We first searched the GEO database GSE6740 and downloaded the gene expression profiles from five individuals each with acute HIV infection, chronic progressive HIV infection, and no HIV infection. These data were reported and deposited to the GEO database by Hyrcza et al. (2007).

WGCNA is a widely used data-mining method in genomic applications. We used the WGCNA software package in the R environment (R Foundation for Statistical Computing, Vienna, Austria) to construct a co-expression network of differentially expressed genes between HIV-infected and non-infected individuals. The algorithms were used to calculate the correlations between the levels of differentially expressed genes after the selection of an appropriate threshold (β), and then a scale-free network was constructed. We used the minimum value of β greater than 0.85 as the most suitable threshold, and then used the topological overlap matrix (TOM) (direct correlation + indirect correlation) between genes for hierarchical clustering to construct a clustering tree, which contained different gene modules represented by different colours. In this process of module identification, we set the minimum number of genes contained in the module to 50 (Zhang and Horvath, 2005).

The number of metabolic genes enriched in a particular pathway was combined with the expressional values of the genes using the following formulas: E i j = ∑ k = 1 n j g i k n j (1) where E _ij indicates the average expression level of the ith gene in the jth cell type, g _ik indicates the expression level of the ith gene in the kth sample, and n _j indicates the number of samples in the jth cell type; r i j = E i j 1 N ∑ j N E i j (2) where N indicates the number of cell types, and r _ij represents the ratio of the average expression level of the ith gene in the jth cell type to the average level of the gene in all cell types. A ratio greater than 1 indicates that the gene expression level in the cell is higher than the average expression level in all cells, and a ratio below 1 indicates the opposite pattern; and S t j = ∑ i = 1 m t W i × r i j ∑ i = 1 m t W i (3) where s _tj indicates the score value of the tth pathway in the jth cell type (the higher the score, the stronger the significance), m _t indicates the number of genes in the tth pathway, and W _i represents the weighted value of the ith gene (the reciprocal of the tth pathway metabolic gene) (Xiao et al., 2019).

Using clusterProfiler (an R package), we performed pathway enrichment analysis of the differentially expressed genes with respect to Gene Ontology terms (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, using default parameters. We then constructed a protein-protein interaction network based on these data using the STRING database in Cytoscape version 3.8.2. We also identified the chromosomal localisation of all genes in the target pathway by using the Ensembl and Genecards websites.

In our study, the sample characteristics were divided into three stages as follows: Healthy, HIV acute infection, and HIV chronic infection. Then we choose the appropriate threshold (Supplementary Figure S1). We combined the relevant traits and modules of the sample for joint analysis to show the correlation between modules and traits using WGCNA’s systems biology method. Different modules were represented by different colours. Each module contained a set of highly connected genes, and the genes in each module might participate in similar pathways or have the same biological functions. These modules ranged from large to small according to the number of genes that they contained.

The results of co-expression network analysis are shown in Figure 1. The number of genes in the module is shown in Figures 1B,E and Supplementary Table S1. The correlation coefficient of each module is shown in Figures 1C,F. We identified 23 co-expressed gene modules for CD4⁺ T cells (Figure 1A). In terms of the number of genes contained in the module, MEblue was the largest module, containing 2,786 genes, whereas MEgrey was the smallest module, containing 13 genes. From the perspective of the correlation coefficient after infection, the module with the largest negative correlation coefficient was MEskyblue, with a value of −0.65; the module with the largest positive correlation coefficient was MEorangered4, with a value of 0.77. Each module has different functions. For example, Leukocyte transendothelial migration and FoxO signaling pathway are enriched in the MEskyblue.

We confirmed 20 co-expressed gene modules in CD8⁺ T cells (Figure 1D). From the perspective of the number of genes contained in the module, MEturquoise was the largest module, containing 3,214 genes; MEgrey was the smallest module, containing only one gene. From the perspective of the correlation coefficient after infection, the module with the largest negative correlation coefficient was MEdarkgrey, with a value of −0.6; the module with the largest positive correlation coefficient was MEbrown, with a value of 0.8. In addition, the coefficient of the MElightgreen ranked second, reaching 0.72. RNA transport and Viral carcinogenesis are enriched in the MElightgreen.

By calculating the correlation between the gene modules and the phenotype matrix, the key modules were screened out (i.e. those exhibiting higher correlations). The “salmon” module and the “orangered4” module were selected to represent the genes affected during early and chronic infection in CD4⁺ T cells, respectively, whereas the “brown” module and the “skyblue” module were selected to represent early and chronic infection in CD8⁺ T cells, respectively. We draw scatter plots of the relationship between gene saliency and module membership in the four modules respectively (Figure 2). Functional annotation of the key modules showed a distinct biological significance bias for each module (Supplementary Table S2) for example, the “salmon” module (early infection of CD4⁺ T cells) was significantly enriched in inactivation of metabolic and infection-related pathways (Figure 2A), whereas the “orangered4” module (chronic infection of CD4⁺ cells) was most significantly enriched in the TGF-beta signaling pathway and IL-17 signaling pathway, among others (Figure 2B). By contrast, the cell cycle and DNA replication were activated in the early infection CD8⁺ T cells (Figure 2C), whereas the proteasome and sphingolipid metabolism were largely activated in CD8⁺ T cells in the chronic infection stage (Figure 2D).

We focused on the similarities and differences in the activated gene pathways between CD4⁺ and CD8⁺ T cells during early infection. There were 20 special pathways among CD4 cells and 70 special pathways for CD8 cells. In total, 34 gene pathways (Figure 3A) overlapped between the two cell types. Among them, the top pathways mainly included the following: metabolic pathways, human immunodeficiency virus 1 infection, NOD-like receptor signaling pathway, and cAMP signaling pathway, among others. In addition to metabolic pathways and HIV infection pathways, two pathways with broad significance, NOD-like signaling pathways rank the top among other pathways. So we further focused on the NOD-like receptor (NLR) pathway. NLRs are type pattern recognition receptors for the host, which can recognise the pathogen-related molecular patterns of viruses to regulate antiviral innate immune signaling pathways, thereby regulating the innate antiviral immune response (Zheng, 2021). We first identified the chromosomal localisations for all of the genes enriched in the NLR pathway (Figure 3B), as well as the expression modules showing consistent direction of change (down- or up-regulated) in the two cells in the context of early and chronic infection (Figures 3C–F). The modules that were co-up-regulated in CD4 and CD8 cells after infection included IRF9, IRF7, PLCB3, CXCL3, and TXN, among others. The co-down-regulated modules we identified included JAK1, PKN2, TNF, NLRP1, RIPK1, and RIPK2, among others. Among them, three core modules were selected, including NLR family pyrin domain containing 1 (NLRP1), receptor-interacting serine/threonine kinase 1 (RIPK1), and RIPK2. By constructing a protein–protein interaction network, we identified the genes that might interact with each other in the three modules (Figures 3G–I).

Differential activation of cell metabolism between CD4⁺ and CD8⁺ T cells before and after HIV infection.

There are a total of 9,700 genes in our data set, and 1,352 genes related to metabolism are collected. These metabolic genes come from metabolic pathways. We selected 296 genes expressed by both as the most basic metabolic genes in this study. Since CD4⁺ and CD8⁺ T cells play different roles in the immune response, we assessed the differential activation and characteristics of these two cell types and screened their co-expressed metabolic genes. The cell metabolism pathways were differentially activated before and after infection. HIV infection also appears to disrupt the metabolic balance between these two cell types; some metabolic pathways were activated, whereas others remained unchanged. As shown in Figure 4, the pathways that were significantly altered in both types of cells in early infection included valine, leucine, and isoleucine degradation; beta-alanine metabolism; and PPAR signaling pathways. These three pathways also showed the greatest change in CD4⁺ T cells between the early infection and chronic infection stages. Among them, the PPAR signaling pathway also has significant changes in pathogen infections such as ZIKV and Neisseria meningitidis. More HIV-induced metabolic abnormalities were detected in CD8⁺ T cells compared with those occurring in CD4⁺ T cells. Before and after infection, the majority of pathways were changed in CD8⁺ T cells, along with some pathways such as the arachidonic acid metabolism pathway that was unchanged. In addition to the three most significant pathways mentioned above, the oxidative phosphorylation, fatty acid metabolism, and lysine degradation pathways also exhibited relatively large changes in CD8⁺ T cells between the acute and chronic infection stages. Thus, changes in these metabolic pathways may be conducive for these two cell types to cope with HIV infection.