Tuberculosis (TB) is a highly contagious respiratory disease caused by Mycobacterium tuberculosis (M. tb) infection that poses a severe threat to public health globally (Cohen et al., 2019). The interaction between pathogenic virulence and organismal immunity largely dictates TB progression. After M. tb infection, a series of complex immune processes, most cases enter a latent state through the long-term co-existence with the pathogens. M. tb invasion and consequent organism damage initiate inflammatory responses in cases of immune dysfunction, leading the latent state to progress to an active state. Globally, around 5%–15% of latent TB cases eventually progress to active disease (Luo et al., 2019), contributing to one of the leading causes of death worldwide (World Health Organization, 2021). The in-depth investigation of the underlying mechanisms of TB progression has been a research priority in recent decades.

The adaptive immune responses mediated by T cells play a crucial role in the equilibrium of the host against M. tb (Jasenosky et al., 2015; Chai et al., 2020); these responses include antigen recognition, inflammatory homeostasis maintenance, and granuloma formation. Several T-cell biological processes, such as development, differentiation, and sensitization, are involved in suppressing TB activity (Feruglio et al., 2015; Huang et al., 2019). In recent years, studies based on cohort microarray data analysis have identified T-cell-related mRNAs, microRNAs (miRNAs), and long non-coding RNAs (lncRNAs) as critical biomarkers in distinguishing active from latent states (Walzl et al., 2018; Sinigaglia et al., 2020; Kundu and Basu, 2021). However, most of these markers were identified via data-driven strategies; thus, the regulating mechanisms of these genes in TB progression remain under-characterized. The results of the bioinformatics approaches applied in the present study identified a lncRNA/miRNA axis-mediated competing endogenous RNA (ceRNA) network that is involved in TB activation through T-cell regulation.

This study was designed as shown in Figure 1. We used TB samples from the GSE37250 dataset (Kaforou et al., 2013) and randomly selected a portion of them for data mining. Using integrated bioinformatics and statistical approaches, we constructed a ceRNA network mediated by MIR600HG/hsa-mir-21-5p axis that may play an important role in TB activation by regulating some CD4⁺ T subsets. To assess the clinical significance of the regulatory genes in this ceRNA network, we employed an elastic net regression model (Friedman et al., 2010) to construct a diagnostic classifier based on MIR600HG, hsa-mir-21-5p, and the targeting mRNAs. Internal validation confirmed the repeatability and evaluated the generalizability of these markers in HIV-negative and HIV-positive sub-groups of patients. As the external validation should be strictly constrained by the target population size and implementation criterion, we used independent datasets with similar scales and execution strategies to perform this assessment. The results of external validation showed the high applicability of these markers in pediatric cases. The results of this study extend our understanding of the immunological mechanisms involved in TB activation and provide novel potential diagnostic biomarkers for clinical applications (Figure 2).

CD4⁺ T cells play a central role in the defense against TB associated with adaptive immune mechanisms. Investigating the molecular immune mechanisms in the CD4⁺ T regulatory TB process contributes to efforts to individualize TB prevention and management. Herein, we employed integrated bioinformatics approaches to construct a ceRNA network. More importantly, we demonstrated the relationship between this ceRNA network and some CD4⁺ T subsets. This study was performed based on large-scale and widely accepted TB datasets, thus improving the credibility and representation of the present study. With the coronavirus disease 2019 (COVID-19) pandemic, the relationships between COVID-19 and TB have drawn increasing attention from the immune community. The influence of CD4⁺ T subsets on TB and COVID-19 is complex and multidimensional (Yang and Lu, 2020). Our results shed new light on further investigation into the mechanism of COVID-19 and other respiratory infectious diseases.

The limitations of the data mining strategy in this study were also apparent in the number of various T cells that were not directly accessible. Thus, we used the ImmuCellAI platform to estimate the abundance of T cells based on the gene set signature method. However, the most practical experience of this platform was obtained from oncology studies. We calculated the abundance of peripheral blood T cells in TB cases and observed lower levels of CD4⁺ T, Tr1, nTreg, iTreg, and Tfh in active cases. Some of these results were supported by circumstantial evidence. The present study performed an essential attempt to use ImmuCellAI in its investigation of TB immunological mechanisms, the results of which provided a theoretical reference for the analysis of the molecular mechanisms involved in the interactions between pathogens and the immune system.

The protective role of CD4⁺ T is among the most critical for inhibiting TB activity, with the depletion of CD4⁺ T promoting TB progression and increasing the risk of death (Foreman et al., 2016; Santos-Pereira et al., 2021). The number of polyfunctional M. tb antigen-specific CD4⁺ T cells in the peripheral blood demonstrated an inverse relationship with the pathogen burden in the lungs (Day et al., 2011). Tfh is an important CD4⁺ T-cell subset that supports humoral immunity and facilitates neutralizing antibody responses to control pathogens (Crotty, 2019). Although the mechanisms remain unknown, a decrease in peripheral blood Tfh levels in active cases compared to those in latent cases has been reported (Kumar et al., 2014). However, the Tregs estimations in this study were controversial. Previous small-scale studies (Semple et al., 2013; Stringari et al., 2021) suggested an increasing trend of peripheral blood Tregs counts in active TB cases compared to the numbers in latent cases. Due to the methodological limitations of the studies, the abundance of Tregs mainly represented cell activity but not necessarily the exact count. As supporting evidence, the functional enrichment results of genes positively related to Tregs indicated that the decrease of those genes in active TB probably suppressed the activation processes of Tregs rather than proliferation. Tregs attenuate inflammation during chronic infectious disease, which maintains immune homeostasis against host damage from excessive inflammation. However, the adverse effect was that pathogen clearance was suppressed (Stephen-Victor et al., 2017). Tregs play a complex role during M. tb infection; in different stages of TB progression, Tregs are redistributed between the peripheral blood and local inflammatory sites; thus, their regulatory effect might be dual (Boer et al., 2015). The regulatory details of various Treg sub-types in TB progression were not previously elucidated. Our results demonstrated the potential possibilities of Treg-related mechanisms in TB activation and called for larger-scale investigations with precise designs.

The major cluster (cluster I) genes regulated by the ceRNA network were associated with T-cell behaviors in biological processes and pathways. The enriched genes were down-regulated in active cases, which indicated their suppressed activation, differentiation, and development during TB activity, echoing the estimation of T-cell subset abundance in the present study. We found that the PD-1 checkpoint was one of the critical pathways in TB activation. PD-1 inhibitors are widely used in patients with tumors. Latent TB reactivation, as a severe adverse event, has attracted much attention (Picchi et al., 2018; Anand et al., 2020). The PD-1 pathway plays a critical role in various infectious diseases (Sharpe et al., 2007). M. tb infection elevates PD-1 expression in peripheral blood CD4⁺ T cells, which likely limits the host immune response against pathogens (Shen et al., 2016). PD-1 inhibition promotes TNF-α secretion and results in increased M. tb growth. Consistent with this, the peripheral blood expression of CD4⁺ T PD-1 declines (Tezera et al., 2020). However, the mechanisms by which the PD-1 pathway regulates TB progression remain unknown. Therefore, our findings provide novel insight for further exploration.

This study constructed a ceRNA network mediated by the MIR600HG/hsa-mir-21-5p axis and included five targeting mRNAs (BCL11B, ETS1, EPHA4, KLF12, and KMT2A). A few studies have demonstrated the function of MIR600HG, which is considered a lncRNA biomarker for predicting the progression of tumor patients (Song et al., 2018; Cao et al., 2022). Xiao et al. (2021) reported that MIR600HG, which was involved in the ceRNA network in the present study, influenced pancreatic adenocarcinoma progression by regulating immune cell infiltration. The present study investigated the regulatory role of MIR600HG in infectious diseases. Our results demonstrated the great promise of MIR600HG in immunology. hsa-mir-21-5p is considered a hot-spot miRNA in immune research associated with TB (Kozomara et al., 2019). However, opposing views debate the expression difference in active cases. Kleinsteuber et al. (2013) reported decreased hsa-mir-21 levels in active cases compared to those in latent cases. However, two independent studies reported higher hsa-mir-21 expression in active cases compared to that in healthy controls and patients receiving anti-TB treatment (Wang et al., 2018; Kathirvel et al., 2020). Previous studies revealed that hsa-mir-21-5p mitigated inflammatory responses by regulating macrophages (Zhao et al., 2019). Carissimi et al. (2014) identified hsa-mir-21 as a negative modulator of T cells and the T-cell receptor (TCR) as the critical pathway. Nguyen et al. (2021) confirmed that hsa-mir-21 regulated the cellular functions and apoptosis of CD4⁺ T through the TCR pathway in infectious diseases. The results of the present study showed revealed an hsa-mir-21-5p regulatory network presumably linked to the TCR regulating CD4⁺ T subsets during TB activation, which supplemented our understanding of the role of hsa-mir-21-5p in TB progression.

BCL11B, ETS1, EPHA4, KLF12, and KMT2A were down-regulated in active cases. ROC analysis results confirmed the diagnostic prospects. Most of these findings are consistent with those previous independent reports (Jezela-Stanek et al., 2020; Zhao et al., 2020; De Araujo et al., 2021; Natarajan et al., 2022). We also confirmed their accuracy in distinguishing active TB and, more importantly, explored the ceRNA regulatory mechanism and affected T-cell types. Complementarity prediction and expression correlated analysis revealed the potential role of the MIR600HG/hsa-mir-21-5p axis in mediating these genes. The integrated results of the co-expression analysis, PPI network construction, and enrichment indicated that the critical CD4⁺ T subsets likely influenced TB progression via a functional gene cluster containing these five genes. BCL11B and ETS1 play key roles in the development of CD4⁺ T subsets (Liu et al., 2010; Garrett-Sinha, 2013); as transcription factors, they are deeply involved in Treg and Tfh activation and differentiation (Li et al., 2010; Kitagawa et al., 2017; Kim et al., 2018). EPHA4, as a receptor tyrosine kinase, has been implicated in the mediation of cell developmental events and is associated with the maturation and development of CD4⁺ T cells (Munoz et al., 2006). KLF12 and KMT2A are gene expression-regulated transcription factors associated with T cell proliferation (Peterson et al., 2018; Parrado, 2020). The experimental evidence presented above theoretically supported our hypothesis, which proposed that decreased levels of these five mRNA biomarkers suppress T-cell activity in active TB cases.

Compared to previous data-driven studies (Sweeney et al., 2016; Singhania et al., 2018; Natarajan et al., 2022), the present study was driven by immunological mechanisms to screen for gene biomarkers. In practice, this dual strategy approach provides more useful information. The elastic net regression model was used to determine the coefficients of each feature to obtain an integrated model to assess the compounded accuracy. For revealing the mechanisms, despite sacrificing some diagnostic accuracy, our elastic net scoring classifier still showed excellent performance in distinguishing active cases. The validation results in the testing set confirmed that our findings were robust and that the classifier model was fitted appropriately. As demonstrated in this study, the MIR600HG/hsa-mir-21-5p axis ceRNA network regulated TB activation via some subsets of CD4⁺ T cells; thus, the performance of our findings in CD4⁺ T-cell-deficient populations should be assessed. In the real-world TB disease spectra, cases with HIV coinfection are probably the most representative samples with CD4⁺ T deficiency. The cases also have a higher prevalence of hospitalization and mortality compared to HIV-negative cases (Bruchfeld et al., 2015; Subbarao et al., 2015). The HIV-positive internal validation confirmed the adequate diagnostic accuracy of our hypothesis in CD4⁺ T-cell-deficient cases. Further in-depth exploration of the effect of HIV on TB activity was not within the scope of the current study. However, the five critical T-cell subsets and several gene biomarkers identified in the present study are also reportedly affected by HIV infection (Jasenosky et al., 2015; Zhou et al., 2021). The results of the present study also suggest implications for future research on the molecular mechanisms by which HIV facilitates TB activity. Our findings were based on an adult cohort; however, children living with TB deserve special attention (Marais et al., 2014). Constrained by the target population size, we were unable to enroll a new cohort; however, we performed validation in comparable high-quality datasets. The results suggested that our findings were also valid for pediatric cases. It is worth noting that the peripheral T-cell compartment could change with aging (Arsenovic-Ranin et al., 2017; Reynaldi et al., 2019). As shown in Supplementary Table S7, the diagnostic accuracy of nTreg in child cases was low. The immune effect of nTreg on tuberculosis may be affected by aging factors (Namdeo et al., 2020). The issue is worthy of further exploration.

The results of this study identified a T-cell-related MIR600HG/hsa-mir-21-5p axis ceRNA network, which likely revealed the immunological mechanisms associated with TB activation. The results of the internal and external validations confirmed that our findings applied to various populations with different backgrounds. Although some results remain controversial, we believe that this ceRNA network helps uncover the CD4⁺ T subsets associated with the regulatory mechanisms in TB activation and provides prospects for clinical applications.