Induction of the Stem-like Cell Regulator CD44 by Rho Kinase Inhibition Contributes to the Maintenance of Colon Cancer–Initiating Cells

The emerging picture from recent discoveries revealed that, in some types of tumors, only a small fraction of cancer cells is capable of initiating cancer (1–3). These cancer-initiating cells, or cancer stem cells (CSC), as they are often defined because of their associated characteristics with stem cells, are one of the major foci of recent cancer research (1–3). Elucidation of the biologic and biochemical nature of CSCs will be important to understand the mechanisms of cancer development and to devise new strategies for cancer therapy.

It has been reported that CSCs are present in colorectal cancer (4, 5). Colon CSCs were identified as cells that express specific surface markers, including CD133, CD44, CD166, and ALDH1 (4–14).

Several laboratories reported that CSCs from colon cancer proliferate in vitro as spheroids (5, 12, 14–18). Remarkably, the spheres can be maintained in conditions of exponential growth for more than a year without losing the ability to generate tumors (5). Characterization of the spheroids revealed that a fraction of the cells that express surface markers such as CD133 (15, 17) or ALDH (16) are attributed to their features as cancer-initiating cells and that extrinsic factors such as IL-4 (15) or Wnt (12) mediate maintenance of CSC population.

Despite of the progress on the characterization of the colon spheroid cells, their stable culture in vitro can be maintained only from a fraction of primary cancers (12, 15), and it will be instrumental to establish more efficient methods for spheroid cultivation to clarify the common biochemical nature of colon cancer–initiating cells.

In this article, we found that inhibitors of Rho-associated protein kinase (ROCK) or of actomyosin cytokinesis markedly facilitated the formation of spheroids from primary colon cancers, and revealed that the CD44^high cells in the spheroids share common characteristics with CSCs. Unexpectedly, a fraction of CD44^−/low cells was capable of developing into CD44^high cells via the induction of CD44 expression by the ROCK inhibitor. We will discuss the potential significance of the reversible transition between CD44^high and CD44^−/low cells in light of the plasticity of CSCs and of devising a novel therapeutic strategy against colon cancer.

All human colon cancer samples were resected from patients with informed consent at the National Cancer Center Hospital (Chuo-ku, Tokyo, Japan), and all procedures were conducted under the protocol approved by the ethics committee of the National Cancer Center.

Tumor samples were minced and enzymatically dissociated with 1 mg/mL collagenase D (Roche) and 1 μg/mL DNase I (Roche) for 1 hour at 37°C, and then sequentially filtered through 100 and 70 μm cell strainers (BD Falcon). After the lysis of red blood cells with Red Blood Cell Lysis Solution (Miltenyi Biotec), the filtered cells were grown in STEMPRO hESC SFM (Invitrogen) supplemented with 8 ng/mL bFGF (Invitrogen), 20 μmol/L Y-27632 (Wako), and penicillin/streptomycin on ultra-low attachment culture dishes (Corning). For serial passage, spheroid cells were dissociated into single cells with Accutase (Invitrogen) once a week and incubated under the culture conditions described earlier. For differentiation experiments, spheroid cells were cultivated in the presence of 10% FBS on standard tissue culture dishes (BD Falcon).

Dissociated single spheroid cells were filtered, incubated with 7-AAD (BD Pharmingen) for the exclusion of nonviable cells, and double-stained with a phycoerythrin (PE)-conjugated monoclonal antibody against CD44 (G44-26; BD Pharmingen) and an allophycocyanin (APC)-conjugated monoclonal antibody against CD133 (AC133; Miltenyi Biotec). Isotype-matched mouse immunoglobulins were used as controls. Stained cells were then sorted using the FACS Aria II Cell Sorter (BD Biosciences) under the following conditions: nozzle tip diameter (100 μm), pressure (20 psi), and threshold rate (∼2,000 events/s). Subsequently, the sorted cells were analyzed using FlowJo ver.7.6 software. Viability of the sorted cells was examined by Trypan-blue staining.

For analyses of cell-cycle profiles of cells with different levels of CD44 expression, dissociated cells were double-stained with PE-conjugated anti-CD44 and APC-conjugated anti-CD133 antibodies, fixed with 70% ethanol, and incubated with 0.1 mg/mL RNase A (Qiagen) and 25 μg/mL propidium iodide (Sigma), to determine the DNA contents of double-stained cells from each fraction with different levels of CD44 expression.

Accutase-dissociated single cells or fluorescence-activated cell sorting (FACS)-sorted cells were seeded at a density of viable 1,000 cells per well on 96-well ultra-low attachment plates (Corning). Cell growth was quantified by measuring the amounts of cellular ATP from a pool of spheroid forming and nonforming cells (CellTiter-Glo Luminescent Cell Viability Assay; Promega). Spheroid formation was evaluated by counting the number of formed spheres (>∼20 μm in diameter). For quantification of the inhibitory effects of CD44 inhibition on spheroid formation, 10 μg/mL anti-CD44 neutralizing antibodies (IM7; Biolegend and 2C5; R&D Systems) or the control antibody were added at the beginning of spheroid formation. Single-cell dilution assays were conducted as described earlier (12).

For cell transplantation assays of spheroid cells, the spheroids were dissociated into single cells with Accutase, suspended in 50 μL medium containing 50% Matrigel (BD Biosciences), and used for subcutaneous injection with a 27G needle into the flank of NOD/SCID mice (Central Institute for Experimental Animals, Tokyo, Japan). All mouse procedures were approved by the Animal Care and Use Committees of the National Cancer Center and conducted in accordance with Institutional policies.

To isolate and expand cancer-initiating cells from primary colon cancer, we dissected and cultivated primary cancer cells under spheroid culture conditions (5). In an attempt to establish the optimum conditions for the maintenance and growth of cancer-initiating cells, we examined the effects on spheroid formation of chemicals that were reported to be effective in promoting the growth of normal or cancer stem cells (data not shown). It was previously reported that ROCK inhibitor promotes the survival of embryonic stem cells (19–23). After extensive screening, Y-27632, a ROCK inhibitor, stood out as a chemical that greatly facilitated spheroid formation from a colon cancer specimen (Fig. 1A).

ROCKi-IV, another ROCK inhibitor, was also effective for spheroid formation (Fig. 1B). The presence of Y-27632 at 10 to 20 μmol/L, a concentration sufficient for maintenance of embryonic stem cells (21), caused maximum enhancement of spheroid formation (Fig. 1C and D), and extended in vitro cultivation revealed that Y-27632 was required for sustained growth of spheroids (Fig. 1E). There was a striking increase in a fraction of cells with sub-G₁ DNA content in the absence of Y-27632 (Fig. 1F) indicating that Y-27632 protects spheroid cells from apoptotic cell death.

Because Y-27632 was highly effective in establishing spheroids from the aforementioned case (hereafter referred as #6), we tested the same conditions for cultivation from other cases of primary colon cancers. In aggregate, we successfully cultivated spheroids that could be maintained in vitro for 1 month in 10 of 16 cases examined (Supplementary Table S1). Continued passage revealed that 5 of 10 spheroids could be further expanded over a period of more than 6 months (Supplementary Table S1). In all 5 cases, in which spheroid culture could be sustained, the withdrawal of Y-27632 resulted in marked reduction of spheroid formation (Fig. 1A; Supplementary Fig. S1A–S1C). ROCKi-IV was also effective for spheroid formation (#9 and #20; Supplementary Fig. S1D and S1E). As in #6, Y-27632 was required for continued growth of the spheroids (#20; Supplementary Fig. S1F), and for the protection of spheroid cells from apoptotic cell death (#19 and #20; Supplementary Fig. S1G and S1H). Taken together, the presence of ROCK inhibitor markedly facilitates the proper maintenance and growth of spheroids from primary colon cancer.

We next examined whether spheroid cells formed in the presence of Y-27632 share characteristics for CSCs. We determined whether spheroid cells were capable of (i) expressing specific markers, (ii) differentiation, and (iii) tumor formation in immunocompromised mice, based on the proposed criteria for CSCs (3). First, spheroid cells expressed colon CSC–specific markers, CD44 and CD133 (#6 and #20, Fig. 2B; Supplementary Fig. S2B). Second, spheroid cells were capable of differentiating into epithelial-like cells if grown under differentiating conditions, based on their morphology (#6 and #20, Fig. 2A; Supplementary Fig. S2A), reduction of CD44 and CD133, and induction of a differentiation marker, cytokeratin 20 (CK20; #6 and #20; Fig. 2B and Supplementary Fig. S2B). Third, xenograft experiments using immunocompromised NOD/SCID mice showed that the injection of spheroid cells (1 × 10³ cells) could form tumors that were histologically indistinguishable from the original primary tumor or from its mouse xenograft (#6, #19, and #20; Fig. 2C and Supplementary Fig. S2C and S2D). Thus, spheroids included cells that meet the major criteria for CSCs.

To gain insight into whether entire spheroid cells or only a fraction retain characteristics associated with CSCs, we examined the expression of CD44 and CD133 in spheroid cells by flow cytometry analysis. Although levels of CD133 expression did not significantly differ, there were striking differences in CD44 expression among cells (#6 and #20; Fig. 3A and Supplementary Fig. S3A). To determine whether the difference in CD44 expression reflects cellular hierarchy among them, we sorted spheroid cells into CD44^−/low, CD44^med, and CD44^high fractions. As expected, the sorted CD44^high cells expressed high levels of CD44, whereas their expression in CD44^−/low was not detectable by Western blot analyses (Fig. 3C). Analyses of the cell-cycle profile revealed that a significant fraction of CD44^−/low cells underwent apoptosis (#6 and #20; Fig. 3B and Supplementary Fig. S3B). CD44^high cells could form spheroids more effectively than CD44^−/low cells (Fig. 3D), and were capable of differentiation based on the induction of CK20 and downregulation of CD44 and CD133 (#6 and #20; Fig. 3E and Supplementary Fig. S3C). In addition, CD44^high cells gave rise to CD44^med and CD44^−/low cells under the conditions of spheroid culture (#6 and #20; Fig. 3F and Supplementary Fig. S3D), indicating that there is dynamic transition from CD44^high cells to CD44^−/low cells in spheroids.

It was shown that high Wnt activity is associated with the colon CSC phenotype (12). To examine whether high levels of CD44 are associated with the activation of the Wnt pathway, we examined the activating phosphorylation of β-catenin (ser552; refs. 24–26) in the CD44^high and CD44^−/low cells. The fraction of cells with the activating phosphorylation of β-catenin was higher in the CD44^high cells than in the CD44^−/low cells (Fig. 3G), which was corroborated by Western blot analyses (Fig. 3H). Thus, β-catenin was activated in the CD44^high cells.

Enhancement of aerobic glycolysis is another phenotypic hallmark associated with normal and CSCs (27–29). We conducted microarray gene expression analyses of the CD44^high and CD44^−/low cells, and gene set enrichment analyses (30) of the expression profiles revealed that several pathways associated with glycolysis, that is glucose metabolism and the pentose phosphate pathway, were upregulated (Supplementary Fig. S3E). These are in agreement with recent reports (31), and indicate that the glycolytic metabolism is enhanced in the CD44^high cells. Collectively, these results strongly suggest that cells that express high levels of CD44 are associated with the known CSC-like characteristics.

To show that high levels of CD44 expression are required for the CSC-like properties, we inhibited CD44 with the corresponding shRNAs and determined whether inhibition of CD44 thwarts spheroid formation. Knockdown of CD44 by the shRNAs was confirmed by quantitative reverse transcription (qRT)-PCR (#6 and #20; Fig. 4A and Supplementary Fig. S4A) and Western blot analyses (#6 and #20; Fig. 4B and Supplementary Fig. S4B). In accordance with the elevated levels of apoptotic cell death in CD44^−/low cells (Fig. 3B), the inhibition of CD44 by shRNAs induced apoptotic cell death (data not shown). Examination of a remaining infectant indicated that the inhibition of CD44 by shRNAs reduced spheroid formation (#6 and #20; Fig. 4C and Supplementary Fig. S4C) as well as cell growth (Fig. 4D and Supplementary Fig. S4D).

In agreement with the inhibition of spheroid formation by the CD44 shRNAs, spheroid formation and cell growth were compromised by the inhibition of CD44 by neutralizing antibodies (#6 and #20; Fig. 4E and F and Supplementary Fig. S4E and S4F). The results from serial dilution assays confirmed that the anti-CD44 antibody inhibited the clonal growth of #20 spheroid cells (Supplementary Fig. S4G). Combined with the data presented in Fig. 3, these data indicate that CD44 expression is required to maintain the CSC-like characteristics of CD44^high cells.

The functional importance of CD44 presented in Fig. 4 prompted us to examine its expression in spheroids. Of the 5 established spheroids examined, 2 (#6 and #20) expressed higher levels of CD44 than the others (#9, #17, and #19) in the presence of Y-27632 (Fig. 5A). Strikingly, CD44 expression was markedly reduced in the absence of Y-27632 in all the spheroids, whereas CD133 expression was not significantly affected (Fig. 5A). The withdrawal of Y-27632 caused rapid reduction of CD44 within 3 days (#6 and #20; Fig. 5B and Supplementary Fig. S5A).

In addition to the standard form (CD44s), there are several variant isoforms of CD44 (CD44v) as a result of alternative splicing (32). The molecular weight of the major polypeptide detected with the anti-pan-CD44 antibody (∼170 kD) indicated that CD44v was a major form induced by Y-27632 (#6 and #20; Supplementary Fig. S5B and S5C, left columns). Western blot analyses with variant-specific antibodies revealed that CD44v9 and CD44v6 were induced by Y-27632 (Supplementary Fig. S5B and S5C, middle and right columns). The molecular weight of the major band detected with the anti-pan-CD44 antibody approximately matched that of CD44v9 (Supplementary Fig. S5B and S5C), although a longer exposure revealed that the anti-pan-CD44 antibody detected the polypeptide that roughly corresponds to the size of CD44v6 (data not shown). Thus, several variant forms of CD44 were induced by Y-27632.

To determine whether CD44 expression can be reinitiated after readdition of the ROCK inhibitor, dissociated spheroid cells were cultivated in the absence of Y-27632 for 3 days, and then reincubated with the inhibitor. The readdition of Y-27632 resumed CD44 expression, and 10 to 20 μmol/L Y-27632 was sufficient for maximum induction of CD44 (#6; Fig. 5C). Time-course analyses of CD44 induction by Y-27632 showed that CD44 was induced for 3 days after incubation (#6 and #20; Fig. 5D and Supplementary Fig. S5D). ROCKi-IV as well as Y-27632 was capable of inducing CD44 (#6 and #20; Supplementary Fig. S5E). The kinetics of the induction of CD44 were overall in line with those of spheroid formation (Fig. 1E and data not shown). Thus, given the functional role of CD44 expression in spheroid formation, these data strongly suggest that the induction of CD44v contributes to the formation of spheroids by ROCK inhibitors.

In accordance with the induction of CD44 after the readdition of ROCK inhibitor, the incubation of FACS-sorted CD44^−/low cells with Y-27632 caused an increase of CD44^med and CD44^high cells (#6 and #20; Fig. 5E and Supplementary Fig. S5F). The induction of CD44-positive cells by Y-27632 was supported by Western blot analyses (#6 and #20; Fig. 5F and Supplementary Fig. S5G). These results suggest that the ROCK inhibitor, by inducing CD44, induced reverse transition from CD44^−/low to CD44^high cells in spheroids.

qRT-PCR analyses indicated that there was no significant increase of CD44 mRNA over the same time course, indicating that CD44 was posttranscriptionally induced (#6 and #20; Fig. 5G and Supplementary Fig. S5H).

It was reported that the enhanced survival of embryonic stem cells by ROCK inhibitors is mediated via the interference of the cytokinesis pathway, and that survival of dissociated embryonic stem cells was greatly enhanced by Blebbistatin, an inhibitor of myosin hyperactivation, as well as by ROCK inhibitor (20). Treatment of spheroid cells with blebbistatin, as well as with Y-27632, induced CD44 expression (Fig. 5H and Supplementary Fig. S5I) and promoted cell growth of spheroid cells (Fig. 5I and Supplementary Fig. S5H). Thus, it is likely that the ROCK/myosin pathway mediates the CD44-dependent enhancement of spheroid cell growth as well as cell survival of dissociated embryonic stem cells.

To determine the type of CD44 variant isoforms expressed in spheroid cells (Supplementary Fig. S5B and S5C), we conducted RT-PCR analyses using isoform-specific primers (Fig. 6A; ref. 33). Overall profiles of CD44 isoforms were similar between those from the spheroids and HT29 cells (Fig. 6B and C), which were reported to express a variety of CD44 variants including CD44v6 and CD44v9. Of note, CD44v6 mediates the activation of the hepatocyte growth factor (HGF)/c-Met signaling (34). Indeed, CD44 induction by Y-27632 was associated with activating phosphorylation of c-Met (Fig. 6D), suggesting that the induction of CD44v6 augments the HGF/c-Met signaling.

Finally, we examined the capacity of CD44^high and CD44^−/low cells to generate tumors in immunocompromised mice. Both types of cells were capable of forming tumors, although the frequency of tumor formation by CD44^−/low cells was less than that by CD44^high cells (Supplementary Fig. S6A). Hematoxylin and eosin staining showed that tumors derived from CD44^high cells were histologically indistinguishable from the original primary tumor (#6; Figs. 2C and 7A). CD44^−/low cells also formed similar adenocarcinoma, although cribriform and more PAS-positive vacuoles were found in some parts of the tumor (Fig. 7A and data not shown).

Remarkably, some adenocarcinoma cells from CD44^−/low cells as well as from CD44^high cells were CD44-positive (Fig. 7B), suggesting that CD44^−/low cells generated CD44-positive tumors. Supporting this, flow cytometry analyses showed that a fraction of the CD44^−/low cell-derived tumors as well as CD44^high cell-derived ones expressed CD44 (Fig. 7C). Comparison of CD44 expression between CD44^−/low cell-derived tumor cells and the original FACS-sorted cells indicated that approximately 10% of the CD44^−/low cell-derived tumor cells became CD44-positive (Supplementary Fig. S6B). It is unlikely that CD44 expression in tumors from CD44^−/low cells was attributed to cells other than tumors, because CD44 staining was almost exclusively detected in the tubular structure of adenocarcinoma (Fig. 7B). Western blot analyses further confirmed the expression of CD44 in tumors derived from CD44^−/low cells (Fig. 7D). Judging from levels of human Topo I expression, levels of CD44 expression in CD44^high cell-derived tumors were approximately 3-fold higher than those in CD44^−/low cell-derived ones (Fig. 7D).

We also conducted transplantation assays using spheroid cells that were cultivated either in the presence or absence or Y-27632 (Fig. 5). The generated tumor from these cells showed similar CD44 staining (Supplementary Fig. S6C). Taken together, our data indicate that CD44-negative cells are capable of forming CD44-positive tumor.

In this report, we showed that the addition of ROCK inhibitors markedly facilitated the continuous growth of cells with properties for colon CSCs in vitro. Characterization of the cultivated cells showed that ROCK inhibitors were capable of sustaining spheroids that contained cells that met the major criteria for CSCs. In the formed spheroids, CD44^high cells shared characteristics for CSCs, and were capable of generating their differentiated progeny, CD44^−/low cells, as well as CD44^high cells themselves.

Recently, it was reported that CD44v contributes to the ROS resistance of gastrointestinal cancer stem-like cells (35), and associates with enhanced aerobic glycolysis (31). CD44 expression in spheroid cells is associated with the enhanced glycolytic pathways (Supplementary Fig. S3E), which may be attributed to CD44v expressed in the presence of ROCK inhibitor (Figs. 6A–C and Supplementary Fig. S5B and S5C).

Of particular interest among the CD44 variants is CD44v6, because it was reported that CD44v6 mediates the activation of HGF/c-Met pathway (34). Indeed, our data indicate that CD44 expression is associated with c-Met activation (Fig. 6D). Considering that HGF restores the colon CSC phenotype (12), CSC-like characteristics associated with CD44 expression may be induced via the activation of the HGF/c-Met pathway.

What is the mechanism for the induction of CD44 by ROCK inhibitors? It was reported that the functions of ROCK center on the regulation of cytoskeletons via the phosphorylation of its targets (36). Considering that CD44 induction by the ROCK inhibitor is posttranscriptional (Fig. 5D and G and Supplementary Fig. S5D and S5I), alterations in the cytoskeleton caused by ROCK inhibitor may lead to stabilization or enhanced translation of CD44v protein.

Notably, ROCK inhibitors facilitate the in vitro growth of embryonic stem cells by inhibiting dissociation-induced apoptosis (21), and the inhibition of the cell death is mediated via the blockage of ROCK/myosin hyperactivation (20). Similarly, the inhibition of dissociation-induced apoptosis may also be important for the establishment of colon cancer spheroids, and indeed the inhibitor was used to block cell death of dissociated organoids that are derived from colon adenocarcinoma (18). ROCK inhibitors and blebbistatin may contribute to CD44-mediated spheroid formation through the inhibition of actomyosin hyperactivation.

ROCK inhibitor may also regulate other pathways independent of CD44 induction to facilitate the formation of spheroids. In the absence of Y-27632, some spheroids (#6 and #20) express residual amounts of CD44, the levels of which are roughly equivalent to those from other spheroids in the presence of the inhibitor (#9, #17, and #19; ref. Fig. 5A). Because the capacity to form spheroids was lower in the former than the latter (Supplementary Fig. S1C), the CD44 pathway and other unknown pathways may synergize to facilitate the formation of spheroids.

The retrograde transition from CD44^−/low to CD44^high cells by ROCK inhibitor may lead to the formation of a dynamic equilibrium between these cells. Of note, similar dynamic transitions between CSC and non-CSC states were also observed in melanoma cells (37, 38) and in breast cancer cells (39, 40). It was proposed that, through mathematical analyses, stochastic transition between cancer stem-like cells and non-stem-like cells forms a dynamic equilibrium in cancer (39), and such equilibrium may be influenced by extracellular factors (12). Thus, the phenotypic plasticity of CSC-like cells we observed may be a widespread feature of solid cancer cells.

In devising a therapeutic strategy against colon cancer–initiating cells, tumors with high activity for retrograde transition may pose a major problem, because elimination of both CD44^high and CD44^−/low cells may be required for effective therapy; CD44^−/low cells, if not simultaneously neutralized, will revert back to CD44^high cells with cancer-initiating ability. The combination of killing cancer-initiating cells and blocking retrograde differentiation may be considered an effective therapy in the future.

Y. Aihara is employed (other than primary affiliation; e.g., consulting) in Sysmex Corporation as a Researcher. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H. Ohata, H. Nakagama, K. Okamoto

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Ohata, T. Ishiguro, Y. Aihara, A. Sato, H. Sakai, S. Sekine, T. Akasu, S. Fujita

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Ohata, H. Nakagama, K. Okamoto

Writing, review, and/or revision of the manuscript: H. Ohata, H. Nakagama, K. Okamoto

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Taniguchi, H. Nakagama

Study supervision: H. Nakagama, K. Okamoto

The authors thank the laboratory members for critical reading of the manuscript, Ibuki Kobayashi, and Naoaki Uchiya for technical assistance, Yasuhide Yamada for clinical suggestion, and Hideyuki Saya for anti-CD44v9 antibody. SCADS inhibitor kit, which includes Y-27632, was a gift from the Screening Committee of Anticancer Drugs supported by Grant-in-Aid for Scientific Research on Priority Area “Cancer” from MEXT, Japan.

This study was supported by a Grant-in-Aid for the Third-Term Comprehensive 10-Year Strategy for Cancer Control from the MHLW (K. Okamoto); the Program for Promotion of Fundamental Studies in Health Sciences of the National Institute of Biomedical Innovation (NiBio; K. Okamoto); Grant-in-Aid for Scientific Research in Innovate Areas from MEXT (K. Okamoto); Grant-in-Aid for Scientific Research (C) from Japan Society for the Promotion of Science (JSPS; K. Okamoto); Grant-in-Aid for Cancer Research from Foundation for Promotion of Cancer Research (K. Okamoto); Research Resident Fellowship from the Foundation for Promotion of Cancer Research (FPCR, Japan; T. Ishiguro). Grant-in-Aid for Young Scientist (B) from JSPS (H. Ohata and T. Ishiguro).

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.