The KCl cotransporter (KCC) is a major determinant of osmotic homeostasis and plays an emerging role in tumor biology. This study stresses the important role of KCC4 in tumor malignant behavior. Real-time reverse transcription-PCR on samples collected by laser microdissection and immunofluorescent stainings with different KCC isoform antibodies indicate that KCC4 is abundant in metastatic cervical and ovarian cancer tissues. Insulin-like growth factor I (IGF-I) and epidermal growth factor (EGF) stimulate KCC4 recruitment from a presumably inactive cytoplasmic pool of endoplasmic reticulum and Golgi to plasma membrane along actin cytoskeleton that is significantly inhibited by LY294002 and wortmannin. Throughout the trafficking process, KCC4 is incorporated into lipid rafts that function as a platform for the association between KCC4 and myosin Va, an actin-dependent motor protein. KCC4 and ezrin, a membrane cytoskeleton linker, colocalize at lamellipodia of migratory cancer cells. Interference with KCC activity by either an inhibitor or a dominant-negative loss-of-function mutant profoundly suppressed the IGF-I–induced membrane trafficking of KCC4 and the structural interaction between KCC4 and ezrin near the cell surface. Endogenous cancer cell invasiveness was significantly attenuated by small interfering RNA targeting KCC4, and the residual invasiveness was much less sensitive to IGF-I or EGF stimulation. In the metastatic cancer tissues, KCC4 colocalizes with IGF-I or EGF, indicating a likely in vivo stimulation of KCC4 function by growth factors. Thus, blockade of KCC4 trafficking and surface expression may provide a potential target for the prevention of IGF-I– or EGF-dependent cancer spread. [Cancer Res 2009;69(22):8585–93]
- KCl cotransport
- membrane trafficking
KCl cotransport is the coupled electroneutral movement of K and Cl ions carried out by four protein isoforms, KCC1 to KCC4. These transporters play an important role in epithelial ion transport and osmotic homeostasis (1). The activities of KCC1, KCC3, and KCC4 are osmotically sensitive and involved in cell volume regulation (2). The neuron-specific KCC2 is critical for the maturation of inhibitory γ-aminobutyric acid responses in the central nervous system by the control of intracellular Cl− concentration (3, 4). KCC activity in RBC was undiminished in KCC1 knockout mice, decreased in KCC3 knockout mice, and almost abolished in mice lacking both isoforms (5). Loss of KCC3 caused deafness and neurodegeneration and reduced seizure threshold (6). KCC3 also plays an important role in regulating cell proliferation (7). Deafness and renal tubular acidosis were noted in mice lacking KCC4 (8).
Our previous studies highlight the important role of KCC in tumor development and progression. The malignant transformation of cervical epithelial cells is associated with the differential expression of volume-sensitive KCC activities (9). The loss-of-function KCC mutant cervical cancer cells exhibit inhibited cell growth accompanied by decreased activities of the cell cycle gene products and matrix metalloproteinase (MMP; ref. 10). In addition, KCC activation by insulin-like growth factor I (IGF-I) stimulation plays an important role in IGF-I signaling to promote the growth and spread of gynecologic cancers (11). However, most functional studies were done on KCl cotransport fluxes without knowing the molecular details. By overexpression or knockdown of specific KCC isoforms, we showed that IGF-I upregulates KCC3 and KCC4, which are differentially required for breast cancer cell proliferation and invasiveness (12). In addition, KCC3 overexpression downregulates E-cadherin/β-catenin complex formation by inhibiting transcription of the E-cadherin gene and accelerating degradation of β-catenin protein, thereby promoting epithelial-mesenchymal transition of cervical cancer cells (13).
IGF-I and epidermal growth factor (EGF) are known to be overexpressed in most types of cancer tissues and notably contribute to cancer resistance to existing treatments (14–16). Our previous study also showed that IGF-I and EGF are two most potent stimulators for gynecologic cancer cell invasiveness (17). This study aims to investigate the contribution of individual KCC isoforms in cancer metastasis. The results indicate that metastatic cancer tissues express abundant KCC4, which benefits cancer cells in invasiveness. IGF-I and EGF stimulate the membrane recruitment of KCC4, in which KCC4 interacts with ezrin, an actin-binding protein, at lamellipodia. We thus propose that, in addition to ion transport, KCC4 can function as a membrane scaffold in the assembly of signal complexes.
Materials and Methods
Primary antibodies and reagents
Antibodies against KCC1, mannosidase II, β1 integrin, and αvβ3 integrin and function-blocking antibodies against α1, α4, β1, and αvβ3 integrins were purchased from Chemicon. KCC3 and KCC4 polyclonal antibodies were generated with the epitopes of KKARNAYLNNSNYEEGDEY and AERTEEPESPESVDQTSPT, respectively (18, 19). KCC1 polyclonal antibody was also generated against KCC1 COOH-terminal amino acids 1074–1085 (10). Antibodies against KCC4, EGF, calnexin, myosin Ib, myosin Va, and myosin VI were obtained from Santa Cruz Biotechnology. Antibody against IGF-I was from Upstate Biotechnology. Antibodies against E-cadherin and ezrin were from BD Transduction Laboratories. Antibodies against phospho-p44/42 mitogen-activated protein (MAP) kinase (Thr202/Tyr204) and phospho-Akt (Ser473) were from Cell Signaling Technology. IGF-I, EGF, methyl-β-cyclodextrin (MβCD), PD98059, LY294002, wortmannin, [(dihydroindenyl)oxy]alkanoic acid (DIOA), and antibody against β-actin were from Sigma-Aldrich.
Cell cultures, invasion assay, and MMP zymography
Human cervical cancer SiHa cell line, ovarian cancer OVCAR-3 cell line, lung cancer AS-2 cell line, and breast cancer T47D cell line were prepared as previously described (11, 20, 21). The clones of cervical cancer SiHa cell lines with KCC1, KCC3, or KCC4 overexpression were established previously (13). Invasive migration was done in the BD Matrigel invasion chamber (BD Biosciences) for 6 h for cervical cancer SiHa cells and for 12 h for ovarian cancer OVCAR-3 cells and lung cancer AS-2 cells in serum-free culture medium at 37°C, as an index of invasive activity of tumor cells (17, 22). Conditioned media from the invasion assays were cleared of cells and debris by centrifugation at 3,000 × g for 10 min. MMP-2 activity was measured in conditioned media by gelatin zymography (23).
Surgical specimens, laser microdissection, and real-time reverse transcription-PCR
From January 1999 to November 2001, 150 cervical cancer patients with International Federation of Gynecology and Obstetrics (FIGO) staging Ib to IIa were scheduled for radical hysterectomy and pelvic lymphadenectomy at National Cheng Kung University Hospital, Taiwan. Patients who had undergone the loop electrosurgical excision for the transformation zone of uterine cervix before radical hysterectomy and who had unusual histopathology such as small-cell carcinoma and adenosquamous carcinoma were excluded. Paraffin sections of surgical specimens were immunostained with polyclonal KCC4 antibody, followed by Alexa 488–labeled secondary antibody and Hoechst 33258 (Molecular Probes). At least five sections of each specimen were analyzed for histology and graded for KCC4 expression over 15 to 20 high-power fields, examined by two investigators trained in gynecologic pathology. Low grade indicates that the distribution of KCC4 staining is less than 50% of tumor area, whereas high grade indicates that the distribution of KCC4 staining is more than 50% of tumor area. For laser microdissection, frozen specimens of cervical squamous carcinoma or ovarian serous adenocarcinoma were obtained at the same hospital. These cases contained the following tissues representing cancer metastatic process: (a) normal tissues, (b) primary tumor tissues, and (c) metastatic tumor tissues. The preparation protocol for laser microdissection and the detail of amplification primers and probes were described elsewhere (13). The RNA levels of individual KCC isoforms in the surgical specimens were examined by real-time quantitative reverse transcription-PCR with the use of ABI Prism 7900 Analyzer and TaqMan probes (Applied Biosystems). cDNA was normalized against glyceraldehyde-3-phosphate dehydrogenase (GAPDH).
Small interfering RNA
Cells were transfected with 100 nmol/L of either targeting or control nontargeting small interfering RNA (siRNA) using Lipofectamine 2000 (Invitrogen) for 48 h. The functional assays were subsequently done after validation of KCC4 or myosin Va knockdown. The siRNA targeting exon 24 of human KCC4 (NM_006598; sense, 5′-GGUUGUCCAUAAAGAGGUUtt-3′; antisense, 5′-AACCUCUUUAUGGACAACCtt-3′) was from Ambion. The myosin Va siRNA pool of three duplexes [5′-GCAAGAAUGUGUUAGAGAAtt-3′ (sense) and 5′-UUCUCUAACACAUUCUUGCtt-3′ (antisense); 5′-CAAGUGGCCUAUCUAGAAAtt-3′ (sense) and 5′-UUUCUAGAUAGGCCACUUGtt-3′ (antisense); 5′-CAAGCGUCAAGAACUAGAAtt-3′ (sense) and 5′-UUCUAGUUCUUGACGCUUGtt-3′ (antisense)] targeting exons 18, 19, and 28 of human myosin Va (NM_000259) were from Santa Cruz Biotechnology.
Immunofluorescent images and quantitative analyses
Immunofluorescent studies were done using affinity-purified antibodies against various molecules. The fluorophores were excited by laser at 405, 488, or 543 nm and detected by a scanning confocal microscope (FV-1000, Olympus). Positive KCC4 membrane staining was defined as the distribution of KCC4 staining in more than one third of the plasma membrane area. A pixel-by-pixel analysis by FV-1000 software was used to assess the colocalization of KCC4 with other molecules in confocal images.
Immunoblotting, surface biotinylation, raft fractionation, and dot blot
For Western immunoblotting, the preparation of cell lysate was described in detail elsewhere (13). Plasma membrane proteins for surface biotinylation assay were isolated with the cell surface protein isolation kit (Pierce Biotechnology) according to the manufacturer's instructions. Bands in the Western blots were quantified using ImageJ 1.30 software (NIH) and normalized to the total protein fraction. For raft fractionation and dot blot, cells were harvested in ice-cold protein lysis solution [50 mmol/L Tris (pH 7.5), 150 mmol/L NaCl] containing 1% Triton X-100 and a protease inhibitor mixture (Roche Diagnostics), and lysates were adjusted to contain 40% iodixanol gradient medium (Opti-prep, Sigma-Aldrich). The lysates (2 mL) were placed in an ultracentrifuge tube and overlaid with 4 mL of lysis solution containing 30% iodixanol gradient medium and 2 mL of lysis solution containing 5% iodixanol gradient medium. The lysates were subjected to ultracentrifugation in a SW40Ti rotor (Beckman) for 18 h at 40,000 rpm at 4°C. Eight fractions were collected. Proteins in each fraction were concentrated and precipitated with trichloroacetic acid, redissolved in lysis solution, and subjected to immunoblotting.
All values were reported as mean ± SEM. Student's paired t test, unpaired t test, or χ2 test was used for statistical analyses. Survival data were assessed by Kaplan-Meier methods and differences were compared by the log-rank statistic. Differences between values were considered significant when P < 0.05.
KCC4 is abundant in metastatic tumor tissues
We first analyzed the expression patterns of KCC isoforms in the surgical specimens of cervical and ovarian carcinomas. We selected the cases of cervical cancer containing the following tissues for laser microdissection: (a) normal or noncancerous squamous epithelia, (b) primary tumor tissues, and (c) metastatic tumor tissues in the parametrium and pelvic lymph nodes (Fig. 1A). These tissues represented the progression of cervical carcinoma. Compared with normal or noncancerous squamous epithelia, mRNA levels of KCC3 and KCC4 were significantly increased 25 ± 3-fold and 7 ± 3-fold in the primary tumor (n = 8), respectively (Fig. 1A). In contrast, KCC4 was the most abundant KCC isoform in the metastatic tumor tissues, such as invaded parametrium or metastatic pelvic lymph nodes. Similar results were observed in the cases of ovarian cancer (Fig. 1B). The immunofluorescent stainings by different KCC isoform antibodies confirmed that KCC4 is abundant in metastatic tumor tissues (Fig. 1C).
The association between KCC4 expression and clinical outcome was studied. KCC4 protein was scanty in noncancerous cervical squamous epithelial tissues (n = 80). In contrast, primary cancerous tissues of cervix clearly expressed KCC4 protein at different levels (Supplementary Fig. S1). The patients were grouped by KCC4 grading and the survival data were analyzed accordingly. Compared with low-grade KCC4 expression, primary tumor with high-grade KCC4 expression presented the significantly higher percentage of parametrium invasion and pelvic lymph node metastasis, which are two major poor prognostic factors for early-stage cervical cancer (Table 1). Consistently, increased KCC4 expression was associated with the poor clinical outcome (Fig. 1D).
To study KCC-mediated tumor behavior, we have established previously various clones of cervical cancer SiHa cell lines with KCC1, KCC3, or KCC4 overexpression (13). Among these clones, KCC4 was dominant in enhancing cancer cell invasiveness (Supplementary Fig. S2A). In addition, KCC4 knockdown by siRNA reduced the invasive migration of ovarian cancer OVCAR-3 cells in parallel with decreased MMP-2 activity (Supplementary Fig. S2B).
IGF-I and EGF stimulate KCC4 membrane trafficking
Interestingly, in 80% of surgical specimens we examined, KCC4 protein colocalized with IGF-I or EGF in metastatic ovarian cancer tissues and metastatic lymph nodes of cervical cancer (Supplementary Fig. S3A and B). We studied whether KCC4 expression is important for IGF-I– or EGF-stimulated cancer cell invasiveness. As shown in Supplementary Fig. S3C, endogenous invasiveness of ovarian cancer OVCAR-3 cells was significantly attenuated by KCC4 knockdown, and residual invasiveness was much less sensitive to either IGF-I or EGF stimulation.
We have previously shown that chronic treatment (>12 hours) with IGF-I increased KCC4 biosynthesis in cancer cells (11, 12). Here we studied the dynamic process of KCC4 in the acute response to growth factor. Most ovarian cancer OVCAR-3 cells (>85%) displayed the quiescent phenotype in the absence of growth factor, in which KCC4 randomly scattered in the cytosol with punctuate spots representing KCC4-containing vesicles (Fig. 2A). On IGF-I or EGF stimulation for 30 minutes, more than 45% of OVCAR-3 cells exhibited pronounced KCC4 staining with continuous distribution or punctuate clusters in the juxtamembrane region of lamellipodia (Fig. 2A). IGF-I and EGF also induced the membrane recruitment of KCC4 in migratory lung, cervical, and breast cancer cells (Supplementary Fig. S4). In a surface biotinylation experiment, IGF-I increased membrane KCC4 abundance by 50%, but total KCC4 amount was not changed (Fig. 2B). These results indicate that IGF-I and EGF induce KCC4 membrane trafficking.
We analyzed the juxtanuclear location of KCC4 aggregation. A large proportion of juxtanuclear KCC4 colocalized with endoplasmic reticulum (ER) marker protein calnexin in the quiescent cells (Fig. 2C). A small proportion of juxtanuclear KCC4 was located at the Golgi, as shown by the colocalization of KCC4 with Golgi marker protein mannosidase II. IGF-I increased a large proportion of juxtanuclear KCC4 accumulated at the Golgi. The redistribution of KCC4 between ER and Golgi became apparent in a time-dependent manner (Fig. 2D), which was consistent with the increased KCC4 abundance at the plasma membrane. This suggests that IGF-I stimulation of KCC4 membrane trafficking is via a direct transport pathway (i.e., from ER to Golgi), then targeting to the plasma membrane.
The regulatory mechanisms of KCC4 membrane trafficking
Phosphoinositide 3-kinase (PI3K) and MAP kinase represent the distinct signaling pathways that mediate the biological functions of IGF-I (24). In ovarian cancer OVCAR-3 cells, IGF-I treatment resulted in extracellular signal–regulated kinase-1/2 and Akt phosphorylation that were abolished by specific inhibitors (Supplementary Fig. S5). These specific inhibitors decreased IGF-I stimulation of KCC4 trafficking to different extents (Fig. 3A), indicating that PI3K activation is the dominant signaling controlling KCC4 recruitment.
Lipid rafts, the membrane microdomain rich in cholesterol and sphingolipid, have been implicated in recruiting several ion transport systems to form signaling complexes (25, 26). KCC4 mainly distributed in the lipid raft fractions after IGF-I treatment (Fig. 3B). Moreover, membrane recruitment of KCC4 was markedly reduced by MβCD-induced cholesterol depletion (Fig. 3C). Integrin signals may regulate the location of lipid rafts and thereby control domain-specific signaling events in anchorage-dependent cells (27, 28). Expression of integrins such as αvβ3, β1, α4, and β4 has been reported in human ovarian cancer cells (29). Function-blocking monoclonal antibody against integrin αvβ3 or β1, but not α4 or β4, significantly inhibited IGF-I– or EGF-stimulated KCC4 membrane trafficking (Fig. 3D). Thus, in addition to EGF and IGF-I receptor signaling, integrin signaling participates in regulating membrane recruitment of KCC4.
KCC4 trafficking depends on actin-associated myosin
Protein trafficking between organelles usually depends on actin filaments with myosin motors or microtubules with associated motors (30). On IGF-I stimulation, the complex networks of actin filament and microtubule were apparent at juxtamembrane area, where KCC4 seemed to colocalize with actin filament (Fig. 4A). Cytochalasin D disrupted actin filaments into discontinuous spots and, as a consequence, KCC4 aggregated in the cytosol (Fig. 4A). This indicates that KCC4 membrane trafficking was abolished by cytochalasin D. In contrast, the collapse of microtubule complex induced by colcemide showed little effect on KCC4 membrane expression.
We dissected the actin-associated motors in charge of KCC4 membrane trafficking. Myosins Ib, Va, and VI are well known for powering various intracellular trafficking among actin-associated motors (30). The dynamics of myosins Ib, Va, and VI were sensitive to IGF-I stimulation (Fig. 4B). Among them, KCC4 apparently colocalized with myosin Va at lamellipodia of migratory OVCAR-3 cells (Fig. 4B). The EGF and IGF-I effect of increasing the surface expression of KCC4 was significantly attenuated in the presence of myosin Va–specific siRNA (Fig. 4C and D), indicating that myosin Va motor protein powers dominantly KCC4 membrane trafficking along actin cytoskeleton.
KCC4 functions as a plasma membrane scaffold
We are interested in knowing whether KCC4 functions as a membrane scaffold protein to facilitate the cytoskeletal reorganization that is required for the invasive migration. The neuron-specific KCC2 has been shown to have a structural interaction with actin-binding ezrin/radixin/moesin proteins that is important for the maturation of dendritic spines (31). The ezrin/radixin/moesin family is associated with a number of membrane proteins at the cytoplasmic face of the plasma membrane and is involved in epithelial cell migration and adhesion (31). As shown in Fig. 5A, IGF-I induced the extensive formation of lamellipodia, where KCC4 was associated with ezrin, as shown by quantitative pixel-by-pixel analyses of confocal images. DIOA, a KCC inhibitor, inhibited the membrane recruitment of KCC4 as well as ezrin in a concentration-dependent manner (Fig. 5A). We also overexpressed the loss-of-function KCC mutant in lung cancer AS-2 cells to study the functional role of KCC4 as a membrane scaffold. The previous studies have shown that removal of the NH2-terminal 117 amino acids (ΔN117KCC1) from KCC1 produced a dominant-negative loss-of-function phenotype for KCl cotransport in different types of cells (10–12, 32). Overexpression of ΔN117KCC1 in AS-2 cells did not change the cellular levels of KCC4 and ezrin (Fig. 5B). On IGF-I stimulation, more than 40% of AS-2 cells exhibited pronounced KCC4 distribution at or near the surface area, where ezrin apparently colocalized with KCC4 at lamellipodia (arrows in Fig. 5C). The loss-of-function KCC mutant cells presented a striking contrast that KCC4 membrane trafficking and ezrin recruitment were almost abolished regardless of IGF-I stimulation (arrowheads in Fig. 5C and D). More importantly, endogenous invasiveness of AS-2 cells was significantly attenuated in loss-of-function KCC mutant cells and the residual invasiveness was much less sensitive to IGF-I stimulation. These results indicate that KCC activity is necessary for the membrane recruitment of KCC4 as well as ezrin.
This study stresses the important role of KCC4 in tumor malignant behavior. We show that ovarian and cervical cancer cells benefit from KCC4 overexpression in the invasive migration. This conclusion is supported by the following evidence: (a) KCC4 is overexpressed in the metastatic ovarian and cervical cancer tissues at both mRNA and protein levels; (b) KCC4 overexpression enhances, whereas KCC4 knockdown attenuates, cancer cell invasiveness. IGF-I and EGF are two potent growth factors that stimulate cancer invasiveness and metastasis (17). Our previous studies showed that chronic treatment with IGF-I upregulates KCC biosynthesis, which plays an important role in IGF-I signaling to promote growth and spread of gynecologic cancers (11, 12). Here we highlight the novel mechanisms by which IGF-I and EGF increase KCC4 surface expression. IGF-I and EGF stimulate KCC4 recruitment from a presumably inactive cytoplasmic pool, such as ER and Golgi, to the active membranous target pool along the actin cytoskeleton. PI3K activation is the dominant signal controlling KCC4 trafficking. Throughout the trafficking process, KCC4 is incorporated into lipid rafts, which function as a platform for the association between KCC4 and lipid raft–associated myosin motor. The high correlation of KCC4 expression with growth factors in metastatic cancer tissues suggests a likely in vivo stimulation of KCC4 function and biosynthesis by IGF-I or EGF, which contributes to cancer invasiveness and metastasis. Taken together with our previous studies, the mechanisms by which growth factors increase KCC4 surface expression could be summarized as follows. The increased cellular functions of KCC4 are highly dependent on the upregulation of de novo protein synthesis, which occurs within hours, as well as the regulation of dynamic protein trafficking from inactive cytoplasmic pools to the active membranous ones, which mostly takes several minutes and can respond to stimulation immediately. The membrane dynamics of KCC abundance have been described in neurons and lens. Oxidative stress and seizure attack induced a rapid loss of tyrosine phosphorylation of KCC2 that resulted in translocation of the protein and functional loss of transport activity (33). Osmotic stress could modulate the subcellular localization of KCC isoforms to different extents in peripheral fiber cells of lens (34).
We considered three possible explanations for the potent effect of KCC4 overexpression on cancer invasion and metastasis. First and most simply, KCC4 abundance changes the regulatory mechanisms of cell volume control. Cancer invasion is a complex process of cell adhesion, migration, and secretion of different classes of enzymes (35). Cell migration involves substantial alterations of cell volume, and thus volume regulatory mechanisms are certainly activated (36, 37). Volume-sensitive KCC activity significantly contributes to the volume regulation of cervical and ovarian cancer cells (10, 11). KCC4 is one of the KCC isoforms critically involved in cell volume regulation (6). Cellular invasiveness is thus affected by KCC4 overexpression. Second, KCC activity seems to be associated with MMP activation (10). An important event in cancer invasiveness is the activation of MMP cascade to dissolve the basement membrane matrix (38, 39). We previously showed that the loss-of-function KCC mutant cervical cancer cells exhibited inhibited cellular invasiveness accompanied by decreased MMP activity (10). Here we show that KCC4-specific siRNA reduced cellular invasiveness in parallel with decreased MMP-2 activity. Third, and perhaps more importantly, KCC4 may function as a plasma membrane scaffold with ezrin in the assembly of cytoskeletal reorganization complex. The association of actin cytoskeleton with the plasma membrane regulates the integrity of the cortical cytoskeleton and maintains cell shape and adhesion, which are required for cellular invasive migration. The cytoskeletal and signaling functions of plasma membrane ion transporters have been described for KCC2 (31) and NHE1 (40–42). Further work will be required to determine whether the KCC4-associated changes in intracellular Cl− and K+ concentration and ion transporter–independent function of KCC4 act cooperatively to regulate cell invasion.
In conclusion, we show that IGF-I and EGF stimulate the recruitment of KCC4 from the inactive storage pool in the cytosol to the active membranous target pool at lamellipodia through a mechanism involving PI3K activation and myosin Va-actin trafficking routes. KCC4 is incorporated into lipid rafts throughout the trafficking process. Furthermore, KCC4 functions as a plasma membrane scaffold protein to facilitate the modulation of cytoskeletal reorganization via association with ezrin, which is required for cellular invasive migration (summarized in Supplementary Fig. S6). Blockade of KCC4 recruitment to the plasma membrane may provide a potential target for the prevention of IGF-I– and EGF-dependent invasive metastasis in cancers of epithelial origin.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: National Science Council, Taiwan; Center of Excellence for Clinical Trial and Research (DOH-TD-B-111-004), Department of Health, Executive Yuan, Taiwan; Center for Gene Regulation and Signal Transduction Research and Center for Micro/Nano Science and Technology, National Cheng Kung University, Taiwan; Medical Research Council (MRC grant G0700759), United Kingdom; and NIH grant (DK57708).
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.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received June 25, 2009.
- Revision received August 19, 2009.
- Accepted September 10, 2009.
- ©2009 American Association for Cancer Research.