Differential Role of Human Choline Kinase α and β Enzymes in Lipid Metabolism: Implications in Cancer Onset and Treatment

Background The Kennedy pathway generates phosphocoline and phosphoethanolamine through its two branches. Choline Kinase (ChoK) is the first enzyme of the Kennedy branch of synthesis of phosphocholine, the major component of the plasma membrane. ChoK family of proteins is composed by ChoKα and ChoKβ isoforms, the first one with two different variants of splicing. Recently ChoKα has been implicated in the carcinogenic process, since it is over-expressed in a variety of human cancers. However, no evidence for a role of ChoKβ in carcinogenesis has been reported. Methodology/Principal Findings Here we compare the in vitro and in vivo properties of ChoKα1 and ChoKβ in lipid metabolism, and their potential role in carcinogenesis. Both ChoKα1 and ChoKβ showed choline and ethanolamine kinase activities when assayed in cell extracts, though with different affinity for their substrates. However, they behave differentially when overexpressed in whole cells. Whereas ChoKβ display an ethanolamine kinase role, ChoKα1 present a dual choline/ethanolamine kinase role, suggesting the involvement of each ChoK isoform in distinct biochemical pathways under in vivo conditions. In addition, while overexpression of ChoKα1 is oncogenic when overexpressed in HEK293T or MDCK cells, ChoKβ overexpression is not sufficient to induce in vitro cell transformation nor in vivo tumor growth. Furthermore, a significant upregulation of ChoKα1 mRNA levels in a panel of breast and lung cancer cell lines was found, but no changes in ChoKβ mRNA levels were observed. Finally, MN58b, a previously described potent inhibitor of ChoK with in vivo antitumoral activity, shows more than 20-fold higher efficiency towards ChoKα1 than ChoKβ. Conclusion/Significance This study represents the first evidence of the distinct metabolic role of ChoKα and ChoKβ isoforms, suggesting different physiological roles and implications in human carcinogenesis. These findings constitute a step forward in the design of an antitumoral strategy based on ChoK inhibition.


Introduction
Human choline kinase alpha (ChoKa) and beta (ChoKb) are members of the choline kinase family. In mammals this family is encoded by two separate genes, CHKA and CHKB, resulting in three different proteins with a choline/ethanolamine kinase (ChoK/EtnK) domain: ChoKa1 (NP_001268), ChoKa2 (NP_997634) and ChoKb1 (NP_005189) [1]. ChoKa1 differs from ChoKa2 in only an extra stretch of 18 amino acids, while ChoKb differs from ChoKa1 and ChoKa2 in approximately 40%. The presence of the ChoK/EtnK domain confers the capacity to catalyze the phosphorylation of choline (Cho) to phosphocholine (PCho) [1]. This constitutes the first step in the biosynthesis pathway of phosphatidylcholine (PC) [2]. PC is the major phospholipid in eukaryotic membranes and plays a critical role in membrane structure and also in cell signalling [2]. ChoK enzymes could be implicated also in the synthesis of phosphatidylethanolamine (PE), using as substrate ethanolamine to render phosphoethanolamine (PEtn) [1,[3][4][5].
Previous studies suggest that ChoK acts as a dimeric protein [6,7] and the proportion of the different homo-or hetero-dimer population has been proposed to be tissue-specific [8]. Furthermore, the combination between choline kinase isoforms results in a different level of ChoK activity in vitro under cell-free systems conditions. Thus, the a/a homodimer is the most active choline kinase form, the b/b homodimer the less active, and the a/b heterodimer has an intermediate phenotype [8].
The specific phospholipase D-driven hydrolysis of PC generates choline and signal transduction metabolites such as PA (phosphatidic acid), and its derivatives LPA (lysophosphatidic acid) and DAG (diacylglycerol), that are important in mitogenesis and cellular transformation [9][10][11]. Choline gets further converted by ChoK into PCho, an stable metabolite able to induce mitogenesis in murine fibroblasts [12]. In addition, several oncogenes such as RAS or RHOA increase ChoKa activity resulting in higher intracellular levels of PCho [13][14][15][16]. Magnetic resonance spectroscopy (MRS) studies have revealed abnormal phospholipid metabolism in cancer cells [17,18]. Furthermore, high levels of PCho have been found in tumoral cells as well as in tumour samples of cancer patients compared with the normal counterparts in breast, prostate, brain and ovarian cancer [19][20][21][22][23][24][25]. The increase in phosphoethanolamine has been also reported in transformed cells or tumor samples using MRS, although it contributes to a much lesser extent, compared with the increase of PCho, to the phosphomonoester peak [26].
The implication of ChoKa in cell growth, proliferation, initiation and progression of cancer is well documented. ChoKa is overexpressed in some of the most common cancers, such as breast, lung, colorectal, prostate [27][28][29][30] and bladder (unpublished observations). Furthermore, it has oncogenic activity when overexpressed in human cells [15]. Increased ChoKa mRNA levels have been recently described as an independent prognostic marker in non small cell lung cancer patients [30]. Also, human breast cancer cell lines show upregulation of ChoKa but not ChoKb, when compared with normal mammary epithelial cells [21]. In that sense, it has been recently reported that in the prostate tumor mouse model (TRAMP), using IHQ immunodetection of ChoKb, a low level expression of ChoKb was found in tumoral samples compared with wild type tissues [31].
The implication of members of the Rho GTPase family in cancer onset and progression has been extensively described [32][33][34][35]. Evidence that ChoK is involved in malignant transformation along with Ras and RhoA has been provided [14,15,36]. Moreover, the activity of ChoKa is modulated by known effectors of both Ras and RhoA [15,29].
Pharmacological inhibition of ChoKa has been proposed as a novel antitumoral strategy [2]. A strong support to this strategy is based on the use of MN58b, a well-characterized ChoKa inhibitor, which displays a potent antiproliferative effect in several tumoral cell lines in vitro, and a strong reduction of tumor growth in nude mice xenografts [14,[37][38][39][40].
We have compared the biochemical and biological properties of ChoKa and ChoKb, and demonstrate that besides their homology, the latter is not able to induce cell transformation in HEK293T or MDCK cells. Moreover, we suggest differential behaviour between a and b isoforms in phospholipids metabolism. Finally, the antitumoral properties of MN58b can be attributed to its specific inhibition of ChoKa. The implications of these findings are discussed.

Characterization of enzymatic properties of ChoKa1 and ChoKb isoforms
The activity of both choline kinase isoforms, ChoKa1 and ChoKb, has been described previously in different mammalian tissues, however their choline kinase and/or ethanolamine kinase activities are not fully characterized [1,4,[41][42][43]. Thus, we first carried out a comparative analysis of the in vitro kinase activity of the two human ChoKa1 and ChoKb isoforms, regarding their ability to phosphorylate choline and ethanolamine. HEK293T cells were transfected with appropriate expression vectors carrying the human CHKA or CHKB genes or an empty vector, and tested for ChoK and EtnK activity. Expression levels achieved were checked by Western Blot (Figure 1a), and the relative enzymatic activity from cell extracts estimated. Both ChoKa1 and ChoKb displayed choline and ethanolamine kinase activities under these experimental conditions, but the rate of conversion was apparently different (Figure 1b).
We further analyze ChoK and EtnK kinetic activities for ChoKa1 and ChoKb isoforms using the recombinant enzymes. DH5a Escherichia coli were transformed with the gene encoding for human ChoKa1 or ChoKb and an enzymatic in vitro assay for either ChoK or EtnK activities performed. Michaelis constants (Km) for each isoform for the different substrates were obtained (Table 1). ChoKb showed a Km for choline 2.8 times higher than ChoKa1. By contrast, the Km of ChoKb for ethanolamine was lower than ChoKa1. These results suggest that in cell-free systems choline is a better substrate for ChoKa1 (2.85 fold) than ChoKb, whereas ethanolamine is a better substrate for ChoKb (5.83 fold) than ChoKa1.

ChoK isoforms are differentially regulated by Ras and Rho GTPases
Since Ras and RhoA GTPases are upstream regulators of ChoKa1 [15,28], some of the most studied proteins of the small GTPase family including H-Ras, and the Rho-family members RhoA and Cdc42 were tested as potential upstream modulators of ChoKb. To that end, an in vitro ChoK and EtnK activity assay was performed in HEK293T cells transfected with the constitutive active mutants of each GTPase (Fig. 2a) and either ChoKa1 and ChoKb. As shown in Figure 2, none of the GTPases tested were able to significantly increase choline or ethanolamine kinase activity of ChoKb. By contrast, under similar conditions, Ras and RhoA were able to induce a statistically significant change in the activation of both ChoK and EtnK activities of ChoKa1 ( Fig. 2b  and 2c).

Differential role between ChoK isoforms in cell transformation and tumorigenesis
The implication of ChoKa1 in cell transformation and human carcinogenesis has been extensively studied, and it has been shown to display oncogenic activity [15,36]. Due to its extensive homology, we investigated if ChoKb could induce cellular transformation when overexpressed in non-tumorigenic cells. First, the ability of ChoKb-transfected cells to promote anchorage independent cell growth was determined using the cell line HEK293T. As previously reported, ChoKa1 induced a significant increase in the number of soft agar colonies [15]. By contrast, under similar conditions, ChoKb overexpression had no significant effect on the number nor the size of the colonies when compared to those generated by control, empty vector transfected HEK293T cells (Fig. 3). These results were confirmed using another non-tumorigenic cell line from a different species such as MDCK, obtaining similar results despite of the lower level of overexpression of both ChoK isoforms compared to the HEK293T cells. The differential level of expression obtained was due to the rather different efficiency in transfection for each cell lines (data not shown).
The ability of ChoKb to induce tumour growth was also investigated. Both ChoKa1-and ChoKb-transfected HEK293T or MDCK cells were subcutaneously inoculated into athymic mice. As shown in Fig. 4, the overexpression of ChoKa1 was sufficient to induce tumour growth in immunosuppressed mice in the both cell systems analyzed. By contrast, cells overexpressing ChoKb were not able to induce tumor growth under similar conditions. As indicated above, comparison in tumor growth rates between both cell lines is related to the diverse efficiency of transfection, much higher in HEK293T (aproxm. 80-90%) than MDCK (aproxm. 15-20%). These results are in keeping with those obtained in the soft-agar experiments, and strongly indicate that overexpression of ChoKb is not sufficient to induce tumor growth under conditions where ChoKa1 does.

ChoKb isoform shows preferential EtnK activity in vivo
We next investigated if there was any difference in the enzymatic activity of both ChoK isoforms under in vivo conditions that could explain their differential biological activity. Thus, the activity of ChoKa1 and ChoKb as ChoK and/or EtnK in a whole-cell system was tested. Intracellular lipids of human HEK293T overexpressing ChoKa1, ChoKb or transfected with an empty vector, were extracted an analyzed. Both, the insoluble lipid fraction containing the hydrophobic lipids and total protein content were used as loading control obtaining similar results. As shown in Figure 5A, the intracellular phosphoethanolamine levels were increased to a similar extent in ChoKa12 or ChoKbtransfected cells. However, the intracellular phosphocholine levels of ChoKb-transfected cells were not significantly different to that of control cells, while ChoKa1-transfected cells showed an increased intracellular PCho levels ( Figure 5B). Similar results were obtained using epithelial cells of different origins: human breast adenocarcinoma cells SK-Br-3 ( Fig. 5C, D) or the human Non Small Cell Lung Cancer (NSCLC) cell line H1299 (Fig. 5E, F). In addition, in all cell lines analyzed, protein expression was also determined by Western Blot analysis (Fig. 5G).
Thus, the differential enzymatic activity of ChoKa1 and ChoKb found in HEK293T is not cell line specific. These results indicate that ChoKa1 but not ChoKb, is able to induce increased intracellular levels of PCho under in vivo conditions. However, under the same conditions both enzymes are able to generate phophorylethanolamine. Thus, even though ChoKb displays both ChoK and EtnK activity under in vitro conditions, it shows preferential EtnK activity in vivo.

Increased levels of ChoKa mRNA but not ChoKb in tumor-derived cell lines
In order to further investigate the relevance of each ChoK isoenzyme in tumorogenesis, we determined the levels of both ChoKa and ChoKb mRNA in a panel of human tumour-derived cell lines by quantitative PCR technology. A panel of breast and lung cancer cell lines were compared with the primary, senescent, non-tumoral cell line HMEC (Human Mammary Epithelial Cells) or the immortalized, non-tumorogenic MCF10A cells as control breast cell lines, and primary Bronchial Epithelial Cells (BEC) as control lung cell line. All tumoral cell lines tested significantly overexpress ChoKa mRNA compared with the normal cell lines, whereas no changes were found for ChoKb mRNA levels (Fig. 6). These results indicate that ChoKb is not specifically upregulated in breast or lung cancer cells, suggesting that high levels of ChoKb are not required for the promotion of cancer.

MN58b is a specific inhibitor of ChoKa isoform
The implication of ChoKa in human carcinogenesis has been used for the design of specific inhibitors of this enzyme as a novel anticancer strategy [2,14,37]. MN58b is the leading compound to support this novel strategy since it has shown a potent antitumoral activity under in vivo conditions versus human colon, lung, breast and bladder tumor models [15,37]. The primary structure of mammalian ChoKb displays an overall 60% homology with that of ChoKa1 [1]. The higher degree of homology lies within the choline/ethanolamine kinase domain. This homology could make it susceptible to a similar inhibition for both ChoKa1 and ChoKb with the same inhibitors. Thus, we have investigated the specificity of MN58b towards these two ChoK isoenzymes in order to verify if its antitumoral effect is specifically related to the inhibition of the alpha isoform, that is the one implicated in tumor progression, or the drug similarly affects both isoforms. Using the human ChoKa1 and ChoKb recombinant proteins, we performed an in vitro ChoK activity inhibition assay using increasing MN58b concentrations, determining the IC 50 for each enzyme. As expected, despite the high similarity displayed between ChoK isoforms, MN58b showed a much higher specificity against ChoKa1 (IC 50 = 5 mM) than against ChoKb (IC 50 = 107.5 mM). Thus, MN58b is 21.5 times more potent against ChoKa1 than ChoKb isoform.

Discussion
The family of human choline/ethanolamine kinases comprises two genes, CHKA and CHKB that codify for three enzymes, ChoKa1 (52 kDa), ChoKa2 (50 kDa) and ChoKb1 (45 kDa). ChoKa1 and ChoKa2 are almost identical, except for a stretch of 18 extra amino acids in ChoKa1, as they result from differential splicing from the same gene, CHKA. While the implication of ChoKa1 in the regulation of cell growth and cancer has been extensively demonstrated [13,15,27,28,37,38,44], preliminary evidence suggest that ChoKb may not be involved in carcinogen-esis since it is not overexpressed in breast cancer cell lines [21] nor in the TRAMP mouse prostate cancer model [31].
A distinct human gene family has been described that codifies for EtnK activity [45], suggesting that EtnK1 is the main enzyme involved in PE homeostasis. The ChoK/EtnK domain confers to ChoKa and ChoKb the ability to function as both ChoK and EtnK activity under cell-free conditions [1,[3][4][5], but it is still unknown if these previously characterized enzymes showed any selectivity to each branch of the Kennedy pathway in the de novo synthesis of PC or PE. It has been recently described the different specificity towards Cho and Etn of the two isoforms of Cho/EtnK of Tripanosoma brucei. Whereas TbCho/Etn1 displays only EtnK activity, TbCho/Etn2 displays both ChoK and EtnK activities in vitro [46]. These results are in keeping with those described for murine EtnK1 that is Etn specific and EtnK2 that displays a dual ChoK/EtnK function [45,47]. On the other hand, whereas murine Pcyt1a and Pcyt1b are involved in PC biosynthesis, Pcyt2 is focused to PE only [48].
However, it is not yet fully understood which ChoK isoform, if any, contributes in vivo in each pathway to maintain the normal homeostasis of both PC and PE in biological membranes. The results shown here confirm that both enzymes have the ability to phosphorylate choline and ethanolamine under cell-free conditions, either as recombinant proteins produced in E. coli, or in cell extracts from mammalian cells. However, we found that under whole cell conditions ChoKa1 has the ability to function as both ChoK and EtnK, but ChoKb only affects the production of PEtn. These findings of different roles for a and b isoforms are in keeping with the information from the recently generated Knock Out (KO) mice for ChoKa and ChoKb genes [49,50]. Thus, ChoKb KO mice (rmd mice) are viable, but develop a rostrocaudal muscular dystrophy, while normal PC lipid levels are found in most tissues analyzed except in hindlimb skeletal muscle [49]. Therefore, ChoKa is sufficient to maintain normal PC levels in most tissues. By contrast, the lack of ChoKa results in embryonic lethality, and ChoKa +/2 heterozygous mice display an accumulation of Cho and a reduction in PCho in liver and testis, suggesting that there is no ChoKb compensation for PC biosynthesis in vivo. These results suggest different roles in vivo for both ChoKa and ChoKb isoforms. Furthermore, the attenuated levels in PE found in ChoKa +/2 heterozygous mice suggest the involvement of ChoKa not only in the biosynthesis of PC but also in the PE pathway. This is also consistent with the fact that in ChoKb KO mice, PE levels are unaffected, indicating that PE homeostasis is fully maintained with the EtnK1 and ChoKa proteins intact.
The group of Ishidate has provided valuable information about the in vitro activity of ChoK from different mouse tissues, and they have postulated that the most active form for choline kinase activity is the a/a homodimer followed by a/b heterodimers, being the b/b homodimer the least active form [8]. The in vivo results shown here are partially in keeping with this hypothesis. The in vivo activity of ChoKb is focused in PE biosynthesis and displays higher Km for choline than ChoKa, this could be the reason why b/b dimers show low ChoK activity. When ChoKb was overexpressed we observed an increase in the intracellular levels of PEtn but not of PCho. However no differences were found between both isoforms for the generation of PEtn, since both display similar EtnK activity.
PCho has been proposed to promote mitogenesis in mammalian cells [12]. In keeping with this, magnetic resonance spectroscopy techniques have revealed higher levels of phosphomonoesthers in tumoral samples when compared to their normal counterparts [19][20][21][22][23][24]. Moreover, overexpression of ChoKa1 is oncogenic [15], and enhanced ChoKa activity is a frequent feature in tumoral samples compared to normal tissues [27,28]. Taken together all these results strongly indicate that ChoKa1 activity and PCho levels have a strong implication in cancer. Furthermore, overexpression of ChoKa1 results in an increase in EtnK activity and PEtn levels. However the latter effect by itself is not sufficient to induce cell transformation, since overexpression of ChoKb does not induce enhanced colony formation in soft-agar or tumor growth in nude mice. These results are consistent with the hypothesis that it is the production of PCho what is linked to cell proliferation and transformation mediated by ChoK, and that the production of PEtn may not be sufficient or relevant in this process.
The above results suggest that the two ChoK isoforms investigated, besides their similarity in their primary sequences, are implicated in different metabolic pathways. Thus while ChoKa1 impinges into both PC and PE synthesis, ChoKb affects only PE synthesis. Furthermore, the transformation capacity seems to be exclusive to the ChoKa isoform. However, since in the human HEK293T, Sk-Br-3 and H1299 cell lines, ChoKa1 overexpression produces elevated levels of both PCho and PEtn, while similar ChoKb overexpression results only in higher levels of PEtn but normal levels of PCho, we can not rule out the possibility that cell transformation requires both ChoK and EtnK activities.
Consistent with the idea that links oncogenic activity to the function of ChoKa, but not ChoKb, the antiproliferative and antitumoral activity of MN58b was only associated to the activity of ChoKa. Furthermore, as previously reported, the in vivo treatment with MN58b results in a specific decrease of PCho levels in the tumours but no significant effect on the levels of PEtn [51]. Previous results from our group have demonstrated that MN58b also inhibits choline transport [38]. However this effect has a little influence in the antiproliferative and antitumoral activity of the drug since HC-3, a much more potent inhibitor of choline transporters, is far less potent as an antiproliferative agent than MN58b [2]. Furthermore, MN58b has a differential effect on either normal or tumor cells, a strong demonstration of a differential activity due to ChoK inhibition [38][39][40]. These results are also in keeping with the observation that ChoKa but not ChoKb is a downstream target of oncogenic molecules such as Ras and RhoA. Thus, while Ras activates ChoKa through Ral-GDS and PI3K [29], and RhoA activates ChoKa through ROCK [15], none of these oncogenic GTPases affect ChoKb activity under similar conditions. Again, these results indicate that although both ChoK enzymes are able to phosphorylate both choline and ethanolamine under cell-free systems conditions, they display different affinities for these substrates, and in whole-cell assays conditions they are governed by distinct regulatory pathways.
Finally, the involvement of ChoKb in breast and lung cancer has been studied, ChoKa and b mRNA levels were determined by Q-PCR. All tumour-derived cell lines assayed significantly overexpress ChoKa mRNA, while no changes were found in the expression of ChoKb. Similar results have been recently reported, For the comparison between tumoral and non-tumoral cell lines, the 2 2DDCt method was applied and log 10 RQ is represented. Note that the data are referred to the Human Mammary Epithelial Cells (HMEC) mRNA levels in breast cell lines and no significant difference in the level of both ChoK isoforms mRNA was found with the normal MCF10A cell line. The reference for lung cancer cells was the primary Bronchial Epithelial Cells (BEC). doi:10.1371/journal.pone.0007819.g006 using semi-quantitative PCR in breast cancer cell lines [21]. In addition, it has been recently described that elevated mRNA levels of ChoKa is a poor prognostic factor in lung cancer [30]. These results suggest that ChoKa but not ChoKb is playing a crucial role in human carcinogenesis.
The results shown here suggest that ChoKb and its produced metabolites are not implicated in human cell transformation. Therefore, all the efforts aimed at dilucidating the involvement of ChoK activity in the diagnosis, prognosis and treatment of cancer have to be focused in the ChoKa isoform. In keeping with this, recently the use of specific monoclonal anti-ChoKa antibodies has been proposed as a diagnostic tool in human cancer [52]. Furthermore, the newly designed antitumoral agents are expected to be more specific and hence less toxic than the actual drugs used in conventional chemotherapy. Due to the high structural homology displayed by both choline kinase proteins, the search for new anticancer agents based on their ability to interfere with ChoK activity, must exhibit stronger antiproliferative activity based on their specificity towards the ChoKa isoform.
The lack of specific inhibition of ChoKb by these newly designed compounds represents a new feature to take into account for the chemical improvement of ChoKa inhibitors with potential antitumoral activity. Furthermore, non-specific drugs affecting ChoKb may result in a muscular disease produced by the lack of cell membrane lipid reparation in muscle tissue.
In addition, it has been recently demonstrated using a genetic approach that the specific inhibition of ChoKa by shRNA displays antiproliferative and antitumoral activity The high specificity of this technology provides definitive evidence of an antitumoral strategy based on ChoKa inhibition, supporting previous results with the pharmacological inhibitors. [53][54][55] Thus, despite the high homology and similar activity displayed under cell-free conditions, ChoKa1 and ChoKb isoforms show a different substrate ability and behave very differently under in vivo conditions, suggesting that in human cells, ChoKb behaves as an EtnK and its overexpression is not able to induce higher intracellular levels of PCho. As a consequence, ChoKb has no effect on cell proliferation and does not contribute to oncogenic transformation. Finally, ChoKa but not ChoKb, should be used as the molecular target for the design of anticancer drugs aiming at interfering with choline kinase acitivity.
For Michaelis-Menten kinetics assay, extracts of E. coli expressing human ChoKa1 or ChoKb1 were used. A time course reaction was performed for each substrate, the reaction was stopped each 2 minutes and a logarithmic scale was represented to obtain the Km. Specific ChoK inhibitor MN58b was kindly provided by TCD Pharma SL (Madrid, Spain).

Analysis of phosphocholine and phosphoethanolamine production in whole cells
HEK293T, SK-Br-3 and H1299 cells were transfected as described above. After transfection, cells were labelled to equilibrium adding 1 mCi/ml [methyl-14 C] choline chloride or 1 mCi/ml 2-[ 14 C]-ethan-1-ol-2-amine hydrochloride to the culture media for 24 h. Samples were resolved and quantified as previously described [37]. Briefly, cells were rinsed with PBS and fixed with 16% ice-cold trichloroacetic acid (TCA). TCAsoluble material containing choline and phosphocholine was washed three times with four volumes of diethyleter, dryed under vacuum and resuspended in water. Samples were resolved by Thin Layer Chromatography (TLC) on 60Å Silica gel plates (Whatman, NJ), using as liquid phase 0,9% NaCl: methanol: ammonium hydroxide (50:70:5; v/v/v). Radioactivity was automatically quantified by an electronic radiography system (Instantimager, CT). TCA insoluble material containing hydrophobic lipids was used as a loading control. The insoluble fraction was dissolved in 0,25N sodium hydroxide and total lipids were analyzed by scintillation counting.

Analysis of protein expression by Western blot
Equal amounts of cell lysates were resolved by electrophoresis in a 10% SDS-PAGE. ChoKa expression levels were detected using monoclonal antibody AD3 as previously described [52]. ChoKb1 was detected using the polyclonal antiserum described above as primary antibody at a standard dilution of 1/5000. GTPases antibodies: RhoA (Santa Cruz Biotechnology, Germany), Rac1 (Upstate, NY), pan-Ras (Santa Cruz Biotechnology), Cdc42 (BD Biosciences, transduction labs) and GAPDH antibody (Chemicon International, CA) were used following manufacturers recommendations. HRP-conjugated secondary antibodies were from Santa Cruz Biotechnology. The specific binding was detected using the chemioluminiscence detection kit ECL (Amersham Bioscience) following the manufacturer indications.
Soft agar anchorage-independent growth and in vivo tumorigenic assays Anchorage-independent growth assays were performed at 5-6 weeks of incubation in HEK293T and 6-10 weeks in MDCK, as previously described [15], plating 10 5 cells per 60-mm dish. Crystal violet staining was performed to enhance colony contrast. The number of colonies was automatically quantified by computer based image software (Image J). For the in vivo assays, cells were injected (10 6 cells per mouse flank) s.c. in nu/nu athymic mice. Mice were kept under standard laboratory conditions according to the guidelines of Spanish Government.

Quantitative PCR
ChoKa1 or b1 mRNA levels were quantified by real-time reverse transcriptase PCR. The RNA was extracted from the cell lines using QIAshredder following the RNeasy Mini kit (Qiagen, Inc.) according to the manufacturer's instructions. The amount of 0.9 mg of total RNA, in a final concentration of 10 ng/ml per reaction, was retro transcribed by High-Capacity cDNA Archive Kit (Applied Biosystems) to the cDNA preparation. The reverse transcription conditions were 25uC for 10 min and 37uC for 2 h. Then, each cDNA sample was analyzed in triplicate using the ABI PRISM 7700 Sequence Detector (PE Applied Biosystems). Realtime PCR was carried out using Taqman Universal PCR Master Mix (Applied Biosystems), containing ROX to normalize emis-sions. Primers used for amplification of ChoKa, ChoKb and 18S ribosomal RNA were purchased from Applied Biosystems as Taqman Gene Expression Assays (ChoKa1 ID: Hs00608045_m1, ChoKb1 ID: Hs01925200_s1 and 18S ribosomal RNA ID: Hs99999901_s1). For thermal cycling, the following conditions were applied: 10 min at 95uC and 40 cycles of 15 sec at 95uC and 1 min at 60uC. The data are presented as Log 10 RQ (Relative Quantity).

Statistical analysis
Mean comparisons of ChoK activity or in-cell PCho increase between different groups were performed. Continuous variables with normal distribution were compared by T-test and nonnormal distribution variables were compared by means of the Kruskal Wallis and Mann-Whitney U tests. Statistical significance was defined as p#0.05. The statistical analyses were performed using SPSS software, version 13.0 (Inc, Chicago, Illinois). The method used to analyze data from real-time PCR experiments was 2 2DDCt method [58], comparing the relative gene expression between normal and tumoral cell lines normalized to 18S ribosomal RNA as endogenous reference gene.