THP-1 cells, which are a human monocytic cell line, were cultured in a T25 cell culture flask in Roswell Park Memorial Institute Medium (RPMI)-1640 (LTC Tech, 21875034) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (LTC Tech, 10493106), incubated under 5% CO₂ and at 37°C and transferred after 3 days to a T75 culture flask for further expansion inside a biosafety level 2 laboratory (BSL2).

The clinical multi drug-resistant (MDR) R179 M. tb strain and the M. Bovis BCG strain were cultured separately in Middlebrook 7H9 (Sigma-Aldrich, M0178) with 10% OADC (Albumin, Dextrose, Oleic Acid, Catalase and Sodium Chloride) (Becton Dickinson, 212240) and 0.5% glycerol (Merck Millipore, Germany, G5516-1L) without Tween 80, as this detergent is known to affect macrophage uptake and the host response to M. tb (Leisching et al., 2016b),. The bacteria were cultured as previously described. Briefly, a starter culture was set up in T25 cell culture flask at a volume of 10 ml and incubated for 2-3 days at 37°C until an optical density (OD) of less than 0.4 was reached. Subcultures, also in T25 cell culture flasks, were grown to a maximum OD of 0.4 and stored at -80°C until needed.

For infection experiments, THP-1 cells were seeded in 24-well plates (CELLSTAR^®, P1PLA044C-000048) at 1 × 10⁵ cells/well with Phorbol 12-Myristate 13-Acetate (PMA) (Sigma Aldrich, USA) at a final concentration of 100 nM in growth medium (RPMI) supplemented with 10% FBS and incubated for 72 h to differentiate THP-1 cells into macrophage cells. The differentiated THP-1 (dTHP-1) cells were then transferred to a CO₂ incubator in a biosafety level 3 (BSL3) laboratory and infected.

Mycobacterial culture stocks (1ml) were thawed, and clumps were broken and loosened by pipetting ten times with a 1000 μl pipette. Subsequently, the mycobacterial culture was syringed 10 times through a 25 gauge (G25) needle to further break the clumps and obtain single bacteria (Leisching et al., 2016a). This technique is referred to as the syringe-settle-filtrate (SSF) method (Madhvi et al., 2020). The vial was immediately allowed to stand for 30 s to allow the large clumps to settle after syringing. The top 750 μl was carefully removed, mixed with 4.25 ml macrophage growth medium and the 5 ml bacterial suspension filtered through a 5.0 μm pore size filter (Merck Millipore, Germany). Filtering through a 5.0 μm pore size filter eliminates all large clumps, leaving only small clumps of approximately 2–3 bacteria. More than 90% of the filtered bacteria have been determined to be single (Leisching et al., 2016a). The exact same method (SSF) was used prior to infection to determine the titer of the filtered bacteria. The required volume of filtered bacteria was mixed with growth medium and added to the dTHP-1 cells to give a Multiplicity of Infection (MOI) of one. The 24-well plates were incubated for 4 h to allow for bacterial uptake.

To remove extracellular mycobacteria, the infected dTHP-1 cells were washed three times with 800 μl 1x Phosphate Buffer saline (PBS) (LTC Tech, 14190144) and prepared for vector-based overexpression of the target gene (IFIT2) by adding 250 μl of growth medium per well. Cells that served as a negative control for vector treatment, received 300 μl of growth medium per well. Uninfected THP-1 cells were used as controls. For basal measurements, infected and uninfected cells were collected 4 h post-infection to capture baseline gene expression prior to vector-based treatments.

Plasmid transfection was performed using Mission siRNA liposome-based transfection reagent (Sigma-Aldrich, S1452). The transfection of the IFIT2 gene into THP-1 cells for its overexpression (knock-up) was done in 24 well plates with 1 × 10⁵ cells per well in 250 μl of growth medium. For each well containing THP-1 cells (infected or uninfected) that had to be treated with vector, 50 μl of transfection Master Mix was added to the 250 μl of growth medium already present in the wells. Transfection Master Mix, per well, consisted of 1 μl of transfection reagent, 0.3 μl (equating to 60 ng at the final dilution) of plasmid DNA vector and 48.7 μl of Dulbecco’s Modified Eagle’s Medium (DMEM) which was mixed, vortexed and allowed to stand at room temperature for 15 minutes. The final volume in each well was 300 μl.

The plasmid DNA vectors used included IFIT2 with a FLAG tag placed at the N-terminus [GenScript, SC1200_OHu31069C], IFIT2 with a FLAG tag placed at the C-terminus (GenScript, SC1200_OHu31069D), IFIT2 with no FLAG tag (GenScript, SC1200_OHu31069E), a non-specific gene vector, Creatinine Kinase (CKB) (GenScript, OHu09573), and Empty Vector (GenScript, SC1822). The pcDNA 3.1+C/-(K)-DYK plasmid was used to generate each vector. Uninfected THP-1 cells, mycobacteria-infected THP-1 cells, uninfected-IFIT2 induced THP-1 cells, mycobacteria-infected CKB-induced THP-1 cells, and mycobacteria-infected THP-1 cells transfected with the empty vector served as control samples. The cells were then incubated for 8 h at 37°C in an incubator with 5% CO₂.

Infected THP-1 cells after 4 h of uptake (4 h post infection) and uninfected THP-1 cells were prepared for RNA extraction after washing thrice with PBS. RNA was extracted immediately after collecting the 4-hour uptake samples to accurately capture baseline gene expression. Total RNA from human macrophage-like cells was extracted using a kit (RNeasy Plus Mini Kit) according to the manufacturer’s instructions. Extraction was also performed immediately following the 12 h infection period. Genomic DNA (gDNA) was removed by column filtration using the “gDNA eliminator” column included in the kit. For each experiment, the RNA quantity and quality were assessed using an Agilent 2100 Bioanalyzer. Only RNA samples with an RNA integrity number (RIN) ≥ 9 were used for AmpliSeq (Targeted Transcriptome Sequencing) and RT-qPCR experiments. Total RNA was extracted and frozen immediately at −80°C until use.

AmpliSeq, a targeted transcriptome sequencing technique, was performed at the Central Analytical Facilities (CAF) at Stellenbosch University, South Africa. The Ion AmpliSeq™ Transcriptome Human Gene Expression Panel Chef-Ready Kit was used to prepare Ion AmpliSeq transcriptome libraries on the Ion Chef™ System according to the manufacturer’s protocol. 10 ng total RNA was reverse transcribed using the SuperScript™ VILO™ cDNA Synthesis Kit with incubation at 42°C for 30 minutes and enzymatic inactivation at 85°C for 5 minutes on a SimpliAmp™ Thermal Cycler (Thermo Fisher Scientific). The reverse-transcribed samples were loaded onto the Ion Chef™ System for library construction, with thirteen amplification cycles consisting of 16 minutes of annealing and extension for amplification of the primer targets. The libraries were diluted to a target concentration of 60 pM (picomolar) for template preparation using the Ion 540 Chef Kit (Thermo Fisher Scientific). 25 μl of diluted, pooled barcoded library were loaded onto the Ion Chef liquid handler for template preparation and enrichment using Ion 540™ Chef Reagents, Solutions and Supplies according to the protocol, MAN0010851 F.0. The enriched spherical ion particles were loaded onto an Ion 540 chip. Massively parallel sequencing was performed on the Ion Torrent™ GeneStudio™ S5 Prime System using Sequencing Solutions and Reagents, according to the protocol MAN0010851 F.0. Flow space calibration and base-caller analysis were performed using standard analysis parameters in Torrent Suite Version 5.18.1 Software. The reads were quality trimmed during the base-calling step using a moving window of 30 base pairs and a cutoff quality value (QV) of 15. Mapping was performed using the Torrent Mapping Alignment Program (TMAP) included in Torrent Suit software V5.18.1. Reads were mapped to the hg19_AmpliSeq_Transcriptome_v1.1 reference genome, and only reads with an alignment length of 17 or more were retained. Read counts were generated using the AmpliSeq RNA plugin.

The analysis of differentially expressed genes (DEGs) was performed in R version 4.3.1, using the DESeq2 R package, which uses a negative binomial distribution model (Love et al., 2014). The Benjamini-Hochberg method was used to correct for multiple testing. P-adjusted (or false discovery rate) values less than 0.05 and absolute log₂(fold change) > 1.5 were used to screen out the highly significant DEGs.

The top 70 genes with log₂(fold change) >1.5 extracted were selected and GO functional analysis and KEGG pathway enrichment analysis (P < 0.05) were performed using the clusterProfiler package (Tianzhi. et al., 2021) in R software.

Functional analysis was performed by first performing over-representation analysis using Gene Ontology (GO) terms to identify which GO terms were overrepresented or underrepresented in the DEGs. The GO terms were grouped into biological processes, molecular functions, and cellular components. The clusterProfiler R package was used for over-representation analysis by performing statistical analyses using hypergeometric testing (Tianzhi. et al., 2021). All genes tested for differential gene analysis were used as background genes. Genes with p-adjusted values less than 0.05, and log₂(fold change) > 1.5 or < –1.5, were used as significant gene lists. Bar plots were used to visualize over-representative GO terms in the differentially expressed genes. Functional class scoring analysis was performed using gene set enrichment analysis (GSEA) and gene sets from the Kyoto Encyclopedia of Genes and Genomes (KEGG). GSEA evaluates whether gene sets for certain biological pathways are enriched in the larger absolute log₂(fold changes) by performing analysis using the log₂ (fold changes) generated by DEGs analysis from DESeq2 for each gene. The clusterProfiler R package and Pathview tools were used to perform gene set enrichment and pathway analyses in R Studio. Here, the coordinated differential expression of a group of genes that share functional similarities was tested instead of single genes. The NA values and duplicates were removed prior to the analysis. The Pathview R package was used to transform the KEGG pathway data from clusterProfiler into pathway images. Finally, pathway topology analysis was performed using the Signaling Pathway Impact Analysis (SPIA) tool. The SPIA R package in R was used to include all substantial information that was not obtained from the GO and GSEA analysis (Tarca et al., 2009).

To investigate the relationships between the highly significant DEGs extracted with log₂(fold change) > 1.5, protein-protein interaction networks were constructed using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database version 12.0 with default parameters.

To validate the transcriptomic data, we selected five key DEGs and performed RT-qPCR. Good quality RNA (RIN>9, 0.8µg) was converted to cDNA using the Quantitect^® Reverse Transcription Kit. To ensure the removal of genomic DNA, “gDNA wipe-out buffer” was used. RT-qPCR amplification was performed using a LightCycler 96 system (Roche, Germany). The LightCycler^® 480 SYBR Green I Master and QuantiTect^® primers were used.

Hs-GAPDH and Hs-UBC were selected as reference genes. The amplification process involved 45 cycles at 95°C for 10s followed, 60°C for 10s and finally 72°C for 10s. Gene expression fold-changes were computed for infected macrophages without treatment and infected macrophages with treatment using calibrated, normalized relative quantities using the equation N =N₀×2^Cp. All RT-qPCRs were performed on RNA extracted from three separate experiments, with three technical replicates for each set of experiments. All biological replicates, a positive control, and a non-reverse transcription control were run in triplicate (along with the calibrator) according to MIQE Guidelines (Bustin et al., 2009).

Data generated from RT-qPCR were analyzed using Light Cycler 96 SW 1.1 software to obtain the relative mRNA expression. The relative mRNA expressions were further analyzed using One-Way ANOVA and Tukey’s test for multiple test corrections to generate p-values and graphs in GraphPad Prism V8.

GraphPad Prism V8 was used to analyze the CFU count using One-Way ANOVA and Tukey’s test for multiple test corrections to generate p-values and graphs.

In constructing plasmid vectors for overexpression analysis, DNA sequences of target genes are often altered to include tags for purification and detection purposes. Commonly used mammalian expression vectors, such as pcDNA3.1, can incorporate tags such as FLAG, Poly histidine tag (His-tag) and the glutathione S-transferase (Booth et al., 2018). This study used the FLAG tag for IFIT2 vector construction, which is advantageous due to its small size, hydrophilicity, and ease of detection (Yadav et al., 2016). The positioning of tags can affect protein stability, activity, and structure, necessitating investigation to ensure functionality. Therefore, the current study evaluated the effects of FLAG tag positioning on IFIT2 expression and anti-mycobacterial activity against M. tb in THP-1 cells by comparing N-terminal FLAG, C-terminal FLAG, and no FLAG vector constructs. The anti-mycobacterial effects were evaluated by recording CFUs, and gene expression levels were measured using RT-qPCR.

The positioning of the FLAG tag significantly influenced IFIT2 expression levels and anti-mycobacterial activity. The C-Flagged IFIT2 vector exhibited the lowest gene expression and highest CFU counts against M. bovis BCG ( Supplementary Figures 1A, B ), indicating reduced anti-mycobacterial activity. In contrast, both N-Flagged and No-Flag IFIT2 vectors demonstrated higher expression levels and significantly lower CFU counts, suggesting improved anti-mycobacterial activity. Interestingly, the empty vector control also showed some anti-mycobacterial effects and IFIT2 expression, likely due to a non-specific immune response. However, the anti-mycobacterial activity of the empty vector-transfected control group was lower than that of the N-Flagged and No-Flag vectors though the observed difference were not statistically significant.

To explore this further, we tested the N-Flagged and No-Flag constructs using the multi-drug-resistant R179 M. tb strain, which is closer to real-life TB infections. A non-specific gene vector, Creatine Kinase (CKB), was included as a control, based on previous data showing unchanged expression in infected macrophages. The results showed that the N-Flagged IFIT2 vector had the lowest CFUs, and higher IFIT2 gene expression compared to the No-Flag vector ( Supplementary Figures 2A, B ). Interestingly, both the empty vector and CKB also showed some anti-mycobacterial activity and IFIT2 expression. Collectively, these results highlight the N-Flagged IFIT2 construct as the most effective and suitable for further high throughput studies, while the C-Flagged construct was deemed unsuitable due to poor antimycobacterial activity.

RNA-seq analysis of the samples from uninfected cells, uninfected cells with IFIT2 treatment, infected cells with treatment (IFIT2, Empty vector, and non-specific gene vector CKB), and infected cells without treatment yielded an average of 63 million reads aligned to the hg19 AmpliSeq Transcriptome v1.1. Principal Component Analysis (PCA) showed distinct clustering of the different groups – uninfected from infected cells and treated from untreated cells ( Figure 1 ). Infected, untreated THP-1 cells served as the reference for all treatments. The pairwise comparisons included IFIT2-transfected M. tb-infected THP-1 cells vs. M. tb-infected THP-1 cells only, empty vector-transfected M. tb-infected THP-1 cells vs. M. tb-infected THP-1 cells only, and CKB-transfected M. tb-infected THP-1 cells vs. M. tb-infected THP-1 cells only. Venn diagrams of the differentially expressed genes from the three treatment groups revealed both shared and unique genes in each treatment group ( Supplementary Figure 7 ).

Using R Studio and the DESeq2 R package, we conducted a thorough analysis to identify differentially expressed genes (DEGs) and compared all experimental treatments with the control groups. We employed a cutoff value of false discovery rate (FDR), also known as the p-adjusted value < 0.05, to extract significant DEGs. An MA-plot was generated to visualize genome-wide transcriptome expression at 12h post-infection in the treated groups compared to that in the control groups ( Supplementary Figure 3 ). A total of 282 DEGs were identified, with 189 upregulated and 90 downregulated DEGs observed in IFIT2-induced M. tb-infected THP-1 cells compared to those in M. tb-infected THP-1 cells only. Due to the anti-mycobacterial activity and IFIT2 gene expression levels observed in the control (empty vector-transfected M. tb-infected THP-1 cells, and the non-specific gene (CKB)-transfected M. tb-infected THP-1 cells) in the preliminary (gene optimization) experiments, we sought to identify the DEGs expressed ( Supplementary Figures 1 , 2 ). For the Empty Vector, 194 DEGs were identified and compared to the M. tb-infected THP-1 cells only. Of these, 123 DEGs were upregulated and 71 DEGs were downregulated. In addition, 249 DEGs were identified for the non-specific gene vector CKB, of which 175 DEGs were upregulated and 74 DEGs were downregulated.

Volcano plots and heatmaps were used to provide detailed insights into the expression patterns of the DEGs. Hierarchical clustering of DEGs shown in the heatmap in Figure 2 , revealed distinct clustering patterns between the treatment and control groups, underscoring the reliability of the samples. We used the gene expression data from each sample to determine the distance between them and their correlations. As shown in Figure 2 , the treatment groups consistently formed separate clusters from those of the control groups, emphasizing the precision and reliability of the samples. Additionally, volcano plots in Figures 3 , 4 illustrate the top 30 significant gene expression changes downstream of IFIT2 and the Empty Vector, respectively.

We applied stringent filtering criteria to extract top-ranked significant DEGs, resulting in 70 ( Table 1 ) genes with an absolute log₂(fold change) > 1.5 and FDR < 0.05, among which only two were downregulated. The top 30 significant DEGs are represented in a volcano plot ( Figure 3 ).

Gene Ontology (GO) enrichment analysis for biological processes highlighted enrichment in GO terms, including response to virus, defense response to virus, regulation of innate immune response, and cytokine-mediated signaling pathways. Molecular function analysis indicated enrichment in functions, such as double-stranded RNA binding, nucleotidyltransferase activity, GTP binding, and cytokine receptor binding. Notably, no GO terms were enriched for cellular components ( Figure 5 ). A table showing GO terms with associated gene lists and statistical values can be found in Tables 2 , 3 .

Additionally, Gene Set Enrichment Analysis (GSEA) using the KEGG database revealed enrichment in pathways related to viral infections, including influenza A, Hepatitis C, measles, and COVID-19, as well as pathways involved in host response to various pathogens, such as NOD-like receptor signaling and cytokine-cytokine receptor signaling pathways. The overlap of genes in these pathways was depicted using an Upsetplot, illustrating the shared genes among certain pathways ( Supplementary Figure 4 ). A table showing specific KEGG pathways with associated gene lists, statistical values and enrichment score can be found in Table 4 .

Signaling Pathway Impact Analysis (SPIA) further supported these findings, indicating significant enrichment and activation of pathways related to cytokine-cytokine receptor interaction and chemokine signaling. Evidence suggested inhibition of TB and activation of apoptosis pathways ( Supplementary Figure 5 ; Table 5 ).

As seen from the preliminary experiment, the empty vector control revealed unexpected observations and outcome by showing an appreciable increase in IFIT2 gene expression and revealing antimycobacterial effects. We therefore hypothesized that the empty vector elicits a strong non-specific host-immune response in TB infection. As a result, we carried out a transcriptomic investigation to understand this outcome and tested out this hypothesis. It can be seen from Table 6 that 23 DEGs were identified with 4 being downregulated and 19 upregulated after extracting genes with absolute log₂(fold change) > 1.5. The top 30 significant genes with FDR <0.05 affected by empty is represented in a volcano plot ( Figure 4 ).

Predictably, the GO terms were primarily enriched in response to virus, defense response to virus, defense response to symbiont, cellular response to zinc ion, cellular response to copper ion, antiviral innate immune response, response to copper ion, response to zinc ion, negative regulation of viral genome replication, regulation of viral genome replication, negative regulation of viral process, cytosolic pattern recognition receptor signaling pathway, cellular response to exogenous dsRNA, regulation of nuclease activity ( Figure 6 ). A table showing the top 15 GO terms and all KEGG with associated gene lists and statistical values can be found in Tables 7 , 8 .

The SPIA results for the empty vector, as shown in Table 9 , Supplementary Figure 6 , revealed intriguing findings that underscore the distinct effectiveness of IFIT2 induction in anti-mycobacterial activity compared to the empty vector. Notably, the analysis demonstrated that pathways such as the cytokine-cytokine mediated interaction pathway, chemokine signaling pathway, and apoptosis were inhibited by the empty vector but were activated downstream by IFIT2 induction.

In comparison to IFIT2, the treatment with empty vector and CKB which served as controls for the present study revealed unexpected results in their transcriptomic profiles. To identify the commonalities and potential specific genes to each treatment, Venny online software v 2.1 (https://bioinfogp.cnb.csic.es/tools/venny/) was used to obtain the intersection of IFIT2/empty vector, IFIT2/CKB, and IFIT2/empty vector/CKB. Venn diagrams showing number and percentage of overlapping genes and genes exclusive to IFIT2 M. tb infected THP-1 cells can be found in Supplementary Figure 7 . For IFIT2/empty vector comparison, the result showed that 20 genes were common to both, 50 genes were exclusive to IFIT2 and 3 genes exclusive to Empty Vector ( Supplementary Figure 7A ). IFIT2/CKB comparison results also revealed that 50 genes overlapped, 20 genes were IFIT2 specific genes and 5 were specific to CKB ( Supplementary Figure 7B ). Lastly, for the three comparisons, the results showed that there were 20 DEGs common to all the three groups, 20 specific genes in IFIT2, 3 specific genes in Empty Vector and 3 specific genes for CKB ( Supplementary Figure 7C ).

A protein-protein interaction (PPI) network was constructed using the STRING database to assess functional associations among the proteins encoded by the DEGs of IFIT2 induction in M. tb-infected THP-1 cells. Each node represented a DEG, while connections between nodes indicated established interactions ( Figure 7 ). This analysis provided insight into the functional relationships among the DEGs, enhancing our understanding of their roles in TB infection. Within the PPI network, several hub genes were identified based on their high degree of connectivity. Key hub genes include ISG15, RSAD2, IFIT3, IFI44L, suggesting their central roles in mediating interactions among DEGs induced by M. tb infection.

Gene expression levels were quantified using RT-qPCR for five selected key DEGs identified downstream of IFIT2 which includes, CMPK2, RSAD2, ISG15,IFI44L and IFI44. The qPCR results were consistent with the AmpliSeq data, demonstrating robust fold changes and statistical significance (p < 0.05) albeit not for ISG15 and RSAD2 ( Figure 8 ). This concordance reinforces the reliability of our transcriptomic analysis and highlights the potential role of these interferon-stimulated genes (ISGs) in the antimycobacterial activity of IFIT2. The enhanced expression of these genes suggests their potential contribution to the host’s immune defense mechanisms, offering promising targets for future therapeutic strategies against TB.