Cytoplasmic accumulation of connexin32 expands cancer stem cell population in human HuH7 hepatoma cells by enhancing its self-renewal

Authors

  • Yohei Kawasaki,

    1. Department of Molecular and Tumour Pathology, Akita University Graduate School of Medicine, Akita, Japan
    2. Department of Otorhinolaryngology, Akita University Graduate School of Medicine, Akita, Japan
    Search for more papers by this author
  • Yasufumi Omori,

    Corresponding author
    1. Department of Molecular and Tumour Pathology, Akita University Graduate School of Medicine, Akita, Japan
    • Department of Molecular and Tumour Pathology, Akita University Graduate School of Medicine, 1-1-1 Hondo, Akita 010-8543, Japan
    Search for more papers by this author
    • Tel: +81-18-884-6061, Fax: +81-18-836-2601

  • Qingchang Li,

    1. Department of Molecular and Tumour Pathology, Akita University Graduate School of Medicine, Akita, Japan
    2. Department of Pathology, College of Basic Medical Sciences, China Medical University, Shenyang, China
    Search for more papers by this author
  • Yuji Nishikawa,

    1. Department of Molecular and Tumour Pathology, Akita University Graduate School of Medicine, Akita, Japan
    Search for more papers by this author
  • Toshiaki Yoshioka,

    1. Department of Molecular and Tumour Pathology, Akita University Graduate School of Medicine, Akita, Japan
    Search for more papers by this author
  • Masayuki Yoshida,

    1. Department of Molecular and Tumour Pathology, Akita University Graduate School of Medicine, Akita, Japan
    Search for more papers by this author
  • Kazuo Ishikawa,

    1. Department of Otorhinolaryngology, Akita University Graduate School of Medicine, Akita, Japan
    Search for more papers by this author
  • Katsuhiko Enomoto

    1. Department of Molecular and Tumour Pathology, Akita University Graduate School of Medicine, Akita, Japan
    Search for more papers by this author

Abstract

Although the connexin32 (Cx32)-mediated gap junction is abolished in hepatocellular carcinoma (HCC), the expression of cytoplasmic Cx32 tends to increase in correspondence with the grade of malignancy. Establishing a Tet-off expression system in human nonmetastatic HuH7 HCC cells where cytoplasmic Cx32 was overexpressed by doxycycline (Dox) withdrawal, we previously demonstrated that overexpression of cytoplasmic Cx32 made HuH7 cells metastatic in mice. In our study, hypothesizing that the cytoplasmic Cx32-induced metastasis may involve expansion of the cancer stem cell (CSC) population, we examined whether cytoplasmic Cx32 controlled the size of the side population (SP) in HuH7 Tet-off Cx32 cells. Fluorescence-activated cell sorting revealed that SP was expanded in a Dox-free medium compared with a Dox-supplemented one. Although cytoplasmic Cx32 did not block maturation from SP to non-SP, purified SP reconstituted a larger SP fraction in the Dox-free medium than in the Dox-supplemented one. Furthermore, although SP from HuH7 Tet-off mock cells formed a similar number of CSC spheres of a similar size whether with or without Dox, SP from HuH7 Tet-off Cx32 cells developed a greater number of larger CSC spheres in the Dox-free medium than in the Dox-supplemented one. Taken together, these results suggest that accumulation of cytoplasmic Cx32 should enhance self-renewal of CSC to expand the CSC population in HCC.

Connexin proteins are exclusive components of gap junction, through which 2 adjacent cells exchange small (Mr < 1,000) water-soluble molecules directly between cytoplasms of each other.1 A hexamer of connexin molecules functions as a hemichannel called “connexon,” and the connexons provided by 2 adjacent cells dock with each other in a cell–cell contact area to complete a gap junction channel. At present, the connexin gene family comprises more than 20 members in humans, and their expression patterns vary from tissue to tissue.2, 3

Gap junctional intercellular communication (GJIC) plays pivotal roles in tissue homeostasis, and its downregulation or disruption causes several human diseases, including peripheral neuropathy, deafness and cataract.4 Moreover, a considerable number of studies have revealed that GJIC suppresses cancer promotion during carcinogenesis and, thus, that GJIC is commonly downregulated in cancerous lesions5, 6 with some exceptions.7, 8 In the liver, normal hepatocytes express connexin26 (Cx26) and connexin32 (Cx32), both of which form gap junctions at cell–cell contact areas.9, 10 In hepatocellular carcinoma (HCC), the expression of these 2 connexin proteins is altered differently, i.e., while the expression of Cx26 is abolished, Cx32 remains expressed but is translocated into cytoplasm instead of plasma membrane, resulting in failure of the formation of a gap junction.11, 12 Intriguingly, as noted in various human cancers, the amount of connexin proteins that accumulate in cytoplasm tends to increase in correspondence with the grade of malignancy.11, 13–17

We previously explored the contribution of such a cytoplasmic Cx32 protein to progression of HCC by using human nonmetastatic HuH7 hepatoma cells in terms of intrahepatic metastasis.18 In that study, severe combined immunodeficiency (SCID) mice orthotopically xenografted with HuH7 cells developed intrahepatic and peritoneal metastatic foci only when a high level of Cx32 expression was induced in cytoplasm of the grafts, indicating that cytoplasmic accumulation of Cx32 protein could induce a metastatic ability in HuH7 cells.

For the last few years, an avalanche of studies has revisited a classical idea based on the hierarchy model in cancer tissues,19–21 and many of them have successfully confirmed that cancer stem cells (CSCs) could be enriched in a distinct population.22 In this model, a cancer tissue consists of a heterogeneous cell population organized in a hierarchical manner sustained by CSCs at the apex. CSCs have the potential to self-renew extensively as well as that to produce non-CSCs and are considered to be tumor-initiating cells. On the other hand, although non-CSCs account for the absolute majority of the population within a cancer tissue, their proliferation is self-limiting and, thus, non-CSCs fail to develop a tumor. Strikingly, Chiba et al.23 demonstrated that a CSC population existed even in HuH7 cell line,24 which had long been believed to be an absolutely homogenous population. According to the hierarchy model, each metastatic lesion arises from a single CSC but not from non-CSCs. The size of a CSC population is thus considered to be one of the determinants defining the incidence of metastases.21, 25, 26

From this point of view, cytoplasmic Cx32-mediated enhancement of metastasis may involve expansion of the CSC population. In our study, we examined the effects of cytoplasmic Cx32 protein on regulation of the size of the CSC population by using a Tet-off system27 where overexpression of Cx32 was induced by withdrawal of doxycycline (Dox) from the medium and found that cytoplasmic Cx32 enhanced the self-renewal of CSC without having any effect on non-CSCs.

Abbreviations:

APC: allophycocyanin; CSC: cancer stem cell; Cx26: connexin26; Cx32: connexin32; Dox: doxycycline; FACS: fluorescence-activated cell sorting; FBS: fetal bovine serum; GAPDH: glyceraldehyde-3-phosphate dehydrogenase; GJIC: gap junctional intercellular communication; HBSS: Hank's balanced salt solution; HCC: hepatocellular carcinoma; HRP: horseradish peroxidase; mAb: monoclonal antibody; MP: main population; pAb: polyclonal antibody; PBS: phosphate-buffered saline; PI: propidium iodide; SCID: severe combined immunodeficiency; SP: side population

Material and Methods

Cell culture

HuH7 Tet-off Cx32 cell clone 16-4 and HuH7 Tet-off mock cells were previously generated18 and used for all the experiments. For monolayer culture, both of them were maintained in RPMI1640 medium (Nissui Pharmaceutical, Tokyo, Japan) supplemented with 10% Tet-system-approved fetal bovine serum (FBS) (Clontech Laboratories, Mountain View, CA), 2 mM L-glutamine, 2.4 g/l NaHCO3, 100 U/ml penicillin, 100 μg/ml streptomycin and 4 μg/ml Dox (Sigma-Aldrich, St. Louis, MO). The medium was changed every other day. For several experiments, serum-free RPMI1640 medium containing 2 mM L-glutamine, 2.4 g/l NaHCO3, 100 U/ml penicillin, 100 μg/ml streptomycin, 5.2 ng/ml sodium selenite, 0.025 mg/ml insulin, 20 ng/ml EGF, 10 ng/ml FGF-2 and 4 μg/ml Dox was also used. To induce overexpression of Cx32 protein, the medium was substituted with that containing no Dox. In every culture condition, the cells were incubated at 37°C in a humidified atmosphere containing 5% CO2 in air.

To evaluate plating efficiency, 1 × 105 cells were seeded onto a 60-mm dish containing RPMI1640 medium with 10% FBS in the presence or absence of Dox. Three days after incubation, the cells on culture dishes were stained with hematoxylin. The cells in 5 areas (1 cm2 each) were counted in triplicate plates. Each value was converted to that for a 60-mm dish.

Flow cytometry and cell sorting

For side population (SP) analysis, trypsinized cell pellets were incubated at 37°C for 90 min with 20 μg/ml Hoechst 33342 (Sigma-Aldrich) in 1% FBS/Hank's Balanced Salt Solution (HBSS) in the presence or absence of 50 mM verapamil (Sigma-Aldrich), further supplemented with 4 μg/ml Dox for the cells having been maintained in the presence of Dox. After washing with HBSS, the cells were resuspended in 2 μg/ml propidium iodide (PI)/HBSS, filtered through a 40-μm cell strainer (BD Biosciences Discovery Labware, Bedford, MA) and then applied to MoFlo Cell Sorter (Beckman Coulter, Fullerton, CA). Hoechst 33342 was excited with the UV laser at 350 nm, and fluorescence emission was measured with the 405/BP30 (Hoechst blue) and the 570/BP20 (Hoechst red) optical filters. PI labeling was measured with the 630/BP30 filter for elimination of dead cells. The SP fraction was defined as the fraction that was resistant to Hoechst staining and was made undetectable by verapamil treatment.

For analysis of CD133-positive fraction, cell pellets were incubated with allophycocyanin (APC)-conjugated anti-human CD133/2 monoclonal antibody (mAb) clone 293C3 (Miltenyi Biotec, Bergisch Gladbach, Germany) at the dilution of 1:11 in the presence or absence of 4 μg/ml Dox at 4°C for 15 min. After washing with HBSS, the cells were resuspended in 2 μg/ml PI/HBSS, filtered through a 40-μm cell strainer and then applied to MoFlo Cell Sorter. To determine the negative fraction, APC-conjugated mouse IgG2b isotype antibody (Myltenyi Biotec) was excited with a helium-neon laser at 633 nm, and the fluorescence emission was measured with the 660/BP optical filter.

Sphere formation assay and immunohistochemistry

To avoid adhesion and subsequent maturation to non-CSCs, cells were cultured in a serum-free semisolid medium. A total of 1 × 103 cells each of the SP and the main population (MP) fractions were seeded in 80 μl of serum-free RPMI1640 medium, the recipe of which was mentioned above, containing 0.33% agar on 100 μl of a solidified serum-free RPMI1640 basal layer containing 0.5% agar in wells of a 96-well plate. Fifty microliters of serum-free RPMI1640 was further poured on the top of each well as a reservoir of fresh medium and growth factor, and this top liquid layer was changed every 3 days. The cells were incubated for 20 days in the same atmosphere as that for the monolayer culture. The number of formed spheres was counted, and the dimension of each sphere was analyzed by imageJ.

The spheres derived from SP cells were embedded in Tissue-Tek O.C.T. Compound (Sakura Finetek, Tokyo, Japan), frozen and sectioned in a cryostat. The 6-μm cryosections were fixed in pure acetone for 7 min at −20°C and incubated with anti-Cx32 mAb clone CX-2C2 (Invitrogen, Carlsbad, CA) diluted at 1:200. Specific signals were revealed with diaminobenzidine chromogen by using Envision System (DAKO, Carpinteria, CA). Nuclei were stained with Meyer's hematoxylin.

Xenograft into mice

Male SCID mice (C.B-17/Icr-scid/scid) were purchased from CLEA Japan (Tokyo, Japan) and maintained under a specific pathogen-free condition. The mice received humane care and were fed with autoclaved water and chow. The protocol of the animal work was approved by the Animal Research Committee, Akita University, and was in accordance with the Regulation for Animal Experimentation of the University. After cell sorting by fluorescence-activated cell sorting (FACS), each of the SP and the MP fractions was resuspended in HBSS containing 1% FBS and 4 μg/ml Dox at the density of 5 × 105 cells/ml. Each suspension of SP and MP cells was injected subcutaneously into the right and left flanks of a SCID mouse, respectively, at 200 μl per site. Apart from this experiment, 1 × 105 unsorted cells each were xenografted into SCID mice. For the group to which Dox was administered, drinking water was supplemented with 2 mg/ml of Dox, given to the mice beginning 1 week before implantation of the cells, and a new bottle was provided every 3 days. The mice were euthanized with CO2 inspiration 12 weeks after implantation, and autopsies were performed immediately. The developed tumors were fixed in 20% formalin/phosphate-buffered saline (PBS) for hematoxylin-eosin staining.

Immunoblotting analysis

Immunoblotting analysis was performed mostly as described previously.18 Anti-Cx32 polyclonal antibody (pAb) (Sigma-Aldrich) and antiglyceraldehyde-3-phosphate dehydrogenase (GAPDH) mAb clone 6C5 (HyTest, Turku, Finland) were diluted at 1:500 and 1:10,000, respectively. Horseradish peroxidase (HRP)-conjugated anti-rabbit and anti-mouse IgG antibodies (GE Healthcare Bio-Sciences, Piscataway, NJ) were diluted at 1:2,000 and 1:5,000, respectively. Finally, the protein–antibody complex was visualized with a WEST-one Western Blot Detection System (iNtRON Biotechnology, South Korea) following the protocol provided by the manufacturer.

Indirect immunofluorescence

Indirect immunofluorescence was performed as described previously.18 Anti-Cx32 pAb, anti-Golgi 58K protein mAb clone 58K-9 (Sigma-Aldrich) and anti-mannosidase II mAb clone 53FC3 (Covance, Emeryville, CA) were diluted at 1:400, 1:50 and 1:5,000, respectively. After incubation with HRP-labeled anti-rabbit IgG antibody (KPL, Gaithersburg, MD) and biotin-labeled anti-mouse IgG antibody (KPL), specific signals were revealed by Fluorophore Tyramid Amplification Reagent (Perkin Elmer Life Sciences, Boston, MA) and rhodamine-labeled streptavidin (KPL). Nuclei were stained with diamidine phenylindole dihydrochloride (KPL) at a concentration of 0.5 μg/ml.

Statistical analysis

Student's t-test was performed for the estimation of statistical significance. p values are 2 tailed. All experiments were independently repeated at least 3 times except for the tumorigenicity assay of xenografts in mice, which was performed only once.

Results

Inducible overexpression of cytoplasmic Cx32 protein in HuH7 Tet-off Cx32 cells

To assess the effects of cytoplasmic Cx32 in human HCC-derived HuH7 cells without any necessity to consider clonal diversity, we used HuH7 Tet-off Cx32 cells along with HuH7 Tet-off mock cells as a control, both of which had previously been generated.18 As shown in Figure 1a, HuH7 Tet-off Cx32 and the mock cells express similar basal levels of endogenous Cx32 protein when cultured in RPMI1640 medium supplemented with 4 μg/ml of Dox. On the other hand, the Dox-free medium induces overexpression of Cx32 protein in HuH7 Tet-off Cx32 cells but not in the mock cells. A densitometric analysis reveals an ∼5-fold induction compared with the expression level in an uninduced state (Fig. 1b). In these 2 types of HuH7 cells, as is the nature of HCC-derived cells, neither endogenous nor exogenous Cx32 protein is capable of forming any gap junction channels at a cell–cell contact area, and, instead, fluorescent signals for Cx32 protein are merged with those for Golgi markers such as Golgi 58K protein and mannosidase II (Fig. 1c and Supporting Information Fig. 1), indicating that Cx32 protein is localized in cytoplasm, notably in Golgi apparatuses.

Figure 1.

(a) Immunoblotting of Cx32 protein expressed in HuH7 Tet-off Cx32 and mock cells. The cells were cultured in the presence or absence of Dox. After incubation for 48 hr, the cells were lysed and subjected to immunoblotting to detect Cx32 protein, and GAPDH as a loading control. (b) Densitometric analysis of 3 independent immunoblottings for Cx32 protein expressed in HuH7 Tet-off Cx32 and mock cells. Error bars represent the SD (n = 3). *p < 0.01. (c) Indirect immunofluorescence of Cx32 protein (left) and Golgi 58K protein (middle) in HuH7 Tet-off Cx32 and mock cells in the presence or absence of Dox. Nuclei were stained with diamidine phenylindole dihydrochloride. Note that signals of both Cx32 and Golgi 58K proteins are colocalized (right). Scale bar, 20 μm.

Therefore, cultured in a Dox-free medium, HuH7 Tet-off Cx32 cells mimic a high-grade HCC with respect to the Cx32 expression pattern such as cytoplasmic accumulation of overexpressed Cx32 protein.

Cytoplasmic accumulation of Cx32 protein expands the SP fraction representing CSCs

A small subset of cells presenting a highly active efflux of Hoechst 33342 dye is resistant to the dye, as revealed by FACS, and is thus called “the side population (SP)” against “the main population (MP),” which is stained with the dye. It is well known that SP is the fraction into which stem cells in normal tissues are efficiently enriched.28–30 As is the case not only with normal tissues, a number of recent studies on various, otherwise not all, malignant tumors have revealed that the cells from SP but not from MP display a series of phenotypes signifying CSC.31–36 Furthermore, Chiba et al.23 found SP within the cell line HuH7 and proved that SP in HuH7 cells was the exclusive fraction containing CSCs.

Considering SP to be a CSC marker of HuH7 cells, we analyzed the effects of cytoplasmic Cx32 protein on the CSC population by measuring the size of SP with FACS. In an experiment (Fig. 2a), when HuH7 Tet-off Cx32 cells were cultured continuously for 10 days in the presence of 4 μg/ml of Dox, the verapamil-sensitive SP fraction accounted for 0.59% of the whole population. In contrast, the SP fraction in the cells cultured in a Dox-free medium expanded to up to 8.0% of the whole population. Figure 2b shows that, in HuH7 Tet-off Cx32 cells, the proportion of the SP fraction to the whole population is ∼10 times higher in a Dox-free medium than in a Dox-supplemented one. Such a Dox-dependent alteration of the size of SP is not observed in HuH7 Tet-off mock cells. Therefore, it is concluded that the cell population expressing a higher level of cytoplasmic Cx32 protein contains a larger SP fraction.

Figure 2.

(a) A representative result of FACS analysis of the SP fraction in HuH7 Tet-off Cx32 and mock cells. The cells were cultured in the presence or absence of Dox for 10 days. The trypsinized cells were stained with Hoechst 33342 dye and applied to the MoFlo cell sorter. The SP fraction was defined as the fraction that was resistant to Hoechst staining and was made undetectable by verapamil treatment. (b) Diagrammatical presentation of 5 independent FACS analyses of the SP fraction. Error bars represent the SD (n = 5). *p < 0.001. (c) Tumorigenicity of SP and MP cells in SCID mice. A total of 1 × 105 cells each of the SP and MP fractions from HuH7 Tet-off Cx32 cells were subcutaneously implanted into the right and left flanks of SCID mice, respectively. Note that a tumor developed only at the right flank, into which SP cells were injected. (d) Histology of the developed tumors from SP cells (left) and the unsorted cells (right). Hematoxylin-eosin staining. Scale bar, 100 μm.

To confirm that CSCs of our HuH7 Tet-off Cx32 and mock cells are contained exclusively in the SP fraction, we xenografted SP and MP fractions subcutaneously into a flank of SCID mice after cell sorting and examined tumorigenicity of the grafts (Table 1). As presented in Figure 2c, although 1 × 105 cells from the SP fraction developed a tumor at each inoculated site in all the mice examined, 1 × 105 cells from the MP fraction could form no palpable tumors in any mice, indicating that the CSC population resides only in the SP fraction in our HuH7 Tet-off Cx32 and mock cells as well as in parental HuH7 cells.23 It should be noted that the SP fraction-derived tumors are histologically identical to tumors raised from unsorted HuH7 Tet-off Cx32 cells (Fig. 2d).

Table 1. Tumorigenicity of xenografts subcutaneously implanted into SCID mice
inline image

Cytoplasmic accumulation of Cx32 protein elevates self-renewal rate of SP cells

CSCs have 2 distinct functions as their essential nature. One is to mature to tumor progenitor cells that then produce a great majority of the cells constituting a tumor mass. The other is to self-renew to maintain the tumor-initiating ability.21 Cytoplasmic Cx32 protein-induced expansion of SP may be explained by the enhancement of self-renewal of CSCs. We purified SP cells by FACS from HuH7 Tet-off Cx32 cells and the mock cells, which had been maintained in a Dox-supplemented medium, and cultured them in the presence or absence of Dox. Maturation from CSCs to non-CSCs is not avoidable in a monolayer culture. Consistently, although the purified SP cells were inoculated into the culture, the vast majority of the population is occupied by MP cells after a 1-week culture whether with or without Dox (Fig. 3a). However, the SP fraction from HuH7 Tet-off Cx32 cells but not from the mock cells is kept much larger in a Dox-free medium than in a Dox-supplemented one (Figs. 3a and 3b), suggesting that cytoplasmic Cx32 protein might play roles in elevating the self-renewal rate of CSCs.

Figure 3.

(a) FACS analysis of the SP fraction in the population reconstituted by purified SP cells. SP cells purified from HuH7 Tet-off Cx32 and mock cells were cultured in the presence or absence of Dox. After incubation for 1 week, the trypsinized cells were stained with Hoechst 33342 dye and applied to the MoFlo cell sorter. (b) Diagrammatical presentation of 4 independent FACS analyses of the SP fraction in the reconstituted population. Error bars represent the SD (n = 4). *p < 0.001. (c) FACS analysis of the CD133-positive fraction in the population reconstituted by purified SP cells. SP cells purified from HuH7 Tet-off Cx32 and mock cells were cultured in a serum-free medium either with or without Dox. After incubation for 1 week, the trypsinized cells were reacted with APC-conjugated anti-human CD133/2 mAb and applied to the MoFlo cell sorter. To determine CD133-negative fraction, nonspecific reaction to APC-conjugated mouse IgG2b isotype antibody was also measured (left panel). (d) Diagrammatical presentation of 5 independent FACS analyses of the CD133-positive fraction in the reconstituted population. Error bars represent the SD (n = 5). *p < 0.001.

Several studies have reported that CSCs of HuH7 cells are contained in the CD133-positive fraction.37, 38 Our pilot experiment showed that more than 80% of the SP fraction in HuH7 Tet-off Cx32 cells are CD133-positive cells in whether the presence or absence of Dox when cultured stably (Supporting Information Fig. 2). We thus evaluated the effects of cytoplasmic Cx32 protein on SP cells by analyzing the size of the CD133-positive fraction after a 1-week culture of purified SP cells in the presence or absence of Dox. As shown in Figures 3c and 3d, the CD133-positive fraction of HuH7 Tet-off Cx32 cells is significantly larger in a Dox-free medium compared with in a Dox-supplemented one.

Cytoplasmic accumulation of Cx32 protein enhances sphere formation of CSCs

Similarly to stem cells in normal tissues, when CSCs are cultured in a serum-free medium on a nonattachment dish or in a serum-free semisolid medium, they proliferate as sphere-like cellular aggregates and sustain their undifferentiated state without maturing to non-CSCs, resulting in a pure culture of CSCs.25, 39, 40 To define the roles of cytoplasmic Cx32 protein in self-renewal of CSCs, we isolated SP and MP cells separately by FACS and incubated each of them in a serum-free semisolid medium with or without Dox for 20 days. As shown in Figures 4a and 4b, MP cells have almost no capacity for sphere formation, whereas SP cells exhibit an efficient ability to develop numerous large spheres, confirming that the SP fractions of our HuH7 Tet-off Cx32 and the mock cells represent the CSC population. As expected, spheres of SP cells from HuH7 Tet-off Cx32 cells are increased in both number and size in a Dox-free medium compared with in a Dox-supplemented one (Figs. 4b and 4c). On the other hand, SP cells from HuH7 Tet-off mock cells show similar capacities for sphere formation regardless of the presence or absence of Dox (Figs. 4b and 4c). As shown in Figure 4d, Cx32 protein remains localized in cytoplasm of the sphere-forming CSCs and constitutes no gap junction plaques in a cell–cell contact area. These results clearly indicate that cytoplasmic accumulation of Cx32 protein enhances self-renewal of CSCs in HuH7 cells.

Figure 4.

CSC sphere formation in serum-free semisolid medium. A total of 1 × 103 cells each of SP and MP isolated from HuH7 Tet-off Cx32 and mock cells were seeded in a serum-free semisolid medium either with or without Dox. After incubation for 20 days, the number and the size of developed CSC spheres were recorded. (a) A representative result. (b) The number of developed CSC spheres per well in 5 independent experiments. Error bars represent the SD (n = 5). *p < 0.001. (c) The size of developed CSC spheres. The dimension of each sphere was measured and presented in pixel. Error bars represent the SD (n = 25). *p < 0.001. (d) Cytoplasmic expression of Cx32 in the spheres derived from SP cells. Scale bar, 20 μm. Proper reactivity of the antibody was confirmed by staining cryosections of a normal mouse liver, in which Cx32 gave punctate signals in a cell–cell contact area (Supporting Information Fig. 3).

Cytoplasmic accumulation of Cx32 protein is not involved in reverse conversion from non-CSCs to CSCs

Unlike normal tissues where the cellular hierarchy is rigorously organized unidirectionally from stem cells toward terminally differentiated cells, it has been assumed that the cellular hierarchy within tumors should be relatively shallow, allowing non-CSCs to revert to CSCs to a limited extent.26 Therefore, cytoplasmic Cx32 protein may enhance not only self-renewal of CSCs but also reverse conversion from non-CSCs to CSCs, contributing in part to expansion of the CSC population. To examine such a possibility, MP cells sorted from HuH7 Tet-off Cx32 and the mock cells were introduced into a 1-week monolayer culture with or without Dox. Interestingly, Figure 5a demonstrates that a small SP fraction has indeed arisen from MP cells. There is, however, no significant difference in the size of the SP fraction between the presence and the absence of Dox (Fig. 5b), reflecting that overexpression of cytoplasmic Cx32 protein is not involved in reverse conversion from non-CSCs to CSCs. Although overexpression of Cx32 protein is induced equally in both SP and MP fractions of HuH7 Tet-off Cx32 cells by Dox withdrawal (Fig. 5c), only SP cells are susceptible to cytoplasmic Cx32 protein and enter self-renewal cycles.

Figure 5.

(a) FACS analysis of the SP fraction in the population reconstituted by isolated MP cells. MP cells isolated from HuH7 Tet-off Cx32 and mock cells were cultured in the presence or absence of Dox. After incubation for 1 week, the trypsinized cells were stained with Hoechst 33342 dye and applied to the MoFlo cell sorter. (b) Diagrammatical presentation of 5 independent FACS analyses of the SP fraction in the reconstituted population. Error bars represent the SD (n = 5). (c) Immunoblotting of Cx32 protein expressed in each fraction of SP and MP from HuH7 Tet-off Cx32 and mock cells. The cells were cultured in the presence or absence of Dox. Immediately after cell sorting, the cells were lysed and subjected to immunoblotting to detect Cx32 protein, and GAPDH as a loading control. (d) Plating efficiency of SP and MP cells isolated from HuH7 Tet-off Cx32 and mock cells. A total of 1 × 105 cells each of SP and MP were seeded in a growth medium in the presence or absence of Dox. Three days after incubation, the cells on culture dishes were stained with hematoxylin and counted in 5 areas (1 cm2 each). Each value was converted to that for a 60-mm dish. Error bars represent the SD (n = 3). *p < 0.001.

A similar finding was also seen in a plating efficiency assay. As presented in Figure 5d, SP cells show a higher plating efficiency in a Dox-free medium than in a Dox-supplemented one, whereas the plating efficiency of MP cells remains pretty low in both media.

Taken together, it has become apparent that cytoplasmic Cx32 protein forces only SP cells to self-renew extensively, resulting in expansion of the CSC population in HuH7 Tet-off Cx32 cells.

Discussion

Aberrant localization of Cx32 protein in cytoplasm is rather common in HCCs and is one of the mechanisms underlying downregulation of Cx32-mediated GJIC during HCC development.6, 11 A considerable number of multidisciplinary studies have revealed that expression of cytoplasmic connexin protein tends to increase in correspondence with the grade of malignancy.13, 41 We previously reported that such cytoplasmic accumulation of Cx32 protein was not merely a cause of “loss of function” but a pathogenic factor involved in metastasis of developed HCCs.18 In our study, we hypothesized that expansion of the CSC population should mediate the prometastatic function of cytoplasmic Cx32 protein and found that cytoplasmic accumulation of Cx32 protein enhanced self-renewal of CSCs and expanded the CSC population in HuH7 cells. It is incontestable that cell motility and invasiveness are essential factors for metastasis. What these 2 factors modulate most directly is the length of the latent period between development of a primary tumor and that of its first metastatic focus. On the other hand, the number of CSCs in a tumor should be related closely to the number of metastatic lesions that develop because CSCs are so-called functional seeds that are tumorigenic at a destination site of cell migration while non-CSCs are sterile.21 A recent report clearly demonstrated that tumor cells began to circulate in peripheral blood even at the early phase of cancer development, when no risk for metastasis was clinically estimated.42, 43 Taken together, the proportion of CSCs to the whole population in a tumor should be more relevant to the extent of metastasis than the bulk of tumor cells in migration. Therefore, it is quite reasonable that cytoplasmic Cx32 protein should enhance the metastatic potential of HCC by expanding its CSC population.

A growing body of recent studies has explored specific markers for CSCs in various human cancers.22, 44 There has, however, been a much smaller number of articles describing the factors that modulate the size of the CSC population. Among them are the proteins determining or maintaining the undifferentiated state of stem cells such as BMI1, NANOG and SOX2.45–49 Their expression is upregulated in CSCs and downregulated or abolished in non-CSCs, overall mimicking the expression patterns in stem and nonstem cells in normal cell society. It is thus expected that these factors block maturation from CSCs to non-CSCs, resulting in expansion of the CSC population. On the other hand, the expression level of cytoplasmic Cx32 protein does not vary during maturation from CSCs to non-CSCs of HuH7 cells (Fig. 5c). It is obvious that cytoplasmic Cx32 protein is not a marker for CSC. Nevertheless, overexpression of cytoplasmic Cx32 protein could have no significant effects on cell growth or dedifferentiation of non-CSCs (Figs. 4 and 5). Therefore, cytoplasmic Cx32 protein does not modulate the differentiation state of CSCs or non-CSCs but leads CSCs to choose symmetric cell division to upregulate self-renewal. As a portion of the CSC population escaped from such self-renewal pressure by cytoplasmic Cx32 protein, MP cells could arise in a pure culture of SP cells even in the absence of Dox (Fig. 3). We have not found a good explanation for the fact that CSCs but not non-CSCs can respond to overexpression of cytoplasmic Cx32 protein. Susceptibility to cytoplasmic Cx32 protein may be an important feature of CSC in HCC.

One of our most intriguing findings is that the SP fraction reappeared in a cell culture into which pure MP cells had been inoculated, suggesting that reverse conversion from non-CSCs to CSCs occurred (Fig. 5a). We initially thought that this was due to an inevitable error of FACS allowing contamination of an extremely small number of SP cells into “pure” MP cells. However, it was confirmed that the performance of our device satisfied the accuracy required for our experiments (Supporting Information Fig. 4). Furthermore, when we cultured pure MP cells derived from several different human cell lines, not all of them showed reverse conversion (data not shown). Therefore, the reverse conversion from non-CSCs to CSCs we observed is real. The cellular hierarchy in immortalized cell lines may be much shallower than that in human cancers. Or reverse conversion may be seen only in monolayer culture in vitro and may not be significant in SCID mice in vivo. Consistently, even 2 × 106 MP cells failed to result in a tumor in SCID mice (data not shown). At any rate, the existence of reverse conversion would not affect the conclusion in our study.

Following up our previous study,18 this article has provided further evidence for a pathogenic function of cytoplasmic Cx32 protein favorable for progression of HCC. We recognize that several important issues still remain to be elucidated. The molecular mechanism underlying cytoplasmic Cx32-induced expansion of the CSC population is now under investigation. Relocation of Cx32 protein to the plasma membrane should not only restore tumor-suppressive GJIC but also abrogate an oncogenic function of cytoplasmic Cx32 protein. Restoration of normal subcellular sorting of connexin protein may become a novel area of research to pursue from the aspect of cancer treatment.

Acknowledgements

The authors are very grateful to Ms. Fujiko Hirasawa and Mr. Masayuki Sone for their helpful suggestions, Ms. Reiko Ito, Ms. Yuko Doi and Ms. Makiko Kawamata for their technical assistance and Ms. Eriko Kumagai for her secretarial work. This work was supported in part by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan (Y. Omori).

Ancillary