Mtb H37Rv was obtained from BEI Resources (USA) and propagated in the biosafety level 3 (BSL-3) facilities at the University of Texas Health Science Center at Tyler (UTHSCT). It was cultured in 7H9 broth with 10% OADC supplement following standard Mtb culture procedures (30). After 7 days of growth, the bacteria were collected and subjected to sonication three times, at an amplitude of 38%, for 10 seconds/each, with a 5-second interval, followed by low-speed centrifugation (1,100 RPM). Bacteria were diluted to an optical density (OD) value of ≈ 1 in sterile NaCl 0.9% and aliquots were made and frozen at -80°C to be used as inoculum. Two weeks later, one aliquot was thawed, and the bacterial content was evaluated by plating ten-fold serial dilutions in 7H10 agar, supplemented with OADC. After 3 weeks of incubation, the colony forming units (CFU) per mL were calculated.

HIV-1 BaL strain was obtained from NIH AIDS Reagent Program, also prepared in the BSL-3 facilities at UTHSCT, following standard procedures (31). Briefly, frozen human PBMCs (STEMCELL Technologies, Vancouver, Canada) were thawed and seeded in a 75 cm² flask at a concentration of 5 × 10⁶ cells/mL in RPMI 1640 media (Corning Inc., Corning, NY) supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, 1 µg/ml of PHA and 2 µg/ml polybrene (MilliporeSigma, Burlington, MA). After 3 days of stimulation, 4 × 10⁷ cells were centrifuged and infected with HIV-1 BaL using an MOI (multiplicity of infection) of 0.1 (4 × 10⁶ TCID₅₀) in two adsorption cycles. Following the second adsorption cycle, the cells were seeded in two 75 cm² flasks with 30 ml of media supplemented with FBS, antibiotics, and human IL-2 (20 Units/ml). Cell culture supernatant was collected every three days, with fresh media being added, until day 21 of culture and stored at -80°C. A small aliquot from each collection will be used to titrate the virus using quantitative RT-PCR.

All animal procedures were approved by the UTHSCT Institutional Animal Care and Use Committee (IACUC) (Protocol #707). NOD.Cg-Prkdc^scid Il2rg^tm1Wjl Tg(CMV-IL3,CSF2,KITLG)1Eav/MloySzJ (NSG-SGM3) mice were purchased from The Jackson laboratory (Bar Harbor, ME) and bred in the Vivarium facilities at UTHSCT. Pups were weaned at 21 days after birth and, 1–3 weeks after that, they were irradiated at a dose of 100 cgy/mouse, followed by intravenous injection with 2 × 10⁵ CD34⁺ stem cells/mouse at 12 h post-irradiation. Humanization was monitored starting at 12 weeks after stem cell transplantation and again at 14 and 16 weeks. For this purpose, blood was drawn from the submandibular vein (100–150 µl, based on animal weight) and PBMCs were collected through density gradient centrifugation using Ficoll Paque (Cytiva, Marlborough, MA). After erythrocyte lysis, the PBMC from each animal were stained for human (hu) and mouse (mo) hematopoietic cell surface marker (CD45⁺), as well as lymphocytic and myeloid markers. Animals that showed a positive huCD45⁺/moCD45⁺ ratio, accompanied by differentiation of various immune cell populations, were selected for experimental infection.

Mice were randomly divided into four experimental groups: Uninfected (n=5), HIV-infected (n=8), Mtb-infected (n=8) and HIV/Mtb co-infected (n=7). Mtb infection was performed using aerosolized Mtb H37Rv through a Madison chamber, as previously described (32), using an infection dose of 100 CFU/mouse. Three additional mice were included in the Madison chamber at the time of infection and were euthanized 24 hours after infection. The lungs were collected, macerated and plated on 7H10 agar to confirm the initial bacterial implantation (33).

One day after Mtb infection, the mice for the HIV alone and HIV/Mtb co-infection groups were subjected to intraperitoneal (IP) inoculation with 10⁵ TCID₅₀ of HIV_BaL. Blood samples from all experimental groups were collected on the day of infection and at 15-, 28- and 35-days post infection (dpi). Serum samples from all the animals were separated and stored at -80 °C until further use. PBMCs were isolated and stained for flow cytometry analysis. At 35 dpi, the animals were terminally anesthetized, using a Ketamine/Xylazine mixture, in order to perform computed tomography (CT) scan and pulmonary function (PF) tests. Afterwards, the animals were euthanized and whole blood samples were collected through cardiac punction. During necropsy, lung and spleen samples were collected and macerated through a 70 μM cell strainer (Thermo Fisher scientific) in a final volume of 2 ml of PBS. Serial ten-fold dilutions of the organ macerates were plated in 7H10 agar, supplemented with OADC, to assess the bacterial load. The remaining volume of lung and spleen macerates were stored at -80°C for further analysis.

For each experimental group, lung samples from one animal were selected for histopathological analysis and, therefore, not subjected to maceration and bacterial culture. Lungs were filled with 10% formalin, before being removed from the animal, and stored in the same media after the necropsy (34, 35). Sample processing and Hematoxylin-Eosin (HE) staining was conducted at the histopathology core of UT southwestern.

Mice were intraperitoneally injected with ketamine/xylazine (100 mg/kg Ketamine, 20 mg/kg Xylazine). Once the correct anesthetic plane was achieved, the mice were intubated with a sterile, 20-gauge intravenous cannula through the vocal cords into the trachea. Following intubation, anesthesia was maintained using isoflurane.

Pulmonary function test was performed using the FlexiVent system (SCIREQ, Tempe, AZ). This system artificially ventilates the animal with short pressure-volume measurement maneuvers, and then measures the resulting expiratory pressure and volume changes as a function of time against a physiological positive end expiratory pressure of 2.5–5 cmH₂O (36). Elastance (Ers), compliance (Crs), and total lung resistance (Rrs) have been widely used to assess pulmonary function when using the FlexiVent system (36). Therefore, we measured these three parameters for each mouse through the snapshot perturbation method, as previously described (37). Measurements were performed in triplicates for each animal, using the FlexiVent system, with a tidal volume of 30 mL/kg at a frequency of 150 breaths/min against 2–3 cm H2O positive end-expiratory pressure.

After PF testing, the mice were subjected to CT scans for the measurements of lung volume, using the Explore Locus Micro-CT Scanner (General Electric, GE Healthcare, Wauwatosa, WI). CT scans were performed during full inspiration and at a resolution of 93 μm. Lung volumes were calculated from lung renditions collected at full inspiration. Microview software 2.2 (http://microview.sourceforge.net) was used to analyze lung volumes and render three-dimensional images.

Serum samples from all experimental groups were extracted using the NucleoSpin RNA isolation kit (Macherey-Nagel, Allentown, PA). Following viral RNA extraction, samples were evaluated using RT-qPCR to determine the viral RNA load in each animal (38). Control standards (obtained from NIH AIDS Reagent Program) with known quantities of HIV-1 genome copies were used as amplification controls, as well as to stablish a standard curve that was used to determine the viral RNA load, based on the cycle threshold (Ct) value.

Flow cytometry was performed using the PBMCs from all experimental animals at the specified sampling timepoints. In all cases, the PBMCs isolated from each animal were divided into two wells of a 96-well U-shaped bottom plate (Corning Inc., Corning, NY), used for staining with two separate flow cytometry panels. Cells were washed and inoculated with Fc block (Biolegend, San Diego, CA) at 4°C for 20 minutes, followed by another wash. Afterwards, cells were incubated with fluorescence-conjugated monoclonal antibodies. For the first flow cytometry panel, cells were incubated with antibodies against the following human surface markers: Alexa Fluor™ 421-CD45 (Cat# 368522), FITC-CD3 (Cat# 300406), APC-CD4 (Cat# 317416), PE-CD8 (Cat# 344706), PerCP-CD56 (362526), Alexa Fluor™ 510-CD19 (Cat# 302242) (All antibodies were purchased from Biolegend Inc., San Diego, CA). For the second flow cytometry panel, the antibodies against human cell surface markers were as follows: Alexa Fluor™ 421-CD45 (Cat# 368522), Alexa Fluor™ 510-CD86 (Cat# 305432), APC-CD11b (Cat# 301310), PE-CD11c (Cat# 301606), PerCP-HLA-DR (Cat# 307628), Alexa Fluor™ 700-CD14 (Cat# 325614) (Biolegend Inc., San Diego, CA). Additionally, for the second panel, the cells were also incubated with an FITC-labelled antibody against moCD45. After staining, the cells were washed and fixed for 1 hour, followed by another wash. Flow cytometry was performed using the Attune NxT flow cytometer (Invitrogen, Waltham, MA), including the corresponding isotype controls for each antibody. Analysis was conducted with the FlowJo software v10.6.1 (BD life sciences), using the isotype controls as guidelines for gating.

Paraffin-embedded lung sections were used for immunofluorescent staining against human immune cell subsets (39). Samples were deparaffined by submerging the slides in Xylene (Fisher bioreagents), followed by sequentially lower concentrations of ethanol. Afterwards, antigen retrieval and blocking of non-specific binding were performed, using 10mM sodium citrate buffer and PBS with 0.4% triton and 5% FBS, respectively. Primary antibody incubation was conducted overnight at 4°C with human-CD68 monoclonal antibody (cat. No. 14–0688-82, Invitrogen) and CD19 Rabbit polyclonal antibody (cat. No. 27949–1-AP, Proteintech, Rosemont, USA), diluted in PBS + 0.4% triton + 1% FBS at the recommended dilutions. The following day, samples were incubated for 2 hours at room temperature with goat anti-mouse IgG1-Alexa Fluor™ 568 (cat. No. A21124, Invitrogen) and goat anti-rabbit IgG-Alexa Fluor™ 488 (cat. No. A11008, Invitrogen), at the recommended dilutions. The slides were mounted using DAPI-supplemented mounting medium (Abcam, Cambridge, UK) and images were captured with a LionheartLX automated microscope (Biotek, Winoovski, VT). Images were processed with the GEN5 software version 3.09 (Biotek) and the ImageJ software (NIH).

The cytokine profile in lung and spleen tissue macerate, as well as serum samples at 35 dpi, from all experimental groups were evaluated in duplicates using the Bio-Plex Pro™ Human Cytokine panel (Bio-Rad, Hercules, CA), according to the manufacturer’s instructions. Briefly, 50 µL of filtered tissue homogenate, or 1:4 diluted serum, were dispensed in a 96-well plate containing magnetic beads conjugated with antibodies for the detection of 27 different cytokines. Following incubation with detection antibodies and streptavidin-PE, the samples were analyzed in the Bio-Plex MAGPIX multiplex reader (Bio-Rad Laboratories Inc., CA). A regression curve, based on the values obtained from a set of standard dilutions, was used to convert the fluorescence values reported by the machine into cytokine concentrations (expressed as pg/mL).

The 27 cytokines and chemokines reported by the Bio-Plex Pro™ Human Cytokine panel were: Basic FGF, Eotaxin, G-CSF, GM-CSF, IFN-γ, IL-1β, IL-1Ra, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-10, IL-12, IL-13, IL-15, IL-17, IP-10, MCP-1, MIP-1α, MIP-1β, PDGF-BB, RANTES, TNF-α and VEGF.

Whole blood sample was collected from mice in all the experimental groups at the end of the study and plasma was separated through centrifugation. The samples were processed for collection of the metabolite pellet as follows: 50 μl of plasma were mixed with 950 μl of 80% ice-cold methanol, followed by centrifugation at >20.000 G for 15 minutes in a refrigerated centrifuge. Afterwards, the supernatant was transferred to a new tube and vacuum dried, using no heat. The metabolite pellet was analyzed at the metabolomic core facility at the Children’s Medical Center Research Institute at University of Texas Southwestern Medical Center (Dallas, TX, USA) using liquid chromatography–mass spectrometry (LC-MS), as previously described (40).

Statistical analysis of metabolome profiles was performed in R environment (R version 4.1.0). Raw abundance values of metabolites were used as input for statistical analysis. The raw data was log2 transformed and normalized across the samples using ‘limma’ package (41) by cyclically applying fast linear loess normalization with a 0.3 span of loess smoothing window and 10 iterations wherein each sample was normalized to pseudo-reference sample which was computed by averaging all samples. Principal components analysis was performed using ‘PCAtools’ package. Orthogonal partial least squares discriminant analysis (OPLS-DA) was performed and variable importance on projection (VIP) score were computed using ‘ropls’ package. VIP score of >1 is considered for feature selection. Hierarchical clustering was performed on normalized data after univariate scaling. Hierarchical clustering was performed using correlation to calculate clustering distance with averaging method for clustering. Differentially abundant metabolites (DAMs) were identified using student t test. The correlation between metabolite abundances and Mtb or HIV loads were analyzed using Pearson correlation method. For all hypothesis testing analyses, statistical significance was set 5% (p value = 0.05) to reject null hypothesis.

Statistical differences between groups were assessed using the Prism software version 8.3.0. for Windows (GraphPad Software, San Diego, California USA, www.graphpad.com). Unpaired, non-parametric, t-tests were employed for different comparisons between groups. Sample size of the animal experiment is estimated by assuming moderate to large Cohen’s effect sizes (0.6 - 0.8) between groups with constant variance across groups using semi-parametric bootstrap tests at a 1% level of significance. Based on these tests, 5–10 mice will be sufficient to detect significant differences across groups with 80% power using a semi-parametric bootstrap test for Student’s t-tests.

After 16 weeks of humanization, PBMCs from the hCD34⁺ HSCs-transplanted mice were evaluated by flow cytometry for human lymphoid and myeloid cell surface markers. The NSG-SGM3 mice allow stem cells to develop into human lymphoid lineages, such as T cells (CD3⁺, between 10–90%, including both CD4+ T cells and CD8+ T cells) and B cells (CD19⁺, between 7–60%) (42) ( Figure 1 ). Additionally, differentiation of human myeloid subsets (CD14⁺) was also observed, ranging between 1 and 25%. Within the myeloid lineage, we also detected CD11b+ macrophages (43) ( Figure 1 , Gating strategy is shown in Supplementary Figure 1 ). Since neutrophil population plays a protective role during HIV infection (22), we also evaluated the differentiation of neutrophils in these humanized mice by staining typical neutrophil markers CD15 and CD66b within the CD14 negative population, and we found ~4% CD14-CD15+ cells and ~3% CD14-CD66b+ cells, demonstrating the differentiation of neutrophils in the humanized mice ( Supplementary Figure 2 ).

After HIV/Mtb infections, HIV viral RNA was detected in serum samples from the infected mice starting at 15 dpi, with most animals in the HIV single-infection group being positive at this time, while only two out of the seven mice in the HIV/Mtb co-infection group showed viral RNA ( Figure 2A ). The viral RNA load detected in the positive animals at 15 dpi was between 2×10⁵ and 2.2×10⁶ copies/ml. However, all the HIV-infected animals were positive in subsequent samplings at 28 and 35 dpi. The HIV RNA load was between 3.7×10⁴ and 6.8×10⁵ copies/ml for animals with single HIV infection and between 4.1×10⁴ and 7.7×10⁵ copies/ml for the HIV/Mtb co-infected mice. No significant differences were detected in the viral RNA load between the two HIV-infected groups at these timepoints.

The Mtb bacterial load was assessed in lung and spleen samples after euthanasia in the Mtb single infection group and the HIV/Mtb coinfected mice ( Figure 2B ). In both groups, a higher bacterial load was found in lungs than in spleens. Moreover, the mean CFU count in the lungs and spleens from Mtb single infection group (7.3×10⁶ and 1.4×10⁶, respectively) was higher than the animals co-infected with HIV (5.8×10⁶ for lung and 9.2×10⁵ for spleen), even though their differences are not significant ( Figure 2B ).

We also monitored the human immune cell population changes over time after HIV/Mtb infections. Starting from 15 dpi, huCD45⁺/moCD45⁺ ratio was significantly decreased (p<0.05) in the two HIV-infected groups (HIV single infection and HIV/Mtb co-infection), and the huCD45⁺/moCD45⁺ ratio decrease was sustained until the late stage of the experiment. Conversely, the Mtb single infection group showed similar or even increased huCD45⁺/moCD45⁺ ratio after infection ( Figure 3A ). We also evaluated the human and mouse immune cell changes (hCD45+ and mCD45+ cells) in different treatment groups. The human immune cell population in the HIV-infected groups (HIV-infected and HIV/Mtb co-infected groups) decreases with the time, likely due to the CD4+ T cell depletion ( Supplementary Figure 3A ), while the mouse immune cell population remain the same, because the mouse cells in the humanized mice are mostly immature and not functional, and these cells cannot be infected by HIV ( Supplementary Figure 3B ). As expected, the human immune cells in the Mtb-infected mice showed an increase after Mtb-infection, indicating the T cells activation and proliferation. We also analyzed human CD4+ and CD8+ T cells separately at different time points and found that the CD4+ T cells reduced with the time after HIV infection ( Supplementary Figure 4A ), whereas CD8+ T cells remained comparable ( Supplementary Figure 4B ).

We further used CD4⁺/CD8⁺ T cell ratio as an indicator for CD4⁺ T cell depletion because this parameter was always used in clinical setting to diagnose HIV infection (44, 45), and we found a ~10-fold CD4⁺/CD8⁺ T cell ratio reduction (p<0.05) in the HIV/Mtb co-infected mice as early as 15 dpi, and this trend remained until the end of the experiment. In the single infection group, we also found a lower mean CD4⁺/CD8⁺ T cell ratio since 15 dpi, while the subsequent samplings at 28 and 35 dpi showed significant decreases on CD4⁺/CD8⁺ T cell ratio values. In contrast, there was no significant difference detected over time in the Mtb alone infection group ( Figure 3B ).

In serum sample, significant increases in G-CSF, MCP-1 and MIP-1α was detected in the Mtb single infection group, in comparison with both HIV-infected groups ( Figure 4A ). Additionally, the serum concentration of IL-2 and IL-8 were also significantly increased in the Mtb single infection group, compared to the HIV/Mtb co-infection. The HIV/Mtb co-infected mice analyzed showed higher IP-10 than both the HIV and Mtb single infection mice ( Figure 4A ).

Lung macerate supernatants showed an increase in the concentration of IL-6, RANTES and TNF-α in the HIV single infection group compared to the uninfected control animals, as well as the Mtb single infection group ( Figure 4B ). Additionally, IL-2 concentrations were also higher in the HIV-infected animals than in the uninfected mice. Moreover, HIV single infection also induced statistically higher levels of Eotaxin, MIP-1α and MIP-1β than single Mtb infection. Statistical analysis also revealed a decrease in MCP-1 and PDGF concentration in lung samples from Mtb infected mice, compared to the remaining three experimental groups ( Figure 4B ).

In the case of spleen samples, macerates from the Mtb single-infection group were found to have significantly higher concentrations of IL-1β, G-CSF and MIP-1β than the HIV single-infection group ( Figure 4C ). Similarly, the levels of IL-8 and MIP-1α were higher in the Mtb group than in both HIV-infected groups. In contrast, both the HIV and Mtb single infection groups showed lower concentrations of GM-CSF than the HIV/Mtb co-infected animals, while this group also had statistically higher amounts of IFN-γ than the Mtb group. All the infected groups showed a decrease in IL-1Rα and IL-13, compared to the uninfected control animals ( Figure 4C ).

Given that cytokines are critical components for HIV and Mtb regulation, we examined whether these humanized mice can secrete comparable levels of cytokines as in humans. We found that some of the cytokines, such as IL-4, IL-8, IL-10, and IFN-γ are relatively lower in humanized mice compared to humans with HIV and Mtb infections (46, 47) ( Supplementary Figures 5–7 ).

We stained the lung section with H&E staining, and we observed diffuse immune cell infiltration in lung sample from Mtb-infected mice. In some cases, immune cell infiltration was observed around a necrotic nucleus, in structures similar to TB granulomas. No such cellular aggregates were detected in either the uninfected or the HIV single infection groups ( Figure 5A ). We stained lung sections from Mtb-infected humanized mice by immunofluorescent staining, and the result showed that the cell populations surrounding the necrotic area mostly corresponded with macrophages (CD68⁺), though other immune cell types, such as CD19⁺ B cells, were also found. However, no granuloma structure was observed in the lung section of the uninfected mice, even though a low proportion of cells expressing the human CD68⁺ and CD19⁺ surface markers was observed in the lung sections from uninfected mice ( Figure 5B ).

The CT scan showed an increase in high density areas in the Mtb-infected animals, regardless of their HIV-infection status, indicating the occurrence of inflammation and other pathological changes in the lungs ( Figure 5C ). However, no significant differences were detected in the pulmonary function parameters between the experimental groups ( Figure 5D ).

Plasma metabolome profiling was performed for a total of 10 samples including no infection (n=3), Mtb infection (n=3), HIV infection (n=2), and HIV/Mtb co-infection (n=2). Abundances of 175 metabolites were estimated. To enable comparison of metabolite abundances between different samples, data was normalized across the samples. To investigate differences in plasma metabolome landscape among the four categories of infection, principal components analysis (PCA) was performed. PCA is an unsupervised learning method suitable for dimensionality reduction of high dimensional metabolome data. Interestingly, the plasma metabolome profiles are stratified according to infection status in PCA ( Figure 6A ). Mice with no infection appeared distinct from all infected mice. While the mice with infections were clustered separately from healthy mice, there was a clear distinction among HIV infection alone, Mtb infection alone, and HIV/Mtb co-infection. This suggests that the global plasma metabolome is distinctly altered based on infection status and type. Interestingly, the samples from HIV/Mtb co-infected mice clustered in between HIV infection alone and Mtb infection alone suggesting they show metabolic changes common for individual infections.

To identify metabolites varying across the four categories, we performed OPLS-DA followed by computation of VIP scores on all 175 metabolites. OPLS-DA is a supervised analysis which helps in identifying variables that discriminate different categories of samples based on VIP score. There were 75 metabolites with a VIP score >1 ( Supplementary Table 1 ). The abundances of these metabolites across all four categories were shown with hierarchical clustering (an unsupervised algorithm) in Figure 6B . As expected, in concordance with PCA, dendrogram of hierarchical clustering showed that infection and no infection categories are distinct, while co-infection stratified between the two individual infections ( Figure 6B ).

To identify metabolites that are differentially abundant in HIV infection, we compared healthy mice (n=3) to HIV infection mice (n=4; HIV infection alone and HIV/Mtb co-infection). We identified 8 DAMs in HIV infection with a p value <0.05 ( Figure 6C , Table 1 ). Similarly, we compared healthy mice (n=3) to Mtb infection mice (n=5; Mtb infection alone and HIV/Mtb co-infection) to identify metabolites differentially abundant in Mtb infection which yielded 13 DAMs ( Figure 6D , Table 2 ). Interestingly, three fatty acids, namely dodecanoic acid, palmitic acid and myristic acid were less abundant in HIV infected mice as well as Mtb infection mice ( Tables 1 , 2 ).

To identify metabolites correlating with HIV or Mtb load with metabolites, we used Pearson correlation analysis. HIV infection load (as detected by RNA copies/ml plasma) positively correlated with diethanolamine (r=0.99), and negatively correlated with glucose 6-phosphate/mannose 6-phosphate (r=-0.95) and imidazole acetic acid (r=-0.92) ( Figure 7A ).

Next, we observed that Mtb-infected mice did not show a strong correlation (r=0.68) between pathogen load (as measured by colony forming units per organ) in the lungs and spleens ( Supplementary Figure 8 ), underscoring the heterogeneity of Mtb distribution in these organs of the humanized mice. This is consistent with an earlier report (48) showing that increase of Mtb load in the lungs and spleens follow different trajectories over the course of infection. Therefore, we separately analyzed the correlation between the abundance of metabolites and Mtb load in spleens and lungs.

Interestingly, none of the metabolites correlated with the HIV load (shown in Figure 6C ) exhibited correlation either positively or negatively with Mtb load in lung or spleen. However, PC(16:0/18:1(11Z)) and lysoPC(16:0/0:0) positively correlated with Mtb load in lung as well as spleen ( Figure 7B ). In addition, 3-hydroxyheptanoic acid exhibited a strong negative correlation with Mtb load in lung ( Figure 7B ). Similarly, LysoPC(18:1/0:0) showed strong positive correlation, and myristic acid, PC(20:3(5Z,8Z,11Z)/20:3(5Z,8Z,11Z)) and PC(18:2(9Z,12Z)/16:0) showed strong negative correlation with Mtb load of the spleens ( Figure 7C ).