Cancer Stem-like Properties in Colorectal Cancer Cells with Low Proteasome Activity

This article is featured in Highlights of This Issue, p. 5159

Colorectal cancer is one of the most common human malignancies worldwide. Despite recent advances in treatment with surgery, chemotherapy, and radiotherapy, colorectal cancer remains a major cause of cancer-related deaths (1). According to recent reports, the existence of cancer stem-like cells (CSCs) is one of the main reasons for treatment resistance (2, 3). CSCs are defined as cancer cells with a self-renewal capacity and multilineage potency that exist as a small population within bulk tumors and play a critical role in tumorigenicity, cancer progression, treatment resistance, metastasis, and recurrence (4). Briefly, the CSC theory is that, although conventional cancer therapy, such as chemotherapy and/or radiotherapy, can kill non-CSCs, it will not fully eradicate CSCs. To thoroughly analyze CSCs, a specific marker that can correctly select for and identify these cells in a bulk tumor is required. CSC markers were identified for the first time in the hematopoietic system (e.g., CD34⁺/CD38⁻; ref. 5). Using similar methodology, several markers have been reported for the identification of CSCs in solid cancer, depending on the type of cancer: CD44⁺/CD24^−/low in breast cancer (6); CD24, CD26, CD44, CD133, and CD166 in colorectal cancer (7–9); CD24, CD44, and ESA in pancreatic cancer (10); and CD44, CD133, CD90, and ESA/EpCAM in hepatocellular carcinoma (11).

On the other hand, specific characteristics of CSCs based on the common biologic properties of their intracellular activities have broad utility in identifying CSCs in solid tumors (12). The relationship between proteasome activity and CSCs is receiving a lot of attention. Generally, proteasome activity is significantly elevated in human cancer cells (13, 14). Previous reports have shown that glioma and breast cancer CSCs can be purified by utilizing low proteasome activity cells (LPACs; refs. 15–19).

In the current study, the relationship between LPACs and colorectal cancer CSCs was elucidated. In addition, we explored the mechanism underlying treatment resistance to establish a new treatment strategy.

The cell culture technique and quality maintenance were described previously (20, 21). Four human colorectal cancer cell lines were used in this study. HCT116 and SW480 cells were purchased from the ATCC. DLD1 cells were purchased from the Cell Resource Center for the Biomedical Research Institute of Development, Aging, and Cancer at Tohoku University (Sendai, Japan). KM12SM cells were a kind gift from Prof. T. Minamoto (Kanazawa University, Ishikawa, Japan). All cells were grown in DMEM supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin at 37°C in a humidified incubator with 5% CO₂. All experiments were performed with cells passaged <8 times.

The isolation system for LPACs was established by engineering cells to stably express ZsGreen fused to the carboxyl terminal degron of ornithine decarboxylase (ODC) as described previously (17). The degron sequence of ODC is degraded directly by proteasomes. Therefore, cells with low proteasome activity accumulate the fluorescent fusion protein and can be detected by fluorescent microscopy or flow cytometry (FITC channel). In this study, detected cells were termed LPACs and others were non-LPACs. The LPACs were detected after 24 hours from transduction. Stable transfectants were selected using G418. This stable expression was confirmed using MG132 as proteasome inhibitor at 10 days after it was transfected (Supplementary Fig. S1).

We used FACSAria II (BD Biosciences) for cell sorting and FACSDiva software (BD Biosciences) for analysis. CD24, CD26, CD44, CD133, and CD166 were previously reported to be typical markers of CSCs in colorectal cancer (7–9). CD24 (PE/Cy7 channel; Biolegend), CD26 (PE channel; BD Biosciences), CD44 (Alexa Fluor 700 channel; BD Biosciences), CD133 (APC channel; Miltenyi Biotec), and CD166 (PE channel; BD Biosciences) were used for the analysis of CSC markers according to the manufacturers' instructions.

Chymotryptic, tryptic, and caspase activities were measured as proteasome activity (22) in the colorectal cancer cell lines using the Proteasome-Glo Assay (Promega) according to the manufacturer's instructions.

Radio- and chemotherapy analysis was performed as described previously (15, 17, 18, 23). Briefly, the percentage of LPACs after treatment was compared with the percentage before treatment. In radiotherapy, each cell line was irradiated with 8 Gy in a 137Cs Gamma Cell 40 Exactor (MDS Nordion). For chemotherapy, 5-fluorouracil (5-FU) and oxaliplatin (L-OHP), key drugs in colorectal cancer treatment, were used at their IC₅₀ dose (HCT116 5-FU; 3.90 μmol/L, L-OHP; 1.19 μmol/L, SW480 5-FU; 12.47 μmol/L, L-OHP; 0.37 μmol/L, DLD1 5-FU; 2.90 μmol/L, L-OHP; 14.16 μmol/L, KM12SM 5-FU; 3.41 μmol/L, L-OHP; 4.46 μmol/L). These assays were performed at 72 hours after treatment.

The sphere formation assay was performed as described previously (23). Briefly, sorted cells were plated in 96-well ultralow attachment plates (Corning Inc.) at a density of 100 cells per well and grown in tumorspheric culture medium (DMEM/F-12) supplemented with 20 ng/mL human platelet growth factor, 20 ng/mL epidermal growth factor, G418, and 1% antibiotic–antimycotic solution at 37°C in a humidified atmosphere of 95% air and 5% CO₂. The number of spheres (≥100 μm) was counted in all wells and differences in the average number per well evaluated.

The reactive oxygen species (ROS) activities of LPACs and non-LPACs were measured using CellROX (Molecular Probes/Life Technologies) as described previously (24). Briefly, the cells treated with CellROX dye (5 μmol/L) were incubated at 37°C for 30 minutes and analyzed by flow cytometry (BD Biosciences) using the APC channel.

Four-week-old female NOD/ShiJic-scid Jcl (NOD/SCID) mice were purchased from Clea Japan Inc. Colorectal cancer cells were sorted by FACS based on proteasome activity (LPACs and non-LPACs) as described previously and the amount of cells adjusted (2 × 10¹–1 × 10⁴ cells/25 μL medium). Each adjusted sample was mixed with 25 μL Matrigel (BD Biosciences), and a total volume of 50 μL was injected subcutaneously into both flanks of the mice under anesthesia. Tumor formation was monitored weekly after inoculation according to guidelines for endpoints in animal study proposals (http://oacu.od.nih.gov/ARAC/index.htm). The tumor volume was estimated using the following equation: volume = (length) × (width)²/2.

To elucidate the role of LPACs in colorectal cancer, microarray analysis was performed using the Agilent Microarray Platform (Human 8×60K ver.2.0). LPACs and non-LPACs were isolated from four cell lines (HCT116, SW480, DLD1, and KM12SM) using the FACSAria II (BD Biosciences). Total RNA was extracted from isolated cells using the RNeasy Micro Kit (Qiagen).

LPACs and non-LPACs were sorted from each cell line using the FACSAria II (BD Biosciences). The isolated cells were treated with 10 μg/mL Hoechst 33342 (Sigma-Aldrich) in staining medium at 37°C for 60 minutes according to the manufacturer's instructions. Cell-cycle analysis was then carried out using BD FACSDiva software (BD Biosciences; ref. 25).

For microarray analysis, 190 colorectal cancer clinical samples, excluding stage I and stage IV, were collected consecutively from Osaka University Hospital and its nine associated hospitals between 2003 and 2006. The mean follow-up time was 42.9 ± 28.9 months for patients with disease-free survival (DFS) as described previously (20, 23). In this study, no patient received preoperative chemotherapy or radiation. After surgery, patients with Dukes stage C tumors were generally treated with 5-FU–based chemotherapy. Reporting recommendations for tumor Marker prognostic studies criteria for tumor marker studies was used for the preparation of this article (26).

After reverse transcription, real-time monitoring of PCR was performed using the LightCycler System (Roche Applied Science) for quantification of mRNA expression (20). The housekeeping gene β-actin (ACTB) was used as an internal control (27).

Colorectal cancer cell lines were transfected with an EID3 expression vector (Kazusa DNA Research Institute, Chiba, Japan) using the FuGENE 6 Transfection Reagent (Promega) according to the manufacturer's protocol. Control cells were transfected with an empty control vector using the same method. The efficiency of overexpression was validated by mRNA (Supplementary Fig. S2A). The EID3 assays were performed at 48 hours after transfection.

For siRNA inhibition, the Silencer Select siRNA Kit (Applied Biosystems/Life Technologies) was used with double-stranded RNA duplexes targeting human EID3 and negative control siRNA. Colorectal cancer cell lines were transfected with siRNA using Lipofectamine RNAiMAX (Invitrogen/Life Technologies) according to the manufacturer's instructions. The efficiency of downregulation was validated by mRNA (Supplementary Fig. S2B). The EID3 assays were performed at 48 hours after transfection.

The number of living cells was counted using Cell Counting Kit-8 (DOJINDO) according to the manufacturer's instructions. After a 2-hour preincubation in the assay solution, the viable cell number in each well was determined from the absorbance at 450 nm (OD 450) as measured by a microplate reader (BIO-RAD Model 680 XR).

Statistical analyses were performed with the Student t test or Fisher exact test for categorical data and the Mann–Whitney U test for nonparametric data. Correlations were assessed with Pearson correlation coefficient test. P < 0.05 indicated a significant difference. The Kaplan–Meier method was used to examine DFS, and the log-rank test was used to determine significance. A Cox proportional hazard model was used to assess the risk ratio with simultaneous contributions from several covariates. The StatView 5.0 program (Abacus Concepts, Inc.) was used for statistical analysis.

After stable transfection, LPACs were detected by fluorescent microscopy and flow cytometry (Fig. 1A). The LPAC population was small (HCT116 0.43 ± 0.25%, SW480 0.4 ± 0.26%, DLD1 0.37 ± 0.25, KM12SM 0.57 ± 0.21%).

To validate the reliability of this system, proteasome activity was examined in purported LPACs. The chymotrypsin-like, trypsin-like, and caspase-like activities were significantly lower in purported LPACs compared with non-LPACs (Supplementary Fig. S3).

To assess the radioresistance and chemoresistance of LPACs, radiation and drug treatments were performed on colorectal cancer cells. After irradiation with 8 Gy, the LPAC population was significantly increased in HCT116 and KM12SM (HCT116 P = 0.0003, KM12SM P = 0.0003; Fig. 1B) and SW480 and DLD1 cells (SW480 P = 0.0027, DLD1 P = 0.0031). After chemotherapy (5-FU and L-OHP), the LPAC population was significantly increased in three of four cell lines (HCT116 5-FU P = 0.0004, L-OHP P < 0.0001, KM12SM 5-FU P < 0.0001, L-OHP P = 0.0010; Fig. 1C), and there was no significant difference in SW480 (5-FU P = 0.2606, L-OHP P = 0.1210). The LPAC population of DLD1 was significantly increased after L-OHP (P = 0.0173); however, there was no significant difference after 5-FU (P = 0.4918).

Furthermore, LPACs were not detected transitioning from non-LPACs after each treatment (Supplementary Fig. S4), this result supported that LPACs were enriched and selected after treatment, 5-FU, L-OHP, and irradiation with 8 Gy.

The appearance of spheres is considered to be indicative of the ability to self-renew. This phenomenon would be consistent with the development of a CSC phenotype (28). We conducted a sphere formation assay to evaluate whether LPACs possess CSC properties. The sphere formation ability of LPACs was significantly augmented compared with non-LPACs (SW480 number/well 1.71 ± 1.11 vs. 0.43 ± 0.54, P = 0.0174; DLD1 1.39 ± 0.78 vs. 0.87 ± 0.63, P = 0.0163; Fig. 2A); however, HCT116 and KM12SM did not demonstrate statistical difference (HCT116 number/well 1.54 ± 0.74 vs. 1.60 ± 0.82, P = 0.6658; KM12SM 1.49 ± 0.72 vs. 1.51 ± 0.61, P = 0.8848). Low ROS activity (24) and a dominant G₀–G₁ phase of the cell cycle (12) are also known as general characteristic of CSCs. The ROS activity of LPACs was significantly lower than that of non-LPACs (HCT116 P = 0.0006, KM12SM P = 0.0008; Fig. 2B), as was that of SW480 and DLD1 cells. In the cell-cycle assay, LPACs were enriched with cells in the G₀–G₁ phase compared non-LPACs (HCT116 P = 0.0133, KM12SM P < 0.0001; Fig. 2C; no data for SW480 and DLD1).

Next, the CSC markers of LPACs and non-LPACs were examined. Several markers were significantly enriched in LPACs compared with non-LPACs (HCT116: CD24 P = 0.0004, CD26 P = 0.0023, KM12SM: CD24 P < 0.0001, CD44 P = 0.0363; Fig. 2D; SW480: CD26 P = 0.0001, CD133 P = 0.0003, CD166 P < 0.0001, DLD1: CD26 P = 0.0016, CD133 P = 0.0015).

However, authentic identification of CSCs still relies on the in vivo tumorigenicity assay (17). The LPACs formed tumors with only 2 × 10¹ cells, whereas the non-LPACs failed to form tumors with <1 × 10² cells, suggesting that LPACs had at least 5-fold greater tumorigenicity than non-LPACs (Fig. 2E). The subcutaneous inoculation of 10⁴ cells from each population revealed more rapid tumorigenicity for LPACs than non-LPACs (P = 0.0485; Fig. 2F). This indicates that CSCs are enriched in LPACs.

Microarray analysis was performed to investigate the mechanism underlying treatment resistance and factors leading to cancer stemness in LPACs. Several cancer-related gene candidates were identified using microarray data from the four cell lines (HCT116, SW480, DLD1, KM12SM; Supplementary Table S1). EID3 was identified as a potential factor leading to treatment resistance and cancer stemness. EID3 is responsible for inhibiting differentiation in cultured cells (29, 30). However, no function of EID3 in cancer has been reported. Because of the relationship with cell differentiation, we hypothesize that EID3 may be one of the causes of treatment resistance in LPACs and may be prognostic factor.

After validation of EID3 expression in LPACs (Supplementary Fig. S5; refs. 31, 32), EID3 assays via overexpression and siRNA were performed in HCT116 and KM12SM, because SW480 and DLD1 could not downregulate EID3 using three different siRNAs.

To assess the prognostic impact of EID3 expression in patients with colorectal cancer, patients were stratified as low EID3 or high EID3. The cut-off value of low/high EID3 expression was determined using the minimum P value approach for DFS (33). No significant differences were found between the two groups regarding clinicopathologic factors, such as age, gender, tumor location, depth of tumor invasion, lymph node metastasis, histologic grade, Dukes stage, and vessel invasion (Supplementary Table S2).

In univariate analysis of DFS, EID3 expression (P = 0.0400; Fig. 3), lymph node metastasis (P = 0.0031), histologic grade (P = 0.0366), and vessel invasion (P = 0.0003; Table 1) were significant prognostic factors. A multivariate analysis of DFS demonstrated that only vessel invasion (P = 0.0116) is a significant prognostic factor (Table 1).

Furthermore, prognostic analysis of patients with colorectal cancer (n = 177) stratified by EID3 mRNA expression using publicly available datasets from PrognoScan (http://www.abren.net/PrognoScan/; ref. 34) revealed that the EID3 expression level is a significant prognostic factor (P = 0.0028) for patients with colorectal cancer after curative resection.

To explore the function of EID3, we transfected colorectal cancer cells with a plasmid-encoding EID3 or empty control vector. In the proliferation assay, forced expression of EID3 elicited significant cell proliferation at 72 hours compared with cells transfected with the empty control vector (HCT116 P = 0.0124, KM12SM P = 0.0114; Fig. 4A). Regarding treatment resistance, upregulation of EID3 in colorectal cancer cell lines conferred stronger radioresistance compared with cells transfected with empty control vector (HCT116 P = 0.0012, KM12SM P = 0.0014; Fig. 4B). Similarly, colorectal cancer–derived cells transfected with EID3 exhibited much greater resistance to the anticancer drugs 5-FU and L-OHP than cells transfected with empty control vector (HCT116 5-FU P = 0.0129, L-OHP P = 0.0256, KM12SM 5-FU P < 0.0001, L-OHP P = 0.0056; Fig. 4C).

To assess the potential relevance of EID3 as a therapeutic target, in vitro knockdown experiments were performed. Significant growth inhibition was observed in siRNA-treated colorectal cancer cells (HCT116 siRNA#1 P = 0.0003, KM12SM siRNA#1 P < 0.0001, similar results for siRNA#2; Fig. 5A). The siRNA treatment also significantly increased sensitivity to radiation (HCT116 siRNA#1 P = 0.0010, KM12SM siRNA#1 P = 0.0002, same results for siRNA#2; Fig. 5B). In addition, sensitivity to chemotherapy was significantly increased in siEID3-treated colorectal cancer cells compared with controls (HCT116 5-FU siRNA#1 P = 0.0307, KM12SM 5-FU siRNA#1 P = 0.0413, similar results for siRNA#2; Fig. 5C, left; HCT116 L-OHP siRNA#1 P = 0.0546, KM12SM L-OHP siRNA#1 P = 0.0048, similar results for siRNA#2; Fig. 5C, right).

Novel targeted therapies are required for colorectal cancer in addition to conventional therapy. Despite the growing evidence in this field, the results of current studies are not satisfying in terms of evaluating therapeutic strategies to target CSCs (35). To establish a treatment strategy for CSC, a highly sensitive method for identifying CSCs is needed.

Several CSC markers have been reported in colorectal cancer (7–9), and the population of CSCs in colorectal cancer has been reported to be 1.6% to 8.1% (8, 36). On the other hand, in the field of hematologic malignancies, in which CSCs have been well studied, the population of CSCs has been reported to be <1% (35, 37). Furthermore, some studies suggested that CSCs were enriched using two surface epitopes, such as CD44 and CD166 or CD44 and CD133 (7, 38–40). We have analyzed the expression of CD44 and CD133, but there was considerable variation in population of CD44 and CD133 between cell lines (HCT116; 35.5 ± 3.3%, SW480; 0.1 ± 0%, DLD1; 0.1 ± 0%) similar to previous reports (7, 38–40). This discrepancy in the colorectal cancer population indicates that cell-surface markers cannot be accurately used to purify true CSCs but result in enrichment of the CSC population.

On the other hand, identifying CSCs based on specific features of CSCs, such as the G₀–G₁ phase of the cell cycle, quiescence with low protein turnover, reduced metabolism, and downregulation of proteasome activity, may be one solution to these difficulties (19, 41). Among the several properties of CSCs, we focused on proteasome activity. Proteasome-dependent protein degradation is important for the regulation of cell cycle, DNA repair, apoptosis, and protein quality control (42). In general, human cancer cells use the proteasome system for survival, and proteasome activity is significantly activated in human cancer cells (13, 14). Although proteasome activity in CSCs is still controversial depending on the site of cancer origin (43, 44), several recent studies have strongly indicated a close relationship between low proteasome activity and CSCs (15–17).

From CRC cells, we isolated LPACs using FACS using a retroviral expression vector. In agreement with previously reported ratios of CSCs in the field of hematologic malignancies, our results showed that the ratio of LPACs in colorectal cancer was 0.43 ± 0.25% (HCT116), 0.4 ± 0.26% (SW480), 0.37 ± 0.25 (DLD1), and 0.57 ± 0.21% (KM12SM). Sphere formation assay, ROS activity, cell-cycle analysis, and CD markers indicated a stemness of colorectal cancer LPACs. LPACs exhibited significantly increased tumorigenicity in vivo and could form tumors in immunologically incompetent mice from as few as 20 cells (Fig. 2E). For the first time, we elucidated that LPACs may have the capacity to be CSCs in gastrointestinal tumors.

In addition to CSC markers, which we assayed, more described markers, such as CD133, CD166, Lgr5, and EpHB2 (36), were confirmed. The LPACs clearly demonstrated upregulation of several CSC markers, including CD133, CD166, Lgr5, and EpHB2. Especially, Lgr5 was enriched in all cell lines (Supplementary Fig. S6). This also supports our results.

Therapeutic resistance is a key fundamental feature of CSCs. In the treatment of CRC, 5-FU, L-OHP, and radiation treatment are important milestones (45, 46). Our results indicate that LPACs confer significant resistance to colorectal cancer cells against 5-FU, L-OHP, and radiotherapy. Understanding this resistance mechanism should be necessary to establish a novel cancer therapy targeting CSCs. Thus far, the detailed therapeutic resistance mechanism conferred by LPACs has not been elucidated. Microarray analysis using LPACs and non-LPACs revealed an upregulation of several interesting genes related to cancer survival, such as matrix metallopeptidase 1 and ATP-binding cassette subfamily B member 1 (ABCB1). We focused on EID3 because a unique function of EID3 on cell differentiation was reported recently.

EID3 was identified as a negative regulator of cellular differentiation, as it blocks cellular differentiation by binding to class I histone deacetylases (HDAC) or CBP/p300 (29, 30). The relationship between EID3 and cancer has not been reported previously. We hypothesized that colorectal cancer LPACs acquire the capacity for treatment resistance by using some function of EID3. Confirming our hypothesis, colorectal cancer–derived cells transfected with EID3 exhibited much greater resistance to radiotherapy and anticancer drugs (i.e., 5-FU and L-OHP) than cells transfected with empty control vector (Fig. 4B and C). These results were also supported by clinicopathologic data (Fig. 3; Table 1). To probe the potential of a novel targeted therapy, an EID3 knockdown assay was performed using siRNA. Knockdown of EID3 improved radiation and chemosensitivity (Fig. 5B and C). Accordingly, EID3 may be a novel target gene in colorectal cancer. Interestingly, expression of EID3 did not regulate proteasome activity itself in colorectal cancer (data not shown). Therefore, EID3 may be one of the downstream factors regulating treatment resistance under low proteasome activity. Here, this characteristic and novel function of EID3 was shown for the first time.

EID3 has the ability to bind HDACs or CBP/p300 known as one of the important molecule in wingless-type MMTV integration site family (WNT)–β-catenin pathway (29, 30). WNT–β-catenin signaling plays crucial roles in cell fate, proliferation, survival, and migration, and it has been found that WNT is enriched in CSCs (47, 48). Interestingly, β-catenin, one of the key molecules in WNT signaling, is a target of the proteasome degradation system (49). Taking all of these things together, EID3 may play a crucial role in the maintenance of cancer stemness, although additional verification experiments are required same as previous report (50).

In conclusion, our data indicate that colorectal cancer LPACs exhibit treatment resistance and have the capacity for cancer stemness. Moreover, we identified EID3 as one of the potential regulators of treatment resistance in LPACs for the first time. Although further analyses are required, EID3 may be a potential therapeutic target for CSC treatment in colorectal cancer.

No potential conflicts of interest were disclosed.

Conception and design: K. Munakata, M. Uemura, J. Nishimura, K. Murata, H. Yamamoto, M. Mori

Development of methodology: K. Munakata, M. Uemura

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Munakata, S. Nishikawa, I. Takemasa, T. Kato

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Munakata, M. Uemura, Y. Takahashi, J. Nishimura, J.M. Carethers, H. Yamamoto, M. Mori

Writing, review, and/or revision of the manuscript: K. Munakata, M. Uemura, T. Mizushima, K. Murata, J.M. Carethers

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Munakata, M. Uemura, Y. Kano, S. Nishikawa, T. Fukusumi, M. Mori

Study supervision: M. Uemura, S. Tanaka, T. Hata, J. Nishimura, I. Takemasa, T. Mizushima, M. Ikenaga, H. Yamamoto, Y. Doki, M. Mori

Other (additional experiment): K. Kawai, T. Kitahara, M. Miyo

This work was supported by JSPS KAKENHI grant number 15K10140 and the Takeda Science Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.