Miltefosine enhances infectivity of a miltefosine-resistant Leishmania infantum strain by attenuating its innate immune recognition

Background Miltefosine (MIL) is currently the only oral drug available to treat visceral leishmaniasis but its use as first-line monotherapy has been compromised by an increasing treatment failure. Despite the scarce number of resistant clinical isolates, MIL-resistance by mutations in a single aminophospholipid transporter gene can easily be selected in a laboratory environment. These mutations result in a reduced survival in the mammalian host, which can partially be restored by exposure to MIL, suggesting a kind of drug-dependency. Methodology/Principal findings To enable a combined study of the infection dynamics and underlying immunological events for differential in vivo survival, firefly luciferase (PpyRE9) / red fluorescent protein (DsRed) double-reporter strains were generated of MIL-resistant (MIL-R) and syngeneic MIL-sensitive (MIL-S) Leishmania infantum. Results in C57Bl/6 and BALB/c mice show that MIL-R parasites induce an increased innate immune response that is characterized by enhanced influx and infection of neutrophils, monocytes and dendritic cells in the liver and elevated serum IFN-γ levels, finally resulting in a less efficient establishment in liver macrophages. The elevated IFN-γ levels were shown to originate from an increased response of hepatic NK and NKT cells to the MIL-R parasites. In addition, we demonstrated that MIL could increase the in vivo fitness of MIL-R parasites by lowering NK and NKT cell activation, leading to a reduced IFN-γ production. Conclusions/Significance Differential induction of innate immune responses in the liver was found to underlie the attenuated phenotype of a MIL-R parasite and its peculiar feature of drug-dependency. The impact of MIL on hepatic NK and NKT activation and IFN-γ production following recognition of a MIL-R strain indicates that this mechanism may sustain infections with resistant parasites and contribute to treatment failure.


Introduction
Visceral leishmaniasis (VL) is a fatal vector-borne disease that is caused by transmission of the protozoan parasite Leishmania donovani in Asia and parts of Africa and by L. infantum in the Mediterranean area and Latin America [1][2][3]. In 2017, 90 000 new cases of VL were reported with another 300 million people at risk of acquiring the disease [4][5][6]. Leishmania parasites are transmitted through the bites of infected female sand flies. During blood feeding, extracellular promastigotes are injected into the dermis of the host where they are taken up by cells of the phagocytic system. Inside these cells, promastigotes differentiate into obligate intracellular amastigotes that will disseminate and infect macrophages in the major target organs liver and spleen [7][8][9]. In most VL mouse models, different outcomes of infection can be distinguished depending on the tissue [10]. In the liver, an acute infection leads to the development of granulomas around infected Kupffer cells (resident macrophages of the liver) resolving the infection. The cytokine interferon (IFN)-γ plays a crucial role in this process by stimulating the production of tumor necrosis factor (TNF)-α and inducible nitric oxide synthase (iNOS) in infected Kupffer cells [11]. On the other hand, the immune response in the spleen is characterized by high TNF-α levels and a delayed or absent granuloma formation leading to a chronic infection accompanied by architectural damage and immunological dysfunction [10,[12][13][14][15][16][17][18]. Early immunological events play a crucial role in shaping these organ-specific immune responses. In the liver, one of those events is the recruitment and activation of natural killer (NK) and NKT cells within hours of infection, as seen during experimental infections with L. donovani and L. infantum [19][20][21][22]. Whereas IFN-γ-producing CD4 + and CD8 + T cells are crucial for parasite clearance at a later stage of the infection, IFN-γ production by NK and NKT cells during the early onset is critical for efficient granuloma formation and early parasite control [19,22]. While several reports have shown that NK cells are not essential in shaping the early immune response, the absence of NKT cells or NKT cell activation result in decreased IFN-γ levels, impaired granulomatous response and higher parasite burdens in both the liver and spleen [21,22].
A successful vaccine against VL has not been developed to date and drug treatment has been hampered by widespread resistance, treatment failure (TF) and toxicity of the current antileishmanial drugs [23]. Since its introduction in 2002, miltefosine (MIL) has remained the only oral drug available for VL treatment [23][24][25]. Although the exact mode of action of MIL is not entirely understood, changes in the parasite's lipid metabolism, reduction of mitochondrial cytochrome c oxidase and induction of apoptosis-like cell death are at the basis of its activity [26][27][28][29][30][31]. Rather soon after its introduction, relapse rates increased up to 20% [32,33] with indication of an elevated drug-tolerance [34] but without the emergence of full and stable MIL-resistance. To date, only few fully MIL-resistant (MIL-R) isolates have been described, including two MIL-R L. infantum strains isolated from French HIV-infected patients and two MIL-R L. donovani strains recovered from patients in India [35][36][37]. Despite the scarce number of MIL-R field isolates [33][34][35][36][37][38][39], resistance can easily be selected experimentally arising from mutations in the MIL transporter (MT) or the Ros3 gene [37,40]. The encoded MT/Ros3 transporter complex facilitates MIL import over the parasite's cell membrane [41]. Mutations in either the MT or ROS3 gene have been reported in the four above mentioned MIL-R field isolates [35][36][37]. Re-introduction of functional MT or ROS3 gene copies fully restores MILsusceptibility [37], linking the defective subunits to the MIL-R phenotype. Former research by our lab and others has indicated that MT mutations provoke a severe fitness loss of Leishmania strains in mice and sand flies [40,[42][43][44][45]. This fitness loss was characterized by a reduced in vitro growth rate and metacyclogenesis and a reduced IL-10 production by infected macrophages [43]. Furthermore, MIL-R parasites were unable to cause a typical visceralization pattern in mice, with rapid clearance from the viscera [42]. In the sand fly vector, these parasites showed a reduced colonization of the stomodeal valve by metacyclic promastigotes, a process that is essential for transmission to the vertebrate host [45]. These observations were recently confirmed by our lab using a natural MIL-R L. infantum strain (LEM5159) which has a defective LiROS3 gene caused by a frameshift mutation and early stop codon [46]. Surprisingly, the reduced infectivity in mice was partially restored either following in vivo MIL treatment or by drug pre-exposure of parasites [42]. To gain better insights into the impact of resistance acquisition and mechanisms underlying apparent MIL-dependency of MIL-R parasites, the present study evaluated the immunological characteristics of early infections with and without MILexposure.

Ethics statement
The use of laboratory animals was carried out in strict accordance to all mandatory guidelines (EU directive 2010/63/EU on the protection of animals used for scientific purpose and the declaration of Helsinki) and was approved by the Ethical Committee of the University of Antwerp (UA-ECD 2015-90 and 2017-04).

Animals
Female Swiss, BALB/c and C57Bl/6 mice (6-8 weeks old) were purchased from Janvier Laboratories (Le Genest-Saint-Isle, France). C57Bl/6 IFN-γ reporter mice with endogenous polyA transcript (GREAT) and were obtained from Prof. C. De Trez (VUB, Belgium). Mice were randomly allocated to groups of 3 to 5 mice and housed in individually ventilated cages with laboratory rodent food (Carfil, Belgium) and drinking water ad libitum.

Parasites and transfections
The clinically MIL-resistant L. infantum strain LEM5159 (MHOM/FR/2005/LEM5159) [36,37], the MIL-susceptible L. infantum strain LEM3323 MIL-S (MHOM/FR/96/LEM3323) and its MIL-resistant isogeneic counterpart LEM3323 MIL-R were previously rendered bioluminescent by transfection with the PpyRE9 gene [42]. Both strains were additionally transfected with the DsRed gene that was cloned from the pssuINTDsRed vector which was obtained through Prof. C. De Trez (VUB, Belgium). A homologous overlap for Gibson assembly in the pLEXSY-Sat2.1 vector (Jena Bioscience, Germany) was added to the DsRed gene during cloning using the primers FP 5'-CGC TCC TGC TTT CCT TGC TGT GCC TTG CCA CCA GAT CTG CAT GGC CTC CTC CGA GG-3' and RP 5'-AGA AGG CGA CGT GAA AGG AAC AAG AAA GGA GGA GGA GGG CCT ACA GGA ACA GGT GGT GG-3'. pLEX-SY-Sat2.1-DsRed was digested using SwaI restriction enzyme and the 2.8 kb fragment was gelpurified using the QIAquick Gel Extraction Kit (Qiagen, Germany) followed by ethanol precipitation and resuspension in cytomix transfection buffer (120 mM KCl; 0.15 mM CaCl 2 ; 10mM KH 2 PO 4 ; 25mM HEPES; 2mM EDTA; 5mM MgCl 2 ; pH 7.6) at 1 μg/μL. For each strain, 10 μg of the linearized construct was used for the electroporation of 1×10 8 procyclic promastigotes (twice at 25 μF and 1500V with 10s interval) using the Bio-Rad GenePulse Xcell electroporation unit (Bio-rad Laboratories, USA). MIL-S PpyRE9/DsRed and MIL-R PpyRE9/DsRed transfectants were selected under Nourseothricin (Jena Bioscience, Germany) pressure and subcultured twice weekly in HOMEM promastigote medium. To assure that transfection did not impact the resistance phenotypes of the parasites, MIL-susceptibility assessment was performed using the standard intracellular amastigote assay. Briefly, primary peritoneal macrophages were harvested after starch stimulation of Swiss mice. The macrophages were seeded in a 96-well plate and kept at 37˚C and 5% CO 2 . Twenty-four hours after seeding, macrophages were infected with stationary-phase promastigotes at an infection ratio of 10:1. After 24 hours of incubation, residual promastigotes were removed after which two-fold dilutions of MIL were added to the infected macrophages. Plates were incubated for 96 hours and subsequently stained with Giemsa solution. Finally, 50% inhibitory concentrations (IC 50 ) were determined microscopically by determining the intracellular amastigote burdens of treated and untreated macrophages.

Animal infections
Mice were infected with MIL-S PpyRE9/DsRed or MIL-R PpyRE9/DsRed double reporter parasites by intravenous (i.v.) injection of 1×10 8 stationary phase promastigotes. In some experiments, MIL-R PpyRE9/DsRed parasites were preconditioned for 5 days with 40 μM MIL prior to infection. To assess the viability of stationary phase MIL-S and MIL-R parasites prior to infection, parasites were stained with propidium iodide (Thermo Fisher scientific, USA) and analyzed on a MACSQuant 10 Analyzer (Miltenyi Biotec, Germany) and using FlowLogic software (Miltenyi Biotec, Germany). As a positive control, heat-killed (15 min. at 65˚C) were included. Parasite burdens were monitored in vivo using bioluminescent imaging (BLI) of mice with the IVIS System (PerkinElmer, Belgium) throughout early infection. In some experiments, 1×10 7 stationary phase promastigotes were inoculated intradermally (i.d.) in the ear of BALB/c mice that were treated for 5 consecutive days, starting 1 day before infection, with 40 mg/kg of MIL via oral gavage. Mice were imaged for 15 minutes starting 3 minutes after i.p. injection of 150 mg/kg D-luciferin (Beetle Luciferin Potassium Salt, Promega, USA). Images were analyzed using LivingImage v4.3.1 software (PerkinElmer, Belgium) by quantifying the signal as relative luminescence units (RLU) within regions of interest corresponding to the ear, liver, spleen and bone marrow. At each time point, blood was collected via the tail vein and plasma was stored at -80˚C until further analysis. Experiments were performed in duplicate or triplicate depending on the experiment with 3 to 5 mice per infection group.

Cell isolation
Infected C57Bl/6 and/or C57Bl/6 IFN-γ reporter mice were sacrificed at 1, 7 and 21 dpi. Spleen and liver were isolated after transcardial perfusion with 10 mL KREBS Henseleit solution supplemented with 10 U/mL heparin (Sigma-Aldrich, USA) [47]. Tissue imprints of infected livers and spleens were made on glass slides, fixed with methanol, Giemsa-stained and evaluated microscopically to determine the level of infection. Livers were homogenized using a liver dissociation kit (Miltenyi Biotec, Germany) according to the manufacturer's instructions. Briefly, livers were mechanically disrupted in 5 mL DMEM medium (Thermo Fisher scientific, USA) containing liver dissociation enzymes, using the gentleMACS Dissociator (Miltenyi Biotec, Germany) and incubated at 37˚C for 30 min. The digested tissue was homogenized and filtered through a 100 μM filter. Spleens were dissociated in 10 mL FACS medium (PBS, 5% FCS, 2mM EDTA, pH 7.2) by mechanical disruption. Subsequently, spleen cells were centrifuged (10 min, 400×g, 4˚C) and the pellet was treated with ACK lysis buffer (0.15 M NH 4 Cl, 1 mM KHCO 3 , 0.1 mM Na 2 -EDTA). All cell suspensions (spleen and liver) were centrifuged (10 min, 400×g, 4˚C) and the pellet was resuspended in FACS medium at a concentration of 2×10 7 cells/ml for flow cytometry analysis.
At several time points over the course of infection, different DsRed + cell populations were sorted using a BD FACS Melody apparatus (BD Biosciences, Belgium). Sorted cells were centrifuged for 10 minutes at 1,000×g and pelleted on a glass slide using a cytospin 4 cytocentrifuge (Thermo Fisher scientific, USA). Glass slides were air dried, subjected to a 5-minute fixation in methanol and staining with Giemsa solution for 15 minutes and analyzed microscopically.

Serum analysis
Serum levels of IFN-γ, IL-10 and TNF-α were determined using the V-PLEX Custom Mouse Cytokine assay (Meso Scale Discovery, Rockville, USA) according to the manufacturer's instructions.

RNA extraction
Dissected liver and spleen samples were treated with RNALater (Qiagen, Germany) according to the manufacturer's instructions and stored at -80˚C. Tissue samples were thawed and RNA extraction was performed using the Qiagen RNA extraction kit (Qiagen, Germany). Briefly, samples were homogenized in 100 μL RLT buffer using a pestle rotor. After homogenization, 500 μL RLT buffer was added and samples were centrifuged for 3 minutes at maximum speed after which the supernatant was loaded on a gDNA Eliminator spin column and centrifuged for 30 seconds at 10,000×g. 1 volume of 70% ethanol (50% for liver) was added to the flow through and loaded onto a RNeasy spin column followed by centrifugation for 15 seconds at 10,000×g. The flowthrough was discarded and columns were washed with RW1 and RPE buffer. RNA was eluted in RNA-free water following a centrifugation for 1 minute at 10,000×g. Concentration of RNA samples was measured by absorbance at 260 nm using a NanoDrop spectrophotometer (Isogen Life Science, The Netherlands). Absorbance ratios at 260/280 nm and 260/230 nm were determined as a measure for RNA purity and integrity. Samples were stored at -80˚C until qPCR analysis.

Real-time quantitative PCR
For real-time quantitative PCR (RT-qPCR) detection of IFN-γ, IL-10, TNF-α and IL-1β encoding transcripts in liver and spleen, validated primers were obtained from an online repository (http:// www.rtprimerdb.org) ( Table 1). 1:10 dilutions of RNA samples were analyzed using SensiFast SYBR Hi-ROX One-Step kit (Bioline, UK) and 400 nM primer concentrations. PCR conditions comprised an initial polymerase activation at 96˚C followed by 40 cycles, each consisting of a denaturation step at 95˚C for 15 seconds, 30 seconds annealing at 60˚C and 30 seconds elongation at 72˚C. The specificity of the primers was assessed by verifying the presence of a single amplicon by melting curve analysis. Eukaryotic translation elongation factor 2 (eef2) [55] was used as reference gene for normalization in the Q base Plus Software (Biogazelle, Belgium). The relative expression levels were analyzed using Graphpad Prism software v7 (GraphPad, USA).

Statistical analysis
Statistical analysis was performed using the Mann-Whitney U test in IBM SPSS Statistics v25 (IBM, Armonk, USA). Values are expressed as mean ± standard deviation (SD) and values of p �0.05 are considered statistically significant.

Parasite transfection with PpyRE9 and DsRed did not alter the parasite's MIL susceptibility
MIL-susceptibility assessment using the intracellular amastigote assay did not reveal any differences between transfected (IC 50 values 0.26 ± 0.03 μM and � 20 μM for MIL-S and MIL-R respectively, with 20 μM being the toxic concentration of MIL) and untransfected parasites (IC 50 values 0.38 ± 0.08 μM and � 20 μM for MIL-S and MIL-R respectively) indicating that the transfection did not interfere with the MIL resistance profile. Furthermore, both MIL-S PpyRE9/DsRed and MIL-R PpyRE9/DsRed stationary phase cultures had similar percentages of dead parasites (1.10 ± 0.29% and 0.97 ± 0.59%, respectively) (S2 Fig) indicating that any differences observed during the in vitro and in vivo infections could not be attributed to differences in inoculum viability. Previous enrichment studies with peanut agglutinin also already demonstrated that interstrain differences of in vivo infectivity were not related to differences in metacyclogenesis (42).

MIL-R parasites fail to disseminate during the first week of the infection
Compared to susceptible (MIL-S) parasites, isogenic MIL-R parasites were rapidly cleared from the liver in C57Bl/6 and BALB/c mice with a failure to successfully disseminate to the spleen and bone marrow (Fig 1). MIL-S PpyRE9/DsRed showed a typical visceralization pattern with a rapid increase of hepatic parasite burdens during the first weeks of the infection followed by dissemination of parasites to the spleen and bone marrow. In contrast, hepatic MIL-R PpyRE9/DsRed burdens were the highest at day 1 of infection after which they rapidly subsided. Spleen and bone marrow signals were only rarely detected during onset of MIL-R PpyRE9/DsRed infection. Flow cytometry (Fig 2A)   bioluminescence data clearly indicated a parasite expansion. This increase of MIL-S PpyRE9/DsRed parasite burdens was confirmed by Giemsa counting of liver and spleen tissue imprints at 1, 7 and 21 dpi ( Fig 2B). The lower DsRed reporter signals are therefore most likely associated with the increasingly quiescent state of intracellular amastigotes.

Myeloid cells are rapidly recruited to the liver and spleen during early MIL-R infection resulting in an altered cellular distribution
An increased influx of monocytes, neutrophils and dendritic cells was observed in MIL-R PpyRE9/DsRed infected livers at 1 dpi ( Fig 3A) with a different cellular distribution of infection ( Fig 3B). In MIL-R PpyRE9/DsRed infected livers, monocytes (CD11b + Ly-6G -Ly-6C High F4/80 -), neutrophils (CD11b + Ly-6G + ) and dendritic cells (CD11b + CD11c + ) represented a higher percentage of the total number of DsRed + CD11b + cells when compared to the MIL-S PpyRE9/DsRed (28.0 ± 10.6% vs 16.3 ± 3.9%; 6.6 ± 3.4% vs 3.1 ± 0.7%; 8.3 ± 1.7% vs 3.6 ± 0.9%; p �0.05) while the relative importance of macrophages (CD11b + Ly-6G -Ly-6C low F4/80 + ) diminished (43.7 ± 11.2% vs 68.7 ± 6.7%; p�0.05). At 7 and 21 dpi, the differences in cellular distribution between MIL-S PpyRE9/DsRed and MIL-R PpyRE9/DsRed became less prominent; however, monocytes, neutrophils and dendritic cells still represented a higher percentage of the total DsRed + CD11b + cells in MIL-R PpyRE9/DsRed infected liver. In the spleen no major differences in myeloid cell influx and cellular tropism were observed (S4A and S4B Fig). These data show that MIL-R PpyRE9/DsRed induces a stronger influx of myeloid cells in the liver during the initial phase of infection with concomitant change of initial cell tropism. At 7 and 21 dpi, different infected cell populations in the liver (macrophages, monocytes, neutrophils and dendritic cells) were sorted and analyzed microscopically for differences in the intracellular parasite burdens between MIL-S PpyRE9/DsRed and MIL-R PpyRE9/DsRed (Fig 4). By 21 dpi, liver macrophages were more heavily infected by MIL-S PpyRE9/DsRed parasites compared to MIL-R PpyRE9/DsRed (5.52 ± 3.18 amastigotes/cell vs 3.08 ± 2.34 amastigotes/cell; p�0.0001) whereas no differences could be observed in the other myeloid cell types.

MIL-R parasites induce stronger early IFN-γ responses in hepatic NK and NKT cells
At three time points during early infection (1, 7 and 21 dpi), serum levels of several infectioninduced key inflammatory cytokines were determined (Fig 5A). No major differences in TNF- α, IL-10 or IL-1β levels were observed. Both parasite strains induced a moderate TNF-α response at 1 dpi (43.5 ± 32.1 pg/mL and 51. 8 ± 29.6 pg/mL, respectively) and TNF-α remained elevated throughout the early infection with a second peak at 21 dpi (62.6 ± 12.8 pg/ mL and 45.0 ± 26.9 pg/mL, respectively). IL-10 levels were moderately elevated throughout infection with both strains, with an initial increase at 1 dpi (29.4 ± 14.0 pg/mL and 37.7 ± 27.8 pg/mL for MIL-S PpyRE9/DsRed and MIL-R PpyRE9/DsRed , respectively) and a stronger increase at 3 wpi in MIL-S PpyRE9/DsRed compared to MIL-R PpyRE9/DsRed (59.8 ± 11.3 pg/mL and 27.8 ± 13.6 pg/mL, respectively). For IL-1β, levels did not exceed 10 pg/mL and no important interstrain differences were observed. As for IFN-γ, a prominent response (p <0.01) was observed following MIL-R PpyRE9/DsRed infection at 1 dpi (213.1 ± 123.1 pg/mL) in comparison to the moderate MIL-S PpyRE9/DsRed -induced IFN-γ response (62.0 ± 53.3 pg/mL). In both infection groups, the early IFN-γ surge subsided by 4 dpi. These results were confirmed by quantitative RT-qPCR of the gene expression levels of IFN-γ, TNF-α, IL-10 and IL-1β in the liver and spleen (S5A Fig). Enhanced transcription of the IFN-γ gene was observed at each time point for MIL-R PpyRE9/DsRed compared to MIL-S PpyRE9/DsRed in the liver and spleen, with statistical differences in the liver at 1 and 7 dpi. TNF-α transcription was significantly increased in MIL-R PpyRE9/DsRed infected livers at 1, 7 and 21 dpi and at 1 and 7 dpi in the spleen. Less interstrain variation was observed for the IL-10 and IL-1β transcript levels.
The increased IFN-γ levels seen at 1 dpi suggest an enhanced infiltration and/or activation of IFN-γ producing lymphocytes upon MIL-R PpyRE9/DsRed infection. It has been shown that different lymphocyte subsets produce IFN-γ throughout a VL infection in mice [56]. While CD4 + and CD8 + T cells are an important source of IFN-γ during the later stages of infection when fully mature granulomas develop to clear infected Kupffer cells, NK and NKT cells were shown to produce IFN-γ during the initial stages of infection [19][20][21][22]. Given the early IFN-γ response at day 1 of infection, NK (TcRβ -NK1.1 + ), NKT (TcRβ + CD1d tetramer + ), CD4 + T (TcRβ + CD4 + ) and CD8 + T (TcRβ + CD8 + ) cell populations were examined in the liver and spleen of MIL-S PpyRE9/DsRed and MIL-R PpyRE9/DsRed infected C57Bl/6 mice. Both strains induced infiltration of similar numbers of NK, NKT, CD4 + T and CD8 + T cells at 1 dpi (Fig 5B). IFN-γ reporter (GREAT) mice [57,58] were used next to investigate the activation of these cells and to identify the source of differential IFN-γ production at 1 dpi, (Figs 5C and S5B). Compared to an uninfected reporter control, both strains induced IFN-γ production by NK and NKT cells in the liver whereby MIL-R PpyRE9/DsRed induced more IFN-γ-producing NK and NKT cells. No differential induction of the IFN-γ reporter was observed in splenic NK and NKT cells at 1 dpi, nor was there any induction of IFN-γ in CD4 + and CD8 + T cells in the liver and spleen at this time point (Fig 5C). These data clearly indicate that MIL-R parasites induce an enhanced activation of hepatic NK and NKT cells leading to an increased IFN-γ production at the very early onset of infection.

NK and NKT cells contribute to early MIL-R parasite clearance
To investigate whether the early IFN-γ burst and increased NKT cell activation contributes to the decreased survival of MIL-R PpyRE9/DsRed , infections were carried out in BALB/c mice and/ or C57Bl/6 treated with an anti-IFN-γ or an anti-CD1d antibody to neutralize CD1d-mediated lipid antigen presentation to NKT cells. BALB/c mice treated with an anti-IFN-γ neutralizing antibody 1 day before MIL-R PpyRE9/DsRed infection showed increased liver burdens as compared to isotype treated mice (Fig 6). No effect was observed on spleen and bone marrow burdens. The contribution of NKT cells to the increased IFN-γ production during the first 24h of infection was confirmed by a drastic drop in IFN-γ serum levels as a result of anti-CD1d treatment (Fig 7C). Bioluminescent imaging of the MIL-R PpyRE9/DsRed infection dynamics revealed significantly increased liver burdens in anti-CD1d as compared to isotype treated BALB/c and C57Bl/6 mice (Fig 7A and 7B). In most anti-CD1d-treated mice, some dissemination to the spleen and bone marrow was observed. Similar results, but less pronounced, were obtained in anti-NK1.1-depleted C57Bl/6 mice showing an increased survival of MIL-R PpyRE9/DsRed during the first weeks of infection compared to isotype control-treated mice (S6 Fig). These data indicate the importance of early NK and NKT cell responses and the corresponding IFN-γ production in the rapid clearance of MIL-R parasites from the liver preventing further dissemination to other target organs. It must be noted that IFN-γ and NKT cell neutralization and NK cell depletion does not fully restore parasite survival, indicating that still other factors contribute to the decreased in vivo infectivity.

Neutrophils have a dual role during the MIL-R infection onset
Given the infection-induced neutrophil influx, depletion was performed in BALB/c mice to study the impact on MIL-R PpyRE9/DsRed infection and early IFN-γ burst (Fig 8). MIL-R PpyRE9/ DsRed parasites failed to establish following transient neutrophil depletion. Dissemination to the bone marrow and spleen was seen in the IgG2a isotype-treated mice but not in the neutrophil-depleted mice, indicating that neutrophils support MIL-R PpyRE9/DsRed dissemination. Furthermore, IFN-γ levels were significantly decreased in the absence of neutrophils during the first 24h of the infection compared to the levels in IgG2a-treated control mice. These data indicate that MIL-R parasites show a more reduced survival in neutropenic mice, in line with a protective role of neutrophils as initial host cell while contributing to triggering the initial IFN-γ response.

MIL increases MIL-R parasite survival by reducing the activation of NK and NKT cells
We recently described that the MIL-R parasite developed a drug dependency that translates into an increased in vivo survival in the presence of MIL, either by in vivo treatment or in vitro pre-incubation of the parasite [42]. Using bioluminescent imaging, increased survival was observed during the first weeks of infection with MIL-R PpyRE9/DsRed pre-incubated with MIL compared to a control MIL-R PpyRE9/DsRed infected group (Fig 9A and 9B). After 2 weeks, parasite numbers rapidly dropped in both groups until no differences could be observed by week 4 of infection. In the previous sections, it was shown that MIL-R PpyRE9/DsRed induced a strong activation of NK and NKT cells leading to an increased IFN-γ response. The pre-incubation of MIL-R PpyRE9/DsRed with MIL resulted in a 6-fold reduction of serum IFN-γ levels at 1 dpi (135.6 ± 40.0 pg/ml vs 21.4 ± 12.0 pg/ml; Fig 9C). This reduced IFN-γ production was found to be associated with a reduced activation of liver NK (p = 0.1) and NKT cells (p< 0.05; Figs 9D and S7). No induction of the IFN-γ reporter was observed in splenic NK and NKT cells at 1 dpi, nor was there any induction of IFN-γ in CD4 + and CD8 + T cells in the liver and spleen.

MIL increases the infectivity of MIL-R parasites in the skin after intradermal inoculation
To further confirm MIL-dependency of MIL-R parasites in more natural settings, the effect of MIL treatment was evaluated in BALB/c mice following intradermal (i.d.) inoculation with a natural MIL-R (LEM5159 PpyRE9 ) parasite, harboring a mutation in the LiROS3 subunit of the MT transporter. This i.d. infection model is a better representation of the natural infection route, although parasites in this model fail to disseminate to the visceral organs. Therefore, the effect of MIL treatment was studied at the inoculum site in the ear (Fig 10). In vivo MIL treatment of LEM5159 PpyRE9 infected mice resulted in higher parasite burdens at all time points during the first 4 weeks of the infection indicating MIL dependency of this natural MIL-R strain.

Discussion
Miltefosine (MIL) is currently the only oral drug available to treat VL with initial cure rates above 94% upon introduction in the Indian subcontinent [24,59]. Within one decade of use treatment failure rates during a 12 month follow-up increased to 10% in India and up to 20% in Nepal [32,33]. Despite this evolution, only a limited number of fully resistant field isolates have been described, of which two MIL-R L. infantum isolated from French HIV-infected VL patients who received several successive treatments with MIL and liposomal amphotericin B [35,36,39,60]. This might suggest a connection between the emergence of MIL-resistance in L. infantum parasites and the immunocompromised state of the host. Furthermore, the unresponsiveness of Brazilian L. infantum strains to MIL was recently attributed to the absence of a MIL-sensitivity locus that was shown to be present in the genomes of all sequenced L. infantum and L. donovani isolated in the Old World [61]. The scarce number of resistant field isolates is surprising considering that full resistance can be obtained via single mutations in a aminophospholipid transporter complex, encoded by the MT/Ros3 genes [37,40]. It was shown that experimentally induced MIL-R L. major and L. infantum are hampered by a severe in vivo fitness loss which renders them unable to mount a successful infection in mice [40,42]. Given this limited information, the present study investigated the impact of acquired resistance on in vivo host cell infection and initiation of immune responses using the experimentally induced MIL-R L. infantum LEM3323 line carrying an inactivating 2 bp deletion in the LiMT gene [37].
By generating fluorescent/bioluminescent reporter strains, the in vivo fitness loss of an experimental MIL-R parasite in BALB/c and C57Bl/6 mice was confirmed by a rapidly decreasing liver signal and the absence of successful dissemination to the spleen and bone marrow. When exploring the host cell infectivity, a higher number of infected cells was detected at 1 dpi in the liver and spleen of MIL-R infected mice. The number MIL-R infected cells progressively decreased by 21 dpi whereas the number of MIL-S infected cells increased as shown by tissue imprints and bioluminescent imaging corresponding to parasite expansion in the liver. Flow cytometric detection of the DsRed reporter suggests the occurrence of a progressively quiescent state and decreased reporter gene expression over the course of infection. Flow cytometry analysis of different infected cell populations further revealed differences in cellular tropism between MIL-S and MIL-R which were correlated with an increased influx and concomitant infection of monocytes, neutrophils and dendritic cells by MIL-R. As a result, MIL-R less efficiently established in F4/80 + macrophages.
Compared to MIL-S, MIL-R parasites also induced elevated IFN-γ levels at 1 dpi which was attributed to an increased activation of hepatic NK and NKT cells. This activation contributed to the reduced survival of MIL-R parasites in vivo. Both NK and NKT cells have previously been shown to produce IFN-γ during VL infection onset in mice: NKT cells are activated in a CD1d-and IgG1 isotype-treated BALB/C and C57Bl/6 mice. (C) Comparison between the serum IFN-γ levels during the first 24h of MIL-R PpyRE9/DsRed infection in anti-CD1d-and IgG1 isotype-treated BALB/c and C57Bl/6 mice. For the experiments in BALB/c mice, each infection group consisted of 6 BALB/c mice and for the experiments in C57Bl/6 mice, two independent repeat experiments were performed with 3 mice in each infection group. Results are expressed as mean ± SD ( � p�0.05 and �� p�0.01). https://doi.org/10.1371/journal.pntd.0009622.g007 CD1d-dependent manner by parasite lipid antigens, such as LPG and glycosphingolipid (GSL) presented on dendritic cells and Kupffer cells [21, [62][63][64]. Moreover, GSL-induced relocation of CD1d molecules into lipid rafts facilitates the interaction with NKT cells [62]. NK cells are known to become activated by LPG through engagement of this phosphoglycan with TLR2 [65]. Mutations in the LiMT/LiRos3 transporter disrupts the transportation of lipids affecting the lipid asymmetry of the plasma membrane by exposing phosphatidylethanolamine and phosphatidylcholine on the outer membrane [44]. MIL-R parasites were shown to contain more saturated phospholipid acyl chains and more short alkyl fatty acids, resulting in a higher membrane rigidity [66,67]. We hypothesize that this altered lipid composition of the MIL-R membrane directly results in increased NK and NKT cell activation and enhanced presentation of lipid antigens by CD1d molecules on dendritic cells and Kupffer cells following infection with the experimental MIL-R strain. However, despite the increased survival of the MIL-R strain in the absence of activated NK and NKT cells, fitness was not fully restored indicating the contribution of additional factors to its decreased in vivo survival. Our results also indicate that neutrophils contribute to the early IFN-γ burst against the experimental MIL-R strain. Several other studies already suggested the involvement of neutrophils in regulating the IFN-γ production during VL infection [68][69][70] although the exact mechanisms remain to be elucidated. Neutrophils are known to influence NKT cell activation. In neutrophilic mice, the number of hepatic NKT cells was shown to be reduced as well as the proportion of IFN-γ-producing NKT cells in response to αGalCer [71]. This reduced IFN-γ production was correlated with a reduced T-bet expression in NKT cells. Neutrophils can also indirectly influence IFN-γ production by NKT cells. In the skin, L. major infected neutrophils were shown to be efficiently captured by dendritic cells [72]. Similar interactions have not yet been addressed during VL infection but may contribute to increased CD1d presentation capacity of dendritic cells resulting in an increased NKT cell activation. The observation of reduced parasite numbers in the absence neutrophils corroborates the concept of a dual role for neutrophils during MIL-R infection. Neutrophils were shown to be the major infected population during the first 24 hours of intradermal inoculation with L. infantum and are regarded as an important transient safe haven that facilitates passage of parasites to macrophages [73][74][75]. However, they are also known to contribute to protective antileishmanial immune responses as evidenced by a delayed hepatic granuloma formation, decreased expression of iNOS and IFN-γ and increased production of IL-4 and IL-10 in the absence of neutrophils [69,[74][75][76][77].
Our lab recently described drug-dependency of the experimental MIL-R parasite whereby the in vivo fitness loss in BALB/c mice was partially restored by exposure to MIL [42]. Analogous experiments in C57Bl/6 mice confirmed this drug effect. The present results showed that MIL-exposure lowers NK and NKT cell activation resulting in a decreased IFN-γ production early in the infection. The exact mechanism behind this MIL-induced modulation remains unclear, although interactions of MIL with the parasite's plasma membrane are likely to play a role. The interaction and incorporation of MIL in lipid monolayers have been described [78,79] and it was shown that MIL can increase the lipid dynamics of L. amazonensis membranes [80]. MIL is known as a lipid raft modulating agent [81] where the condensed microdomains of lipid rafts are suggested to incorporate high quantities of MIL [66,67]. This incorporation may affect the presentation of parasite antigens. The effect of MIL on other antiparasitic mechanisms should also be considered. For example, a recent study with MIL-tolerant L. donovani parasites from relapse patients showed a decreased production of ROS by infected macrophages following MIL exposure [34]. The drug-dependency of MIL-R parasites was further confirmed using the natural MIL-R LEM5159 L. infantum strain via an intradermal infection model which better represents the natural route of infection. In this model with the natural MIL-R parasite, MIL treatment also resulted in higher dermal parasite burdens. Altogether, these observations show that the experimental MIL-R L. infantum LEM3323 strain induces an enhanced antileishmanial innate immune response. Together with a different cellular tropism and the inability to successfully survive and multiply inside macrophages, this results in the rapid clearance from the liver and spleen. This increased immunogenicity might account for the scarce number of MIL-R field isolates and possibly explains why the few known L. infantum MIL-R field isolates were isolated from HIV-infected VL patients. Furthermore, these observations emphasize the risk of MIL treatment in sustaining infections with MIL-R parasites and provide an immunological basis for the attenuated and drug-dependent phenotype of resistant parasites.