NOX2-Derived Reactive Oxygen Species Control Inflammation during Leishmania amazonensis Infection by Mediating Infection-Induced Neutrophil Apoptosis

Leishmaniasis is a neglected tropical disease affecting ∼350 million people in 88 countries worldwide (1). Disease is transmitted by the bite of a Leishmania-infected sand fly of the genus Phlebotomus in the Old World or of the genus Lutzomyia in the New World. Infection can cause a range of clinical manifestations depending upon which species of Leishmania is involved. Infection with Leishmania amazonensis, a major cause of cutaneous leishmaniasis in South America, is often associated with nonhealing dermal lesions in people. Infection of C57BL/6 mice is also characterized by development of nonhealing lesions with persistent parasitism (2–4). Interestingly, this experimental infection is associated with a relatively weak Th1 immune response that is characterized by low IFN-γ production not accompanied by a reciprocally strong nonprotective Th2 response, which is associated with progressive disease in other models of Leishmania infection (3, 4). We have previously shown that, during the early stages of L. amazonensis infection, parasite growth is equivalent in IFN-γ^−/− mice and wild-type (WT) mice (5). Thus, we hypothesized that low parasite loads during early L. amazonensis infection may be due to the production of IFN-γ–independent reactive oxygen species (ROS).

ROS are a group of molecules that include highly reactive radicals or nonradicals derived from oxygen (6). The NADPH oxidase isoform of phagocytes (NOX2), expressed principally in neutrophils and macrophages, is activated by phagocytosis. NOX2 is the major enzyme involved in the rapid cellular response against intracellular pathogens. NOX2 consists of a catalytic transmembrane heterodimer, gp91^phox and p22^phox, and four regulatory cytosolic subunits, p40^phox, p47^phox, p67^phox, and Rac2 (7). Phagocytosis triggers the assembly of the transmembrane and cytosolic parts of the NOX2 complex, and the gp91^phox subunit transfers electrons from NADPH to molecular oxygen, forming the anion superoxide (O₂^•−) (7). Therefore, gp91^phox is critical to NOX2 function. P47^phox- or gp91^phox-deficient mice develop chronic granulomatous disease (8, 9), which is associated with susceptibility to microbial infections, especially fungal (8–10) and bacterial infections (9, 11–13), similar to what is observed in humans. ROS also have an important role in the inflammatory processes, including activation of inflammatory genes, such as NF-κB and AP-1 (14, 15), and induction of apoptosis in neutrophils and T lymphocytes under oxidative stress (16–19). The role of ROS during Leishmania infection is poorly understood, because most studies focus on the role of NO in parasite killing (20, 21). NO, in particular, is critical for IFN-γ–mediated Leishmania killing in mice, because mice lacking inducible NO synthase (iNOS^−/− or NOS2^−/−) present with uncontrolled infection due to enhanced parasite replication (5). In contrast, although ROS are essential to kill many intracellular pathogens, susceptibility to IFN-γ–independent ROS-mediated killing appears to be Leishmania species specific. For example, L. guyanensis, but not L. amazonensis, is highly susceptible to ROS-mediated killing following in vitro infection of BALB/c-derived macrophages and also induces higher levels of ROS compared with L. amazonensis (22), whereas L. donovani, L. major, and L. enriettii inhibit ROS generation (23–25). Because of this inhibition, and in contrast to iNOS^−/− mice, gp91^phox−/− mice ultimately control infection with L. donovani or low-dose L. major infection in the footpad, similar to WT mice (26, 27). In studies using L. major, inhibition of ROS is largely mediated by lipophosphoglycan and gp63 localized on the parasite surface (28, 29). Lastly, ROS may be more important for parasite elimination by human, rather than mouse, phagocytes because in vitro treatment of human macrophages with SOD1, a depleting enzyme of anion superoxide, enhanced L. amazonensis replication, whereas pharmacological inhibition of ROS in mouse macrophages did not (30, 31).

In this study, we investigated the role of ROS using an intradermal inoculation model of cutaneous infection with purified L. amazonensis metacyclic promastigotes in mice deficient in the gp91^phox subunit of NOX2 (gp91^phox−/−). In contrast to our prediction that ROS play a role in L. amazonensis control, we observed that ROS deficiency did not alter parasite loads. However, ROS were critically important for neutrophil programmed cell death (PCD) and the subsequent control of inflammation. Neutrophils are the first cells recruited to acute sites of infection, and L. amazonensis infection of human or mouse neutrophils induces ROS production in vitro (32, 33). Recent studies have identified a critical role for neutrophils in L. major infection, including providing an intracellular “safe haven” for parasites and downmodulation of the adaptive immune response (34, 35). In this study, we found that the absence of ROS induced by NOX2 alters neutrophil PCD, leading to the accumulation of necrotic neutrophils and significantly enhanced pathology, including extensive loss of the infected tissue. These observations highlight the role of ROS-dependent PCD in modulating inflammation in settings of chronic infection.

All mice were bred and maintained at the University of Calgary Animal Resource Centre or at the National Institute of Allergy and Infectious Diseases animal care facility under specific pathogen–free conditions. Male or female C57BL/6 mice and gp91^phox null allele-deficient mice (B6.129S6-Cybb^tm1Din - gp91^phox−/−) (9) were purchased from The Jackson Laboratory or provided by Dr. R. Yates (University of Calgary) and cohoused.

L. amazonensis PH8 (IFLA/BR/1967/PH8) and a stable transfected line of L. amazonensis PH8 expressing an RFP (La-RFP) were used for infections. La-RFP was generated similarly as described (34).

Parasites were grown at 26°C in medium 199 supplemented with 20% heat-inactivated FCS (Gemini Bio Products, San Francisco, CA), 100 U/ml penicillin, 100 μg/ml streptomycin, 2 mM l-glutamine, 40 mM HEPES, 0.1 mM adenine (in 50 mM HEPES), 5 mg/ml hemin (in 50% triethanolamine), 1 mg/ml 6-biotin (M199/S), and 50 μg/ml Geneticin (Life Technologies, Waltham, MA). Infective-stage metacyclic promastigotes were isolated from stationary cultures (4–5 d old) in Ficoll gradient, as described previously (36).

Soluble L. amazonensis Ag was obtained from 100 ml of stationary cultures. Parasites were lysed by five cycles of freezing and thawing, and Ag quantification was made using a BCA Protein Assay Kit (Pierce, Waltham, MA), according to the manufacturer’s instructions.

For long-term infection studies, mice were infected with 5 × 10³ metacyclic promastigote forms of L. amazonensis PH8, and the lesion growth was followed weekly by lesion diameter measurements using a caliper. Ears presenting with tissue loss were also determined weekly. For short-term infection studies, mice were infected with 5 × 10⁵ metacyclic promastigote forms of L. amazonensis PH8. For both conditions, the parasites were diluted in 10 μl of DMEM (Life Technologies) and injected intradermally in the ears in 10 μl.

Ear tissue was prepared as previously described (37). Briefly, the two sheets of infected ear dermis were separated, deposited in DMEM containing 100 U/ml penicillin, 100 μg/ml streptomycin, and 0.2 mg/ml Liberase CI purified enzyme blend (Roche Diagnostics, Rotkreuz, Switzerland), and incubated for 1.5 h at 37°C and 5% CO₂. Digested tissue was placed in a Medicon tissue grinder and processed in a tissue homogenizer (Medimachine; Becton Dickinson, San Jose, CA). Tissue homogenates were filtered through a 70-μm cell strainer (Falcon Products, San Jose, CA).

Parasite titrations were performed as previously described (38). Tissue homogenates were serially diluted in 96-well flat-bottom microtiter plates containing biphasic medium, prepared using 50 μl of Novy–MacNeal–Nicolle medium containing 20% of defibrinated rabbit blood and overlaid with 100 μl of M199/S. The number of viable parasites in each ear was determined from the highest dilution at which promastigotes could be grown out after 7–10 d of incubation at 26°C.

Single-cell suspensions were incubated with an anti-Fc III/II (CD16/32) receptor Ab (2.4G2; BD Biosciences, San Jose, CA) in RPMI 1640 without phenol red (Life Technologies) containing 1% FCS and stained with fluorochrome-conjugated Abs. The following Abs were used: allophycocyanin anti-mouse CD11c (HL3; BD Biosciences, San Jose, CA), PE-Cy7 anti-mouse CD11b (M1/70; eBioscience, San Jose, CA), PerCP-Cy5.5 anti-mouse Ly6C (HK1.4; eBioscience), allophycocyanin-Cy7 anti-mouse Ly6C (AL-21; BD Pharmingen, San Jose, CA), FITC anti-mouse Ly6G (1A8; eBioscience), eFluor 450 anti-mouse F4/80 (BM8; eBioscience), allophycocyanin–eFluor 780 anti-mouse MHC II (M5/114.15.2; eBioscience), PE anti-mouse CD86 (GL1; eBioscience), FITC anti-human Ki-67 (B56; BD Biosciences), PE anti-mouse Foxp3 (FJK-16S; eBioscience), eFluor 660 anti-mouse T-bet (4B10; eBioscience), PE-Cy7 anti-mouse CD4, allophycocyanin–eFluor 780 anti-mouse CD44 (IM7; eBioscience); eFluor 450 anti-mouse CD8 (53-6.7; eBioscience), V500 anti-mouse CD3 (500A2; BD Biosciences), FITC anti-mouse IFN-γ (S54411; BD Biosciences), PE anti-mouse IL-10 (JES5-16E3; BD Biosciences), PerCP-Cy5.5 anti-mouse IL-17A (17B7; eBioscience), allophycocyanin anti-mouse IL-2 (JES6-5H4; eBioscience), Alexa Fluor 700 anti-mouse TNF (MP67X22; BD Pharmingen), Brilliant Violet anti-mouse TCRβ (H57-597; BioLegend, San Diego, CA), and allophycocyanin anti-mouse CD95 (15A7; eBioscience). The isotype controls used (all obtained from BD Biosciences) were rat IgG1 (R3–34) and rat IgG2b (A95-1). Neutrophils, dendritic cells (DCs), macrophages, monocytes, and T lymphocytes from the ear dermis were identified based on size (forward scatter) and granularity (side scatter) and by surface phenotype, as indicated in the text and figure legends. Forward-scatter and side-scatter width were used to eliminate cell doublets or triplets from analysis.

The staining of surface and intranuclear markers (Ki-67, Foxp3, T-bet) was performed sequentially: cells were initially stained for surface markers, followed by fixation and permeabilization with the eBioscience Foxp3 stain kit, as previously described (37).

For intracellular detection of cytokines, ear cell suspensions were stimulated with 5 × 10⁵ naive spleen cells depleted of CD4 T cells using MACS Separation Columns and anti-mouse CD4 (L3T4) MicroBeads (Miltenyi Biotec, San Diego, CA) plus 50 μg of soluble L. amazonensis Ag for 10 h. During the final 4 h of culture, 1 μg/ml brefeldin A (Sigma-Aldrich, St. Louis, MO) was added. After brefeldin A incubation, the cells were washed and labeled with a LIVE/DEAD Fixable Aqua Dead Cell Stain Kit, for 405 nm excitation (Invitrogen, Waltham, MA) to exclude dead cells from analysis. Following surface staining, cells were permeabilized with BD Cytofix/Cytoperm (BD Biosciences) and stained for intracellular cytokines (IFN-γ, TNF, IL-2, IL-10, and IL-17A). Intracellular staining was carried out for 1 h on ice. Data were collected and analyzed using FACSDIVA software and a FACSCanto flow cytometer (BD Biosciences).

Ear cell suspensions were stained for neutrophil markers, as described previously (39). Cells were stained with Annexin-V–allophycocyanin and 7-AAD (BD Biosciences), as recommended by the manufacturer. The data were collected and analyzed using FACSDIVA software and a FACSCanto flow cytometer (BD Biosciences). Controls were stained with buffer lacking Mg²⁺ and Ca²⁺.

Bone marrow neutrophils were collected and purified from C57BL/6 mice and gp91^phox-deficient mice. Bone marrow was flushed from femurs and tibias of mice; after centrifugation (400 × g, 10 min, 4°C), cells were resuspended in 2.5 ml of PBS. Cells were then laid onto a layer of 3 ml of Ficoll Histopaque-1077 (Sigma-Aldrich) and centrifuged at 400 × g for 30 min at room temperature. After centrifugation, we carefully discarded the upper layer and transferred the pellet to a clean 15-ml conical tube and washed the cells with PBS (400 × g, 10 min, 4°C). Bone marrow neutrophils were cultured at 1 × 10⁶ cells per milliliter in RPMI 1640 (Life Technologies) supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM l-glutamine. Cells were infected or not with metacyclic promastigotes of La-RFP (1:4) and washed after 1 h of infection (400 × g, 10 min, 4°C) to remove the parasites that were not phagocytosed. After 2 and 10 h of infection, we determined the frequency of apoptotic cells, as described above. In some experiments, neutrophils were treated with 100 μM H₂O₂ (31, 40, 41) 2 h before, immediately postinfection (p.i.), and 2 h p.i, when the samples were washed to remove extracellular parasites. At 10 h p.i., we determined the frequency of apoptotic and necrotic cells.

Statistical significance between groups was determined by the unpaired two-tailed Student t test or the Mann–Whitney U test. Two-way ANOVA with the Tukey posttest was performed when comparing three or more groups. Contingency table analysis employing the Fisher exact test was used to calculate significant differences between the number of ears with or without tissue loss. Statistical tests were performed using Prism software (GraphPad, La Jolla, CA).

All experiments were approved by the University of Calgary Animal Care Committee (Protocol number AC14-0142) in compliance with the Canadian Council for Animal Care or by the National Institutes of Allergy and Infectious Diseases Animal Care and Use Committee (Protocol number LPD68E) in compliance with the Animal Welfare Act and the Public Health Service Policy on Humane Care and Use of Laboratory Animals.

To address the role of NOX2 during L. amazonensis infection, we inoculated 5 × 10³ L. amazonensis metacyclic promastigotes into the ear dermis of C57BL/6 (WT) or gp91^phox deficient (gp91^phox−/−) mice. Although lesions were detectable in both groups as early as 3 wk p.i., gp91^phox−/− mice presented with significantly larger lesions starting at week 5 (Fig. 1A, left axis). We also observed open wounds at the lesion site in the ears of gp91^phox−/− mice, as indicated by the arrows in Fig. 1B. At 8 wk p.i., ears in both groups started to display tissue loss (Fig. 1A, right axis, 1B). Between 10 and 15 wk, gp91^phox−/− mice had an increase in the frequency of ears with tissue loss of between 30 and 45% compared with WT mice. Tissue destruction observed in gp91^phox−/− mice was so significant that, at 11 wk p.i., lesion size could not be accurately determined, and we consider the severity of tissue loss a more accurate reflection of disease at later time points (Fig. 1B). In contrast to high-dose inoculation of the footpad with 1 × 10⁶ L. amazonensis, in which disease is nonhealing but also nonprogressive (31), WT and gp91^phox−/− mice infected in the ear ultimately presented with progressive disease, as illustrated by loss of ear tissue, although this was more severe in the gp91^phox−/− group (Fig. 1A, right axis, 1B).

To determine whether the enhanced lesion diameters and pathology in gp91^phox−/− mice were associated with uncontrolled parasite replication due to the lack of gp91^phox-dependent ROS production, we determined the parasite load at the site of infection. Unexpectedly, and despite the differences observed in lesion size, tissue destruction, and the potential role of ROS in parasite killing, deletion of the gp91^phox subunit did not impact parasite numbers, as demonstrated by equivalent parasite loads in the ears of WT and gp91^phox−/− mice (Fig. 1C). Although tissue loss at later time points may influence parasite loads, this is not the case during the first 8 wk of infection. Therefore, enhanced parasite numbers at or before 8 wk of infection are not responsible for enhanced disease in gp91^phox−/− mice. Subsequent experiments were designed to determine the cause of the increased tissue necrosis observed in gp91^phox−/− mice.

Because gp91^phox−/− and WT mice presented with similar parasite loads over the course of infection, we analyzed whether the composition of phagocytic cells in the ear contributed to the necrotic phenotype observed in gp91^phox−/− mice. We restricted our analysis to the first 12 wk of infection, because of the extensive tissue loss at the 16-wk time point. Using flow cytometry (Fig. 2A, Supplemental Fig. 1), phagocytic myeloid cells were identified as neutrophils (CD11b⁺Ly6G⁺Ly6C^int), Ly6C^hi monocytes (or inflammatory monocytes) (CD11b⁺Ly6G⁻Ly6C^hi), DCs (CD11b⁺Ly6G⁻Ly6C⁻CD11c⁺MHCII⁺), and macrophages (CD11b⁺Ly6G⁻Ly6C⁻CD11c⁻F4/80⁺). Corresponding to the increased lesion size in gp91^phox−/− mice, we observed an increase in the total number of CD11b⁺ cells per ear at 8 and 12 wk p.i. in these mice (Fig. 2B). Remarkably, the increase in total phagocytic cell numbers was attributed solely to a 3-fold increase in neutrophil numbers in the ears of gp91^phox−/− mice compared with WT mice at 8 wk p.i. and an almost 5-fold increase at 12 wk p.i. (Fig. 2C). This increase corresponded with higher frequencies of neutrophils within the total CD11b⁺ population in gp91^phox−/− mice (Fig. 2D), with neutrophils representing >40% of CD11b⁺ cells by 8 wk p.i. Surprisingly, the number of other myeloid cell populations was not affected by the absence of the gp91^phox subunit (Fig. 2E, 2G, 2I). The increased number of neutrophils in gp91^phox−/− mice resulted in lower frequencies of Ly6C^hi monocytes and DCs within the total CD11b⁺ population at 8 and 12 wk p.i. (Fig. 2F, 2H). Our analysis of innate immune cells in the ear suggests that neutrophils are the predominant cells associated with the exacerbation of disease in gp91^phox−/− mice.

CD4⁺ T cells play a central role in immunity against Leishmania. Therefore, we also analyzed the possible contribution of CD4⁺ T cells to the enhanced disease observed in gp91^phox−/− mice. We followed the activation, proliferation, cytokine production, and transcription factor profile of CD4⁺ T cells at the inoculation site during infection. The total numbers of CD4⁺ T cells increased in both groups over time (Fig. 3A). Although gp91^phox−/− mice demonstrated a trend toward higher numbers at 4 and 8 wk p.i., this did not reach significance. gp91^phox−/− and WT mice showed a similar increase in activated CD4⁺CD44^hi T cell numbers (Fig. 3B) and recently proliferated cells based on Ki-67 expression (Fig. 3C), indicating that the increased number of neutrophils in gp91^phox−/− mice was not associated with a change in the CD4⁺ T cell response. Even in the presence of intense inflammation, gp91^phox−/− mice did not have increased numbers of CD25⁺Foxp3⁺CD4⁺ regulatory T cells (Fig. 3D). We also analyzed IFN-γ and TNF-α production by CD4⁺ T cells at the lesion site, because these cytokines are involved in the activation of infected phagocytes and parasite killing. Similar to parasite loads in the two groups of mice, gp91^phox−/− and WT mice had equivalent numbers of IFN-γ⁺ single cytokine–producing (Fig. 3E), IFN-γ⁺TNF-α⁺ double cytokine–producing (Fig. 3F), and IFN-γ⁺TNF-α⁺IL-2⁺ triple cytokine–producing (Fig. 3G) cells over the course of infection. In addition, analysis of IL-17 production, which has been shown to be directly correlated with neutrophil migration in some inflammatory processes (42, 43), revealed no association between the number of IL-17⁺CD4⁺ T cells and the increased neutrophil numbers in gp91^phox−/− mice (Fig. 3H). Similarly to our previous observations using L. major infection (44), we could not detect many IL-17a⁺CD4⁺ T cells. In contrast, we observed higher numbers of IL-10⁺CD4⁺ T cells at 12 wk p.i. in WT mice compared with gp91^phox−/− mice (Fig. 3I). Although decreased numbers of IL-10⁺CD4⁺ T cells in gp91^phox−/− mice at 12 wk p.i. were associated with increased pathology compared with WT mice, the late manifestation of this decrease suggests that it is not responsible for the enhanced pathology in gp91^phox−/− mice or the enhanced numbers of neutrophils starting at week 8 of infection.

NOX2 is known to have a role in PCD, also known as apoptosis (45–49). Therefore, we wished to determine whether the increased number of neutrophils and enhanced disease in the ears of gp91^phox−/− mice were associated with altered neutrophil apoptosis at the site of infection. Because neutrophils are preferentially infected compared with other phagocytic cell types during acute L. major inoculation (50), we used L. amazonensis metacyclic promastigotes expressing RFP to track the phenotype of infected cells (Fig. 4A). We analyzed the neutrophil population at 8 wk p.i., 2 wk before the onset of significantly increased tissue loss in gp91^phox−/− mice and WT mice (Fig. 4). We once again found increased total numbers and frequencies of neutrophils at the site of infection in gp91^phox−/− mice versus WT controls (Fig. 4B, 4C). In agreement with our limiting dilution analysis of parasite numbers reported in Fig. 1C, we observed similar total numbers of RFP⁺CD11b⁺ infected cells per ear in WT and gp91^phox−/− mice (Fig. 4D). However, analysis of the phenotype of infected cells revealed a significant shift toward infection of neutrophils in gp91^phox−/− mice (Fig. 4E). Therefore, the increased total number of neutrophils in gp91^phox−/− ears was associated with a shift in the phenotype of infected cells to neutrophils (Fig. 4E), but it did not change the total number of infected CD11b⁺ cells (Fig. 4D).

We used 7-aminoactinomycin D (7-AAD), a measure of membrane permeability, and annexin V, a measure of phosphatidylserine on the cell surface indicating apoptosis, to analyze PCD by neutrophils. Because Leishmania infection has been shown to influence neutrophil apoptosis (39), we first determined the necrotic (7-AAD⁺AnnV⁻), late apoptotic (7-AAD⁺AnnV⁺), apoptotic (7-AAD⁻AnnV⁺), or living (alive 7-AAD⁻AnnV⁻) phenotype of infected (RFP⁺) or uninfected (RFP⁻) neutrophils from the dermal site of infection at 8 wk p.i. (Fig. 4G). In WT mice, the infected neutrophil population had a significantly lower frequency of viable 7-AAD⁻AnnV⁻ (alive) cells and an increased frequency of late apoptotic and necrotic cells compared with uninfected neutrophils, demonstrating that infection was associated with increased cell death during chronic disease (Fig. 4G, left panel). Although neutrophils from gp91^phox−/− mice revealed a similar pattern of 7-AAD and annexin V staining, only death by necrosis was significantly increased in infected cells compared with uninfected cells.

We also directly compared the frequency of necrotic, late apoptotic, apoptotic, or living neutrophils among total, infected, or uninfected neutrophil populations in WT and gp91^phox−/− mice. Our analysis revealed that the frequency and number of necrotic cells were significantly higher in each population of gp91^phox−/− neutrophils versus WT neutrophils (Fig. 4H, 4I). The higher percentage of necrotic neutrophils, combined with the larger number of neutrophils per ear in gp91^phox−/− mice (Fig. 4B), resulted in a very large total number of necrotic neutrophils in gp91^phox−/− ears that was 4-fold higher than in WT animals (Fig. 4I, left panel). The increase in necrotic neutrophils in gp91^phox−/− versus WT mice was associated with a significant decrease in the frequency of apoptotic cells (Fig. 4H). The pattern of 7-AAD and annexin V staining suggests that neutrophils from gp91^phox−/− mice are more likely to be 7-AAD⁻AnnV⁻, having failed to initiate PCD, or necrotic, having undergone nonprogrammed cell death or failed PCD (Fig. 4H). Therefore, gp91^phox appears to be required for efficient apoptosis of neutrophils during L. amazonensis infection, resulting in an accumulation of viable and necrotic neutrophils at 8 wk p.i. The frequency of neutrophil necrosis was significantly enhanced upon infection (Fig. 4G) and gp91^phox deficiency (Fig. 4H), with infected gp91^phox−/− neutrophils displaying the highest frequency of 7-AAD⁺AnnV⁻ necrotic cells (mean 25.3%). These data demonstrate that the dramatic increase in the number of uninfected and infected necrotic neutrophils in the ears of gp91^phox−/− mice at 8 wk p.i. was associated with the enhanced disease in these mice at later time points.

At 8 wk p.i., the number of neutrophils in the dermis may be influenced by factors related to chronic infection rather than PCD. Because neutrophils are also the first cells to arrive at the site of inoculation during acute infection, we hypothesized that if altered cell death was the prime factor controlling neutrophil accumulation in gp91^phox−/− mice, then neutrophils should also accumulate during acute infection in these mice. We analyzed the recruitment, infection, and phenotype of neutrophils shortly p.i. with 5 × 10⁵ La-RFP. This higher inoculation dose was used to allow tracking of infected cells ex vivo (50). We observed an intense recruitment of neutrophils 10 h p.i. in gp91^phox−/− and WT mice, followed by a drop in the number and percentage of these cells at 36 and 60 h p.i. (Fig. 5A, left panel, and 5B), suggesting the kinetics of neutrophil recruitment is similar in these two strains. However, at 10 and 36 h p.i., there were significantly higher numbers and frequencies of neutrophils in gp91^phox−/− mice, comparable to what we observed during chronic infection. Similar to what has been reported previously by Pollock et al. (9) for chemically induced peritonitis, the increase in neutrophil numbers in the skin was not due to higher numbers of neutrophils available for recruitment from the blood of gp91^phox−/− mice (mean number of neutrophils per naive blood sample; 1275 (gp91^phox−/−, n = 6) versus 1457 (WT, n = 5), p = 0.74). Therefore, the enhanced frequencies and numbers of neutrophils at sites of L. amazonensis infection in gp91^phox−/− are not limited to chronic infection but are a general property of infection in these mice at early and late time points. Analysis of infected CD11b⁺RFP⁺ cells at acute time points following L. amazonensis infection revealed that the phenotype of infected cells over time was similar in WT and gp91^phox−/− mice (Fig. 5A, right panel). However, the higher number of total neutrophils in the site (Fig. 5B) was associated with more infected neutrophils in gp91^phox−/− mice at 10 and 36 h p.i. (Fig. 5C).

To determine whether neutrophil accumulation during acute infection was also associated with alterations in PCD, we analyzed the apoptotic and necrotic phenotype of infected and uninfected neutrophils at 36 h p.i. (Fig. 5D–F). Because the majority of neutrophils at 10 h p.i. are recently recruited, and neutrophil recruitment is significantly reduced at 60 h, we used 36 h as the most likely time point to observe the accumulation of cells with an altered cell death phenotype. Similar to our observations at 8 wk p.i., infection of neutrophils was associated with increased cell death in WT and gp91^phox−/− mice (Fig. 5E). However, in WT mice, dying RFP⁺ neutrophils were apoptotic or late apoptotic, with virtually no necrotic cells, whereas RFP⁺ neutrophils from gp91^phox−/− mice were more likely to be late apoptotic or necrotic versus RFP⁻ cells from the same site of infection. Similar to our observations at 8 wk p.i., comparison of cells from gp91^phox−/− and WT mice revealed a significantly smaller frequency of apoptotic cells among total and infected neutrophils in gp91^phox−/− mice, suggesting a failure of infection-induced apoptosis in gp91^phox−/− neutrophils (compare Figs. 4H and 5F). This was associated with a corresponding increase in the frequency of infected necrotic cells. In contrast, gp91^phox deficiency did not influence the pattern of cell viability among uninfected neutrophils at this early time point. Decreased apoptosis, combined with a corresponding increase in AnnV⁻7-AAD⁻ alive cells within the infected neutrophil population in gp91^phox−/− mice, likely contributes to the higher number and percentage of total neutrophils per ear at this time point in these mice (Fig. 5B).

The frequency of infected neutrophils that were necrotic was much lower at 36 h. p.i. versus 8 wk p.i. (compare Fig. 5F, middle panel, with Fig. 4H, middle panel). This may be due to efficient clearance of necrotic cells at 36 h p.i., when the total number of neutrophils is lower, versus the 8-wk time point. Therefore, to determine whether changes in neutrophil PCD were cell intrinsic or environmental, and to exclude the potential influence of in vivo cell clearance and/or loss of cells during tissue processing, we infected bone marrow–derived neutrophils in vitro (Fig. 6A, 6B). After 1 h of coincubation, excess parasites were removed, and neutrophils were phenotyped after an additional 2 or 10 h of incubation. At 2 h post–in vitro infection, L. amazonensis–infected neutrophils were significantly more likely to be apoptotic compared with uninfected cells from the same culture wells (Fig. 6B, % Apoptotic), once again confirming the observation that Leishmania infection of neutrophils induces apoptosis. Interestingly, and in contrast to the in vivo data, gp91^phox deficiency did not alter the frequency of infected apoptotic cells at this early time point, suggesting that early induction of phosphatidylserine exposure on the cell surface is gp91^phox−/− independent. However, at 10 h p.i., only infected neutrophils from gp91^phox−/− mice presented with dramatically decreased frequencies of apoptotic cells and increased frequencies of necrotic cells (Fig. 6B, % Necrotic), similar to our in vivo observations. Moreover, infected gp91^phox−/− neutrophils were once again more resistant to cell death during in vitro infection, as indicated by higher frequencies of 7-AAD⁻AnnV⁻ cells at the 10-h time point (Fig. 6B, % Alive), similar to what we observed in vivo (Figs. 4H, 5). Therefore, using an in vitro infection system, we were able to demonstrate impaired apoptosis and enhanced necrosis of infected neutrophils from gp91^phox−/− mice at 10 h post–in vitro infection, independent of unknown in vivo environmental cues. Importantly, we were able to use control neutrophils from parasite-free culture wells to demonstrate that, although induction of apoptosis at the 2-h time point is due to direct neutrophil infection, uninfected neutrophils from wells containing infected neutrophils also undergo enhanced apoptosis at later (10 h) time points compared with controls. Of note, gp91^phox−/− neutrophils alone or RFP⁻gp91^phox−/− neutrophils from neutrophil + L. amazonensis cultures did not undergo necrosis in vitro, suggesting the necrotic phenotype under these conditions is infection dependent.

Our observations suggest that, in the absence of gp91^phox-dependent ROS production, infected neutrophils die by necrosis rather than apoptosis. Therefore, we decided to examine whether administration of exogenous ROS in the form of H₂O₂ would reverse this phenotype. Treatment of neutrophils alone with 100 μM/ml H₂O₂ (31, 40) resulted in neutrophil PCD, and this occurred in WT and gp91^phox−/− groups (Fig. 7A), with the majority of neutrophils presenting with a late apoptotic phenotype and a negligible amount of necrosis. Following neutrophil + L. amazonensis coculture, infected gp91^phox−/− neutrophils once again presented with a high degree (∼35%) of infection-dependent necrosis (Fig. 7B, % Necrotic, upper panel), which was completely reversed following treatment with H₂O₂. This observation formally demonstrates that gp91^phox-dependent ROS production prevents the infection-dependent necrosis observed in gp91^phox−/− neutrophils in vitro and emphasizes the role of ROS in mediating PCD. In addition, under conditions of H₂O₂ treatment, RFP expression was associated with higher frequencies of late apoptotic cells versus uninfected cells (% Late Apoptotic H₂O₂ treated, WT and gp91^phox−/−, RFP⁺ ∼ 90% versus RFP⁻ ∼ 60%, p ≤ 0.0001; see Fig. 7B, % Late Apoptotic, upper panel versus lower panel). Infection was also associated with lower frequencies of apoptotic cells in the presence of H₂O₂ (% Apoptotic H₂O₂ treated, WT and gp91^phox−/−, RFP⁺ ∼ 5–10% versus RFP⁻ ∼30–35%, p ≤ 0.0001; see Fig. 7B, % Apoptotic, upper panel versus lower panel). This occurred with WT and gp91^phox−/− neutrophils, suggesting that, in neutrophil + L. amazonensis coculture wells, infection is associated with enhanced and/or faster progression of neutrophils to a late apoptotic phenotype in the presence of H₂O₂. Taken together, these observations suggest that ROS production is likely the mechanism by which gp91^phox mediates PCD following neutrophil infection.

Neutrophils from patients with chronic granulomatous disease have an impairment in ROS production and a deficit in CD95 expression (51, 52), a death receptor from the TNFR family that can trigger apoptosis following activation by CD95 ligand (53). We sought to determine whether neutrophils from gp91^phox−/− mice also showed a deficit in CD95 expression following infection with L. amazonensis. We did not find differences in the basal level of CD95 expression between WT and gp91^phox−/− neutrophils (Fig. 7C, Neutrophils alone). However, 10 h following infection with L. amazonensis, we observed a significant increase in the frequency of infected WT cells expressing CD95, which was not observed in uninfected WT cells or infected gp91^phox−/− neutrophils (Fig. 7C, right panel), suggesting that infection and gp91^phox are required for CD95 expression. Remarkably, treatment of neutrophils alone with exogenous H₂O₂ did not lead to an increase in the frequency of CD95⁺ neutrophils (Fig. 7D), whereas the combination of infection and exogenous H₂O₂ resulted in a huge increase in the frequency of CD95⁺ neutrophils in WT and gp91^phox−/− cells. Therefore, CD95 expression is likely dependent upon infection and gp91^phox-dependent ROS production, and this can be bypassed in gp91^phox−/− neutrophils via the provision of exogenous ROS in the form of H₂O₂.

In this study, we used low-dose intradermal infection of the ear to analyze the role of NOX2 during L. amazonensis infection in C57BL/6 mice. gp91^phox−/− mice presented with significantly enhanced pathology compared with WT mice, including increased lesion size and complete loss of the ear by 16 wk p.i., despite equivalent parasite loads. Although infection-related pathology was more severe in gp91^phox−/− mice, we ultimately observed progressive disease over time in both strains, confirming our previous observation that C57BL/6 mice are susceptible to infection with L. amazonensis following intradermal inoculation (3). Previously, we addressed the role of ROS from NOX2 in C57BL/6 mice following s.c. infection of the footpad with L. amazonensis (31) and found a limited role for gp91^phox in disease. However, L. amazonensis infection of the footpad does not result in progressive disease in WT mice, a phenotype that may be related to the failure of the s.c. injection site to replicate many of the characteristics of natural inoculation by the bite of an infected sand fly, including robust recruitment of neutrophils (50). Although needle inoculation of the mouse ear does not replicate all aspects of transmission by sand fly bite, it has been shown to be the most representative needle inoculation model available (50). Despite the differences between intradermal and s.c. inoculation sites, the degree to which gp91 deficiency increased the severity of disease following intradermal infection was surprising, given the comparable parasite loads compared with WT mice. In vitro studies have shown intracellular production of ROS following infection of macrophages (31, 54) and neutrophils (32, 33) by L. amazonensis and partial dependency on ROS for parasite killing in murine and human phagocytic cells (30, 55, 56). Moreover, the production of ROS during in vivo infection has been demonstrated indirectly by the presence of protein nitration by peroxynitrite (55, 57). Ultimately, our results in mice indicate a limited role for ROS derived from NOX2 in the control of parasite numbers in vivo. Rather, enhanced disease severity in gp91^phox−/− mice at later time points correlated with an accumulation of neutrophils at the site of infection. In contrast, the number of CD4⁺ T cells, their activation state, and their production of cytokines were similar. This occurred despite the fact that ROS have also been shown to promote changes in T cell responses. For example, ROS modify MHC class I and MHC class II Ag presentation through oxidation, thereby compromising T cell activation (10, 58). Moreover, ROS also can downregulate IL-12p70 production by DCs through regulation of p38-MAPK activity (59). Surprisingly, we did not observe any differences in T cell activation or production of cytokines between gp91^phox−/− or WT mice. Considering the relatively weak Th1 response observed following L. amazonensis infection and the significant impact of ROS deficiency on neutrophils and disease pathology, any potential effects of ROS deficiency on T cells may be negligible in vivo. Analysis of other innate immune cells also revealed that only the neutrophil population was significantly altered. The increase in neutrophil numbers correlated with a higher frequency of neutrophils within the total infected CD11b⁺ population without a change in the total number of infected cells, demonstrating that L. amazonensis preferentially infects, or is preferentially phagocytized by, neutrophils if they are present, similar to what we have reported during L. major infection (50).

Previous observations demonstrating a role for ROS in neutrophil apoptosis or PCD suggested that the accumulation of neutrophils and associated tissue necrosis in gp91^phox−/− mice may be due to changes in neutrophil cell death at the site of infection. For example, ROS have been shown to induce neutrophil PCD via the activation of caspases (60), and patients with chronic granulomatous disease, in which the NADPH oxidase NOX2 is defective, have neutrophils with impaired CD95-mediated and endogenous H₂O₂-dependent apoptosis (51). Lastly, H₂O₂-mediated apoptosis can be inhibited by catalase (51) and because catalase impairs spontaneous neutrophil apoptosis, it is suggested that endogenous NOX2-dependent H₂O₂ plays a critical role in the spontaneous apoptosis of mature neutrophils. These observations suggest that ROS play an important role in neutrophil apoptosis during the inflammatory processes, participating in an intimate and integrated way with apoptotic signaling cascades. Following L. amazonensis infection, we found that gp91^phox−/− neutrophils migrated to the site of infection but did not undergo normal cell death, as evidenced by the accumulation of necrotic cells at early and late time points. We suggest that, over time, this leads to an accumulation of live neutrophils that eventually die by necrosis rather than undergoing spontaneous apoptosis or cells that are stimulated to initiate apoptosis following infection cannot complete the process due to the absence of ROS. The release of cell contents during the necrotic processes would cause release of cytoplasmic proteins to the extracellular environment, including proteases that can degrade the extracellular matrix, leading to the tissue destruction (61) observed in the ears of gp91^phox−/− mice (Fig. 1A, 1B). Although neutrophil depletion using the Ly6G Ab would appear to be an attractive approach to formally demonstrate the role of neutrophils in enhanced pathology during chronic infection in gp91^phox−/− mice, the highly significant shift in the frequency of infected cells that are neutrophils compared with WT mice precludes this approach, because it is likely to differentially affect parasite loads in the two strains. In addition, our experience with neutrophil depletion during L. amazonensis infection has shown that depletion is invariably followed by a nonphysiological rebound in neutrophil numbers in excess of what is observed over the normal course of disease (62).

Analysis of RFP⁻ and RFP⁺ neutrophils following in vivo and in vitro infection revealed that the necrotic phenotype associated with gp91 deficiency was most prominent among infected cells. Infection by the Leishmania parasite has been extensively shown to influence neutrophil apoptosis. For example, ex vivo analysis of 7-AAD⁻ neutrophils from a dermal site of L. major infection revealed increased frequencies of AnnV⁺ cells within the infected versus uninfected neutrophil population in C57BL/6 mice (39). In contrast, in vitro studies concluded that L. major inhibited apoptosis through ERK1/2 phosphorylation and enhanced expression of Bcl-2 and Bfl-1 (63). L. braziliensis has also been reported to induce neutrophil apoptosis concomitant with ROS production (64) and L. amazonensis does induce macrophage apoptosis through activation of caspase-3, -8, and -9 (65). In our study, L. amazonensis infection of WT neutrophils resulted in higher frequencies of cells in the later stages of cell death compared with RFP⁻ cells in vivo (Fig. 4G, left panel, Fig. 5E, left panel) and in vitro (Fig. 6B, 2 h). Infected WT neutrophils also had higher frequencies of CD95⁺ cells (Fig. 7C), confirming our interpretation of PCD using 7-AAD and annexin V staining. In contrast, gp91^phox deficiency was characterized by a shift of infected neutrophils toward a late apoptotic or, more significantly, a necrotic phenotype. Remarkably, our in vitro analysis revealed this shift was time dependent. Analysis at 2 h post–in vitro infection revealed that the initial induction of apoptosis, as indicated by the expression of phosphatidylserine, associated with RFP⁺ infected cells was equivalent in WT and gp91^phox−/− neutrophils and, therefore, was ROS independent (Fig. 6B). However, at 10 h p.i., a high percentage of infected gp91^phox−/− neutrophils had shifted directly to a necrotic, rather than an apoptotic, phenotype compared with WT cells. Adding back exogenous H₂O₂ to infected gp91^phox−/− neutrophils was enough to revert the cell death pathway from necrosis to apoptosis in these cells (Fig. 7B), emphasizing the importance of ROS in carrying out PCD over time. The provision of H₂O₂ also rescued the expression of CD95 by infected gp91^phox−/− neutrophils. Notably, ROS production has been implicated in the induction of CD95-independent apoptosis via caspase-3 activation (41) and the release of the CD95 signaling pathway from negative regulation (66). These observations may explain why RFP⁻ neutrophils treated with H₂O₂ die by apoptosis without a corresponding increase in CD95⁺ expression, whereas, at the same time, exposure of infected neutrophils to H₂O₂ resulted in remarkably enhanced CD95 expression. In the latter case, infection and ROS appear to be required. Of interest, at 10 h post–in vitro infection, uninfected gp91^phox−/− neutrophils underwent apoptosis rather than the necrotic phenotype triggered by infection of neutrophils from the same culture wells, thereby demonstrating that bystander apoptosis can occur but that this “early” 7-AAD⁻AnnV⁺ apoptotic phenotype is largely ROS independent, much like the early 7-AAD⁻AnnV⁺ apoptotic phenotype observed in infected cells at 2 h p.i.

The exact mechanism by which gp91^phox−/− neutrophils undergo nonapoptotic cell death, presenting with the 7-AAD⁺AnnV⁻ phenotype that we have referred to as necrosis, is not known. It is possible that, in the absence of ROS, cells undergo necroptosis, pyroptosis, or some other caspase-independent form of cell death (67). Further studies addressing the mechanisms of ROS-dependent and -independent cell death after L. amazonensis exposure are necessary to define the effector molecules leading to apoptosis and necrosis under these conditions.

Our observations demonstrate that ROS from the NADPH oxidase NOX2 is not essential, in vivo, for parasite control. However, ROS plays a crucial role controlling the inflammatory response by mediating appropriate neutrophil apoptosis in response to infection. So far as we are aware, our observations are the first to demonstrate the role of appropriate PCD in a setting of chronic inflammation initiated by an infectious disease in the skin.

We thank Dr. Robin Yates (University of Calgary) for B6.129S6-Cybb^tm1Din mice.

The authors have no financial conflicts of interest.