HCT116 cells were obtained from the Cell Resource Center, Peking Union Medical College (which is the headquarters of the National Infrastructure of Cell Line Resource, NSTI), which were cultured in Dulbecco’s Modified Eagle Medium (DMEM)/high glucose (11965092, Gibco, USA) supplemented with 10% fetal bovine serum (FBS; 16140071, Gibco, USA) and penicillin (100 IU)/streptomycin (100 µg/ml) at 37°C in a 5% CO₂ incubator. Cells are available within 15 generations. The extraction and culture methods of MSC refer to previous studies.

We collected adult fat samples from plastic surgery hospitals after obtaining informed consent from the donors. Using D-Hanks’ buffer, the adipose tissue was washed twice with two antibiotics (penicillin and streptomycin) and centrifuged at 800g for 3 min. The upper layer was transferred to a new 50-ml centrifuge tube. Then, 0.2% collagenase P (Life Technologies Corporation) was added to the tubes containing the pelleted tissue for enzymatic digestion followed by incubation at 37°C for 30 min. Subsequently, the digested adipose tissue was filtered with a 100-µm cell strainer. The sample was centrifuged at 1,500g for 10 min. Next, 2 × 10⁶ cells were seeded in T75 flasks and incubated at 37°C and 5% CO₂ in a cell incubator.

DMEM (Life Technologies Corporation) was replaced with human adipose-derived MSC (hAD-MSC) culture medium without FBS, 36−48 h before exosome extraction. Supernatants were harvested after culture and centrifuged at 3,000g for 10 min to remove dead cells and cell debris. The sample was transferred to the ultrafiltration apparatus (Life Technologies Corporation) with a 100,000-kDa–molecular weight ultrafiltration membrane. Exosomes were resuspended in D-Hanks’ buffer, and the suspension was filtered with a 0.2-µm microporous membrane filter, dispensed in 1.5-ml sterile microcentrifuge tubes, and preserved at −80°C.

The purified exosomes were diluted and dropped onto a copper mesh for 5 min for precipitation. Then, filter paper was used to absorb excess liquid, and the sample was air dried. Subsequently, 3% phosphotungstic acid in water was used to counterstain the sample for 2 min. Finally, exosomes were observed using a transmission electron microscope (Olympus, Japan) and photographed.

1,1-Dioctadecyl-3,3,3,3-tetramethylindotricarbocyaine iodide (DiR; 1 µM) (Life Technologies Corporation) is a lipophilic carbon cyanine dye that can bind lipoproteins in a manner similar to phospholipids and is embedded in the membrane of the biomass and oriented within the membrane. Diffusion movement can be used to observe cell-bound or endocytic lipoproteins under a fluorescence microscope, and this allows for semi-quantitative analysis (27). Purified exosomes were exposed to 1 µM DiR for 10 min. After incubating with MSCs for 10 h, the cells were washed with PBS three times, and the nuclei were stained with Hoechst 33342 (10 µg/ml) for 15 min at room temperature and washed with Phosphate Buffered Saline (PBS) three times. The cells were observed under a fluorescence microscope (OLYMPUS) and photographed.

A Transwell^® chamber (0.4 µm) (Corning) was used to co-culture the HCT116 cells with the hAD-MSCs at 1:1 ratio. The cells were passaged when cell density is 90%. Cell–cell co-culturing samples were collected at days 0, 3, 5, 7, and 9.

Small interfering RNA (siRNAs) were used to knockdown ICAM-1 mRNA and synthesized by GenePharma company (China). The sequence of knocking down ICAM-1 is GGCTGGAGCTGTTTGAGAACA. Specific operation of virus infection was described as previous report (28).

Proteins were extracted from cells using IP lysis buffer (87787, Thermo Fisher Scientific) with a cocktail (4693116001, Roche, Basel, Switzerland) and PhosSTOP (4906845001, Roche). The proteins were denatured in SDS (Sigma-Aldrich) with loading buffer and boiled for 10 min at 100°C. Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) was used to separate the proteins followed by the transfer of protein bands onto polyvinylidene fluoride (PVDF) membranes (Merck Millipore, Billerica, MA, USA). The membranes were then blocked with 5% milk in Tris-buffered saline–Tween 20 followed by overnight incubation at 4°C with primary antibodies. They were then washed and incubated with appropriate secondary antibodies for 1 h at room temperature, and bands were visualized using the enhanced chemiluminescence detection kit Life Technologies Corporation. The ICAM-1 (5915, 1:1,000), IL-6 (12912, 1:1,000), AKT (9272,1:1,000), Extracellularr regulated protein kinases (ERK) (4695, 1:1,000), p-ERK1/2 (4370, 1:1,000), p-JNK (9251, 1:1,000), anti-rabbit Horseradish Peroxidase labeled Anti-mouse IgG (IgG-HRP) (14708,1:2,000), Phospho-Stat3 (Tyr705, 9145, 1:1,000), Stat3 (D3Z2G, 12640, 1:1,000), Jak2 (D2E12, 3230, 1:1,000), Phospho-Jak2 (Tyr1007, 3771, 1:1,000), and anti-mouse IgG-HRP (14709, 1:2,000) antibodies were obtained from Cell Signaling Technology (Danvers, MA, USA); IL-8 (500-M08, 1:1,000) and p-Akt (ser473, 66444-1-IG, 1:2,000) antibodies were purchased from ProteinTech (Chicago, IL, USA).

RNA was extracted from cell samples using TRIzol (Thermo Fisher Scientific). RNA was thawed in 30 μl of RNA free water (Applygen) and reverse-transcribed (60 µl) according to the protocol recommended for the TaKaRa M-MLV reverse transcriptase (Takara). Amplification of the gene fragment was performed. To amplify the ICMA1, IL-6, IL-8, and Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) genes, one-step real-time quantitative polymerase chain reaction (RT-PCR) was performed as follows: 95°C for 5 min, 95°C for 10 s, 60°C for 40 s, 40 cycles. Reactions were performed in triplicate, and independent experiments were repeated three times. The RT-PCR data were analyzed using StepOne Software 2.1, and primers are presented in Table 1 .

The levels of soluble IL-6/8/TNFα in the supernatant of primary MTCAFs and the supernatant were measured using an enzyme-linked immunosorbent assay (ELISA) kit (Jiangsu Meimian industrial Co., Ltd., TNFα: MM-0132M1, IL-6: MM-0163M1, and IL-8:MM-0123M1), according to the manufacturer’s instructions. The absorbance (450 nm) of each sample was detected on a standard automatic microplate reader (BioTek, USA).

Colorectal cancer cell migration and invasion assay was conducted using 24-well Matrigel-coated Transwell inserts (BD Biosciences, San Diego, CA, USA). Approximately 2 × 10⁵ cells were seeded in serum-free medium in the upper chamber. Next, DMEM with FBS was added to the lower chamber, and after incubation at 37°C with 5% CO₂ for 24–48 h, the non-filtered cells were removed using a cotton swab and the migratory cells were stained with 0.1% crystal violet solution. The invasive cells attached to the bottom surface of the filter were quantified under a light microscope (200×). The data are presented as the average number of cells from randomly chosen fields. Each treatment condition was assayed using triplicate filters, and all filters were counted in five areas.

Using marker pen is to marker the 6-well plate with the ruler, which draw horizontal lines evenly (0.5-1cm). Each hole have to pass through at least 3 lines. Cell density is about 5*105 cells/pole. Next day, holding the head of the spear against the ruler and trying to keep it to the horizontal line in order to scratch. Wash the cells three times with PBS, remove the suspending cells, and add serum-free medium. Putting it into an incubator at 37°C with 5%CO2. Sampling at different hours and taking photos.

This is a retrospective cohort study. Colorectal cancers were obtained with informed consent from patients in Peking Union Medical College Hospital (Beijing, China) during January 2014 to December 2016. All specimens were collected using the protocols approved by the Ethics Committee of Peking Union Medical College Hospital. All patients were R0 resected and pathologically diagnosed with CRC.

The resected tissue samples were fixed with formaldehyde, embedded in paraffin, and prepared into 4-m-thick sections. The slides were then dewaxed and hydrated. Next, we decreased the peroxidase activity by treating with 3% H₂O₂. The sections were blocked by using 10% normal goat serum and incubated with appropriate primary antibody overnight at 4°C. Then, PBS diluted secondary antibody at 1:100 was added followed by incubation at room temperature for 2 h. All immunostained sections were then lightly restained with hematoxylin. The results of immunohistochemistry (IHC) were evaluated by two pathologists independently. If the results were inconsistent, the final result would be judged by the third pathologist. The membrane staining of cells >5% was defined as ICAM-1 positive.

The total RNA of the sample was quantified by NanoDrop ND-2000 (Thermo Scientific), and then, the RNA integrity was checked by Agilent Bioanalyzer 2100 (Agilent Technologies). After passing the RNA quality inspection, the labeling of the sample, the hybridization of the chip, and the elution refer to the standard process of the chip. First, total RNA is reverse-transcribed into double-stranded cDNA and then cRNA labeled with Cyanine-3-CTP (Cy3) is synthesized. The labeled cRNA is hybridized with the chip, and the original image is obtained by scanning with Agilent Scanner G2505C (Agilent Technologies) after elution.

All mice were maintained and manipulated according to the guidelines established by the Medical Research Animal Ethics Committee of Peking Union Medical University. The samples were randomly assigned. A mixture of 5 × 10⁶ HCT116 cells were re-suspended with 1 × 10⁶ cells or PBS (5:1) in 100 μl of PBS and subcutaneously injected into 6-week-old female athymic nude mice (BALB/C). Tumor formation was examined after 7 days. We detected the tumor size every three days, recorded the data, and finally calculated the volume (1/2 *the long side*the short side squared). When tumor volume reached 1–1.5 cm, the animals were sacrificed. Tissues were collected and sectioned followed by some sections being fixed with 10% buffered formalin for IHC analysis, whereas the others were preserved at −80°C for Western blotting.

Participants have provided written informed consent to take part in the study.

All data are expressed as means ± SD from at least three independent experiments. The statistics were analyzed by SPSS 25.0 statistical software (IBM, Armonk, USA). The relationship between the expression of ICAM-1 and disease-free survival (DFS) was evaluated by the Kaplan–Meier method. DFS was defined as the time from complete resection of tumor to disease recurrence. Statistical analysis was performed using two-tailed t-tests and one-way ANOVA. P < 0.05 was considered statistically significant. Each experiment was repeated at least three times to obtain a P-value and to control for systematic errors.

Previous studies have shown that MSCs can differentiate into CAFs (4, 29). First, we applied H-Exos to induce MSCs differentiate into CAFs, which is named MTCAFs, and we found that MTCAFs have the higher CAF-specific gene expression (α-SMA and FAPA) ( Figure 1A ). The characteristics of H-Exos are presented in Supplementary Figure 1 . Next, we evaluated the transcriptomic alterations and identified activated proteins in MTCAFs compared with MSCs; MTCAFs was kept into a transcriptionally active state, which was demonstrated by an increased number of upregulated genes ( Figure 1B ). Meanwhile, clustering identified upregulation of gene markers related to cell secreted inflammatory factors and immune regulation in MTCAFs compared with MSCs ( Figures 1C, D ). To be similar to the physiologic al conditions, we applied the co-culturing system, and we found that the co-culturing effect with HCT116 cells and MSCs is the same as that in exosomes secreted from HCT116 cells with MSCs ( Figure 1E ). Western blotting analysis showed an increase in the inflammatory and angiogenesis proteins from transcriptome analysis results ( Figures 1F, G ). In addition, the Agilent expression profiling gene chip results showed that ICAM-1 expression increased gradually during MSCs differentiation, and Western blotting analysis showed the same effect ( Figures 1H, I ). In conclusion, HCT116 cells can promote the differentiation of MSCs into MTCAFs and screen key gene ICAM-1 during the differentiation process.

To explore the correlations between ICAM-1 expression and progression and prognosis of patients with colon cancer, we collected patients samples with colon cancer from Oncomine Database, which includes paracarcinoma tissue and colorectal cancer tissue. The Oncomine analysis showed that α-SMA, ICAM-1, and LFA-1 exhibited a higher expression in colorectal cancer compared with colon tissue ( Figures 2A–C ), and ICAM-1 was positively correlated with α-SMA and LFA-1 ( Figures 2D, E ). To further clarify the function of these genes, we used clinical specimens for further analysis. The expression of ICAM-1 and α-SMA in tumor tissue of patients with stage I, II, and III CRC were obtained by using immunofluorescent staining. The result showed that ICAM-1 and α-SMA were co-expressed in clinical samples ( Figure 2F ). Next, we enrolled 72 patients (n = 72), 38 samples showed ICAM-1 positive, and 34 samples showed ICAM-1 negative ( Figure 2G ). On basis of this, we analyze the relationship between ICAM-1 expression and patient survival, and the result showed that the DFS of patients with positive ICAM-1 expression was significantly shorter than that of ICAM-1–negative patients [(28.06 ± 1.47) months vs. (38.87 ± 3.35) months, P = 0.013] ( Figure 2H ). These results suggest that ICAM-1 is inversely associated with survival in patients with colorectal cancer.

To further validate the function of ICAM-1 in the process of MSC differentiating into MTCAFs, we detected the invasion and migration abilities of MTCAFs, and the results showed that MTCAFs with ICAM-1 knockdown presented with significantly decreased abilities of migration ( Figures 3A, B ) and invasion ( Figures 3C, D ) compared with MTCAFs, which means that the ICAM-1 may play an important role in the MTCAFs. To further understand the situation of MTCAF homing, we built a nude mouse xenograft tumor model (mice, n = 10). HCT116 cells were subcutaneously co-implanted with MSCs (S1), MSCs with ICAM-1 knockdown (S2) and PBS (NC) at a ratio of 5:1. We next detected the abilities of distant migration by in vivo fluorescence image, and we found that homing to the lungs in S1 group was more stronger than that in S2 group ( Figure 3E ). To determine which cells migrate to the lungs, we next detected the CAF-specific marker genes (α-SMA and FAPA) by immunofluorescence staining, and the results showed that MTCAFs migrate to the lung in the S1 group compared with the other two groups ( Figure 3F ). These results show that the ICAM-1 gene mediates the movement of MTCAFs, which may influence the progression of colon cancer cells.

In our study, the transcriptomic analysis indicated that H-Exos activated different signals including TNFα and IL6 signaling pathways in MTCAFs ( Figure 4A ). Among the inflammatory factors, Western blotting analysis revealed that the expression of IL-6 and IL-8 was decreased, whereas MTCAFs were knocked down by ICAM-1 ( Figure 4B ). This result was further verified by ELISA to detect ICAM-1, IL-6, and IL-8 concentration of serum on day 7 ( Figure 4C ). In addition, by accessing immune cell infiltration in vivo, we showed that the number of F4/80 macrophages was lower in tumors in S2 group compared with S1 group ( Figures 4D, E ). Some studies suggest that the inflammatory factors, IL-6, IL-8, and TNFα, are major regulators of tumor stroma interaction in the cancer microenvironment (32–35). We examined their expression levels in mice serum by ELISA (n = 5) and found increased levels of IL-6 and IL-8 ( Figures 4F, G ). In conclusion, MTCAFs with ICAM-1 are able to mediates the inflammatory microenvironment.

In our study, we further explored the effect of MTCAF-derived ICAM-1 on the HCT116 cells, and we found that migration abilities ( Figures 5A, B ) and invasion abilities ( Figures 5C, D ) of HCT116 cells were obviously weakened at day 7 when MTCAFs were knocked down. We next analyzed whether MTCAFs with ICAM-1 knockdown affected the tumor progression and immune environment by using a nude mouse xenograft tumor model (mice, n = 10). HCT116 cells were subcutaneously co-implanted with MSCs (S1), ICAM-1 knockdown MSCs (S2), or PBS (NC) at a ratio of 5:1. Mice in S1 group promoted the growth of tumor compared with the other two groups in vivo ( Figures 5E, F ). Next, we assessed the tumor weight by excising the tumor from mice, and the results were similar to those observed for tumor growth ( Figures 5G, H ). Meanwhile, Ki67 staining was performed, and the results showed that the proliferation capacity of the S1 group was significantly higher than NC group, whereas knocking down ICAM-1 significantly decreased the ability of proliferation ( Figures 5I, J ). We found that HCT116 cells can also migrate into the lungs in S1 group but not in S2 group and NC group ( Figure 5K ). These results show that ICAM-1 mediates progression of colon cancer cells.

LFA-1 has been reported to be the most important ICAM-1 receptor (31). In our study, wound healing assay confirmed that ICAM-1 secreted by MTCAFs regulates migration of HCT116 cells by interacting with LFA-1 expressed on HCT116 cells ( Figure 6A ). Western blotting results showed that JAK, STAT3, and AKT were also activated in HCT116 cells co-cultured with MTCAFs after day 3 ( Figure 6B ), whereas MTCAFs were knocked down by ICAM-1, and the phosphorylation of JAK, STAT3, and AKT in HCT116 cells was significantly decreased ( Figure 6C ). Next, we detected the migration and invasion ability of HCT116 cells in different treatment groups, including HCT116 and MSC co-culture group (group A), HCT116 and MSC co-culture plus STAT3 inhibitor group (group B), and HCT116 and MSC co-culture plus AKT inhibitor group (group C). Transwell results showed that HCT116 cell invasion was significantly reduced in groups B and C compared with group A ( Figures 6D, E ). Wound healing test results showed that the migration ability of HCT116 cells in groups B and C was significantly weakened compared with group A ( Figure 6F ). Finally, immunohistochemistry demonstrated that MTCAFs can activate AKT and STAT3 signaling pathways in tumor tissues in nude xenograft tumor models ( Supplementary Figure 2 ). These results suggest that MTCAF-derived ICAM-1 promotes the progression of colon cancer cells by binding LFA-1 to activate STAT3 and AKT signaling pathways.