In the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/), use “Lupus nephritis” as the search term, limit entry type to series, and detection method/data type to expression profiling by array, the organization source is homo sapiens, and all gene expression data related to LNs were retrieved as of October 3rd, 2022. The data sets were differentiated for tissue origin: peripheral blood and kidney tissue. Three eligible gene expression datasets were screened in the renal group: GSE32591 (64 LN, 29 healthy controls), GSE113342 (56 LN, 16 healthy controls), and GSE200306 (79 LN, 19 healthy controls), enrolling a total of 199 patients with LN and 64 healthy controls. Three eligible gene expression data sets were screened in the peripheral blood group: (GSE72798 (30 LN, 17 healthy controls), GSE99967 (29 LN, 17 controls), and GSE81622 (14 LN, 24 controls)’s, which included a total of 73 LN patients and 58 healthy controls.

Firstly, convert the probe matrix into a gene matrix according to the probe annotation file. For multiple probes corresponding to the same gene, the average value of the probes is calculated as the final expression value of the gene. Secondly, the two data sets with larger sample size were pooled and used as training groups, GSE99967, GSE72798 and GSE32591, GSE113342 for peripheral blood and kidney tissues respectively, and GSE81622 and GSE200306 for the validation group. The sva package was used to process the batch effect, the limma package was used to analyze theDEGs in the matrix, and the DEGs were selected with the gene expression | log2 fold change, logFC | > 1 and the adjusted P value <0.05 as the screening conditions. Finally, ggplot was used to visualize the DEGs and plot the corresponding volcanoe map.

The “clusterProfiler” and “DOSE” packages of the R package were used to perform disease ontology (DO) enrichment analysis of DEGs. And gene set enrichment analysis (GSEA) was also used to identify significant pathways between the LN and control groups. The “c2.cp.kegg.v7.4.symbols.gmt” from the Molecular Signatures Database (MSigDB) was used as the reference gene set. If p. adjust < 0.05, the pathways were considered significantly enriched.

Machine learning is a novel algorithm analysis tool, and in this study, the least absolute shrinkage and selection operator (LASSO) combined with support vector machine-recursive feature elimination (SVM-RFE) was applied to the screening of LN biomarkers to predict the status of the disease. LASSO is a regularized regression algorithm using the “glmnet” package in R (15). SVM-RFE is a supervised machine learning technique that can rank features based on recursion to avoid overfitting (16, 17). The genes screened by the two algorithms are intersected and the overlapping genes are the candidate bio-diagnostic markers.

In order to test the accuracy of the biodiagnostic markers screened by machine learning, we generated ROC curves using the “diff gene xp” between the LN group and the normal sample group. The greater the area under curve (AUC), the higher the accuracy of the gene as a diagnostic marker in LN. In the same method, its effectiveness was further verified in the validation group GSE81622/GSE200306.

The CIBERSORT algorithm is a reliable machine learning method based on linear support vector regression (SVR), which is widely used to assess the relative content and dynamic regulatory process of 22 immune cells and is superior to other methods for identifying human immune cell phenotypes in terms of noise and unknown mixed content (18, 19). In order to understand the relative proportions of different immune cells in the LN sample, this study used the R language program and linked CIBERSORT to simulate and calculate the transcription feature matrix of 22 immune cells, and the number of calculations was set to 1,000. Based on the P value<0.05, use the “corrplot” package to draw correlation heat maps to visualize the correlation between 22 infiltrating immune cells; use “vioplot” to draw a violin chart to visualize the differences between the immune cells of the LN group and the control group.

Correlation analysis between genes and immune cells was performed using Spearman correlation coefficient. For those with p < 0.05, the results were visualized using the “ggplot2” package.

To explore the correlation of candidate gene markers with different pathological types, relapse frequency, renal tubulointerstitial inflammation, complex immune accumulation, and renal function in patients with lupus nephritis based on current relevant literature and the Nephroseq V5 tool (https://nephroseq.org/).

For continuous variables between two groups, t-test was applied if they conformed to normal distribution; Mann-Whitney U-test was used for non-normal distribution. ANOVA was applied for continuous variables between three groups. Pearson’s analysis was used to analyze the correlation between gene expression and immune cell fraction. ROC curve analysis was used to determine the diagnostic performance of the diagnostic indicators identified in the study. All statistical analyses were performed using R software (version 4.1.2) and SPSS (version 27.0) software. All statistical analyses were two-sided with P < 0.05 were regarded statistically significant.

According to the screening criteria adjusted P value <0. 05 and |log fold change(FC)| >1, a total of 30 DEGs were screened in the LN kidney group, of which 25 were significantly up-regulated, including C1QB, MX1, C1QA, IL32, etc.; 5 were significantly down-regulated, including ZBTB16, FKBP5, DEFB1, RORC, EGR1. Volcano map of DEGs ( Figure 1A ). 21 DEGs were obtained from LN peripheral blood, all significantly up-regulated, volcano map drawn for DEGs ( Figure 1B ).

A total of two machine learning algorithms, LASSO regression and SVM-RFE, were used to screen biomarkers in this study. In the renal group, LASSO regression identified 16 biomarkers ( Figure 3A ), SVM-RFE algorithm established 6 biomarkers ( Figure 3B ), and the overlapping genes C1QA, C1QB, MX1, RORC of both algorithms were used as diagnostic biomarkers ( Figure 3C ). To further validate the accuracy of the diagnostic biomarkers, their expression levels were further verified in the validation set GSE200306. C1QA, C1QB, MX1, and RORC were expressed at significantly higher levels in LN than in controls (p < 0.05) ( Figure 4A–D ). In the peripheral blood group, LASSO regression identified 7 biomarkers ( Figure 3D ) and SVM-RFE algorithm established 4 biomarkers ( Figure 3E ). Eventually, three overlapping genes CD177, DEFA4, and HERC5 from both algorithms were taken as diagnostic biomarkers ( Figure 3F ). To further validate the accuracy of diagnostic biomarkers, they were further validated for differences in the validation set GSE81622. All three diagnostic genes had significantly higher expression levels in LN than in controls (p < 0.05) ( Figure 5A-C )

As shown in Figure 6 , all six biomarkers screened had a high diagnostic value in LN. In the kidney tissue training group, the AUCs of the selected biomarkers were: C1QA (0.948, 95% CI: 0.909−0.978), C1QB (0.926, 95% CI: 0.885 − 0.962), MX1 (0.881, 95% CI: 0.823 − 0.925), and RORC (0.819, 95% CI: 0.740 − 0.889), as shown in Figure 6A ; in the validation group, the AUCs of the biomarkers were C1QA(0.741, 95% CI: 0.634−0.836), C1QB(0.758, 95% CI: 0.646−0.856), MX1(0.865, 95% CI: 0.775−0.944), and RORC(0.911, 95% CI: 0.846−0.962), as shown in Figure 6B . The AUCs of the selected biomarkers in peripheral blood in the training group were: CD177 (0.791, 95% CI: 0.685 − 0.878), DEFA4 (0.812, 95% CI: 0.707 − 0.905), and HERC5 (0.775, 95% CI: 0.654 − 0.884), respectively ( Figure 6C ); the AUCs of the selected biomarkers in the validation group were: CD177 (0.885, 95% CI: 0.752 − 0.979), DEFA4 (0.843, 95% CI: 0.685 − 0.965), and HERC5 (0.880, 95% CI: 0.725 − 0.987), respectively ( Figure 6D ). The biomarkers showed a high diagnostic value in both the LN training and validation groups, with AUCs largely ≥0.8.

Differential analysis of immune cells in LN and normal samples showed that T cells regulatory (Tregs) (p = 0.017) were significantly lower and NK cells resting (p = 0.019) were significantly higher in LN kidney tissue than in the normal group (see Figure 7A ). In LN blood tissues, B cells naive (p<0.001), B cells memory (p<0.001), T cells CD4 memory resting (p=0.007), Monocytes (p=0.006), Macrophages M1 (p=0.011), Macrophages M2 (p=0.039), Dendritic cells activated (p<0.001) were significantly higher than the control group ( Figure 7C ).

In addition, the correlation of 22 immune cells in all samples was analyzed ( Figure 7B, D ). There was a significant positive correlation between NK cells resting and T cells CD4 memory activated (r = 0.66) and a significant negative correlation between Mast cells activated and MacrophagesM1 (r = − 0.59) in LN kidney tissue and normal group samples ( Figure 7B ). In LN peripheral blood and normal group samples, T cells CD8 showed a significant negative correlation with Neutrophils (r=-0.49); T cells CD4 memory activated showed a significant positive correlation with T cells gamma delta (r=0.57) ( Figure 7D ).

In kidney tissue, C1QA was positively correlated with T cells CD8 (R= 0.56, P= 7.94E-05)、Macrophages M1(R= 0.47, P= 0.0016) and Mast cells resting(R= 0.32, P= 0.032); and negatively correlated with Mast cells activated (R= -0.40, P= 0.0087), T cells CD4 naïve (R= -0.39, P= 0.0088) and NK cells activated(R= -0.32, P= 0.033) ( Figure 8A ). C1QB was positively correlated with MacrophagesM1 (R = 0.39, p = 0.0099), Neutrophils (R = 0.34, p= 0.024), and TcellsCD8 (R = 0.44, p = 0.0029); and negatively correlated with Dendritic cells resting (R = − 0.36, p= 0.017) and Mast cells activated (R= − 0.39, p= 0.01) ( Figure 8B ). MX1 was positively correlated with Eosinophils (R = 0.52, p = 0.00037), T cells CD8 (R = 0.35, p = 0.023), T cells follicular helper (R = 0.45, p = 0.0023); and negatively correlated with Dendritic cells resting (R = − 0.33, p = 0.032), TcellsCD4naive (R= − 0.53, p = 0.00026) ( Figure 8C ). RORC was positively correlated with Dendritic cells resting (R = 0.49, p = 0.00084); negatively correlated with B cells naive (R= − 0.42, p= 0.0056) ( Figure 8D ), and C1QB, MX1, and RORC were all correlated with Dendritic cells resting.

In peripheral blood tissues, CD177 was positively correlated with Macrophages M0 (R = 0.38, p = 0.00037), Macrophages M1 (R = 0.3, p = 0.0059), Macrophages M2 (R = 0.3, p = 0.0054), Monocytes (R = 0.47, p = 6.8e − 06), Neutrophils (R = 0.33, p = 0.002), and negatively correlated with T cells CD4 memory activated (R= − 0.31, p = 0.0042), T cells CD4 memory activated (R = − 0.39, p = 0.00026), and T cells CD8 (R = − 0.22, p = 0.042) ( Figure 8E ). DEFA4 was positively correlated with MacrophagesM0 (R = 0.32, p = 0.0032), Macrophages M2 (R = 0.3, p = 0.0051), Monocytes (R = 0.42, p = 6.1e − 05), and negatively correlated with T cells CD4 memory activated (R = − 0.24, p = 0.03), T cells CD4 memoryresting (R = − 0.31, p = 0.0039) ( Figure 8F ). HERC5 was positively correlated with B cells memory (R= 0.42, p=7.6e-05), Dendritic cells activated (R = 0.65, p = 2.2e − 11), Plasma cells (R = 0.4, p = 0.00017), and negatively correlated with B cells naive (R= − 0.3, p = 0.0056), T cell CD4 memoryresting (R = − 0.34, p = 0.0016) ( Figure 8G ). CD177, DEFA4, HERC5 were all associated with T cells CD4 memory resting.

To determine the clinical applicability of these markers, we consulted the Nephroseq database and related literature to obtain correlations between seven immunomarkers and clinicopathology. The summary is shown in Table 1 and is confirmed below.

In this study, DEGs in kidney tissue and peripheral blood of LN were analyzed by combined GEO database multi-chip. A total of 30 DGEs were screened in kidney tissues, 25 of which were upregulated, mainly CD163, CFH, CX3CR1, HLA-DPB1, HLA-DQB1, IFITM1, IL10RA, etc.; 5 were downregulated, mainly RORC, etc. A total of 21 DGEs were screened in LN peripheral blood, all of which were significantly upregulated. The main ones were IFI44L, LCN2, CD177, etc. DEGs were all closely related to LN.

Professor Xueqing Yu of Sun Yat-sen University found that several independent susceptibility loci such as HLA-DPB1 and HLA- DQB1 are closely associated with the development of LN (118). Also, HLA- DQB1 is a susceptibility gene for SLE (119). Differentiated antigen cluster 163 (CD163) is expressed in monocytes/macrophages, a marker of macrophage activation, and has the ability to mediate the clearance of free hemoglobin by macrophages, thereby reducing kidney damage from the reaction of heme iron with endogenous hydrogen peroxide to generate free radicals (120, 121). It has the ability to modulate immunity, scavenge tumor necrosis factor-like weak apoptosis-inducing factors, activate inflammatory responses, and induce cell proliferation and apoptosis (122, 123). Several studies have shown that CD163+ macrophages infiltrate significantly in patients with LN (124, 125), and the number of CD163+ macrophage infiltration is closely related to the activity of LN (125). Studies on other macrophages are relatively scarce and need to be further explored in the future. CX3CR1 is widely expressed in the renal tissue of patients with LN and can be used as one of the alternatives to renal biopsy (126). CX3CR1 is involved in the pathogenesis of human type IV LN (127). Administration of an analogue of fractalkine as an antagonist in MRL/lpr lupus mice significantly inhibited the development and progression of their LN (128). An investigation by Professor Wang Fengmei of Peking University showed that serum CFH levels were closely related to LN disease activity and that CFH deficiency accelerated the development of LN (129). CFH-related genes are also associated with SLE susceptibility (130). IFN is a major pathogenic factor in systemic lupus erythematosus (SLE) and LN (131). More than 20 cross-sectional case-control studies have shown that the expressions of IFI27, IFI44, IFI44L, IFIT1, PRKR and RSAD2 are proportionally clustered in LN (131). IFI44L promoter methylation can be used as a blood biomarker for LN (132–134). The protein encoded by IL10RA is a receptor for interleukin 10 (IL-10) (135). Overexpression of IL-10 in LN renal tissue and peripheral blood is highly correlated with the onset and activity of LN (136, 137). FKBP5 and EGR1 are closely associated with SLE, but the direct correlation with LN needs to be further confirmed (138, 139).

The interferon-inducible transmembrane protein (IFITM) gene family includes IFITM1, which plays a role in a number of biological activities such as interferon-homotypic cell adhesion function and anti-proliferative activity of cells (140–143). Lipocalin-2 (LCN2) is an indicator of the severity of LN and plays a key role in the immune response, which promotes Th1 cell differentiation exacerbating LN in an autocrine or paracrine manner, mainly through the IL-12/STAT4 pathway (144). Urinary LCN2 can be used as an early biomarker of LN (145, 146). RSAD2 is an interferon-inducible iron-sulfur cluster-binding antiviral protein. In a study of genome-wide DNA methylation in LN, methylation of the CpG site of the RSAD2 gene was shown to be highly correlated with LN (147). TGFB1 and MMP8 may be involved in LN through a mediated matrix pathway (148). The relationship between diagnostic biomarker genes and LN is detailed in the biomarkers section.

The disease enrichment analysis of LN kidney tissue DGEs was mainly enriched in immune class, inflammatory, renal and cardiovascular diseases, such as: primary immunodeficiency disease, vasculitis, renal failure, renal cancer, and renal disease. This fits in previous studies. In terms of the pathogenesis of SLE, multiple factors such as multiple genes and environmental factors are involved (149), and viral infections are often used as environmental triggers of SLE (150). In other words, LN is essentially an immune-mediated inflammatory response. In terms of the symptoms and outcomes of LN, there is a significant correlation between LN and cardiovascular disease (151), and 20.8% of deaths in LN patients are caused by cardiovascular disease (152). Many LN patients end up with chronic kidney disease (CKD) and ESRD despite receiving potent anti-inflammatory and immunosuppressive therapy.

Differential genetic disease enrichment in peripheral blood is mainly associated with hematologic disorders, bone marrow. is mainly associated with hematologic disorders, bone marrow. For example, erythrocytosis, thrombocytosis, bone marrow disorders, oral diseases, etc. SLE often involves the blood system, and hematological abnormalities are a common manifestation of SLE (153). Hypercoagulation is often accompanied in LN, which may be related to its immune complexes which can promote platelet aggregation, form thrombus, and cause microcirculation disorders. One of the more controversial points is that thrombocytosis may be an indicator of disease activity and responsiveness. Because about a quarter of SLE patients usually show thrombocytopenia, thrombocytosis is an unusual finding. LN can also be accompanied by oral ulcers. For example, a retrospective study conducted by a tertiary medical center in Assam, India, showed that 176 patients with LN had a 31.8% probability of having a concomitant oral ulcer (154). SLE, especially patients with LN, is often accompanied by osteoporosis, which may be related to the long-term use of prednisone (155). Ramsey et al., 1994, in the US National Health Survey, found that the incidence of fractures in patients with SLE was five times higher than in the normal population (156).

GSEA results indicate that the immune response plays a crucial role in LN. GSEA enrichment of kidney tissue mainly involved immunity with inflammation. For example, the five pathways of autoimmune thyroid disease, cell adhesion molecules cams, systemic lupus erythematosus, type 1 diabetes mellitus, and viral myocarditis were activated. Cell adhesion molecules (CAM) play a key role in inflammation, immune response, and thrombosis (157). Intercellular adhesion molecule-1 (ICAM-1) and vascular cell adhesion molecule-1 (VCAM-l) are two CAMs that are currently receiving the most attention. In the autoimmune MRL/lpr model in LN mice, it was found that the severity of renal histopathology was correlated with the expression of VCAM-1, and ICAM-1 was also used as a major marker of renal inflammatory activity (158). Blocking this pathway may inhibit the pathological infiltration of LN inflammatory cells (159). In clinical trials, the expression of ICAM-1 and VCAM-1 in glomerular endothelial cells and cells, serum and urine of LN patients was significantly increased, which was positively correlated with LN activity and played an important role in the development of LN (160–162). At present, anti-adhesion molecule antibodies have achieved a certain effect in rheumatism (163), vascular remodeling, heart and cerebral ischemia-reperfusion (164, 165) and other aspects, and the field of nephropathy urgently needs to be developed. The results of LN peripheral blood GSEA enrichment analysis showed that the pathways were mainly involved in immune, cell cycle. For example, cell cycle, cytosolic DNA sensing pathway, NOD-like receptor signaling pathway, proteasome, and RIG-1-like receptor signaling pathway were activated. Inflammatory response or cellular injury triggers abnormal proliferation of LN tethered cells, activates the cell cycle and ultimately leads to glomerulosclerosis and end-stage renal failure. Interference with the cell cycle can prevent cell proliferation or promote apoptosis (166). For example, tacrolimus and motilmic acidophilus, commonly used drugs in LN, inhibit the proliferation of thylakoid cells by acting on different phases of the cell cycle of thylakoid cells and thus slow down the progression of glomerular disease (167, 168). The transmission mechanism of cytoplasmic DNA is the molecular basis for the immune system to produce an immune response. Aberrant self-DNA accumulation and cellular type I interferon (IFN-I) secretion are the two main parts in the pathogenesis of LN (169, 170), while cellular sensors play an important role in the process of extracellular DNA triggering IFN-I (171). Various DNA sensors have been reported so far, but specific cytoplasmic DNA sensors involved in LN’s own DNA-induced autoimmune response are yet to be further investigated. Nucleotide-binding oligomerization domain NLR and RIG-1 like receptors are involved in the pathogenesis of SLE/LN, which mainly stimulate cytokine production by activating innate immunity through the uptake of immune complexes (ICs) between antinuclear antibodies and nucleic acids into cells (172, 173). In addition, the NLR family also drives key molecules of the inflammatory response through the formation of “inflammasomes”, resulting in transcriptional induction of type 1 IFNs and other inflammatory cytokines (150). This information is essential to improve therapeutic strategies for LN, such as the production of drugs that target interferon or interferon receptors. Proteasome inhibition is a new therapy for LN and plays an important role in the treatment of LN, such as cyclophosphamide, rituximab, and bortezomib (174).

Based on two machine learning algorithms, LASSO and SVM-RFE, four biomarkers with diagnostic value were screened in kidney tissues: C1QA, C1QB, MX1, and RORC. C1q (C1qA, C1qB and C1qC) is the first component of the classical pathway of complement activation and plays an essential role in SLE. In recent years, domestic and international studies have shown a significant correlation between C1q deficiency and abnormal C1qAb and the onset, disease activity and pathological type transformation of SLE, especially LN (42, 175, 176). Genetic defects in C1q are susceptibility factors for SLE, and close association of LN with C1q deficiency has been demonstrated in both human SLE and C1q-deficient animal models (4, 175, 175, 177–179). Based on a systematic review of reported cases, approximately 75-93% of patients with C1q deficiency have SLE or SLE-like disorders (74, 180). In 1998, in a review of patients with C1q deficiency, Walport et al. found that 38 of 41 subjects (93%) exhibited a clinical syndrome closely related to SLE (57). In 2011, Schejbel et al. found that 88% of the 63 patients reported with C1q deficiency had SLE/SLE-like disease (181). In 2014, Mihaela Stegert et al. reviewed the overall prevalence of SLE/SLE-like syndrome in all 71 patients with C1q deficiency reported as 77.5% (182). In the C1q-deficient SLE mouse model (MRL/Mp-+/+), C1q deficiency in mice can also lead to autoimmune diseases such as SLE (177). On the one hand, it is due to the impaired phagocytic clearance of apoptotic cells, and on the other hand, C1q deficiency significantly enhances the production of anti-C1q antibodies, leading to the development of proliferative immune complex glomerulonephritis (183, 184). It explains the lack of association of C1q with SLE/LN (177). The association between C1QA and C1QB in the renal biomarker screen may further confirm the importance of the complement cascade in LNs. According to related studies complement cascade activation in SLE patients leads to hypocomplementemia and deposition of complement components at sites of tissue damage (4, 184). Human mucovirus resistance protein 1 (Mx1), one of the type I interferon (IFN)-inducible genes (27), is a type I IFN-dependent transcript that plays a major role in apoptosis and cytokine-mediated cell signaling. In LN, MX1 is used as a potential marker for diagnosing peripheral blood LN activity (25, 28) and is also considered as a susceptibility gene for SLE (25, 28). RORC is a key factor in coordinating the transcription of genes encoding IL17 and plays a key role in the regulation of inflammatory response (185). Studies have confirmed that it plays an important role in the dysregulated immune response associated with SLE (50).

Three biomarkers with diagnostic value were screened in peripheral blood: CD177, DEFA4, HERC5. CD177 is an important neutrophil gene encoding neutrophil membrane glycoprotein, but its function in neutrophils is not fully understood. Neutrophils play an active role in driving autoimmune response and tissue damage in SLE. We speculate that CD177 is mainly involved in the pathogenesis of SLE/LN by regulating NET and/or neutrophils (186). Current studies of CD177 are mainly focused on anti-neutrophil cytoplasmic antibody (ANCA)-related systemic vasculitis (187). In the future, we should pay attention to related research to fill the gap in LN. DEFA4 is expressed at a higher level in patients with active SLE (188), and is involved in immune-mediated tissue damage caused by SLE autoantibody deposition (189). In addition, DEFA4 is related to early granulocyte production and plays a role in the maturation of neutrophils and the formation of neutrophil extracellular traps (NETs). In other words, DEFA4 may be involved in the pathogenesis of LN by affecting neutrophils (37). HERC5 is an interferon-inducible gene (190–193). CoitP et al. found that HERC5 is hypomethylated in lupus patients with renal involvement (194), and Lingling Shen et al. further validated the crucial role of HERC5 in the pathogenesis of LN by PCR (qRT-PCR) and enzyme-linked immunosorbent assay (ELISA) (134).

Analysis of LN samples and normal samples using CIBERSORT revealed that a variety of immune cell subtypes were closely related to important biological processes in LN. In kidney tissue, T cells regulatory (Tregs) infiltration was reduced; NK cells resting infiltration was increased; in blood samples, B cells memory, T cells CD4 memory resting, Monocytes, Macrophages M1 and Macrophages M2 Dendritic cells activated infiltration was increased; B cells naive infiltration was decreased. The screened biomarkers were correlated with infiltrating immune cells, and the biomarkers C1QB, MX1, and RORC were correlated with Dendritic cells resting in kidney tissues. Peripheral blood tissue CD177, DEFA4, HERC5 were correlated with T cells CD4 memory resting. The specific mechanism can be traced back to the fact that LN is often triggered by abnormal immune system, involving a variety of immune cells, cytokines and related pathways (195). T cells e.g., CD4⁺, CD8⁺) play a central role in the pathogenesis of LN (196). In animal experiments, CD4 was shown to induce LN when infused into lupus-prone mice with CD4⁺ T cell lines (197). In the clinic, Masutani et al. found that CD4+ cells were the predominant cell type in grade I, IV and V renal infiltrates in patients with lupus nephritis (198, 199). CD4⁺ T is involved in LN pathogenesis and has been demonstrated to mediate LN inflammation (200), which exerts its function mainly by secreting cytokines (e.g., IFN-γ, IL-17 and IL-10, etc.) upon antigen activation and by transmitting tissue inflammation (201). In addition, it was reported that CD8+ T lymphocyte infiltration was predominant around the glomeruli of patients with LN, and the infiltrating interstitial cells were mainly CD4+ve T cells or CD8+ve (199, 202–205). Nataly Manjarrez-Orduño et al. found that a subpopulation of LN exhibited enhanced CD8+ terminal differentiation, suggesting a pathological role for CD8+ T cells in a specific subpopulation of LN and laying the groundwork for CD8-targeted therapy (206).

At the same time, relevant studies have shown that RNA sensing by traditional dendritic cells (DC) is central to LN development (207). Dendritic cells are the most potent antigen presenting cells known so far and are essential in the induction and maintenance of immune tolerance (208). DC have two subpopulations of myeloid dendritic cells (mDCs) and plasmacytoid dendritic cells (pDCs) (209). Regardless of which subpopulation is dysregulated, it will lead to tolerance disruption and self-attack, resulting in the development of LN (210). Both peripheral DC subpopulations are expected to be alternative diagnostic tools for invasive biopsy. Current SLE therapies, both conventional and biological, target specific subsets of adaptive immune cells, primarily B cells, T cells, and B and T cells are still considered to be the culprits of immune dysregulation in SLE (211), but DCs are paramount in establishing and maintaining peripheral tolerance (212, 213). DCs have a superior ability to acquire and process antigen for delivery to T cells and express high levels of co-stimulatory or co-inhibitory molecules that determine immune activation or non-response. This further highlights the importance of developing DCs in LN therapy, such as the administration of DC vaccines that may bring new hope for the treatment of LN (214).

Although this study used a large GEO dataset of LN kidney tissue and peripheral blood to explore immune cells and immune-related genes in LN and identified six diagnostic gene biomarkers with diagnostic and predictive value in patients with LN, there are still some limitations to be addressed. First, the study was retrospective and did not allow for first-hand, clinically important information. Second, the CIBERSORT algorithm has limitations in distinguishing Th17 from other CD4+ T cell subpopulations and cannot further analyze the correlation between the CD4+ T subpopulation Th17 and LN biomarkers. Third, bioinformatic determination of LN biomarkers and the function of immune cell infiltration requires further confirmation in a larger sample size prospectively.