Microbe-Specific Unconventional T Cells Induce Human Neutrophil Differentiation into Antigen Cross-Presenting Cells

Neutrophils are the first cells that are recruited to sites of microbial infection. Although classically viewed as terminally differentiated cells, there is emerging evidence that neutrophils represent key components of the effector and regulatory arms of the innate and adaptive immune system (1–3). As such, neutrophils regulate the recruitment and function of various cell types and interact with immune and nonimmune cells. Intriguingly, neutrophils directly affect Ag-specific responses by facilitating monocyte differentiation and dendritic cell maturation, and by interacting with T cells and B cells (4–10). Murine neutrophils have been shown to present Ags to both CD4⁺ and CD8⁺ T cells (11–13), and to differentiate into neutrophil–dendritic cell hybrids in vitro and in vivo (14, 15). In humans, neutrophils with a phenotype consistent with a possible APC function, including expression of MHC class II, have been found in diverse inflammatory and infectious conditions (16–22). This notwithstanding, direct Ag presentation by neutrophils has to date not been demonstrated in patients, especially with respect to an induction of Ag-specific CD8⁺ T cell responses upon cross-presentation of exogenous proteins.

The physiological context underlying the differentiation of neutrophils into APCs and the implications for Ag-specific immune responses remain unclear. Unconventional T cells such as human γδ T cells, NKT cells, and mucosal-associated invariant T (MAIT) cells represent unique sentinel cells with a distinctive responsiveness to low m.w. compounds akin to pathogen and danger-associated molecular patterns (23–25). Such unconventional T cells represent a substantial proportion of all T cells in blood and mucosal epithelia, accumulate in inflamed tissues, and constitute an efficient immune surveillance network in inflammatory and infectious diseases as well as in tumorigenesis. Besides orchestrating local responses by engaging with other components of the inflammatory infiltrate (26–29), unconventional T cells are also ideally positioned in lymphoid tissues to interact with freshly recruited monocytes and neutrophils (30–32). We previously showed that human γδ T cells enhance the short-term survival of neutrophils but did not characterize these surviving neutrophils on a phenotypical and functional level (28). In this work, we studied the outcome of such a crosstalk of human neutrophils with both γδ T cells and MAIT cells in vitro and translated our findings to patients with severe sepsis. We demonstrate that neutrophils with APC-like features can be found in blood during acute infection, and that the phenotype and ex vivo function of circulating sepsis neutrophils was replicated in vitro upon priming of neutrophils by human γδ T cells and MAIT cells. Our findings thus provide a possible physiological context and propose a cellular mechanism for the local generation of neutrophils with APC functions, including their potential to cross-present soluble Ags to CD8⁺ T cells, in response to a broad range of microbial pathogens.

This study was approved by the South East Wales Local Ethics Committee under reference numbers 08/WSE04/17 and 10/WSE04/21 and conducted according to the principles expressed in the Declaration of Helsinki and under local ethical guidelines. Sampling of adult patients with sterile systemic inflammatory response syndrome (SIRS) or with acute sepsis (defined as patients with SIRS in conjunction with a proven or suspected infection) was carried out within the United Kingdom Clinical Research Network under study portfolio UKCRN ID 11231, “Cellular and Biochemical Investigations in Sepsis.” All study participants provided written informed consent for the collection of samples and their subsequent analysis. A waiver of consent system was used when patients were unable to provide prospective informed consent due to the nature of their critical illness or therapeutic sedation at the time of recruitment. In all cases, retrospective informed consent was sought as soon as the patient recovered and regained capacity. In cases in which a patient died before regaining capacity, the initial consultee’s approval would stand.

Sepsis patients had a proven infection as confirmed by positive culture of at least one relevant sample according to the local microbiology laboratory overseen by Public Health Wales, and developed at least three of the four following SIRS criteria over the previous 36 h: 1) temperature from any site >38°C or core <36°C; 2) heart rate of >90 beats/min (unless individual had a medical condition or was receiving treatment preventing tachycardia); 3) respiratory rate of >20 breaths/min, arterial PaCO₂ <32 mmHg, or mechanical ventilation for an acute process; and 4) total WBC >12,000 cells/mm³ or <4,000 cells/mm³ or differential WBC count showing >10% immature (band) forms (n = 37; age range 35–82 y, median 63 y; 51% female). Patients with sterile SIRS developed at least three of the four SIRS criteria but had no suspected or proven microbial infection (n = 14; age range 26–70 y, median 48 y; 21% female). All patients with sepsis or SIRS had at least one documented organ failure on recruitment to the study and were either mechanically ventilated, on inotropic support, or received acute renal replacement therapy. Healthy donors served as controls for the patient cohorts (n = 10; age range 31–68 y, median 59 y; 20% female). Individuals were excluded from the study if pregnant or breastfeeding; suffering from documented severe immune deficiency or severe liver failure; admitted postcardiac arrest; treated with high-dose steroids or immunosuppressant drugs for the last 6 mo; or unlikely to survive for the duration of the study period regardless of treatment.

Culture medium was RPMI 1640 medium supplemented with 2 mM l-glutamine, 1% sodium pyruvate, 50 μg/ml penicillin/streptomycin, and 10% FCS (Invitrogen). Synthetic (E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate (HMB-PP) was provided by H. Jomaa (Justus-Liebig University Giessen); synthetic 6,7-dimethyl-8-d-ribityllumazine (DMRL) was provided by B. Illarionov (Hamburg School of Food Science). Staphylococcus aureus toxic shock syndrome toxin-1 (TSST-1) was purchased from Toxin Technology; Mycobacterium tuberculosis purified protein derivate (PPD) was purchased from Statens Serum Institut (Copenhagen, Denmark). Salmonella abortus equi LPS, brefeldin A, and BSA-FITC were purchased from Sigma-Aldrich. Recombinant IFN-γ, TNF-α, and GM-CSF was purchased from Miltenyi Biotec. Human T-activator CD3/CD28 Dynabeads, CFSE, and 10-kDa dextran-FITC were purchased from Life Technologies.

The following mAbs were used for surface labeling: anti-CD3 (UCHT1, SK7, HIT3a), anti-CD4 (SK3, RPA-T4), anti-CD8 (SK1, HIT8a, RPA-T8), anti-CD11c (S-HCL-3), anti-CD14 (M5E2, MOP9), anti-CD15 (HI98), anti-CD16 (3G8), anti-CD25 (M-A251), anti-CD27 (M-T271), anti-CD31 (WM-59), anti-CD32 (FLI8.26), anti-CD45RO (UCHL1), anti-CD49d (9F10), anti-CD50 (TU41), anti-CD54 (HA58), anti-CD56 (B159), anti-CD62L (DREG-56), anti-CD64 (10.1), anti-CD69 (FN50), anti-CD70 (Ki24), anti-CD71 (M-A712), anti-CD72 (J4-112), anti-CD83 (HB15e), anti-CD86 (2331), anti-CD209 (DCN46), anti–HLA-DR (L243), anti–TCR-Vδ2 (B6.1), anti-CCR4 (1G1), anti-CCR5 (2D7), anti-CCR7 (3D12), and anti-CXCR3 (1C6) from BD Biosciences; anti–TCR-Vβ2 (MPB2D5), anti–TCR-Vγ9 (Immu360), and anti-CD40 (mAB89) from Beckman Coulter; anti-CD11a (HI111), anti-CD66b (G10F5), anti-CD154 (24-31), anti-CD161 (HP-3G10), anti–HLA-ABC (w6/32), and anti–TCR-Vα7.2 (3C10) from BioLegend; anti-CD11b (ICRF44), anti-CD14 (61D3), anti-CD19 (SJ25C1), anti-CD25 (BC96), anti-CD45RA (HI100), and anti-CD80 (2D10.4) from eBioscience; anti–HLA-A2 (BB7.2) from Serotec; and anti-CCR9 (248601) and anti-CCR10 (314305) from R&D Systems; together with appropriate isotype controls. Intracellular cytokines were detected using anti–IFN-γ (B27, BD Biosciences; 4S.B3, eBioscience) and anti–TNF-α (6401.1111, BD Biosciences; 188, Beckman Coulter). Blocking reagents used included anti-Vα7.2 (3C10; BioLegend); anti–TCR-Vγ9 (Immu360; Beckman Coulter); anti-TLR4 (HTA125; eBioscience); anti-CD277 (103.2; D. Olive, Université de la Méditerranée, Marseille, France); anti–MHC-related protein 1 (MR1) (26.5; T. Hansen, Washington University School of Medicine, St. Louis, MO); anti–IFN-γ (25718) and anti–GM-CSF (3209) (BioLegend); and soluble TNFR (sTNFR) p75-IgG1 fusion protein (etanercept/Enbrel; Amgen).

Total leukocytes from healthy donors and patients were isolated from heparinized blood by mixing with HetaSep (StemCell Technologies), followed by sedimentation of RBCs (Supplemental Fig. 1A). Neutrophils were purified from whole blood or Lymphoprep (Axis-Shield) separated granulocytes by HetaSep sedimentation, followed by negative selection using the EasySep neutrophil enrichment kit (StemCell Technologies) (33), resulting in purities of >99.2% CD14⁻CD66b⁺CD15⁺ and <0.1% contaminating monocytes (Supplemental Fig. 1B). Total CD3⁺ T cells (>98%) were isolated from PBMC using the pan T cell isolation kit II (Miltenyi Biotec); CD4⁺ and CD8⁺ T cells (>98%) were obtained using the corresponding EasySep kits (StemCell Technologies). Vγ9⁺ T cells (>98%) were purified using anti–Vγ9-PE-Cy5 (Immu360; Beckman Coulter) and anti-PE microbeads (Miltenyi Biotec); Vα7.2⁺ T cells (>98%) were purified using anti–Vα7.2-PE (3C10; BioLegend) and anti-PE microbeads. Alternatively, Vγ9⁺ CD3⁺ γδ T cells or Vα7.2⁺ CD161⁺ CD3⁺ MAIT cells were sorted to purities >99% using a FACS-Aria II (BD Biosciences).

Clinical isolates of Enterobacter cloacae, Enterococcus faecalis, Klebsiella pneumoniae, and S. aureus (28) were grown in liquid Luria-Bertani broth and on solid Columbia blood agar (Oxoid). The distribution of the nonmevalonate and riboflavin pathways across microbial species was determined based on the absence or presence of the enzymes HMB-PP synthase (EC 1.17.7.1) and DMRL synthase (EC 2.5.1.78) in the corresponding genomes, according to the Kyoto Encyclopedia of Genes and Genomes (http://www.genome.jp/kegg).

PBMC were stimulated with 0.1–100 nM HMB-PP or 0.1–100 μM DMRL. Vγ9⁺ T cells or Vα7.2⁺ T cells were cocultured with autologous monocytes at a ratio of 1:1, in the presence of 25% (v/v) cell-free supernatants from neutrophils that had phagocytosed live bacteria, as described previously (28). For blocking experiments, anti–TCR-Vα7.2, anti–TCR-Vγ9, anti-CD277, and anti-MR1 were used at 1–20 μg/ml.

Freshly isolated neutrophils were cultured for up to 48 h in the absence or presence of autologous Vγ9/Vδ2 T cells or MAIT cells at a ratio of 10:1, and 10 nM HMB-PP or anti-CD3/CD28 dynabeads (1 bead per T cell). Alternatively, neutrophils were cultured with 25–50% (v/v) conditioned medium obtained from purified Vγ9/Vδ2 T cells or MAIT cells stimulated for 24 h with anti-CD3/CD28 dynabeads (1 bead per cell) or 100 nM HMB-PP. Other stimuli included 100 ng/ml LPS and 100 U/ml recombinant IFN-γ, TNF-α, and/or GM-CSF. sTNFR p75-IgG1 fusion protein, anti–IFN-γ, and anti–GM-CSF were used at 10 μg/ml. Neutrophil survival and activation were assessed by flow cytometry, after gating on CD15⁺ cells and exclusion of Vγ9⁺ or Vα7.2⁺ cells where appropriate. For morphological analyses, neutrophils were centrifuged onto cytospin slides, stained with May-Grünwald-Giemsa solution, and analyzed by light microscopy.

Endocytosis and APC functions were assessed as before (34–38). Freshly purified neutrophils and neutrophils cultured for 24 h in the presence or absence of unconventional T cell–conditioned medium were incubated with 500 μg/ml 10-kDa dextran-FITC or BSA-FITC for up to 60 min at 4°C or 37°C. Endocytic uptake was measured immediately by flow cytometry; the specific uptake of each reagent was calculated by subtracting the background MFI at 4°C from the MFI obtained at 37°C.

For MHC class II–restricted presentation of Ags, activated neutrophils were generated by 48-h culture with a combination of IFN-γ, GM-CSF, and/or TNF-α, or with unconventional T cell–conditioned medium. Neutrophils were pulsed with 10 ng/ml TSST-1 for 1 h. After extensive washing, neutrophils were mixed with autologous CD4⁺ T cells at a ratio of 1:1; 1 h later 10 μg/ml brefeldin A was added and cultures were incubated for an additional 4 h. Activation of TSST-1–responsive Vβ2⁺ CD4⁺ T cells was assessed by intracellular cytokine staining and analysis by flow cytometry (35). To assess CD4⁺ and CD8⁺ T cell responses to complex Ag preparations, neutrophils were pulsed with 1–10 μg/ml PPD for the last 18 h of the 48-h culture phase. After extensive washing, neutrophils were mixed with CFSE-labeled autologous CD3⁺ T cells at a ratio of 1:1 and incubated for 7 d. CFSE dilution in the CD4⁺ and CD8⁺ T cell populations was assessed by flow cytometry, after exclusion of CD66b⁺ cells.

For MHC class I–restricted Ag presentation, Ag-specific HLA-A2–restricted CD8⁺ T cell lines were generated using the immunodominant peptide of influenza matrix protein, M1(p58–66) (GILGFVFTL), as described before (37, 38). M1(p58–66)-specific responder CD8⁺ T cells used in APC assays were >95% pure, as confirmed by tetramer staining (data not shown). Activated neutrophils from HLA-A2⁺ donors were generated as above, using unconventional T cell–conditioned medium or recombinant cytokines, and pulsed for 1 h with 0.1 μM peptide. For cross-presentation assays, 0.01–1 μM recombinant influenza M1 protein was added during the last 18 h of the 48-h neutrophil culture period. Fresh neutrophils from HLA-A2⁺ sepsis patients were incubated with 0.01–1 μM recombinant influenza M1 protein for 18 h or cultured in medium for 17 h prior to addition of 0.1 μM M1(p58–66) peptide for an additional 1 h. In each case, following extensive washing, neutrophils were incubated with HLA-A2⁺ peptide-specific CD8⁺ T cells at a ratio of 1:1; after 1 h, 10 μg/ml brefeldin was added and cultures were incubated for an additional 4 h. Activation of CD8⁺ T cells was assessed by intracellular cytokine staining and analyzed by flow cytometry, after exclusion of CD66b⁺ cells.

Cells were acquired on an eight-color FACSCanto II (BD Biosciences) and analyzed with FlowJo (Tree Star). Single cells of interest were gated based on their appearance in side and forward scatter area/height, exclusion of live/dead staining (fixable Aqua; Invitrogen), and surface staining. Apoptotic cells were identified using annexin-V (BD Biosciences).

Cell culture supernatants were analyzed on a Dynex MRX II reader, using ELISA kits for IL-17A (R&D Systems) as well as IFN-γ and TNF-α (eBioscience). Cell-free plasma samples and unconventional T cell–conditioned media were analyzed on a SECTOR Imager 6000 using the ultrasensitive human proinflammatory 9-plex kit (Meso Scale Discovery).

Data were analyzed using two-tailed Student t tests for normally distributed data and Mann–Whitney tests for nonparametric data (GraphPad Prism). Differences between groups were analyzed using one-way ANOVA with Bonferroni's posttests or with Kruskal–Wallis and Dunn’s posttests; two-way ANOVA was used when comparing groups with independent variables.

Vγ9/Vδ2⁺ γδ T cells recognize the isoprenoid precursor HMB-PP, which is produced via the nonmevalonate pathway by a broad range of Gram-negative and Gram-positive bacteria (27, 39). Vα7.2⁺ CD161⁺ MAIT cells show a very similar responsiveness to an overlapping, but distinct spectrum of microorganisms by sensing intermediates of the microbial vitamin B2 biosynthesis (Table I) (40–43). We therefore sought to investigate the antimicrobial responses of these two types of unconventional T cells side by side. In this study, Vγ9/Vδ2 T cells, but not MAIT cells, responded to HMB-PP, as judged by induction of CD69 expression (Fig. 1A). In contrast, the riboflavin precursor DMRL induced a dose-dependent activation of MAIT cells, but not Vγ9/Vδ2 T cells. Blocking experiments confirmed a requirement for butyrophilin 3A/CD277 for Vγ9/Vδ2 T cells and the MHC-related protein MR1 for MAIT cells (Fig. 1B), in support of current models of Ag recognition (41–45).

We previously identified a crucial role for neutrophils in facilitating access to HMB-PP by Vγ9/Vδ2 T cells (28). As control, purified Vγ9⁺ T cells readily responded to neutrophils after phagocytosis of clinically relevant bacteria, in accordance with the distribution of the nonmevalonate pathways across the different pathogens (Table I). Strikingly, purified Vα7.2⁺ T cells showed very similar responses depending on the utilization of the riboflavin biosynthesis pathway by the phagocytosed species. Activated Vγ9⁺ T cells and Vα7.2⁺ T cells upregulated CD69 (Fig. 1C) and secreted IFN-γ (Fig. 1D), but not IL-17A (data not shown). The response of Vα7.2⁺ T cells to microbial compounds was confined to the CD161⁺ bona fide MAIT cell population (Fig. 1C, 1D). Both Vγ9/Vδ2 T cells and MAIT cells failed to respond to neutrophil-released microbial compounds in the presence of anti-CD277 and anti-MR1, respectively (Fig. 1D), and in the absence of autologous monocytes (Fig. 1E), highlighting a requirement for presentation by accessory cells. These findings reveal a remarkable similarity in the responsiveness of Vγ9/Vδ2 T cells and MAIT cells to microbial metabolites.

To resolve the existence of APC-like neutrophils in human infectious disease and determine a possible link with antimicrobial unconventional T cell responses, we recruited adult patients with newly diagnosed severe sepsis and characterized their circulating leukocytes phenotypically and functionally. As proof of principle for the involvement of unconventional T cells in early inflammatory responses, patients with acute sepsis revealed a substantial activation of Vγ9/Vδ2 T cells, as judged by CD69 expression, but not SIRS patients who served as noninfected controls (Fig. 1F, Supplemental Fig. 1A). Of note, we found a significant increase in the absolute counts and the proportion of Vγ9/Vδ2 T cells among all circulating T cells between patients with microbiologically confirmed infections caused by HMB-PP–producing as opposed to HMB-PP–deficient species (Fig. 1F). These clinical findings evoke earlier studies in patients with acute peritonitis (28) and further support the notion of a differential responsiveness of unconventional T cells to defined pathogen groups that can be detected both locally at the site of infection (46) and systemically in blood (Fig. 1F).

We recently showed that Vγ9/Vδ2 T cells trigger short-term (<20 h) survival of autologous neutrophils (28). In this study, highly purified neutrophils cocultured for extended periods with activated Vγ9/Vδ2 T cells or MAIT cells displayed a prolonged survival, as judged by exclusion of amine reactive dyes and retention of surface CD16 (FcγRIII) for at least 48 h (Fig. 2A). A similar effect was observed when incubating purified neutrophils with Vγ9/Vδ2 T cell or MAIT cell–conditioned culture supernatants, indicating a significant contribution of soluble factors in mediating the observed effects (Fig. 2B). In contrast to the highly active metabolite HMB-PP as specific activator of Vγ9/Vδ2 T cells, the MAIT cell activator used in the current study, DMRL, only possesses a relatively modest bioactivity. The true MAIT cell activator is far more potent than DMRL and active at subnanomolar concentrations, but not commercially available and difficult to synthesize chemically (41, 43). Most stimulation experiments with purified MAIT cells were therefore conducted with anti-CD3/CD28–coated beads. Importantly, use of either anti-CD3/CD28 beads or HMB-PP to activate Vγ9/Vδ2 T cells elicited identical neutrophil responses (Fig. 2A–C and data not shown). Surviving neutrophils possessed a highly activated morphology, as judged by the presence of hypersegmented nuclei (Fig. 2C). The antiapoptotic effect of unconventional T cells was confirmed by the preservation of the total number of neutrophils present after 48 h of culture and the lack of annexin-V binding (Fig. 2D). As confirmation of their activated status, surviving neutrophils showed pronounced upregulation of CD11b and CD66b expression and complete loss of CD62L (Fig. 2E).

Circulating neutrophils in healthy people do not express CD40, CD64 (FcγRI), CD83, or HLA-DR, yet all these surface markers were found on unconventional T cell–primed neutrophils (Fig. 2F). Moreover, these neutrophils also showed a marked upregulation of CD54 (ICAM-1) and HLA-ABC (Fig. 2F), suggestive of a possible function of unconventional T cell–primed neutrophils as APCs for both CD4⁺ and CD8⁺ T cells.

The chemokine receptors CCR7, CCR9, and CCR10 remained undetectable under those culture conditions (data not shown), arguing against trafficking of APC-like neutrophils to noninflamed lymph nodes, the intestine, or the skin. In contrast, APC-like neutrophils displayed enhanced expression levels of CXCR3 and CCR4 (data not shown), indicative of an increased responsiveness to inflammatory chemokines and supporting a local role during acute inflammation.

Neutrophils stimulated with defined microbial compounds on their own, in the absence of Vγ9/Vδ2 T cells or MAIT cells, failed to acquire a similar phenotype. Most notably, neutrophils cultured for 48 h in the presence of LPS did not show increased levels of HLA-ABC, HLA-DR, CD40, CD64, or CD83 compared with neutrophils cultured in medium alone (data not shown), emphasizing the crucial and nonredundant contribution of unconventional T cells and their specific ligands to the acquisition of APC characteristics by neutrophils.

To resolve the existence of APC-like neutrophils in human infectious disease, we characterized circulating leukocytes in sepsis patients as a means to access neutrophils that had recently been activated in different infected tissues. Sepsis neutrophils displayed a strikingly altered phenotype compared with neutrophils from healthy individuals and SIRS patients and were characterized by markedly higher expression of CD40, CD64, and CD86 (Fig. 3A). We also found increased surface levels of CD83 and HLA-DR on circulating neutrophils in some patients with sepsis, although this was not significant across the cohort as a whole. Of note, there was a correlation between the expression of CD64 and HLA-DR on sepsis neutrophils, supporting a link between neutrophil activation and APC phenotype (Fig. 3B). These findings indicate the presence of APC-like neutrophils in sepsis patients, despite the generally presumed immune suppression in those individuals, as judged by reduced HLA-DR expression levels on monocytes (data not shown) (47).

To identify the unconventional T cell–derived factor(s) exerting the observed effects on neutrophils, we quantified proinflammatory mediators in the culture supernatants. These experiments revealed a dominant production (>1000 pg/ml on average) of GM-CSF, IFN-γ, and TNF-α by activated Vγ9/Vδ2 T cells and MAIT cells, but only very low levels (<25 pg/ml) of IL-1β, IL-6, and CXCL8, indicating that both unconventional T cell populations share a similar cytokine profile (Fig. 4A). Experiments using blocking reagents identified an involvement of GM-CSF, IFN-γ, and TNF-α in promoting neutrophil survival by both Vγ9/Vδ2 T cells and MAIT cells (Fig. 4B). Whereas neutralization of each individual cytokine on its own had a partial effect, combined blocking of GM-CSF and IFN-γ was most effective in inhibiting neutrophil survival, with blocking of TNF-α having little additive effect. In contrast, CD66b upregulation was mainly triggered by TNF-α (Fig. 4B). Of note, the effect of unconventional T cells on neutrophils could be mimicked in part by using recombinant GM-CSF, IFN-γ, and TNF-α. In this respect, only neutrophils cultured with a combination of all three cytokines exhibited a morphology characterized by hypersegmented nuclei (Fig. 4C). TNF-α was particularly important for the induction of CD40, CD54, CD66b, and MHC class I expression (Fig. 4D). Taken together, these experiments identify microbe-responsive unconventional T cells as a rapid physiological source of GM-CSF, IFN-γ, and TNF-α and imply that the unique combination of cytokines secreted by unconventional T cells is key for the observed impact on neutrophils.

The particular requirement for TNF-α in the acquisition of the full APC phenotype is especially noteworthy when considering the cytokine/chemokine profiles in acutely infected patients. Plasma proteins that were highly elevated in sepsis patients included TNF-α as well as IL-6 and CXCL8 (Fig. 4E). Of note, there was a trend toward higher levels of TNF-α in patients with HMB-PP–positive infections (p = 0.09; data not shown). A proportion of individuals with sepsis also had increased plasma levels of GM-CSF, IFN-γ, and IL-1β, although this was not significant across the whole cohort (Fig. 4E). These findings confirm that the blood of sepsis patients contains proinflammatory mediators implicated in driving survival and activation of neutrophils, including their differentiation into APCs.

We next tested the capacity of APC-like neutrophils to take up soluble Ags. Although freshly isolated neutrophils were not very efficient at endocytosing FITC-labeled BSA and dextran (10,000 Da) as model compounds, short-term exposure to Vγ9/Vδ2 T cell–conditioned medium led to a greatly enhanced uptake (Fig. 5A). With unconventional T cell–primed neutrophils kept in culture for 24 h before addition of BSA-FITC, Ag endocytosis was confined to the CD16^high APC-like population, whereas no such uptake was seen in the apoptotic CD16^low population (Fig. 5B). In contrast, neutrophils cultured in medium alone showed no specific uptake of BSA in the CD16^high population. These data indicate that unconventional T cells promote the uptake of exogenous Ags as a prerequisite for Ag processing and presentation by neutrophils.

The functionality of cell surface–expressed HLA-DR on activated neutrophils was confirmed using the S. aureus superantigen, TSST-1, which cross-links MHC class II molecules with the TCR of CD4⁺ T cells expressing a Vβ2 chain (35). Neutrophils exposed to Vγ9/Vδ2 T cell–conditioned medium or to a combination of GM-CSF, IFN-γ, and TNF-α were both capable of presenting TSST-1 to autologous Vβ2⁺ CD4⁺ T cells (Fig. 5C). When using the complex M. tuberculosis Ag, PPD, which requires intracellular processing, unconventional T cell–primed neutrophils displayed a striking capacity to trigger proliferation of both CD4⁺ and CD8⁺ T cells (Fig. 5D). Sequestration of TNF-α during the neutrophil-priming period by addition of sTNFR diminished both CD4⁺ and CD8⁺ T cell responses (Fig. 5E) as further confirmation of the key role for unconventional T cell–derived TNF-α in the acquisition of APC features by neutrophils.

Following up from the striking induction of PPD-specific CD4⁺ and CD8⁺ T cell responses, we assessed the potential of APC-like neutrophils to trigger CD8⁺ T cell responses, by taking advantage of HLA-A2–restricted responder T cell lines specific for M1(p58–66), the immunodominant epitope of the influenza M1 protein (36–38). Using the M1(p58–66) peptide, which can be pulsed readily onto cell surface–associated MHC class I molecules for direct presentation to CD8⁺ T cells, unconventional T cell–primed neutrophils showed a significantly improved Ag presentation, compared with freshly isolated neutrophils (Fig. 6A) and in agreement with the elevated levels of MHC class I molecules on APC-like neutrophils. Importantly, only unconventional T cell–primed neutrophils, but not freshly isolated neutrophils, were also able to induce robust responses by M1(p58–66)-specific responder CD8⁺ T cells when utilizing the full-length M1 protein (Fig. 6A), a 251-aa–long Ag that requires uptake, processing, and loading of M1(p58–66) onto intracellular MHC class I molecules for cross-presentation to CD8⁺ T cells (36–38). Control experiments supported the need for Ag uptake and processing, as recombinant M1 protein could not be pulsed directly onto neutrophils, demonstrating the absence of potential degradation products in the M1 protein preparation that might be able to bind directly to cell surface–associated MHC class I molecules on neutrophils or CD8⁺ T cells (Fig. 6B). Neutrophils cultured for 48 h in the presence of GM-CSF and IFN-γ were also capable of enhanced presentation of M1(p58–66) peptide to M1-specific CD8⁺ T cells. However, only neutrophils generated by incubation with a combination of GM-CSF, IFN-γ, and TNF-α readily processed the full-length M1 protein (Fig. 6A), demonstrating that TNF-α plays a pivotal role in the acquisition of a fully competent APC phenotype and function by neutrophils.

It has not yet been established whether neutrophils are capable of triggering Ag-specific T cell responses in vivo. To translate our findings on APC-like neutrophils to the situation in acute infections, we isolated untouched neutrophils from sepsis patients to purities of 99.2–99.8%. Our experiments show that sepsis neutrophils and control neutrophils had a similar capacity to activate M1-specific responder CD8⁺ T cells when pulsed with the peptide itself (Fig. 6C). Strikingly, only sepsis neutrophils, but not control neutrophils, were also able to take up the full-length M1 protein and cross-present the M1(p58–66) peptide to responder CD8⁺ T cells (Fig. 6C, 6D), consistent with the differences in APC marker expression between patients and healthy individuals. These findings indicate that in acute sepsis neutrophils acquire an APC-like phenotype with the capacity to induce Ag-specific CD8⁺ T cell responses that is reminiscent of neutrophils primed by unconventional T cells (Fig. 7).

To our knowledge, the present study is the first demonstration that human neutrophils can assume Ag cross-presenting properties. Although our work does not formally demonstrate a causal link for the interaction of unconventional T cells and neutrophils in vivo, it does suggest a plausible scenario for the generation of APC-like neutrophils during acute infection. Our data support a model in which different types of unconventional T cells respond rapidly to neutrophils after phagocytosis of a broad range of bacteria at the site of infection, and in turn mediate the local differentiation of bystander neutrophils into APCs for both CD4⁺ and CD8⁺ T cells (Fig. 7). APC-like neutrophils may be particularly relevant for local responses by tissue-resident memory and/or freshly recruited effector CD4⁺ and CD8⁺ T cells at the site of infection, rather than the priming of naive CD4⁺ and CD8⁺ T cells in secondary lymphoid tissues. Expression of the lymph node homing receptor CCR7 by activated neutrophils was reported before (30) but could not be confirmed in the current study (data not shown). Still, APC-like neutrophils may also gain access to inflamed draining lymph nodes through the action of inflammatory chemokines (7–10). Irrespective of the anatomical context, APC-like neutrophils may contribute to protective immune responses, by fighting the “first hit” infection as a result of inducing Ag-specific CD4⁺ and CD8⁺ T cells and by harnessing the T cell compartment against potential “second hit” infections. However, it is also thinkable that such an early induction of cytotoxic CD8⁺ T cells may add to the systemic inflammatory response and ultimately lead to tissue damage and organ failure. Whereas the generation of APC-like neutrophils is likely to occur locally in the context of infected tissues, in severe inflammatory conditions, including sepsis, such APC-like neutrophils may eventually leak into the circulation and become detectable in blood. Larger stratified approaches are clearly needed to define the role of APC-like neutrophils in different infectious scenarios, locally and systemically, in clinically and microbiologically well-defined patient subgroups.

The presence of cross-presenting neutrophils in patients with sepsis is intriguing and may point to an essential role of APC-like neutrophils in acute disease. Sepsis patients who survive the primary infection often show signs of reduced surface expression of HLA-DR on monocytes and a relative tolerance of monocytes to LPS stimulation (47). As consequence of what is generally perceived as a loss of immune function, many patients are susceptible to subsequent nosocomial infections, including reactivation of latent viruses that are associated with high mortality rates (48, 49). Trials specifically targeted at reversing this apparent monocyte deactivation have shown promising clinical results (50). However, our present findings suggest that HLA-DR expression by circulating monocytes is a poor surrogate marker for a systemic immune suppression and rather indicate that, contrary to the proposed general loss of function, certain cells such as neutrophils may actually assume APC properties under such conditions, as evidence of a gain of new function. Yet, with a complex and multilayered clinical phenomenon such as sepsis it is challenging to dissect the relevance of APC-like neutrophils for infection resolution and clinical outcome in vivo.

With their unique ability to recognize microbial metabolites in a non-MHC–restricted manner, unconventional T cells such as Vγ9/Vδ2 T cells and MAIT cells greatly outnumber Ag-specific conventional CD4⁺ and CD8⁺ T cells at the site of infection and represent early and abundant sources of proinflammatory cytokines (23, 27), among which GM-CSF, IFN-γ, and TNF-α each make key contributions. Although conventional T cells may produce a similar combination of cytokines and provide similar signals to neutrophils, preliminary findings in our laboratory indicate that Vγ9/Vδ2 T cells and MAIT cells represent together up to 50% of all TNF-α–producing T cells among peritoneal cells stimulated with bacterial extracts, suggesting that these two cell types are indeed major producers of proinflammatory cytokines in response to microbial stimulation (A. Liuzzi and M. Eberl, unpublished observations). Although we cannot rule out a further contribution of contact-dependent mechanisms, this observation builds upon earlier studies describing the generation of human neutrophils expressing MHC class II through the action of recombinant cytokines in vitro (22, 51–56) and in vivo (57–59). Previous investigations reported an upregulation of MHC class II on activated neutrophils under the control of GM-CSF and IFN-γ, albeit the physiological source of those mediators during acute infection was not defined. Most importantly, in this study, we describe a direct role for TNF-α in the efficient induction of MHC class I–restricted CD8⁺ T cell responses by neutrophils. Of note, plasma from sepsis patients was previously shown to induce some (upregulation of CD64), but not other features (upregulation of CD11b, loss of CD62L) (60) that are characteristic for unconventional T cell–primed neutrophils, indicating that circulating cytokines alone do not confer APC properties. In support of local cell-mediated processes at the site of inflammation, our findings evoke earlier descriptions of APC-like neutrophils characterized by MHC class II expression in infectious and noninfectious inflammatory scenarios such as periodontitis (17) and tuberculous pleuritis (20), in which locally activated Vγ9/Vδ2 T cells were found (61–63). These associations lend further support to the existence of a peripheral immune surveillance network comprised of distinct types of unconventional T cells and their crosstalk with local immune and nonimmune cells. In the absence of unconventional T cell–derived signals, such as during sterile inflammation induced by LPS administration (64), neutrophils may not become fully primed, in accordance with our failure to induce APC-like neutrophils using LPS alone. Of note, a possible feedback regulation may require the activation of unconventional T cells to reach a certain threshold to overcome the inhibitory effect of bystander neutrophils (65–67).

Our present data demonstrate that both isoprenoid and riboflavin precursors are released by human neutrophils upon phagocytosis of live bacteria and depend on uptake by monocytes and loading onto butyrophilin 3A and MR1, respectively. The surprising similarities between Vγ9/Vδ2 T cells and MAIT cells illustrate their overlapping, yet distinct roles. Given the broad distribution of the nonmevalonate and riboflavin pathways across pathogenic, opportunistic, and commensal species, the vast majority of invading microbes is likely to be detected by either Vγ9/Vδ2 T cells or MAIT cells, or both. Our analysis of sepsis patients identified a systemic mobilization of Vγ9/Vδ2 T cells in response to HMB-PP–producing species, in support of their differential responsiveness to distinct groups of bacteria (28, 46). Because the present clinical study was conceived before information about the responsiveness of MAIT cells for riboflavin metabolites became available in the literature, we did not conduct a differential analysis for MAIT cells during acute sepsis. Of note, except for two cases of streptococcal infections, all bacterial and fungal pathogens identified in this patient cohort in fact possessed the riboflavin pathway, that is, were theoretically capable of stimulating MAIT cells. Intriguingly, Grimaldi et al. (68) recently reported a specific depletion of peripheral MAIT cells in sepsis patients with nonstreptococcal (i.e., riboflavin-producing) bacteria compared with infections caused by riboflavin-deficient species, which may indicate differences in the recruitment and retention of different types of unconventional T cells at sites of infection, depending on the nature of the causative pathogen and the underlying pathology (69–72). The contribution of tissue-resident and freshly recruited unconventional T cells to acute inflammatory responses has implications for clinical outcome and for the development of novel diagnostics and therapeutic interventions (46).

Taken together, our present study provides evidence 1) that Vγ9/Vδ2 T cells and MAIT cells respond similarly to microbial pathogens that produce the corresponding ligands when phagocytosed by human neutrophils, 2) that, once activated, both types of unconventional T cells trigger longer-term survival and differentiation of neutrophils into APC-like cells, 3) that unconventional T cell–primed neutrophils readily process exogenous Ags and prime both CD4⁺ and CD8⁺ T cells, and 4) that circulating neutrophils from patients with acute sepsis possess a similar APC-like phenotype and are capable of cross-presenting soluble proteins to Ag-specific CD8⁺ T cells ex vivo. These findings define a possible physiological context for the generation of APC-like neutrophils in response to a broad range of microbial pathogens and imply a unique and decisive role for human unconventional T cells in orchestrating local inflammatory events and in shaping the transition of the innate to the adaptive phase of the antimicrobial immune response, with implications for diagnosis, therapy, and vaccination.

We are grateful to all patients and volunteers for participating in this study and we thank the clinicians and nurses for cooperation. We also thank Mark Toleman for clinical pathogens, Hassan Jomaa and Boris Illarionov for HMB-PP and DMRL, Andrew Thomas for recombinant M1 protein and HLA-A2 tetramers, Ted Hansen and Daniel Olive for mAbs, Ann Kift-Morgan for multiplex ELISA measurements, Catherine Naseriyan for cell sorting, Chia-Te Liao for help with cytospins, and Marco Cassatella, Adrian Hayday, Ian Sabroe, and Phil Taylor for stimulating discussion.

The authors have no financial conflicts of interest.