Monocytic Niches Escape T cell Surveillance and Promote Mycobacterium tuberculosis Persistence in Lymph Nodes

Summary:

Lung-draining mediastinal lymph nodes (medLNs) are critical for Mycobacterium tuberculosis (Mtb) pathogenesis, serving both as sites of T cell priming and, paradoxically, reservoirs for long-term bacterial persistence. To understand this dichotomy, we examined myeloid and CD4⁺ T cell dynamics in medLNs after aerosol Mtb infection. Early bacterial dissemination occurred via monocytes and IL-12-producing conventional dendritic cells (cDCs), which initiated Th1 cell priming within the T cell zone. Over time, cDC migration and T cell activation declined, and medLNs became dominated by heavily infected monocyte-derived aggregates that persisted into late infection. Despite inducing proinflammatory and bactericidal pathways, these aggregates escaped recognition by Mtb-specific T cells and failed to clear Mtb, thereby forming an immunologically ‘blind’ niche. Bacille Calmette-Guérin (BCG) vaccination reduced Mtb burden and niche establishment, without altering myeloid trafficking or T cell priming. Thus, Mtb persists in medLNs within monocytic niches that elicit classical antimicrobial programs which are insufficient for sterilizing control.

Graphical Abstract

eTOC:

Lymph nodes initiate protective T cell responses to Mycobacterium tuberculosis yet also serve as sites of chronic infection. Shamskhou et al. shed light on this paradox by showing a shift from early dendritic cell–driven Th1 cell priming to monocyte-derived bacterial niches in the T cell zone that evade recognition and promote persistence.

Introduction:

Tuberculosis remains a leading cause of global mortality caused by a single pathogen, Mycobacterium tuberculosis (Mtb)¹. Although pulmonary infection has dominated research focus, Mtb pathogenesis extends beyond the lung to secondary lymphoid organs and other tissues². Studies in mice have shown that Mtb first disseminates to the lung-draining mediastinal lymph node (medLN) beginning around eight days post-infection (dpi)^3–5. This triggers local initiation of T cell responses, which are essential for immune control^3–8. Paradoxically, lung-draining lymph nodes often become a preferred site for Mtb persistence⁹. Human post-mortem studies from the pre-antibiotic era frequently recovered Mtb from lung-draining lymph nodes even when lung lesions appeared sterilized¹⁰. In human TB and non-human primate models, Mtb can form granulomatous lesions within LNs that resemble those in the lung^11,12. How these opposing processes — robust immune activation and persistent infection – can be supported within the same lymphoid organ represents an understudied aspect of TB pathogenesis.

Previous studies have suggested that initial transport of bacteria from lungs to lymph nodes is mediated by inflammatory monocytes (iMOs)⁷, whereas conventional dendritic cells (cDCs) elicit T cell priming after acquiring Mtb antigens from these transporting iMOs^6,7,13,14. However, the classical phenotypic markers used for classification or depletion of specific innate populations have proven insufficient to faithfully capture myeloid cell diversity or isolate effects to select subsets^14–18. Emerging evidence suggests that inflammatory conditions further blur conventional phenotypic boundaries, with cDC2 populations acquiring monocyte-like characteristics upon IFN-I stimulation, termed inflammatory cDC2s (inf-cDC2s)^17–19. These inf-cDC2s express markers typically associated with iMOs, while also acquiring IL-12 production typically associated with cDC1s¹⁸, challenging simplified models of myeloid cell function during infection.

This cellular complexity is particularly relevant given the heterogeneity of CD4⁺ T cell responses during Mtb infection^20–24. While cDCs are recognized as principal antigen-presenting cells for T cell priming^6,7,13,25, which specific subsets instruct different effector T cell fates remains unresolved. Moreover, vaccination and cutaneous infection studies demonstrated that spatial organization of innate cells within lymphoid organs generates microenvironments specialized in promoting effector versus memory precursor T cell development^25–28, suggesting that tissue architecture shapes adaptive immune outcomes. Whether similar microenvironments arise during Mtb infection in the medLN and how they influence T cell priming and differentiation remains unknown.

The temporal dimension likewise requires consideration. T cell priming occurs within the first few weeks of infection⁸. These activated T cells can sustain responses throughout infection²⁹, indicating potent progenitor generation. Yet, certain vaccination studies have suggested that effector versus memory progenitor states are temporally encoded³⁰, and whether these programs arise at distinct timepoints during Mtb infection, when antigen and cytokine availability is altered, remains unknown. Furthermore, how prior immune status, such as BCG vaccination, modulates this balance has not been studied. While BCG limits disseminated Mtb infection^31,32, its effects on LN-localized myeloid cell states and T cell priming within infected LNs have not been defined.

The limited understanding of how this critical immunological site functions during infection highlights a key paradox: how can lung-draining LNs both initiate protective immunity and support long-term pathogen persistence? One explanation is that T cell priming and microbial persistence are spatiotemporally distinct processes supported by different cellular niches. To address this, we integrated mouse Mtb infection models with imaging, single-cell analysis, and mathematical modeling to define the spatiotemporal dynamics of infection and T cell priming in the medLN. We find that Mtb infection undergoes a temporal shift from early immune activation to the establishment of a persistent, monocyte-derived niche that resists bacterial clearance and evades T cell recognition. BCG immunization perturbs this niche and Mtb replication without altering initial Mtb trafficking or T cell priming. Together, these findings demonstrate that despite a brief but robust window of T cell priming, Mtb exploits the lymph node environment to establish a persistent, immunologically privileged niche, reflecting an advanced strategy for immune evasion.

Results:

Spatial and Temporal Analysis of medLN Infection Reveals a Monocytic Mtb Niche

To define the myeloid landscape harboring Mtb in the medLN, we infected C57BL/6 mice with mCherry-expressing Mtb (H37Rv; 50–100 CFU) and performed longitudinal analysis of infected cells by flow cytometry. Myeloid populations were defined as neutrophils (Ly6G⁺), inflammatory monocytes (iMOs; CD88⁺SIRPα⁺CD64⁺CD26^LOLy6C^HI), differentiated monocyte-derived cells (MCs; CD88⁺SIRPα⁺CD64⁺CD26^LOLy6C^INT-LO), and cDCs (CD88⁻CD26⁺CD11c⁺MHCII⁺), including resident (CD11c^HIMHCII^INT) and migratory (CD11c^INTMHCII^HI) subsets, further subdivided into cDC1 (XCR1⁺) and cDC2 (SIRPα⁺CCR2^INT-LO) (Fig. S.1A).

Mtb dissemination from lungs to the medLN has been reported as early as 8 dpi when assessed by culture^4,5. By flow cytometry, however, Mtb-infected (mCherry⁺) cells in the medLN were reliably detected by 15 dpi (Fig. 1A, S.1A). Between 15 and 22 dpi, infected medLN cells increased approximately ten-fold before plateauing at later timepoints, consistent with prior CFU studies⁴. At 15 dpi, Mtb was detected in cDCs (both cDC1 and cDC2) and iMOs (Fig. 1B). By contrast, at 22–28 dpi, the infected cell compartment shifted, with reduced cDC representation and a corresponding increase in iMOs and MCs (Fig. 1B). This redistribution was not explained by changes in overall composition of innate cells in the medLN, as total numbers of cDCs, iMOs, and MCs were maintained or increased across these timepoints (Fig. S.1B). In contrast, lung infection was characterized by preferential early infection of alveolar macrophages (AMs) and neutrophils³³, underscoring tissue-specific infection dynamics. Notably, we observed a similar temporal transition in the representation of iMOs and MCs within Mtb-infected lungs, though this shift was less pronounced than in medLNs, and cDCs remained a consistently minor infected population (Fig. S.1C). Consistent with prior reports^7,16–18,34, these myeloid populations exhibited distinct mean fluorescence intensity (MFI) of CCR2 expression (Fig. S.1A), highest in iMOs and variable in MCs and cDC2s.

Figure 1: Mtb infection is initially harbored in cDCs and monocytes, and later in multicellular monocytic aggregates in the medLN.

(A-F) C57Bl/6 mice infected with Mtb H37Rv.mCherry (50–100 CFU) (A-B) Flow cytometry of medLNs, 15 to 28+ (28–33) dpi. (A) Total Mtb⁺ cells per medLN (Fig. S1A: gating strategy). (B) Mtb⁺ innate cells as frequency of B220⁻CD3⁻IV⁻Mtb⁺ cells, and absolute numbers. Graphs reflect mean±95%CI. N=26–30 mice/group. (C-F) Confocal microscopy of myeloid cells in naïve and Mtb-infected medLNs at D15-D28+ (C); Mtb density (D), myeloid aggregate density and volume (E), and Mtb/Mtb⁺ myeloid aggregate quantifications (F). N=7–11 mice/group. T/B border (dotted line) defined by B220 staining (bottom-left). Arrows indicate infected cells/aggregates. Scale-bars: 300μm (zoom-out); 20μm (zoom-in). (G-J) Mtb H37Rv.mCherry (50–100 CFU) infected Zbtb46-GFP mice (N=5). (G-H) D15 and D28+ p.i. medLN microscopy. Frequency of Mtb surfaces assigned to cDCs (Zbtb46-GFP⁺), MOs (Zbtb46-GFP⁻SIRPα⁺), or Not determined (N.D.; Zbtb46-GFP^dim/SIRPa^dim). Scale-bars: 10μm. (I-J) Flow cytometry of D15 and D25 p.i. Zbtb46-GFP medLNs, identifying Mtb-infected cDCs (Zbtb46-GFP⁺CD88⁻CD26⁺) and Monocytes (CD88⁺Zbtb46-GFP⁻). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Significance measured by Kruskal-Wallis Test with Dunn’s Multiple Comparisons (A, D) or ordinary one-way ANOVA with Tukey’s Correction (B, E, F). Graphs reflect mean±SD, unless noted. Data represent 3 (C-J) or 6 (A-B) independent experiments, identified by symbol shape. See also Figure S.1.

We next studied the spatiotemporal dynamics of medLN infection using multiparameter confocal microscopy. At 15 dpi, individual Mtb bacilli were detected within dispersed myeloid cells (CD11c^+/−SIRPα^+/−), which were primarily localized within the T cell zone (Fig. 1C). By contrast, just one week later (day 22 and after), we observed an increase in bacterial burden (Fig. 1D), with bacilli now predominantly contained within large multicellular myeloid cell aggregates (CD11c⁺SIRPα⁺) distributed throughout the T cell zone (Fig. 1C–F). These aggregates persisted for at least 110 dpi, suggesting establishment of a long-lived niche for Mtb within the medLN (Fig. S.1D). Similar myeloid cell aggregates containing intracellular Mtb antigens (PPD staining) also formed in a more physiologically relevant ultra-low-dose (ULD; 1–3 CFU per mouse) infection model³⁵ (Fig. S.1E), indicating that this phenotype is independent of infectious dose.

We further examined medLNs at 34 dpi in C3HeB/FeJ mice infected with a W-Beijing Mtb strain, SA161. In this model, mice infected with 50–100 CFU develop at least one large necrotic lung granuloma^36,37, characterized by dense MPO⁺CD11b^HI neutrophil aggregates and loss of p120⁺ alveolar epithelium (Fig. S.1F). We observed similar Mtb-antigen-containing myeloid aggregates throughout the medLN T cell zone (Fig. S.1F), suggesting this biology is conserved across mouse and Mtb genotypes with different pulmonary pathologies.

Because CD11c and SIRPα can be expressed by both cDCs and monocyte-derived populations, we used Zbtb46-GFP reporter mice to trace cDC-lineage cells³⁸. At 15 dpi, approximately 45% of Mtb bacilli in the medLN localized within GFP⁺ cDCs, whereas ~20% of bacilli were contained within GFP⁻SIRPα⁺ cells, indicating monocytic/macrophage origin cells (Fig. 1G–H). By 22–28 dpi, however, the majority of Mtb bacilli were found within GFP⁻SIRPα⁺ cells, while fewer than 10% of bacilli were within GFP⁺ cDCs. This temporal redistribution was also observed by flow cytometry (Fig. 1I–J).

Together, these data define two stages of Mtb infection within medLN myeloid cells. The first (~15 dpi) is characterized by dispersed infiltration of the T cell zone by cDCs and MOs harboring single or few bacilli. This is followed by a second phase marked by formation of monocyte-derived aggregates throughout the medLN parenchyma, particularly the T cell zone, with increased bacterial burden. The persistence of these aggregates suggests they represent a stable niche for Mtb.

A Cross-Organ, Cross-Timepoint Atlas of Innate Cell Dynamics Defines Myeloid Cell Transitions During Mtb Infection

To better define how innate cells change across tissues during Mtb infection, we generated a single-cell transcriptional atlas spanning lung, medLN, and blood at 15 and 28 dpi, timepoints with distinct cellular and structural features in the medLN, alongside naïve baselines (0 dpi). We profiled both Mtb-infected (mCherry⁺) and uninfected bystander (mCherry⁻) parenchymally localized (IV⁻) innate immune cells from the lung and medLN. In the blood, we assessed only uninfected cells due to the rarity of Mtb-infected populations. This yielded 93,145 high-quality cells for analysis. Cell identities were assigned using ImmGen-based references with manual refinement informed by curated lineage markers consistent with flow cytometry (Fig. 2A, Fig. S2A; Table S.1).

Figure 2: Transcriptional remodeling reveals inflammatory programs in Mtb-infected myeloid cells.

Single cell RNAseq of sorted Mtb-mCherry^+/− myeloid cells from lungs, medLNs, and blood (D15/D28 H37Rv.mCherry 50–100 CFU: N=16; D0: N=3). (A) UMAP clustering by myeloid phenotype faceted by tissue, timepoint, and infection status. (B) Proportion plots of Mtb-infected cell types in medLN and lung. (C) mSigDB.Hallmark GSEA faceted by cell type, infection, and timepoint, as compared to D0 for each cell type/tissue. (D) GO:BP GSEA of Mtb-infected iMO2s and MCs at D28, comparing lung versus medLN. Leading-edge genes shown under each parent term. (E) Nos2 expression in select Mtb^+/− medLN populations and microscopy of iNOS expression D22 p.i. Scale-bar: 20μm. See also Figure S.2.

In uninfected (mCherry⁻) cells, each tissue exhibited distinct and temporally dynamic phenotypes. In the lung, naïve samples contained relatively few monocytes defined by expression of Nr4a1, Csf1r, Ly6c2, Ccr2, Fcgr1, whereas by 28 dpi there was pronounced enrichment of inflammatory monocyte 2 (iMO2) cells (Ly6c2^IntCcr2^IntLy6i^Hi) and emergence of monocyte-derived cells (MCs). MCs expressed Mitf and Sirpa alongside MHC and C1q-associated genes, a signature linked to tissue-regenerative interstitial macrophages during Mtb infection^39–41 and a biomarker of TB disease in patients⁴² (Fig. 2A, Fig. S2A–C).

In the medLN, naïve samples were dominated by cDC populations, including resident (Ccr7⁻Fscn1⁻) and migratory (Ccr7⁺Fscn1⁺) cDC1s (Irf8⁺Xcr1⁺Clec9a⁺) and cDC2s (Irf4⁺Xcr1⁻), and Irf4⁻Irf8⁻Cd11b⁻ DCs (here termed transitional DCs)⁴³ (Fig. 2A, S.2C). By 15 dpi, this composition shifted toward monocytic dominance, with expansion of iMO1, iMO2, and MC populations, suggesting early recruitment of blood-derived myeloid cells into the medLN. Consitent with this, blood cells were dominated by circulating monocyte subsets, including iMO1 (Ly6c2^HiCcr2^Hi), iMO2, iMO3 (Ly6c2^LOCcr2^LO), and patrolling monocytes (pMOs, Pou2f2⁺Ly6c2⁻), alongside neutrophils and fewer cDC2s. iMO3 cells increased at 15 dpi, followed by iMO2 expansion at 28 dpi, suggesting sustained monocytosis throughout infection (Fig. 2A, S.2B).

Analysis of Mtb-infected (mCherry⁺) cells revealed tissue- and time-dependent transitions. In the lung, infected cells at 15 dpi were dominated by AMs and neutrophils, along with monocyte-derived populations, but with minimal representation of cDCs. By 28 dpi, infected lungs showed expansion of iMO2s and, to a lesser extent, MCs (Fig. 2A–B). In contrast, infected medLNs at 15 dpi contained a heterogeneous mixture of iMOs, MCs, and cDCs, consistent with flow cytometry and imaging analyses. By 28 dpi, the infected medLN compartment became dominated by iMO2s and MCs, with a marked reduction in infected cDCs, indicating a transition from cDC involvement to monocyte-derived dominance.

To confirm that the monocyte-derived populations observed by scRNA-seq corresponded to the monocytic aggregates observed by imaging and flow cytometry, we examined lineage-associated transcriptional features. Compared to iMO1 and iMO2 cells, MCs showed reduced Ly6c2 and Ccr2 expression (Fig. S.2C), consistent with loss of circulating monocyte characteristics and flow cytometry-based surface marker profiles (Fig. S.1A). In contrast, infected MCs, and to a lesser extent iMO2s, exhibited increased expression of Cd68, Vcam1, and Mertk, genes associated with phagocytosis and tissue residency^44,45 (Fig. S.2D). These transcriptional features were recapitulated at the protein level, with CD68, VCAM1, and MerTK enriched within the medLN aggregates by confocal microscopy (Fig. S.2E).

To assess developmental relationships within the monocyte lineage, we performed Slingshot trajectory analysis⁴⁶ on monocytic cells across tissues (Fig. S.2F). This revealed a continuum from circulating to tissue-associated states. Pseudotime was anchored to naïve pMOs and iMO1s⁴⁷. iMO3 cells, enriched in blood 15 dpi, showed low pseudotime, consistent with an early myelopoietic response. iMO2s occupied intermediate positions, whereas MCs displayed the highest pseudotime, suggesting they represent the most differentiated monocyte-derived population.

Together, these data define temporal remodeling of Mtb-infected myeloid niches across lung and medLN, with early recruitment of monocytes and monocyte-derived cells that dominate infected compartments at later stages. This transition occurs more rapidly in the medLN than in the lung. While cDCs transiently contribute to early medLN infection, they are progressively excluded as monocyte-derived aggregates establish dominance.

Differential Gene Expression and Pathway Enrichment Reveal Functional Remodeling of Monocytic Populations

To define infection-associated programs across tissues, we examined differentially expressed genes (DEGs) relative to naïve baselines in Mtb-infected (mCherry⁺) and uninfected bystander (mCherry⁻) cells, followed by Gene Set Enrichment Analysis (GSEA). This revealed tissue- and time-dependent remodeling of inflammatory and metabolic pathways (Fig. 2C; Data S.1–2). In the lung, uninfected myeloid populations showed limited transcriptional changes at 15 dpi, followed by enrichment of interferon-associated and inflammatory pathways at 28 dpi. A similar pattern was observed in the blood, with minimal transcriptional changes at 15 dpi but prominent IFN and inflammation pathway signatures at 28 dpi. In contrast, uninfected medLN bystander cells exhibited robust enrichment of inflammatory pathways, including IFN-I, IFN-II, and TNF, already at 15 dpi. By 28 dpi, several pathways, including TNF signaling and p53-associated programs, were attenuated relative to naïve controls, indicating a transient inflammatory state within the medLN.

Comparing Mtb-infected populations to naïve cells revealed induction of inflammatory pathways in both lungs and medLNs at 15 and 28 dpi (Fig. 2C), indicating robust pathogen sensing and cytokine responses in both organs. Hypoxia pathways were also induced in infected MCs and iMO2s. Tissue-specific differences emerged, with infected medLN cells showing enrichment of xenobiotic metabolism and p53-associated pathways, as well as tissue remodeling and repair pathways at 28 dpi. We also observed temporal shifts in metabolic pathways. Oxidative phosphorylation and mTORC1 signaling were preferentially enriched in infected MCs at 15 dpi, whereas glycolysis, hypoxia, fatty acid metabolism, and heme metabolism pathways predominated at 28 dpi.

These pathway-level findings were supported by DEG analysis of iMO2s and MCs (Fig. S.2G). Compared to naïve cells, infected cells in both the lung and medLN upregulated genes associated with interferon responses (Nos2, Gbp2, Gbp7, Cxcl9, Cxcl10), antigen presentation (H2-K1, H2-Ab1, H2-M2, H2-Q4), hypoxia and metabolic adaptation (Hif1a, Slc7a11, Fabp5, Acod1), and immune regulation (Il18bp, Cd274, Inhba). Comparison of DEGs between timepoints demonstrated that in the lung, infected iMO2s and MCs exhibited relatively limited transcriptional differences between 15 and 28 dpi, with the later timepoint characterized by modest increases in interferon-associated genes, particularly within MCs (Fig. S.2H). In the medLN, infected cells displayed a temporal shift in transcriptional programs (Fig. S.2I). Genes enriched at 15 dpi included interferon-stimulated genes (Isg15, Irf1, Irf7, Oasl2, Zbp1), inflammatory mediators (Il6, Il1rn, Il15ra), and activation and survival markers (Cd40, Bcl2a1b). By 28 dpi, infected medLN cells upregulated genes associated with wound healing (Vegfa, Mertk, Sash1), phagocytosis and efferocytosis (Cd68, Lyz2, Ctsd, Ctss, Lamp1), and metabolic adaptation (Apoe, Slc7a11, Nos2).

We also examined tissue-specific differences between lung and medLN infected iMO2s and MCs at a matched timepoint, 28 dpi (Fig. 2D). Gene Ontology (Biological Process; GO:BP) analysis demonstrated that in the lung, MCs preferentially engaged cell signaling and cell cycle–associated pathways, while iMO2s were biased toward tissue regenerative pathways. In contrast, in the medLN, infected iMO2s and MCs were enriched for metabolic pathways, including NAD metabolism, monocarboxylic acid processing, and neutral lipid metabolism, consistent with metabolic reprogramming during Mtb infection⁴⁸.

We observed robust induction of IFN-γ–associated genes in infected medLN populations. In particular, Nos2 (encoding iNOS), an IFN-γ–induced effector that mediates bactericidal activity and Mtb control but also regulates inflammation^49–51, was upregulated in infected MCs and iMO2s at 15 and 28 dpi (Fig. 2E, S.2G). Immunofluorescence microscopy confirmed abundant iNOS within Mtb-containing myeloid aggregates in the medLN (Fig. 2E). Thus, despite the well-established role of IFN-γ signaling in Mtb control^52–54, its induction in the medLN is insufficient for bacterial clearance.

Consistent with our findings, prior single-cell analyses of lung myeloid cells during Mtb infection identified an infected interstitial macrophage population (IM_3) characterized by high Nos2 expression and iron metabolism signatures⁴¹. We found transcriptional concordance between the MC population defined here and the previously described IM_3s, and these cells constituted the dominant infected cell population at 28 dpi (Fig. S.2J–K). Notably, the prior study demonstrated that IM_3 cells harbored Mtb expressing a DosR stress-response reporter^55,56, suggesting that the MCs identified here may similarly support bacterial persistence.

Collectively, these data demonstrate that Mtb infection induces a temporally structured myeloid response within the medLN. Early infection (15 dpi) is characterized by large-scale induction of inflammatory cytokine and interferon-associated programs across both infected and bystander populations. By 28 dpi, this inflammatory state is attenuated and replaced by programs of metabolic adaptation, hypoxia, and tissue remodeling within infected myeloid aggregates, consistent with establishment of a persistent, pathogen-permissive niche. These shifts are distinct from those in the lung, which show progressive increases in inflammation across timepoints.

Tracking Infected Cell Origin Reveals Local Mtb Replication in the medLN

We next aimed to define the anatomical source of Mtb-infected myeloid cells in the medLN, particularly during the exponential phase of bacterial growth and emergence of myeloid aggregates. To track myeloid cell migration from the lung and blood, we combined intratracheal⁵⁷ and serial-intravascular (sIVs)^58–60 labeling strategies to distinguish lung-derived and blood-derived cells within the medLN parenchyma (Fig. 3A). Ovalbumin–Alexa-Fluor-488 (OVA-488) was administered intratracheally 24-hours prior to harvest to label lung myeloid cells that subsequently migrated to the medLN, while two temporally separated intravenous CD45.2 antibody injections (24-hours and 5-minutes before euthanasia) marked cells entering via the blood vasculature. This enabled identification of medLN parenchymal cells (BV421⁻) that had recently migrated from the lung (OVA-488⁺) or circulation (APC⁺) (Fig. 3B, Fig. S3A–C)^58,59,61.

Figure 3: Myeloid Cell Migration Dynamics Support Local Mtb Replication in the medLN.

(A) Experimental schematic. (B) Concatenated flow plots of lung/blood-derived Mtb⁻ and Mtb⁺ myeloid cells in medLNs. C-D) Lung/blood-derived Mtb^+/− innate cell numbers at D15-D28+ (28–33) p.i. (E-G) Cell type pie charts for total or lung/blood-derived Mtb-infected cells in medLN. (B, E-G) reflect concatenated data from each experiment. 5 mice concatenated/timepoint, including iMO, MC, cDC1, and cDC2 populations. (H) Mathematical model schematics. (Bone marrow (M), blood (B), lung (L), medLN (N), immune response (E)) (H.i) Modeling of blood monocyte migration to medLN: monopoeisis rate p, medLN circulation rate m_BN, cell death/migration d (eqns. (1)–(8)). (H.ii) Modeling of Mtb⁺ myeloid cells replication (r), migration (m), immune response (E) at rate ρ, and Mtb⁺ cell death rate (δ)(eqns. (11)–(15)). (I) Model fit predictions of Mtb⁺ myeloid cells in the medLN at different Mtb replication rates (r_N). Lines represent model predictions (Table S.3: model parameters). Data represents 3 (D-G) or 6 (B-C, I) independent experiments. See also Figure S.3.

Infection resulted in a 5–10-fold increase in OVA-488⁺ myeloid cells migrating from the lung to the medLN over 24 h. These cells rose modestly, remained relatively stable, and peaked by 30 dpi at ~10⁴ cells per 24 h (Fig. 3C), with only a small fraction (1–2%) harboring Mtb (mCherry⁺). In contrast, blood-derived myeloid cell (APC⁺) entry increased ~1,000-fold relative to naïve controls, becoming evident by 15 dpi, peaking at 22 dpi, and declining by 30 dpi (Fig. 3D). A modestly larger fraction of these cells was Mtb-infected (~0.5–10%) than lung-derived cells. Because most blood-derived cells are unlikely to be infected before entering the medLN, these data suggest that blood-derived myeloid cells are recruited and infected locally.

Analysis of lung-derived populations revealed temporal shifts in composition. Early (day 15) seeding involved both iMOs and cDCs carrying Mtb from the lungs, similar to total Mtb-infected cells in the medLN (Fig. 3E–G). However, after 22 dpi, most lung-derived Mtb-infected cells were MCs (Fig. 3E), while uninfected lung-derived cells still included large numbers of cDC2s (Fig. S.3D). sIVs-labeled blood-derived Mtb-infected cells showed similar patterns, with early involvement of both cDCs and iMOs and greater MC representation at later timepoints (Fig. 3F, 3G).

These findings indicate that bacterial transport to the medLN occurs relatively infrequently and does not correlate with the rapid increase in medLN bacterial burden. To quantitatively assess the contribution of lung-derived Mtb-infected myeloid cells, we developed a mathematical model incorporating cell migration and bacterial replication (Fig. 3H, Methods eqns (11)–(15)). The initial model permitted Mtb replication in the lung and migration of Mtb-infected cells from the lung to medLN, and was fit to experimental cell counts.

The best-fitting model recapitulated the accumulation, peak, and early decline of Mtb⁺ myeloid cells in the lung (Fig. S.3E), and required lung-to-medLN migration rates to vary over time in a step-like manner (Fig. S.3F). However, a medLN model where Mtb⁺ cells accumulated solely through migration from the lung without local replication (i.e., replication rate r_N=0/day) failed to reproduce the rapid rise in Mtb⁺ cells from 15–22 dpi (Fig. 3I). Allowing Mtb replication within the medLN at relatively high rates (r=0.5–0.7/day or doubling times of 29–33h) enabled the model to closely match the observed medLN infection dynamics (Fig. 3I). This estimated replication rate was similar to that in the lung (Fig. 3I, S.3E).

These results align with our microscopy observations, demonstrating early detection of single Mtb bacilli within individual cells, followed by accumulation of myeloid aggregates containing numerous bacilli. This supports a model in which initial Mtb dissemination to medLNs occurs via relatively low-frequency trafficking from the lung by cDCs and iMOs, followed by monocyte recruitment from the blood and local bacterial replication within monocyte-derived aggregates.

Mtb-Specific T Cell Priming Occurs Within a Narrow Temporal Window in the medLN

Effector CD4⁺ T cells are critical for Mtb control⁶², with ESAT-6–specific CD4⁺ T cell priming in the medLN initiating ~10 dpi and peaking around three weeks, coincident with Mtb dissemination^3–5. Despite stable medLN bacterial burdens thereafter (Fig. 1A), T cell priming declines²⁹, raising the question of how evolving changes in cDCs and monocytic aggregates influence T cell priming and effector differentiation.

To investigate this, we adoptively transferred naïve CD4⁺ TCR-transgenic cells specific for ESAT-6_3–15 (C7 line)⁶³ into congenic recipients at defined timepoints post-infection, with tissues analyzed 2–3 days later by confocal microscopy or flow cytometry (Fig. 4A). This timed transfer approach captures the capacity of the medLN microenvironment to support de novo T cell priming, rather than cumulative responses. At 15 dpi, C7 cells exhibited robust proliferation, with extensive CellTrace Violet (CTV) dye dilution and clonal expansion (Fig. 4B–C), accompanied by increased IRF4 and pS6 expression, indicating enhanced TCR signaling^64,65 (Fig. 4D–E). These cells also efficiently differentiated into Th1 cell effectors, denoted by increased T-bet and CD25 expression (Fig. 4F–G, S.4A). However, by 22–28 dpi, T cell proliferation and IRF4/pS6 expression were markedly reduced, indicating impaired priming (Fig. 4B–E), accompanied by diminished Th1 cell differentiation and PD1 expression ((Fig. 4F–G, S.4A, S.4B). Activated T cells across timepoints did not extensively upregulate KLRG1 or BCL6 expression, indicating limited differentiation into terminal effector^20,23 or T follicular helper⁶⁶ lineages within this timeframe (Fig. S.4C–E). We observed similar temporal patterns in adoptively transferred Ag85B-specific (P25)⁶⁷ CD4⁺ T cells (Fig. S.4F), and in endogenous CD4⁺ and CD8⁺ T cell responses (Fig. S.4G), indicating that these changes were not restricted to a single antigen specificity. Similarly, limited T cell activation and differentiation were observed in the spleen (Fig. S.4H), suggesting a broader failure of lymphoid organs to support effective priming at later stages of infection.

Figure 4: Mtb-specific T cell priming declines as myeloid aggregates and Mtb burden increase in the medLN.

(A) Experimental schematic. (B-G) Flow cytometry analysis of C7 (Live,DUMP⁻CD3⁺CD4⁺CD8⁻CD90.1/CD45.1⁺) responses in medLNs D15-D28+ (28–33) dpi, subgated for divided cells (1+ division) to examine pS6, IRF4, and Tbet expression, and percent Tbet⁺CD25⁺ or Tbet^+/−TCF1^+/−. (H-L) Microscopy of medLNs demonstrating C7 activation (Ki67, pS6) with respect to infected myeloid cells, quantifying C7 cluster density (I, K), and Ki67/pS6 (J, L). Scale-bar: 200μm (zoom-out); 10μm (zoom-in). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Significance measured by ordinary one-way ANOVA with Tukey’s Correction. Graphs reflect mean±SD. Data represent 4–6 (B-C,F-G, N= 21–26 mice; D-E, N=8–14 mice), 3 (H-J, N=9–14 mice) or 1 (K-L, N=31 mice) independent experiments, identified by symbol shape. See also Figure S.4.

Visualization of medLNs confirmed these temporal changes in T cell activation. At 15 dpi, when innate cells harbored single bacilli, we observed robust C7 clustering with strong pS6 and Ki67 staining, indicating potent T cell activation (Fig. 4H–J). In contrast, later timepoints dominated by heavily-infected monocytic aggregates showed reduced T cell numbers, diminished clustering, and lower Ki67 expression. While C7 cells were directly juxtaposed to the infected aggregates, they lacked pS6 staining, indicating ineffective antigen recognition (Fig. 4H). Reduced T cell priming at 26 dpi was also observed in the ULD infection model (Fig. 4K–L, Fig. S.4I). Impaired T cell activation persisted until at least 50 dpi (Fig. S.4J), indicating a prolonged shutdown of the tissue’s capacity to support T cell responses. Although B cell infiltration occurred around some aggregates, as previously described⁶⁸, most infected aggregates remained in direct contact with both endogenous and C7 T cells, suggesting physical access was not the dominant limiting factor (Fig. S.4J). These findings demonstrate a narrow temporal window for effective T cell priming in the medLN, occurring prior to the emergence of Mtb-infected myeloid niches and when the bacterial burden is relatively low.

Mtb-Specific CD4⁺ T Cell Priming is Driven by the Sequential Activity of cDC1s and Inflammatory cDC2s

To identify myeloid cell types supporting T cell responses, we distinguished cDCs from monocyte-derived populations using Zbtb46-GFP reporter mice. At 15 and 22 dpi, timepoints associated with T cell priming, most C7 cells were clustered around Mtb⁺Zbtb46-GFP⁺ cells (Fig. 5A–B, S.5A), suggesting that initial activation is driven by infected cDCs. Given that IL-12 is critical for Th1 cell polarization^69–71, we examined the distribution of IL-12p40, a subunit of IL-12p70. Activated C7 T cells frequently clustered around IL-12p40⁺ cells, nearly all of which were Zbtb46-GFP⁺ cDCs (Fig. 5A–B). By 30 dpi, T cell clustering and presence of IL-12p40⁺ cDCs associated with clusters diminished (Fig. 4I, 5A–B), mirroring the loss of effector T cells and reduced proliferation observed by flow cytometry.

Figure 5: cDC1s support T cell effector differentiation in the medLN.

(A-B) Microscopy of C7 clusters surrounding IL-12p40⁺ cells in Zbtb46-GFP mice. Scale bar: 10 μm (A). C7 cluster-associated IL-12p40 signal enumerated and phenotyped for cell type by Zbtb46-GFP/SIRPα (B). Data representative of 3 independent experiments.

(C-D) Microscopy of XCR-Venus medLNs showing C7 clusters around XCR-Venus⁺ cDC1s (arrows). Scale bar: 100 μm (zoom-out), 10 μm (zoom-in) (C). C7 cluster-associated IL-12⁺ cells quantification for XCR1-Venus/SIRPa expression (D).

(E-G) Microscopy of C7 clusters associated with IL-12⁺ cells in wild-type (WT) and BATF3^−/− mice. Scale bar: 10 μm (E). C7 cluster density quantification (F). Fraction of C7 clusters associated with IL-12⁺ cells (G).

(H and I) Tbet/TCF1 expression in medLN C7 cells. *p < 0.05, ***p < 0.001. Significance measured by Student’s unpaired t test. Graphs reflect mean ± SD. Data represent 3 independent experiments.

Since cDC1s are potent producers of IL-12, we used XCR1-Venus cDC1 reporter mice⁷² to visualize their distribution relative to activated T cells in medLNs. We observed individual cDC1s (Venus⁺) harboring single bacilli (Fig. 5C), supporting their role in bacterial transport from infected lungs (Fig. 3E). Many infected cDC1s expressed IL-12p40 and co-localized with clusters of activated C7 cells. However, we also noted IL-12p40-expressing Venus⁻SIRPα⁺ cells associated with C7 clusters, suggesting that additional myeloid populations contribute to this process (Fig. 5C–D).

To assess the role of cDC1s in T cell priming, we performed timed C7 cell transfers into BATF3^−/− mice⁷³, which lack normal cDC1 development (Fig. S.5B). Flow cytometry confirmed reductions in cDC1s in lungs and medLNs, with no changes to bacterial burden (Fig. S.5C–D). We observed comparable C7 proliferation and expansion in BATF3^−/− and wild-type mice at 15 and 22 dpi (Fig. S.5E–F), as well as similar antigen sensing (pS6 expression) and clustering (Fig. 5E–F), indicating that cDC1s are not required for T cell activation. However, the frequency of IL-12p40⁺ cells associated with T cell clusters was reduced in Batf3^−/− mice at 15 dpi (Fig. 5G), which corresponded with partially diminished Th1 cell differentiation, marked by lower frequencies of T-bet⁺TCF1⁻ cells and a trend toward decreased T-bet expression (Fig. 5H–I)⁷⁴. These impairments were not evident at 22 dpi, suggesting a role for cDC1s in early Th1 cell differentiation, and compensatory mechanisms at later timepoints.

To this end, we observed strong induction of IFN pathways in cDC2s (Fig. 2C, 6A), suggesting conditions conducive to inf-cDC2 development^{17–19,75,76}. Indeed, flow cytometric analysis of cDC2s (CD88⁻CD26⁺CD11c⁺MHCII⁺SIRPα⁺XCR1⁻) identified cells co-expressing Ly6C and CD64, phenotypic characteristics of inf-cDC2s (Fig. S.1A, S.6A). A significant proportion of Mtb-infected cDC2s (mCherry⁺) exhibited this phenotype (Fig.S.6B), indicating that infection promotes inf-cDC2 development. To test whether IFN-I signaling promotes inf-cDC2 generation and subsequent T cell priming, we infected wild-type or Ifnar1^−/− mice and performed timed adoptive C7 cell transfers. In this model, Ifnar1 deficiency is restricted to host cells, while transferred T cells retain intact IFN-I signaling. Compared to infected wild-type controls, Ifnar1^−/− mice exhibited a marked reduction in total and Mtb-infected inf-cDC2s at 15 and 22 dpi (Fig. 6B–D), despite comparable lung bacterial burdens (Fig. S.6C), numbers of Mtb-infected cells in the medLN (Fig. S.6D), and myeloid aggregate formation (Fig. S.6E). A modest reduction in iMO cellularity was observed 22 dpi (Fig. S.6F), suggesting a role for IFN-I in monocyte recruitment. Assessment of C7 responses revealed that T cell activation and Th1 cell differentiation were intact at 15 dpi, consistent with cDC1 driving T cell priming at this timepoint. However, by 22 dpi, Ifnar1^−/− mice exhibited a notable decline in T cell clustering around IL-12⁺ myeloid cells (Fig. 6E–F), accompanied by impaired T cell expansion and Th1 cell differentiation (Fig. 6G–H, S.6G).

Figure 6: Inf-cDC2s sustain effector T cell responses during Mtb infection.

(A) GO:BP GSEA of D15 versus D28 for Mtb⁻ medLN resDC2s. (B-H) WT and Ifnar^−/− mice infected with Mtb H37Rv.mCherry (50–100 CFU), harvested D15–28+ (28–33) p.i. (B-D) Flow cytometry of inf-cDC2s (CD88⁻CD26⁺CD11c⁺MHCII⁺SIRPa⁺CD64⁺Ly6C⁺). (E) Microscopy of C7 clusters associated with IL-12⁺ cells. Scale-bar: 10μm. (F) C7 cluster density and percent clusters associated with IL-12-expressing cells. (G-H) Flow cytometric analysis of C7 responses. Data representative of 2 independent experiments, indicated by symbol. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Significance measured by student’s unpaired T test. Graphs reflect mean±SD. See also Figure S.6.

Collectively, these findings support a temporal model, where Mtb-specific CD4⁺ T cell priming in the medLN is mediated by sequential cDC activity (Fig. S6H). Early after infection, cDC1s drive robust IL-12-dependent effector differentiation, wherease at later stages, IFN-I–dependent inf-cDC2s sustain T cell responses, contributing to limited effector activation that eventually dissipates. Both processes appear largely independent of expanding Mtb-infected monocytic aggregates, which fail to elicit potent T cell activation.

BCG Immunization Limits Monocytic Niche Establishment but not T cell Priming in the medLN

BCG immunization reduces Mtb dissemination from the lung to distal sites^31,32, but its impact on lung-to-medLN transport and subsequent dissemination remains unclear. BCG also modulates IFN-I/-II responses in myeloid cells^77,78, cytokines abundantly sensed in the medLN (Fig. 2, 6), yet its effects on myeloid cell dynamics and de novo T cell priming are not defined. To address this, we immunized mice with BCG eight weeks prior to Mtb infection and assessed medLN immune responses over time. BCG markedly reduced myeloid aggregate formation in the medLN at both early (18 dpi) and late (28 dpi) timepoints (Fig. 7A–B.I), and decreased the number of bacilli per LN section (Fig. 7B.II). Notably, the bacillary load per infected myeloid aggregate was lower at 18 but not 28 dpi (Fig. 7B.III), indicating that once infected, aggregates eventually accumulate similar burdens. Similar decreases in aggregate formation with BCG immunization were observed in the ULD infection model (Fig. 7C–D). Flow cytometry confirmed a ~10-fold reduction in total Mtb⁺ cells in medLNs of BCG-immunized mice (Fig. 7E). These reductions were most notable for Mtb⁺ iMOs and MCs at 22 and 28 dpi, but not at 15 dpi (Fig. 7F), suggesting comparable initial seeding but reduced later expansion of the monocytic niche and Mtb. Parallel reductions in Mtb⁺ iMOs and MCs were observed in the lung (Fig. S.7A). Phenotypic analysis further revealed that BCG immunization altered medLN iMO and MC maturation, marked by elevated MHC-II early and reduced CD11c at later timepoints (Fig. S.7B–C). In contrast to monocyte-derived populations, Mtb⁺ cDCs were present in similar numbers in the medLN regardless of immunization status (Fig. 7F, S.7D), indicating that BCG did not impact cDC trafficking from the lung or their cellularity.

Figure 7: BCG immunization enhances Mtb control in the medLN.

(A) Microscopy of D18 and D28 medLNs from unimmunized and BCG-immunized mice, and (B) quantifications of myeloid aggregate density (I), Mtb density (II), and total Mtb bacilli per Mtb-infected myeloid aggregate (III). N=4–5/group. Scale-bars: 200μm (zoom-out), 10μm (zoom-in). (C-D) Microscopy of D37–38 ULD medLNs with myeloid aggregate quantifications. Scale-bars: 200μm (zoom-out), 10μm (zoom-in). (E-F) Flow cytometry quantification of total Mtb⁺ innate cells in medLN at D15-D28+ (28–33) p.i., and (G) total or Mtb-infected inf-cDC2s (N=10–15 mice/group). (H) Bulk RNAseq IFN-b and IFN-g signature scores of bystander medLN MOs (CD3⁻B220⁻Ly6G⁻CD26⁻CD88⁺SIRPα⁺Mtb⁻). (I-J) ULD Mtb infection +/− BCG immunization (I) Quantification of D28 lung and medLN CFUs. (J) Mtb barcodes from lung and medLN. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Significance measured by Mann-Whitney Test (B, D), student’s unpaired T test (E-G), or Fischer’s Exact Test (J). Graphs reflect mean±SD. Data represent 2–3 independent experiments. See also Figure S.7.

We next evaluated de novo Mtb-specific T cell priming in the medLN following adoptive transfer of naïve C7 cells (ESAT6 is expressed by Mtb but not BCG⁷⁹). Across timepoints, T cell proliferation and expansion were comparable between non-immunized and BCG-immunized groups (Fig. S.7E), indicating that early cDC-mediated T cell priming was preserved despite markedly different Mtb burdens. T cell activation declined by 28 dpi in both groups, consistent with reduced Mtb-infected cDC2 numbers at this timepoint (Fig. 7F). In addition, we observed reduced C7 effector differentiation in BCG-immunized mice 22 dpi (Fig. S.7F), suggesting altered quality of priming conditions. Supporting this, total inf-cDC2 numbers in the medLN decreased following BCG immunization (Fig. 7G), implicating diminished IFN signaling. To assess this, we isolated monocytes from the medLN, lung, and blood of BCG-immunized and control Mtb-infected mice at 15 and 25 dpi and performed bulk RNA sequencing. Analysis of IFN-I and IFN-II–inducible gene signatures in the medLN⁸⁰ revealed that BCG immunization broadly suppressed both pathways at 15 dpi, with further dampening at 25 dpi in both groups (Fig. 7H, S.7G). In contrast, lung and blood monocytes were enriched for IFN signature genes 15 dpi in BCG-immunized mice (Fig. S.7G), consistent with findings that BCG accelerates immunity⁸¹ and that concomitant infection alters IFN responses in the lungs⁸². Together, these data align with prior reports that BCG modulates IFN-driven responses in myeloid cells^77,78, and demonstrate downstream effects on innate cells and T cell priming within the medLN (Fig. 6).

Collectively, these findings suggest that BCG limits monocytic niche formation in the medLN and reduces Mtb burden. Using intratracheal and intravascular labeling, we found that although BCG increased circulating monocytes in the blood (Fig. S.7H), it resulted in ~2-fold fewer blood-derived monocytes in medLNs at 15 and 22+ dpi (Fig. S.7H), suggesting reduced recruitment. To quantify these dynamics, we developed a mathematical model of monocyte trafficking fitted to these data (eqns. (6)–(8)). The model captured increased monocyte migration into the medLN in both groups (Fig. S.7H), and demonstrated a ~2–3-fold reduction in per-capita monocyte entry rate in BCG-immunized mice (Fig. S.7I and Table S.2). Total numbers of blood-derived Mtb-infected cells in the medLN were unchanged (Fig. S.7J), indicating that altered monocyte recruitment does not substantially contribute to reduced bacterial burdens.

Additionally, the model estimated comparable lung-to-medLN migration rates of Mtb-infected cells in both groups (Fig. S.7K), suggesting that BCG does not alter initial Mtb dissemination to the medLN. Inferred rates of Mtb replication in lung and medLN were also comparable (Fig. S.7L and Table S.3). Instead, reduced Mtb burden in BCG-immunized mice was best explained by earlier immune control of Mtb growth, reflected by a shorter time to half-maximal control (parameter T_E; Fig. S.7M). This is consistent with earlier Mtb control observed in lungs of BCG-immunized mice⁸¹.

Finally, to assess how BCG impacts Mtb dissemination to the medLN versus other organs, we employed the ULD model, infecting mice with a pool of 50 barcoded H37Rv strains^31,35. This allowed tracking dissemination of individual barcoded strains across tissues. BCG immunization reduced Mtb burden in lungs and medLNs at 28 dpi (Fig. 7I). As previously shown³¹, BCG also restricted the dissemination of barcoded Mtb strains from the initially infected lung to the contralateral lung. In contrast, the barcodes found in the medLN largely matched those in the primary infected lung, regardless of immunization status (Fig. 7J, S.7N). These findings suggest that BCG-mediated immunity does not prevent initial Mtb access to the medLN, but instead curtails local replication and myeloid niche formation, while limiting establishment of distal infection.

Discussion:

Lymph nodes are critical sites for initiating adaptive immune responses, yet Mtb and related mycobacteria effectively infect and persist within lymphatic tissues across species^9,10. In primary TB, particularly in children, this can manifest as lymphadenopathy and pathology⁸³. While prior mouse studies defined the onset of T cell responses in medLNs^4,5, how the same organ supports both robust T cell priming and long-term bacterial persistence has remained unclear. Here, we leveraged tools uniquely available in murine models to construct a spatiotemporal map of innate and adaptive responses to Mtb within the medLN. We show that infection proceeds through distinct stages. An initial phase is characterized by the induction of Mtb-specific CD4⁺ T effector responses driven by sequential waves of cDC activity. This phase is transient and coincides with the emergence of an immuno-evasive monocytic niche that supports local bacterial replication and persistence. This niche forms within the T cell zone despite strong induction of inflammatory and microbicidal pathways in infected cells. These findings indicate that Mtb is highly adapted for persistence within the LN microenvironment and underscore the limits of Th1-mediated immunity in achieving sterilizing protection.

Through blood and lung tracing experiments, paired with mathematical modeling, we demonstrate that both cDCs and monocytes initially transport Mtb from the lung to the medLN, contrasting with earlier work that attributed this solely to monocytes⁷. We believe this discrepancy reflects reliance on CCR2-based depletion in the earlier study, as during inflammation and Mtb infection, CCR2 is expressed by multiple immune populations, including cDCs. Over time, however, migration of Mtb-infected cDCs declines, whereas monocyte trafficking persists, mirroring shifts in lung innate cell composition. We also observe a burst of iMO entry from the blood, likely representing a major source of cells driving monocytic niche formation. Following niche establishment, medLN bacterial burden appears largely sustained by local replication rather than continued trafficking from the lung.

We further uncover a division of labor among cDC subsets in T cell priming. cDC1s, canonically associated with IL-12 production and T cell differentiation^25,73, are indeed critical early mediators of Th1 cell responses. However, as infection progresses, elevated IFN-I signaling induces inf-cDC2s, which become the dominant IL-12-producing population sustaining T cell responses, albeit less robustly than cDC1s. This reveals a nuanced role for IFN-I across organs. While IFN-I promotes pathogenic outcomes in the lung^80,84–88, it appears partially beneficial in medLNs by sustaining effector T cell differentiation. Over time, however, we observe reduced migration of Mtb-infected cDCs from the lungs, accompanied by decreased IFN-I sensing in medLNs, which together likely contribute to waning T cell priming.

Although T cell activation in the medLN declined at later timepoints, it was not abolished. T cells activated at later phases exhibited reduced effector differentiation but retained elevated TCF1 expression, consistent with generation of stem-like cells that sustain long-term surveillance and provide partial immune control^{8,20,23,24,52,62,89}. These findings suggest that T cell response heterogeneity is, in part, temporally imprinted in the medLN, dictated by the myeloid and inflammatory context during priming. Of note, although the timed adoptive transfers employed here do not recapitulate physiological priming kinetics, our observations mirror endogenous polyclonal responses and align with prior studies^4,29.

Our cross-organ single-cell RNA-sequencing atlas enabled us to assess tissue-specific myeloid dynamics during early Mtb infection. In the medLN, infected cells shifted from cDCs and iMOs to MC predominance by 28 dpi, whereas in the lung, infection transitioned from AMs and neutrophils to iMO2-dominated populations with fewer MCs. Infected cells in both tissues upregulated inflammatory transcriptional programs, including TNF and IFN-I/-II signaling, indicating robust immune sensing. Despite this, Mtb persisted, underscoring its ability to withstand classical inflammatory and antimicrobial responses.

In addition to shared inflammatory programs, Mtb-infected cells in the lung and medLN exhibited dynamic metabolic profiles. At 15 dpi, infected monocyte-derived cells in the medLN were enriched for oxidative phosphorylation, whereas by 28 dpi, glycolytic programs predominated in both tissues. This shift toward glycolysis supports rapid energy production and inflammatory responses against pathogens⁹⁰. Infected cells also upregulated hypoxia-associated pathways and acquired tissue-regenerative features resembling non-canonical, pro-inflammatory wound-healing macrophages that promote cellular survival during bacterial infection⁹¹. Together, these features suggest that Mtb persistence in both organs reflects a combination of resilience to inflammatory and hypoxic stress, metabolic adaptation, and engagement of tissue-repair programs within infected myeloid cells.

One factor likely contributing to Mtb persistence is the inability of Mtb-specific T cells to efficiently recognize infected monocyte-derived aggregates despite their close physical proximity. Similar defects in T cell sensing of infected macrophages in liver and lung granulomas^65,92 suggest a conserved limitation across organs. Although Mtb can suppress antigen presentation^93–97, this alone is unlikely to explain the near-complete lack of T cell activation, as infected MCs demonstrated high surface MHC-II protein expression, while other infected populations, such as cDCs, effectively primed T cells. Instead, infected aggregates may establish a locally immunosuppressive environment analogous to tumors^2,33,98. Supporting this, infected medLN MCs upregulated hypoxia-associated programs and inhibitory molecules such as PD-L1 and iNOS, which can suppress TCR signaling^{50,51,99–101}.

A non-mutually exclusive explanation is that infected cDCs and MCs present distinct Mtb peptide repertoires. Immunodominant T cell responses primed by cDCs are typically directed against secreted antigens from actively replicating bacteria^102,103. By contrast, induction of iNOS and hypoxia in infected MCs likely promotes DosR regulon activation⁴¹ and presentation of stress-response antigens^{55,56,104,105}. Accordingly, infected monocytic aggregates may present antigens that are not recognized by cDC-primed T cells^106,107, resulting in ineffective immune surveillance. While recent studies have begun to define Mtb antigens presented in vitro^108,109, which antigens are presented under in vivo stress conditions remains unclear. Notably, infected MCs upregulated non-classical MHC-I molecules, including H2-T23 (Qa-1), the murine homolog of HLA-E. Qa-1/HLA-E–restricted Mtb peptides are largely conserved across species. Loss of Qa-1 increases susceptibility to Mtb infection^110–112, while targeting Mtb–peptide HLA-E complexes can reduce bacterial burden in vitro¹¹³. Thus, vaccines targeting stress-induced and non-classical MHC-restricted epitopes may help enhance immunity against persistent Mtb.

The only approved TB vaccine, BCG, has limited efficacy against pulmonary disease but restricts disseminated Mtb infection in humans³² and mice³¹. We show that BCG immunization limits monocytic niche establishment in the medLN and reduced local Mtb burden without altering bacterial trafficking from the lung or substantially affecting de novo T cell priming, even with markedly reduced bacterial loads. These findings highlight the distinction between mechanisms that initiate T cell responses (cDC-mediated antigen presentation, unchanged by BCG) and those supporting Mtb persistence (infected monocytes that do not activate T cells). Consistent with prior work³¹, BCG immunization also prevented infection in the contralateral lung following ULD infection. Future studies are needed to determine whether this reflects reduced hematogenous dissemination or enhanced Mtb clearance after distal seeding.

Altogether, our work reveals dynamic remodeling of the medLN during Mtb infection, which rapidly transitions from a site of potent adaptive immune induction to an immunologically inert reservoir of infection. These findings have direct implications for TB vaccine design, particularly for strategies targeting lymphoid tissues and monocytic niches that sustain bacterial persistence. Ultimately, effective TB control strategies must account for the dynamic, tissue-specific nature of host-pathogen interactions rather than focusing exclusively on pulmonary infection or early phases of rapid bacterial replication.

Limitations of the study

This study was performed using a limited set of host and pathogen strains, which do not encompass the full spectrum of TB pathogenesis. In some cases of human and NHP infection, lung-draining LNs can develop large necrotic granulomas distinct from those observed here. Future studies, including human post-mortem analyses and diverse animal models, will be needed to more fully characterize the spectrum of LN pathology.

Resource Availability

Lead Contact

Requests for information and resources should be directed to and will be fulfilled by the Lead Contact M.Y.G. (gernermy@uw.edu).

Materials Availability

This study did not generate new reagents.

Data and Code Availability

Transcriptomic data have been deposited in GEO and https://eshamskh-mouse-mtb-innate-atlas.share.connect.posit.cloud and are publicly available from the date of publication. Accession numbers are listed in the Key resources table. Code has been deposited in Zenodo (10.5281/zenodo.20144043).

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
BUV563 anti-mouse CD11c	BD (clone HL3)	749040; RRID: AB_2873434
BUV805 anti-mouse CD86	BD (Clone 2331 (FUN-1))	741946; RRID:AB_2871258
BV421 anti-mouse CD45.2	Biolegend (clone 104)	109832; RRID:AB_2565511
BV480 anti-mouse SiglecF	BD (Clone E50-2440)	746668; RRID:AB_2743940
BV510 anti-mouse SIRPa	BD (Clone P84)	740159; RRID:AB_2739912
BV605 anti-mouse Ly6G	Biolegend (clone A20)	127639; RRID:AB_2565880
BV650 anti-mouse XCR1	Biolegend (clone ZET)	148220; RRID:AB_2566410
BV711 anti-mouse PDL1	BD (Clone 10F.9G2)	563369; RRID:AB_2738163
BV785 anti-mouse B220	Biolegend (clone RA3-6B2)	103246; RRID:AB_11218795
PE anti-mouse CD88	Biolegend (Clone 20/70)	135806; RRID:AB_2243735
PerCP-ef710 anti-mouse CD64	Invitrogen (clone (X54-5/7.1))	46-0641-82; RRID:AB_2735016
PE-Cy7 anti-mouse CD26	Biolegend (Clone H194-112)	137810; RRID:AB_2564312
APC anti-mouse CD45	Biolegend (clone 30-F11)	103112; RRID:AB_312977
AF700 anti-mouse Ly6C	Biolegend (Clone HK1.4)	128024; RRID:AB_10640119
APCeF780 anti-mouse MHCII	Ebio (clone M5/114.15.2)	47532182; RRID:AB_1548783
APC anti-mouse CD88	Biolegend (Clone 20/70)	135807; RRID:AB_10899415
AF488 anti-mouse CCR2	R&D (clone 475301)	FAB5538G; RRID:AB_3651044
PerCP-Cy5.5 anti-mouse CD88	Biolegend (Clone 20/70)	135813; RRID:AB_2750209
BV421 anti-mouse BCL6	BD (Clone K112-91)	563363; RRID:AB_2738159
BUV395 anti-mouse KLRG1	BD (clone 2F1)	740279; RRID:AB_2740018
BUV661 anti-mouse CD8	BD (clone 53-6.7)	750023; RRID:AB_2874241
BUV737 anti-mouse CD3	BD (clone 17A2)	612803; RRID:AB_2870130
BUV805 anti-mouse CD4	BD (clone GK1.5)	612900; RRID:AB_2827960
BV510 anti-mouse CD45.1	Biolegend (clone A20)	110741; RRID:AB_2563378
BV510 anti-mouse CD90.1	Biolegend (clone OX-7)	202535; RRID:AB_2562643
BV605 anti-mouse Ki67	Biolegend (Clone 16A8)	652413; RRID:AB_2562664
BV711 anti-mouse CD44	Biolegend (clone IM7)	103057; RRID:AB_2564214
BV785 anti-mouse Tbet	Biolegend (clone 4B10)	644835;RRID:AB_2721566
Primary anti-mouse IRF4	Biolegend (clone IRF4.3E4)	646402; RRID:AB_2563110
AF555 anti-mouse TCF1	CST (Clone C63D9)	17404S; RRID:AB_2798895
PE-Cy7 anti-mouse PD1	Biolegend (clone RMP1-30)	109110; RRID:AB_572017
AF700 anti-mouse pS6	CST (Clone D57.2.2E)	27036S; RRID:AB_2798937
APC ef780 anti-mouse CD25	Invitrogen (clone PC61.5)	47-0251-82;; RRID:AB_1272179
BV711 anti-mouse MHCII	BD (clone M5/114.15.2)	563414; RRID:AB_2738191
PerCP-Cy5.5 anti-mouse B220	Biolegend (clone RA3-6B2)	103236; RRID:AB_893354
PerCP-Cy5.5 anti-mouse NK1.1	Biolegend (clone PK136)	108728; RRID:AB_2132613
PerCP-Cy5.5 anti-mouse CD11b	Biolegend (clone M1/70)	101228; RRID:AB_893232
PerCP-Cy5.5 anti-mouse CD11c	Biolegend (clone HL3)	117328; RRID:AB_2129641
AF488 anti-mouse TCF1	Cell Signaling (clone C63D9)	6444s; RRID:AB_2797627
APC anti-mouse CD45.2	Biolegend (clone 104)	109814; RRID:AB_389211
Biotin anti-mouse SIRPa	Invitrogen (clone P84)	13-1721-82; RRID:AB_2573352
BV421 anti-mouse CD45.1	Biolegend (clone A20)	110732; RRID:AB_2562563
BV480 anti-mouse CD11c	BD (clone N418)	746392; RRID:AB_2743706
eF506 anti-mouse Ki67	Invitrogen (clone SolA15)	69-5698-82; RRID:AB_2574235
PE anti-mouse CD64	Biolegend (clone X54-5/7.1)	139304; RRID:AB_10612740
BV421 anti-mouse CD64	Biolegend (clone X54-5/7.1)	139309; RRID:AB_2562694
R718 anti-mouse Ki67	BD (clone SolA15)	567292; RRID:AB_2916041
APC anti-mouse IL-12p40/IL-23	Biolegend (clone C15.6)	505206; RRID:AB_315370
R718 anti-mouse B220	BD (clone RA3-6B2)	567381; RRID:AB_2916576
R718 anti-mouse CD3	Biolegend (clone 17A2)	567296; RRID:AB_2916539
APCeF750 anti-mouse CD3	Biolegend (clone 17A2)	100247; RRID:AB_2572117
BV421 anti-mouse CD3	Biolegend (clone 17A2)	100228; RRID:AB_2562553
AF488 anti-mouse CD11c	Biolegend (clone N418)	117311; RRID:AB_389306
APCeF750 anti-mouse B220	Biolegend (clone RA3-6B2)	103259; RRID:AB_2572108
Primary anti-mouse pS6	CST (clone 2F9)	4856S; RRID:AB_2181037
Secondary rabbit anti-mouse AF750	Invitrogen	A-21039; RRID:AB_2535711
APC anti-mouse MerTK	Biolegend (clone 2B10C42)	151507; RRID:AB_2650738
AF488 anti-mouse VCAM1	Biolegend (clone 429 (MVCAM.A))	105710; RRID:AB_493427
BV421 anti-mouse CD68	BD (clone FA-11)	566389; RRID:AB_2739021
AF555 anti-mouse CD68	Abcam (clone EPR20545)	AB280860
Primary anti-mouse MHCII	Biolegend (clone M5/114.15.2)	107602; RRID:AB_313317
FITC anti-mouse Mtb	Abcam	ab20962; RRID:AB_445945
AF488 anti-mouse p120	Santa Cruz (Clone 6H11)	sc23873; RRID:AB_2086394
primary anti-iNOS	Thermo Fisher (clone C-11)	14-5920-82; RRID:AB_2572890
RB704 anti-mouse Ki67	BD (clone SolA15)	570280; RRID:AB_3685635
Rb744 anti-mouse CD3	BD (clone 17A2)	570561; RRID:AB_3685845
Rb780 anti-mouse B220	BD (clone RA3-6B2)	569207; RRID:AB_3684868
MPO	Abcam (CloneEPR20257)	ab208670; RRID:AB_2864724
Bacterial and virus strains
Mycobacterium tuberculosis H37Rv	Obtained from Joel Ernst (NYU)	ATCC 27294
M. tuberculosis H37Rv mCherry	Generated in house	N/A
M. tuberculosis SA161	Provided by Ian Orme	Clinical Isolate from patient
Barcoded M. tuberculosis H37Rv	Obtained from David Sherman (UW)	N/A
BCG-Pasteur	ATCC	ATCC 35734
Chemicals, peptides, and recombinant proteins
Mix-n-Stain CF Dye Antibody Labeling Kits	Biotium	Cat# 92433-92339
DY-395XL-NHS-Ester	Dyomics	295XL-01A
Ovalbumin-488	Invitrogen	O34781
Cell Trace Violet	Invitrogen	C34557
Middlebrook 7H9 broth	BD Difco	Cat# 271310
Middlebrook 7H10 agar	BD Difco	Cat# 262710
OADC supplement	BD Difco	Cat# 212240
Tween-80	Sigma-Aldrich	Cat# 9005-65-6
DNase I	Sigma-Aldrich	Cat# 10104159001
Dispase II	Sigma-Aldrich	Cat# 04942078001
Collagenase P	Sigma-Aldrich	Cat# 11213857001
Isoflurane	Henry Schein	N/A
Zombie UV Fixable Viability Dye	BioLegend	Cat# 423107
BD Cytofix	BD Biosciences	Cat# 554655
OCT Compound	Tissue-Tek, Fischer Scientific	Cat# 23-730-571
Sucrose	Sigma-Aldrich	Cat# 57-50-1
Critical commercial assays
MagniSort Mouse Naïve CD4+ T Cell Isolation Kit	Thermo Fisher Scientific	Cat# 8804-6824-74
CellTrace Violet	Thermo Fisher Scientific	Cat# C34557
eBioscience Intracellular Fixation & Permeabilization Kit	Thermo Fisher Scientific	Cat# 88-8824-00
NEBNext Ultra DNA Library Prep Kit for Illumina	New England Biolabs	Cat#E7103
NEBNext Multiplex Oligos for Illumina	New England Biolabs	Cat#E6440S
AMPure XP beads	Beckman Coulter	Cat# A63880
SMART-Seq v4 Ultra Low Input RNA Kit	Takara Bio	Cat# 88-8824-00
Nextera XT DNA Library Prep Kit	Illumina	Cat# FC-131-1096
LymphoLyte	Cedarlane	Cat# CL5035
Deposited data
Mycobacterium tuberculosis Co-opts Monocytic Niches to Evade Immunity in Lymph Nodes	Gene Expression Omnibus (GEO)	GSE309019
Experimental models: Organisms/strains
Mouse: C57BL/6J	Jackson Laboratory	JAX:000664
Mouse: C3H/FeJ	Jackson Laboratory	JAX:000658
Mouse: B6.129S(C)-Batf3tm1Kmm/J (BATF3−/−)	Jackson Laboratory	JAX:013755
Mouse: B6.SJL-Ptprca Pepcb/BoyJ (CD45.1)	Jackson Laboratory	JAX:002014
Mouse: B6.Cg-Xcr1tm2(HBEGF/Venus)Ksho (XCR1-Venus)	RIKEN BRC	RBRC09485
Mouse: B6.129S6(C)-Zbtb46tm1.1Kmm/J (Zbtb46-GFP)	Provided by Dr. Mark Headley	JAX:027618
Mouse: ESAT-6 TCR Transgenic (C7)	Provided by Dr. Eric Pamer	JAX:035728
Mouse: Ag85B TCR Transgenic (P25)	Previously described	JAX:011005
Mouse: B6(Cg)-Ifnar1tm1.2Ees/J (IFNaR−/−)	Provided by Dr. Michael Gale	JAX:028288
Mouse: ESAT-6 TCR Transgenic (C7) x Bt.SJL (C7 CD45.1)	This study	NA
Software and algorithms
Deposited Code	Zenodo	10.5281/zenodo.20144043
Cell Ranger v7.1.0	10x Genomics	RRID:SCR_017344
Seurat	Satija Lab	RRID:SCR_016341
SingleR	Bioconductor	RRID:SCR_023730
clustifyr	Bioconductor	RRID:SCR_023718
rrvgo	Bioconductor	RRID:SCR_023725
slingshot	Bioconductor	RRID:SCR_017012
STAR aligner v2.4.2a	Dobin et al.	RRID:SCR_004463
HTSeq	Anders et al.	RRID:SCR_005514
Picard Tools	Broad Institute	RRID:SCR_006525
Imaris	Bitplane	RRID:SCR_007370
FlowJo	BD Biosciences	RRID:SCR_008520
Other
Glas-Col Aerosol Infection Chamber	Glas-Col	N/A
gentleMACS Dissociator	Miltenyi Biotec	N/A
Cytek Aurora	Cytek Biosciences	N/A
Leica SP8 Confocal Microscope	Leica Microsystems	N/A
10x Chromium Controller	10x Genomics	N/A
Illumina NextSeq 500	Illumina	N/A
Illumina NovaSeq X	Illumina	N/A
Illumina NextSeq 2000	Illumina	N/A
Glas-Col Aerosol Infection Chamber	Glas-Col	N/A

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Star Methods:

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mice

C57BL/6, C3H/FeJ, B6.129S(C)-Batf3^tm1Kmm/J (BATF3^−/−) and B6.SJL mice were purchased from Jackson Laboratories (Bar Harbor, ME). BATF3^−/− mice have been previously described⁷³. B6.Cg-Xcr1^{tm2(HBEGF/Venus)Ksho} (XCR1-Venus)^72,114 mice were provided by the RIKEN BRC through the National BioResource Project of the MEXT, Japan, and were redistributed by Dr. Mark Headley (Fred Hutchinson Cancer Center) with permission from RIKEN BRC. B6.129S6(C)-Zbtb46^tm1.1Kmm/J (Zbtb46-GFP) mice^38,72 were also kindly provided by Dr. Mark Headley. ESAT-6 TCR Transgenic (C7) mice were provided by Dr. Eric Pamer (Memorial Sloan Kettering Cancer Center, New York, NY) and have been described previously⁶³. Ag85B TCR Transgenic (P25) mice have been previously described⁶⁷. B6(Cg)-Ifnar1^tm1.2Ees/J (IFNaR^−/−) mice were provided by Dr. Michael Gale (University of Washington). Mice with ESAT-6 TCR-Transgenic T cells on a CD45.1 background were generated in house by crossing ESAT-6 TCR Tg mice with B6.SJL mice, with genotypic confirmation through blood phenotyping by flow cytometry.

All mice were housed in specific pathogen free conditions in individually ventilated cages with a maximum density of 5 mice per cage, and negative pressure ventilation and air filtering provided by either Seattle Children’s Research Institute (SCRI) or the University of Washington, South Lake Union campus. Animals were monitored by full-time staff, and given access to food and water ad libitum, with a 12 hour light/dark cycle, maintained at 22–25 degrees Celsius. All mice were of normal health and immune status, and were naïve of treatment, procedure, or invasive testing prior to experiment start. Experiments were performed in compliance of the Animal Care and Use Committee at SCRI or the University of Washington, as appropriate, and all experiments were conducted with sex and age-matched mice (male and female mice between the ages of 8–16 weeks old). The influence of sex was not assessed.

Mycobacterium tuberculosis (Mtb)

For use in murine infections, Mtb H37Rv infections were performed with and without a transformed mCherry reporter plasmid driven by the pMSP12 promoter, as previously described^33,115; Mtb SA161 was provided by Ian Orme (Colorado State University)¹¹⁶. Virulence of these strains are maintained by periodic passage in mice. For the presented datasets, ~50–100 CFU were used in C57Bl/6 background mice, unless otherwise noted (e.g. ULD infections or C3HeB/FeJ - SA161 studies).

Method Details:

Conventional dose Mtb aerosol infections

All infections were done with stocks of either Mtb H37Rv mCherry or Mtb SA161, as previously described¹¹⁷. To perform conventional dose aerosolized infections, mice were placed in a Glas-Col aerosol infection chamber, and ~100 CFU were deposited directly into their lungs. The nebulization cycle was set to 45 minutes followed by 45 minutes of cloud decay. To confirm infectious dose, 2 mice from each infection were immediately sacrificed. Lungs were harvested and homogenates were plated on 7H10 plates for CFU counting after ~3 weeks of growth.

Ultra-low dose Mtb Aerosol Infections

Bar-coded H37Rv Mtb was used for all infections³⁵. Mtb stocks were grown in Middlebrook 7H9 with OADC supplement and 0.05% Tween-80 at 37°C with constant agitation to an OD = 1. Cultures were filtered through a 5μm filter to remove clumps and aliquots were frozen at −80°C. Frozen filtered stocks were thawed and titered side by side with stocks used for conventional dose infection to determine how to dilute the ULD stocks with the goal of leaving 37% of mice uninfected. Mice were placed in a Glas-Col aerosol infection chamber and 5mL of diluted Mtb was injected into the nebulizer of the infection chamber. The nebulization cycle was set to 45 minutes followed by 45 minutes of cloud decay. For analysis of ULD data, uninfected mice were excluded from plots shown.

Barcoded Sequencing

Mice were infected with a pool of 50 bar-coded strains. Sequencing of bacterial bar-codes has been previously described³⁶. Briefly, genomic DNA was pre-amplified with pooled barcoded primers before libraries were prepared with NEBNext Ultra DNA Library Prep Kit for Illumina (New England Biolabs) using the AMPure XP reagent (AgenCourt Bioscience) for size selection and cleanup. The NEBNext Multiplex Oligos for Illumina (New England Biosciences) were used to barcode DNA libraries and enabled multiplexing of 96 libraries per sequencings run. Samples were sequenced using the NextSeq 500 Mid Output v2 kit (Illumina) at the University of Washington Northwest Genomics Center. Read alignment was carried out using a custom processing pipeline that has been previously described.

CFU Determination

Mouse tissues (right lung, left lung, LN) were individually homogenized in an M tube (Miltenyi) containing BS+0.05% Tween-80. Homogenates were diluted and plated onto 7H10 plates. Plates were incubated at 37 degrees Celsius for a minimum of 21 days before CFU enumeration.

Naïve cell isolation and adoptive transfers

Spleens and lymph nodes were harvested from C7 mice and enriched for naïve CD4⁺ T cells using the MagniSort^™ Mouse Naïve CD4+ T cell isolation Kit (Thermo Fisher, #8804–6824-74). For flow cytometry experiments, isolated naïve CD4⁺ T cells were labeled with Cell Trace Violet (Thermo Fisher, #C34557) following manufacturer kit protocol. Cells were confirmed for naïve phenotype and congenic markers, counted, washed and resuspended in sterile PBS, then intravenously injected into recipient mice at indicated timepoints. In all adoptive transfer experiments, host mice received ~10⁶ naïve C7 cells.

OVA-488 Intratracheal Labeling

OVA-488 was used from Thermo Fisher (O34781) and reconstituted to 2 mg/mL according to manufacturer recommendation in sterile PBS. Reagent was diluted fresh 1:2 in PBS on the day of administration, to yield a dose of ~33ug/50 uL. Mice were anesthetized via isoflurane until sufficient to induce deep breathing, after which they were hung up on a rubber band by their teeth to displace their tongue by forceps and administer the reagent. Mice were held in this position until the reagent solution was inhaled and monitored until awakened.

BCG Immunizations

BCG-Pasteur was used for all studies. BCG was cultured in Middlebrook 7H9 with OADC supplement plus 0.05% Tween-80 at 37°C with constant agitation for five days. BCG was back diluted in 7H9 for two days and grown to an OD of 0.2–0.5. Bacteria were diluted in PBS and mice were injected subcutaneously with 10⁶ CFU in 200μL. BCG was diluted and plated on 7H10 plates and plates were counted for CFU after 21 days to confirm dose of immunization. After immunization, mice were rested for 8 weeks prior to Mtb infection.

Single cell suspensions from mouse tissue

Lungs and LNs were harvested from sacrificed mice at indicated timepoints and placed into cold RPMI buffer containing 10% fetal bovine serum (FBS) with DNase I (100 μg/ml; Sigma-Aldrich), Dispase II (800 μg/ml; Sigma-Aldrich), and Collagenase P (200 μg/ml; Sigma-Aldrich). Lungs were homogenized using a gentleMACS dissociator (MilTenyi Biotec), while LNs were individually dissociated by sharp forcep disruption in a 24 well plate. Homogenized samples were incubated for 30 min at 37C with periodic manual disruption, then filtered through a 100μM cell strainer and RBC lysed using RBC lysing buffer (Thermo), then resuspended in FACS buffer (PBS +2.5% FBS +0.1% NaN3). Blood cells were harvested by cardiac puncture, and isolated using LymphoLyte, following manufacturer protocol.

Antibody Staining

Single cell suspensions were first washed in PBS and then incubated with 50 μl Zombie UV viability dye (BioLegend) for 10 min at room temperature in the dark. Viability dye was immediately quenched by the addition of 100 μl of a surface antibody cocktail diluted in 50% FACS buffer/50% 24G2 Fc block buffer using saturating levels of antibodies. Surface staining was performed for 20 min at 4°C. Then, the cells were washed once with FACS buffer and fixed overnight with the eBioscience Intracellular Fixation and Permeabilization kit (Thermo Fisher). The following day, cells were permeabilized with the provided permeabilization buffer, incubated for 60 min at 4°C with 100 μl of an intracellular antibody cocktail diluted 1:100 in permeabilization buffer, and washed with FACS buffer. Cells were analyzed on the Cytek Aurora.

Confocal Microscopy

For confocal imaging, PFA-fixed and sectioned LN tissues were imaged as previously described using a Leica SP8 microscope¹¹⁸. Briefly, isolated LN tissues were fixed using BD Cytofix (BD Biosciences) diluted 1:3 in PBS for 20–24 h at 4°C then dehydrated with 30% sucrose solution for 24–48 h at 4°C. LNs were then embedded in an OCT compound (Tissue-Tek) and stored at −20°C. LNs were sectioned on a Thermo Fisher Scientific Micron HM550 cryostat into 20-μm sections and stained as previously described¹¹⁸. A Leica SP8 tiling confocal microscope equipped with a 40× 1.3 NA oil objective was used for image acquisition. All raw imaging data was processed and analyzed using Imaris (Bitplane).

Histo-cytometry

Histo-cytometry analysis was performed as previously described^26,118. Briefly, multiparameter confocal images were corrected for fluorophore spillover using the built-in Leica Channel Dye Separation module. Single stained controls were acquired using UltraComp eBeads (Invitrogen) that were incubated with fluorescently conjugated antibodies and then mounted on slides with Fluoromount-G slide mounting media (Thermo Fisher Scientific). All images were visualized and analyzed using Imaris (Bitplane). For analysis of myeloid aggregates, a combinatorial myeloid channel was created using CD11c and Sirpα signal with the Imaris XT channel arithmetic module, and this sum myeloid channel was used for myeloid isosurface object creation without cell splitting, but including region growing. Aggregates were then filtered by volume, where surfaces below 10000μM³ were excluded. Mtb surfaces were created by using mCherry signal, where these isosurface objects were used for visual clarity in representative overview images. C7 isosurface objects were created on congenic CD45.1 signal for analysis of Ki67, pS6, and Tbet expression by mean fluorescence intensity. Clustering analysis was performed by creating isosurfaces on CD45.1 or CD45.2 congenic signal without cell splitting. Surfaces were then filtered by volume, and surfaces exceeding 2500 voxels (defining dense clusters representing ~3 or more cells) were included for density analysis. IL-12 surfaces were determined by masking IL-12p40 signal in T cell cluster surfaces, then analyzed for Zbtb46-GFP, SIRPa, or XCR1-Venus expression. Density analyses were performed by calculating number of isosurfaces per LN volume, where LN volume was calculated by manually creating a surface over the entire LN area across Z stacks.

Single cell RNA-sequencing

Single cell suspensions were generated from medLN, blood, and lung samples as described above prior to Mtb infection and at days 0, 15 and 28 post-Mtb infection from pools of 8–10 mice for each suspension/condition. Cells were resuspended in 200 μl MACS buffer (PBS containing 2.5% FBS plus 1 mM EDTA), filtered through a 70 μm filter, and run on a FACS AriaII (BD) sorter. To collect parenchymal cells for single-cell RNA sequencing, AMs (SiglecF⁺ Autofluorescence⁺) were sorted separately into one collection tube to account for autofluorescence in the IV label channel, and all other IV-negative cells were sorted into a separate collection tube. After sorting, AMs were included in respective lung samples (Mtb-infected or uninfected). Mtb-infected and uninfected bystander (mCherry⁺ and mCherry⁻) samples were harvested separately. After one round of washing with ice-cold DPBS, cells were resuspended to 1000 cells/μl in DPBS, and ~8000 cells were inputted into the 10X Genomics pipeline following the manufacturer’s recommendations. After the generation of cDNA following the manufacturer’s protocol, samples were centrifuged through two sequential rounds of 0.2 μm SpinX (Costar) columns to sterilize the sample for removal from the BSL3 facility and subsequent library generation. Libraries were sequenced by Psomagen (Rockville, MD) on an Illumina NovaSeq X platform, with 300M reads per sample.

Alignment and processing of single cell RNAseq data

10X chromium 3′-derived scRNAseq sequence reads were aligned to the 10X Genomics pre-built mouse reference genome mm10–2020-A, assigned to individual cells by barcode, and unique molecular identifiers (UMI) summarized using the 10X Cell Ranger 7.1.0 software package.

The Seurat R package was used for initial QC filtering and integration. First, a filtering step was applied across all samples, requiring all passing cells to have UMIs mapped to at least 500 distinct genes and fewer than 5% of UMIs mapped to mitochondrial genes. Genes detected in fewer than three cells per mouse were excluded from further analysis.

The Seurat integration pipeline¹¹⁹ was then applied to correct for batch effects, normalizing for cell numbers via sctransform for findmarkers within Seurat¹²⁰, and aligning cells across conditions, including all combinations of tissues, timepoints, and infection statuses. UMAP clustering was determined using Seurats FindNeighbor and FindCluster functions using a resolution of 0.5. Cell type annotations were refined using SingleR package and ImmGen Fine reference annotations^121,122 by mapping cluster-specific gene expression profiles to curated immune cell gene signatures, enabling high-resolution classification of immune cell subsets in the single-cell RNA-seq dataset. These annotations were then refined using curated gene set lists and top markers and aligned with our flow cytometry cell type annotations (iMO, MC, cDC). Correlation annotation gene expression analysis and plots were generated using clustifyr package¹²³. Pearson (centered delta of gene expression) correlation was performed on genes excluding common low-informative genes (mt-, ribosomal, hemoglobin, etc.) to best identify correlations and differences between cell types.

To quantify changes in cell type proportion over time, total numbers of cells per sample were calculated and normalized to cells per thousand per sample. Negative-binomial linear models, appropriate for zero-inflated count data, were fit and used to calculate P values using the R glm.nb function.

DEG analysis was performed using sctransform values to normalize for different read depth between samples, using cutoffs of log2FC > 0.5 and adjusted p value (FDR) < 0.05. These DEGs were then run for mSigDB Hallmark and Gene Ontology: Biological Processes (GO:BP) gene set enrichment analysis (GSEA). GO:BP GSEA was then reduced to parent terms using rrvgo package to minimize redundancy of terms¹²⁴. For GO:BP reduced bubble plots, up to 5 leading edge genes were shown based on over-representation across GO terms within each parent term. Pseudotime slingshot analysis was performed on monocyte subsets as shown (S.2F) to determine trajectory, with pMO and iMO1 cell clusters set as origins based on prior annotations and timepoint meta data⁴⁶. Pseudotime scores were then mapped onto the original UMAP space to represent trajectories.

The mouse Mtb Innate Atlas DEG and GSEA data sets are publicly available for exploration and use at https://eshamskh-mouse-mtb-innate-atlas.share.connect.posit.cloud. Primary data can be found at GSE309019.

Bulk RNA sequencing

MedLNs, lungs, and blood of interest were harvested and processed into single cell suspensions, stained with cell surface markers, and sorted (gating schemes in Fig. S6F) from individual mice on days 0, 15, and 25 post-Mtb infection. 200 cells for each condition was sorted on an Aria II (BD Biosciences) directly into reaction buffer from the SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Takara). Reverse transcription was performed followed by PCR amplification to generate full-length amplified cDNA, as previously described²⁷. Sequencing libraries were constructed using the NexteraXT DNA sample preparation kit with unique dual indexes (Illumina) to generate Illumina-compatible barcoded libraries. After the generation of libraries following the manufacturer’s protocol, samples were centrifuged through two sequential rounds of 0.2 μm SpinX (Costar) columns to sterilize the sample for removal from the BSL3 facility. Libraries were pooled and quantified using a Qubit Fluorometer (Life Technologies). Sequencing of pooled libraries was carried out on a NextSeq 2000 sequencer (Illumina) with paired-end 59-base reads using NextSeq P2 sequencing kits (Illumina) with a target depth of 5 million reads per sample. Base calls were processed to FASTQs on BaseSpace (Illumina) and a base call quality-trimming step was applied to remove low-confidence base calls from the ends of reads. The FASTQs were aligned to the GRCm38 mouse reference genome using STAR v.2.4.2a, and gene counts were generated using htseq-count. QC and metrics analysis was performed using the Picard family of tools (v1.134). Gene counts were filtered and normalized from raw counts by trimmed-mean of M values (TMM) normalization and filtered for genes expressed with at least one count per million total reads in at least 10% of the total number of libraries. Further downstream analysis was performed using publicly available RNAseq toolkits. Gene signatures were derived from previous work⁸⁰, and generated alongside heatmap data tables by inputting DEG data into the BIOMEX toolkit¹²⁵.

Primary data can be found at GSE309019.

Quantification and Statistical Analysis

Statistical tests were selected based on appropriate assumptions with respect to data distribution and variance characteristics. Statistical details of experiments can be found in the figure legends. No statistical methods were used to predetermine sample size. For timepoint comparisons the group “Day 28+” is defined as samples from mice at days 28 (N=2 independent experiments), 30 (N=2 independent experiments), or 33 (N=1 independent experiments) post-infection. These groups were combined due to similarity in results at these timepoints. The statistical significance of differences in mean values was determined by the appropriate test, as denoted in the figure legends ****, p ≤ 0.0001; ***, p ≤ 0.001; **,p ≤ 0.01; and *, p ≤ 0.05 or otherwise noted.

Mathematical Modeling Details:

Calculating total number of circulating monocytes

Total concentration of white blood cells in mice is assumed at 5 × 10³ cell/ul of blood based on prior study¹²⁶. Monocytes represent about 1% of WBCs¹²⁶. The total blood volume in mice is dependent on weight and has been calculated to be 7.8 mL per 100 g of body weight¹²⁶. A mouse of 20g weight then has 1.5 mL of blood, and thus, ~0.8 × 10⁶ monocytes in circulation. Based on our measurements (we did not weigh the mice), this is a slight overestimate as we find 0.3 × 10⁵ monocytes in B6 mice (Fig. S7H). The half-life time of murine circulating monocytes is estimated at about 17 hours or removal rate of $d_B=\mathit{\text{ln}}\left(2\right)/(17/24)\approx 1/\text{day}$¹²⁶.

Modeling sIVs of monocytes in circulation and their migration to medLN

To quantify kinetics of monocyte migration from the blood to medLN we developed a mathematical model that tracks labeling of cells with different Abs administered at different times. In our basic model (Fig. S3E), we assume that monocytes in circulation are produced at a rate $p$ from the bone marrow, and that cells in circulation migrate to various tissues and/or die at a rate $d_B$ with $m_{BN}$ being the rate of cell migration from blood to medLN. In our experiments, we only label circulating cells with one Ab, administered at ${\tau }_1$ time prior to sampling time $T$; therefore, in this model we track the dynamics of cells unlabeled and labeled with sIVs Ab in the blood ($B_u$ and $B_l$, respectively) and in the LN ($N_u$ and $N_l$, respectively):

$$\frac{d{B}_u}{\mathit{\text{dt}}}=p-\left(d_B+m_{\mathit{\text{BN}}}\right)B_u-l\left(t\right)B_u,$$ (1)

$$\frac{d{N}_u}{\mathit{\text{dt}}}=m_{\mathit{\text{BN}}}{B}_u-d_N{N}_u,$$ (2)

$$\frac{d{B}_l}{\mathit{\text{dt}}}=l\left(t\right)B_u\left(t\right)-\left(d_B+m_{\mathit{\text{BN}}}\right)B_{l,}$$ (3)

$$\frac{d{N}_l}{\mathit{\text{dt}}}=m_{\mathit{\text{BN}}}{B}_l-d_N{N}_l$$ (4)

where initially all cells are unlabeled and in the simplest model, at the steady state, $B_u\left(0\right)=p/\left(m_{BN}+d_B\right),N_u\left(t0\right)=mp/\left(d_N\left(m_{NB}+d_B\right)\right)$, and $B_l\left(0\right)=N_l\left(0\right)=0$ where time starts from Mtb infection and sampling is done at time T. This is different from a typical set-up of the original sIVs study where time is counted backwards from the time of sampling⁵⁹. In the model, $l\left(t\right)$ is the function describing pulse labeling by sIVs Abs:

$$\begin{array}{c}l\left(t\right)=\left\{\mathit{\text{rl}},\text{if}T+{\tau }_1\le t\le T+{\tau }_1+\Delta {\tau }_1,\right.\\ 0,\text{otherwise}\}\end{array}$$ (5)

where $T+{\tau }_1$ is the time of 1st Ab injection, $\Delta {\text{T}}_1$ is the duration of labeling (typically, 20–30 min given the half-life time of the Ab), and r₁ is the rate of labeling of the cells by the Ab. In our experiments ${\tau }_1=-24\text{h}=-1\text{d}$.

In previous studies involving sIVs, Ab-labeled cells were typically sampled once; however, in our experiments, we injected labeling Ab at several times, typically 1 day prior to tissue sampling, e.g., at $T=0,14$, etc days since Mtb infection. Therefore, the basic model (eqns. (1)–(4)) needs to be solved for every sampling time T. By assuming that labeling of circulating cells with the sIVs Ab occurs rapidly and by knowing the number of circulating cells at time of Ab injection ($T+{\tau }_1$) given in eqn. (1) (with $l\left(t\right)=0$) the dynamics of Ab-labeled cells in circulation and LN is then:

$$\frac{d{B}_l}{\mathit{\text{dt}}}=-m_{\mathit{\text{BN}}}{B}_l$$ (6)

$$\frac{d{N}_l}{\mathit{\text{dt}}}=m_{\mathit{\text{BN}}}{B}_l-d_N{N}_l$$ (7)

where $t\in \left(0,-{\tau }_1\right)$ and $B_l\left(0\right)=B_u\left(T+{\tau }_1\right)$ (from eqn. (1)) and $N_l\left(0\right)=0$. The number of sIVs+ myeloid cells in the LN is $N_l\left(-{\tau }_1\right)$ and can be calculated iteratively for different values of the sampling time $T$.

In our analyses, we found that a model that assumes a constant rate of monocyte migration from the blood to the medLN does not adequately describe the sIVs data; to allow for the rate of monocyte migration to change with time, we adopt a previous approach that assumed that the migration rate is constant in a given time interval but may vary between time intervals:

$$\begin{array}{lll}{m}_{\mathit{\text{BN}}}\left(t\right)=\hfill & \left\{m_{\mathit{\text{BN}}1}\right.,\hfill & \text{if}0\le t<10,\hfill \\ \hfill & m_{\mathit{\text{BN}}2},\hfill & \text{if}10\le t<25,\hfill \\ \hfill & m_{\mathit{\text{BN}}3},\hfill & \left.\text{otherwise}\right\}\hfill \end{array}$$ (8)

The choice of the time intervals where the rate of migration is changing is semi-arbitrary as long as changes occur in between experimental measurements; we chose them to allow for change in the rapid increase in the migration rate from uninfected state to 10–14 days post Mtb infection, and then small decline in the migration rate after 22–25 days post-infection.

Modeling Mtb dissemination from the lung to medLN

To describe the dynamics of Mtb-infected cells in the lung and mediastinal medLN (medLN) we adopted mathematical models recently developed to describe dissemination of Mtb bacilli from the lung to medLN and spleen after a conventional dose infection^{(Cross et. al., In Prep)}. In the simplified version of the model, we track the number of Mtb-infected (myeloid) cells in the lung and medLN assuming that infected cells in the medLN migrate from the lung, and that growth in the number of Mtb-infected cells is described by logistic model:

$$\frac{d{I}_L}{\mathit{\text{dt}}}=r_L\left(1-I_L/K_L\right)-m_{\mathit{\text{LN}}}{I}_L$$ (9)

$$\frac{d{I}_N}{\mathit{\text{dt}}}=r_N\left(1-I_N/K_N\right)I_N\theta \left(I_N-I_{\mathit{\text{Ncr}}}\right)+m_{\mathit{\text{LN}}}{I}_L$$ (10)

where $I_L$ and $I_N$ are the number of Mtb-infected myeloid cells in the lung and medLN at time $t$ after infection, respectively, $r_i$ and $K_i$ denote the rate of Mtb replication and carrying capacity of Mtb in the $i^{\text{th}}$ tissue, $\theta \left(x\right)$ is a shifted Heaviside function defined as $\theta \left(x\right)=0$ if $x<1$ and $\theta \left(x\right)=1$, otherwise, medLN is the rates of Mtb-infected cell migration from the lung to medLN. In the model, infection starts with infected cells in the lung $\left(I_L\left(0\right)=I_0\right)$ but not the medLN $\left(I_N\left(0\right)=0\right)$.

We found that the logistic model (eqns. (9)–(10)) did not accurately describe the data, in part, because the logistic model predicts reaching a steady state over time while the actual number of Mtb-infected cells peaked at 21 days and declined afterwards (S. 7J), most likely due to an immune response. We therefore formulated an alternative model that takes this into account (Fig. S3F).

Adaptive immune responses can control Mtb numbers several weeks post-infection⁶². Because in our studies, we lacked fine-grained measurements of Mtb- specific T cell response dynamics in the lung and medLN (only timed transfers were used), here we propose a simple model previously shown to describe immune dynamics following vaccination or influenza virus infection in humans¹²⁷. We thus assume that immune response $E$ expands exponentially after exposure and suppresses Mtb replication as $f\left(E\right)=1/\left(1+(hE{)}^n\right)$:

$$\frac{d{I}_L}{\mathit{\text{dt}}}=\frac{\left(r_L+{\delta }_L\right)I_L}{1+h^{n}E(t{)}^n}-{\delta }_L{I}_L-m_{\mathit{\text{LN}}}{I}_L$$ (11)

$$\frac{d{I}_N}{\mathit{\text{dt}}}=\left[\frac{\left(r_N+{\delta }_N\right)I_N}{1+h^{n}E(t{)}^n}-{\delta }_N{I}_N\right]\theta \left(I_N-I_{\mathit{\text{Ncr}}}\right)+m_{\mathit{\text{LN}}}{I}_L$$ (12)

where $E=e^{\rho t}$ and for simplicity we set $\rho =0.6/\text{day}$¹²⁸. We define time $T_E=-\text{ln}\left(h\right)/\rho$ as the time when immune response is at half maximal capacity to inhibit growth of the bacteria.

To describe migration of Mtb+ myeloid cells from the lung to medLN that had been stained with OVA-488 we used model-predicted number of Mtb+ cells in the lung (eqn. (11)) at time $T+{\tau }_o$ where ${\tau }_o=-24\text{h}$ is the time of OVA-488 administration prior to sampling at time $T$ :

where $t\in \left(0,-{\tau }_0\right)$ and ${\tau }_o=-24\text{h}$ is the time of labeling of lung-localized myeloid cells with OVA-488 and ${I_L}^o(0)=I_L\left(T+{\tau }_o\right)$ (from eqn. (11)) and ${I_N}^o\left(0\right)=0$. The number of Mtb+ myeloid cells that had migrated from the lung to medLN in $-{\tau }_o$ time is ${I_N}^o\left(-{\tau }_o\right)$ and can be calculated iteratively for different values of the sampling time $T$.

$$\frac{d{I}_L^o}{\mathit{\text{dt}}}=-m_{\mathit{\text{LN}}}{I}_L^o$$ (13)

$$\frac{d{I}_N^o}{\mathit{\text{dt}}}=m_{\mathit{\text{LN}}}{I}_L^o-{\delta }_N{I}_N^o$$ (14)

In fitting models to data on migration of Mtb+ myeloid cells, we found that models with a constant rate of migration of cells from the lung to the medLN generally did not fit the data very accurately. This was in part because the ratio of lung-to-medLN migrating myeloid cells to total number of myeloid cells in the lung was non-linearly dependent on the time since infection. Therefore, we assumed that the rate of migration of Mtb+ cells from the lung to medLN change with time since infection. Similarly to our previous work¹²⁸, we allow for the migration rate to be constant in a given time interval ($t_i,t_{i+1}$) but change between different time intervals that are determined by times when experimental measurements were taken:

$$\begin{array}{ll}{m}_{\mathit{\text{LN}}}\left(t\right)=& \{m_{\mathit{\text{LN}}1},0\le t<15;\\ & m_{\mathit{\text{LN}}2},15\le t<22;\\ & m_{\mathit{\text{LN}}3},22\le t<30;\\ & \left.m_{\mathit{\text{LN}}4},\text{otherwise}\right\}\end{array}$$ (15)

Fitting models to data

To fit models to data, we log-transformed the data and model predictions and minimized the sum of squared residuals (SSR). Data or model predictions on the number of cells that were below a limit of 1 we set to be equal to this limit. We compared alternative models using either F-test (for nested models) or AIC (for non-nested models). In most cases we attempted to fit the least number of parameters, e.g., by fixing some parameters to values found in the literature (e.g., the rate of monocyte migration/death from the blood to other tissues). Note that because we assumed that the rates of cell migration are constant in a given time period but change between time periods, the model predicted changes in the number of myeloid cells migrating from the blood (Fig. S.7H) or the lung (Fig. S.7J) or to medLN to appear oscillatory. Modeling the migration rate to be dependent on time would require additional parameters and was not pursued.

Supplementary Material

Supplemental References

Document S1. Figures S.1–S.7, Tables S.1–S.3.

Data S.1. DEG timepoint comparisons, related to Figure 2.

Data S.2. Pathway timepoint comparisons, related to Figure 2.

Highlights:

• Mtb trafficking is mediated by inflammatory monocytes and dendritic cells

• cDC1 and inf-cDC2 subsets drive transient Th1 cell priming in mediastinal lymph nodes

• Monocyte-derived niches evade T cell recognition and enable Mtb persistence

• BCG vaccination restricts niche formation and reduces lymph node Mtb burden