This study was approved by the ethics committee of the Changzhou Tumor Hospital (Changzhou, China). The consent form was signed by patients according to the International Ethical Guidelines for Human Biomedical Research standards. The clinical staging of CRC patients was carried out following the American Joint Committee on Cancer (AJCC) standard, meaning that clinical staging was based on cancer tissue invasion, lymph node metastasis, and distal metastasis [1]. All patients were treated according to the CRC guidelines of the Chinese Society of Clinical Oncology. During follow-up, once recurrence was found, it was recorded as the endpoint of disease-free survival (DFS). If the patient died for any reason other than CRC, the data were deleted. From 2015 to 2020, a total of 323 patients participated in the study, and 292 patients were ultimately included in the analysis (Supplementary Table 1). Cancer and paracancer tissues from patients were treated with liquid nitrogen for at least 1 h before being stored at –80 °C for subsequent Western blotting (WB) and immunohistochemistry (IHC) analyses.

In this study, we used the following antibodies: rabbit anti-MEIS1 (1:500; Atlas Antibodies, Bromma, Sweden), rabbit anti-MEIS2 (1:500; Proteintech Group, Rosemont, IL, USA), rabbit anti-MEIS3 (1:500 for WB, 1:100 for IHC; Atlas Antibodies), rabbit anti-laminin beta 1 (anti-LamB1) (1:500; Proteintech Group), rabbit anti-MMP2 (1:500; Proteintech Group), rabbit anti-E-cadherin (1:1000; Cell Signaling Technology [CST], Danvers, MA, USA), mouse anti-beta catenin (ACTB, 1:1000; Proteintech Group), mouse anti-vimentin (VIM, 1:2000; Proteintech Group), rabbit anti-ACTB (1:3000; Proteintech Group), horseradish peroxidase (HRP)-conjugated Affinipure Donkey Anti-Mouse IgG (H+L) (1:5000; Proteintech Group), HRP-conjugated donkey anti-rabbit IgG (H+L) (1:5000; Proteintech Group), rabbit anti-E-cadherin (1:500; Thermo Fisher Scientific, Waltham, MA, USA), and mouse anti-vimentin (1:500; CST). Dulbecco’s Modified Eagle Medium (DMEM) and fetal bovine serum for cell culture were purchased from Thermo Fisher Scientific. The CRC strains SW480 and SW1116 were from Shanghai ZJ Bio-Tech Co., Ltd. (Shanghai, China) and Fuheng Biology Co., Ltd. (Shanghai, China). All cell lines were validated by STR profiling and tested negative for mycoplasma. Cells were all cultured in a humidified incubator at 37 °C and 5% CO${}_2$.

The total RNA in cells was extracted with the Trizol method, and reverse transcription was performed with AMV Reverse Transcriptase (Promega, Madison, WI, USA). Quantitative PCR (qPCR) was performed using a SYBR Green PCR Master Mix (Toyobo, Tokyo, Japan) according to the manufacturer’s instructions. The 2${}^{-\mathrm{\Delta }\mathrm{\Delta }\text{Ct}}$ method was used to quantify gene expression.The detailed methods and primers for qPCR are shown in Supplementary Methods and Supplementary Table 2, respectively.

Protein extraction and WB were carried out according to the protocols described in MolecularCloning: A Laboratory Manual, with modification [27]. Briefly, RIPA buffer was used to lyse the tissues or cells, and the Bradford method was used to quantify the protein. After boiling in Laemmli buffer for 10 min, protein samples were resolved on a 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis gel, followed, and electrophoreticallytransferred to a PVDF membrane. After blocking in 5% fat-free milk powder in Tris-buffered saline with 0.1% Tween20 detergent (TBST) for 30 min, the membrane was incubated with primary antibody overnight at 4 °C. The next day, the membrane was washed three times with TBST for 5 min, and then incubated with a secondary antibody at a dilution of 1:5000 at room temperature for 2 h. The film was scanned after developing, and its gray-scale value was analyzed using Quantitative One 4.40 software to evaluate the relative protein expression. The ratio = (MEIS3c/ACTB c)/(MEIS3p/ACTBp) was employed to determine the protein abundance of MEIS3 in the CRC tissue, and the ratio = (target/ACTB) was used to determine the abundance of the target protein in the cell sample.

The tissues were dehydrated in various alcohol concentrations (i.e., 70%, 80%, 95%, 100%, and 100%), made transparent with xylene, fixed in 4% paraformaldehyde, and finally embedded in paraffin wax for 24 h. Then slices were prepared by cutting the tissues at a thickness of 8 µm using a microtome (Leica, Wetzlar, Germany). The tissue samples were reverse-treated with 100% to 50% alcohol from xylene, dipped in sodium citrate antigen repair solution, and repaired with the microwave method. Following a 10-min soak in 5% hydrogen peroxide, the tissue samples were rinsed under running water for 1 min before being incubated in 10% rabbit serum for 1 h. Then the serum was replaced with MEIS3 antibody (1:100) and incubated overnight at 4 °C. The next day, slices were washed three times with phosphate-buffered saline (PBS) and incubated in HRP-conjugated IgG (1:500) for 1 h at room temperature. The color was developed with the SABC method. Slices were sealed with neutral gum after being dehydrated with a gradient of ethanol and xylene, and the images were photographed using a microscope.

The lentiviral particles for MEIS3 silencing were packaged with the assistance of Shanghai ZJ Bio-Tech Co., Ltd. The knockdown expression vector was composed of three U6 promoters connected in series, each of which directed a short hairpin MEIS3 RNA (shMEIS3) sequence. The three shRNA sense sequences were as follows: 5${}^{\prime }$-CUUGGAAGGAGAAUGGCAUUAUCTA-3${}^{\prime }$, 5${}^{\prime }$-CUGCAAGUCAACAACUGGUUCAUTA-3${}^{\prime }$, and 5${}^{\prime }$-CUGGUGGAGAAGAUGAGGACUUGGA-3${}^{\prime }$.

The cells were first cultured in DMEM without FBS for 24 h. During this time, the Transwell chambers were soaked and moistened with 1 $\times$ PBS. Then matrigel was added and incubated in a CO${}_2$ incubator at 37 °C to congeal the gel. Next, the lower chamber was filled with 600 µL complete medium containing FBS and 200 µL cell suspension (density 5 $\times$ 10${}^5$ cells/mL) was added to the upper chamber, followed by incubation for 24 h. Cells that migrated to the lower surface of the membrane were fixed, stained with 0.5% crystal violet, and counted under a microscope.

The patient’s MEIS3 mRNA level, DFS, and overall survival (OS) analysis of The Cancer Genome Atlas (TCGA) database were performed on the Gene Expression Profiling Interactive Analysis (GEPIA) portal (http://gepia.cancer-pku.cn/) [28]. The data of GSE17537 in the Gene Expression Omnibus (GEO) database was downloaded from GenomicScape (http://www.genomicscape.com/) and analyzed again with GraphPad Prism 8.0 (GraphPad Software, San Diego, CA, USA) [29, 30]. Independent OS analysis was also implemented on this platform.

For statistical analyses, we used GraphPad Prism 8.0 (GraphPad Software, Boston, MA, USA). To examine the differences between comparisons in protein level, cell migration, and invasion, the t-test and/or one-way analysis of variance were utilized. The 5-year DFS and OS of patients with different factors were estimated with Kaplan–Meier curves and log-rank tests.

First, we analyzed the localization pattern of MEIS3-positive cells in CRC tissues by IHC. Although regulated by Wnt/$\beta$-catenin protein, the distribution pattern of MEIS3-positive cells was not consistent with the characteristics of $\beta$-catenin [31]. As shown in Fig. 1, MEIS3 protein was not found in bulk cells, which were well differentiated (Fig. 1A,B) but were present in the cell nucleus of the cancer invasion front (Fig. 1C,D). The invasion fronts dominate the process of cancer invasion into paracancer and then metastasis into distant organs [10]. The second regions containing abundant strong MEIS3-positive cells were the tumor–stroma interface (Fig. 1E,F), which was similar to VIM in tumor budding [32, 33]. By morphological characteristics, we found tumor budding cells in these MEIS3-positive cells, which have active migration ability (Fig. 1D,F). MEIS3-positive cells were also widespread in undifferentiated regions composed of cells that proliferate vigorously (Fig. 1G,H), yet the staining intensity was much weaker than the invasion fronts and tumor–stroma interface(Fig. 1C–F). Thus, the IHC results suggest that MEIS3 is related to tumor budding cells.

Fig. 1.

Localization of myeloid ecotropic viral insertion site 3 (MEIS3) protein in colorectal cancer (CRC) tissues. MEIS3 was not detected in well-differentiated CRC tissue areas (A,B) but was detected in the cancer nucleus of the growth front (C,D) and tumor–stroma interface regions (E,F) containing an abundance of budding cells (black arrows, D, F), and was widely distributed in poorly differentiated cells, although the intensity was weak (G,H). Among them, B, D, F, and H are the screenshots within the boxes of A, C, E, and G diagrams, respectively. Scale bar = 50 µm.

We employed the ratio of protein abundance between cancer and paracancer to evaluate MEIS3 protein expression. We found that the MEIS3 expression was significantly increased in CRC compared to paracancer (2.81 $\pm$ 1.45 fold; p 0.01) (Fig. 2A,B), which incrementally increased with CRC progression (stage I/stage II/stage III/stage IV: 1.71 $\pm$ 0.94/2.52 $\pm$ 0.97/3.16 $\pm$ 1.43/3.42 $\pm$ 1.31 fold, respectively; p 0.01) (Fig. 2C). Considering the association between clinical stage and cancer metastasis, we used clinical stage as an indicator of CRC metastasis.

Fig. 2.

Expression level of MEIS3 in CRC. The MEIS3 protein expression in CRC cancer and paracancer tissues was detected by Western blotting (WB) (A) based on which we analyzed the ratios of MEIS3 protein abundance in CRC tissues to paracancer tissues in our cohort (B) and GSE4107 cohort (D) and the tendency of MEIS3 expression in cancer tissues according to clinical stage in our cohort (C) and The Cancer Genome Atlas (TCGA) cohort (E). The relative protein expression of MEIS3 in CRC tissues was also analyzed according to sex (F). P, Paracancer tissue; C, Cancer tissue. **p 0.01.

To verify the results, we employed the GSE4107 cohort data to analyze MEIS3 expression in CRC tissues [34]. As shown in Fig. 2D, MEIS3 mRNA expression in cancer tissues was significantly increased to about 3 fold that in normal control tissue (p = 0.015). By employing the GEPIA platform, we found that MEIS3 mRNA levels gradually increased with clinical staging progression (Fig. 2E). Although the expression of some genes was found to be sex-related, we did not find this trend for MEIS3 in our cohort (female vs. male: 2.86 $\pm$ 1.29 vs. 3.05 $\pm$ 1. 19; p = 0.43) (Fig. 2F). Considering the relationship between clinical stage and cancer metastasis, these results show that MEIS3 expression is positively correlated with the progression of CRC metastasis.

Based on the distribution characteristics of MEIS3-positive cells (Fig. 1), the correlation between MEIS3 expression and CRC progression in tissues (Fig. 2), and the role of MEIS3 in promoting cell migration into gut tissue, we hypothesized that MEIS3 may contribute to CRC cell migration and invasion [22, 23]. We constructed lentiviral particles carrying three tandem U6-promoter-MEIS3-shRNA, which were employed to silence MEIS3 expression in SW480 and SW1116 cells (Supplementary Fig. 1A). By analyzing the mRNA and protein levels of MEIS1, MEIS2, and MEIS3, we determined that the virus particle could specifically and effectively silence MEIS3 expression (Supplementary Fig. 1B–E, Supplementary Table 2).

The scratch wound healing assay was used to analyze cell migration ability. When the wound area of the SW480 cells was covered by 67.1 $\pm$ 7.82%, only 34.3 $\pm$ 4.73% of the region was covered by MEIS3-silenced cells (p 0.001; Fig. 3A). The migrated area of SW1116 cells was also significantly decreased in MEIS3-silenced cells (wound area closure, shCTL vs. shMEIS3: 63.7 $\pm$ 5.12% vs. 41.5 $\pm$ 3.10%; p 0.001) (Supplementary Fig. 2B).

Fig. 3.

Silencing of MEIS3 expression results in a significant decrease in SW480 cell metastasis. After treatment with lentivirus particles expressing negative control shRNA or shMEIS3, we analyzed the wound closure area of SW480 cells (A) with the scratch wound healing assay, and crystal violet-stained SW480 cells (B) in transwell analysis and counted the cells. MEIS3 silencing also resulted in the decreased protein expression of VIM, LamB1, and MMP2, and increased expression of E-cadherin in SW480 cells (C). **p 0.01. Scale bar = 200 μm.

The invasion process of cancer cells relies on degrading the surrounding matrix and penetrating adjacent tissues; thus, we employed Matrigel-covered transwells to analyze the role of MEIS3 in invasion. When MEIS3 was silenced, the penetration level of SW480 cells was reduced by 61.8% (Transwell cell number, shCTL vs. shMEIS3: 101.3 $\pm$ 6.03 vs. 38.7 $\pm$ 3.05; p 0.001) (Fig. 3B). The same treatment reduced the invasion level of SW1116 cells by 44.2% (shCTL vs. shMEIS3: 97.3 $\pm$ 5.03 vs. 54.3 $\pm$ 6.11; p 0.001) (Supplementary Fig. 2B). Thus, MEIS3 silencing significantly reduced the migration and invasion ability of CRC cells.

We analyzed the protein expression of LamB1, E-Cadherin, VIM, MMP2, and $\beta$-catenin, which are responsible for the migration and invasion of tumor cells in the EMT and tumor budding process [13, 17, 33]. We found that LamB1, VIM, and MMP2 expression was significantly decreased; E-Cadherin expression was increased; and $\beta$-catenin expression did not significantly change upon MEIS3 silencing (Fig. 3C, Supplementary Fig. 2C). Therefore, high MEIS3 expression may increase CRC metastasis by enhancing tumor budding and/or the EMT.

Then we made an expression association analysis between MEIS3 and genes that regulate cancer cell metastasis and proliferation in TCGA cohort on the GEPIA platform [28]. MEIS3 expression was positively correlated with genes promoting cell migration and invasion such as VIM (R = 0.78; p 0.001), MMP2 (R = 0.71; p 0.001), fibronectin (FN1) (R = 0.63; p 0.001), and LamB1 (R = 0.15; p = 0.013); genes regulating the EMT and tumor budding process including Twist-related protein 1 (TWIST1) (R = 0.66; p 0.001), TWIST2 (R = 0.72; p 0.001), Snail family transcriptional repressor 1 (SNAI1) (R = 0.53; p 0.001), SNAI2 (R = 0.71; p 0.001), and transforming growth factor beta (R= 0.77; p 0.001). However, there was a weak or even no correlation with the genes regulating cell proliferation including cyclin D1 (R = 0.15; p = 0.011), proliferating cell nuclear antigen (R = –0.12; p = 0.023), and MYC (R = –0.1; p = 0.083) (Supplementary Fig. 3). These results suggest that MEIS3 might lead to cancer metastasis by regulating the migration and invasion abilities of CRC cells.

Based on the MEIS3 protein ratio of cancer/para-cancer, CRC patients were stratified into two groups with equal numbers. We used the Kaplan–Meier curve to analyze the 5-year DFS of CRC patients. As shown in Fig. 4A, the 5-year DFS of MEIS3-high and MEIS3-low cohorts were 61.7% and 40.6%, respectively (hazard ratio [HR] = 2.441, 95% confidence interval [CI]: 1.493–3.989; p 0.0001) (Fig. 4A, Table 1). However, when grouped by sex, age, tumor volume, or location, the 5-year DFS of patients did not show significant differences (Supplementary Table 1).

Fig. 4.

Disease-free survival (DFS) and MEIS3 level in CRC patients. The 5-year DFS of our cohort was stratified by MEIS3 level (A) and verified through independent analysis in the Gene Expression Profiling Interactive Analysis (GEPIA) portal (B) and GSE17537 dataset (C). The relationship of 5-year OS of patients and MEIS3 level are shown from our cohort (D), GEPIA platform of TCGA cohort (E), and GSE17537 by GenomicScape portal (F). The blue line indicates the MEIS3-low group, and the red line indicates the MEIS3-high group.

To verify these results, we analyzed the DFS of CRC patients from TCGA cohort on the GEPIA platform [28]. The 5-year DFS of the MEIS3-high cohort was significantly lower than that of the MEIS3-low cohort with a cutoff of 50% (p = 0.0026; Fig. 4B). Then we analyzed the GSE17537 cohort from the GEO database and found the same trend (p = 0.0073; Fig. 4C). Therefore, MEIS3 can be employed to independently assess the recurrence risk of CRC patients after surgery.

We also analyzed the 5-year OS rate grouped as the 5-year DFS. The 5-year OS of the MEIS3-high cohort was 42.6%, which was significantly lower than that of the MEIS3-low cohort (62.3%; p 0.001) (Fig. 4D). Regarding TCGA cohort, the 5-year OS of the MEIS3-high cohort was also significantly lower than that of the MEIS3-low cohort with a cutoff of 45% (p = 0.0084; Fig. 4E). Another independent survival analysis of the GSE17537 cohort also showed similar results (p = 0.00012; Fig. 4F). In conclusion, high MEIS expression was strongly correlated with a poor prognosis in CRC patients.

Currently, the prognosis for the postoperative recurrence of CRC is predominantly based on clinical stage [1]. We determined whether a more precise prognosis could be made when MEIS3 expression is introduced into the prognosis system. To this end, we performed multivariate analysis according to MEIS3 expression with The American Joint Committee on Cancer (AJCC) stage and other independent factors (Table 1). As shown in Fig. 5, the 5-year DFS of the stage II MEIS3-high cohort was significantly lower than that of the MEIS3-low cohort (53.4% vs. 67.3%, HR = 2.38, 95% CI: 0.9843–5.786; p = 0.0123) (Table 1, Fig. 5B). The 5-year DFS of the stage III MEIS3-high cohort was also significantly lower than that of stage III MEIS3-low cohort (30.9% vs. 49.5%, HR = 2.817, 95% CI: 1.370–5.792; p = 0.0038) (Table 1, Fig. 5C). Moreover, the 5-DFS was comparable between the stage II MEIS3-high cohort and stage III MEIS3-low cohort (53.4% vs. 49.5%; p = 0.23) (Fig. 5E), and between the stage III MEIS3-high cohort and overall stage IV cohort (29.6% and 30.9%; p = 0.7844) (Fig. 5F). However, when grouped by MEIS3 expression, there was no prognostic difference in the sub cohort of stage I and IV patients (Fig. 5A,D, Table 1).

Fig. 5.

DFS in different clinical stages of CRC patients. The 5-year DFS of postoperative CRC patients was stratified by MEIS3 expression according to clinical stage I (A), stage II (B), stage III (C), and stage IV (D). Comparison between the stage II MEIS3-high cohort and stage III MEIS3-low cohort (E), and between the stage III MEIS3-high cohort and stage IV cohort (F).