Heterogeneity of F4/80⁺ macrophages during VL defined by the differential expression of CLEC4F and TIM-4

Kupffer cells in naïve mice are characterized as F4/80^hiCD11b^intCD64⁺CLEC4F⁺TIM-4⁺ (Supplementary Fig. 1a). At 42 days post L. infantum infection (42 d.p.i.), we identified 4 subsets of F4/80^hiCD11b^intCD64⁺ liver macrophages in C57BL/6 mice. The frequency of CLEC4F⁺TIM-4⁺resKCs was reduced, while the proportions of CLEC4F⁺TIM-4^–moKCs, and CLEC4F^–TIM-4⁺ and CLEC4F^–TIM-4^– macrophages were increased compared to naïve mice (Supplementary Fig. 1b–d). To investigate these subsets in situ, we performed confocal microscopy on liver sections at multiple time points (Fig. 1a). In naïve and 19 d.p.i. mice (“early-stage infection”), the majority of F4/80⁺ macrophages were CLEC4F⁺TIM-4⁺resKCs. By 42 d.p.i. (“late-stage infection”), these cells were significantly reduced in frequency and number, coinciding with a peak in the CLEC4F^–TIM-4⁺ and CLEC4F^–TIM-4^– populations. CLEC4F⁺TIM-4^–moKCs continued to accumulate during the “resolving phase” at 72 d.p.i. (Fig. 1a, b). Consistent with whole-liver observations, analysis of the cellularity within the granulomas revealed that CLEC4F⁺TIM-4⁺resKCs were the principal subset found within early-stage granulomas at 19 d.p.i. (Fig. 1c, d). However, by 42 d.p.i., these cells were significantly reduced in granulomas, which were predominantly populated by CLEC4F^–TIM-4⁺ and CLEC4F^–TIM-4⁻ cells. By 72 d.p.i., granulomas, while few in number, contained relatively equal proportions of all four F4/80⁺ subsets (Fig. 1c, d). Overall, resKCs were markedly reduced during late VL, in favor of F4/80⁺ macrophages lacking CLEC4F and/or TIM-4 expression. While early-stage granulomas were largely composed of resKCs, later infection resulted in granulomas containing heterogeneous macrophage populations.

a Top: Representative confocal microscopy images showing wild-type naïve, 19-, 42-, and 72-day infected livers stained with anti-F4/80 (cyan), anti-CLEC4F (white), and anti-TIM-4 (red). Bottom: Surface rendered F4/80⁺ cells classified according to their CLEC4F and TIM-4 expression, CLEC4F⁺TIM-4⁺ (yellow), CLEC4F^–TIM-4⁺ (red), CLEC4F⁺TIM-4^– (magenta), and CLEC4F^–TIM-4^– (cyan). Scale bars, 200 μm. b Frequency and number of F4/80⁺ cells classified according to their CLEC4F and TIM-4 expression, in naïve and infected mice at different days post-infection. Data pooled from 2 independent experiments (n = 6 for naïve and n = 7 for infected mice). c Granulomas identified in (a) are shown in the inset and were defined as clusters of F4/80⁺ cells with volumes greater than 1.03 × 10⁴ μm³. Scale bars, 40 μm. d Proportion of CLEC4F⁺ and/or TIM-4⁺ cells in F4/80⁺ granulomas at different times post-infection. Data pooled from 2 (19- and 72 d.p.i.) and 4 (42 d.p.i.) independent experiments (n = 7 for 19- and 72 d.p.i. and n = 15 for 42 d.p.i.). Values in (b and d) represent the mean ± SD. Source data are provided as a Source Data file.

TIM-4^– macrophages are monocyte-derived and KCs downregulate CLEC4F in late-stage granulomas

To confirm the origin of the macrophage subsets identified in the liver at 42 d.p.i., we generated congenically paired parabiotic mice, which share a chimeric blood supply. In naïve mice, nearly all F4/80^hiCD11b^intCD64⁺ macrophages carried the congenic marker of the host parabiont, corroborating their identity as TRMs. At 42 d.p.i., 10.5 ± 7.8% of the cells originated from the congenic partner (Fig. 2a, b). Further classification of CD45.2⁺ cells within infected livers of the CD45.1 parabiont revealed that TIM-4^– macrophages and monocytes contained cells from the congenic partner (Fig. 2c). In contrast, TIM-4⁺ macrophages were primarily from the host parabiont (Fig. 2c). These findings provide strong evidence that the absence of TIM-4 expression identifies macrophages of monocytic origin during VL.

a Representative contour plots showing the percentages of chimerism in F4/80⁺CD11b^intCD64⁺ cells from naïve and 42-day infected CD45.1⁺ parabiont. b Percentages of chimerism in F4/80⁺CD11b^intCD64⁺ cells from uninfected and infected CD45.1⁺ parabiotic partners at 42 d.p.i. c Frequency of CD45.2⁺ cells in infected CD45.1⁺ parabionts, gated on F4/80⁺CD11b^intCD64⁺CLEC4F^+/-TIM-4^+/- macrophages or Ly6C⁺CD11b⁺Ly6G^–SiglecF^– monocytes. Data pooled from two independent experiments (n = 3 for uninfected and n = 4 for infected pairs). d, e Number of monocytes in the livers of naïve (d) and 19- and 42-day infected WT and Ccr2^−/− mice (e) determined by flow cytometry. Data pooled from 5 independent experiments (naïve, n = 14 for WT and n = 17 for Ccr2^−/−; 19 d.p.i., n = 7 for WT and n = 9 for Ccr2^−/−; 42 d.p.i., n = 12 for WT and n = 14 for Ccr2^−/−). f Number of F4/80⁺ cells in WT and Ccr2^−/− at 42 d.p.i., quantified from confocal microscopy images. g Frequency of macrophage subsets based on CLEC4F and TIM-4 expression in WT and Ccr2^−/− mice at 42 d.p.i., quantified from confocal microscopy images. Data pooled from 2 independent experiments (n = 8 for WT and n = 9 for Ccr2^−/−). In (b–g) values show mean ± SD. In (d, f ) for data that passed the normality test, P values were obtained using a two-tailed unpaired t test. For data that did not pass the normality test, P values were obtained using a two-tailed Mann-Whitney test. In (c) P-values were obtained using ordinary one-way ANOVA with Tukey’s multiple comparisons, in (e) P-values were obtained using ordinary one-way ANOVA with Sidak’s multiple comparisons. h Representative dot plots from confocal microscopy images showing the frequency of macrophage subsets in infected WT and Ccr2^−/− mice at 42 d.p.i. Numbers indicate mean ± SD percentage of cells in the gate. i Representative rendered confocal microscopy images of 42 d.p.i. livers from WT and Ccr2^−/− mice showing CLEC4F⁺TIM-4⁺ (yellow) and CLEC4F^–TIM-4⁺ (red) macrophages, CLEC4F⁺TIM-4⁻ (magenta) moKCs and CLEC4F^–TIM-4^– (cyan) momacs. Scale bars, 200μm. Source data are provided as a Source Data file.

We also used Ccr2^−/− mice, in which recruitment of Ccr2⁺ monocytes to the liver was markedly reduced in both naïve (Fig. 2d) and infected mice (Fig. 2e). Confocal microscopy showed that although F4/80⁺ macrophage numbers were comparable between infected WT and Ccr2^−/− mice (Fig. 2f), the frequency of TIM-4^– subsets were significantly reduced in Ccr2^−/− mice (Fig. 2g). Conversely, the proportions of TIM-4⁺ subsets were increased compared to WT mice (Fig. 2g). These changes reflected overall differences in macrophage heterogeneity, with infected WT mice exhibiting all four F4/80⁺ subsets, whereas infected Ccr2^−/− mice predominantly showed TIM-4⁺ macrophages (Fig. 2h, i and Supplementary Fig. 2a).

It has been reported that residual monocyte infiltration in Ccr2^−/− livers is enough to repopulate an open KC niche due to increased proliferation of the few engrafted monocyte-derived cells⁶. In L. infantum-infected WT mice, around 60% of the proliferating cells were F4/80^– (Supplementary Fig. 2b). Among the proliferating F4/80⁺ cells, the majority were the CLEC4F^–TIM-4^–momacs, while resKCs corresponded to 8%, and CLEC4F^–TIM-4⁺ cells to 30% of the proliferating cells (Supplementary Fig. 2c, d). Thus, proliferation might support increased frequencies of CLEC4F^–TIM-4⁺ cells within WT granulomas. In infected Ccr2^−/− mice, 60% of the proliferating cells were again F4/80^– (Supplementary Fig. 2b), but proliferation was increased 4.6-fold in resKCs, and 1.7-fold in CLEC4F^–TIM-4⁺ cells compared to WT mice (Supplementary Fig. 2c, d), presumably to compensate the reduced monocyte infiltration in these mice.

The parabiosis experiments showing that the CLEC4F^–TIM-4⁺ cells were host-derived, in addition to their presence in granulomas of the Ccr2^−/− mice, suggested that these cells could be KCs that downregulated CLEC4F expression within granulomas. Breeding Clec4f^Cre-TdT mice with RCL-ZsGreen mice, which have a loxP-flanked STOP cassette preventing transcription of CAG promoter-driven ZsGreen, allowed us to distinguish CLEC4F active expression, marked by tdTomato, from its prior expression, indicated by ZsGreen. Live imaging of naïve Clec4f^Cre-TdTZsGreen mice confirmed that KCs were tdTomato⁺ZsGreen⁺ (Fig. 3a). At 19 d.p.i., some clusters of tdTomato⁺ZsGreen⁺KCs were observed (Fig. 3b, d). By 42 d.p.i, tdTomato⁺ZsGreen⁺KCs appeared as individual cells, while KC clusters were predominantly tdTomato^–ZsGreen⁺ (Fig. 3c, d). These results confirm that CLEC4F^–TIM-4⁺ cells are KCs that downregulate CLEC4F expression in late-stage granulomas.

a–c Representative live images from naïve (a), 19-day (b) and 42-day (c) infected Clec4f^Cre-TdTZsGreen mice, showing Clec4f active expression (red) and Clec4f previous expression (green). Scale bars, 100 μm. d Frequency of tdTomato⁺ cells identified in ZsGreen⁺ granulomas at 19 and 42 d.p.i. Data pooled from live imaging carried out in 4 independent experiments (n = 2 for 19 d.p.i. and n = 2 for 42 d.p.i.). Values represent mean ± SD. P-values were obtained using a two-tailed unpaired t test. e Sinusoid distribution from images (a–c), visualized by CD31 staining (magenta). Scale bars, 100 μm. f Representative images showing a tdTomato⁺ZsGreen⁺ cluster (red and green) and the sinusoids (magenta) at 19 d.p.i. Scale bars, 20 μm. g Representative images showing tdTomato^–ZsGreen⁺ clusters (green) and the sinusoids (magenta) at 42 d.p.i. Scale bars, 50 μm. h Images of Clec4f^Cre-TdTZsGreen mice at 19 and 42 d.p.i., showing Clec4f active expression (red), Clec4f previous expression (green), F4/80 (cyan), and sinusoids (magenta). Scale bars, 200 μm. Inset 1, the dotted line indicates a small cluster of F4/80⁺tdTomato⁺ZsGreen⁺KCs in contact with intact sinusoids. Scale bars, 20 μm. Insets 2 and 3 show tdTomato^–ZsGreen⁺ clusters outside the sinusoids. Arrows point to tdTomato⁺ZsGreen⁺ F4/80⁺ or F4/80⁻ KCs within granulomas, next to or in contact with the sinusoids. Scale bars, 15 μm. Source data are provided as a Source Data file.

Taken together, infected livers exhibited an ontogenically heterogeneous macrophage population consisting of CLEC4F⁺TIM-4⁺resKCs and CLEC4F^–TIM-4⁺KCs, while CLEC4F⁺TIM-4⁻moKCs and CLEC4F^–TIM-4^–momacs were monocyte-derived and absent in infected Ccr2^−/− livers.

Late-stage granulomas remodel the sinusoidal network without causing vascular damage

Previous VL studies have shown that KCs can redistribute to form granulomas²³, and based on H&E staining it was suggested that granuloma expansion causes loss of the physical association between KCs and the sinusoids²⁶. These studies did not directly observe the sinusoidal network and could not determine whether KCs redistribution to form granulomas occurred inside or outside the sinusoids. By using CD31 to label the sinusoids in Clec4f^Cre-TdTZsGreen mice, we observed that L. infantum infection caused remodeling of the sinusoidal network (Fig. 3e). At 19 d.p.i., small early clusters were identified partially outside the sinusoids, but they did not alter sinusoid distribution or integrity (Fig. 3f and Supplementary Movie 1). At 42 d.p.i., changes in the sinusoidal network became evident around large ZsGreen clusters, which were mostly outside the blood vessels (Fig. 3g and Supplementary movie 2). The sinusoids appeared to be displaced, pushed outward and downward by the expanding granuloma in the perisinusoidal space (Fig. 3g, arrows, and Supplementary Movie 3). Importantly, there were no signs of damage to the vessel walls or changes in blood vessels diameter (Fig. 3g, arrows, and Supplementary Movie 3).

To further investigate if the formation of KC clusters was associated with damage or rupture to the sinusoids, we stained red blood cells (RBCs) with Ter119 and performed live imaging in WT mice at 3- and 6-w.p.i. RBCs were observed flowing within the sinusoids, however, no blood flow was detected in areas where granulomas were present (Supplementary Movies 4–7). Consistently, tracking of RBCs revealed their movement within the sinusoids but not in regions containing granulomas (Supplementary Fig. 3a). We did, however, detect RBCs within some F4/80⁺ clusters at 3 w.p.i. (Supplementary Movies 4 and 5) and in areas corresponding to granulomas at 6 w.p.i. (Supplementary Movies 6 and 7). However, unlike the dynamic streaming of RBCs observed within the sinusoids, these RBCs were mostly stationary. In addition, static RBCs were observed in association with several F4/80⁺ cells inside the sinusoids (Supplementary Movie 7).

Intravascular staining allows the distinction between tissue and blood-borne cells²⁸. By intravenous (i.v.) administration of anti-F4/80, we detected small F4/80⁺ tdTomato⁺ZsGreen⁺ clusters at 19 d.p.i., suggesting these cells remained at least partially in contact with the sinusoids/circulating blood (Fig. 3h, inset 1). By contrast, larger ZsGreen⁺ clusters at 19- and 42-d.p.i. were not stained following i.v. anti-F4/80 administration (Fig. 3h), suggesting a loss of contact with the sinusoids and the absence of blood flow into the perisinusoidal space. When detected inside granulomas, F4/80⁺ and/or tdTomato⁺ZsGreen⁺KCs were located closer to the sinusoids (Fig. 3h, arrows in insets 2 and 3).

To address the clonality of the KCs within granulomas, we bred Clec4f^Cre-TdT mice with R26R-Confetti mice in which Cre recombinase causes permanent expression of one of four possible fluorescent proteins, allowing us to trace clonal lineages. At 42 d.p.i., ~ 70% of the granulomas contained two or more colors of KCs (Fig. 4a, b). Thus, most granulomas were formed by the redistribution of polyclonal KCs and were not solely aggregates of self-proliferating KCs.

a Intravital microscopy images representing a 42-day infected liver from a Clec4f^Cre-TdTConfetti mouse, stained with anti-CD31 (magenta), and showing tdTomato (red-nucleus), RFP (red-cytoplasm), YFP (yellow-cytoplasm), GFP (green-nucleus), and CFP (blue-membrane) KCs. Scale bars, 50 μm. b Frequency of KC clones in granulomas based on KC colors. Data pooled from live imaging performed in 2 independent experiments (n = 2 and quantification of 284 granulomas). c Representative rendered immunofluorescence image of a 42-day infected, wild-type liver showing the spatial distribution of CLEC4F⁺TIM-4⁺resKCs (yellow), CLEC4F^–TIM-4⁺KCs (red), CLEC4F⁺TIM-4^– moKCs (magenta), CLEC4F^–TIM-4^–momacs (cyan), and sinusoids (green). Scale bars, 30 μm. d Bar graphs showing the localization of each macrophage subset according to their interaction with the sinusoids in 42-day infected and naïve, wild-type mice. Data pooled from 2 independent experiments using 4 naïve mice and 8 infected mice (42 d.p.i.). Frequencies were calculated from four regions of interest (ROIs) per infected mouse (n = 32 ROIs) and 2-3 ROIs per uninfected mouse (n = 11 ROIs). e Scatter plots from immunofluorescence images showing the frequency of each F4/80⁺ population based on their distribution outside or inside granulomas at 42 d.p.i. Data pooled from 3 independent experiments (n = 11). In (d, e) for data that passed the normality test, P-values were obtained using a two-tailed unpaired t test. For data that did not pass the normality test, P-values were obtained using a two-tailed Mann-Whitney test. Values from (b, d, e) represent mean ± SD. Source data are provided as a Source Data file.

Live imaging of Clec4f^Cre-TdTZsGreen and Clec4f^Cre-TdTConfetti infected livers allowed us to track KCs, but it did not discriminate resKCs from moKCs, both of which are CLEC4F⁺, or CLEC4F^–TIM-4^–momacs within granulomas. Using confocal microscopy on 42 d.p.i. wild-type livers, we showed that CLEC4F⁺TIM-4⁺resKCs and CLEC4F⁺TIM-4^–moKCs maintained their contact with the sinusoids. In contrast, nearly 60% of the CLEC4F^–TIM-4⁺KCs and CLEC4F^–TIM-4^–momacs had lost contact and were located outside the sinusoids (Fig. 4c, d). Consistently, most CLEC4F^–KCs and momacs were observed within granulomas, while resKCs and moKCs were predominantly outside granulomas (Fig. 4e).

Taken together, our findings suggest that resKCs cross the sinusoidal endothelium to form granulomas in the perisinusoidal space, creating an open niche that is subsequently filled by moKCs. As granulomas expand and other cells, such as momacs, surround the KC core, the sinusoidal network is displaced, leading to resKCs losing their contact with the sinusoids and downregulating their CLEC4F expression.

BACH1 regulates resKC proliferation and macrophage lipid peroxidation during VL

The replacement of resKCs by moKCs may also occur due to KC niche availability following cell death^8,9,15,17,29. Murine models of Listeria monocytogenes⁹ infection and viral hepatitis¹¹ have demonstrated a massive loss of resKCs, with rapid replacement by moKCs. In our chronic VL model, resKCs replacement was gradual, with CLEC4F⁺moKCs occupying the sinusoidal space and coexisting with remaining resKCs at 42 d.p.i. To assess the potential role of cell death in our VL model, we used mice deficient in MLKL, a key protein in the necroptotic cell death pathway, and mice deficient in Caspase 1, involved in the lytic, pyroptotic cell death pathway³⁰. At 42 d.p.i., no differences were observed in the frequencies of resKCs or moKCs in Mlkl^−/− (Supplementary Fig. 4a) or Casp1^−/− livers (Supplementary Fig. 4b). Using cleaved caspase 3 staining to detect apoptotic cells in situ at 42 d.p.i., we observed an increased number of clv-casp3⁺ cells in all F4/80⁺ subsets and in F4/80⁻ cells, except in resKCs (Supplementary Fig. 4c). During infection, although the majority of clv-casp3⁺ cells were F4/80^–, the frequency of apoptotic resKCs decreased, while clv-casp3⁺ cells increased among all other F4/80⁺ macrophages (Supplementary Fig. 4d, e).

Ferroptosis, an iron-dependent oxidative cell death characterized by lipid peroxidation and damage to biological membranes³¹, has been implicated in various hepatic diseases, including malaria^29,32. By flow cytometry, we found evidence of ferroptotic cell death in both TIM-4⁺ and TIM-4^– macrophages at 42 d.p.i. (Supplementary Fig. 5a). Glutathione peroxidase 4 (GPX4) mediates the reduction of phospholipid hydroperoxides using glutathione (GSH) as a co-factor³¹, and lower levels of GPX4 and/or GSH are commonly associated with ferroptosis^33,34,35,36. At 42 d.p.i., GSH levels were reduced compared to uninfected controls (Supplementary Fig. 5b). BACH1, a pro-oxidant factor that represses NRF2, a master regulator of host antioxidant responses^37,38, has been shown to reduce oxidative stress-mediated ferroptosis when deficient^38,39. TIM-4⁺ and TIM-4⁻ cells from Bach1^−/− mice at 42 d.p.i. exhibited reduced levels of lipid peroxidation (Supplementary Fig. 5c). Furthermore, in infected Bach1^−/− mice, the frequency and number of resKCs (Supplementary Fig. 5d–f, h) and the total number of F4/80⁺ macrophages (Supplementary Fig. 5g, h) were increased. In contrast, the frequencies of moKCs and momacs were reduced (Supplementary Fig. 5d, e), although their absolute numbers were the same compared to WT mice (Supplementary Fig. 5f, h). The frequency of infiltrating monocytes remained unchanged (Supplementary Fig. 5i). In infected Bach1^−/−, resKCs showed a 2.9-fold increase in proliferation compared to WT mice, while no difference in proliferation was observed among other F4/80⁺ (Supplementary Fig. 5j) or F4/80^– cells (Supplementary Fig. 5k). Despite the accumulation of resKCs, parasite loads did not change between WT and Bach1^−/− mice (Supplementary Fig. 5l).

Collectively, apoptosis was increased during VL in F4/80^– cells and in all F4/80⁺ cells except resKCs, while evidence of ferroptosis was observed in resKCs and other F4/80⁺ macrophages. In addition, the absence of BACH1 resulted in increased resKCs proliferation. The findings suggest that BACH1 negatively regulates resKC numbers through mechanisms involving reduced proliferation, increased lipid peroxidation, and potentially ferroptosis during the peak immune response at 42 d.p.i., without affecting parasite loads.

Macrophage heterogeneity during VL revealed by single-cell RNA sequencing

To further investigate the cell heterogeneity identified by confocal microscopy and flow cytometry, we sorted F4/80^hiCD11b^intCD64⁺ macrophages and performed single-cell RNA sequencing (scRNA-seq) on 7673 cells, of which 7352 were macrophages. The macrophage identification and nomenclature dataset from Remmerie et al.¹⁴ was used as a reference for mapping our single-cell data onto the UMAP structure. UMAP projection and clustering revealed 3 main clusters in naïve and 6 clusters in infected mice (Supplementary Fig. 6a). Re-clustering and manual annotation identified two distinct clusters of Transitioning monocytes in infected mice (Fig. 5a). The top 10 differentially expressed genes (DEGs) for all clusters are shown in a heatmap (Supplementary Fig. 6b), and a full list of DEGs for each cluster identified in our scRNA-seq dataset is presented in Supplementary Data 1. The conserved KC transcriptomic signature described by Guilliams et al.¹⁹ confirmed that cells in ResKCs, MoKCs, and Transitioning monocytes1 clusters expressed KC signature genes (Fig. 5b). In naïve mice, the ResKCs cluster contained only CLEC4F⁺TIM-4⁺resKCs, which lacked monocytic markers. In infected mice, the ResKCs cluster also included monocyte-derived macrophages, identified by Ccr2/Cxc3cr1 and/or lack of Clec4f/Timd4 expression, suggesting these were moKCs transcriptionally similar to resKCs that still had not acquired CLEC4F/TIM-4 expression (Fig. 5c, d and Supplementary Fig. 6b). Proliferating macrophages were also increased in infected compared to naïve mice (Fig. 5a, e and Supplementary Fig. 5a). Remmerie et al.¹⁴ identified the Mac1 cluster as pre-moKCs due to Clec1b expression, a marker for moKCs expressed earlier than Clec4f ⁸. We did not detect Clec1b or KC signature genes in cells from this cluster (Fig. 5b, d). However, given that these cells expressed several KC-associated transcription factors (Supplementary Fig. 6c), they might eventually differentiate into KCs in infected mice. The Mac2 cluster, previously identified as hepatic lipid-associated macrophages (hep-LAMS)¹⁴, showed higher expression of Spp1, Cd9, and Trem2, consistent with this macrophage subtype (Fig. 5f). Finally, the term Transitioning monocytes was used to describe cells exhibiting features of both monocytes and macrophages¹⁴. The expression of KC signature genes and multiple KC transcription factors in the Transitioning monocytes1 cluster (Fig. 5b and Supplementary Fig. 6c) may also suggest that these macrophages were committed to a KC fate.

a UMAP plot of scRNA-seq data from sorted live, single, CD45.2⁺F4/80⁺CD11b^intCD64⁺ cells from livers of uninfected and 42 d.p.i. mice, showing three main clusters for naïve and seven clusters for infected mice. b Annotated UMAP plot showing the single-cell expression of conserved KC signature genes as described by Guilliams et al.¹⁹ c Single-cell expression of Ccr2, Clec4f, and Timd4 in CD45.2⁺F4/80⁺CD11b^intCD64⁺ cells from naïve and 42 d.p.i. mice. d Average gene expression and the percentage of cells expressing each gene within the identified clusters in naïve and 42-day infected CD45.2⁺F4/80⁺CD11b^intCD64⁺ cells. e Frequency of the different subsets identified by scRNA-seq in naïve and 42 d.p.i. mice. f Heatmap showing the differential expression of Mac2-specific genes identified by Remmerie et al.¹⁴. Data include 1200 naïve cells and 6152 cells from 42 d.p.i. mice after QC filtering.

Of note, we could not identify a defined cluster containing CLEC4F^–TIM-4⁺ cells, which were well represented in 42-day infected livers by confocal microscopy. This absence may be due to cell recovery bias during ex vivo analysis, resulting from liver disaggregation, digestion⁴⁰, or high death rates among resident macrophages, causing an underrepresentation of resKCs and CLEC4F^–KCs vs monocyte-derived populations. These cells could also be dispersed within the clusters identified in our scRNA-seq. To reveal their transcriptional program(s), we sorted F4/80^hiCD11b^intCD64⁺CLEC4F^–TIM-4⁺ cells from 42-day infected mice liver samples and performed a new scRNA-seq analysis. The list of DEGs for each cluster identified is presented in Supplementary Data 2. A pseudo bulk comparison with our previously identified clusters showed that more than half of the sorted cells had gene expression profiles similar to resKCs and proliferating cells, while around 20% were grouped in the Transitioning Monocytes clusters (Supplementary Fig. 6d). The CLEC4F^–TIM-4⁺ sorted cells shared expression of several KC-associated transcription factors, although Nr1h3, Irf7, and Mafb expression was lower than compared to cells in the ResKCs cluster (Supplementary Fig. 6e). In addition, cells in the ResKCs cluster and CLEC4F^–TIM-4⁺ sorted cells displayed significant transcriptional similarity regarding KC identity genes¹⁹ (Supplementary Fig. 6f).

Functional heterogeneity of macrophage subsets and their contribution to L. infantum control

Functional analysis of our scRNA-seq revealed that in infected mice, the ResKCs cluster expressed genes associated with KC functions, such as lipid and iron metabolism¹⁹ (Fig. 6a). In addition, cells in this cluster expressed Cxcl13 and Ccl24 (Fig. 6b). While most resKCs retained their putative homeostatic roles, they were also responsive to the Th1 inflammatory environment of infected livers^41,42, as evidenced by their expression of Cxcl10, Cxcl9 and Il1a (Fig. 6b, g). CLEC4F^–TIM-4⁺ cells expressed the highest levels of Il10 and Il1a, and of some inflammatory chemokines, such as Cxcl2, Ccl3, and Ccl4 (Supplementary Fig. 7a). MoKCs and Mac1 clusters showed a chemokine-cytokine profile that resembled more resKCs than macrophages in other clusters but did not express lipid and iron metabolism-associated genes (Fig. 6b). Transitioning monocytes1 had the highest levels of pro-inflammatory chemokines and cytokines, including Tnf, Il1b, Cxcl9, Cxcl10, Cxcl12 (Fig. 5b). Confocal microscopy revealed that iNOS expression peaked at 42 d.p.i. (Fig. 6c, d and Supplementary Fig. 7b) and was confined to CLEC4F^–KCs and momacs within 42 d.p.i. granulomas (Fig. 6e, f). This suggests that Nos2-expressing cells in Transitioning monocytes1 and 2, and Mac2 clusters were likely localized within granulomas (Fig. 6g and Supplementary Fig. 7b). The highest expression of Cxcl10 and Tnf was also detected in the same Nos2 expressing clusters (Fig. 6g), reinforcing the pro-inflammatory nature of the granulomas. By contrast, Il10 was detected mostly in cells from ResKCs and MoKCs clusters (Fig. 6g), and in CLEC4F^–TIM-4⁺ sorted cells (Supplementary Fig. 7a).

a, b Relative expression of lipid and iron metabolism (a), and chemokines and cytokines (b) genes from sorted CD45.2⁺F4/80⁺CD11b^intCD64⁺ cells at 42 d.p.i. Data was downsampled to a maximum of 500 cells per cluster. c Number of iNOS⁺F4/80⁺ granulomas in WT mice, quantified from immunofluorescence images. Data pooled from 2 independent experiments (n = 7 for 19- and 42 d.p.i. and n = 6 for 72 d.p.i.). d Representative images showing F4/80(cyan), CLEC4F(white), TIM-4(red), and iNOS(green) in WT livers. Scale bars, 30 μm. e Number of F4/80⁺iNOS⁺ macrophages in WT livers at 42 d.p.i. Data pooled from 5 independent experiments (n = 19). f Representative images showing F4/8(cyan), CLEC4F(white), TIM-4(red), and iNOS(green) (top), F4/80⁺iNOS⁺ granulomas (bottom) in WT liver at 42 d.p.i. Scale bars, 100 μm. g Single-cell expression of Nos2, chemokines, and cytokines in CD45.2⁺F4/80⁺CD11b^intCD64⁺ cells at 42 d.p.i. h Frequency of infected cells within F4/80⁺ granulomas, obtained from 2-3 regions from each liver (ROIs=24 for 19 d.p.i., ROIs=23 for 42 d.p.i.). Data pooled from 2 independent experiments for each time point, 8 mice at 19- and 42 d.p.i., 7 mice at 72 d.p.i. i Representative images of F4/80⁺ macrophages in WT livers, stained with anti-F4/80(cyan), anti-CLEC4F(white), anti-TIM-4(red), and anti-L. infantum(yellow). Scale bars, 30 μm. j Number of F4/80⁺ granulomas quantified from immunofluorescence images. Data pooled from 2 independent experiments (n = 7 for 19- and 42 d.p.i., n = 6 for 72 d.p.i.). k Distribution of parasites inside and outside F4/80⁺ granulomas. Data pooled from 2 independent experiments (n = 7). l Parasite loads in WT and Ccr2^-/- mice at 42 d.p.i. Data pooled from 3 independent experiments (n = 12 for WT and n = 13 for Ccr2^−/−). m Number of L. infantum amastigotes per field in 42-day infected WT and Ccr2^−/− mice, quantified from confocal microscopy images. Data pooled from 2 independent experiments (n = 8). n Representative images showing F4/80(cyan), CLEC4F(white), TIM-4(red), and L. infantum(yellow) in WT and Ccr2 ^−/− livers at 42 d.p.i. Scale bars, 20 μm. Values in (c, e, h, j–m) represent the mean ± SD. P values were obtained in (c, e) using ordinary one-way ANOVA with Tukey’s multiple comparisons test, in (h, j) using Kruskal-Wallis test with Dunn’s multiple comparisons tests, in (k–m) using a two-tailed unpaired t test. Source data are provided as a Source Data file.

Altogether, cells from Transitioning monocytes1 and 2 and Mac2 clusters expressed iNOS and high levels of pro-inflammatory cytokines and chemokines. Cells from the ResKCs cluster and CLEC4F^–TIM-4⁺ sorted cells expressed some pro-inflammatory chemokines but also IL-10. MoKCs showed a more homeostatic and regulatory profile, expressing IL-10 and fewer chemokines and cytokines. Our data indicates significant functional heterogeneity among KCs and recruited macrophages, likely due to differences in ontogeny, length of time since engraftment, and spatial localization within the infected liver.

Previous studies using colloidal carbon or fluorescent nanobeads to label KCs before infection showed that L. donovani amastigotes were present in KCs within granulomas from 8-28 d.p.i^23,26. Using confocal microscopy, L. infantum was found mostly inside CLEC4F⁺TIM-4⁺resKCs at 19 d.p.i. (Fig. 6h, i and Supplementary Fig. 7c). Forty-two d.p.i. was the peak of granuloma formation (Fig. 6j), and amastigotes were mostly found in CLEC4F^–KCs within granulomas, with minimal presence in momacs (Fig. 6h, i, k and Supplementary Fig. 7c). This finding suggests that resKCs are the primary infected cells during early infection, which become CLEC4F^–KCs within granulomas. This raises the question as to what role the uninfected momacs are playing in the development of the anti-parasitic response in the liver.

Deficient monocyte recruitment during L. donovani infection has been previously shown to result in disorganized granulomas and increased parasite burdens^43,44,45. Similarly, we found that Ccr2 ^−/− mice had 16.8-fold more liver parasites than WT mice at 42 d.p.i. (Fig. 6l). Quantification by confocal microscopy showed that resKCs and CLEC4F^–KCs of the Ccr2^−/− mice harbored 4-fold more parasites than WT mice (Fig. 6m, n and Supplementary Fig. 7d). In the absence of recruited macrophages, CLEC4F^–KCs and some resKCs from Ccr2^−/− mice both expressed iNOS (Supplementary Fig. 7e, f) but could not control the infection as effectively as WT mice. Liver homogenates from infected WT and Ccr2^−/− mice showed a different cytokine and chemokine environment (Supplementary Fig. 7g). Chemokines associated with granuloma formation, such as CCL2, CCL3 and CXCL10 were increased in Ccr2^−/− compared to WT, as was IFN-γ. However, disease-promoting cytokines such as IL-10 and IL-6⁴⁶ were also increased. The contribution of momacs to hepatic resistance is likely due to their cooperative functions with CLEC4F^–KCs in the effector response. However, the increased proliferation of resKCs observed in infected Ccr2^−/− mice (Supplementary Fig. 2c) suggests that monocyte recruitment into the open niche might also serve to limit the homeostatic proliferation of resKCs, which better supports parasite survival and growth. In either case, a heterogeneous population containing both KCs and momacs is important for hepatic immunity in VL.