This case-control study included consecutive cases of neonates with CTB and pHC neonates from four level 4 neonatal intensive care units (NICUs) in four tertiary hospitals in Guangdong, China. Written informed consent was obtained from the legal guardians of all participants. Investigators that analyzed the routine blood tests, immune cell subsets, and inflammatory biomarker levels were blinded to the characteristics of patients from whom the blood samples were obtained.

The medical ethics committee of Nanfang Hospital, Southern medical university approved the study procedures (ID: NFEC-2022-446). The implementations were in concordance with the International Ethical Guidelines for Research Involving Human Subjects as stated in the Helsinki Declaration.

Abandoned blood samples from route blood tests of two extremely preterm infants enrolled in control group in our registered randomized controlled trial- autologous cord blood mononuclear cells for prevention of bronchopulmonary dysplasia program (NCT04440670, https://clinicaltrials.gov/) on the first week after birth were subjected to performing single-cell RNA-seq. After complement, we found one of them was diagnosed with CTB on 22 days after birth. No symptoms of infection or positive culture pathogen was found in another patient. We therefore compared the single-cell RNA-seq data thus to investigate the immune cell characters of the early CTB infection stage in preterm neonates. Further, open access single-cell RNA-seq result of PBMC from an adult diagnosed with active TB drawn from https://www.ncbi.nlm.nih.gov/sra/?term=SRR11038989 was also introduced to compare with the CTB. To compare the basic characters of CTB, we also enrolled another 8 CTB patients from neonatal intensive care unit (NICUs) from 4 tertiary hospitals who were diagnosed between January, 2015 and May, 2022, and had intact demographics information, routine blood tests results, confirmed diagnosis evidence and chest imaging data. The maternal TB (MTB) diagnosis was also confirmed. 9 healthy control infants without confirmed pathogen infection or congenital malformation were paired with the CTB patients by the following characters: born during similar time duration (± 1 year) and in the same hospital, comparable gestational age (GA: ± 4 weeks), with available routine blood tests results during the similar periods after birth that were comparable to CTB cases (± 7 days). When there were several paired infants, we preferred to choose the infant with more comparable GA and testing time points to the CTB infant and who had available lymphocyte immune cells flow-cytometric analysis result. “No confirmed infection” was defined as no positive blood or cerebral spinal (or the normally sterile body sites) culture. The schematic representation of subject cohorts and work flow was showed Figures 1A, B.

The data harmonized across all sites were collected in uniform file where final curation was performed by the clinical research team. CTB was diagnosed as defined by the primary criterion (the presence of proven TB disease) and at least one of the secondary criteria as defined by Cantwell’s criteria (17). These secondary criteria are as follows: (i) lesions in the newborn during the first week of life; (ii) a primary hepatic complex or caseating hepatic granulomata; (iii) TB infection of the placenta or the maternal genital tract; and (iv) exclusion of the possibility of postnatal transmission by investigation of contacts. All of our CTB patients were isolated from their mothers and admitted to the neonatal intensive care unit right after birth. The CTB infection was confirmed by the sputum smear acid-fast staining or culture, positive polymerase chain reaction (PCR) or next generation sequencing (NGS) results for M.tb infection from sputum or blood or a positive interferon gamma release assay (IGRA) result. MTB diagnosis was based on TB infection symptoms and typical chest radiography or computerized tomography results with at least one of the following positive results: IGRA, tuberculin skin test (TST), sputum smear acid-fast staining and culture, PCR or NGS for M.tb from sputum or blood or vaginal secretion. The diagnosis of MTB was all confirmed.

Symptoms of CTB infection deterioration was defined as the following: a.infants who did not need tracheal intubation initially but symptoms deteriorated and needed tracheal intubation or developed shock; b. infants needed reintubation while the symptoms alleviated or were stable previously; c. infants presented as dyspnea who initially did not. Basic clinical demographics information, TB test methods, development and outcomes of CTB and MTB infection characters were presented.

PBMCs were isolated from whole blood through a histopaque density gradient. Cell viability was assessed by trypan blue staining and the samples (cell viability >90%) were prepared and loaded on a 10 X Genomics GemCode Single-cell instrument that generates single-cell Gel Bead-In-EMlusion (GEMs). Full-length, barcoded cDNAs were then amplified by PCR to generate sufficient mass for library construction. Libraries were generated and sequenced from the cDNAs with Chromium Next GEM Single Cell 3’ Reagent Kits v3 according to the manufacturer’s instructions. Single-cell libraries were sequenced on NovaSeq 6000 system (Illumina, USA). Raw sequencing reads are available at the National Genomics Data Center (https://ngdc.cncb.ac.cn/gsa-human/) under the accession numbers HRA015988.

For each sample, the cleaned data filtered for the low quality reads and unrelated sequences were imported to Cell Ranger (10x Genomics, version 2.2.0) and aligned to human reference genome (Ensembl_release 106). Cells were sorted according to the barcodes and the unique molecular identifiers (UMIs) were counted per gene for each cell. Cells with unusually high number of UMIs (≥25000) or mitochondrial gene percent (≥10%) were filtered out. We also excluded cells with less than 500 or more than 4000 genes detected. Additionally, doublet GEMs also should be filtered out. It was achieved by using the tool Doublet Finder (v2.0.3) by the generation of artificial doublets, using the PC distance to find each cell’s proportion of artificial k nearest neighbors (pANN) and ranking them according to the expected number of doublets. Seraut software was used to cluster and group the filtered cells, and R Harmony software was used to merge data and for batch effect correction. Cell type identification was performed using SingleR software.

After scaling the data, dimension reduction was performed using principal component analysis, cell clustering was performed using a graph-based method, and visualization was achieved with uniform manifold approximation and projection (UMAP) considering the top 50 principal components in Seurat. Then, to identify the differently expressed genes (DEGs) and marker genes under a particular condition, we selected the UMAP results as the final visualization of the 22 cell clusters in Seurat. Typically, cell types within a cluster can be directly identified based on canonical cell types using canonical marker genes that strongly indicate which clusters represent the corresponding cell types. For novel clusters or those clusters that lacked canonical marker genes, we used the DEGs as marker genes for cell type identification. We compared gene expression in cells belonging to each cluster with that in all other cells to obtain markers for each cluster. Specifically, the DEGs were identified by comparing cells in a particular cluster with all cells of the other 21 clusters. Then, using the graph-based cluster method, the unsupervised cell cluster result was acquired, and the marker genes were calculated by the FindAlIMarkers function under the following criteria: |log2FC|>0.5, adjusted P-value <0.05 (Ben-jamini-Hochberg correction for multiple testing), and pct>0.25. All marker genes within any particular cluster had to be in the top of the up-regulated or down-regulated genes in that cluster.

Genes were ranked using the moderated T statistics for the relevant coefficient from the limma voom model. Enriched gene sets were identified using the pre-ranked GSEA algorithm implemented in the fgsea R package. Gene set lists were used for kyoto encyclopedia of genes and genomes (KEGG) enrichment assessment. The significance of DEG was analyzed by Model-based Analysis of Single-cell Transcriptomics (MAST) following the criteria including |log2FC|≥1, p_value_adjust ≤ 0.05. P values were adjusted using the Benjamini–Hochberg method for the whole gene set list.

Selected pathways shown in figures were manually curated to select gene sets relevant to immunology and often enriched in several cell types across the various differential expression comparisons. The normalized enrichment score was determined for pathway in each cluster.

Pathway analyses were performed on the 50 hallmark pathways annotated in the molecular signature database, which was exported using the GSEA Base package (version 2.2.4). The GSVA package (version 1.30.0) was applied with default settings to assign pathway activity estimates to individual cells¹⁹. Briefly, the activity scores for each cell were compared using a generalized linear model (GLM). The outputs of these GLMs were visualized in heat maps to assess differential activities of pathways between sets of cells.

Destiny software package (v.2.6.2) was used to perform the trajectory analysis based on dimensionality reduction using diffusion maps. We used diffusion maps as a nonlinear dimensionality reduction technique to find the major non-linear components of variation across T cells. We computed diffusion components in each cell type separately using the Charlotte Python package, which implements diffusion maps. To account for differences in cell density and cluster size, we used a fixed perplexity Gaussian kernel with perplexity 30, with symmetric Markov normalization and t=1 diffusion steps. We selected t=1 diffusion steps because this approximates diffusion of information for each cell through its 20 nearest neighbors in our data. To avoid technical biases in cell type proportions and focus on processes explaining variation specifically within major cell types, we performed diffusion map analyses separately on all cells labeled as T cells.

Single-cell data from the cohort of Yi Cai et al. were downloaded from https://www.ncbi.nlm.nih.gov/sra/?term=SRR11038989 (10x genomics single-cell RNA-seq of Homo sapiens: TB infection adult male peripheral blood dataset-Name: PBMC_TB_3) (18). Using the pre-annotated cell clusters from the original publication, single-cell gene expression data were pooled into pseudo-bulk libraries, and differential gene expression and GSEAs of patients with adult tuberculosis (ATB) versus CTB were done.

The UMAP plots before and after integration, and analysis scripts of all the scRNA-seq in this study are available in Supplementary File 2.

As this was an observational study, no statistical method was used to predetermine sample size. Means (standard deviation) and Student’s t-test were used for continuous variables with normal distribution, and median and interquartile range (IQR) and non-parametric analysis was used for data with non-normal distribution. The variables’ distribution characteristics were estimated with Kolmogrov-Smirnov test. The number and percentage were reported for categorical variables. Group comparisons of categorical variables were performed using the Fisher’s exact test, or chi-square test, as appropriate. For continuous variables, the mean values at each time point in each group for variables were derived from a generalized linear model and differences in values at each time point were compared using an non-parametric analysis between the two groups for data with non-normal distribution and Student’s t-test were used for data with normal distribution, since the small n and non-normal distribution of the values, median was showed for each time point and non-parametric analysis was used for statistical tests. For the time-anchored blood data, analyses were performed using linear mixed-effects models with patient ID as a random intercept to account for intra-individual correlation in repeated measurements. Each time bin contains only one measurement per infant. No repeated-measures methods were applied. Given the exploratory nature of this analysis and small n, we did not apply formal multiple testing correction (e.g., Bonferroni), as this would overly inflate Type II error. All statistical analyses were performed in GraphPad Prism (version 5.0). Two-tailed statistical tests were conducted and a P <0.05 was considered statistically significant.

We included nine neonates with confirmed CTB and nine pHC infants between January 2015 and May 2022 from four NICUs in four tertiary hospitals in Guangdong, China (Figure 1A). The demographic, clinical, and laboratory detection characteristics of the patients with CTB and their mothers are showed in Figures 1B, C; Supplementary Tables 1, 2. The median deterioration time was 24 DAB, and the median diagnosis time 31 DAB. None of the infants received preventive anti-TB therapy. All neonates received anti-TB treatment after diagnosis, and the median number of days of treatment was 31 DAB. A total of 77.7% (7/9) of the patients died at a median age of 42.5 days. The clinical characteristics including gender, GA, birth weight, delivery mode, Apgar scores, respiratory support condition, sepsis evaluations and BPD severity were comparable in two groups, except that infants in CTB group received longer antibiotic therapy (P = 0.001, Supplementary Table 2). DAB for which routine blood tests were conducted in the patients with CTB and pHC infants are shown in Supplementary Table 3.

Routine blood tests during hospitalization in the two groups were compared, and the results shown in Figure 1D. Significant time × group interactions were detected for L% (β=-0.849, P<0.001), platelet count (PLT; β=-6.570, P = 0.002), monocyte percentage (MN%; β=-0.190, P = 0.001), CRP (β=2.886, P<0.001), and N% (β=1.067, P<0.001), demonstrating distinct temporal trajectories between CTB and pHC neonates. Specifically, compared with pHC, lymphocytes, monocyte and platelet count (PLT) in CTB showed a downward trend over time. On the contrary, the proportion of neutrophil cells and CRP levels showed an upward trend and increased before the symptoms worsened. No significant time × group interaction was found for WBC (β=0.0743, P = 0.333). The de-identified raw longitudinal complete blood counts were available in Supplementary Table 5.

As T cells were the major cell type identified in lymphocytes, we also collected the lymphocyte immunophenotype of five HCs and three patients with CTB (one of which had two repeated results recorded at different time points) with available results (Supplementary Table 5). We found the total T, B, and natural killer (NK) cell frequencies were comparable between the two groups, but the FoxP3⁺ regulatory T cell (Treg) frequency in CTB neonates (2.06%) was lower than that in pHC neonates (9.39%; p = 0.016) (Figure 1E). The de-identified raw flow cytometry percentages were available in Supplementary Table 5.

For this pilot, hypothesis-generating scRNA-seq study, we analyzed peripheral blood mononuclear cells (PBMCs) from one CTB neonate, one paired healthy control (pHC) neonate, and one adult tuberculosis (ATB) patient, representing single biological replicates per group. We initially isolated and sequenced 24,273 cells from PBMC suspensions derived from two infants: one with CTB (3,597 cells) and one paired control (9,787 cells), as well as one patient with active ATB (10,878 cells). After removing 12.92%, 13.22%, and 15.73% of the cells that represented doublets, empty droplets, and low-quality cells, respectively, we selected 21,421 cells for further analysis (Supplementary Table 6).

We applied principal component analysis across all the cells and classified them into 22 groups of cell types using graph-based clustering of the informative principal components. The 22 expression groups were further grouped by hierarchical clustering, leading to the characterization of five major cellular compartments: T, B, NK, myeloid, and hematopoietic stem cells (HSCs), as well as other cell types (20-mast cells, 8,19-platelets, and 18-erythrocytes) (Figure 2A; Supplementary Table 6). Cluster-1 (24.63% of all cells) comprised T cells expressing CD3D, CD3E, CD3G, and TRAC; Cluster-2 (51.47% of all cells) comprised myeloid cells expressing S100A8, S100A9, CD14, and FCN1; Cluster-3 (3.88% of all cells) comprised B cells expressing CD79A, CD79B, IGHM, and IGHd; Cluster-4 (14.99% of all cells) comprised NK cells expressing GZMA, GZMB, NKG7, and GNLY; and Cluster-5 (1.00% of all cells) comprised HSCs expressing STMN1, MYB, IKZF1, and CD34 (Figures 2B–D; Supplementary Figure 1 and Supplementary Table 7).

The control patient had fewer T cells at this stage (four DAB); however, regarding the upward trend of the lymphocyte frequency and flow cytometry analysis results, the lack of T cells was transient. In particular, we showed the trend in lymphocyte frequency for the two patients by performing scRNA-seq, which was similar to the trend observed for the two groups, indicating the representativeness of the two samples (Supplementary Figure 2).

We also noticed the presence of a specific type of cell (i.e., HSCs) whose characteristics have rarely been reported. In this limited single-case comparison, HSCs appeared less abundant in the ATB patient than in the CTB and pHC infants, and relatively more abundant in the CTB infant compared with the pHC infant. These HSCs showed low expression of inflammatory markers, S100A8 and S100A9, and high levels of STMN1, IKZF1, MYB, SPINK2, and CD34 (Figure 2D). These marker genes play roles in the regulation of hematopoiesis, lymphoma progression, and stem cell attachment to the extracellular matrix in bone marrow. The relative enrichment of HSCs in the single CTB infant at this early stage may suggest a compensatory lymphopoietic response, which we interpret as a hypothetical early defensive mechanism against Mycobacterium tuberculosis infection.

For the differentially expressed genes (DEGs) of each cell type observed between CTB and the control or ATB, we used the Kyoto Encyclopedia of Genes and Genomes (KEGG) gene set enrichment analysis and weighed the gene effect on the gene ontology to define the pathways related to CTB. Relative to the pHC infant, DEGs in the CTB infant were enriched in apoptosis, endocytosis, Salmonella infection, and Th17 cell differentiation pathways. Compared with the ATB patient, ubiquitin-mediated proteolysis, endocytosis, and Salmonella infection pathways were more prominent in the CTB infant (Figure 2E; Supplementary Table 8).

We then used gene set enrichment analysis (GSEA) to weigh the gene effect to define the enrichment pathways involved in cellular and environmental information processes, human infectious diseases, and the immune system, and systemically assessed cell-type-specific significant enrichment pathway activity among HCs and patients with ATB and CTB based on the DEGs measured between samples. In this single-case comparison, myeloid cells in the CTB infant exhibited high expression of inflammatory factors including S100A8 and S100A9, alongside strong T cell activation signatures and elevated antigen processing and presentation signatures relative to the pHC infant. By contrast, NK and T cell effector functions appeared globally repressed in the CTB infant (Figure 2F; Supplementary Figures 3A–C, Supplementary Table 9). Compared with the ATB patient, antigen processing and presentation pathways in myeloid, NK, and T cells were transcriptionally less activated in the CTB infant (Figure 2F). These preliminary observations suggest a potential imbalance characterized by excessive innate inflammation but impaired adaptive immune activation in this CTB case. B cell subset distribution and functional signatures were comparable among the three individuals (Supplementary Figure 4, Supplementary Table 10).

The antigen processing and presentation ability plays a critical role in the ultimate outcome of TB infection (19). Therefore, we compared the gene expression signatures of the major histocompatibility complex (MHC) in the antigen presentation pathway of the three samples. We found that in myeloid cells, less expression of MHC II genes was observed in CTB compared with that in ATB, which indicated that although TB infection could induce myeloid cell activity in the early stage of infection, compared with adults, these activated cells showed decreased up regulation of co-stimulatory molecules (MHC II) that were necessary for activation of T and B lymphocytes. For NK cells, MHCI molecular expression in CTB and HC was comparable, but both exhibited lower expression of MHC II molecules than that in ATB. In T cells, MHC II expression in CTB showed reduced activation compared with that in both HC and ATB (Figure 2G; Supplementary Table 11). While these observations are limited to a single biological replicate per group, they raise the hypothesis that attenuated antigen presentation may contribute to the severe clinical course and high mortality of CTB.

On the basis of clinical and flow cytometry data, CTB neonates showed a declining lymphocyte trend, suggestive of insufficient anti-tuberculosis adaptive immunity. our exploratory scRNA-seq analysis detected 5,275 T cells in all three donor groups that could be subclustered into eight subsets (Figure 3A). The T cell subsets according to cell lineage and functional state were identified as CD4+ T (naïve CD4+ T, Th1/17, Tfh-follicular helper T, and Treg cells), CD8+ T (naïve CD8+ T, exhausted T, and cytotoxic T cells), and proliferated T cells (Figure 3A).

T0, T1, T2, T4, and T10 expressed high CCR7 levels, indicative of naïve-like CD4 T cells. Both T6 and T13 expressed high levels of activated CD4 T-cell markers, such as KLRB1, IL17R, NKG7, and GZMA, and were identified as Th1/17 cells (Figure 3B). In this single-case comparison with the pHC infant, Th1/17 cells in the CTB infant displayed transcriptional downregulation of multiple immune-related pathways (Figure 3C; Supplementary Figure 3B).

Proliferating T cells showed robust activation of TNF signaling, reactive oxygen species, IL6-JAK-STAT3, IL2-STAT5, and inflammatory response pathways, indicating strong activation (Figure 3D). The proportion of proliferating T cells appeared lower in the CTB and ATB samples than in the pHC sample (Figure 3E). Pathway analysis further suggested global downregulation of immune and infectious disease-related pathways in T cells from the CTB infant relative to the pHC infant (Figure 3C; Supplementary Figure 3B, Supplementary Tables 12, 13).

Cluster T7 expressed FOXP3, CTLA4, IL2RA, and IKZF2, consistent with Tregs (Figure 3B). The frequency of Tregs appeared lower in the CTB and ATB samples than in the pHC sample (Figure 3E). Combined with longitudinal clinical data showing decreasing Treg frequency and rising CRP levels in CTB neonates, these single-sample observations support the hypothesis of disrupted immune homeostasis and excessive inflammation during CTB progression.

We identified three diverse CD8 T cell subsets: naïve CD8 T (T5: CD8A, CD8B, and CCR7), cytotoxic CD8 T (T3 and T12: GZMA, GZMB, NKG7, KLRB1, and PRF1), and exhausted CD8 T cells (T8: weaker expression of GZMA, GZMB, NKG7, KLRB1, and PRF1) (Figures 3A, B). The cytotoxic CD8+ T cell subsets displayed up-regulated pathways associated with peroxisomes, oxidative phosphorylation, fatty acid metabolism, and interferon alpha and interferon gamma (IFN-γ) responses, indicating that these cells had a high energy metabolism and were highly active in infectious diseases and immune systems (Figure 3D). In this limited comparison, the proportion of cytotoxic CD8+ T cells was lower in the CTB infant than in the pHC and ATB individual (Figure 3E), which may suggest a hypothetical defect in controlling mycobacterial replication. Exhausted CD8+ T cells represented a group of dysfunctional T cells, which are present in chronic infections or tumors (20, 21). We identified more exhausted T cells in patients with both CTB and ATB, which may indicate a dysfunctional state of T cells during TB infection (Figure 3E).

To understand the state transitions among T cell subtypes, we applied destiny in R software to draw diffusion maps to construct potential developmental trajectories of T cell subtypes. The developmental trajectory suggested that naïve CD8+ T cells may eventually enter a cytotoxic CD8+ T cell state (Figure 3F). In CD4+ T cells, naïve T cells may eventually enter a proliferated T cell state through Tfh cells (Figure 3F). These diffusion map analysis identified model-based trajectory patterns of T cell subset differentiation, which should be interpreted as computational inferences rather than definitive lineage evidence.

Our exploratory scRNA-seq analysis identified 11,026 myeloid cells in all three donor groups that could be subclustered into 10 subsets (Figure 4A; Supplementary Table 14). The myeloid cell subsets according to cell lineage and marker genes, as well as gene set variation analysis (GSVA) hallmark pathway characteristics, were identified as three subsets of monocytes [classical (CD14+CD16), nonclassical (CD14lowCD16+), and intermediate (CD14+CD16+) monocytes], dendritic cells (high expression levels of CLEC4C, IL3RA, and GZMB), five subsets of neutrophils (CAMP+, FCGR3B+, DEFENSIN+, CEACAM1+, and CD3G+), and megakaryocytes (high expression levels of PF4 and PPBP) (Figures 4B–D).

The overall distribution of myeloid cell subsets appeared similar between the single CTB and pHC infants (Figure 4C). Next, we systemically assessed the cell-type-specific significant enrichment pathway activity and found that compared with HCs, myeloid cells in CTB were generally more activated in the immune response, which potentially indicated a stronger T cell activation signature, increased antigen processing and presentation ability. Compared with ATB, the frequencies of the five subsets of neutrophils were higher. These neutrophils showed high expression of antimicrobial genes, such as CAMP, DEFFESIN, FCGR3B, MPO, S100A9, and S100A12. However, compared with ATB, the pathway of antigen processing and presentation was much weaker (Figure 4E; Supplementary Figures 3C, Supplementary Table 13). Antigen processing and presentation is critical for T cell activity (3, 19, 22, 23). These single-case observations raise the hypothesis that CTB is characterized by robust innate activation but impaired ability to prime TB-specific adaptive immunity.

The scRNA-seq analysis detected 3,211 NK cells in all three donor groups that could be subclustered into three subsets (Figure 5A, Supplementary Table 10). The NK cell subsets according to marker genes and GSVA hallmark pathway characteristics were identified as memory-like NK (high expression of CXCR4 and TRBC2) and cytotoxic NK cells (high expression levels of GZMB and GNLY) (Figures 5A–C).

Cytotoxic NK cells displayed strong IFN−γ response and immune effector pathway activation (Figure 5D). NK cells play a significant role in the host defense to a variety of infections, with the capacity to secrete IFN-γ and perform cytolytic functions (21). In this exploratory single-replicate comparison, the frequency of cytotoxic NK cells was lower in the CTB infant than in the pHC and ATB individual (Figure 5C). Cell-type-specific pathway analysis further suggested reduced immune and anti-infectious activity in NK cells from the CTB infant compared with the pHC infant (Figure 2F). The functionally significant reductions in IFN-γ and TNF-α production activity, as well as the reduced cytotoxic function of NK cells in preterm infants may be one of the immune factors contributing to the gradual deterioration of symptoms under CTB infection.

This study has several important limitations, particularly regarding the epistemological constraints of scRNA-seq sample size. First, scRNA-seq was performed on only one individual per group (one CTB neonate, one pHC neonate, one ATB patient). With single biological replicates, it is impossible to determine whether the observed transcriptomic shifts reflect universal pathophysiological features of CTB or individual-specific genetic and developmental idiosyncrasies. As highlighted by Conti et al., host genetic background, including common polymorphisms in HLA and innate immune signaling pathways and rare inborn errors of immunity, profoundly shapes susceptibility and immune responses to mycobacterial infection (24). Such inter-individual genetic variability severely limits the generalizability of n=1 transcriptomic data. Accordingly, all scRNA-seq findings in this study are explicitly interpreted as exploratory and hypothesis-generating, not definitive population-level conclusions. To mitigate this limitation, we leveraged flow cytometry data from our larger 9−patient CTB cohort to orthogonally validate key scRNA-seq observations, including Treg depletion and trends in lymphocyte subset dynamics. However, additional protein-level validation of other critical findings-such as reduced cytotoxic CD8+ T cells and altered NK cell frequencies-could not be performed due to the lack of remaining clinical samples, which is acknowledged as a major limitation. Second, critical consideration must be given to neonatal immune ontogeny when interpreting immune differences between congenital tuberculosis and adult tuberculosis. Preterm infants exhibit profound developmental immaturity in antigen−presenting cell function, including intrinsically lower baseline expression of MHC−II molecules, diminished co−stimulatory capacity, and reduced Th1−polarizing potential relative to fully mature adults (28, 29). These developmental differences are independent of infectious or disease status and represent inherent features of the neonatal immune system. Accordingly, direct cross−age comparison between the CTB neonate and the ATB adult is inherently confounded by developmental immune immaturity, and observed differences in MHC−II expression, antigen processing, and T cell function cannot be solely attributed to disease−specific immune evasion mechanisms. To isolate true CTB−specific immune signatures, all mechanistic conclusions in this study should be anchored on the internal, age−matched comparison between the CTB neonate and the pHC neonate. Transcriptomic differences between the CTB neonate and the ATB adult should therefore be interpreted as exploratory and illustrative. This rigorous focus on the internally controlled contrast strengthens the interpretability of our findings and ensures that conclusions regarding CTB pathogenesis are not conflated with inherent developmental immaturity of the neonatal immune system. Third, most patients deteriorated rapidly despite early intervention, precluding analysis of immune dynamics following anti-tuberculosis therapy.