Alternatively activated monocyte-derived myeloid cells promote extracellular pathogen persistence in pulmonary fungal granulomas

Abstract

Inhaled fungal pathogens often generate granuloma-contained latent infections that can reactivate to cause invasive disease. However, why protective pathways fail to achieve sterilizing immunity in this setting is unclear. Here, we identify type 2 inflammation as a major arm of immunosuppression during latent, granulomatous fungal infection. Using reporter mice, we found that T_H2 cells were the dominant source of type 2 cytokines and deletion of IL-4/IL-13, Stat6, or T_H2 cells drove fungal clearance. Type 2 signaling acted on monocyte-derived myeloid cells that formed an ARG1⁺ ring around the granuloma core. STAT6 was required cell-intrinsically in this compartment, and its loss reduced lung burden. Contrary to an intracellular-niche model, we found that Cryptococcus was predominantly extracellular in vivo. STAT6 deficiency did not enhance macrophage intracellular killing, but instead antagonized IFNγ-dependent protection against extracellular yeast. These findings identify a spatially organized T_H2-STAT6 checkpoint that limits IFNγ-mediated extracellular killing and maintains cryptococcal latency.

Short summary

During latent pulmonary cryptococcal infection, Zheng et al. show that T_H2-driven STAT6 signaling in monocyte-derived myeloid cells suppresses extracellular fungal clearance, revealing a spatially organized immune checkpoint that promotes pathogen persistence within granulomas.

A key feature of many fungal pathogens that cause human disease is their ability to drive latent, persistent infections (Brunet et al., 2018). Fungi such as Cryptococcus, Histoplasma, and Coccidioides, can establish latent infections contained within pulmonary granulomas (Ristow and Davis, 2021, Demkowicz and Procop, 2021, Miranda and Hoyer, 2023), microanatomical structures composed of heterogenous immune cells (Pagán and Ramakrishnan, 2018). Although asymptomatic in healthy individuals, these infections can reactivate to cause invasive, disseminated disease in patients with suppressed immune systems (May et al., 2016). For instance, Cryptococcus neoformans (C. neoformans) causes asymptomatic latent pulmonary infections in immunocompetent people but life-threatening meningitis in patients with HIV infection, which accounts for 19% of AIDS-related mortality (Rajasingham et al., 2022).

Interferon gamma (IFNγ) is essential for host protection against C. neoformans; patients with neutralizing anti-IFNγ autoantibodies are highly susceptible, and mice deficient in IFNγ or treated with IFNγ-blocking antibodies succumb rapidly to serotype D infections (Zeng et al., 2019, Davis et al., 2023, Kawakami et al., 1996, Chen et al., 2005). IFNγ can drive sterilizing immunity in some contexts, as C. neoformans strains engineered to overexpress IFNγ are rapidly cleared, and IFNγ is indispensable for vaccine-induced protection (Hardison et al., 2010, Wang et al., 2024). However, despite evidence that it is induced during latent infection (Betancourt et al., 2025), IFNγ fails to eliminate the pathogen. A central unresolved issue is why this protective axis does not achieve sterilizing immunity, and specifically whether there are active immunosuppressive circuits that blunt IFNγ-effector functions.

Infections with Cryptococcus serotype A and D are uniformly fatal in immunocompetent C57BL/6J (B6) mice, and the immune response is overwhelmingly biased towards type 2 inflammation, characterized by T helper (T_H) 2 cells, eosinophilia, and alternative activation of macrophages (Lionakis et al., 2023, Wiesner et al., 2017, Wiesner et al., 2015). Genetic disruption of type 2 signaling in this setting reduces lung fungal burden (Müller et al., 2013, Dang et al., 2022). However, these lethal models poorly mirror human infection, which in immunocompetent individuals is either rapidly cleared or follows a latent course.

Granulomas are a hallmark of latent cryptococcosis (Ristow and Davis, 2021). Once viewed primarily as barriers to dissemination, they are increasingly linked to pathogen persistence (Pagán and Ramakrishnan, 2018). In zebrafish infected with Mycobacterium marinum, granuloma macrophage recruitment sustains bacterial persistence by providing a renewable niche (Flynn et al., 2011, Berg et al., 2016, Davis and Ramakrishnan, 2009, Pagán et al., 2015). Infections using a clinical isolate of Cryptococcus neoformans (UgCl223) in C3HeB/FeJ (FeJ) mice, which carry a susceptibility mutation affecting responses to several infections (Pan et al., 2005, He et al., 2013, Bergmann et al., 2013), achieve cryptococcal latency with organized granulomas (Betancourt et al., 2025). In this model, granulomas feature a myeloid core with a neutrophil-rich center and a peripheral T cell cuff, reminiscent of human Mycobacterium tuberculosis granulomas. Interestingly, the immune response to UgCl223 transitions from an early wave of type 2 inflammation to a later type 1 response (Betancourt et al., 2025). UgCl223 can induce latency without organized granuloma formation in C57BL/6J (B6) mice (Ding et al., 2022), the background on which most murine genetic tools exist, and no clinical isolate consistently produces both latency and fully organized granulomas in this background. By contrast, B6 infection with a C. neoformans strain lacking glucosylceramide synthase (gcs1Δ) yields robust pulmonary granulomas, latent, asymptomatic infection, and reactivation upon blockade of lymphocyte recirculation (Rittershaus et al., 2006, Bryan et al., 2020, McQuiston et al., 2010, Farnoud et al., 2015). This enables the use of B6 genetic tools with a tractable pathogen to dissect protective versus detrimental pathways in granulomatous cryptococcosis (Ristow and Davis, 2021, Chun and Madhani, 2010).

Here, we show that type 2 inflammation is an immunosuppressive feature of latent, granulomatous Cryptococcus infection that limits pathogen clearance. Using spatial transcriptomics to assess the cellular and molecular features of gcs1Δ cryptococcal granulomas, we found that these structures contain an inner core of type 2 cytokine-responsive myeloid cells, which we subsequently identify as monocyte-derived macrophages that expand during infection and are found in both gcs1Δ and UgCl223 infection. These macrophages are alternatively activated by type 2 cytokines derived from T_H2 cells, and this circuit antagonizes protective immunity during latent Cryptococcus infection. Finally, we find that the detrimental effects of type 2 inflammation act on extracellular fungal persistence, which contrasts with the prevailing view of Cryptococcus as a dominantly intracellular pathogen.

Results

B6 mice can control KN99α gcs1Δ to form a latent infection contained by pulmonary granulomas.

To investigate the mechanisms underlying Cryptococcus latency, we sought to establish latent, granulomatous infection in B6 mice, the background on which most murine genetic tools exist. Consistent with previous reports, wild type (KN99α) Cryptococcus neoformans induced a lethal infection, where we could not observe clear protective roles for IFNγ (Ifng^−/−) or adaptive immunity (Rag1^−/−) (Davis et al., 2023) (Fig. S1 A). Moreover, even Il4/Il13^−/− mice had a mortality rate as high as 70% (Fig. S1 A). Rather than displaying anatomical confinement, wild type KN99α Cryptococcus infection was diffuse across the pulmonary tissue (Fig. S1 B). Additionally, we found that IFNγ production by lymphocytes was unable to be maintained as infection progressed (Fig. S1 C), and iNOS production by myeloid cells could not be clearly observed at any timepoint during lethal infection (Fig. S1 D), demonstrating a failure to induce protection in wild type hosts, which contrasts with human Cryptococcus infection where immunocompetent hosts normally do not succumb to infection.

On the other hand, gcs1Δ Cryptococcus was reported to be capable of inducing a latent infection in B6 mice (Rittershaus et al., 2006). Using gcs1Δ on a KN99α background, we found that lung lesions at 21 dpi showed non-caseating granulomas composed of cohesive epithelioid macrophage aggregates with a thin lymphocytic rim (Fig. 1A). By 35 dpi, lesions matured into organized nodules with clear zoning, a compact epithelioid core encircled by a peripheral lymphocytic cuff, with compression of adjacent alveoli (Fig. 1A). This pattern is broadly consistent with clinicopathologic descriptions of cryptococcal granulomas (Ristow and Davis, 2021, Betancourt et al., 2025), while allowing for model-specific differences. We performed thick-tissue imaging with CD11c^Yfp reporter mice infected with a gcs1Δ mCherry fluorescent reporter strain to visualize localization of myeloid cells and yeast during latent infection. We observed CD11c⁺ cells surrounding C. neoformans and forming a spatially confined structure (Fig. 1B).

Fig. 1. B6 mice can control KN99α gcs1Δ to form a latent infection contained by pulmonary granulomas.

(A) Representative images for H&E staining of lung tissues from mice infected with KN99α gcs1Δ at 21 dpi and 35 dpi. Scale bar, 100 μm.

(B) Representative images for thick-tissue imaging of lungs from KN99α gcs1Δ-infected CD11c^Yfp reporter mice at 35 dpi. Scale bar, 400 μm.

(C) Colony-forming unit (CFU) analysis of lung tissue in mice infected KN99α gcs1Δ infection across time courses (5 mice per time point).

(D) UMAP representation of merged images from all replicates at each time point. 4 to 5 mice per time point. 17 sections from 17 mice in total.

(E) Dot plots for top differentially expressed genes by different types of cells acquired by spatial transcriptomics.

(F) Representative Images for different types of cells within granulomatous area in the lung tissue from mouse at 56 dpi (D56#3). Scale bar, 1000 μm.

(G) Representative images for the spatial expression of Cxcl12, Cxcl10, Cxcl16, and Cxcl13 in sample D56#3. Scale bar, 1000 μm.

To confirm that gcs1Δ can drive a latent infection, we infected B6 mice and collected lungs for colony-forming unit assays (CFUs) across a time course. Pulmonary fungal loads peaked at 35 dpi at which point they began to contract, but CFUs remained detectable out to 140 days (Fig. 1C). To further test if gcs1Δ infection was a good mimic of latent infection controlled by adaptive immunity, we administered anti-CD4 antibodies to deplete CD4⁺ T helper cells starting from the peak of infection, 35 dpi (Fig. S1 E). Using intravascular labeling with anti-CD45-APC, flow cytometry analysis at 49 dpi showed a successful CD4 depletion both in circulation and in the lungs, although there were more residual CD4⁺ T cells in the pulmonary interstitium (Fig. S1, F and G). Consistent with a protective role for helper T cells, the pulmonary fungal burden in CD4-depleted mice was markedly higher than that in isotype control-treated mice (Fig. S1 H). Taken together, and consistent with previous reports using this model (Rittershaus et al., 2006, Bryan et al., 2020), these data show that gcs1Δ induces a latent infection in B6 mice.

Spatial transcriptomics reveal the cellular architecture of cryptococcal granuloma in mouse models.

Although our histological analysis showed structures consistent with granulomas, we wanted to obtain a more comprehensive understanding of the cellular and molecular microanatomy of these lesions. Therefore, we conducted spatial transcriptomics utilizing the 10x Genomics Xenium Prime 5K kit, which utilizes an in situ hybridization strategy to provide an accurate spatial gene expression pattern using a panel of 5006 mouse gene probes. This kit also contains multiple staining (DAPI, boundary (CD45, E-cadherin, and APT1A1), 18S, α-SMA, and vimentin) to aid in cell segmentation, thus enabling single-cell resolution for gene expression and allowing for annotation of cell types in addition to revealing spatial gene expression patterns (Fig. S2 A).

We harvested lung tissues from infected mice at 28, 35, 42, and 56 dpi (4 to 5 mice per time point) and performed Xenium single-cell analysis to annotate cell types within the acquired sections. We were able to identify clusters pertaining to Mac/Mon, T lymphocytes, B lymphocytes, epithelial, endothelial, and stromal cells from all samples (Fig. 1, D and E). Immune cells (Mac/Mon, T and B cells) formed small aggregates at 21 dpi; by 35 dpi, tight myeloid clusters appeared in some mice; at 42 dpi, myeloid clusters were observed in every mouse although an organized architecture was evident in only one; by 56 dpi, all samples exhibited an organized structure: an inner myeloid-rich core, a contiguous intermediate “buffer” band of phenotypically distinct myeloid and stromal cells arranged as a ring, and an outer lymphocyte cuff enriched for T and B cell programs (Fig. 1 F; Fig. S1 B).

Consistent with highly organized microanatomy, we also observed spatially distinct chemokine expression. While there has been speculation that the lymphocyte cuff observed in Mtb granulomas is due to exclusion from the core, we found clear induction of Cxcl13 (Fig. 1G), which guides B cells into secondary lymphoid follicles, in the region of the cuff, suggesting that lymphocytes may be actively recruited to the ring rather than excluded from the center (Ansel et al., 2000).

The overall architecture of gcs1Δ-induced granulomas resembled the granulomas observed in UgCl223-infected FeJ mice. Another defining feature of UgCl223-induced granulomas in FeJ mice is a neutrophil-rich center within the myeloid core (Betancourt et al., 2025). By immunofluorescence, we found that gcs1Δ-induced granulomas likewise exhibited a neutrophilic center surrounded by CD11c⁺ myeloid cells at 35 dpi (Fig. 2A), arguing that this is a conserved feature of murine Cryptococcal granulomas. Taken together, gcs1Δ Cryptococcus induces granulomas in B6 mice that share key features with human Mtb granulomas and with UgCl223-induced granulomas in FeJ mice.

Fig. 2. Neutrophilic core in KN99α gcs1Δ-induced granulomas and structure of cryptococcal granuloma in human lymph node sample.

(A) Representative immunofluorescent images for lung tissue sections from KN99α gcs1Δ-infected mice at 35 dpi. Scale bar, 100 μm.

(B) H&E staining of human lymph node sample with cryptococcal granulomas. Scale bar, 2 mm. Box indicates the filed presented in (C).

(C) Representative images for granulomatous area with IHC staining of CD14, CD68, CD3, and CD4 on human lymph node sample with cryptococcal granulomas. Scale bar, 2 mm.

While pulmonary Cryptococcal granulomas from non-HIV patients are difficult to source, we were able to obtain a HIV patient-derived lymph-node cryptococcal granuloma to test whether the features observed in the murine setting extend to humans. H&E showed lobulated granulomas under the capsule with hypocellular centers (Fig. 2B). Staining with CD14/CD68 marked a macrophage-rich core. CD3 highlighted a peripheral T-cell cuff, with enrichment of CD4 around and between lobules (Fig. 2C). This zoning mirrors our mouse lung lesions, suggestive of a conserved Cryptococcal granuloma structure consisting of a macrophage core with a T lymphocyte-rich cuff, similar to what has been observed with Mtb granulomas. However, future studies on non-HIV pulmonary Cryptococcus granulomas are needed to fully dissect the conservation of microanatomy across species.

Distinct waves of type 1 and 2 immune responses characterize Cryptococcus latent infection.

In the UgCl223-induced latent model with FeJ mice, the immune response features an early type 2 inflammatory phase followed by delayed type 1 immunity (Betancourt et al., 2025). To test whether gcs1Δ infection induced a similar response, we performed flow cytometry analysis to systematically investigate T cell responses across a time course of gcs1Δ latent infection. Consistently, GATA3^hi T_H2 cells peaked in the lungs at 21 dpi, whereas T-bet⁺ T_H1 cells expanded strikingly from 35 to 42 dpi (Fig. 3A; Fig. S3 A). To assess cytokine production competency, we stimulated T cells with phorbol myristate acetate (PMA) and ionomycin and found that the dynamics of cytokine production generally followed the same kinetics as T cell transcription factor expression (Fig. 3, B and C). Intriguingly, we observed strikingly high IFNγ production by T_H1 cells at 42 dpi (Fig. 3C), when the pulmonary fungal burdens began to contract. Consistent with the flow cytometry data, multiplex cytokine assays showed that IL-4 and IL-13 levels peaked around 30 dpi, whereas IFNγ peaked at 40 dpi (Fig. 3, D–F). The kinetics of eosinophils and neutrophils further support early type 2 inflammation and a type 1 response that peaks at 42 dpi (Fig. 3, G and H). Overall, this pattern broadly parallels the immunity course reported for the UgCl223-FeJ model (Betancourt et al., 2025), suggesting an immune trajectory during latent Cryptococcus infection that is conserved across fungus and host genetic backgrounds.

Fig. 3. Type 2 immunity is detrimental during latent infection.

(A) Quantification of lung T_H1 and T_H2 cells in mice with KN99α gcs1Δ-induced latent infection across time courses (5 mice per time point).

(B) Representative flow cytometry plot showing IL-13 versus IFNγ production upon restimulation within CD4+FOXP3− T cell compartment in lung tissue from mice infected with KN99α gcs1Δ at 21 and 42 dpi.

(C) IFNγ⁺ and IL-5/IL-13⁺ frequency of CD4⁺FOXP3⁻ T cell compartment in lung tissue from mice infected with KN99α gcs1Δ across time courses (5 mice per time point).

(D) Quantification of IL-4 levels in lung tissues from KN99α gcs1Δ-infected mice across time courses (6 mice per time point).

(E) Quantification of IL-13 levels in lung tissues from KN99α gcs1Δ-infected mice across time courses (6 mice per time point).

(F) Quantification of IFNγ levels in lung tissues from KN99α gcs1Δ-infected mice across time courses (6 mice per time point).

(G) Quantification of eosinophils in lung tissue from KN99α gcs1Δ-infected mice across time courses (5 mice per time point).

(H) Quantification of neutrophils in lung tissue from KN99α gcs1Δ-infected mice across time courses (5 mice per time point).

(I) Colony-forming unit (CFU) analysis of lung tissue from WT and Il4/Il13^−/− mice infected with KN99α gcs1Δ at 140 dpi.

(J) CFU analysis of lung tissues from KN99α gcs1Δ-infected mice with full bone marrow chimeras (bone marrow donors: WT, Il4^−/−, Il13^−/−, and Il4/Il13^−/−) at 35 dpi. **P < 0.01 and ***P < 0.001 by One-way ANOVA

(K) Representative flow cytometry plots for hCD2 (reporting IL-4 translation) expression in eosinophil, basophil, and T_H2 cells in lung tissues from WT and Il4^KN2/+ mice infected with KN99α gcs1Δ at 35 dpi.

(L) Representative flow cytometry plots for hCD4 (reporting IL-13 transcription) expression in T_H2 cells and ILC2s in lung tissues from WT and Il14^KN4/+ mice infected with KN99α gcs1Δ at 35 dpi.

(M) CFU analysis of lung tissues from KN99α gcs1Δ-infected WT and Epo^DTA mice at 35 dpi. ns = no significance by unpaired two-sided Student’s t-test

(N) CFU analysis of lung tissues from KN99α gcs1Δ-infected Mcpt8^Cre and Mcpt8^CrexRosa26^Lsl-DTA mice at 35 dpi. ns = no significance by unpaired two-sided Student’s t-test

(O) Representative flow cytometry plots showing T_H2 cells and ILC2s in lung tissues from KN99α gcs1Δ-infected Gata3^fl/fl, Klrg1^Cre x Gata3 ^fl/fl, and hCD2^Cre x Gata3^fl/fl mice at 35 dpi. Top, gated from CD90.2⁺TCRβ⁻. Bottom, gated from CD90.2⁺TCRβ⁺CD4⁺FOXP3⁻.

(P) CFU analysis of lung tissues from KN99α gcs1Δ-infected Gata3^fl/fl, Klrg1^Cre x Gata3^fl/fl, and hCD2^Cre x Gata3^fl/fl mice at 35 dpi. Data are pooled from two independent experiments. **P < 0.01 and ns = no significance by One-way ANOVA.

(Q) Schematic of GATA3 temporal depletion in mice infected with KN99α gcs1Δ. Tamoxifen was administrated orally from 28 dpi for 5 days in KN99α gcs1Δ-infected Gata3fl/fl and Rosa26CreERT2 x Gata3 fl/fl mice.

(R) Representative histogram for GATA3 expression within CD90.2⁺ lymphocyte compartment in lung tissue from KN99α gcs1Δ-infected Gata3^fl/fl and Rosa26^CreERT2 x Gata3 ^fl/fl mice.

(S) CFU analysis of lung tissues from KN99α gcs1Δ-infected Gata3^fl/fl and Rosa26^CreERT2 x Gata3 ^fl/fl mice at 35 dpi. **P < 0.01 by unpaired two-sided Student’s t-test.

Type 2 immunity is detrimental during latent infection

Next, we leveraged the plethora of available B6 genetic tools to ask whether type 2 immunity is functionally important during latent fungal infection. First, we infected Il4/Il13^−/− mice and found they cleared pulmonary fungi at 140 dpi, whereas wild-type (WT) mice maintained low-level infectious burdens (Fig. 3I). To assess the relative contributions of IL-4 vs IL-13, we generated bone marrow chimeras with WT, Il4^−/−, Il13^−/−, and Il4/Il13^−/− donors transplanted into congenic irradiated recipients. After reconstitution and infection, we found that while both single cytokine knockouts resulted in significantly lower fungal burdens at 35 dpi, Il4^−/− had a more striking phenotype, albeit not as strong as the double cytokine knockout (Fig. 3J), arguing that there is a dominant immunosuppressive role for IL-4. These data suggest that hematopoietic IL-4 and IL-13 both contribute to fungal persistence to slightly different degrees; this is in line with previous reports that IL-13 receptor expression is more restricted to non-immune cells that would be less likely to influence pathogen burdens compared to the hematopoietic compartment, albeit there are clear examples of functional IL-13 signaling in immune cells (Bao and Reinhardt, 2015).

To determine the relevant producers of type 2 cytokines during latent infection, we next infected Il4^KN2/+ and Il13^KN4/+ mice to report endogenous IL-4 and IL-13 production, respectively (Mohrs et al., 2005, Liang et al., 2011). Consistent with previous studies on cytokine-producing cells during Nippostrongylus brasiliensis infection, we observed that eosinophils and basophils exclusively express IL-4, ILC2s exclusively express IL-13, while T_H2 cells produce both (Liang et al., 2011) (Fig. 3, K and L). We next asked whether there was a dominant type 2 cytokine producing cell (as opposed to a redundant network) antagonizing fungal clearance. To address this, we infected Epo^DTA (eosinophil-deficient), Mcpt8^Cre x Rosa26^LSL−DTA (basophil-deficient), Klrg1^Cre x Gata3^fl/fl (ILC2-deficient (Gurram et al., 2023)), hCD2^Cre x Gata3^fl/fl (T_H2-deficient (Gurram et al., 2023)), and paired controls with gcs1Δ to assess CFUs at day 35. Eosinophil and basophil deficiency did not show a difference in pulmonary fungal burden (Fig. 3, M and N), nor did ILC2 deficiency (Fig. 3, O and P). On the other hand, T_H2-deficient mice phenocopied the IL-4 and IL-13 double-knockout animals in displaying a dramatic reduction in pulmonary fungal burden (Fig. 3, O and P), suggesting a dominant role for T_H2 cells in establishing a type 2 inflammatory environment during latent infection that prevents fungal clearance.

To test whether immunosuppressive type 2 immunity is a conserved feature of latent Cryptococcus infection, we infected Stat6^−/− and wildtype B6 mice with UgCl223. UgCl223 infection induced eosinophils and T_H2 cells in wildtype B6 mice but not in Stat6^−/− mice (Fig. S3, B–D). Importantly, Stat6^−/−, Il4/Il13^−/−, and T_H2-deficient B6 mice infected with UgCl223 showed significantly reduced fungal burdens in their lung tissues compared to controls (Fig. S3, E–G), indicating that the detrimental effect of T_H2-driven type 2 immunity is not limited to gcs1Δ-induced latent infection.

To address the potential caveat that fungal infection fails to establish in the absence of type 2 inflammation, we asked whether temporal deletion of GATA3 during on ongoing infection would result in decreased pulmonary burdens. To test this, we infected Rosa26^CreERT2 x Gata3^fl/fl versus Gata3^fl/fl control mice, started tamoxifen administration from 28 dpi, and performed CFU assays one week later at 35 dpi (Fig. 3Q). Flow cytometry showed a successful ablation of GATA3 expression in T lymphocytes (Fig. 3R), which resulted in a significantly lower pulmonary fungal burden after just 7 days of deletion (Fig. 3S), arguing that an ongoing type 2 response actively antagonizes fungal clearance.

STAT6 activation induces type 2 responsive ARG1⁺ myeloid cells during latent infection

Macrophages and monocytes (Mac/Mon) are innate immune cells that can shape their effector outputs in response to lymphocyte-derived cytokines (Lawrence and Natoli, 2011). ARG1 is a well-characterized readout of type 2 signaling in macrophages (often alternatively activated macrophages, AAMs). By flow cytometry, ARG1 was detected in CD11b⁺CD64⁺ myeloid cells including IMs, moDCs, and CD64⁺ monocytes (Fig. 4A; Fig. S4, A–C). Because ARG1 can be induced by other cues such as STAT3-activating cytokines or hypoxia (Lin and Simon, 2016, Colegio et al., 2014, El Kasmi et al., 2008), we tested whether ARG1 induction during Cryptococcus latent infection was due to cell-intrinsic STAT6 signaling by generating mixed bone marrow chimeras whereby congenic markers can be utilized to recognize the genotype of cells competed in the same recipient. To minimize any potential decrease in fungal burden in Stat6^−/− bone marrow mixes from cells that may gain increased killing capacity, we mixed 30% Stat6^+/+ (WT, CD45.2) or Stat6^−/− (Stat6 deficient, CD45.2) bone marrow with 70% WT bone marrow from BoyJ mice (WT, CD45.1) (Fig. 4B). We then transplanted these mixes into irradiated recipients (WT, CD45.1/2). After reconstitution, we infected these chimeras and performed flow cytometry analysis. Within CD64⁺CD11b⁺ myeloid cells, which we found to represent the major ARG1 expressors, Stat6^+/+ cells expressed the same level of ARG1 compared to CD45.1 WT cells (Fig. 4C). In contrast, Stat6^−/− cells lost almost all ARG1 expression, while CD45.1 WT cells in the same recipient still expressed ARG1 (Fig. 4C), albeit to slightly diminished levels compared to the control mixes. We further gated on all ARG1⁺ myeloid cells to see the contribution from CD45.1 and CD45.2. The contribution of CD45.2 in BoyJ:Stat6^+/+ group was around 30%, whereas in BoyJ:Stat6^−/− mixes the representation of knockout cells was significantly decreased (Fig. 4D). We further included PD-L2 as another type 2 marker (Loke and Allison, 2003) and found the same intrinsic loss in Stat6^−/− cells compared to WT cells (Fig. 4, E and F). These data suggest that the ARG1⁺ myeloid cells are induced by type 2 cytokine signaling in a STAT6-dependent manner. We also observed a similar ARG1⁺PD-L2⁺ AAM population across multiple latent infection models at 35 dpi including UgCl223-infected B6 mice (Ding et al., 2022), UgCl223-infected FeJ mice (Betancourt et al., 2025), and gcs1Δ-infected CBA/J mice (Rittershaus et al., 2006, Bryan et al., 2020) (originally used for gcs1-related studies) (Fig. S5 A), consistent with a conserved type 2 program in myeloid cells during latent Cryptococcus infection.

Fig. 4. ARG1 expression marks a spatially distinct myeloid niche.

(A) Representative histogram for ARG1 expression within eosinophil, IM, cDC2, moDC, CD64⁺ monocyte, and CD64⁻ monocyte in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi. Data are representative of >3 independent experiments.

(B) Schematic for mixed bone marrow chimeras. Bone marrows were harvested, mixed as indicated and transplanted into irradiated recipients. After 8 weeks, mice were infected.

(C) Representative flow plots for ARG1 versus CD45.2 expression within CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice with BoyJ:Stat6^+/+ and BoyJ:Stat6^−/− chimeras at 35 dpi. Data are representative of 2 independent experiments (n = 5 mice for each group in each experiment).

(D) Quantification for contribution of CD45.1⁺ versus CD45.2⁺ to ARG1⁺ CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice with BoyJ:Stat6^+/+ and BoyJ:Stat6^−/− chimeras at 35 dpi. Data are representative of 2 independent experiments (n = 5 mice for each group in each experiment). ****P < 0.0001 by unpaired two-sided Student’s t-test.

(E) Representative flow plots for PD-L2 versus CD45.2 expression within CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice with BoyJ:Stat6^+/+ and BoyJ:Stat6^−/− chimeras at 35 dpi. Data are representative of 2 independent experiments (n = 5 mice for each group in each experiment).

(F) Quantification for contribution of CD45.1⁺ versus CD45.2⁺ to PD-L2⁺ CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice with BoyJ:Stat6^+/+ and BoyJ:Stat6^−/− chimeras at 35 dpi. Data are representative of 2 independent experiments (n = 5 mice for each group in each experiment). **P < 0.01 by unpaired two-sided Student’s t-test.

(G) Dot plots for top differentially expressed genes by different Mac/Mon subsets acquired by spatial transcriptomics.

(H) UMAP representation of Mac/Mon subsets from all replicates at each time point. 4 to 5 mice per time point. 17 sections from 17 mice in total.

(I) Representative spatial image for Mac/Mon subsets within granuloma area from D56#3 sample. Scale bar, left, 1000 μm; right, 200 μm.

(J) Overview of cellular neighborhood clusters showing the mean frequency of each cell subset (left) and mean expression of selected genes (right) across each neighborhood cluster. Heatmap values are z-scored by column and hierarchically clustered.

(K) Frequency of cellular neighborhood clusters 3 (left) and 8 (right) across time courses. *p < 0.05 by One-way ANOVA.

ARG1 expression marks a spatially distinct myeloid niche

To define the spatial distribution of ARG1⁺ cells, we first subclustered myeloid cells in the spatial transcriptomic dataset by gene-expression profiles. An unbiased analysis identified Arg1 as a defining marker of the Mac/Mon3 subset (Fig. 4, G and H). In organized granulomas, Mac/Mon 3 formed a ring encasing the Mac/Mon 1 and 2 subsets, whereas Mac/Mon 4, Chil3^high, consistent with an alveolar-macrophage-like identity, contributed minimally to the granuloma core (Fig. 4I; Fig. S4 D). Cellular neighborhood analysis showed that macrophages segregated into a coherent Mac/Mon 1, 2, and 3 cluster within granulomas, while AM-like M4 clustered with excluded epithelial regions (Fig. 4J). Neighborhood clusters 3 and 8 were enriched for granuloma-associated Mac/Mon. Neighborhood 3 comprised mostly Mac/Mon 1 and 2 with a minor Mac/Mon 3 component (Fig. 4J) that appeared by day 35 and persisted in the center (Fig. 4K; Fig. S5 B). Neighborhood 8 was dominated by Mac/Mon 3 (Fig. 4J) and, from day 42, became spatially segregated from other neighborhoods, consolidating into a discrete zone. In some organized granulomas (D42#2 and D56#3), Neighborhood 8 formed a distinct zone between the myeloid core and surrounding neighborhoods (Fig. S5B), matching the Arg1+ ring validated by immunofluorescent staining at 56 dpi (Fig. 5, A and B).

Fig. 5. Type 2 signaling drives spatial organization of myeloid cells.

(A) Representative image for spatial ARG1 expression witnin granulomatous area on lung section (D56#3). Scale bar, 500 μm.

(B) Representative immunofluorescent images for spatial ARG1 expression within granulomatous area on lung section from KN99α gcs1Δ-infected mice at 56 dpi. Scale bar, top, 150 μm; bottom, 20 μm. Data are presentative of 2 independent experiments (n = 2 mice for each experiment).

(C) Schematic of mixed bone marrow chimera for Stat6 reporter imaging. Mice were infected 8 weeks post irradiation and transplantation.

(D) Representative immunofluorescent images for granulomatous areas on lung sections from KN99α gcs1Δ-infected mice with WT:Stat6^+/+xCD11c^Yfp and WT:Stat6^−/−xCD11c^Yfp chimeras at 35 dpi (n = 3 mice per group). Scale bar, 100 μm.

To test whether this spatial pattern depends on type 2/STAT6 signaling, we generated CD11c^YfpStat6^−/− mice and then made 30%:70% mixed bone marrow chimeras (CD11c^Yfp:WT vs CD11c^YfpStat6^−/−:WT) for imaging after infection (Fig. 5C). This system allowed us to see whether STAT6 intrinsically controlled myeloid cell positioning. CD11c^YfpStat6^−/− cells were more localized within the granuloma core (Fig. 5D), suggesting that STAT6 activation contributes to ring-like structure formation, consistent with zebrafish Mycobacterial granulomas (Cronan et al., 2021).

In summary, during latent infection, type 2 signaling drives STAT6-dependent induction of ARG1⁺ myeloid cells and may promote their zonation in the myeloid center of cryptococcal granulomas.

Type 2 cytokines signal to monocyte-derived myeloid cells to antagonize fungal clearance.

Next, we sought to investigate the ontogeny of the ARG1⁺ myeloid cells and whether these cells antagonize fungal clearance. Tissue myeloid cells can derive from embryonically seeded precursors that self-renew and persist through adulthood, or from newly recruited bone marrow-derived monocytes that differentiate in the tissue (Ginhoux and Guilliams, 2016, Haldar and Murphy, 2014). To assess myeloid cell ontogeny during latent Cryptococcus infection, we utilized Ms4a3^Cre x Rosa26^LSL-tdTomato fate mapper mice to see whether type 2 cytokine-responsive myeloid cells originate from adult bone marrow granulocyte monocyte progenitors (GMPs) (Liu et al., 2019). During infection, the majority of ARG1 expressing cells were labeled by tdTomato (Fig. 6, A and B). We further gated on ARG1⁺PD-L2⁺ cells from the entire resident immune cell population and compared the tdTomato expression with the remaining cells (Fig. 6C). We observed that almost all PD-L2⁺ARG1⁺ were labeled by tdTomato (Fig. 6, D and E), arguing for an adult GMP monocyte-derived origin of these cells. To rule out contributions from classical dendritic cells (cDCs), since many of the CD11b⁺CD64⁺ cells in the lung are also CD11c⁺MHCII⁺, we infected Zbtb46^Gfp reporter mice to assess whether cDCs were contaminating our myeloid gates, especially the moDC gate (Satpathy et al., 2012). Only ~10% of CD64+ moDCs showed GFP expression, while as a positive control almost all classical DCs (cDC1 and cDC2) were GFP⁺ (Fig. 6, F and G). In summary, we identified that type 2 responsive myeloid cells during latent infection were mostly derived from GMPs with minimal contribution from cDCs.

Fig. 6. Type 2 cytokines signal to monocyte-derived myeloid cells to antagonize fungal clearance.

(A) Representative histogram for Ms4a3-tdTomato expression within eosinophil, IM, cDC2, moDC, CD64⁺ monocyte, and CD64⁻ monocyte in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi. Data are representative of two independent experiments with n > 3 mice for each experiment.

(B) Representative flow cytometry plot showing ARG1 versus Ms4a3-tdTomato expression within CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi.

(C) Representative flow cytometry plot showing ARG1 versus PD-L2 expression within CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi.

(D) Representative histogram of Ms4a3-tdTomato expression within (Gate 1) PD-L2⁺ARG1⁺ myeloid cells and (Gate 2) other hematopoietic cells in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi.

(E) Frequency of Ms4a3-tdTomato⁺ and Ms4a3-tdTomato⁻ within PD-L2⁺ARG1⁺ myeloid cells and other hematopoietic cells in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi (n = 5 mice).

(F) Representative histogram for Zbtb46-GFP expression within cDC1, cDC2, moDC, IM and CD64⁺ monocyte in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi.

(G) Frequency of Zbtb46-GFP⁺ within cDC1, cDC2, moDC, IM and CD64⁺ monocyte in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi (n = 4 mice).

(H) Schematic of mixed bone marrow chimera for Stat6 specific depletion in CCR2⁺ cells.

(I) Schematic for temporal depletion of Stat6 in CCR2⁺ cells using mixed chimera strategy. Mice were irradiated and transplanted with different bone marrow combinations as indicated. DT (100 ng/mouse) was injected intraperitoneally from 21 dpi for 6 times at indicated timepoints.

(J) CFU analysis of lung tissues from KN99α gcs1Δ-infected mice with WT, Ccr2^DTR, Stat6^−/−, WT: Stat6^−/−, Ccr2^DTR:Stat6^+/−, Ccr2^DTR:Stat6^−/− chimeras at 35 dpi (n = 5 mice per group). Data are representative of two independent experiments. *P < 0.05, **P < 0.01 and, ns = no significance by One-way ANOVA.

We then asked whether STAT6 expression in monocyte-derived cells antagonized fungal clearance, presumably downstream of type 2 cytokine signals from T_H2 cells. We mixed Ccr2^DTR bone marrow with Stat6^−/− bone marrow 1:1 and transplanted into irradiated recipients to specifically ablate STAT6 from the monocyte-derived compartment (Fig. 6H). To control for losing 50% of monocytes, and for any dominant effect from 50% Stat6^−/− bone marrow, we set up Ccr2^DTR:Stat6^+/− and WT:Stat6^−/− control mixes. We started diphtheria toxin (DT) injection from 21 dpi to deplete type 2 signaling in CCR2⁺ cells after the initial T_H2 peak (Fig. 6I). From this experiment, we found that Stat6-deletion in CCR2⁺ cells almost fully phenocopied Stat6^−/− full chimeras in their reduction in lung fungal burden (Fig. 6J), suggesting that type 2 cytokines dominantly act through monocyte-derived myeloid cells to antagonize protective immunity to Cryptoccocus.

Type 2 immunity antagonizes extracellular killing of cryptococcus.

In bacterial infection, the canonical model for type 1 immunity is that IFNγ signaling to macrophages increases their cell-autonomous killing capacity, thereby restricting intracellular pathogen replication (Shtrichman and Samuel, 2001). For intracellular pathogens such as Toxoplasma, Leishmania, and Salmonella, published data argue that AAMs may function as an intracellular replication niche due to their dampened expression of antimicrobial effectors (Marshall et al., 2011, Rodríguez-Sosa et al., 2006, Saliba et al., 2016). This argument is consistent with in vitro studies showing that STAT6 can repress STAT1 target genes (Ohmori and Hamilton, 2000). Therefore, it is assumed that during fungal infection, type 2 signaling may also antagonize macrophage intracellular fungicidal capacity. To formally test whether this is the case during latent fungal infection, we again used a mixed chimera strategy to compare the intracellular fungal burden of WT and Stat6-deficient immune cells in the same microenvironment. We infected the Stat6^+/+:BoyJ and Stat6^−/−:BoyJ chimeras with Δgcs1-mCherry reporter strain and gated out cryptococcus anti-capsule (glucuronoxylomannan, GXM) antibody positive yeast, which we presume to be extracellularly bound to host cells but not phagocytosed, as intracellular Cryptococcus should be shielded from the anti-GXM antibody. By gating on mCherry⁺GXM⁻ cells, we identified cells containing intracellular Cryptococcus (Fig. 7A). Within the immune compartment, the majority of these events were CD11b⁺CD64⁺ myeloid cells (Fig. 7B), which overlaps with the population we identified earlier as being type 2 cytokine responsive. If STAT6-deficient effector cells were more capable of killing intracellular Cryptococcus, the representation of CD45.2⁺ in Cryptococcus-containing effector cells should significantly decrease in the BoyJ:Stat6^−/− chimeras but not in BoyJ:Stat6^+/+. However, when we gated on total effector cells as well as Cryptococcus-containing effector cells and compared the CD45.2 representation within the two gates, there was no difference (Fig. 7, C and D), suggesting identical intracellular killing capacity, or lack thereof, between WT and STAT6-deficient myeloid cells.

Fig. 7. Type 2 immunity antagonizes extracellular killing of cryptococcus.

(A) Representative flow cytometry plot of mCherry versus GXM within singlets in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi.

(B) Distribution of intracellular Cryptococcus (mCherry⁺GXM⁻) to different cell types in lung tissues from KN99α gcs1Δ-infected mice with BoyJ:Stat6^+/− and BoyJ:Stat6^−/− chimeras at 35 dpi (n = 5 mice per group). Data are presentative of 2 independent experiments.

(C) Contribution of CD45.1⁺ and CD45.2⁺ to intracellular Cryptococcus-containing (mCherry⁺GXM⁻) CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice with BoyJ:Stat6^+/− and BoyJ:Stat6^−/− chimeras at 35 dpi (n = 5 mice per group). Data are presentative of 2 independent experiments. ns = no significance by unpaired two-sided Student’s t-test

(D) Frequency of intracellular Cryptococcus-containing (mCherry⁺GXM⁻) within CD45.1⁺ versus CD45.2⁺ CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice with BoyJ:Stat6^+/− and BoyJ:Stat6^−/− chimeras at 35 dpi (n = 5 mice per group). Data are presentative of 2 independent experiments. ns = no significance by paired two-sided Student’s t-test.

(E) Frequency of iNOS+ within CD45.1⁺ versus CD45.2⁺ CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice with BoyJ:Stat6^+/− and BoyJ:Stat6^−/− chimeras at 21 dpi(n = 5 mice per group). Data are presentative of 2 independent experiments. *P < 0.05 and ns = no significance by paired two-sided Student’s t-test.

(F) Frequency of iNOS+ within CD45.1⁺ versus CD45.2⁺ CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice with BoyJ:Stat6^+/− and BoyJ:Stat6^−/− chimeras at 35 dpi (n = 5 mice per group). Data are presentative of 2 independent experiments. ns = no significance by paired two-sided Student’s t-test

(G) CFU analysis of lung tissues from KN99α gcs1Δ-infected mice with BoyJ:Stat6^+/− and BoyJ:Stat6^−/− chimeras at 21 and 35 dpi (n = 5 mice per group). ns = no significance by unpaired two-sided Student’s t-test

(H) Representative flow cytometry plots of CD45 versus GXM expression gated from total mCherry⁺ singlets in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi and uninfected mice.

(I) Frequency of CD45⁺ and CD45⁻ within total mCherry⁺ singlets in lung tissues from KN99α gcs1Δ-infected mice across time courses (n = 5 mice per group).

(J) Frequency of intracellular Cryptococcus-containing (mCherry⁺GXM⁻) within CD45.1⁺ versus CD45.2⁺ CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α wildtype-infected mice with BoyJ:Stat6^+/− and BoyJ:Stat6^−/− chimeras at 10 dpi (n = 5 mice per group). Data are presentative of 2 independent experiments. ns = no significance by paired two-sided Student’s t-test.

(K) CD45.2⁺ frequency of total IM and Cryptococcus containing-IM (mCherry⁺GXM⁻) in lung tissues from KN99α WT-infected mice with BoyJ:Stat6^+/− and BoyJ:Stat6^−/− chimeras at 10 dpi (n = 5 mice per group). Data are presentative of 2 independent experiments. ns = no significance by paired two-sided Student’s t-test.

(L) Representative flow cytometry plots of GXM expression gated from total mCherry⁺ singlets in lung tissues from KN99α wildtype-infected mice at 35 dpi and uninfected mice.

(M) H&E staining of human lymph node sample with cryptococcal granulomas. Scale bar, 2 mm.

(N) Two high-power fields from H&E staining of human lymph node sample with cryptococcal granulomas. Arrows, extracellular Cryptococcus. Scale bar, 1 μm.

As STAT6 is reported to be able to repress induction of STAT1 target genes, we also analyzed whether iNOS production was affected by type 2 signaling. At both 21 dpi and 35 dpi in the mixed chimeras, we observed an unremarkable, albeit statistically significant at 21 dpi, increase in iNOS expression in Stat6^−/− myeloid cells (Fig. 7, E and F). At 35 dpi, overall iNOS production by effector cells in BoyJ:Stat6^−/− group was significantly lower than that in BoyJ:Stat6^+/+ group, likely due to the lower fungal burden caused by 30% Stat6^−/− bone marrow mix (Fig. 7G). These results suggest that in vivo, AAMs are not a replication niche for Cryptococcus, and the major mechanism whereby type 2 cytokines impair myeloid effector function is not repressing interferon-responsiveness.

Given that AAMs do not appear to be a Cryptococcus replication niche, we next asked whether the majority of the yeast were intracellular or extracellular. When assessing GXM staining on mCherry⁺ events (Cryptococcus), we noticed that most Cryptococcus yeasts during latent infection were not intracellular. Most yeasts were labeled by anti-GXM antibody and not associated with CD45⁺ cells (Fig. 7, H and I). As a second complementary approach, we utilized live imaging on vibratome thick-section lung slices to observe fungal-immune interactions in real time. Throughout the 90-minute imaging window, we did not observe any active phagocytic uptake of mCherry+ Cryptococcus by CD11b⁺, CD11c⁺, or CX3CR1⁺ myeloid cells (Video S1). Yeasts were predominantly extracellular; rare stationary overlap of CD11b and mCherry signals likely reflects previously internalized organisms rather than ongoing engulfment (Video S2), consistent with the reported antiphagocytic role for the GXM capsule (Kozel and Gotschlich, 1982). To make sure that the observations on the extracellular nature of Cryptococcus were not selective to the gcs1Δ mutant, we did the same STAT6-mixed chimera experiment infected with a WT KN99α-mCherry strain, where we observed an identical phenotype (Fig. 7, J–L). In cryptococcal granulomas from a human lymph node biopsy, most of Cryptococcus observed under high-power field reside extracellularly (Fig. 7, M and N), further supporting an extracellular dominancy.

Type 2 immunity antagonizes IFNγ-mediated protective immunity

Ifng^−/− mice succumbed to gcs1Δ infection by day 60, whereas ~80% of WT mice survived (Fig. 8A), confirming an essential role for IFNγ in host protection. Therefore, we tested whether type 2 signaling limits killing by repressing protective IFNγ responsiveness in monocyte-derived macrophages (Fig. 8B). First, we found that the majority of iNOS⁺ macrophages derived from GMPs as evidenced by Ms4a3 fate mapping (Liu et al., 2019) (Fig. 8, C and D), similar to ARG1⁺ cells. Additionally, we found that monocyte-derived cells were protective during gcs1Δ latent infection as Ccr2^−/− mice uniformly succumbed to infection with similar kinetics as Ifng^−/− mice (Figure 8E), suggesting an overall protective role for this cell type. Next, we infected mixed bone marrow chimeras (Ccr2^DTR:WT, Ccr2^DTR:Stat6^−/−, Ccr2^DTR:Ifngr1^−/−, Ccr2^DTR:Ifngr1^−/−xStat6^−/−) with gcs1Δ and found that double deficiency of Ifngr1 and Stat6 in monocyte-derived cells partially rescued the decreased CFUs observed in Ccr2^DTR:Stat6^−/− chimeras (Fig. 8F), indicating that type 2 signaling impairs monocyte-derived macrophage IFNγ responses that contribute to fungal control. However, the CFUs observed in the Ifngr1/Stat6 dKO animals were still lower than those observed in WT controls, suggesting additional IFNγ-independent functions for type 2 cytokines in promoting fungal persistence. Conditioned supernatants from IFNγ-stimulated bone marrow-derived macrophages (BMDMs) killed Cryptococcus under mammalian culture conditions (Fig. 8G), consistent with previous findings and supporting the existence of IFNγ-induced soluble fungicidal factors (Flesch et al., 1989).

Fig. 8. Type 2 immunity antagonizes IFNγ-mediated killing.

(A) Survival curve for WT, Il4/Il13^−/−, and Ifng^−/− mice infected with KN99α gcs1Δ until 140 dpi. Data are pooled from two independent experiments.

(B) Representative histogram for iNOS expression within eosinophil, IM, cDC2, moDC, CD64⁺ monocyte, and CD64⁻ monocyte in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi. Data are representative of >3 independent experiments.

(C) Representative flow cytometry plot showing iNOS versus Ms4a3-tdTomato expression within CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi (n = 5 mice per group). Data are representative of 2 independent experiments.

(D) Frequency of Ms4a3-tdTomato⁺ within iNOS⁺ and iNOS⁻ CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α gcs1Δ-infected mice at 35 dpi.

(E) Survival curve for WT and Ccr2^−/− mice infected with KN99α gcs1Δ until 78 dpi (n = 5 mice per group).

(F) CFU analysis of lung tissues from KN99α gcs1Δ-infected mice with Ccr2^DTR:WT, Ccr2^DTR:Stat6^−/−, Ccr2^DTR:Ifngr1^−/−, and Ccr2^DTR:Stat6^−/−xIfngr1^−/− chimeras at 35 dpi (n = 5 mice per group). DT was injected peritoneally following the same schedule as Fig. 6I.

(G) Quantification of OD value changes in Cryptococcus cultured with indicated conditions. SUP, supernatant from BMDM culture. Data are representative of 2 independent experiments.

Data are present by mean ± SD. *P < 0.05 by One-way ANOVA in (f).

Discussion

In this study, we demonstrate that type 2 inflammation actively antagonizes protective immunity to Cryptoccocus during non-lethal granulomatous infection. We show that the type 2 response is mainly established by T_H2 cells via suppression of monocyte-derived myeloid cells in a manner that impairs extracellular fungal clearance. These observations are consistent with studies that have found detrimental roles for type 2 responses during acute lethal murine Cryptococcus infection; however, the acute models do not appropriately mimic human infection (Müller et al., 2013, Dang et al., 2022). While type 2 signaling deletion results in a lower pulmonary fungal burden during acute infection, it does not significantly protect animals from mortality. Meanwhile, Ifng^−/− and Rag1^−/− mice have the same survival rate compared to WT mice, suggesting the induction of a completely maladaptive T cell response during lethal infection, which is a very different infectious course compared to humans. Therefore, it is not a priori predicted that type 2 inflammation would play a role in scenarios where hosts do not succumb to latent, subacute infection. Consistent with recent studies using clinical Cryptococcus isolates, we find that the type 2 response is induced early but declines to a low level in this latent infection model. Our data argue that a long-lasting type 2 response can antagonize fungal clearance without causing fungal overgrowth that leads to host fatality. These findings raise the possibility of utilizing IL-4Rα blocking antibodies to pre-clear hosts of latent fungi before going on immunosuppressants for transplant surgery or chemotherapy.

In parallel with immune circuitry, our histology, thick-tissue imaging, and spatial transcriptomics show that gcs1Δ in B6 mice drives lesions that evolve from early epithelioid aggregates into non-caseating, zoned granulomas with a myeloid core (including a neutrophil-enriched center) encased by a T/B-cell cuff; epithelial cells are largely excluded from the core. This organization is broadly consistent with the non-gelatinous cryptococcal granulomas described in clinical settings and with the architecture reported in the UgCl223-FeJ model, while allowing for strain- and host-specific differences. High-throughput spatial transcriptomics further resolves this microanatomy, delineating discrete myeloid and lymphoid zones and identifying a STAT6-linked ARG1⁺ myeloid niche within the granuloma core. These data provide a structural and molecular framework for mechanistic tests of cell positioning and cytokine circuits in cryptococcal granulomas. Although this work delineates the microanatomy, cellular composition, and key gene programs of cryptococcal granulomas in mice, clinical isolates that reproducibly establish granulomatous latency in B6 mice are still needed to strengthen generalizability. Given the scarcity of human material (granulomas often reside in asymptomatic individuals), mouse models will remain essential to uncover granuloma heterogeneity and to define how the immune system maintains or breaks latency.

Across infectious and non-infectious contexts, granulomas are increasingly recognized as structured, multicellular assemblies rather than amorphous inflammatory masses (Gideon et al., 2022, Jiang et al., 2025, Sawyer et al., 2023, Krausgruber et al., 2023, Pyle et al., 2025, Harvest et al., 2023, Pessenda et al., 2025, Fonseca et al., 2025, McCaffrey et al., 2022). Consistent with this view, our data reveal coordinated contributions from myeloid, lymphoid, and stromal populations and their spatial organization within fungal lung lesions. These observations support the notion that granuloma formation represents an active host strategy rather than a passive consequence of inflammation. In gcs1Δ-B6 granuloma model, STAT6 activation drives macrophage repositioning and promotes the emergence of ARG1⁺ myeloid niches distal to the granuloma core, indicating that type 2-associated pathways can shape tissue architecture, conceptually similar to the macrophage- and monocyte-dependent structural programs described in bacterial and hepatic leishmanial granulomas (Cronan et al., 2021, Pessenda et al., 2025, Sorobetea et al., 2023). Conversely, loss of STAT6 may relax this segregated architecture and permit greater lymphocyte access to fungus-containing regions, thereby increasing local cytokine exposure within granulomas. Although this remains to be tested directly at higher spatial resolution, our data are consistent with a model in which STAT6-dependent tissue organization limits productive contact between antifungal lymphocytes and their myeloid effector targets. The spatial organization of ARG1⁺ myeloid cells may also explain why T_H2-driven pathology can emerge despite a concurrent type 1 wave: rather than globally extinguishing interferon responses, type 2 signaling appears to act locally on monocyte-derived myeloid cells within granulomas, where it can dominate effector programming at the fungus-containing site. Thus, the balance between type 1 and type 2 immunity may be determined less by their overall abundance than by their spatial proximity to shared myeloid target cells. We further found that alternative activation of myeloid cells by type 2 cytokines within granulomas is detrimental to pathogen clearance, consistent with findings from macaque tuberculosis, where early type 2-skewed granulomas exhibited poor bacterial control (Gideon et al., 2022). Whereas that study implicated mast cells as a potential cytokine source (Gideon et al., 2022), our genetic model identifies T_H2 cells as the dominant producers driving STAT6 activation and fungal persistence. Despite these parallels, granulomas induced by different pathogens display distinct spatial organizations and immunosuppressive niches (Fonseca et al., 2025, McCaffrey et al., 2022), implicating that the architectural logic of granulomas is context-dependent and should be further dissected using tractable genetic systems.

During type 2 biased lethal serotype D Cryptococcus infection, it has been shown that LysM^Crex Il4ra^fl/fl mice phenocopy Il4ra germline knockouts in displaying infection resistance (Müller et al., 2013), arguing that macrophages are the major targets of type 2 cytokines that antagonize sterilizing immunity. However, LysM^Cre has broad activity, being active in neutrophils, tissue resident macrophages, monocytes, some neurons, and airway epithelium (Spella et al., 2019, Clausen et al., 1999, Orthgiess et al., 2016). Utilizing genetic fate mapping and genetic depletion, we demonstrate that monocyte-derived myeloid cells including IMs, moDCs, and monocytes are the major target cells of type 2 signaling that impair fungal clearance. Why AAMs, and type 2 immunity in general, are detrimental during fungal infection has been unclear. The most common argument is that since AAMs downregulate intracellular killing molecules (iNOS, etc), they may have impaired intrinsic fungicidal capacity, thus generating an intracellular replication niche. Utilizing a Cryptococcus reporter strain and extracellular capsule antibody staining, we were able to test this hypothesis in vivo. We did not observe a difference in intracellular fungicidal capacity between WT and Stat6-deficient macrophages. Instead, we find that the majority of yeasts during both lethal and latent infection reside extracellularly in vivo, indicating that Cryptococcus clearance may happen in the extracellular space.

It has been reported that Cryptococcus can be rapidly internalized by alveolar macrophages (AMs) after intratracheal infection and extracellular predominance occurs by 24 hours post-infection (Feldmesser et al., 2000). In our study, we found that AMs did not significantly expand during latent infection (Figure S4 B) and in our spatial transcriptomic analysis, we observed that cells in the Mac/Mon 4 cluster were mostly excluded by granuloma. These cells highly expressed Chil3, indicating an AM identity. Generally, we did not observe a significant contribution of the AM compartment to granulomas during latent infection and most immunosuppressive myeloid cells derived from monocytes. It has been shown that AMs have a higher phagocytic capacity than IMs (Wizemann and Laskin, 1994), which may explain the extracellular fungal dominance in granulomas. Our study opens new avenues in studying host clearance of Cryptococcus infection and raises the question of how IFNγ might provide protection against an extracellular pathogen. Our spatial transcriptomics data provide gene expression patterns at single-cell resolution within cryptococcal granulomas and reveal highly heterogenous myeloid cells in the granuloma center, some of which appear highly interferon-responsive based on expression of Il1rn, Nos2 and Acod1. These genes are candidate effectors of particular interest. Nos2/Acod1 expression is consistent with an activated antimicrobial program that could restrict fungal proliferation in the extracellular environment, whereas Il1rn may mark a counter-regulatory state that restrains IL-1-driven inflammation, highlighting that granuloma-core myeloid cells likely integrate both antifungal and immunoregulatory modules rather than adopting a uniform activation state. Deeper mechanistic dissection of these discrete Mac/Mon subsets may provide more insights into how myeloid cells directly or indirectly mediate extracellular fungicidal effects downstream of IFNγ.

The existence of IFNγ-induced soluble fungicidal factors was initially identified in 1989, consistent with our in vivo data and in vitro observations using BMDMs (Flesch et al., 1989). However, Ifngr1/Stat6 double deficiency in CCR2⁺ cells only partially rescued the reduced fungal burden in Stat6 single knockout, suggesting the detrimental impact of type 2 immunity is only partly explained by antagonizing IFNγ responsiveness in macrophages. We find that mice with Ifngr1/Stat6 double knockout in CCR2⁺ cells have a lower fungal burden than control chimeras, suggesting the existence of additional IFNγ-independent antifungal effectors that operate in the extracellular space. Our immune profiling, together with spatial maps that resolve cell types, zones, and candidate chemokine niches, provides a practical platform to dissect these mechanisms in vivo. More broadly, STAT6 may antagonize antifungal control by suppressing fungicidal factors or altering metabolic reprogramming required for fungal clearance. Defining how STAT6 reshapes these transcriptional and metabolic programs in granuloma-associated myeloid cells will be important goal for future work.

Taken together, our results identify the cell types that produce and respond to type 2 cytokines during latent granulomatous cryptococcal infection. We propose that in contrast to its previously described role in promoting intracellular fungal growth, type 2 cytokines impair the extracellular killing capacity of monocyte-derived cells.

Materials & Methods

Human samples

Lymph node tissue used in this study was obtained under an NIH Institutional Review Board (IRB)-approved protocol (14-I-0124, PET Imaging and Lymph Node Assessment of IRIS in People With AIDS (PANDORA); ClinicalTrials.gov Identifier: NCT02147405). The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. Written informed consent was obtained from the participant prior to all procedures.

The sample analyzed was an inguinal lymph node biopsy from a 26-year-old White Hispanic male with HIV infection. At the time of biopsy, the patient was not receiving antiretroviral therapy. He presented with a CD4+ T cell count of 15 cells/mm³ (2%) and a plasma HIV viral load of 863,557 copies/ml. The patient had disseminated cryptococcal disease, including cryptococcal meningitis and cryptococcemia, previously confirmed by positive blood cultures. At the time of biopsy, he had been treated with amphotericin B and flucytosine and was maintained on oral fluconazole. His serum cryptococcal antigen titer was 1:8192, although blood cultures were negative at the time of lymph node sampling.

Immunohistochemistry of human tissue

This human lymph-node cryptococcal granuloma was sectioned by pathologists into five serial sections at 5 μm thickness and mounted on Superfrost Plus slides (Fisher Scientific). Slides were baked at 70°C overnight, followed by deparaffinization and rehydration with washes in xylene (3X), 100% ethanol (2X), 95% ethanol (2X), 80% ethanol (1X), 70% ethanol (1X) and ddH₂O with a Leica ST4020 Linear Stainer (Leica Biosystems). Tissues next underwent antigen retrieval by submerging sides in Target Retrieval Solution (pH 9, DAKO Agilent) and incubating at 97°C for 40 min in an IHC-Tek Steamer (IHC World). After cooling to room temperature, slides were washed in 1X PBS IHC Washer Buffer with Tween 20 (Cell Marque) with 0.1% (w/v) bovine serum albumin (Thermo Fisher). The slide utilized for histology was stained with hematoxylin and eosin. The remaining tissues were mounted in Sequenza staining cartridges (Epredia) underwent two rounds of blocking, the first to block endogenous peroxidase activity with 3% H₂O₂ (Sigma-Aldrich) in ddH2O for 30 minutes at room temperature. Tissues were then washed with wash buffer and blocked for 1 h at room temperature with 1× TBS IHC Wash Buffer with Tween 20 with 3% (v/v) normal donkey serum (Sigma-Aldrich), 0.1% (v/v) cold fish skin gelatin (Sigma-Aldrich), 0.1% (v/v) Triton X-100, and 0.05% (v/v) Sodium Azide. Antibodies were diluted in 1x TBS IHC Wash Buffer with Tween 20 with 3% (v/v) normal donkey serum (Sigma-Aldrich) prior to incubation with tissue overnight at 4°C. The following antibodies were used: 1) CD3e clone D7A6E, CST, 1:50 dilution 2) CD4 clone EPR685, Abcam, 1:100 dilution 3) CD14 clone D7A2T, CST, 1:100 dilution 4) CD68 clone D4B9C, CST, 1:200 dilution. Following the overnight incubation, slides were washed twice for 5 min in wash buffer. Chromogenic signal was developed using the ImmPRESS universal (Anti-Mouse/Anti-Rabbit) kit (Vector Laboratories). Slides were then counterstained with hematoxylin. Lastly, slides were dehydrated by washing in 70% ethanol (1×), 80% ethanol (1×), 95% ethanol (2×) and 100% ethanol (2×) and coverslipped. Images from slide scans were processed and exported using QuPath.

Mice

Male and female mus musculus were used. In bone marrow chimera experiments, animals were irradiated at 6- to 8-week-old and analyzed 6–8 weeks following irradiation. Most other experiments were performed on adult animals between 8 and 20 weeks of age. All mice were housed in a specific pathogen–free environment and all mouse experiments were approved by the National Institutes of Allergic and Infectious Diseases Animal Care and Use Committee (NIAID-ACUC) and were performed in accordance with NIAID-ACUC guidelines and under approved protocols (LHIM-4E).

C57BL/6 (JAX: 000664), C57BL/6J-CD45a (JAX: 002014), EPO-DTA (JAX: 036975), Mcpt8-Cre (JAX: 017578), ROSA-LSL-DTA (JAX: 009669), Ms4a3-Cre (JAX: 036382), ROSA-LSL-tdTomato (JAX: 007914), Stat6 knockout (JAX: 005977), and Il13-KN4 (JAX: 031367) mice were purchased from Jackson Laboratory. Il4/Il13 knockout (line 242), Il17a knockout (line 8434), Ifng knockout (line 248), Rag1 knockout (line 146), Il4 knockout (line 46), and R26-CreERT2 x Gata3 flox (line 8445) mice were ordered from NIAID-Taconic Exchange program. Il13 knockout bone marrows were gifted from William Petri’s lab at University of Virginia. Gata3 flox, Klrg1-Cre x Gata3 flox, and hCD2-Cre x Gata3 flox mice were from Jinfang Zhu’s lab at National Institutes of Allergic and Infectious Diseases. CCR2-DTR mice were from Yasmine Belkaid’s lab at National Institutes of Allergic and Infectious Diseases. Il4-KN2 mice were gifted from Richard Locksley’s lab at University of California San Francisco.

Cryptococcus intranasal infection

Cryptococcus was obtained from frozen glycerol stocks and then cultured on a yeast extract peptone-dextrose (YPD) plate for two days, followed by overnight culture at 30 °C in YPD broth with shaking. Mice were anesthetized by 3% isoflurane and infected intranasally with 5×10⁴ CFU (KN99α wildtype or gcs1Δ) or 100 CFU (UgCl223) (Ding et al., 2022) yeast in 25uL 1X PBS. Mice were monitored until recovery from anesthesia and were placed back to cages.

Histological analysis for mouse lung tissues

Lung samples were collected and fixed in 4 % PFA. Sectioning and H&E staining were performed and evaluated by veterinary pathologists at Infectious Disease Pathogenesis Section at NIAID. For immunofluorescent staining, euthanized and perfused mice were inflated with OCT via the trachea, and lung were collected and fast frozen. Frozen sections were made using a CryoStat into 8 μm thickness and air dried for 1 hour. Sections were then fixed with 4% PFA again and blocked with 5% BSA followed by hydrogen peroxide treatment. Fluorophore-conjugated antibodies were incubated at 4 °C overnight. Sections were mounted on slides for the imaging by Leica SP8 conjugated with 690 nm laser at the Bioimaging Research Technologies Branch at NIAID.

Thick-tissue imaging

Mice were euthanized and perfused with 20 ml 1X PBS followed by 20 ml 4% PFA. Inflation with 2 % low-melting agarose was then performed for thick tissue sectioning using a Vibrating Blade Microtome. 500 μm sections were made and processed with Ce3D Tissue Clearing buffer set according to commercial instructions. Briefly, sections were permeabilized at room temperature for 2 days with gentle shaking followed by antibody staining for another 2 days with gentle shaking. Sections were washed 3 times by washing buffer during next 24 hours and then cleared by clearing solution for overnight. Cleared sections were mounted with clearing solution in 4 spacers (120 μm thickness for each) on slides for imaging by Leica SP8 conjugated with 690 nm laser at Bioimaging Research Technologies Branch at NIAID.

Colony forming unit assay

Lung samples were collected and homogenized in sterilized water followed by serial dilution. KN99α Δgcs1 contains a nourseothricin-resistance cassette. Therefore, all samples for CFU were plated on nourseothricin-containing YPD plates for fungal burden detection.

CD4 antibody treatment

Anti-mouse CD4 antibody (GK1.5, BioXCell) or its isotype, rat anti-KLH IgG2b, were injected intraperitoneally with an initial 2 doses of 100 μg per mouse and rest being 4 doses of 200 μg per mouse.

Flow cytometry

All mice for flow cytometry analysis in this study received retroorbital injection of 3 μg CD45 antibody 3 minutes before euthanasia to label intravascular cells. For lung samples, tissues were collected directly into 5 ml digestion buffer (HBSS with 1 mg/ml collagenase II, 1 mg/ml dispase, and 10 μg/ml DNAse I) and then digested using a GentleMACS. Digestion was stopped by adding 5 ml flow buffer (PBS with 2% FBS and 2 mM EDTA) on ice. Samples were immediately filtered using 70um strainers and centrifuged to remove digestion buffer. Red blood cells were removed by ACK lysis buffer for 5 min on ice and cells were then spun, washed, and aliquoted in 96-well round bottom plates for staining. Live/dead staining was performed for 10 min in PBS containing Fc Block before protein staining. Surface proteins were stained for 20 min on ice and then cells were acquired using a Cytek Aurora spectral flow cytometer. Intracellular and intranuclear proteins were stained with Transcription Factor Staining Buffer set. Briefly, cells were fixed and permeabilized on ice for 1 hour and stained in permeabilization buffer containing antibodies for 1 hour at room temperature. Cells were washed and resuspended in flow buffer for acquisition. T cell restimulation was performed for cytokine analysis. Collected cells were cultured in RPMI 1640 culture medium with PMA (50 ng/ml) and ionomycin (500 ng/ml) for 4 hours and then processed for staining. For GXM staining, cells were stained with other antibodies together with anti-GXM antibody (isotype: mouse IgG1) for surface staining for 20 min on ice. Then, after three times washing, cells were stained with FITC anti-mouse IgG1. Data were analyzed using FlowJo.

Survival rate

Survival experiment was conducted by the coordination with animal facility 14DNR at NIAID and was monitored by veterinarians at Comparative Medicine Branch, NIAID. Endpoint was set up when the loss of body weight reached 20% compared to starting point. Animals were euthanized at the endpoint.

Bone marrow chimera

Mice were irradiated at 700 rads in two doses spaced 3hrs apart by 14DNR at NIAID. Bone marrow was prepared and injected the same day at least one hour after the final irradiation dose. Mice were monitored during 6-week reconstitution. Bleeding was performed to determine reconstitution efficiency.

GATA3 temporal depletion

Tamoxifen was administered by oral gavage with the dose of 75 mg/kg.

Spatial transcriptomics

Euthanized mice were perfused with 20 ml PBS followed by 20 ml 4% PFA and then inflated with OCT from trachea. Lung tissues were collected and immediately fast frozen for storage at −80 °C. The granulomatous areas were not selected with any preference. Sections were made with CrypStat by 10 μm directly on slides provided by 10X Genomics. Xenium Prime 5K kit was utilized for spatial transcriptomics according to commercial instructions. Briefly, sections were hybridized with priming oligos and then washed by RNase to release the RNA strand, followed by polishing to allow probe hybridization for the target RNA. We did not add customized probes on pre-designed mouse pan tissue & pathways panel containing 5006 probes. Unbound probes were washed away, and ligation was performed to ensure probe specificity by generating a circular DNA from hybridized probe, which was enzymatically amplified subsequently. Cell segmentation staining and autofluorescence quenching were performed before loading slides into Xenium Analyzer at NIAMS.

The output from the Xenium analyzer was processed with Seurat version 5.1.0.9006(Hao et al., 2024). First, the Xenium data folder from each image was loaded onto Seurat with the ‘LoadXenium’ function and the segmentations parameter set to ‘cell’. Cells without feature counts (nCount_Xenium > 0) were subsequently filtered from the dataset. Each image contained a single section except for one, whose Seurat object was then split into two separate objects based on tissue coordinates using the ‘Crop’ and ‘subset’ functions. Next, all images were combined into a single Seurat object using ‘merge’, and low-quality cells containing less than 3 Xenium features and 6 counts (nFeature_Xenium < 3 & nCount_Xenium < 6). Finally, data was normalized using the SCTransform function (Choudhary and Satija, 2022).

For analysis, the dimensionality of the dataset was determined by running the ‘RunPCA’ and ‘ElbowPlot’ functions, and 25 dimensions were used in subsequent steps. Dimensionality reduction and clustering was performed with ‘RunUMAP’, ‘FindNeighbors’, and ‘FindClusters’, with a range of resolutions. A clustering based on a resolution of 0.3 was chosen to capture the biological diversity in the dataset, resulting in 19 clusters. To perform cluster annotation, the differentially expressed genes per cluster were calculated using the ‘PrepSCTFindMarker’ and ‘FindAllMarkers’ (min.pct = 0.01) functions. Based on these genes and on visualization of cell type specific markers these 19 clusters were annotated into 7 major cell types, which were subsequently used for visualization on the Xenium Explorer (10X Genomics). Finally, these major cell types and the genes used in cluster annotation were visualized using Seurat.

For further analysis of macrophage/monocyte populations from the Xenium dataset, cells annotated as macrophage/monocyte were subset into a new Seurat object. Data processing an analysis was performed as described above for the entire dataset (starting at the normalization step with SCTransform) with the following parameters: 30 dimensions and resolution of 0.2, resulting in 22 macrophage/monocyte clusters. Subsequently these 22 clusters were grouped into four macrophage/monocyte subsets based on shared gene expression patterns and temporal occurrence.

Cellular neighborhoods were computed by utilizing a k-means clustering strategy previously developed to quantify spatial interactions in multiplexed proteomic datasets (Ferrian et al., 2024). For each sample, a neighborhood matrix was generated by enumerating the frequency of each cell subset within a 50-pixel radius of each index cell (centroid-to-centroid distance). The resulting matrix clustered with k-means clustering into 10 neighborhood clusters. While only the neighbor matrix was used to define the clusters, expression of relevant genes was also evaluated in cells based on their number assignment.

Diphtheria toxin (DT) treatment

DT was injected intraperitoneally with the dose of 100 ng per mouse.

Live lung tissue imaging

Mice were euthanized using 5% Isoflurane at induction chamber. After euthanasia, mouse lungs were inflated with 1.5 % of Sea Kem agarose in phenol-red free RPMI at 37 °C. Inflated tissues were kept on ice, in 1 % FBS in PBS, and sliced into 300–350 μm sections using Leica VT1200 S Vibrating Blade Microtome (Leica Microsystems), in ice-cold PBS. Tissue sections were stained with fluorescently labeled antibody cocktail of choice for 2–6 h in the 37 °C incubator. After staining sections were washed 3 times and cultured in complete lymphocyte medium (Phenol Red-free RPMI supplemented with 20 % FBS, 25 mM HEPES, 50 μM β-ME, 1 % Pen/Strep/L-Glu and 1 % Sodium Pyruvate) in humidified incubator at 37°C. Tissues were allowed to completely recover for 12 h prior to recording time-lapse video. Sections were held down with tissue anchors (Warner Instruments) in 2-well imaging chambers and imaged using Leica DIVE inverted 5 channel confocal microscope equipped with an Environmental Chamber (NIH Division of Scientific Equipment and Instrumentation Services) to maintain 37 °C and 5 % CO₂. Microscope configuration was set up for four-dimensional analysis (x,y,z,t) of cell segregation and migration within tissue sections. Diode laser for 405 nm excitation; Argon laser for 488 and 514 nm excitation, DPSS laser for 561; and HeNe lasers for 594 and 633 nm excitation wavelengths were tuned to minimal power (between 0.1 and 2 %). Z stacks of images were collected (10 – 50 μm). Mosaic images of lung sections were generated by acquiring multiple Z stacks using motorized stage to cover the whole section area and assembled into a tiled image using LAS X (Leica Microsystems) software. For time-lapse analysis of cell migration, tiled Z-stacks were collected over time (1 to 4 h). Post-acquisition mages were processed using Imaris software.

Bone marrow–derived macrophages (BMDMs) differentiation.

Bone marrow cells were isolated from the femurs and tibias of adult C57BL/6 mice. Bones were flushed with sterile DMEM supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 1% penicillin–streptomycin and plated at a density of 1 × 10⁶ cells/mL in 10 mL non-tissue-culture-treated Petri dishes. Cells were cultured for 7 days at 37 °C with 5% CO₂ in complete medium supplemented with 10% M-CSF (generated from NIH-3t3-MCSF transduced fibroblast cell line) to promote macrophage differentiation. On day 4, the culture was supplemented with 3mL of 10% M-CSF–containing medium (DMEM). On day 7, adherent cells (BMDMs) were gently washed with PBS and harvested using PBS with 0.5 mM EDTA. Differentiation was confirmed by morphology and expression of CD11b assessed by flow cytometry

BMDM-conditioned medium preparation

Freshly differentiated BMDMs were counted and seeded into tissue culture–treated, flat-bottom 96-well plates (Corning) at a density of 1 × 10⁵ cells per well in 100 μL of Opti-MEM (Gibco). Cells were allowed to adhere and rest for 1 h at 37 °C, after which they were washed once with PBS and replenished with 100 μL of either plain Opti-MEM (unstimulated control) or Opti-MEM containing 100 ng/mL recombinant murine IFN-γ (PeproTech). After 48 h of stimulation, the conditioned media were collected, centrifuged at 300 × g for 5 min to remove debris, and the clarified supernatants were stored at −80 °C for subsequent experiments.

Growth curve assay using BMDM-conditioned media

C. neoformans was counted and resuspended in culture medium at a concentration of 2 × 10⁶ cells mL⁻¹. An inoculum of 3 × 10⁴ CFU was added per well in a final volume of 100 μL, containing either 15% or 70% (v/v) conditioned medium diluted in fresh culture medium. Assays were performed in Corning flat-bottom 96-well plates with lids (tissue culture–treated) to minimize evaporation and contamination. Growth kinetics were monitored for 18 h at 37 °C under continuous shaking, with optical density (OD₆₀₀) recorded every 15 min using a Synergy H1 microplate reader (BioTek Instruments, Winooski, VT, USA).

Supplementary Material

Video 1

Video S1. Live imaging of lung samples in Cryptococcus latent infection. Scale bars are indicated in the video. Data are representative of n = 2 mice.

Video 2

Video S2. Rare phagocytosis events in Cryptococcus latent infection by CD11b⁺ cells. Scale bars are indicated in the video. Data are representative of n = 2 mice.

Figure S1

Fig. S1. B6 mice can control KN99α gcs1Δ to form a latent infection contained by pulmonary granulomas.

(A) Survival curve for WT, Il4/Il13^−/−, Ifng^−/− and Rag1^−/− mice infected with KN99α WT until 40 dpi (n = 10 mice per group).

(B) Representative images for thick-tissue imaging of lungs from KN99α WT-infected CD11c^Yfp reporter mice at 10 dpi. Scale bar, 300 μm.

(C) IFNγ⁺ and IL-5/IL-13⁺ frequency of CD4⁺FOXP3⁻ T cell compartment in lung tissue from mice infected with KN99α WT across time courses (n = 5 mice per group).

(D) Representative histogram for iNOS expression within CD64⁺CD11b⁺ myeloid cells in lung tissues from KN99α WT-infected mice across time courses.

(E) Schematic of CD4 depletion experiment. Mice were injected intraperitoneally with anti-CD4 antibody or isotype control from day 35 for 6 doses at indicated time points.

(F) Representative flow cytometry plots showing CD4 versus CD44 within IV⁺ and IV⁻ CD90.2⁺TCRβ⁺ T cells in lung tissues from KN99α gcs1Δ-infected mice with anti-CD4 or isotype control administration at 49 dpi.

(G) Frequency of CD4⁺ in IV⁺ and IV⁻ CD90.2⁺TCRβ⁺ T cells in lung tissues from KN99α gcs1Δ-infected mice with anti-CD4 or isotype control administration at 49 dpi (n = 5 mice per group). *P < 0.05 and ****P < 0.0001 by unpaired two-sided Student’s t-test.

(H) CFU analysis of lung tissues from KN99α gcs1Δ-infected mice with anti-CD4 or isotype control administration at 49 dpi (n = 5 mice per group). *P < 0.05 by unpaired two-sided Student’s t-test.

Figure S2

Fig. S2. Spatial transcriptomics reveal the cellular architecture of cryptococcal granuloma in mouse models.

(A) Representative images for segmentation staining in Xenium spatial transcriptomics. Scale bar, 1000 μm.

(B) Spatial cell segmentation maps for granulomatous area showing all cell clusters in lung tissue sections from KN99α gcs1Δ-infected mice across time courses. One sample at 35 dpi was not presented due to limited space. Scale bar, 500 μm.

Figure S3

Fig. S3. T cell gating strategy and type 2 immunity is detrimental in UgCl223-induced latent infection in B6 mice.

(A) Gating strategy for T cells.

(B) Representative flow cytometry plots showing GATA3 versus T-bet expression within CD4⁺FOXP3⁻ T cells in lung tissues from UgCl223-infected WT and Stat6^−/− mice at 35 dpi.

(C) Quantification of T_H2 cells in lung tissues from UgCl223-infected WT and Stat6^−/− mice at 35 dpi (n = 5 mice for WT group and n = 4 for Stat6^−/− group). *P < 0.05 by unpaired two-sided Student’s t-test.

(D) Quantification of eosinophils in lung tissues from UgCl223-infected WT and Stat6^−/− mice at 35 dpi (n = 5 mice for WT group and n = 4 for Stat6^−/− group). ns = no significance by unpaired two-sided Student’s t-test.

(E) CFU analysis of lung tissues from UgCl223-infected WT and Il4/Il13^−/− mice at 35 dpi (n = 5 mice per group). **P < 0.01 by unpaired two-sided Student’s t-test.

(F) CFU analysis of lung tissues from UgCl223-infected WT and Stat6^−/− mice at 35 dpi (n = 5 mice per group). *P < 0.05 by unpaired two-sided Student’s t-test.

(G) CFU analysis of lung tissues from UgCl223-infected Gata3^fl/fl and hCD2^Cre x Gata3^fl/fl mice at 35 dpi (n = 5 mice per group). ***P < 0.001 by unpaired two-sided Student’s t-test.

Figure S4

Fig. S4. Myeloid gating strategy and Mac/Mon dynamics in spatial transcriptomics.

(A) Gating strategy for myeloid cells.

(B) Quantification of AM, IM, and Mon in lung tissues from KN99α gcs1Δ-infected mice across time courses.

(C) Quantification of cDC1, cDC2, and moDC in lung tissues from KN99α gcs1Δ-infected mice across time courses.

(D) Spatial cell segmentation maps for granulomatous area showing Mac/Mon clusters in lung tissue sections from KN99α gcs1Δ-infected mice across time courses. One sample at 35 dpi was not presented due to limited space. Scale bar, 500 μm.

Figure S5

Fig. S5. ARG1 expression marks a spatially distinct myeloid niche

(A) Representative flow plots showing PD-L2 and ARG1 expression within CD64⁺CD11b⁺ myeloid cells in lung tissues from indicated mouse strains with indicated infection at 35 dpi (n = 5 per group). Numbers are present by mean ± SD.

(B) Spatial cell segmentation maps for granulomatous area showing neighborhood clusters in lung tissue sections from KN99α gcs1Δ-infected mice across time courses. One sample at 35 dpi was not presented due to limited space. Scale bar, 500 μm.

Figure S1 shows that 1, wild-type Cryptococcus cannot induce granuloma formation or efficient protective responses; 2, temporal CD4 depletion in gcs1Δ-B6 model results in an increased pulmonary fungal burden. Figure S2 shows representative images of segmentation staining used in spatial transcriptomics and granuloma structures over time. Figure S3 shows the gating strategy for T cells, and that type 2 circuit is detrimental in UgCl223-induced latent infection in B6 mice. Figure S4 shows the gating strategy and kinetics of myeloid cells during gcs1Δ-induced latent infection. Figure S5 shows PD-L2+ARG1+ alternatively activated myeloid cells in multiple mouse models of latent cryptococcal infection. Figure S5 also shows the dynamics of different cell clusters in spatial transcriptomics. Video S1 shows extracellular Cryptococcus by live imaging and Video S2 shows in this live imaging that phagocytosis events by CD11b⁺ cells are rarely observed.

Non-standard abbreviation:

AAM(s)

alternatively activated macrophage(s)

AM(s)

alveolar macrophage(s)

ANOVA

analysis of variance

ARG1

arginase 1

B6

C57BL/6

BMDM(s)

bone marrow–derived macrophage(s)

BSA

bovine serum albumin

cDC(s)

conventional dendritic cell(s)

cDC1

type 1 conventional dendritic cell

cDC2

type 2 conventional dendritic cell

Ce3D

Clearing-enhanced 3D

CFU(s)

colony-forming unit(s)

DC(s)

dendritic cell(s)

dpi

days post infection

DT

diphtheria toxin

FBS

fetal bovine serum

GMP(s)

granulocyte-monocyte progenitor(s)

GXM

glucuronoxylomannan

H&E

hematoxylin and eosin

HBSS

Hanks’ balanced salt solution

IFN

interferon

IHC

immunohistochemistry

ILC2(s)

group 2 innate lymphoid cell(s)

IM(s)

interstitial macrophage(s)

iNOS

inducible nitric oxide synthase

IRIS

immune reconstitution inflammatory syndrome

Mac/Mon

macrophages and monocytes

M-CSF

macrophage colony-stimulating factor

MHCII

major histocompatibility complex class II

moDC(s)

monocyte-derived dendritic cell(s)

OCT

optimal cutting temperature compound

PBS

phosphate-buffered saline

PFA

paraformaldehyde

PMA

phorbol myristate acetate

RPMI

Roswell Park Memorial Institute medium

scRNA-seq

single-cell RNA sequencing

SD

standard deviation

STAT6

signal transducer and activator of transcription 6

TH1

T helper 1

TH2

T helper 2

UMAP

uniform manifold approximation and projection

WT

wild type

YPD

yeast extract peptone dextrose

Data availability

All source data are available upon request. Spatial transcriptomic data has been deposited to Zenodo with the following DOI: 10.5281/zenodo.15412158. Codes used for spatial transcriptomic analyses are available through GitHub (https://github.com/ivanzhengnih/Spatial-transcriptomic-analysis_Yufan-Zheng).