Leishmaniasis, vector-borne diseases, are endemic in various countries of the South. These diseases are caused by protozoan parasites with different clinical manifestations. In mammalian hosts, Leishmania are intracellular parasites and the macrophage-Leishmania interaction is critical for the outcome of the infection as macrophages are the main host cells for parasite replication. Parasite pathogenicity and virulence are not the only relevant factors that determine the broad spectrum of disease outcomes. The genetic background, which defines the host immune response, certainly contributes to the diversity of disease phenotypes and clinical manifestations. Studies using animal models such as susceptible BALB/c and resistant C57BL/6 mice have definitely helped better understanding the biological relevance of the host genetic profile in the course of Leishmania infection. Using gene expression data, we have previously investigated these two mouse models to gain a deeper understanding of the macrophage-Leishmania interaction that may influence the fate of the infection (1).

To elucidate mechanistic insights from whole-genome data, recent attempts have incorporated network propagation (2). In this framework, whole genome data is mapped onto networks of interacting proteins and the data is subsequently propagated through the network using stochastic processes such as random walk with restart. This leads to the identification of subnetworks that agglomerate much of the observed signals and identifies structures that may be related to biological function and pathways. Network propagation has been applied in the context of cancer mutations ( (3), multi-omics responses to drugs (4) as well as responses to infections (5).

Previous studies have compared the differential responses of macrophages from resistant and susceptible mice induced by Leishmania (6, 7). However, to the best of our knowledge, no study has focused on the impact of the basal transcriptomic profile of macrophages, from different host backgrounds, on their response to Leishmania infection.

In this study, we used time-resolved expression data combined with network propagation to investigate how differences in host genetic background could determine the outcome of pathogen infection. We therefore first compared the Macrophage Colony-Stimulating Factor (M-CSF)-induced transcriptomic profiles in uninfected monocytes derived from two different mouse strains with contrasting behavior in response to parasite infection and identified DEGs that reveal profound differences in immune-related transcriptomic landscapes. Next, we studied the temporal responses of the two mouse strains after infection with promastigotes of the protozoan parasite L. major and found candidate immune response genes with elevated expression response in the resistant mouse strain compared to the susceptible strain. Using the DEGs from the untreated host transcriptome profiles as seeds, we then performed network propagation with the NetCore method (8), mapped the dynamic infection response signals via homology to a large experimentally validated human protein-protein interaction network integrated from various public resources and identified rather distinct, host-driven network modules. Key genes of these modules were validated with time-resolved qRT-PCR experiments in the two mouse strains. In particular, we found large differences in immune response genes and metabolic processes such as oxidative phosphorylation, lipids and arginine metabolism that highlight the variability in host responses to infection.

In summary, we demonstrate that the host’s gene expression background determines to a large degree its response to L. major infection, and that the gene expression analysis combined with network propagation is an effective approach to help identifying dynamically altered mouse strain-specific networks that hold mechanistic information about these contrasting responses to infection.

In the present work we emphasize the crucial role of macrophage gene expression background in determining its response to L. major infection. Moreover, combining gene expression analysis with network propagation proves to be an effective strategy for identifying dynamically altered mouse strain-specific networks, which contain mechanistic information about these contrasting responses to infection. Comparison of the contrasting response of different mouse strains, reflecting differences in host response, would help define key determinants of an effective response. Our findings showed that the transcriptomic signatures of M-CSF in BALB/c and C57BL/6 BMdMs are significantly different with more than 200 genes differentially expressed between the two mouse strains. The graph-based analysis of the time-resolved gene expression responses upon Leishmania infection with network propagation, and using the DEGs in uninfected macrophages as seeds, identified pathways and biological processes that were activated in infected cells and driven by the host differences.

Global transcriptomic events associated with M-CSF-dependent monocyte-to-macrophage differentiation have been analyzed and associated with major changes in the global transcriptomic profile (17). Moreover, differential basal backgrounds in non-infected BALB/c vs. non-infected C57BL/6 L929 differentiated macrophages have also been previously reported (6). Among the DEGs between the two mouse strains, our analysis showed that Cathepsin E ( Supplementary File 1 ) is one of the most strongly modulated genes in the BALB/c derived macrophages. Indeed, a polymorphism in the promoter of this gene, disrupting the consensus binding sequence of the PU.1. transcription factor has been previously reported in hematopoietic cells from C57BL/6J mice (18). 25% of the DEGs genes belong to the immune response pathway, and half of the remaining genes are linked to genes related to the same pathway ( Figure 1 ). Among the BALB/c DEGs, we found Cxcl14 and Arginase 2 (Arg2). Cxcl14 is involved in antimicrobial immunity preceding the development of the inflammatory response (19) and plays a role in the course of inflammation (20). Indeed, Cxcl14 analogs effectively inhibited the production of nitric oxide, MCP-1, TNF-α and IL-6 from LPS-stimulated RAW264.7 cells (21). Arg2 promotes H. pylori immune evasion by restricting M1 macrophage activation and polyamine metabolism (22), macrophage iNOS protein expression and thus NO production, and pro-inflammatory cytokine responses (23). Its expression, together with Cxcl14 in M-CSF-differentiated BALB/c macrophages, might therefore contribute to limit NO production and cytokines synthesis in these cells in response to Leishmania infection. Furthermore, in a different system, it has been reported that secreted Cxcl14 promotes the recruitment of alternatively activated M2 macrophages (24). In contrast, in C57BL/6 monocytes, M-CSF differentially induces the two mouse Nlrp1 (NLR family pyrin domain containing 1) paralogues, Nlrp1a and Nlrp1b, that are able to assemble into an inflammasome complex (25, 26). NLRP1b is the most well characterized and has been shown to activate the inflammasome during infection with bacillus anthracis (27, 28), Listeria monocytogenes, Shigella flexneri (29) and Toxoplasma gondii infection (30, 31). The ubiquitin system has emerged as fundamental for the control of the inflammatory response and in particular for the activation of the inflammasome (32). Indeed, autoproteolytic processing within the function-to-find domain (FIIND) splits NLRP1b into two fragments that remain non-covalently bound together. The N-terminus is ubiquitinated, releasing the C-terminal fragment. The E3 ubiquitin ligases, which ensure the ubiquitination of the N-terminal part and thus its degradation by the proteasome, are therefore necessary for the activation of NLRP1. Two E3 ubiquitin ligases, Trim12 and ITCH, are differentially expressed in C57BL/6 differentiated macrophages and may play a role in NLRP1 activation.

The immune response plays a central role in determining disease outcome. The differences in the M-CSF induced basal transcriptomic profile could impact the macrophage response to infection. In particular, these profiles could account for the limited amplitude of the inflammatory response observed in Leishmania-infected BALB/c BMdMs, while the conditions that prevail in uninfected C57BL/6 BMdMs could explain the greater magnitude of the immune response induced by Leishmania infection in resistant BMDMs ( Figure 2 ). Similarly, L. amazoniensis has also been reported to induce a stronger immune response activation in BMDMs from C57BL/6 mice as compared to BALB/c mice (6). In vivo, these different amounts of cytokines/chemokines produced by the macrophages in different backgrounds could affect leukocyte recruitment, especially at the site of infection. As L. major proliferation relies on bone marrow-derived cells freshly attracted to the site of infection where they differentiate into macrophages (33), this will affect the proliferation and the spread of the parasite and therefore the disease progresses.

The network propagation approach, using the DEGs in uninfected macrophages as seeds, highlighted pathways and biological processes that depend on the DEG for their activation. The over-represented pathways in each largest module genes led to 65% of pathways that are different between the two mice strains. By analyzing each of these modules, we reveal various genes involved in cellular processes that have already been described as altered during infection as well as novel genes. Indeed, in the C57BL/6 network, we find the macrophage Migrating Inhibitory Factor (MIF) and the endoplasmic reticulum (ER)‐resident protein Stimulator-of-interferon genes (STING). MIF promotes the release of several pro-inflammatory cytokines (TNFα and IL-1), and the activation of inducible NO synthase (iNOS). It also orchestrates the normal flow of leukocytes in inflammatory tissues and is needed for protection against various protozoan infections (34). Indeed, MIF release is induced by T. cruzi infection and leads to parasite killing (35) and MIF-deficient T. brucei-infected mice showed a decreased inflammatory response and reduced infiltration of pathogenic monocytic cells and neutrophils (36). MIF is protective in Leishmania infection. It increases Leishmania elimination (37)while Mif−/− mice are highly susceptible to L. major (38). Furthermore, Mif polymorphism is associated with cutaneous leishmaniasis outcome (39).

In many different cell types, the cyclic GMP-AMP (cGAMP) synthase (cGAS)-STING signaling axis, results in the phosphorylation and dimerization of the transcription factor IRF3, triggering/eliciting the transcription of hundreds of interferon stimulatory genes (ISGs) and therefore a strong type I interferon response against microbial infections (40). The released DNA might also be from host origin and mtDNA, released by immune cells after LPS stimulation is immunostimulatory and able to contribute to NLRP3 activation through cGAS-STING stimulation (41). Furthermore, different nucleic acid sensors belonging to the Ifi-200/HIN-200 gene family are induced by M-CSF and differentially expressed in BMdMs obtained from the two mice strains. In response to Leishmania infection, Ifi proteins, either alone or in cooperation with cGAS-STING, could sense cytosolic DNA and initiate innate immune responses against microbial infection (42). More recently, STING-dependent type I IFN production has been shown to regulate the metabolic changes induced by Mycobacterium tuberculosis infection in BMdMs (43). Activation of STING requires palmitoylation once translocated to the Golgi (44) and an ER-associated Zinc Finger, DHHC-Type Palmitoyltransferase 17 (ZDHHC17), is present in the shared network module.

Thus, Mif and Sting expression in the C57BL/6 module is consistent with the ability of these mice to restrict the parasite infection and to induce the repolarization of M2 macrophages to an M1 phenotype (45).

We checked plausibility of the network approach by validation of well-known marker genes and pathways. The inferred network modules, however, carry also many novelties and potential innovative biomarkers and cross-talk between different pathways. To highlight this, we performed additional pathway over-representation analysis with the 194 genes found either in the C57BL/6 or in the BALB/c network modules and found multiple enriched pathways, in particular metabolic and signaling pathways ( Supplementary Table S3.5 ; Supplementary Figure 4A ). A gene prominently expressed in many of these pathways is Pla2G4a (phospholipase A2 group IVA) which is higher expressed in BALB/c compared to C57BL/6 macrophages ( Supplementary Figure 4B ). Interestingly, in a recent study Pla2G4a among others was found as a signature of infected mouse liver after infection with L. donovani (46). In total, 52 genes associated with metabolism are among the 194 genes of the two network modules ( Supplementary Table 3.6 ). Another metabolic gene that appears higher responsive in infected BALB/c compared to C57BL/6 macrophages is Gbe1 (1,4-Alpha-Glucan Branching Enzyme 1). This enzyme is essential for glycogen production and thus energy storage and its upregulation has only recently been identified as a potential biomarker of early infection with L. donovani (46).

Cellular metabolism plays a crucial role in the functional phenotype acquired by macrophages and in the regulation of macrophage immune response and inflammation. Metabolic reprogramming has therefore emerged as a promising strategy for fighting microbes. Several genes identified in the C57BL/6 and/or BALB/c network belong to cellular metabolism, highlighting the importance of metabolic pathways in macrophage response to Leishmania infection. We and others have previously reported the metabolic reprogramming of infected macrophages toward aerobic glycolysis (47, 48). This change in host energy metabolism is reflected in the transcription upregulation of key glycolytic enzymes and glucose transporters. Macrophage metabolic reprogramming and the induced glycolytic enzymes transcription during infection, is under the control of HIF-1 transcription factor (HIF-1). Although commonly over-represented, the enrichment p-value and HIF-1 signaling target genes are different with respect to the two mouse strains (data not shown). Limiting host glycolysis by targeting metabolic enzymes may be a strategy for Leishmania to facilitate its growth (49). Indeed, as assessed by qRT-PCR, transcript levels of these genes are higher in C57BL/6 BMdMs ( Supplementary Figure 3 ) which may also contribute to the enhanced immune response observed in these cells. Indeed, the macrophage immune response is also metabolically controlled, and various glycolytic enzymes have been shown to play a role in cytokine expression (50). Conversely, cytokines also regulate metabolic pathway (51). Disruption of TCA cycle contributes to accumulation of citrate that is transported from the mitochondria to cytosol, where it is converted by citrate lyase to acetyl CoA and used for FA synthesis. The multifunctional fatty acid synthase enzyme (FASN), present in the BALB/c module network, catalyzes the condensation of acetyl-CoA and malonyl-CoA and leads to the generation of palmitic acid, a fully saturated 16-carbon FA. Condensation of serine and palmitate (PAL; derived from palmitoyl-CoA) initiates the synthesis of sphingolipids that represent a key class of bioactive lipids, such as sphingosine 1P. S1P affects macrophage biology in different ways (52) and its biological effect is regulated by the relative expression level of the different S1P receptors (S1PR) on immune cells. S1PR2, expressed by innate immune cells, is also present in the BALB/c module network. The role of S1P signaling during Leishmania donovani infection has been investigated and has been reported to decrease the expression of S1PR2. This inhibition significantly decreases IL-12 secretion and increases the parasite load as a result of increased IL-10 secretion (53).

Another metabolism highlighted by our analysis is that of arginine, which ultimately leads to the production of a set of biochemically diverse products, including nitric oxide. Network propagation results concerning genes related to this pathway suggest that its regulation is different depending on the genetic origin of the macrophages. Indeed, while Nos2 is present in both BALB/c and C57BL/6 networks, Ass1 is specific to the one of C57BL/6. Arginine can directly be acquired from the extracellular medium but can also be obtained from citrulline recycling via Ass1 expression, when extracellular arginine is depleted. This has been demonstrated during mycobacterial infection of murine macrophages (54). Moreover, recycling activity is required for controlling Mycobacterium bovis infection, as Ass1-deficient macrophages fail to scavenge citrulline under arginine-depleted conditions, rendering them unable to control mycobacterial infection (54). Citrulline recycling via ASS1 could therefore provide the arginine required for the sustained activation of iNOS observed in C57BL/6-derived BMdMs ( Figure 4 ) as arginine from citrulline recycling is the preferred substrate for NO production (55). Therefore, despite the enhanced transcription level of both Arg1 and NOS2 observed in C57BL/6-derived BMdMs, citrulline recovery would benefit NO synthesis.

The activation of ASS1 in BMdMs derived from resistant mice could be driven by a decreased uptake of available extracellular arginine by these macrophages. Indeed, previous studies reported reduced expression of the arginine transporter (Slc7a2), in C57BL/6 macrophages (compared to BALB/c) (7) resulting in decreased uptake of available extracellular arginine in response to cytokines (56). Our transcriptomic results show that Leishmania infection also induced a more robust expression of Slc7a2 mRNA in BMdMs derived from BALB/c (data not shown). Therefore, the reduced expression of the arginine transporter in C57BL/6 BMdMs would result in a low capacity to import extracellular arginine, inducing activation of Ass1 and the recycling of citrulline that maintain arginine availability and sustain NO production.

We have previously reported that several enzymes of the GSH system are differentially regulated in resistant and susceptible BMdMs suggesting glutathione accumulation in macrophages derived from resistant mice (57). In the C57BL/6 network module, we also notice the presence of the multidrug resistance protein 1 (MRP1/ABCC1) which in addition to its role in drug efflux, contributes at the physiological level to GSH export. Moreover, MRP1 has been reported to be involved in the NO-mediated Fe and GSH efflux (58). Enhanced NO production and GSH accumulation induced by the parasite infection in C57BL/6 mice, could result in NO efflux, in the form of dinitrosyl iron complexes (DNICs), mediated by the glutathione (GSH) transporter, MRP1/ABCC1. Especially that in the C57BL/6 module is also present the transferrin receptor (TFRc), implicated in iron uptake. NO storage and transport system protects the cell from large quantities of endogenously generated NO and might be responsible for delivering large quantities of cytotoxic NO as DNICs via MRP1 from M1macrophages to infected cells. Finally, besides its direct antimicrobial activity, NO play a key role in reshaping macrophage metabolism (59) and might also restrict the local enrollment of inflammatory cells and therefore limit the accessibility of cellular niche required for pathogen replication (60).

The genes present in our lists contain also many others whose involvement in the response to Leishmania infection has not yet been investigated. These genes that belong to different pathways and biological processes, could be interesting candidates to explore for a better understanding of the macrophage-Leishmania interaction.

The finding that the networks stemming from these differentially expressed genes yield a compilation of genes and pathways recognized as crucial in the macrophage reaction to Leishmania infection emphasizes the significance of the fundamental transcriptomic profiles of macrophages, and thus, of these DEGs, in driving macrophage polarization and its response to parasite infection. They are therefore, an interesting lead that can yield valuable results such as the identification of potential biomarkers that can improve disease screening and provide drug targets. Further work and in vivo validations in mouse models are required to identify and assess these potential biomarkers. Their applicability to humans should be validated in the human cells and models before being used to guide personalized treatment decisions in the clinic.

Publicly available datasets were analyzed in this study. This data can be found here: GSE31995 and GSE31996.

AB, RH, and LG-T conceived this study and were in charge of overall direction and planning. They also analyzed the data and wrote the manuscript. CB, IR, SR, KT, GB, and HT generated and contributed to the interpretation of the results and provided critical feedback. AB, RH, and CB generated the figures. All authors contributed to the article and approved the submitted version.