Identification of FCN1 as a novel macrophage infiltration-associated biomarker for diagnosis of pediatric inflammatory bowel diseases

RNA-sequencing (RNA-seq) data of rectal biopsies from PIBD and non-IBD pediatric patients was downloaded from Gene Expression Omnibus (GEO) (https://​www.​ncbi.​nlm.​nih.​gov/​geo/​) under the accession number, GSE117993 [27]. Then, we followed the standard workflow of the ‘DESeq2’ package [30] in R software (version 4.0.4) to identify mucosal gene expression changes between 55 non-IBD controls and 75 PIBD patients. Differentially expressed genes were identified with adjusted P values less than 0.05 and Log2FoldChange absolute values greater than 2.

To identify the potential diagnostic biomarkers of PIBD among the differentially expressed genes, we implemented two machine learning algorithms, least absolute shrinkage and select options (LASSO) [31] logistic regression and multiple support vector machine-recursive feature elimination (mSVM-RFE) [32]. We first used the R package ‘glmnet’ (v4.1-1) [33] in for the LASSO logistic regression. Then, mSVM-RFE algorithm was performed by using the R package ‘e1071’ (v1.7-7) and the R implementation of this method available on GitHub (https://​github.​com/​johncolby/​SVM-RFE). Subsequently, the top-ranked biomarkers identified by the above two algorithms were entered into logistic regression models. After that, the area under the ROC curve (AUC) was employed to assess the performance of these models based on the external validation dataset (GSE126124) [28]. The GSE126124 dataset consisted of gene expression microarray data of blood and colon samples from 59 PIBD patients and 39 non-IBD children, so we divided this dataset into two parts (blood and tissue) for validation.

Differentially expressed genes identified above were subjected to Gene Ontology (GO) enrichment analysis using the R package ‘clusterProfiler’ (v3.16.1) [34]. The enriched GO terms were ranked by adjusted P values. Then, we calculated the frequency of FCN1-related GO terms in the top 100 enriched GO terms, which was visualized in a pie chart. Meanwhile, Medical Subject Headings (MeSH) enrichment analysis was performed using the package ‘meshes’ (v1.14.0) [35] based on data source from ‘gene2pubmed’ and ‘C’ (Disease) category.

Differentially expressed genes identified above were uploaded to the search tool for the retrieval of interacting genes/proteins (STRING) (v11.5) database (https://​string-db.​org/​) [36] for PPI network construction. Then, PPI network visualization was performed using Cytoscape (v3.7.2) [37], where node size depends on degree and edge size depends on combined score. To create a new subnetwork associated with FCN1 in Cytoscape, we first selected FCN1 and then performed the ‘From Selected Nodes, All Edges’ function in the ‘New Network’ function group.

To characterize the leukocyte composition of mucosal tissue, we downloaded the CIBERSORT [38] R script (v1.03) from ‘https://​rdrr.​io/​github/​singha53/​amritr/​src/​R/​supportFunc_​cibersort.​R’, and the leukocyte signature matrix (LM22) from ‘https://​github.​com/​zomithex/​CIBERSORT/​blob/​master/​CIBERSORT_​data/​LM22.​csv’. DESeq2-normalized read counts of GSE117993 [27] and normalized microarray data of GSE126124 [28] were taken as the input data for CIBERSORT analysis and Kruskal–Wallis test was used to evaluate the difference between two groups. Then, the R packages ‘ggplot2’ was used for visualization.

Raw expression matrices of mucosal single-cell RNA-seq data from PIBD and control groups were downloaded under the accession number, GSE121380 [29]. Then, the integrated raw data was fed into the established workflow of the R package ‘Seurat’ (v4.0.2) [39]. Following the quality control procedure, cells with 200–5000 detected genes, cells with less than 40000 UMIs and cells with less than 15% UMIs from mitochondrial genes were kept for further analysis. The filtered matrix was processed in Seurat for normalization by using the default method ‘LogNormalize’, followed by dimension reduction. After that, cell clusters were annotated by using both the R package ‘SingleR’ (v1.6.1) [40] with the reference dataset ‘BlueprintEncodeData’ and the known marker genes. The marker genes of the major cell types include PTPRC (immune cells), CD79A (B cells), CD3E (T cells), CD68 (macrophages), EPCAM (epithelial cells), COL1A1 (fibroblasts) [29]. Next, macrophages with above-average FCN1 expression were defined as FCN1^high macrophages, and the rest as FCN1^low macrophages. To investigate the biological pathway activity in FCN1^low and FCN1^high macrophages, hallmark gene sets (v7.5.1) were downloaded from the Molecular Signatures Database (https://​www.​gsea-msigdb.​org/​gsea/​index.​jsp), and then gene set enrichment analysis (GSEA) [41] was performed by using the R package ‘GSEABase’. Intercellular communication network analysis was performed by using the standard workflow of the R package ‘CellChat’ (v1.4.0) [42].

After diagnostic testing, the remaining tissue and blood samples from PIBD patients and non-IBD subjects were collected as our validation cohort. PBMCs were isolated from the whole blood by using Lymphoprep^™ (1858, Serumwerk Bernburg) according to the manufacturer’s instructions. The protocols were approved by the Ethics Committee of Ruijin Hospital, Shanghai Jiao Tong University School of Medicine.

Three-week-old male C57BL/6 mice were purchased from Phenotek Biotechnology (Shanghai) and housed under specific pathogen-free conditions. Ten mice were randomly assigned to two groups and allowed to acclimate for one week. Then, five mice were administrated by 3% (w/v) dextran sodium sulfate (DSS, MB5535, Meilunbio) solution in their water bottle for 7 consecutive days to induce colitis while 5 control mice received water without DSS. The body weight of each mouse was monitored daily. On the 8th day, the mice were sacrificed for further analysis [43, 44]. All animal experiments were approved by the Ethics Committee of Ruijin Hospital, Shanghai Jiao Tong University School of Medicine.

Samples were collected and immediately fixed in 10% neutral buffered formalin. Paraffin-embedded sections of biopsies were subjected to hematoxylin and eosin (H&E), immunohistochemistry and immunofluorescence staining. Primary antibodies for section staining included: FCN1 (10930-R017, Sino Biological), CD68 (97778 s, CST).

Total RNA was isolated using RNAex Pro reagent (AG21102, Accurate Biotechnology) and converted to complement DNA using Evo M-MLV. RT Kit (AG11711, Accurate Biotechnology) according to the manufacturer’s instructions. Quantitative PCR was performed using the SYBR® Green Premix Pro Taq HS qPCR Kit (AG11718, Accurate Biotechnology). Gene expression levels were normalized to B2M or Gapdh in human or mouse samples, and evaluated using the 2^−∆∆Ct method. The primers used are listed in Additional file 4: Table S3.

Human monocyte-like cell line (THP-1) was cultured with RPMI-1640 containing 10% (v/v) fetal bovine serum (FBS) and 1% (v/v) penicillin/streptomycin at 37 °C in a humidified 5% CO₂ air atmosphere. THP-1 cells were differentiated into M0 macrophages after 48 h exposure to 100 ng/ml PMA (phorbol 12-myristate 13-acetate, S1819, Beyotime), and then polarized towards M1 phenotype by incubation with 100 ng/ml LPS (L2880, Sigma) and 20 ng/ml IFN-γ (11725-HNAS, Sino Biological) or towards M2 phenotype by incubation with 20 ng/ml IL-4 (11846-HNAE, Sino Biological) and 20 ng/ml IL-13 (10369-HNAC, Sino Biological) for 48 h [45]. To determine the effect of FCN1 overexpression in macrophages, THP-1 cells were transfected with pCMV3-FCN1-C-His (HG10930-CH, Sino Biological) or pCMV3-C-His control vectors (CV015, Sino Biological) using Lipofectamine 2000 (11668-019, Invitrogen) in 1.5 ml Eppendorf tubes for 24 h as previously reported [46]. After transfection, THP-1 cells were then transferred into a 12-well plate and cultured with fresh medium containing 100 ng/ml PMA for 24 h. These cells were then cultured with or without 100 ng/ml LPS for 2 h.

Protein was extracted from murine tissues and THP-1-derived macrophages of different phenotypes for western blot. The blots were probed with antibodies against IL-6 (DF6087, Affinity Bioscience), FCN1 (10930-R017, Sino Biological), IL-1β (12242S, CST), GAPDH (GB12002, Servicebio), β-actin (GB12001, Servicebio), caspase-1 (3866S, CST), cleaved caspase-1 (4199S, CST), and NLRP3 (15101S, CST).

Statistical analysis was performed using R software (version 4.0.4). Detailed descriptions of statistical tests are provided in the corresponding bioinformatics methods and Figure Legends.

The mucosal expression profiles of 75 PIBD and 55 non-IBD pediatric patients from the GSE117993 dataset [27] were analyzed to identify disease-specific markers of PIBD. A total of 373 differentially expressed genes (DEGs) between PIBD and non-IBD were subjected to marker selection by using least absolute shrinkage and select options (LASSO) [31] and multiple support vector machine-recursive feature elimination (mSVM-RFE)[32] algorithms (Fig. 1A, B, Additional file 1: Table. S1). The LASSO algorithm identified 16 genes from DEGs as diagnostic markers for PIBD by applying the lambda of minimum mean error; the mSVM-RFE algorithm identified 6 markers as top features of PIBD with a low error rate (Fig. 1C, D). Then, we obtained two overlapping markers, FCN1 and LINC01558, from these two algorithms (Fig. 1E). The diagnostic efficacy of FCN1 and LINC01558 was further validated in an independent PIBD dataset (GSE126124) [28], which consisted of the expression profiles of 78 colon biopsies and 98 peripheral whole blood samples. As shown in Fig. 1F, G, FCN1 showed great diagnostic efficacy in both colon and blood validation sets with area under the curves (AUC) values of 0.937 and 0.743, respectively. In contrast, LINC01558 did not achieve good diagnostic performance in the colon and blood validation datasets (AUCs, 0.576 and 0.363). Taken together, the above results indicate that FCN1 may be a good mucosal and circulating biomarker for the diagnosis of PIBD.

The expression level and the underlying role of FCN1 in PIBD remain largely unknown. Using the public RNA-seq dataset (GSE117993), we found that the expression of FCN1 in rectal biopsies from PIBD patients was significantly higher than in non-IBD controls (Fig. 2A). Consistently, in an independent PIBD microarray dataset (GSE126124), we further confirmed that FCN1 expression was increased in both colon biopsies and peripheral whole blood samples from PIBD patients compared with non-IBD subjects (Fig. 2B, C). In addition, DEGs between PIBD and non-IBD identified from the above dataset were subjected to Gene Ontology (GO) analysis and Medical Subject Headings (MeSH) enrichment analysis. The results of GO analysis clearly showed that FCN1 had a high frequency (31%) in the top 100 enriched GO terms. Among the top 10 significantly enriched pathways, FCN1 was involved in humoral immune response, complement activation and phagocytosis (Fig. 2D). The MeSH term enrichment analysis in Fig. 2E showed that FCN1 was closely associated with infection and inflammation. These results suggest that FCN1 is upregulated and associated with immune response in PIBD.

To identify FCN1-related genes from DEGs between PIBD and non-IBD, we performed protein–protein interaction analysis. Among the 14 identified FCN1-related genes (Fig. 2F), S100A8 and S100A9 are the components of calprotectin [47]. Similar to FCN1, the expression of S100A8, S100A9 and other FCN1-related genes were all upregulated in PIBD mucosa compared to non-IBD mucosa (Fig. 2G). Since fecal calprotectin (S100A8/S100A9) is a useful screening biomarker for PIBD diagnosis in clinical practice [12, 13, 16], we compared the diagnostic efficacy of FCN1 with that of S100A8 and S100A9 to determine the potential clinical value of FCN1. In the colon validation set, S100A8 and S100A9 showed efficient diagnostic performance in PIBD, yielding AUCs of 0.920 and 0.877 respectively. Notably, FCN1 demonstrated superior performance for PIBD diagnosis as compared to S100A8 and S100A9 (Fig. 2H). These results further support the potential clinical value of FCN1 in PIBD diagnosis.

Immune cells play a critical role in IBD pathogenesis. To identify changes in the immune cell composition of PIBD mucosa, we applied digital cytometry analysis (CIBERSORT) [38] to mucosal expression profiles from the GSE117993 dataset. We found that the immune cell landscape was altered in PIBD rectal mucosa compared to non-IBD controls, including macrophages, mast cells, dendritic cells, neutrophils, T cells and B cells (Fig. 3A–F, Additional file 2: Fig. S1A). Among them, the inferred mucosal abundance of M0 and M1 macrophages was significantly increased, while the abundance of M2 macrophages was decreased in PIBD compared to non-IBD (Fig. 3A). Both resting mast cells and resting dendritic cells were decreased in PIBD, whereas their activated forms were increased, although not reaching statistical significance (Fig. 3B, C). The abundance of infiltrating neutrophils, activated CD4 memory T cells and plasma cells in PIBD were significantly higher than in controls, whereas abundance of CD8 T cells, naïve B cells and memory B cells were lower in PIBD (Fig. 3D–F). Similar changes in immune cell composition were also observed in another PIBD cohort (GSE126124, Additional file 2: Fig. S1B, S2A–F).

To investigate the association between FCN1 expression and immune cell infiltration, we performed Spearman’s correlation analysis. The strongest correlations were found between FCN1 expression and macrophage subsets, including M0, M1 and M2 states (Fig. 3G). As shown in Fig. 3H, FCN1 expression was positively correlated with the CIBERSORT inferred abundance of M0 macrophages (Spearman’s r = 0.75, P = 2.6 × 10^–24) and M1 macrophages (Spearman’s r = 0.71, P = 3.3 × 10^–21), but negatively correlated with M2-macrophage abundance (Spearman’s r = −0.62, P = 2.5 × 10^–15) in the GSE117993 PIBD cohort. These correlations between FCN1 and inferred macrophage subsets were further corroborated in an additional PIBD cohort (GSE126124, Additional file 2: Fig. S2G).

To further investigate the association between FCN1 expression and macrophage polarization states, we analyzed a single-cell transcriptomic dataset of colon biopsies from non-IBD and PIBD (GSE121380) [29] (Fig. 4A, see Additional file 2: Fig. S3 for cell cluster annotation). FCN1 was almost expressed in CD68⁺ macrophages, and the expression levels of FCN1 and IL1B in macrophages of colon biopsies from PIBD were significantly increased compared to those from non-IBD (Fig. 4B, Additional file 2: Fig. S4). In comparison with FCN1^low macrophages, several of the top 10 upregulated gene sets in FCN1^high macrophages were related to hyper-inflammation, such as TNFα signaling via NF-κB, inflammatory response, interferon gamma response, complement and IL-6 signaling (Fig. 4C–E, Additional file 2: Fig. S5, Additional file 3: Table S2). The classical M1 macrophages are generally characterized by pro-inflammatory responses and high expression of TNF, IL-1, IL6, and interferon-induced genes [48‐50], suggesting that FCN1⁺ macrophages enriched in PIBD mucosa are similar to pro-inflammatory M1 macrophages.

We then applied CellChat [42] to decipher the intercellular signaling networks in PIBD colon mucosa. Interestingly, we found many differential proinflammatory signaling networks between FCN1^low/FCN1^high macrophages and other major cell types, such as pathways related to macrophage migration inhibitory factor (MIF), intercellular adhesion molecules (ICAM), CC chemokine ligands (CCL), and B cell activating factor (BAFF) (Fig. 4F). Compared to FCN1^low macrophages, FCN1^high macrophages received more signals through the MIF signaling pathway from B cells, fibroblasts, and T cells (Fig. 4G). Notably, FCN1^high macrophages also exhibited increased autocrine ICAM and CCL signaling relative to FCN1^low macrophages (Fig. 4H, I). In addition, FCN1^high macrophages were the major ligand sources of BAFF signaling targeting B cells (Fig. 4J). These observations support that the FCN1^high macrophages enriched in PIBD colonic mucosa contribute to mucosal inflammation.

Consistent with the findings of the PIBD single-cell transcriptome analysis, the immunostaining results showed that FCN1 expression was remarkably increased in colon biopsies from PIBD patients compared to non-IBD controls (Fig. 5A), and FCN1 was specifically colocalized with CD68⁺ macrophages in PIBD colon biopsies (Fig. 5B). Considering that macrophages in the intestinal mucosa are continuously renewed by circulating monocytes [51], we hypothesized that the isolation of PBMCs from whole blood samples might improve the blood-based diagnostic performance of FCN1, since PBMCs contain a higher percentage of monocytes than whole blood. In our cohort, the mRNA expression levels of FCN1, S100A8, and S100A9 was significantly higher in PBMCs from PIBD patients than from non-IBD controls, but LINC01558 expression was not different between two groups (Fig. 5C–F). As expected, S100A8 and S100A9 showed efficient diagnostic performance in PIBD, whereas LINC01558 did not perform well, yielding AUCs of 0.843, 0.929, and 0.314, respectively. Notably, FCN1 still showed the best sensitivity and specificity to discriminate PIBD patients from non-IBD controls in our validation cohort with the AUC of 0.986 (Fig. 5G). Collectively, these results further indicate that FCN1 is a promising biomarker for PIBD diagnosis (Fig. 5H).

Our above results suggest that FCN1 expression was increased in both colorectal mucosa, blood, and PBMCs of PIBD patients. Therefore, we further determined involvement of FCN1 in PIBD using DSS-induced murine colitis model (Fig. 6A). Compared with the control group, four-week-old mice exposed to DSS exhibited loss of body weight, a significant decrease in colon length and enlargement of spleen (Fig. 6B–D). Colitis in DSS-treated mice was ascertained by histopathological analysis of epithelial degeneration and immune cell infiltration (Fig. 6E). Mouse FCNB resembles human FCN1 [18]. Consistent with the findings in human PIBD, the mRNA expression of Fcnb was obviously increased in peripheral whole blood of mice with DSS-induced colitis (Fig. 6F), and the protein levels of FCNB were also increased in the colon and spleen tissues of mice with DSS-induced colitis (Fig. 6G, H).

Given that FCN1⁺ macrophage infiltration was increased in human PIBD mucosa, and FCN1^high macrophages had high expression of IL-1β, we then investigated whether there was an association between FCNB (the murine homologue of human FCN1) and IL-1β in the mouse model of colitis. As expected, the expression of IL-1β was markedly upregulated in colon and spleen tissues from mice with colitis compared to the control group, and tissues with high expression of FCNB also showed higher IL-1β expression (Fig. 6G, H), suggesting the association between FCNB and IL-1β. In addition, increased FCNB⁺ cell infiltration was detected in the colon tissues of mice with DSS-induced colitis when compared to the control group, which was consistent with the findings in human PIBD (Fig. 6I). More importantly, FCNB was specifically colocalized with CD68⁺ macrophages recruited into the colon of mice with colitis (Fig. 6J). Taken together, these results indicate that increased FCNB expression is associated with proinflammatory macrophage infiltration in the colon mucosa of mice with colitis.

To further investigate whether changes in macrophage phenotype are associated with FCN1 expression, THP-1-derived M0, M1, and M2 macrophages were generated (Fig. 7A). Compared with M0 macrophages, M1 polarized macrophages exhibited upregulated expression of FCN1 and pro-inflammatory cytokines, including IL-6, pro-IL-1β, and cleaved-IL-1β, whereas M2 polarized macrophages exhibited downregulated expression of FCN1 and the pro-inflammatory cytokines (Fig. 7B).

The proinflammatory cytokine IL-1β was mainly derived from macrophages and its biological activation typically requires NLRP3 inflammasome-dependent caspase-1 cleavage [52]. Given the observed association between the expression of FCN1 and IL-1β, we questioned whether FCN1 could regulate IL-1β activation. To address this, we overexpressed FCN1 in THP-1-derived macrophages, and found that overexpression of FCN1 slightly increased the protein levels of NLRP3, cleaved caspase-1 and mature IL-1β (Fig. 7C, D). Considering the defective intestinal mucosal barrier and the increased systemic exposure to gut-derived endotoxin (LPS) in patients with PIBD [53, 54], we then exposed THP-1-derived macrophages to LPS. Interestingly, in the presence of LPS, overexpression of FCN1 increased the NLRP3 expression, and greatly upregulated the protein levels of cleaved caspase-1 and mature IL-1β (Fig. 7C, D). Consistently, in the presence of LPS, human recombinant FCN1 protein (rhFCN1) could also increase the protein levels of NLRP3, cleaved caspase-1, and IL-1β (Additional file 2: Fig. S6).

Overall, infiltration of FCN1⁺ macrophages are increased in the colon mucosa of PIBD, and macrophage FCN1 promotes the production of mature IL-1β via NLRP3-dependent cleavage of caspase-1, which contributes to the intestinal inflammation (Fig. 7E).