpb, and Δkgp triple mutant, abbreviated as ΔKRAB) for 1 h did not
   externalize PS (Fig. S[65]1A), an ‘eat me signal’ that mediates
   efferocytosis or phagocytosis of dying and distressed cells by a large
   number of efferocytic receptors. Next, we measured whether macrophages
   ingested Pg-infected neutrophils despite the lack of any obvious “eat
   me” signals using well-established in vitro and in vivo efferocytic
   assays [[66]15, [67]16].

   Murine peritoneal exudate macrophages (PEMs) were co-cultured with
   neutrophils and uptake was determined by histochemical staining for
   myeloperoxidase (MPO), which is highly expressed in neutrophils and
   absent in macrophages [[68]15] (Fig. S[69]1B). Despite the absence of
   PS, neutrophils infected with Pg (WT, ΔfimA, and Δppad) were rapidly
   internalized by macrophages with efficiency comparable to the uptake of
   PS-expressing ANs. Interestingly, the phagocytosis of ΔKRAB-infected
   neutrophils was significantly dampened, indicating that
   gingipain-mediated proteolytic processing was essential for the
   sequestration of live neutrophils within macrophages (Fig. S[70]1B, C).
   RgpB, unlike HRgpA, does not induce apoptosis or NETosis in neutrophils
   upon prolonged treatment but instead facilitates their live uptake by
   macrophages [[71]13, [72]14]. Selective inhibition of RgpB activity on
   both live Pg and purified RgpB enzyme using a highly specific
   non-reversible inhibitor, D-Phe-Phe-Arg-chloromethylketone (FFR-CMK),
   significantly dampened uptake (Fig. S[73]1D, E), confirming that the
   proteolytic activity of RgpB was necessary and sufficient for the
   entrapment of live Pg-infected neutrophils within macrophages. While
   the gingipain deficient strains were instrumental in understanding the
   role of gingipains in live neutrophil entrapment, they are rapidly
   cleared in vivo, making direct comparisons of RgpB mutant and WT
   strains improbable in vivo. Thus, we worked with purified RgpB for the
   rest of our experiments and confirmed key findings with Pg infected
   neutrophils. This approach is biologically relevant as gingipains are
   actively secreted and found at sites distant from the oral cavity, such
   as joints, brain, liver, lungs, and blood in humans [[74]17].

   We developed two separate model systems to track neutrophil entrapment
   and its role in modulating macrophage responses in vitro and in vivo.
   The dual-species model, where human neutrophils are fed to murine
   peritoneal macrophages (PEMs)[[75]15, [76]16], allows for the use of
   species-specific reagents to exclude contaminating signals (passenger
   transcripts, proteins, cellular markers) from internalized live or
   apoptotic cells and was used to delineate macrophage transcriptional
   response to live neutrophil entrapment. This approach was complemented
   by using a physiologically relevant single-species model (murine
   neutrophils were fed to murine macrophages) in select experiments.
   Similar to infection with live bacteria, human neutrophils exposed to
   purified RgpB (gLN) did not induce PS expression (Fig. [77]1A, B) and
   were rapidly internalized into early phagocytic compartments by PEMs
   (Fig. [78]1C). The efficiency of uptake was comparable to that of ANs
   as determined by two independent in vitro efferocytosis assays (Fig.
   [79]1D–G). Of note, MPO histochemical staining and flow-based uptake
   assays were highly concordant and showed ~30% uptake of gLNs and ANs
   within 2 h of co-culture with macrophages.

Fig. 1. Purified RgpB mediates entrapment of live neutrophils.

   [80]Fig. 1
   [81]Open in a new tab

   A, B Purified human neutrophils were incubated with P. gingivalis
   protease RgpB (300 nM) for 1 h or with cycloheximide (50 μg/mL) for
   18 h to induce apoptosis. After treatments, phosphatidylserine (PS)
   externalization was assessed by Annexin V staining of apoptotic
   neutrophils (ANs; blue histogram) or RgpB-treated live neutrophils
   (gLNs; red histograms) from 3 independent donors. Relative uptake of
   ANs and gLNs by various macrophage types was determined in vitro and in
   vivo by the following assays: C The phagocytosis of PKH26-green labeled
   gLNs by murine peritoneal exudate macrophages (PEM) after 2 h of
   co-incubation was determined by confocal microscopy. PEM membranes were
   labeled with wheat germ agglutinin (red). D, E To quantitatively assess
   relative uptake rates in vitro, ANs or gLNs were incubated with PEMs
   for 2 h, and uningested neutrophils were removed by washing with PBS.
   Phagocytosing macrophages were determined by histochemical staining for
   myeloperoxidase (MPO), a protein selectively expressed in neutrophils.
   D Representative images of MPO^+ PEMs at 100x magnification are shown.
   E % uptake was determined by counting phagocytosing or MPO^+ (brown
   staining) PEMs and expressed as a percentage value over total PEMs. At
   least 300 macrophages were counted (blinded) for each replicate, and
   data from 3 independent experiments is shownas mean ± SD, and data
   points indicate biological replicates; ****p < 0.0001 (unpaired
   t-test). F, G For flow-based in vitro quantification of phagocytosis,
   PEMs were incubated with PKH26-labeled human gLN or AN. Macrophages
   with attached or incompletely ingested neutrophils were excluded by
   gating out (hCD45^+ F4/80^+, PKH26^+). The relative abundance of
   phagocytosing macrophages (F4/80^+, PKH26^+) is shown as mean ± SD, and
   data points indicate biological replicates. Statistical significance
   was calculated using an unpaired t-test. H The relative phagocytic
   rates of ANs or gLNs were also determined in murine bone marrow-derived
   macrophages (BMDM) and RAW264.7 macrophages by MPO-based phagocytosis
   assay described in (D, E). Averaged data (mean ± SD) from three
   independent experiments is shown. Data points indicate biological
   replicates and statistical significance was calculated by two-way ANOVA
   and Šídák correction (I, J). For in vivo uptake assay, wildtype (WT)
   mice were injected intraperitoneally (i.p.) with 10^7 human neutrophils
   (AN or gLN). 4 h after injection, the peritoneal cells were collected
   by lavage, and phagocytosing macrophages (hCD45^-, F4/80^+, hMPO^+
   population) were determined by flow cytometry. Data are shown as
   mean ± SD, and data points indicate biological replicates. The
   illustration above, data panels in (I), was createdusing Biorender.com.

   To rule out the impact of tissue origin or priming status of
   macrophages in live neutrophil entrapment, we co-cultured gLNs and ANs
   with naïve bone marrow-derived macrophages (BMDMs) and the RAW264.7
   macrophage cell line. Our data show comparable uptake rates of gLN and
   ANs in these cell types (Fig.[82]1H). We injected human gLNs or ANs
   into the inflamed peritoneal cavities of mice and determined uptake by
   flow cytometry using a human-specific anti-MPO antibody. Concordant
   with our in vitro data, gLN entrapment was also observed in vivo and
   occurred with the same efficiency as the uptake of ANs (Fig. [83]1I,
   J). Similar uptake rates were also observed in the single-species
   model, where gingipain-treated murine bone marrow neutrophils (g-BMN)
   were entrapped in vitro and in vivo at similar efficiency as apoptotic
   murine neutrophils, confirming that our observations are independent of
   differences in the species of origin (Fig. S[84]2). We also if
   determined RgpB exposure induced the uptake of other cell types. PEMs
   were co-cultured with PKH-labeled apoptotic or RgpB-treated live
   thymocytes (apop-thym and gL-thym, respectively) and % uptake
   determined by flow cytometry. Interestingly, we observed that
   gL-thymocytes (Fig. S[85]3A–D) were not internalized by macrophages,
   suggesting that RgpB was selectively modifying neutrophil-specific
   protein(s), which might be essential for live neutrophil entrapment.
   Altogether, these data show that Pg, relies on the proteolytic activity
   of RgpB to mediate the entrapment of live neutrophils within
   macrophages.

RgpB cleaves and mobilizes neutrophil granule proteins to the surface

   Since RgpB’s proteolytic activity was central to the entrapment of live
   neutrophils (Fig. S[86]1), we hypothesized that RgpB facilitated the
   generation of neo-epitopes that engaged macrophage phagocytic
   receptors, to drive live neutrophil phagocytosis. To identify RgpB
   protein substrates within live neutrophils, we resorted to the terminal
   amine isotopic labeling of substrates (TAILs) approach, followed by
   liquid chromatography and tandem mass spectrometry (LC-MS/MS) based
   identification. TAILS is a quantitative proteomics approach that
   facilitates the unbiased identification of protease substrates and
   cleavage sites in biological samples [[87]18]. In this approach, all
   naturally occurring N termini and neo-N-termini generated by RgpB are
   blocked by differential isotopic labeling. Briefly, neutrophils from
   control and RgpB exposed samples were isotopically labeled with light
   formaldehyde (+28 Da dimethylation) or heavy formaldehyde (+34 Da
   dimethylation), respectively. Pooled samples were then trypsinized to
   reduce complexity, and then N-termini were enriched by incubation with
   dendritic polyglycerol aldehyde TAILS polymer that removes unlabeled
   proteins as previously described in [[88]19] and illustrated in (Fig.
   [89]2A). The global proteomes were compared by a shotgun
   (pre-enrichment TAILs) proteomic analysis, while TAILS focused on the
   proteomic analysis of N-termini in all proteins [[90]18, [91]20]. After
   sample acquisition and LC-MS/MS analysis, data were analyzed using
   MaxQuant [[92]21] at 1% FDR.

Fig. 2. RgpB proteolytically modifies the intracellular neutrophil proteome.

   [93]Fig. 2
   [94]Open in a new tab

   Live human neutrophils were incubated with active 300 nM RgpB (gLN) for
   buffer alone (Buffer) and subject to terminal amine isotopic labeling
   of substrates (TAILS) mass spectrometry (TAILS-MS). A Schematic
   depicting TAILS-MS workflow (generated using Biorender.com). Briefly,
   after blocking primary amines (not shown), samples underwent isotopic
   labeling with heavy (deuterated) or light formaldehyde and digestion
   with trypsin. After trypsin digestion, a fraction of each sample was
   subject to pre-enrichment TAILS (shotgun analysis or pre-TAILS). The
   rest underwent removal of tryptic N-terminal peptides using a high
   molecular weight dendritic polyglycerol aldehyde polymer, leaving the
   naturally blocked or labeled mature and neo-N-termini unbound via
   negative selection (flow-through). TAILS peptides were recovered by
   size exclusion filtration and analyzed by liquid chromatography coupled
   with tandem mass spectrometry (LC-MS/MS). B The numbers of unique and
   shared peptides between TAILS and preTAILS analysis are shown. C The
   numbers of statistically changing peptides using an interquartile
   boxplot analysis between the pre-TAILS samples are shown. D Left:
   Distribution of N-terminal peptides in the TAILS enrichment. Middle,
   statistically changing peptides using an interquartile boxplot
   analysis. Right, Distribution of post-translational peptide
   modifications as analyzed using TopFINDER. For a complete list of
   peptides in figures (B, C), also see Supplementary Table [95]1. E Left,
   peptide sequence profiles of significantly elevated neo-N-terminal
   peptides in RgpB treated neutrophils identified in the TAILS analysis
   using IceLogo. Right, Cleavage sites identified as RgpB-treated
   neutrophils are depicted as heatmaps from P6 to P6′ residues. Green:
   Upregulated. Red: Downregulated. F Left, peptide sequence profiles of
   significantly elevated neo-N-terminal peptides in untreated (buffer)
   neutrophils identified in the TAILS analysis using IceLogo.
   Significantly (p < 0.05) overrepresented amino acids are shown above
   the x-axis, while the underrepresented residues are shown below the
   x-axis. Statistical analysis was determined by a two-tailed unpaired
   Student’s t-test and adjusted for multiple comparisons. G Metascape
   analysis of the TAILS data of different pathways between buffer- and
   RgpB-treated neutrophils is shown. H STRING-db analysis of RgpB cleaved
   substrates is shown. An enrichment was detected for neutrophil
   degranulation (blue), signaling by interleukins (red) and antimicrobial
   peptides (green). The dotted circle shows an enrichment of neutrophil
   azurophilic granule proteins (cathepsin G (CTSG); neutrophil elastase
   (ELANE); myeloperoxidase (MPO) and azurocidin (AZU1)) as RgpB
   substrates.

   Shotgun (pre-TAILS) analysis of RgpB treated or untreated neutrophils
   from 3 donors yielded 1441 unique peptides, and the TAILS analysis
   yielded 1930 unique peptides where 489 were identical to the
   pre-enrichment TAILS analysis (Supplementary Table [96]1; Fig. [97]2B).
   In the pre-TAILS data, we identified a significant change of 1.8% of
   peptides in the RgpB treated neutrophils and 2.3% in the untreated
   control (Fig. [98]2C). Next, we analyzed the N-terminal processing in
   the RgpB-treated samples. Protease-generated neo-N termini should only
   be present in the protease-exposed samples, and not in the untreated
   (buffer control) samples. Our TAILS analysis identified predominantly
   internal N-termini (93.6%), in addition to other proteoforms, including
   N-termini generated by signal peptide removal (4.9%) and alternative
   start sites (1.5%) (Fig. [99]2D). Next, we generated IceLogos to
   determine cleavage site preferences between RgpB treated and untreated