High Expression of ELFN1 Associates With Poor Prognosis And Immune Cells Infiltration In Colon Adenocarcinoma: A Bioinformatic Analysis


 Background The human ELFN1 encodes a functional protein belonging to leucine-rich repeat (LRR) family neuronal adhesion proteins. However, there remains a lack of research on ELNF1 in tumors, especially in colon adenocarcinoma (COAD). The paper aimed to explore ELFN1 expression and its potential biological function in COAD through bioinformatics analysis. Results ELFN1 expression was significantly upregulated in COAD tissues from The Cancer Genome Atlas (TCGA) database (P＜0.01). The high ELFN1 mRNA expression level in tumor samples was also verified in GSE39582 and GSE146009 cohorts from Gene Expression Omnibus (GEO) website (P＜0.01). Besides, high ELFN1 expression was positively correlated with advanced tumor stage (I vs. IV, II vs. IV), perineural invasion，and vascular invasion (P＜0.05). Moreover, high ELFN1 expression was correlated with worse overall survival (OS) (P＜0.05), which was also verified in GSE17536 and GSE29621 datasets (P＜0.05). Multivariate Cox regression analysis suggested ELFN1 was an independent prognostic predictor in COAD (P＜0.01). Furthermore, the high infiltration levels of M0 Macrophages and Regulatory T cells（Tregs）were found in the high ELFN1 expression COAD tissues, while the infiltration levels of activated CD4 memory T cells, resting or activated Dendritic cells were low in that samples (P＜0.05). The correlation analysis revealed that ELFN1 expression had a positive correlation with the infiltrating levels of Tregs and M0 Macrophages, but inversely correlated with that of activated Dendritic cells and both resting and activated CD4 memory T cells in COAD (P＜0.05). Gene Set Enrichment Analysis (GSEA) exhibited that ELFN1 was enriched in many immunological and cancer-related signaling pathways, like Hedgehog, ECM receptor interaction, focal adhesion, Notch, and MAPK signaling pathways.Conclusions ELFN1 is overexpressed in COAD and which is correlated with worse clinical outcomes. Higher expression of ELFN1 could influence tumor immune cells infiltrating in COAD. ELFN1 could promote progression of COAD through several immunological and cancer-related signaling pathways. ELFN1 may be a new prognostic predictor and a latent therapeutic target for COAD.


Background
Colorectal cancer (CRC) is the third most prevalent malignancy worldwide and the second cause of cancer-related deaths (1). Patients with CRC usually develop to an advanced stage at the rst diagnosis (2), about 20-25% of them with distant metastasis (3). Multiple factors can contribute to colorectal carcinogenesis, but the successive accumulation of genetic and epigenetic alterations remains the fundamental pathogenesis (4). It is reported that adenocarcinoma makes up 98% of newly diagnosed colon cancer (CC) (5). Although signi cant advances have been made in multimodal anticancer strategies, the 5-year OS of metastatic CC is only 13.7% (https://seer.cancer.gov, February 1, 2021). This is mainly due to our limited knowledge of the etiology and pathogenesis of CC, as well as being the shortage of early diagnostic indicators and effective therapeutic targets.
Immunotherapy like PD-1 blockade has been shown a promising response to metastatic CRC with DNA de cient mismatch repair (6). Tumor immune microenvironment (TIME) is tightly associated with tumorigenesis, prognosis, and the e cacy of immunotherapy (7). Tumor-in ltrating immune cells (TIICs), like macrophages, T cells, neutrophils, play a crucial role in modulating anticancer response in the TIME (8). It is primarily because TIICs can secrete cytokines and chemokines, which will further activate or inactivate several signaling pathways related to immune response. Therefore, if the nexus between oncogene and TIICs could be entirely delineated, it would supply a novel perception into the pathogenesis of COAD.
Extracellular leucine-rich repeat and bronectin type III domain-containing 1 (ELFN1) encodes a functional protein belonging to leucine-rich repeat (LRR) family neuronal adhesion proteins (9), which has been validated to perform a crucial role in synapse formation through the coordination of both presynaptic and postsynaptic machinery (10,11). Besides, it has been identi ed in interneurons of the hippocampus, cerebral cortex (12), and cone photoreceptors (13). However, there remains a lack of research on ELNF1 in tumors, especially in COAD.
In this study, it is the rst time to investigate the expression level of ELFN1 in COAD and uncover its potential biological roles, prognostic value, as well as its relationship with the TIICs. We observed that ELFN1 was overexpressed in COAD, which was correlated with worse clinical outcomes. ELFN1 could also in uence TIICs in ltrating in the TIME, particularly Tregs and M0 Macrophages. Moreover, GSEA revealed ELFN1 was linked to many immunological and oncogenic signaling pathways, like the Hedgehog, ECM receptor interaction, focal adhesion, Notch and MAPK signaling pathways. Our work suggests ELFN1 could be a new prognostic predictor and a latent target for COAD therapy.

Expression levels of ELFN1 in COAD
The mRNA expression pro les of ELFN1 in both tumor tissues and normal colon tissues from the TCGA and GEO websites were explored. It was found the ELFN1 mRNA expression level was signi cantly upregulated in COAD tissues from the TCGA database (P 0.01) (Fig. 1a). A similar result could be seen in 41 matched tumor and normal colon tissues from a TCGA cohort (P 0.01) (Fig. 1b). The high ELFN1 mRNA expression level in tumor samples was also veri ed in GSE39582 and GSE146009 cohorts (P 0.01) (Fig. 1c, d). Simultaneously, ELFN1 was also highly expressed in several types of cancers, such as esophageal cancer, stomach cancer, and breast cancer by bioinformatic analysis for the resting 32 cancer types (Additional File. 1a).

Association between ELFN1 mRNA expression and clinicopathological characteristics
The ELFN1 mRNA expression levels of 446 COAD samples from the TCGA database in different clinicopathological parameters were presented ( NO), Vascular invasion (YES vs. NO) and Perineural invasion (YES vs. NO) (P 0.05) ( Fig. 2a-g). However, no association was seen between ELFN1 mRNA expression and the patient's Age, Gender (P 0.05) (Fig.  2h- The potential of ELFN1 to be a diagnostic and prognostic indicator in COAD Due to the high expression of ELFN1 in COAD, its diagnostic and prognostic value for patients with COAD was analyzed. Generally, the high ELFN1 expression of COAD patients in the TCGA-COAD cohort exhibited worse OS (Fig. 3a) and Progression-free survival (PFI) (Fig. 3b) than that in the low ELFN1 expression patients (P 0.05). Simultaneously, the low expression of ELFN1 in the GSE29621 cohort exhibited favorable OS (P 0.01) (Fig. 3c), while ELFN1 expression had no effect on disease-free survival (DFS) of patients with COAD (P 0.05) (Fig. 3d). Then, low ELFN1 expression in the GSE17536 cohort correlated with a favorable OS (Fig. 3e) and DFS (P 0.05) (Fig. 3f). It was also revealed that high ELFN1 expression in stomach cancer, ocular melanomas and cervical cancer patients were correlated with a poor OS (P 0.05) (Additional File. 1b-d).
The clinicopathological parameters, including ELFN1 expression level, age, T, N, M stage, and tumor stage, were thought to in uence patients' OS in COAD by univariate Cox regression analysis (P 0.05) (Fig.   4a). Also, multivariate Cox regression analysis revealed that ELFN1 expression, age and T stage were independent prognostic predictors for COAD patients (P 0.05) (Fig. 4b). Furthermore, ROC curve analysis showed that ELFN1 (AUC=0.900 had a more satisfactory diagnostic value than that of carcinoembryonic antigen (CEA) for COAD (AUC=0.547) (Fig. 4c). The ROC curves of ELFN1 for OS at 3, 5, and 10 years were also plotted ( Fig. 4d).

Relationship between TIICs and ELFN1 expression in COAD
The relative abundances of 22 TIICs in each COAD tumor sample were calculated using the CIBERSORT algorithm. The high in ltration levels of M0 Macrophages and Tregs were found in COAD tissues with the high expression of ELFN1, while the in ltration levels of activated CD4 memory T cells, resting or activated Dendritic cells were low in the high ELFN1 expression COAD tissues (P 0.05) (Fig.5a). The correlation between the relative abundance of TIICs and ELFN1 expression level was also investigated. The outcomes revealed that ELFN1 expression had a positive correlation with the in ltrating levels of Tregs and M0 Macrophages, but inversely correlated with that of activated Dendritic cells and both resting and activated CD4 memory T cells in COAD (P 0.05) ( Fig. 5b-f).
The relationship between in ltrating Tregs, M0 Macrophages, and clinicopathological parameters was analyzed. The higher relative abundance of Tregs in COAD had a positive association with the advanced Tumor stage (III vs. II) (P 0.05) (Fig. 6a), whereas its abundance changes had no in uence on the T stage, N stage, and M stage ( Fig. 6b-d) (P 0.05). The relative abundance of M0 Macrophages was positively correlated with the early N stage (P 0.05), surprisingly, it revealed the opposite trend in the advanced N stage ( Fig.6g) (P 0.05). Besides, the altered abundance of M0 Macrophages had no effect on Tumor stage, T stage and M stage ( Fig. 6 e, f, h) (P 0.05). The changes of Tregs in ltration levels in COAD had no in uence on OS ( Fig. 6i) (P 0.05). However, lower Tregs' relative abundance was associated with better PFI (Fig. 6j) (P 0.05). The changes of relative abundances of M0 Macrophages had no effect on OS and PFI ( Fig. 6 k, l) (P 0.05).
Besides, there was a weak to moderate association among the relative abundance of different TIICs (Fig.  7). For example, the in ltration level of M0 macrophages in COAD negatively correlated with that of M1 Macrophages (r = -0.45) and resting Dendritic cells (r = -0.35).
Association of ELFN1 methylation with clinicopathological features in COAD 11 CpG sites of ELFN1 were found by analyzing the methylation database of TCGA-COAD (Fig. 8a).
Further analyses showed that the ELFN1 hypermethylation was associated with earlier tumor stage, N stage, and M stage ( Fig.8b-d) (P 0.05). Next, the associations of ELFN1 methylation with OS were evaluated. However, the methylation of ELFN1 did not affect the patient's survival (Additional File. 2a-k) (P 0.05).

Analysis of ELFN1 mutation in COAD
The mutation frequency of ELFN1 was too low in TCGA-COAD and ICGC-COAD cohorts ( Fig. 9a, b).

Gene set functional enrichment analysis
Differentially activated signaling pathways related to the high ELFN1 expression in COAD datasets were identi ed by GSEA software (Table. 2). Many signaling pathways were enriched in the ELFN1 high expression phenotype, such as focal adhesion, pathways in cancer, the intestinal immune network for IGA production, Cell adhesion molecules (CAMs), MAPK, Hedgehog, VEGF, ECM receptor interactions, and Notch signaling pathway (Fig. 10).
Table2. Differentially activated signaling pathways related to the high ELFN1 expression in COAD by GSEA. Co-expression analysis of ELFN1 The Pearson correlation coe cients between expression pro les of ELFN1 and PCGs were calculated to determine the co-expression relationships of the ELFN1 and PCGs. Then the Gene Ontology (GO) enrichment analysis was conducted. These genes encoded proteins playing roles mainly in transcription coregulator activity, GTPase regulator activity, growth factor binding, collagen binding crosstalk, extracellular matrix structural constituent, modi cation-dependent protein binding and transcription corepressor activity (Fig. 11a). The Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis re ected the enrichment of ELFN1-related signatures associated with many immunological and cancer-related signaling pathways, like Human papillomavirus infection, PI3K-AKT, CRC, breast cancer, Hippo, ECM receptor interaction, and focal adhesion, Notch, MAPK, mTOR signaling pathways, and so on (Fig. 11b). The STRING and Cytoscape 3.8.2 software were applied to construct the Protein-protein Interaction (PPI) networks among ELFN1-related co-expressed genes. Proteins interaction data in COAD revealed 189 kinds of proteins could interact with ELFN1 protein, among which 21 kinds of proteins had a negative correlation with the expression of ELFN1, whereas the expression of ELFN1 positively correlated with the other 168 kinds of proteins (Fig. 12a). Based on |r|≥0.6, 20 co-expression genes were identi ed to be correlated with ELFN1 expression (Fig. 12b, c).

Discussion
CC is one of the most prevalent malignancies globally, with increasing incidence (14). Despite the rapid progress having been made in screening and therapeutic measures of CC over the years, the patients' clinical outcomes with the advanced CC are still poor (15). That mainly because we have a super cial knowledge of the pathogenesis of CC (16). Thus, it is necessary to uncover novel biomarkers and enunciate the latent molecular mechanism of CC to re ne the patients' prognosis. This study focuses on ELFN1, an adhesion protein of the leucine-rich repeat (LRR) family. ELFN1 is expressed among several neurons and confers target-speci c synaptic properties (17). Given the barely reported correlation of ELFN1 and clinicopathological features of tumors, we aim to dissect the biological functions of ELFN1 in COAD and explore its latent mechanisms of action.
Data analysis in TCGA and GEO databases revealed the expression of ELFN1 was signi cantly upregulated in COAD. The analysis of clinicopathological characteristics manifested that ELFN1 expression in COAD was a positive relationship with the degree of tumor aggression, lymphovascular invasion, distant metastasis, and perineural invasion, which indicated ELFN1 might act as an oncogene to facilitate invasion and migration of COAD. Besides, the ROC curves illustrated that ELFN1 had good diagnostic and prognostic value. Also, Kaplan-Meier analysis displayed that higher ELFN1 expression correlated with worse OS and PFI. Moreover, multivariate Cox regression analysis revealed high expression of ELFN1 might be an independent predictor correlated with worse clinical outcomes of patients in COAD. These results mentioned above suggested that ELFN1 could be a biomarker of diagnosis and prognosis for COAD.
To explore the potential molecular mechanism, we performed GSEA and co-expression analyses. KEGG analysis showed that ELFN1 was associated with several types of tumor signaling pathways, embracing PI3K-AKT, Notch, mTOR signaling pathways, and so on. Also, the results of GSEA exhibited that ELFN1 was enriched in many immunological and cancer-related signaling pathways, like Hedgehog, ECM receptor interaction, focal adhesion, Notch, and MAPK signaling pathways, which indicated that ELFN1 could promote the progression of COAD by diverse biological processes. Additionally, numerous genes encoding several extracellular matrix proteins, like SPON2(18), COL18A1 (19), and SNAI1 (20), were also identi ed to affect the tumor progression. The extracellular matrix proteins encoded by the genes mentioned above could be interacted with the protein encoded by ELFN1. These results implied that ELFN1 can promote tumor progression through various pathways and it might act as a new and potential therapeutic target for COAD.
A growing number of proofs have focused on the interrelationship between tumor and immune stromal cells (21,22). Immune escape (23) and functional inhibition of TIICs in the tumor microenvironment(TME) (24) are two signi cant barriers to cancer immunotherapy.
We utilized CIBERSORT to explore the association of 22 TIICs with ELFN1 in the TME of COAD from the TCGA database. Our results revealed showed a negative nexus between the mRNA expression of ELFN1 and in ltration levels of activated dendritic cells, activated and resting CD4 memory T cells in COAD. Dendritic cells can capture tumor antigens and present them to cytotoxic T lymphocytes (25). Tumor cells can prevent T cells' recruitment and activation via the activated oncogenic pathways in the TME(26). Besides, we observed a positive association between Tregs' in ltration level and ELFN1 expression. We also observed that the high abundance of Tregs is related to the advanced tumor stage. That might imply ELFN1 and Tregs are implicated in the progression of COAD. Additionally, we observed a positive association between M0 in ltration level and ELFN1 expression. we also found that the high relative abundance of M0 Macrophages was positively correlated with the early N stage(N1 vs. N0). However, it revealed the low relative abundance of M0 Macrophages in the advanced N stage N2 vs. N1 . These results imply that different in ltration level of M0 Macrophages in different stages may play different roles in tumor progression. Luo et al. found the increase of Tregs and M0 macrophages in COAD was correlated with worse clinical outcomes (27). In contrast, Forssell et al. found that high macrophages in CC was associated with better patients' survival(28). The reason for this discrepancy may be attributed to cell differentiation with time speci city. M0 macrophages would polarize to M1 and M2 macrophages under different stimulations (29). M1 macrophages act as a protector and appear in the early stage of tumorigenesis (30). However, M2 macrophages play a crucial role in the invasion and immune escape of tumors and weaken the cell attack capacity of natural killer cells and T cells (31). Besides, M2 macrophages reach the peak in the period of tumor progression (32). We all know that most CC belongs to advanced-stage tumors at the diagnosis, so that M0 macrophages might polarize to more M2 macrophages to promote tumor progression. Surprisingly, there was no correlation between M2 macrophages' in ltration level and ELFN1 expression. The number of tumor tissues in our study is not large enough, which may be one of the reasons why there is no relation between ELFN1 expression and the in ltration level of M2 macrophages in the TME. But a positive trend can be observed between ELFN1 expression and M2 macrophages in ltration. Furthermore, we observed the in ltration level of M0 macrophages in COAD was negatively associated with that of M1 Macrophages, activated dendritic cells, and activated CD4 memory T cells, which robustly a rmed the potential biological function of M0 macrophages in the progression of COAD. Among immune cell types, we saw the appreciable nexus between ELFN1 and M0 macrophages, Tregs, which indicated that ELFN1 could regulate immune cell in ltration in the TME and be a latent anticancer target.
This paper has its own limitations. Firstly, our results originated from the analysis of RNA sequencing data of 514 COAD cases from TCGA and which were veri ed in a few GEO databases; thus, that should be veri ed in more and larger cohorts in the future. Secondly, algorithm analysis depended upon RNA-seq, which might not be absolutely accurate to draw the conclusions; so, more corresponding clinical and basic experiments are indispensable to uncover the latent biological mechanisms of ELFN1 in COAD in future period.

Conclusions
ELFN1 is overexpressed in COAD, which is correlated with worse clinical outcomes. Higher expression of ELFN1 could in uence immune cells in ltrating in the TME. ELFN1 could promote progression of COAD through several immunological and cancer-related signaling pathways. ELFN1 may be a new prognostic predictor and a latent therapeutic target for COAD. The mRNA expression analysis of ELFN1

Materials And Methods
The "limma" (33) package was applied to analyze the mRNA expression pro le of ELFN1 in TCGA-COAD via the R 4.0.3 software (34). The variations of ELFN1 mRNA expression between tumor tissue and neighboring normal tissue were visualized via the "ggpubr" (35) package. After deleting 27 samples due to incomplete clinical data, 446 samples were obtained for further analyzing the association of ELFN1 expression with clinicopathological features in TCGA-COAD using the "ggpubr" package. For veri cation, the expression pro le of ELFN1 in GSE39582 (36) (tumor, n=443; normal, n=19) and GSE146009 (37) (tumor, n=33; normal, n=32) datasets came from the GEO website (38) (https://www.ncbi.nlm.gov/geo). Besides, the mRNA expression data of ELFN1, including tumor and normal tissues in other 32 cancer types from the TCGA, were also analyzed using the same methods described above. Matrixes were ltered, depending on analysis requirements via the Perl language (https://www.perl.org/) scripts.

Survival analysis
Survival information of TCGA samples was obtained from TCGA, including PFI and OS. After removing samples with OS less than 30 days, 421 samples were included in the nal analysis. Samples were separated into low or high groups depending on the median of ELFN1 mRNA expression level. The Kaplan-Meier (K-M) curves were plotted via the "survminer" (39) and "survival" (40) package. Univariate and multivariate analyses of prognostic factors were conducted with Cox regression by the "survival" package. The ROC curves were plotted via the "timeROC" (41), "survminer," "survival," "pROC" (42) packages. The survival outcomes of GSE17536 (43)(CC samples, n=177), GSE29621(44) (CC samples, n=65) came from the GEO database to verify the results. Additionally, the same method was used for survival analysis in the other 32 cancer types.

The nexus of ELFN1 expression and TIICs
The "CIBERSORT R script v1.03" (45) was applied to compute the distribution of 22 TIICs in COAD depending upon the transcriptome pro les. Then, the relative abundances of each type of immune cells were calculated separately. Samples were separated into low or high expression groups depending upon the median mRNA expression level of ELFN1 with P < 0.05. The Spearman method was performed to analyze the correlation between the relative abundance of TIICs and mRNA expression level of ELFN1.

ELFN1 methylation analysis
The "plyr(50)", "reshape2(51)", "ggpubr" and "ggplot2" packages were performed to analyze and visualize the methylation of ELFN1. The associations between clinicopathologic variables and the methylation of ELFN1 were analyzed and visualized via the packages as described above. The survival analysis between the methylation of ELFN1 and OS in TCGA-COAD was done via the "survminer" and "survival" packages.

ELFN1 mutation analysis
The "GenVisR" (52) package was applied to analyze the ELFN1 mutation data from the TCGA-COAD and  Figure 1 Analysis of ELFN1 expression levels in COAD. a ELFN1 mRNA level in COAD and adjacent normal colon tissues from the TCGA cohort. b ELFN1 mRNA level in matched COAD tissues and normal colon tissues from the TCGA cohort. c ELFN1 mRNA expression in COAD GSE39582 cohort from GEO database. d ELFN1 mRNA level in COAD GSE146009 cohort from GEO database.   Diagnostic and prognostic value of ELFN1 in COAD patients. a Forest plot in univariate Cox regression showed ELFN1 expression could in uence the OS for COAD patients. b Forest plot in multivariate Cox regression revealed ELFN1 expression was an independent prognostic indicator for COAD patients. c ROC curve showed ELFN1 had a more satisfactory diagnostic value than that of CEA for COAD patients. d ROC curve revealed the prognostic value of ELFN1 at 3, 5, and 10 years after diagnosis for COAD patients.     Analysis of ELFN1 mutation in COAD samples from TCGA and ICGC cohorts. a The very low mutation frequency of ELFN1 gene in 394 COAD tissues from TCGA database. b The very low mutation frequency of ELFN1 gene in 295 COAD tissues from ICGC database.

Figure 10
Differentially enriched pathways associated with phenotype arising from ELFN1 high expression in COAD by GSEA: VEGF, MAPK, pathways in cancer, Hedgehog, Notch, ECM receptor interactions, intestinal immune network for IGA production, CAMs, focal adhesion. GO and KEGG enrichment analyses of ELFN1-related PCGs. a GO analysis of positively related biological processes (top ten) with high ELFN1 expression in bubble map, b KEGG analysis of many signaling pathways enriched in ELFN1-related signatures.

Figure 12
Co-expression analysis of ELFN1. a The PPI network was constructed with 189 nodes and 432 edges using the STRING and Cytoscape software (note upregulated nodes in brown color and downregulated nodes in green color). b Chord diagram depicting the relationships between ELFN1 and co-expression genes, with |r|≥0.6. c Heat map depicting the relationships between ELFN1 and co-expression genes, with |r|≥0.6. r value represented Pearson correlation coe cient, -1 r 1, r 0 meant a positive correlation, while r Page 30/30 0 meant a negative correlation. The line or circle in red color represented a positive correlation between ELFN1 and co-expression genes, while the line or circle in green color represented a negative correlation between ELFN1 and co-expression genes.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.