This study was performed using buffy coats from healthy donors to the blood banks of the Instituto Nacional de Enfermedades Respiratorias Ismael Cosío Villegas (INER) and Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán. Overall, we included 16 healthy subjects. This study was approved by the institutional review boards, which waived informed consent on grounds of donor anonymity and assigned the number C43-22.

Peripheral blood mononuclear cells were obtained using a Lymphoprep gradient (Axis Shield, San Jose, CA, USA). Monocytes were then isolated by positive selection using CD14+ magnetic bead technology (Miltenyi Biotec, CA, USA) conjugated to anti-CD14 antibodies. Monocytes in suspension were adjusted to 1×10⁶ cells/ml in RPMI-1640 (Lonza, Walkersville, MD), supplemented with 200 nM L-glutamine (Lonza, Walkersville, MD), 5 µg/L gentamicin (Lonza, Walkersville, MD), and 10% human serum (Valley Biomedical, Winchester, VA 22602, USA), herein referred to as complete medium. Human monocytes were cultured for 7 days to generate monocyte-derived macrophages (MDMs).

M. bovis BCG (ATCC 35743) and M. tuberculosis H37Ra (ATCC 25177) were grown at 37°C in 7H9 broth (Difco, Detroit, MI) supplemented with ADC Enrichment (Beckton Dickinson, BD, San Jose, CA), 0.2% glycerol, and 0.05% Tween-80 (Sigma-Aldrich, St. Louis, MO USA), and quantitated by the colony forming units method using solid Middlebrook 7H10 medium (Difco) supplemented with 0.5% glycerol, OADC Enrichment (BD) and 0.05% Tween-80. Aliquots were stored at -80°C until use.

Bacterial lysate Pulmonarom PML (Sanofi-Aventis, México) and whole-bacteria immunotherapeutic IPI Bacterial Suspension (ASAC Pharmaceutical Immunology, ASAC S.A., México) were used in their commercial oral formulations, and the CFU counts were used as reported by the manufacturers in their technical sheets.

Monocytes were seeded at 1×10⁶ cells/mL in complete medium, and cultured in the presence of Bacillus Calmette-Guérin (BCG) at an MOI 1, 100 ng/mL lipopolysaccharide (LPS, Sigma-Aldrich, St. Louis, MO USA), bacterial lysate (BL) at 1×10⁸ CFU/ml, and bacterial suspension (BS) at 1×10⁶ CFU/ml, incubated for 24 hours at 37°C in 5% CO₂. After 24 hours, supernatants were removed and stored at –20°C. The plate was washed twice and replenished with complete medium to differentiate monocyte-derived macrophages (MDMs) for an additional 6 days. On day 6, culture supernatants were collected and stored at –20°C, and the medium was replenished. Secondary stimulation with M. tuberculosis H37Ra at an MOI of 5, LPS, or both was carried out for an additional 24 hours, after which the culture supernatants were collected (day 7) and stored at –20 °C until use.

Untrained control macrophages were incubated in complete medium for 6 days and subsequently infected with M. tuberculosis H37Ra at an MOI of 5, with LPS, or with both for an additional 24 hours.

In selected experiments, we used 1mM of methyltransferase inhibitor 5'deoxy-5'-(methylthio) adenosine (MTA, brand, country), which can prevent the induction of a trained phenotype (24), for 1 hour before exposing the monocytes to the training stimuli to prevent epigenetic changes. MTA was replenished after the training stimulus was washed. The cells were cultured and infected with M. tuberculosis H37Ra as described above.

The levels of pro-inflammatory cytokines IL-12p70, TNF-α, IL-10, IL-6, IL-1β, and IL-8 were measured in the supernatants at 24 hours, day 6, and day 7 using the Cytometric Bead Array (CBA) kit from BD (San José, CA, USA), according to the manufacturer’s instructions. Samples were analyzed using a FACS CANTO II flow cytometer (BD, San José, CA, USA), and results were processed with FCAP software (BD, San José, CA, USA). Preliminary pilot measurements of IL-1β were determined by ELISA following the manufacturer’s instructions (Mabtech, Cincinnati, OH, USA), and TNFα was measured as previously reported (25).

MDMs trained for 7 days as described above were lysed with RLT buffer (Qiagen, Germantown, Maryland, USA) for total RNA extraction using the RNeasy Microkit (Qiagen) according to the manufacturer’s instructions. RNA concentration was subsequently measured using a Nanodrop spectrophotometer (Thermo Fisher Scientific, Millersburg, Pennsylvania, USA), and samples were stored at -80°C until use. Complementary DNA (cDNA) was then synthesized using the SuperScript First-Strand Synthesis System kit (Invitrogen, Millersburg, Pennsylvania, USA), according to the manufacturer’s protocol. Gene expression levels of metabolic enzymes PFKP, HK2, GLUD1, and GLS, as well as genes associated with antimicrobial activity such as CAMP, NPC2, MAP1LC3, and ATG16L1, were analyzed by real-time PCR using TaqMan technology and a QuantStudio 12 Flex system (Applied Biosystems, Waltham, USA). Expression differences were calculated using the 2^−ΔΔCt method, with B2M as the housekeeping gene. Details of the TaqMan assays appear in Supplementary Table 1.

Trained MDMs were detached using ice-cold PBS (Corning™Mediatech™ Glendale, Arizona, USA), transferred to polypropylene tubes, washed twice with cold PBS, and centrifuged at 1500 rpm for 5 minutes. The supernatant was discarded, and 1 µL of viability dye AquaDye (Invitrogen) was added and incubated at room temperature in the dark for 20 minutes. After incubation, cells were washed, centrifuged, and the supernatant was discarded. The cell pellet was then resuspended in 2 µL of SBF (Invitrogen) plus 50 µL of Brilliant Stain Buffer, followed by the addition of the antibody cocktail: CXCR4-PE-Cy5 (Biolegend, San Diego, CA, USA) CD3-BV510, CD14 PE-CF594, CD16-APC-H7, CD163-BUV395, TLR2-BUV615, TLR4-PE, TLR9-APC, CD64-PE-Cy7, HLA-DR R718, CD36-BV605, CD80-BV421, CD86-BV650, CCR2-BV711, CD209-FITC, and CD206-BV786 (Becton Dickinson, BD, San José, CA, USA), additional details in Supplementary Table 1). Cells were incubated for 30 minutes at 4°C in the dark. After staining, cells were washed twice with cold PBS and resuspended in a final volume of 200 µL PBS.

Single-stained controls for spectral unmixing were prepared using BD CompBeads Plus (BD) following the manufacturer’s instructions. All samples were acquired using the FACS Discover S8 spectral flow cytometer with integrated CellView and FACS Chorus software (BD). Spectral unmixing was carried out using the single-stained controls and unstained samples, allowing for autofluorescence subtraction and deconvolution of overlapping emission spectra. Data analysis was performed using FlowJo v10.10 (BD). Dimensionality reduction analysis was conducted using t-distributed stochastic neighbor embedding (t-SNE) and FlowSOM for unsupervised quantitative clustering.

The macrophage’s phenotype was only described, not compared. Multiple comparisons were performed using one-way ANOVA followed by Dunnett’s post hoc test. Relative gene expression was assessed with a one-sample Wilcoxon test. Analyses were conducted using GraphPad Prism 8.0, considering p-values < 0.05 statistically significant.

Given that BS and BL were liquid pharmaceutical formulations sold for oral administration, we first evaluated their ability to stimulate human monocytes in vitro. A pilot study was conducted to identify concentrations that elicited immunostimulatory activity without cytotoxicity (Supplementary Figure 1). To further characterize their immunomodulatory potential, we quantified the production of IL-12p70, TNFα, IL-10, IL-6, IL-8, and IL-1β in monocytes stimulated for 24 hours with BS (1 × 10⁶ CFU/mL) or BL (1 × 10⁸ CFU/mL). Lipopolysaccharide (LPS) and M. bovis BCG were included as positive controls, while unstimulated cells cultured in medium served as the negative control.

BS stimulation significantly increased the monocytes’ production of TNF-α and IL-6. In contrast, BL stimulation induced significant levels of IL-6, IL-1β, and IL-10, suggesting the activation of distinct immunological pathways. Surprisingly, all stimulators reduced the expression of IL-8. As expected, LPS elicited robust production of all measured cytokines, whereas BCG infection selectively induced proinflammatory cytokines (Figure 1).

Next, we investigated whether BS and BL induce trained immunity in monocytes, enhancing responses to M. tuberculosis infection. We used a training protocol previously reported for BCG training (24), which involved applying a primary stimulus to monocytes, followed by a washout and a 6-day resting phase (Figure 2A). On day 6, the cells, now differentiated into MDMs, received a secondary stimulus, followed by a 24-hour incubation period. TNF-α production of trained MDMs was measured at day 7 and reported as fold change relative to untrained MDMs’ production (Figure 2B). Untrained MDMs infected with M. tuberculosis were used as the reference henceforth. MDMs trained with BCG exhibited enhanced TNF-α production in response to M. tuberculosis infection, consistent with trained-like signatures. MDMs trained with BS or BCG and infected with M. tuberculosis had the highest response, although it did not reach significance. Training with BL did not affect the TNF-α response (Figure 2B).

Because we did not know whether the large amounts of PAMPs present in BS and BL, many of which originated from gram-negative bacteria, would result in tolerance rather than activation, we included lipopolysaccharide (LPS), a known inducer of tolerance, for comparative purposes. At 24 hours, LPS induced robust TNF-α production, which was lost after the 6-day rest. The LPS-prestimulated MDMs were unresponsive to a secondary stimulation with M. tuberculosis, consistent with tolerance. Regular MDMs produced the expected levels of TNFα when stimulated with LPS for 24 h or infected with M. tuberculosis and then stimulated with LPS. This indicates that LPS produced tolerized macrophages during the 6-day training protocol rather than inducing trained immunity (Figure 2C, Supplementary Figure 2).

To confirm trained immunity and its metabolic regulation, we assessed expression of enzymes involved in glycolysis (Hexokinase 2, HK2, and Phosphofructokinase, PFKP) and glutamine metabolism (glutaminase, GLS, and glutamate dehydrogenase 1, GLUD1) (Figure 2D). Significant increases in PFKP and GLS gene expression were observed with BS and BL treatment compared to untrained MDMs. BL also significantly upregulated GLUD1. No significant differences in HK2 expression were observed. These findings suggest that BS and BL likely induce trained immunity, primarily by enhancing PFKP and glutamine metabolism. BCG tended to increase expression of all four genes, but this did not reach significance; however, BS and BL responses were higher than those during training with BCG, suggesting their role in the metabolic reprogramming underlying training-associated transcriptional and phenotypic changes (Figure 2E).

The morphological and functional diversity of macrophages plays a key role in either controlling or facilitating the dissemination of M. tuberculosis. In simplified terms, M1 macrophages are pro-inflammatory and can eliminate or contain intracellular pathogens, whereas M2 macrophages tend to support intracellular persistence and are often associated with tissue repair and immunoregulation (26). Recent evidence suggests that the phenotype of trained macrophages is strongly influenced by the nature of the training stimulus (27, 28). To explore this, we used spectral flow cytometry combined with dimensionality reduction techniques to characterize macrophage phenotypes in our model.

We defined M1 macrophages by surface expression of CD80, CD86, TLR2, TLR4, HLA-DR, CD64, CCR2, CD14, and CD16, and M2 macrophages by expression of CD206, CD209, CD163, CD36, CXCR4, TLR9, and CD16 (Figure 3A). Cells from trained and untrained macrophages infected with M. tuberculosis were collected from six individuals and concatenated for analysis (see gating strategy in Supplementary Figure 3). We then performed t-SNE on day 7, 24 h post-infection, to visualize the patterns of the 15 surface markers’ expression (Figure 3B). The individual expression profiles for each marker are shown in Supplementary Figure 4.

The t-SNE analysis revealed distinct clustering profiles among untrained MDMs depending on the condition: uninfected, stimulated with LPS, or infected with M. tuberculosis (Figure 3C). In contrast, trained MDMs, whether primed with BCG, BS, or BL, displayed more homogeneous clustering patterns after M. tuberculosis infection, suggesting convergence toward a typical trained phenotype, although each condition retained unique characteristics (Figure 3D). Interestingly, macrophages tolerized with LPS showed an expression pattern that partially overlapped with that of BL-trained MDMs, suggesting that differences between immune tolerance and certain trained states may be in the functionality of signaling pathways (Figure 3E). The violins in the spectral viewer showed that the relative expression of the markers of interest differ between trained and untrained MDMs infected with M. tuberculosis, suggesting that the differences in the expression profile may be attributed to the training (Figure 3F). For instance, in untrained MDMs, M. tuberculosis infection shifted the MDMs general phenotype toward a mixed phenotype with M1 (CD86^hiHLA-DR^hiCD14^hiCD64^hiCXCR4^hi) and M2 (CD163^hiCD36^hi) markers highly expressed, consistent with a flexible state. However, BCG-, BS-, and BL-trained MDMs exhibited the highest expression of M1 markers.

Because MDMs are expected to be a mixture of phenotypically different subpopulations, we performed unsupervised clustering using FlowSOM, followed by metacluster annotation with Cluster Explorer to further characterize phenotypic diversity among macrophages (Figures 4A, B). We identified eight distinct macrophage clusters across all experimental conditions using expression data from 15 surface markers associated with M1 and M2 polarization (Figure 4C). Figure 4D shows the expression level for each marker in each cluster. In general, TLR9 was the lowest expressed marker, followed by CD163. CD86 and HLA-DR were highly expressed, whereas CD14, CD36, CD64, CD80, CCR2, and CXCR4 were differentially expressed. The most abundant macrophage subpopulation (cluster #7, Figure 4C) exhibited high levels of CD86, CD64, and HLA-DR, as well as medium levels of CD80, CD36, CD14, and TLR2, all of which indicate an M1 phenotype.

Next, we investigated which macrophage subpopulation was present across experimental conditions (Figure 4E). The color bar in the heatmap indicates the frequency of each cluster across conditions. Untrained MDMs infected with M. tuberculosis (Med+Mtb) had a higher abundance of clusters 6, 7, and 8, which exhibit a mixed phenotype with high expression of both M1 and M2 markers. Those clusters were less abundant in trained MDMs. BS-trained macrophages showed selective enrichment in M1 phenotype clusters with high expression of CD80, CD86, HLA-DR, TLR2, and CD16, and low expression of CD163, TLR9, CCR2, and TLR4 (clusters 1, 2, 3, and 5). In fact, BS-trained macrophages exhibited a greater diversity of macrophage subpopulations. BL-trained macrophages were mainly enriched in clusters 2 and 5.

To further characterize the trained macrophages’ phenotype, we evaluated the production of IL-1β, TNF-α, IL-6, IL-12p70, IL-8, and IL-10 cytokines in the culture supernatants 24 h after M. tuberculosis infection. Here, macrophages were incubated in the absence or presence of MTA, an inhibitor of histone 3 lysine 4 trimethylation (24), thereby inhibiting epigenetic reprogramming before the training stimulus was administered. BS- and BL-trained MDMs produced higher IL-1β and TNF-α than BCG-trained macrophages, supporting their capacity to induce a trained proinflammatory state, consistent with the M1 phenotype. This phenotype was significantly inhibited by preincubation with MTA before training (Figure 5), highlighting the dependence of epigenetic reprogramming for training. Interestingly, training did not affect IL-6, IL-8, and IL-10 production, suggesting other responses associated with M. tuberculosis infection were not triggered. Levels of IL-12p70 were negligible, likely due to the absence of lymphocytes in the culture (data not shown).

Because M1 macrophages possess antimicrobial properties, we measured the expression of genes involved in macrophage antimicrobial responses (NPC2, CAMP, MAP1LC3, and ATG16L1) in trained MDMs. In untrained or LPS-tolerized MDMs, the infection alone downregulated the expression of all four genes relative to uninfected MDMs (Figures 6A–C). We determined gene expression relative to uninfected, untrained MDMs to assess the effect of M. tuberculosis infection in trained MDMs. BS and BL training restored macrophage ability to upregulate NPC2 and the autophagy-related genes MAP1LC3 and ATG16L1 upon Mtb infection (Figures 6D, F). BCG-trained MDMs only upregulated the expression of the autophagy-related genes upon M. tuberculosis infection (Figure 6F). CAMP expression was downregulated upon M. tuberculosis infection, regardless of the immunomodulation status of the MDMs.