Rapid loss of group 1 innate lymphoid cells during blood stage Plasmodium infection‡

Abstract Objectives Innate lymphoid cells (ILCs) share many characteristics with CD4+ T cells, and group 1 ILCs share a requirement for T‐bet and the ability to produce IFNγ with T helper 1 (Th1) cells. Given this similarity, and the importance of Th1 cells for protection against intracellular protozoan parasites, we aimed to characterise the role of group 1 ILCs during Plasmodium infection. Methods We quantified group 1 ILCs in peripheral blood collected from subjects infected with with Plasmodium falciparum 3D7 as part of a controlled human malaria infection study, and in the liver and spleens of Pc AS‐infected mice. We used genetically‐modified mouse models, as well as cell‐depletion methods in mice to characterise the role of group 1 ILCs during Pc AS infection. Results In a controlled human malaria infection study, we found that the frequencies of circulating ILC1s and NK cells decreased as infection progressed but recovered after volunteers were treated with antiparasitic drug. A similar observation was made for liver and splenic ILC1s in P. chabaudi chabaudi AS (Pc AS)‐infected mice. The decrease in mouse liver ILC1 frequencies was associated with increased apoptosis. We also identified a population of cells within the liver and spleen that expressed both ILC1 and NK cell markers, indicative of plasticity between these two cell lineages. Studies using genetic and cell‐depletion approaches indicated that group 1 ILCs have a limited role in antiparasitic immunity during Pc AS infection in mice. Discussion Our results are consistent with a previous study indicating a limited role for natural killer (NK) cells during Plasmodium chabaudi infection in mice. Additionally, a recent study reported the redundancy of ILCs in humans with competent B and T cells. Nonetheless, our results do not rule out a role for group 1 ILCs in human malaria in endemic settings given that blood stage infection was initiated intravenously in our experimental models, and thus bypassed the liver stage of infection, which may influence the immune response during the blood stage. Conclusion Our results show that ILC1s are lost early during mouse and human malaria, and this observation may help to explain the limited role for these cells in controlling blood stage infection.


INTRODUCTION
Innate lymphoid cells (ILCs) resemble T helper (Th) cells in terms of their characteristic transcription factors and functions. 1 Groups 1, 2 and 3 ILCs make up the ILC repertoire, and these groups are similar to Th1, Th2 and Th17 cells, respectively. In contrast to Th cells of the adaptive immune system, ILCs do not express antigen-specific T-cell receptors. 2 ILCs have both protective and pathogenic roles in infectious and inflammatory diseases. 3,4 However, a recent study has suggested that ILCs are redundant in the presence of a competent B and T-cell response in humans. 5 Group 1 ILCs consist of conventional natural killer (cNK) cells and ILC1s. 1,6 These cell subsets share common developmental requirements for the transcription factor T-bet and their ability to produce the pro-inflammatory cytokine IFNc. 6 However, the relationship between these cells is still widely debated. Liver ILC1s are a specialised subset of group 1 ILCs also known as tissue-resident NK (trNK) cells. 7 Other ILC1s have also been identified in the uterus, 8 spleen, 9 salivary gland, 10 kidney, 11 adipose tissue 12 and gastrointestinal tract. 9 These different tissue-resident subsets have unique cell surface phenotypes and functions. 4,9,13 Malaria is a globally important infectious disease caused by protozoan parasites belonging to the genus Plasmodium. Following infection, a Th1dependent immune response can develop in the mammalian host, aiding clearance of parasites via IFNc-dependent mechanisms. 14,15 IFNc production by antigen-specific CD4 + T cells during P. chabaudi chabaudi AS (PcAS) infection in mice has been reported, 16 and increased parasite growth was observed following IFNc neutralisation. 17 The relationship between IFNc production and control of parasite growth has also been reported in humans during blood stage Plasmodium falciparum (Pf) infection. 18 Additionally, we recently showed an inverse correlation between IFNc levels and parasite burden during controlled human malaria infection (CHMI) with Pf. 19 However, the secretion of IFNc contributes to an inflammatory environment that can also contribute to pathology. 18,[20][21][22][23][24] While the adaptive immune response generated in response to Plasmodium infection has been well characterised, less is known about the innate immune response following infection. Early studies revealed that the depletion of NK cells with anti-asialo GM1 antibody resulted in increased parasitaemia during PcAS infection. 25 However, the effects of this treatment on dendritic cell (DC) function 26 and the depletion of other cell subsets, such as basophils, 27 may impact the interpretation of these results. In fact, depletion of NK cells in mice via administration of anti-NK1.1 antibody resulted in no effect on the course of P. chabaudi adami 556KA infection. 28 However, evidence for direct interactions between human NK cells and Pf parasitised red blood cells (pRBC) in vitro, which stimulates production of IFNc, has been reported. 29 Given that group 1 ILCs function like Th1 cells, and little is known about their roles during Plasmodium infection, we examined these cells, as well as the more well-studied innate-like T cells (including cd T cells, 28 invariant natural killer T (iNKT) cells 30,31 and mucosal-associated invariant T (MAIT) cells 32 ) in volunteers infected with Pf in CHMI studies. Concurrently, we also investigated the role of ILC1s in C57BL/6J mice infected with PcAS, which causes a chronic, nonlethal infection. 33 We report that cNK cells and ILC1s had a limited role in controlling peripheral blood parasitaemia in mice infected with PcAS. However, we found a loss in circulating NK cells and ILC1s in volunteers participating in CHMI studies with Pf, which was independent of parasite burden. A similar loss of ILC1s was also observed in the liver and spleen of mice infected with PcAS. Furthermore, we report a novel NK1.1 + NKp46 + population that expressed both the ILC1 marker CD49a and the cNK marker DX5 (CD49b) in the livers of PcAS-infected mice, suggesting plasticity between these cell populations.

RESULTS
The frequency and number of ILC1s declined during blood stage Pf infection NK and cd T cells produce IFNc in response to Pf infection. [34][35][36] To gain a better understanding of IFNc production by innate immune cells, including more recently identified ILC1s and innate-like T cells, we examined these cell populations during an experimentally induced blood stage malaria infection in healthy volunteers with no prior exposure to malaria or residence in malariaendemic regions. 37,38 Human PBMCs were isolated from blood drawn prior to infection (day 0) and at 7 days postinfection (p.i.), prior to drug treatment ( Figure 1a). We then identified group 1 ILCs (CD56 À CD127 + T-bet + ILC1s and NK cells), group 1 ILC-like cells (CD56 + CD127 + T-bet + ) (Figure 1b figure 1B).
We found the frequency and number of NK cells, CD56 + CD127 + ILC1s and cd T cells were reduced at 7 days p.i., compared to day 0 (Figures 1d and e). NK cells and all innate-like T-cell populations produced IFNc upon PMA + ionomycin restimulation; however, only the frequency of IFNc + iNKT, cd T and MAIT cells was significantly reduced at day 7, compared to day 0 ( Figure 1f). The white blood cell count (WBC) in volunteers was not significantly different between days 0 and 7 p.i. (Figure 1c), indicating that a general loss of blood leukocytes during infection did not account for these results. When the frequency and numbers of these cells were examined in one cohort after volunteers were treated with antiparasitic drug, we found they had recovered and in some cases were increased, relative to pre-infection levels ( Table 1).
Parasite accumulation in volunteers, as measured by the area under the curve (AUC) of blood parasitaemia curves (Figure 1a), was plotted against the frequency or cell number of each cell subset shown in Figure 1 at day 7 p.i. to identify any relationships with parasite burden. However, no significant relationships were found for any ILC or innate-like T cells (P > 0.05 for all cell subsets; data not shown). Similar results were obtained when the parasite multiplication rate (PMR) over time in each volunteer was plotted against corresponding ILC or innate-like T-cell frequencies or cell number. Together, these results show the frequency and cell number of group 1 ILCs were reduced following first exposure to blood stage Pf but this reduction was independent of parasite burden or PMR and recovered following antiparasitic drug treatment. These data suggest that NK cells and ILC1s either have increased cell death, decreased cell proliferation or sequester to tissues following Pf infection.

A loss of liver trNK cells and splenic ILC1s during PcAS infection
To further investigate ILCs in tissues during malaria, we next employed a mouse model of PcAS infection. A novel subset of liver ILC1s (trNK cells) has been reported in mice and humans. 7, 39 We examined these cells, as well as splenic ILC1s, 9 because of the importance of the liver and spleen as blood filtering organs during Plasmodium infection. 40,41 We identified liver ILC1s that were lineage (Lin; CD3, CD5, CD19)-negative, CD45 + NK1.1 + NKp46 + CD49a + DX5 À (Figure 2a). These were distinct from splenic ILC1s, identified as Lin À CD45 + NK1.1 + NKp46 + Eomes À CD127 +9 (Figure 2b). We found a decrease in the frequency and number of liver ( Figure 2c) and spleen ILC1s Figure 1. ILC and innate-like T-cell frequencies decrease following P. falciparum infection. Representative blood parasitaemia curve over the first 7 days of infection from a single cohort (n = 6) (a). Group 1 ILC and group 1 ILC-like subsets were identified by flow cytometry as indicated in the gating strategy (b). White blood cell counts for each patient at days 0 and 7 are depicted (c). The frequencies (d) and cell numbers (e) of group 1 ILC, group 1 ILC-like and innate-like T-cell subsets are shown. The proportion of each subset producing IFNc is shown (f). The data from b-f represent results from one to three cohorts (n = 8-14). Error bars represent the mean AE standard deviation (SD) (a). Comparisons between days 0 and 7 were made using the Wilcoxon (paired, nonparametric) test for NK cells and a two-way ANOVA with Sidak's multiple comparisons test for other subsets. *P < 0.05, **P < 0.01, ***P < 0.001.  (Figures 2c and d). In contrast, liver cNK cells (NK1.1 + NKp46 + CD49a À DX5 + ) increased in cell number over this same time period (Figure 2e).
ILC1s exhibit a more apoptotic phenotype than cNK cells One possible explanation for the reduced ILC1 frequency and number following PcAS infection could be increased apoptosis. To test this, we stained liver ILC1s ex vivo to assess Caspase-3/7 expression as a marker of apoptosis from days 1 to 4 p.i. (Figure 3a). Flow cytometry analysis revealed approximately 20% of liver ILC1s expressing Caspase-3/7 in na€ ıve C57BL/6 mice ( Figure 3b). Following PcAS infection, the frequency of Caspase-3/7-expressing ILC1s increased further at 2 days p.i., compared to na€ ıve cells. Therefore, increased apoptosis may at least partly explain the reduced liver ILC1 frequency early after PcAS infection.

Emergence of a CD49a + DX5 + doublepositive population
We also identified a population of cells within the liver and spleen that were Lin À CD45 + NK1.1 + NKp46 + CD49a + DX5 + (herein referred to as the CD49a + DX5 + double-positive' population) (Figures 4a and b). This population was readily detected 5 days p.i. and increased as infection progressed ( Figure 4c). This CD49a + DX5 + doublepositive population was detected at lower frequencies at 28 days p.i. when the ratio of ILC1s to cNK cells resembled that of na€ ıve samples, although full recovery of ILC1 number or frequency was not evident at this time point ( Figure 4c). Interestingly, the CD49a + DX5 + doublepositive population expressed the cNK cell marker CD62L and the ILC1 marker TNF-related apoptosisinducing ligand (TRAIL) at intermediate levels ( Figure 4d), suggesting they may represent a transitionary population between cNK cells and ILC1s.

Systemic cNK cell and ILC1 depletion do not affect blood parasitaemia
We next investigated the role of ILC1s during Plasmodium infection, given their transcriptional and functional resemblance to Th1 cells, 1,6 and previous reports indicating important roles for NK cells during PcAS infection. 25 WT, Rag1 À/À and Rag2 À/À c c À/À mice were infected with PcAS, which caused a nonlethal, chronic infection in control WT mice. 33 Unexpectedly, immunodeficient Rag2 À/À c c À/À mice (deficient in all lymphocytes) had a delayed peak parasitaemia, compared to Rag1 À/À  mice that were only deficient in B and T cells (Figure 5a). To determine whether the delayed peak parasitaemia observed in Rag2 À/À c c À/À mice could be attributed to the absence of cNKs, we infected Ncr1-iCre x Mcl1 fl/fl mice with PcAS. Myeloid cell leukaemia sequence-1 (Mcl1) is critical for the maintenance of mature NK cells and ILC1s. 42 Therefore, these cells were absent in mice lacking Mcl gene expression in NKp46 (encoded by the Ncr1 gene)-positive cells. Surprisingly, these mice were able to control parasite growth and had similar blood parasitaemia to Ncr1-iCre +/À control mice ( Figure 5b). Hence, the delay in peak parasitaemia in Rag2 À/À c c À/À mice, relative to Rag1 À/À mice, was not likely caused by the absence of NK cells or ILC1s but instead, possibly reflects changes in either the activity of phagocytic cells or alterations to parasite growth in the blood of Rag2 À/À c c À/À mice.
We next focused attention specifically on ILC1s by infecting Ncr1-iCre x Tgfbr2 fl/fl mice with PcAS. These mice lack Tgfbr2 gene expression in NKp46positive cells and have significantly reduced ILC1 numbers. 6 Despite this, these mice were still able to control parasite growth as efficiently as control animals (Figure 5c). Given previous contradictory reports on the role of cNK cells during malaria, [25][26][27][28][29] we next confirmed the above results in 2 other cNK cell-and ILC1-deficient models. First, we treated mice with a-NK1.1 mAb (depletes cNK cell and ILC1s) and found no effect on the ability to control parasite growth, compared with control mice (Figure 5d). Second, we used diphtheria toxin (DT) to deplete cNK cells and ILC1s in Ncr1-iCre x iDTR mice and again found no change in the ability of mice to control infection, relative to controltreated animals (Figure 5e). Together, these data indicate limited roles for cNK cells and ILC1s in antiparasitic immunity during PcAS infection.

DISCUSSION
Here, we show a reduction in the frequency of circulating ILC1s, NK cells and innate-like T cells in healthy volunteers infected for the first time with Pf. This finding was akin to the decrease in frequencies and numbers of peripheral T cells during acute Pf malaria in Ghanaian children. 43 Similar observations were made for spleen and liver ILC1s in C57BL/6 mice infected with PcAS, accompanied by increased apoptosis in the latter cell population. Further studies on the roles of Figure 3. Liver ILC1s exhibit a more apoptotic phenotype than cNK cells. Liver single-cell suspensions were stained for a viability marker, ILC1 surface markers, followed by staining for Caspase-3/7. The representative plots from na€ ıve and day 2 p.i. are shown (a) where viability dye + Caspase-3/7 À = dead cells, viability dye + Caspase-3/7 + = necrotic cells, (b) viability dye À Caspase-3/7 + = apoptotic cells, and viability dye À Caspase-3/7 À = live cells. The relative frequencies of viable, apoptotic and necrotic/end-stage apoptotic ILC1s are shown. Data represent mean AE SEM from one experiment where n = 3 for na€ ıve mice and n = 4 for PcAS-infected mice. Comparisons were made using the Kruskal-Wallis test accompanied by the Dunn's multiple comparisons test. **P < 0.01. ILC1s and NK cells during PcAS infection using cell depletion and genetically modified mice indicated a limited role for these cells in the control of blood parasitaemia. These results contrast earlier findings that reported NK cells confer protection during Plasmodium infection using anti-asialo GM1 to deplete NK cells, 25 but were consistent with more recent findings using anti-NK1.1 mAb for NK cell depletion. 28 Interestingly, a recent study in humans who had received a bone marrow transplant found that ILCs were redundant in the presence of competent B and T cells. 5 Our data indicate this may also be the case in experimental malaria, but we cannot yet make definitive conclusions on the role of these cells in human malaria.
A recent study has reported that ILCs were irreversibly lost during acute HIV-1 infection. 44 Our findings in PcAS-infected mice indicate that liver and splenic ILC1s were lost in the first 5 days of infection but recovered as infection was controlled. Similarly, in volunteers participating in CHMI studies, the frequency and number of circulating ILC1s, NK cells and innate-like T cells all fell in the first 7 days of Pf infection but recovered after drug treatment. This recovery may have been driven, at least in part, by increased levels of parasite molecules being available after drug-mediated killing to promote activation and/ or expansion of certain immune cell subsets. Nevertheless, changes in these cell populations during malaria were transient, possibly reflecting the less persistent nature of Plasmodium infection, compared to HIV. Of note, we cannot exclude the possibility that reduced levels of circulating ILC1s, NK cells, and innate-like T cells following Pf infection or ILC1s following PcAS infection might reflect infection-mediated changes in the marker expression used to define our cell populations and/or sequestration of cells in tissues. This will require further examination.
We identified a cell population in the mouse liver that expressed both CD49a (an ILC1 marker) and DX5 (a cNK cell marker). This population was also NK1.1-and NKp46-positive and expressed intermediate levels of TRAIL (expressed by ILC1s but not cNK cells) and CD62L (expressed by cNK cells but not ILC1s). These CD49a + DX5 + doublepositive cells emerged at 5 days p.i., which coincided with the loss of liver ILC1s. They were also increased in frequency at 14 days p.i., when liver ILC1 frequency had yet to recover but were reduced at 28 days p.i., when the relative proportion of liver ILC1s to cNK cells resembled proportions prior to the loss of liver ILC1s. These cells were not liver-specific, as they were also found in the spleen. A recent report proposed a bidirectional plasticity between cNKs and ILC1s that was mediated by the presence or absence of Eomesodermin (Eomes). 45 Our current findings suggest that this CD49a + DX5 + double-positive cell population may be an intermediate population between cNK cells and ILC1s. Recently, we reported a similar cell population expressing CD49a, CD49b and Eomes in mouse tumour models, and this was found to represent an intermediate population between TGFb-mediated conversion of cNK cells into ILC1s. 46 Whether this plasticity between cNKs and ILC1s also occurs in malaria remains unknown but perturbation of TGFb signalling following infection could help explain the disparity between ILC1 and NK cell frequencies after PcAS infection. Although our results from mouse models of malaria indicate a limited role for cNK cells and ILC1s in control of parasite growth, we cannot exclude a role for these cells in human malaria, based on our data. Previous studies have shown that human NK cells rapidly produce IFNc following exposure to Pf parasitised red blood cells (pRBCs) in vitro. 29,47 Furthermore, depletion of NK cells in humanised mice infected with Pf using anti-CD56 mAb resulted in increased parasite growth, 48 albeit parasite growth appeared relatively modest in this model system. Nevertheless, our investigation of associations between cNK cells, ILC and innate-like T-cell subsets and either parasite burden (AUC) or PMR revealed no relationships. Further research is clearly needed in this area.
Our CHMI and mouse studies were initiated via an intravenous infusion of blood stage parasites, thereby bypassing the liver stage of infection. One advantage of this is that infections are relatively synchronous and hence, early changes in immune cell populations can be readily detected. However, it is possible that host responses to liver stage infections may influence subsequent responses to blood stage infections. 49 Recently, a study of innate-like T cells in CHMI studies in Tanzania using Pf sporozoites to establish infection reported changes in the frequencies of circulating NK and MAIT cells up to 6 months after infection, although reduced frequencies of MAIT and NK cells were observed before this time point. 32 The earliest samples were taken 9 days after infection in this study, making comparisons with our data difficult. Furthermore, the initiation of infection via liver stage, different parasite dose used to establish blood stage infection and the immune status of the Tanzanian volunteers living in a malaria-endemic region make direct comparisons with our data challenging. Nevertheless, comparisons between CHMI studies initiated with sporozoites and blood stage parasites in healthy volunteers and those living in malaria-endemic areas are likely to provide a wealth of future data that may help explain the roles of various immune cell populations during infection and disease, especially if infections initiated by liver and blood stages can be directly compared in the same populations.
In conclusion, we report a decline in the frequency of cNK cells, ILC1s, and innate-like T cells 7 days after blood stage Pf infection in CHMI studies with healthy volunteers. A similar observation was made for liver ILC1s in a mouse model of malaria caused by PcAS. Our data indicate a limited role for cNK cells and ILCs in the control of blood parasitaemia in mice. Together, these data provide novel insights into the responses of innate immune cells in mice and humans during malaria and may help guide future strategies aimed at manipulating host immune responses for clinical advantage during this disease.

Controlled human malaria infection
Experimental procedures were performed as part of a substudy on human blood samples collected from consenting participants enrolled in a drug study (Australian New Zealand Clinical Trials Registry ACTRN12613000565741 and NCT02389348) conducted at Q-Pharm Pty Ltd (Herston, QLD, Australia) under the approval of the QIMR Berghofer Medical Research Institute Human Research Ethics Committee (QIMR-HREC). Volunteers comprised healthy male and female nonsmokers, aged between 18 and 45 in one trial and 18 and 55 in another trial, with no history of malaria or prior exposure to malaria-endemic regions. Experiments were performed on blood samples collected from participants enrolled in 3 cohorts (n = 14 volunteers in total). An additional cohort (n = 8 volunteers) was used to examine NK cell frequency and number. Participants were infected intravenously using a Pf (clone 3D7)-induced blood stage malaria (IBSM) challenge inoculum (1800 parasitised red blood cells [pRBC]), with parasitaemia monitored by realtime quantitative polymerase chain reaction (qPCR), [50][51][52] and blood collected at time points indicated in Results section. Antimalarial drug treatment was administered once parasitaemia exceeded 1000 parasites per mL.
Parasitaemia was plotted over time for each patient, and parasite burden over the course of infection (prior to treatment) was expressed as a measurement of AUC, as previously described. 19 A growth model was also derived from parasitaemia measurements (prior to treatment) over time and fitted to each individual using simple linear regression. The gradient of this growth model (parasite multiplication rate [PMR]) was then estimated using a loglinear model described by the following equation: where ! is parasites per mL measured by qPCR, a is the intercept, m is the gradient of the growth model, and time is the number of days from inoculation.

PcAS infections and measurement of peripheral blood parasitaemia
PcAS was thawed from stabilities and passaged once in vivo in a C57BL/6J mouse, prior to establishing experimental infections. Mice were infected intravenously (i.v.) with 10 5 pRBCs. Parasitaemia was measured by flow cytometry using Hoechst 33342 (Sigma-Aldrich â , St. Louis, MO, USA) and Syto 84 (Invitrogen TM , Life Technologies, Carlsbad, CA, USA), as previously described. 62 Samples were acquired on a BD FACSCanto TM II through BD FACSDiva TM V8.0 (both by BD Biosciences) and analysed on FlowJo v10 OSX (Tree Star, Inc., Ashland, OR, USA).

Preparation of spleen and liver single-cell suspensions
Mice were sacrificed by CO 2 asphyxiation. A mid-sagittal incision was made on the abdominal cavity. The spleen was removed and passed through a 100-lm cell strainer (Corning Incorporated, Corning, NY, USA). The single-cell suspension was washed in Roswell Park Memorial Institute Medium (RPMI; Life Technologies) + 100 lg mL À1 penicillin and streptomycin (ps; Gibco â , Life Technologies) (RPMI/ps) and incubated for 7 min in Red Blood Cell Lysis Buffer Hybri-Max TM (Sigma-Aldrich â ). Cells were washed in RPMI/ps, pelleted by centrifugation and counted on the Countess II FL (Life Technologies), as per manufacturer's protocol. The liver was perfused with 19 phosphate-buffered saline (PBS). The excised liver was collected in 1% (v/v) foetal bovine serum (FBS) in PBS and mechanically passed through a 200-lm square metal mesh. Hepatocytes were separated from lymphocytes and removed using a 33% Percoll TM (GE Healthcare) gradient according to manufacturer's instructions. Red Blood Cell Lysis Buffer Hybri-Max TM was added to each pellet and incubated for 5 min at room temperature, prior to washing in 1% (v/v) FBS in PBS. Cells were pelleted by centrifugation and counted using the Countess II FL, as per manufacturer's protocol.

Flow cytometry
Samples were resuspended in 1% (w/v) paraformaldehyde (PFA) poststaining and stored at 4°C before acquisition on a BD LSRFortessa TM (special order research product; BD Biosciences) through BD FACSDiva TM V8.0 and analysed on FlowJo v10 OSX.

Detection of apoptotic cells
Apoptotic cells were detected by cell surface staining with either Annexin V from the Annexin V FITC Apoptosis Detection Kit I (BD Biosciences) or reagents from the CellEvent â Caspase-3/7 Green Flow Cytometry Assay Kit (Molecular Probes â , Life Technologies), as per manufacturer's instructions.

Depletion of cNKs and ILC1s
C57BL/6J or Rag1 À/À mice were given an intraperitoneal (i.p.) injection with 1 mg per mouse of either anti-NK1.1 (PK136) mAb or InVivo MAb Polyclonal Mouse IgG (both from BioXCell, West Lebanon, New Hampshire, USA) on alternate days, starting from the day before infection.
Ncr1-iCre x iDTR mice were given 8 ng g À1 (based on body weight) of DT from Corynebacterium diphtheria (Sigma-Aldrich â ) per mouse 2 days and 1 day prior to infection and every 2 days after infection. Control mice were given an equal volume of sodium chloride (0.9% [w/v]) for irrigation (Baxter International, Deerfield, IL, USA). Depletion efficacy in both models was determined by flow cytometry analysis using an anti-mouse mAb towards NKp46 (CD335; 29A1.4) (BioLegend).

Statistics
Graphing and statistical analyses were performed on GraphPad Prism 6 (GraphPad, San Diego, CA, USA). A P-value (P) < 0.05 was considered statistically significant. Direct comparisons between two time points in the human trial data were made using the Wilcoxon (paired, nonparametric) test. For studies involving mice, comparisons between time points in time course experiments were made using the Kruskal-Wallis test accompanied by the Dunn's multiple comparisons test, while comparisons between two groups of mice were made using the Mann-Whitney (unpaired, nonparametric) t-test. All data are shown as mean AE standard error of mean (SEM) unless otherwise stated.