Database and correlation analysis

We used The Cancer Genome Altas (TCGA) (https://portal.gdc.cancer.gov/) to do pan-cancer expression analysis. RNAseq data of invasive breast carcer (TCGA-BRCA) were downloaded and collated from the TCGA database and extracted in TPM format. Data is processed and converted using Perl scripts. “ggplot2[3.3.6]”, “stats [4.2.1]” and “car” package were used to analyze the paired and unpaired differential expression of the data. The statistical method of pan-cancer analysis and the unpaired differential expression was Wilcoxon rank sum test. The statistical method of the unpaired differential expression was Wilcoxon signed rank test. Using DESeq2 [1.36.0], edgeR [3.38.2] package for CADM3 single genetic differences, and set up the conditions of P < 0.05 and| log FC| > 2 to further screening.

The R package “ggplot2[3.3.6]” “survminer”, and “survival [3.3.1]” were applied to analyze the patient survival data in TCGA database. The log-rank test was used for all the survival analysis.

Clinical sample data were downloaded from the TCGA database to analysis the differential expression of CADM3. ER, PR, HER2 expression status, age, TNM stage, and PAM50 subtypes were selected as analysis factors. And the data were analyzed by R [4.2.1] and R package “ggplot2 [3.3.6]”. Wilcoxon rank sum test was used for statistical analysis of ER, PR,HER2 status and age group. Kruskal-Wallis Test and Dunn’s test were used for the PAM50 group.

R package “DESeq2 (version 1.26.0)” was used to filter differentially expressed genes (DEGs) [15, 16] (p.adj < 0.05,|log2FoldChange|>2) between high expression group and low expression group of CADM3 divided by the median value in TCGA database. The R package “ggplot2 (version 3.3.6)” was used to plot the volcano figure and the heatmap. The statistical method of the heatmap was Spearman.

The enrichment score was defined by the single sample GSEA to represent the absolute enrichment degree of the gene set in each sample within a given dataset of the R package “GSVA” [17]. We also calculated the standardized enrichment score for each immune category. And we obtained the gene set features of various immune cells from the previous study [18].

We used STRING (https://string-db.org) [19] to determine the interactive protein of CADM3. We constructed the PPI network using Cytoscape software and CytoHubba [20] and identified the first 15 central genes. In addition, the function and mechanism of selected hub genes were predicted by GeneMANIA (https://genemania.org/), which is a flexible user-friendly web site for generating hypotheses about gene function, analyzing gene lists and prioritizing genes for functional assays [21].

Patients and samples

We collected 40 pairs of fresh BC tissues and corresponding adjacent normal tissues surgically removed in 2022, which were transported and preserved in liquid nitrogen tanks as soon as possible after surgical removal for detection of mRNA and protein levels. Paraffin sections of BC tissues from 325 patients surgically removed in 2015 were collected and stored at 4℃ in the dark for immunohistochemical experiments. This study was approved by Ethics Committee of the First Hospital of China Medical University (Number: AF-SOP-07-1.1-01).

RNA transfection

CADM3 overexpression and the negative control lentivirus were bought from Shanghai Genechem Co., LTD Cell planking was performed before transfection, and cell density was determined according to cell growth rate and shape, so that cell coverage reached about 30% on the day of transfection. The number of virus particles was calculated according to the MOI value, and the virus stock solution was diluted. ​The dilute solution of the virus was added to a complete medium free of antibiotics. The culture medium was changed after 24 h in the incubator, and then the medium was changed according to the state of the cells.The sequences of CADM3 and its control for RNA transfection can be seen in supplementary Table 1.

Antibodies

The antibodies used in the experiment are as follows: CADM3 (Affinity, #DF3537); GAPDH (protein tech, #60004-1-Ig); CD3 (Abmart, #T55982); CD8 (Abmart, #T40010); p38 (Abmart, #T40075); Phospho-p38(Thr180/Tyr182) (Abmart, #T40076); ERK1/2 (Abmart, #T55487); Phospho-Erk1(T202/Y204) + Erk2(T185/Y185) (Abmart, #T40072); JNK1/2/3 (Abmart, #T40073); p-JNK1/2/3(Thr183 + Tyr185) (Abmart, #T40074).

Western blot analysis

Cells were lysed for 15 min at 4℃ using an ice cold RIPA buffer. Through a series of operations like sonication and centrifugation, we obtained the total cell extract. Put the cell extract into boiled water for 5 min to make it denatured. Put the protein into SDS–PAGE (10% gels), 30 micrograms per well, and then transferred onto a PVDF membrane. 1 h sealed in 5% milk powder solution, then added diluted primary antibody and incubated overnight at 4℃. The secondary antibody was incubated for 1 h at 25℃. ECL system was used for detecting the samples.

Real time PCR

RNA from tissues and cells was extracted by TRIzol reagent and then cDNA synthesis was performed. Each tubule was added with 10µL SYBR, 8.2µL ddH2O, 0.8µL primer and 1µL template for quantitative real-time PCR. Forty cycles were repeated at 50°C for 2min, 95°C for 10min, 95°C for 15s, and 60°C for 1min. Relative expression levels were calculated by 2-ΔΔCt and standardized by GAPDH. The sequences of the primer pairs are shown below: CADM3 forward, 5’-CTCGGTGACATTCCAGGTTACC-3’, reverse, 5’-GCCTAATCATCGCAGTTGGTG TG-3’; GAPDH forward, 5’-GGAGCGAGATCCCTCCAAAAT-3’, reverse, 5’-GCTGTTGTCATACTTCTCATGGG-3’.

Immunohistochemistry

The tissue sections were deparaffinized and rehydrated, then citrate buffer was used to perform antigen retrieval. Next, the endogenous peroxidase activity was blocked by 3% H2O2. Then, incubated the sections with the CADM3 antibody (Affinity, #DF3537, 1:100) overnight at 4 centigrade. After that, the secondary antibodies incubation, DAB regents (Maxim, DAB-0031/1031) staining, and hematoxylin counterstaining were performed. The CADM3 expression score was measured by multiplying the intensity score of staining (negative, 0; weak, 1; moderate, 2; strong, 3) and percentage score of immunoreactive tumor cells (<1%, 0; 1–25%, 1; 26–50%, 2; 51–75%, 3; >75%, 4).

The CD3 antibody (Abmart, #T55982, 1:100) and CD8 antibody (Abmart, #T40010, 1:100) were used to conduct immunohistochemical experiments related to immune infiltration, and the staining method was the same as above. The correlation between CD3, CD8 and CADM3 expression was detected by Wilcoxon rank sum test.

CCK-8 assay

2 × 103 cells were added to each hole of the 96-well plate and incubated at 37℃ for 24 h to make the cells stick to the wall. Then, 10µL CCK-8 reagent was added to each well for 2 h in the incubator. The absorbance of the cell at wavelength 450 nm was detected and the value was recorded.

Colony formation assay

1 × 103 cells were evenly spread on the six well plates and cultured at 37℃ until the cell colony was formed. The cells were fixed at about 30 min by 95% ethanol, followed by staining over 30 min with Crystal Violet Staining Solution. Clonal formation was counted after staining.

Cell scratch assay

A marker or a steel needle was used to mark the outer side of the six-hole plate with two parallel lines spaced about 1.5 cm apart at the center of each well. Inoculated with 4 × 105 cells at each well and cultured overnight. Then, a straight scratch was made in the middle of each hole in the six-hole plate perpendicular to the marked line. Each well was washed three times with sterile PBS to wash the cells scraped off in the previous step. The intersection of the scratch and the marking line was photographed and recorded as 0 h. Cells were further cultured under serum-free conditions. The above two steps were repeated after 12, 24 and 36 h, respectively. We measured each scratch edge line spacing with ImageJ after setting the scales and denoted as the scratch width. For each picture, the width was measured at three points and their average value was taken. The experimental group and control group were repeated three times by the above method, and the recorded data were quantitatively analyzed by multiple t-tests method. Cell mobility (%)=(0 h scratch width - scratch width after culture)/0 h scratch width 100%.

Transwell assay

Transwell cells with a diameter of 8 μm were placed into a 24-well plate. 1 × 105 cells were added into 150µL serum-free medium, and uniformly inoculated on the upper layer of the cells. Then 500µL complete medium was injected into the lower part of the chamber and cultured in the incubator for 6-12 h. The chambers were removed, the non-penetrating cells in the upper layer were erased. The cells were fixed by 4% paraformaldehyde for 30 min and stained by Crystal Violet for 30 min. The experimental group and control group were repeated three times by the above method. Using ImageJ to count cells, and the recorded data were quantitatively analyzed by t-test method.

Data presentation and statistical analysis

SPSS (version 26.0) and R (version 4.2.1) were used for statistical analysis. 95% CI and HR were calculated using Cox univariate and multivariate models. We analyzed the correlation between clinical features and CADM3 expression using logistic regression. Set P < 0.05,| logFC| > 2 as the threshold to screen differential genes of CADM3. The correlation between CADM3 and immune infiltrated cells was examined by Pearson.