Identification and validation of tissue-based gene biomarkers for acute intestinal graft-versus-host disease(AIGVHD)

Abstract

Background: Acute intestinal graft-versus-host disease (AIGVHD) is a common complication of allogeneic hematopoietic stem cell transplantation (allo HSCT) with a high mortality rate. The primary aim of the present study is to identify tissue-based gene biomarkers pertinent to AIGVHD, thereby facilitating early diagnosis and exploration of potential therapeutic targets.

Method: The dataset was obtained from the GEO database. DEGs were identified, followed by GO and KEGG pathways analysis for the common DEGs. PPI networks and WGCNA analysis were used to identify essential genes, and correlations between critical genes and immune cell infiltration were also examined. The diagnostic efficacy of these essential genes was evaluated using ROC curves, leading to the development of 11 machine learning models based on this gene set. Furthermore, we established a mouse model of aGVHD, which was identified by clinical score, pathological analysis, flow cytometry detection of implantation rate, and immunohistochemical detection of CD4 expression. Finally, we measured the mRNA expression levels of the key genes in the mice's intestinal tissue using real-time PCR.

Result: DEGs showed a marked enrichment in immune and inflammatory response pathways. Our analysis identified three key genes, FCGR3A, SERPING1, and IFITM3, which were positively associated with M1 macrophage and neutrophil infiltration. Subsequently, we developed machine learning models utilizing these three genes and found that the RF model exhibited a robust predictive capacity for AIGVHD occurrence, achieving an AUC of 0.9802 (95% CI: 0.966-0.9945). An aGVHD mouse model was also successfully created, and we discovered that the aGVHD group's mRNA expression levels of three key genes were noticeably higher than the control group's.

Conclusion: In this study, we identified FCGR3A, SERPING1, and IFITM3 as tissue-based gene biomarkers for AIGVHD, highlighting their diagnostic efficacy. Furthermore, we confirmed the association of these genes with AIGVHD through investigations conducted in aGVHD mouse models.

Keywords: Allo HSCT; aGVHD; aGVHD mouse model; gene biomarkers; immune cell infiltration; machine learning.

Figure 1

The differentially expressed genes in the data sets were screened, followed by functional enrichment analysis of the common DEGs. Candidate hub genes were determined through PPI network analysis. (A) The volcano plots of the GSE168116 dataset. (B) The volcano plots of the GSE134662 dataset. (C) The Venn diagram represents the common DEGs between two datasets. (D-F) GO functional analysis of the DEGs. (G) KEGG analysis of the DEGs. (H) PPI network analysis of the DEGs. (I) Identification of candidate hub genes utilizing the CytoNCA plug-in.

Figure 2

Use WGCNA analysis to identify modules associated with AIGVHD. (A) The mean connectivity and scale independence of eigengenes. (B) The gene clustering tree diagram uses clustering to find very similar modules and then dynamically merge them; various clusters are shown by different colors. (C) Eigenmodules adjacency heatmap. (D) The correlation heatmap between the WGCNA module and clinical features. The correlation coefficient and associated p-value are displayed in each column, and positive and negative correlations are denoted by red and blue, respectively. The correlation coefficient increases with color darkness.

Figure 3

Evaluation and validation of critical genes and the construction of machine learning models. (A) Use a Venn diagram to represent the intersection of candidate genes from the PPI network and the WGCNA analysis. Obtain three key genes: IFITM3, SERPING1, and FCGR3A. (B) The ROC curves of the three critical genes in the training cohort. (C) The ROC curves of the three critical genes in the validation cohort. (D) The ROC curves of the 11 machine learning models, which were built using three genes (IFITM3, SERPING1, and FCGR3A).

Figure 4

The landscape of immune infiltration and its relationship to critical genes. (A) The stacking chart depicts the relative proportion of 22 types of immune cells in each sample. (B) The boxplot for expression levels of immune cells between AIGVHD and control group samples. (C) The heatmap depicts the relationship between three critical genes and immune cells. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. ns, no significant.

Figure 5

General Condition and Clinical Scoring. (A) Weight changes were monitored between the two groups. (B) The clinical scoring for aGVHD was assessed within the aGVHD group. (C) Clinical symptoms observed in the two groups revealed that aGVHD mice exhibited hunched posture, alopecia, and the presence of bloody stools. (D) Anatomical images of the colon, liver, and spleen were obtained from both groups. Data are presented as mean ± SEM (n = 6).

Figure 6

HE staining of GVHD pathological tissues in the skin, liver, and intestine and implantation identification. (Aa-Ca) HE staining of skin, liver, and colon from the normal group (×100). (Ab) In the GVHD group, a slight thickening of the epidermal layer was noted, accompanied by an increase in the number of spinous layer cells (black arrow, ×100), evidence of mild cellular necrosis (yellow arrow, ×100), and the presence of numerous ruptured adipocytes (red arrow, ×100). (Ac) Inflammatory cell infiltration was observed, primarily comprising neutrophils (black arrow, ×400). (Bb) In the liver tissue of the GVHD group, minor hepatocyte necrosis was documented (black arrow, ×100); (Bc) along with necrosis of a limited number of bile duct epithelial cells (black arrow, ×400). This region exhibited a small degree of inflammatory cell infiltration, predominantly neutrophils with lobulated nuclei (red arrow, ×400). (Cb) In the intestinal tissue, the structural integrity of the mucosal epithelial cells was compromised (yellow arrow, ×100), with atrophy or complete loss of intestinal glands (black arrow, ×100) and a notable reduction in the number of goblet cells (red arrow, ×100). (Cc) The necrotic region displayed substantial inflammatory cell infiltration (black arrow, ×400) and increased fibrinous exudation (red arrow, ×400). (D) BALB/c mice only expressed H-2Kd histocompatibility antigen; (E) Only H-2Kb histocompatibility antigen was expressed in C57BL/6J mice before transplantation; (F) After transplantation, the GVHD group exhibited expression of the donor histocompatibility antigen H-2Kd.

Figure 7

CD4 Expression Analysis.This figure presents a comparative analysis of CD4 expression between the normal group and the aGVHD mouse model in skin tissue (A-C), liver tissues (D-F), and colon tissues (F-J). ***P<0.001.

Figure 8

mRNA expression levels of FCGR3A, SERPING1, and IFITM3. (A) In murine models, the mRNA expression levels of FCGR3A, SERPING1, and IFITM3 were found to be significantly elevated in the aGVHD group compared to the control group. (B) Similarly, analysis of human colon tissues from the GSE215068 series revealed that the mRNA expression levels of FCGR3A, SERPING1, and IFITM3 were markedly higher in samples from patients with aGVHD than in those without GVHD. *P<0.05, **P<0.01, ***P<0.001.