CD8+ T cells express GzmB in leishmanial lesions but not in the dLNs. Cytotoxic CD8 T cells are pathogenic in cutaneous leishmaniasis (3, 4, 6, 7, 50–54), but their anatomical distribution and mechanisms that regulate their induction remain unclear. In this regard, parasites are found in both dLNs and skin lesions. To determine whether CD8+ T cell cytotoxic effector function is present in both tissues, we assessed the expression of the cytotoxic effector molecule GzmB. C57BL/6 mice were infected with Leishmania in the ear, and 2 weeks after infection, the frequency of GzmBpos antigen-experienced (CD44hi) CD8+ T cells was assessed by flow cytometry. While GzmB-expressing CD8+ T cells were abundant in the infected skin, GzmB expression was nearly absent in the dLNs (Figure 1A). To test whether other effector functions by CD8+ T cells are intact in dLNs, we looked for IFN-γ production, which is required for protection against the parasite (1). As demonstrated previously (55), we detected IFN-γpos CD8+ T cells in dLNs, whereas IFN-γ expression was diminished in lesions (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI177992DS1). Collectively, these data demonstrate a tissue-specific effector response, with CD8+ T cells in dLNs exhibiting a protective phenotype through the production of IFN-γ and CD8+ T cells in lesions exhibiting a pathogenic phenotype by expressing GzmB. Since it is well described that cytotoxic CD8+ T cells are pathogenic in cutaneous leishmaniasis (3–7, 50, 51, 56, 57), we sought to determine how CD8+ T cells become cytotoxic. First, we asked if GzmBpos CD8+ T cells were preferentially recruited to the skin. To address this question, we treated mice with FTY720, a sphingosine 1-phosphate receptor agonist that blocks the egress of T cells from the dLNs. As expected, we observed a defect in the recruitment of T cells to the lesions of FTY720-treated mice compared with vehicle control–treated mice, as observed by a significant decrease in the frequency of CD3pos cells in lesions (Supplemental Figure 1B). Although FTY720-treated mice had fewer CD8+ T cells in lesions, those CD8+ T cells recruited to the inflamed skin still expressed GzmB (Figure 1B). If GzmBpos CD8+ T cells are preferentially recruited to cutaneous lesions, GzmBpos CD8+ T cells should accumulate in the dLNs of FTY720-treated mice. We found no differences in the frequency of GzmBpos CD8+ T cells in the dLNs of mice treated with FTY720 or vehicle (Figure 1B), suggesting that rather than preferential recruitment, exposure to the lesion microenvironment induced a cytotoxic profile in CD8+ T cells. To directly test this, we infected C57BL/6 CD45.1 or CD45.2 congenic mice with Leishmania and purified CD8+ T cells from dLNs of CD45.2 donor mice (GzmBneg) 3 weeks after infection. Donor dLN CD8+ T cells were then transferred directly into the lesions of infected CD45.1 animals (Figure 1C, schematic). Notably, we found that GzmBneg CD8+ T cells became GzmB-expressing CD8+ T cells in the lesions (Figure 1C). In contrast, CD8+ T cells that migrated back to dLNs remained GzmBneg (Figure 1C). These results demonstrated that the lesion microenvironment triggered cytotoxic programs in effector CD8+ T cells.

CD8+ T cell function is tissue specific. (A) GzmB expression in CD8+ T cells from dLNs and lesions from C57BL/6 mice in week 2 after infection with L. major. Data are representative of more than 3 experiments with at least 4 mice per experiment. (B) GzmB expression in CD8+ T cells from dLNs and lesions from L. major–infected C57BL/6 mice treated with FTY720 or vehicle daily for 10 days, 12 days after infection. Data are from 2 independent experiments with 5 mice per group. (C) Schematic representation of the transfer of purified CD8+ T cells from dLNs of CD45.2 mice into the lesions of CD45.1 recipients, both infected for 3 weeks. Flow cytometric plot shows GzmB expression in donor CD8+ T cells from dLNs and lesions from recipient mice 48 hours after transfer. Data were combined from 2 independent experiments. (D–H) RNA-Seq analyses of antigen-experienced (CD44hi) CD8+ T cells purified from dLNs and lesions of L. major–infected mice in week 5. (D) PCA showing PC1 and PC2. (E) Volcano plot highlighting the top 5 DEGs. The orange line indicates an adjusted P value of 0.05. (F) Heatmap showing the expression of genes encoding transcription factors (TFs) associated with effector-like T cell functions and the FC between lesional and dLN CD8+ T cells. (G) GSEA (Biocarta) showing pathways enriched in dLNs (green) and lesions (purple). NES, normalized enrichment score. (H) Heatmap of hypoxia-related pathways enriched in lesional and dLN CD8+ T cells. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001, by 2-tailed Student’s t test (A , C, and F) and 1-way ANOVA (B).

To determine the mechanisms by which CD8+ T cell cytotoxic programs are induced in lesions, we performed RNA-Seq analysis on purified antigen-experienced CD8+ T cells from dLNs and lesions of mice infected with Leishmania for 5 weeks. Principal component analysis (PCA) showed that almost half (principal component 1 [PC1], 47%) of the differences on the whole transcriptional profiles were associated with the organ from which the CD8+ T cells were derived (Figure 1D). Differentially expressed gene (DEG) analysis with thresholds of a FDR of less than 0.05 and a fold change (FC) of greater than 1.5 between tissues revealed overexpression of 118 genes in CD8+ T cells from lesions compared with dLNs and 26 DEGs overexpressed in dLNs compared with lesions (Figure 1E, Supplemental Table 1, and Supplemental Figure 1, C and D). Gzmb was the top DEG in CD8+ T cells from lesions compared with dLNs, with an FDR of 0.002 (Figure 1E and Supplemental Figure 1C). Several transcription factors play important roles in effector CD8+ T cell biology, including BATF, ID-2, IRF4, Stat4, T-bet, Zeb2, and Blimp-1 (encoded by Prdm1) (58–60), and we found that Id2, Irf4, and Prdm1 expression levels were significantly higher in lesions than in dLNs and that Prdm1 had the most significant FC between tissues (>100 FC higher in lesions) (Figure 1F). To further investigate the signals received by CD8+ T cells within lesions, we performed gene set enrichment analysis (GSEA) (Supplemental Table 2), which revealed an array of immune-related pathways, including a hypoxia-related signature (Figure 1G). Hypoxia is a common feature of many inflammatory disorders and can alter CD8+ T cell function (38, 61). To determine whether hypoxia is a critical signature of lesional CD8+ T cells, we assessed other pathway databases, as well as the Harris et al. hypoxia-specific pathway described in ref. 62, to see if there was enrichment for a hypoxia signature in lesional compared with dLN CD8+ T cells. We found enrichment for multiple hypoxia-related signatures in CD8+ T cells in lesions (Figure 1H and Supplemental Table 2). These data indicate that, after their exit from dLNs, CD8+ T cells recruited to cutaneous leishmaniasis lesions were exposed to a hypoxic environment, inducing the expression of key cytotoxicity effector mediators, including Prdm1.

Cutaneous leishmaniasis lesions are hypoxic and alter CD8+ T cell function. Normal skin is naturally low in O2 (63, 64), and since inflammatory environments are frequently hypoxic (65), we hypothesized that the lesions would be hypoxic. To evaluate hypoxia in Leishmania-infected lesions, we used pimonidazole, a 2-nitroimidazole reporter molecule that is reductively activated and forms covalent bonds with macromolecules in hypoxic cells. Two weeks after Leishmania infection, we assessed pimonidazole staining in infected and contralateral ears by confocal microscopy. As expected, we found pimonidazole staining within the epidermis, dermis, and hair follicles of naive mice (Figure 2, A and B, top images). Within lesions, we observed enlargement of the epidermis and dermis associated with robust pimonidazole staining (Figure 2, A and B, bottom images). Pimonidazole staining was absent from ulcerated areas, suggesting tissue heterogeneity in the hypoxic state of lesions. Representative images at lower (Figure 2A and Supplemental Figure 2A) and higher (Figure 2B and Supplemental Figure 2B) magnification show the differences in pimonidazole staining distribution between the naive and infected tissues.

Hypoxia induces GzmB and Prdm1 expression in CD8+ T cells. (A and B) Immunofluorescence staining and confocal microscopy of horizontal sections from lesions of C57BL/6 mice infected with L. major for 2 weeks. Mice received pimonidazole 1 hour before euthanasia. The top panels show contralateral ears (naive skin), and the bottom panels show lesions from infected ears. Pimonidazole is shown in green and nuclear DAPI staining in blue. Representative images are from 2 naive ears and 3 infected ears. Scale bars: 200 μm (A) and 100 μm (B). (C) Partial pressure of oxygen (pO2) in mmHg in dLNs and lesions of C57BL/6 mice infected with L. major for 2 weeks. Data were combined from 2 independent experiments. (D) GzmB expression in CD8+ T cells from dLNs of L. major–infected C57BL/6 mice; cells were cultured with DMOG or vehicle. (E) FC of Prdm1 mRNA over the average expression of vehicle-treated cells measured by qRT-PCR in DMOG- or vehicle-treated purified CD8+ T cells. (F) GzmB expression in CD8+ T cells from dLNs of C57BL/6 mice infected with L. major; cells were cultured under normoxia (21% O2) or hypoxia (1% O2). (G) FC of Prdm1 mRNA over the average expression of normoxic cells measured by qRT-PCR in purified CD8+ T cells cultured in normoxia or hypoxia. (D–G) Data shown are representative of more than 3 experiments with at least 4 mice per experiment. *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001, by 2-tailed Student’s t test (C–G).

To compare O2 levels in lesions and dLNs, we infected mice with Leishmania, and 2 weeks after infection, mice received Oxyphor G4, a phosphorescent probe quenched by O2 (66). Infected skin showed a significant decrease in O2 compared with dLNs (Figure 2C). These data directly demonstrated that skin lesions present a hypoxic microenvironment to CD8+ T cells upon their recruitment from dLNs.

We tested whether exposure to hypoxia is sufficient to promote GzmB expression in dLN CD8+ T cells. We isolated dLN cells from infected mice and stimulated them with anti-CD3 and anti-CD28 antibodies in the presence of dimethyloxalylglycine (DMOG), a compound that mimics hypoxia at normal O2 tension. As shown in Figure 2D, DMOG induced the expression of GzmB in CD8+ T cells, further demonstrating that mimicking hypoxia promotes upregulation of their cytolytic program. Notably, DMOG treatment also increased Prdm1 mRNA expression compared with vehicle-treated CD8+ T cells (Figure 2E). We confirmed these findings by exposing CD8+ T cells to 1% O2 and found that hypoxia induced GzmB and Prdm1 expression in CD8+ T cells (Figure 2, F and G). We also tested whether hypoxia had a similar effect on naive and antigen-experienced CD8+ T cells (Supplemental Figure 3A) and we found that hypoxia enhanced GzmB expression only in antigen-experienced CD8+ T cells (Supplemental Figure 3B). These results raise the question of whether CD8+ T cells in vivo require Leishmania antigen to express GzmB or if sustained hypoxia alone is sufficient to promote GzmB. To test this, we infected mice with a T cell conditional deletion of Von Hippel–Lindau (referred to here as VHLcKO mice), which is necessary for the degradation of hypoxia-inducible factors (HIFs). Consequently, VHLcKO mice have HIF stabilization even in normoxic conditions. VHL deficiency did not increase GzmB expression in naive skin (Supplemental Figure 3C), suggesting that without Leishmania, there was no induction of GzmB when cells are forced to respond to hypoxia. In infected ears, VHL deletion increased GzmB expression in CD8+ T cells (Supplemental Figure 3C). One possible explanation is that at least a portion of CD8+ T cells in lesions were not exposed to hypoxia, which agrees with our observation that there were varied degrees of pimonidazole staining in the infected skin (Figure 2A). Another possibility is that CD8+ T cells newly recruited from the blood were not fully hypoxic and that forcing an immediate hypoxic response induced GzmB in these cells. Nevertheless, these data showed that antigen stimulation was necessary for CD8+ T cells to express GzmB in hypoxic conditions.

While the primary role of CD4+ T cells in our model is the production of IFN-γ and protection (1), CD4+ T cells can also be cytotoxic (67). Therefore, we tested whether hypoxia (1% O2) induces GzmB expression in CD4+ T cells from the dLNs of infected mice stimulated with anti-CD3 and anti-CD28 antibodies. Naive and antigen-experienced CD4+ T cells expressed more GzmB when exposed to hypoxia (Supplemental Figure 3D). However, GzmB expression was lower in CD4+ T cells than in CD8+ T cells (Supplemental Figure 3, B and D), suggesting that the induction of GzmB by hypoxia was shared by CD4+ and CD8+ T cells; however, the magnitude of the response was greater in CD8+ T cells. Given that CD4+ T cells were also exposed to the hypoxic microenvironment of lesions, we analyzed GzmB expression in CD4+ T cells in dLNs and lesions. We found that CD4+ and CD8+ T cells had a similar profile, with higher GzmB expression in lesions, although CD4+ T cells expressed significantly less GzmB than did CD8+ T cells (Supplemental Figure 3E). Collectively, these data demonstrate that the hypoxic state of lesions combined with antigen exposure stimulated the differentiation of recruited CD8+ T cells into cytotoxic effectors, with a moderate effect on the development of cytotoxic CD4+ T cells.

Blimp1 expression is restricted to GzmB-expressing CD8+ T cells. Blimp-1–deficient CD8+ T cells produce less GzmB (68), so we next sought to test cause-effect relationships between hypoxia, Blimp-1, and cytotoxicity. To this end, we infected Blimp-1 yellow fluorescent protein (YFP) reporter mice with Leishmania and assessed Blimp-1 expression in CD8+ T cells. We found that Blimp-1 was highly expressed in lesional CD8+ T cells but not in CD8+ T cells from the dLNs (Figure 3A). We also found that Blimp-1 expression was significantly higher in CD8+ T cells that expressed GzmB (Figure 3B), with similar results in CD4+ T cells (Supplemental Figure 4, A and B), suggesting a link between Blimp-1 and cytotoxicity in both T cell subsets. Since Blimp-1 regulates IL-10 production in CD4+ T cells (69), we measured the expression levels of IL-10 in dLNs and lesions of infected mice and found higher expression of IL-10 in lesional CD4+ and CD8+ T cells compared with dLNs (Supplemental Figure 4, C and D). In lesions, CD4+ and CD8+ T cells had higher Blimp-1 expression in IL-10pos cells than did their IL-10neg counterparts (Supplemental Figure 4, E and F), suggesting a strong relationship between Blimp-1 expression and IL-10, as previously described (69).

Blimp-1 expression is induced by hypoxia within skin lesions. (A) Expression of Blimp-1 in CD8+ T cells from dLNs and lesions from Blimp-1 YFP reporter mice infected with L. major for 1 week. Data from 2 independent experiments combined. (B) GzmB and Blimp-1 YFP expression in CD8+ T cells from lesions from WT or Blimp-1 reporter mice infected with L. major for 1 week. (C) Schematic representation of ODDCre Blimp-1fl/fl mice. (D) ODDCre Blimp-1fl/fl mice had been infected with L. major for 1 week when daily tamoxifen injections commenced. Mice were euthanized 2 weeks after infection. GzmB expression of lesional CD8+ T cells in tamoxifen-treated and untreated mice. Data were combined from 2 independent experiments. *P ≤ 0.05, **P ≤ 0.01, and ****P ≤ 0.0001, by 2-tailed Student’s t test (A, B, and D).

Finally, to assess whether GzmB expression in lesional CD8+ T cells resulted from hypoxic induction of Blimp-1, we used mice expressing a fusion protein bearing an oxygen-dependent degradation (ODD) domain from human HIF1A fused to a tamoxifen-inducible Cre recombinase gene (Cre/ERT2) (70). When expressed under normoxic conditions, the O2CreER fusion protein was degraded rapidly but was stabilized by hypoxia when combined with tamoxifen injection. These mice were crossed with Blimp-1fl/fl mice (referred to here as ODDCre Blimp-1fl/fl mice) to generate mice in which Blimp-1 expression is deleted specifically in hypoxic cells upon tamoxifen injection (Figure 3C). Since the skin has significantly less O2 than dLNs, these mice provide lesion-specific deletion of Blimp-1, thus ensuring the preservation of Blimp-1–dependent development of effector T cells in dLNs (normoxic) and their recruitment to the inflamed tissue. Leishmania-infected ODDCre Blimp-1fl/fl mice were treated with tamoxifen or left untreated for 1 week, and GzmB expression was evaluated. We found that the specific deletion of Blimp-1 in hypoxic cells significantly decreased GzmB expression in lesional CD8+ T cells compared with untreated mice (Figure 3D). Together, these results demonstrate that the upregulation of Blimp-1 by the hypoxic environment led to increased GzmB expression in CD8+ T cells.

A subset of GzmBpos CD8+ T cells express markers of exhaustion. Both hypoxia and Blimp-1 can induce the expression of many coinhibitory receptors involved in the exhaustion program (68, 71–74). To verify whether CD8+ T cells in lesions are exhausted, we reanalyzed a publicly available bulk RNA-Seq data set of 3 subsets of activated (stem-like CD101negTim3neg, transitory CD101negTim3pos and exhausted CD101posTim3pos) PD-1pos CD8+ T cells and naive CD8+ T cells in the context of chronic lymphocytic choriomeningitis virus (LCMV) infection. On the basis these subsets, we generated 3 specific gene signatures and compared them with our data set of CD8+ T cells isolated from the dLN and lesions (Figure 1, D–H). CD8+ T cells isolated from the lesions were enriched for all 3 signatures compared with CD8+ T cells isolated from the dLN (Supplemental Figure 5A). A subset of genes associated with exhaustion was higher in cells from lesions compared with dLNs, such as Pdcd1 (encoding for PD-1), Lag3 (encoding for Lag3), and Havcr3 (encoding for Tim3), while other genes — Tox, Cx3cr1, or Tcf7 — were not differentially represented in either tissue (Supplemental Figure 5B). We confirmed increased protein expression of PD-1, Lag3, and Tim3 in CD8+ T cells from lesions 2 weeks after infection (Supplemental Figure 5C). We sought to determine whether there was preferential expression of PD-1, Lag3, or Tim3 in CD8+ T cells that were GzmBpos and, therefore, had higher Blimp-1 expression and found that only Tim3 was preferentially expressed in GzmB-expressing CD8+ T cells (Supplemental Figure 5D). Collectively, these data suggest that coinhibitory molecule expression was a feature of some CD8+ T cells present in lesions and that a small subset of GzmBpos CD8+ T cells were either terminally exhausted or on their way to becoming so.

Blimp-1 expression is necessary for CD8+ T cells to mediate disease. On the basis of our data, we predicted that Blimp-1 expression is required for cytotoxic CD8 T cell–mediated disease in leishmaniasis. Blimp-1 plays a vital role in CD4+ T cell subsets (69, 75); hence, total T cell deletion of Blimp-1 would complicate interpretations. Therefore, we used our well-characterized mouse model of chronic leishmaniasis, in which severe disease develops following Leishmania infection of RAG–/– mice reconstituted with CD8+ T cells (3, 4, 55, 76). In this model, pathology is dependent on the ability of CD8+ T cells to be cytotoxic (3). RAG–/– mice infected with Leishmania were reconstituted with WT or CD8+ T cells lacking Blimp-1 expression (referred to here as Blimp-1cKO mice) or control animals receiving no T cells. As described previously (3), RAG–/– mice reconstituted with WT CD8+ T cells developed severe pathology characterized by GzmB-expressing CD8+ T cells in lesions, whereas RAG–/– mice that received no cells showed no signs of pathology (Figure 4A). Importantly, RAG–/– mice reconstituted with Blimp-1cKO CD8+ T cells had minimal disease (Figure 4A). As expected, there were similar numbers of parasites (Figure 4B) and a similar frequency of CD8+ T cells (Figure 4C) in lesions of RAG–/– mice reconstituted with WT and Blimp-1cKO CD8+ T cells. Notably, GzmB expression was significantly diminished in cells lacking Blimp-1 (Figure 4D). Collectively, these data indicate that Blimp-1 expression was driven by the hypoxic microenvironment of the lesion, triggering GzmB expression and CD8+ T cell–driven pathology.

Blimp-1 expression is required for CD8+ T cell–mediated disease. RAG−/− mice were infected with L. braziliensis and reconstituted with purified CD8+ T cells from WT or Blimp-1cKO mice. (A) Ear thickness and (B) parasite numbers in lesions at 7 weeks of infection. (C) CD8+ T cell frequency and (D) GzmB expression by CD8+ T cells in lesions were assessed directly ex vivo by flow cytometry 7 weeks after infection. Data represent 3 individual experiments with 3–5 mice per group. *P ≤ 0.05 and **P ≤ 0.01, by 2-tailed Student’s t test (A, C, and D).

Neutrophils generate the hypoxic microenvironment of cutaneous leishmaniasis lesions. Hypoxia occurs in tissues when O2 supply does not meet demand, including scenarios characterized by defective tissue vascularization or increased demand by cells present within a tissue (38). Leishmania-infected lesions are highly vascularized (77), so we hypothesized that O2 consumption by inflammatory cells recruited to lesions promotes hypoxia. Neutrophils are the first cells recruited to the skin upon Leishmania infection (78), but their recruitment does not control L. major parasites. Indeed, their chronic presence is associated with worsening disease (3, 4, 79–83). Therefore, we tested whether neutrophil recruitment contributed to the hypoxic state of inflamed skin. Leishmania-infected mice were injected with pimonidazole 1 hour before euthanasia at multiple time points after infection: (a) soon after parasite challenge (2 and 5 hours), (b) when lesions started to develop and CD8+ T cells were present and expressed GzmB (2 weeks), and (c) when lesions began to heal (9 weeks). As expected (78), neutrophils were quickly recruited after infection and decreased in frequency when lesions resolved (Figure 5A). Surprisingly, neutrophil (CD11bposLy6Gpos) staining by pimonidazole was significantly lower compared with other myeloid cells (CD11bposLy6Gneg) in the skin at all time points analyzed (Figure 5B). To confirm these observations, lesions from pimonidazole-injected mice were stained for Ly6G, revealing intense neutrophil recruitment in ulcered regions in the epidermis and clusters of neutrophils within the skin dermis (Figure 5C and Supplemental Figure 6A). Notably, regions without neutrophils had more pimonidazole staining than did those containing neutrophil clusters. We also observed that pimonidazole staining in regions surrounding neutrophil clusters was more intense. A quantitative assessment of pixel intensities (Supplemental Figure 6B) revealed a negative correlation between pimonidazole and Ly6G expression (Figure 5D). Higher magnification of regions with intense neutrophil recruitment in 2 samples showed a lack of pimonidazole staining in neutrophils (Figure 5E and Supplemental Figure 6C).

Neutrophils consume oxygen from lesions and stimulate GzmB expression in CD8+ T cells. (A and B) C57BL/6 mice infected with L. major received pimonidazole 1 hour before euthanasia at 2 hours, 5 hours, 2 weeks, and 9 weeks. (A) Frequency of neutrophils in lesions and (B) pimonidazole expression in neutrophils (CD11b+Ly6Gpos) and other myeloid cells (CD11b+Ly6Gneg) in lesions. (C) Representative confocal microscopy image of Ly6G (pink) and pimonidazole (green) staining in skin infected for 2 weeks. Scale bar: 200 μm. (D) Pixel intensity of pimonidazole and Ly6G based on 27 regions (Supplemental Figure 6B). (E) Two representative confocal microscopy images of nuclear staining (DAPI, blue), Ly6G (pink), and pimonidazole (green) in lesions infected for 2 weeks. Scale bar: 100 μm. (F–I) C57BL/6 mice infected with L. major were injected every 3 days with anti-Ly6G (clone 1A8) antibody or isotype control until euthanasia 2 weeks after infection. (F) Frequency of neutrophils in lesions, (G) pimonidazole staining in CD8+ T cells in lesions, and (H) GzmB frequency and MFI in CD8+ T cells in lesions. (I) Lesion size and number of parasites in lesions. (J) Lesion size and number of parasites in lesions in WT and Cybb–/– mice infected with L. major for 2 weeks. (K) GzmB frequency and MFI in CD8+ T cells in the lesions of WT or Cybb–/– mice. *P ≤ 0.05, ***P ≤ 0.001, and ****P ≤ 0.0001, by 2-tailed Student’s t test (B, F, H, and K) and 1-tailed Student’s t test (G).

These results suggest that areas surrounding neutrophils were relatively hypoxic. Accordingly, we hypothesized that neutrophils are responsible for promoting the hypoxic state of the skin by competing for oxygen, thereby controlling CD8+ T cell effector function. To test this, we depleted mice of neutrophils for the first 2 weeks of infection using anti-Ly6G antibody (Figure 5F) and assessed pimonidazole and GzmB expression in CD8+ T cells. We found that neutrophil depletion decreased pimonidazole expression (Figure 5G) and the frequency and MFI of GzmB-expressing CD8+ T cells compared with isotype control–treated mice (Figure 5H), without changing lesion size or parasite numbers (Figure 5I). These data suggest that neutrophils consumed O2 in lesions, altering the phenotype of CD8+ T cells.

Activated neutrophils assemble NADPH oxidase to produce ROS, which requires O2 (84). Thus, we tested whether O2 consumption to generate ROS affected CD8+ T cell function. For this purpose, we compared the ability of lesional CD8+ T cells to express GzmB in mice sufficient (WT) or deficient in the gp91phox subunit of the NADPH oxidase (Cybb–/–). Although there was no change in lesion size or parasite numbers in the skin at week 2 after infection (Figure 5J), we found significantly lower frequencies of GzmB-expressing CD8+ T cells in Cybb–/– mice than in WT (Figure 5K), forging a link between NADPH oxidase–dependent ROS production and the development of pathogenic CD8+ T cells in cutaneous leishmaniasis lesions.

The magnitude of hypoxia correlates with the presence of neutrophils in human cutaneous leishmaniasis lesions. To determine whether hypoxia is a feature of human disease, we compared transcriptional signatures from bulk RNA-Seq analysis performed in whole blood and total skin from healthy individuals and patients with cutaneous leishmaniasis using previously published data sets (54, 85) (Figure 6A). To estimate the levels of hypoxia gene expression in clinical samples, we performed single-sample GSEA using the Harris et al. hypoxia gene signature (62) and referred to this variable as the “hypoxia score” (Supplemental Table 3). While there was no difference in the hypoxia scores between blood from patients with cutaneous leishmaniasis and healthy individuals (Figure 6B and Supplemental Table 4), the cutaneous leishmaniasis lesions were significantly more hypoxic than healthy skin (Figure 6C and Supplemental Table 3). Importantly, we observed that the degree of the hypoxia-related gene expression in patients’ lesions was variable. To understand the biological implication of this variability, we performed unsupervised hierarchical clustering (HC) to classify lesions and healthy skin samples according to their hypoxic gene expression (Figure 6D). HC analysis revealed 2 groups of lesion samples: (a) “baseline” (n = 22), with a hypoxia phenotype comparable to that of control skin samples, and (b) “hypoxic” (n = 12), with a high hypoxic phenotype relative to the entire cohort (Figure 6, C and D). The remaining lesions had an intermediate phenotype. PCA revealed that hypoxic lesions were robustly segregated from the baseline counterparts: permutational multivariate ANOVA (PERMANOVA) (Pr), Pr(> F) = 0.001) (Figure 6E). Gene Ontology performed for genes overexpressed in hypoxic versus baseline lesions (FDR > 0.01 and FC = 1.5 thresholds) revealed significant enrichment for neutrophil chemotaxis and signaling (Supplemental Tables 5 and 6). Indeed, estimation of cell abundances from this unstructured bulk RNA-Seq data revealed an increase in neutrophil counts in hypoxic versus baseline lesions (Figure 6F). The top DEGs between hypoxic compared with baseline lesions included genes associated with neutrophil chemotaxis and survival, such as CXCL5, CXCL8, CXCR1, CSF3, and genes encoding proinflammatory cytokines such as IL1B, IL6, and OSM (Figure 6G and Supplemental Table 5). Therefore, our findings in mice are supported by our analysis of patients’ samples, linking the presence of neutrophils to the degree of hypoxia in lesions.

The degree of hypoxia-related gene expression is associated with the presence of neutrophils in patients. (A) Schematics of data collection for the RNA-Seq analysis performed in whole blood and total skin punch biopsies in healthy individuals and patients with cutaneous leishmaniasis. (B) Hypoxia score calculated in blood from healthy individuals (HI) (n = 14) and patients with cutaneous leishmaniasis (CL) (n = 50) and in (C) human intact skin (n = 6) and cutaneous leishmaniasis lesions from patients (n = 51) based on the Harris et al. (62) hypoxia gene signature. (D) Unsupervised hierarchical clustering classification of human skin samples in “baseline” and “hypoxic” lesions according to their hypoxia gene phenotypes. (E) PCA of baseline and hypoxic cutaneous leishmaniasis lesions. (F) Neutrophil abundances estimated by the MCP-counter method. The Wilcoxon test was used for statistical analysis. (G) Scatter plot showing differential gene expression analysis between hypoxic versus baseline lesions (FDR > 0.01 and FC = 1.5). In blue are the top DGEs from this analysis. **P ≤ 0.01 and ***P ≤ 0.001, by 2-tailed Student’s t test (C and F). HI, healthy individuals.