Anal Bioanal Chem (2005) 383: 92–97 DOI 10.1007/s00216-005-3375-1
O R I GI N A L P A P E R
Sandro Hillebrand Æ Wanius Garcia Marcelo Delmar Cantu´ Æ Ana Paula Ulian de Arau´jo Manami Tanaka Æ Tomoo Tanaka Richard Charles Garratt Æ Emanuel Carrilho
In vitro monitoring of GTPase activity and enzyme kinetics studies using capillary electrophoresis Received: 5 January 2005 / Revised: 21 May 2005 / Accepted: 1 June 2005 / Published online: 23 July 2005 Springer-Verlag 2005
Abstract A capillary electrophoresis (CE)-based method for the in vitro detection and monitoring of nucleotidetriphosphatase activity is described. This robust and reproducible method was used to investigate GTPase activity of a recombinant protein construct containing the catalytic domain of Human SEPT4/Bradeion b (GST-rDGTPase). This example application demonstrates that the CE technique can replace classical radioactive methods for GTPase activity assays and may be used as a routine analytical tool. Enzyme kinetics of GST-rDGTPase was studied and yielded the following kinetic parameters: vmax = 1.7 lM min1 ± 0.1, Km = 1.0 mM ± 0.3, and apKcat = 9 · 103 s1. In addition the effect of co-factors such as Mg2+ and Mn2+ on the catalytic activity was investigated. The described analytical method was also shown to be useful to analyze diphosphated and triphosphated forms of other nucleotides. Keywords SEPT4 Bradeion b Æ Septins Æ Guanosine triphosphate Æ Influence of magnesium (Mg2+) Æ Recombinant GTPase
S. Hillebrand Æ W. Garcia Æ A. P. U. de Arau´jo Æ R. C. Garratt Centro de Biotecnologia Molecular e Estrutural, Instituto de Fı´ sica de Sa˜o Carlos, Universidade de Sa˜o Paulo, Sa˜o Carlos, SP, Brazil S. Hillebrand Æ M. Delmar Cantu´ Æ E. Carrilho (&) Laborato´rio de Cromatografia, Instituto de Quı´ mica de Sa˜o Carlos, Universidade de Sa˜o Paulo, Av. Trabalhador Sa˜ocarlense, 400 Caixa Postal 780, CEP 13566-590 Sa˜o Carlos, SP, Brazil E-mail:
[email protected] Tel.: +55-16-33739965 Fax: +55-16-33739983 M. Tanaka National Institute of Advanced Industrial Science and Technology, Higashi, Tsukuba Science City, Ibaraki, Japan T. Tanaka Tokai University School of Medicine, Isehara, Kanagawa, Japan
Introduction Septins are highly conserved cytoskeletal proteins present in animal and fungal cells but absent in plants. About 12 distinct mammalian septins have been identified and a recent report suggests the utilization of a general nomenclature [1]. The Bradeion gene, classified in the SEPT4 category [1], was first identified and characterized by Tanaka et al. [2, 3] and is strongly expressed in the brain, weakly expressed in heart tissue, and is adult stage-specific. This cell-type specificity is also true for its ectopic expression in human cancer, being detectable only in colorectal cancer and malignant melanoma. This unique pathology-related expression profile has never been observed for any other member of the mammalian septin family, and such characteristics satisfy the basic criteria for a target molecule for the monitoring of cancer development. The role of the GTPase activity of SEPT4 (Bradeion) in the cytokinesis of cancer cells is still poorly understood. Therefore, efforts to characterize both the structural and biochemical properties of this class of molecules are highly desirable. The GTPase activity is characterized by the hydrolysis of guanosine-5¢-triphosphate (GTP) producing guanosine-5¢-diphosphate (GDP) and inorganic phosphate. The assay commonly used to characterize GTPase activity utilizes radioisotopically [32P]-labeled GTP associated with a separation technique such as thin-layer chromatography [4, 5]. In spite of being widely used, this method is quite laborious, time consuming, and requires special care for handling radioactive material. Alternative techniques have therefore been explored to avoid the disadvantages of radioactive methods. For instance, an HPLC-based method has been developed and applied to study GTPase activity [6, 7]. Capillary electrophoresis (CE) has been successfully used to separate nucleotides and their monophosphate, diphosphate, and triphosphate derivatives [8–11] with all
93
the advantages attributable to this technique, such as low sample and reagent consumption, reduced analysis time, ease of automation, and high separation power. Kawata et al. [12] reported a method using CE to separate and quantify GTP and GDP after enzymatic reaction. By means of this approach, separation of GTP and GDP, performed in fused silica capillaries, was achieved within 7.2 min. Based on this work, we have optimized the method for both fused silica and coated capillaries to achieve better resolution and fast separations to simplify the whole analytical procedure for enzyme kinetic studies. Our method was applied to characterize the GTPase activity of a recombinant protein construct containing the catalytic domain of Human SEPT4/Bradeion b (rDGTPase) in fusion with glutathione-S-transferase (GST-rDGTPase).
(58,122.8 Da). Protein concentrations were determined by measuring the sample absorbance at 280 nm and relating it to the extinction coefficient (62,900 M1 cm1) obtained by theoretical calculation [13]. GTPase activity assay
All reagents were of analytical grade and were purchased from Sigma (St. Louis, MO, USA). Water used for the buffer preparation was obtained from a Milli-Q water system (Millipore, Bedford, MA, USA). All electrolytes and rinsing solutions used during the CE experiments were filtered through 0.45-lm nitrocellulose filters before use.
The stock solutions of MgCl2 (100 mM), Mn2+ (10 mM), and GTP (10 mM) were prepared by dissolving the compounds in a 50 mM Tris–HCl buffer, pH 8.0, 100 mM. Activity assays for the GST-rDGTPase were performed in 50 mM Tris–HCl buffer, pH 8.0, 100 mM NaCl, in the presence of a given amount of MgCl2 (ranging from 0.5 to 10 mM) or Mn2+ (0.5 mM), and GTP (0.5–4 mM). The typical protein concentration in the reaction mixture was 0.18 mg mL1 (OD = 0.2). The reaction was started by the addition of GTP and subsequently incubated at 28C. Reaction monitoring was performed by collecting 25-lL aliquots from the reaction tube at different time intervals. The aliquots were immediately frozen in liquid nitrogen to stop the enzyme activity. The freezing procedure resulted in irreversible inhibition, as evaluated by following the reaction products for several hours after thawing and observing no alteration to the GDP/GTP ratio. Prior to CE, the samples were equilibrated at room temperature, vortex-mixed, and centrifuged to avoid the injection of any precipitate into the capillary.
Protein expression and purification
Instrumentation and CE procedures
For the expression of the recombinant protein, 5 mL of an overnight culture of E. coli AD202 pGST-DGTPase were inoculated into 500 mL of LB medium containing ampicilin (100 lg mL1). Protein expression was induced at an optical density of 0.7 at 600 nm by addition 0.2 mM isopropyl b-D-thiogalactopyranoside (IPTG). After 16 h incubation at 20C, cells were collected by centrifugation and kept frozen at 80C. The cells were subsequently lyzed by sonication at 4C in 50 mM Tris– HCl pH 8.0 buffer containing 100 mM NaCl and 1 mM phenylmethylsulfonyl fluoride (PMSF). After lysis, the mixture was centrifuged for 20 min at 20.000 g at 4C. The supernatant was mixed with 2.0 mL of sepharoseglutathione 4B fast-flow resin equilibrated in lysis buffer for 30 min at 4C. After incubation, the resin was washed three times with 50 mM Tris–HCl pH 8.0 buffer containing 100 mM NaCl. The fusion protein was eluted from the resin by the addition of 50 mM Tris–HCl pH 8.0 buffer containing 100 mM NaCl and 10 mM reduced glutathione. The resulting GST-rDGTPase was applied to a superdex 200 column for elimination of reduced glutathione. The protein purity was confirmed by gel electrophoresis (SDS-PAGE 15%) in which a single band was observed corresponding to a molecular weigh that agrees with the predicted value for the GST-rDGTPase
CE analyses were performed on an HP 3DCE instrument (Agilent Technologies, Waldbronn, Germany) equipped with a diode-array detector. Analytes were monitored at 254 nm with a bandwidth of 30 nm. The separation conditions used were a modification of those described previously [12]. In the original conditions the separations were performed in fused silica capillaries using 10 mM MES buffer pH 6.5 as the running background electrolyte. This condition was optimized to 50 mM pH 6.5 MES buffer as the electrolyte. In addition, methods were developed for both bare fused silica and PVA-coated capillaries. In the former, electrophoresis was performed in the normal mode (anode at the inlet) using a 50-cm-long (41.7 cm to the detector) fused silica capillary with 50-lm i.d. A 28-kV voltage was applied to achieve separations in about 6 min. For PVAcoated columns, negative polarity was applied (cathode at the inlet) using a 32.5-cm-long capillary (24.2 cm to the detector) with 50-lm i.d. and analytes separated in less than 2 min under a voltage of 25 kV. In both methods, the samples were introduced hydrodynamically using 50 mbar for 5 s (ca. 9 nL). The fused silica capillary was washed with 0.1 M NaOH and deionized water for 10 min each at the beginning of the day. Before injection, a pre-conditioning step was performed by means of a 0.5-min flush with
Experimental Chemicals and solutions
94
0.1 M NaOH followed by water for 1 min, and running buffer for 1.5 min. When using PVA-coated capillaries, these were flushed with water for 10 min at the beginning of the analysis sequence and were pre-conditioned with water for 1 min and running buffer for 2 min prior to each injection. The GTP and GDP concentrations were determined from the relative peak areas. The total area from the GTP and GDP peaks was considered to be 100% of the nucleotide content at any given nominal concentration. The area of the GDP peak relative to the total area was assigned as the proportion of GDP formed (hydrolyzed GTP). This approach prevents possible deviations due to differences in the injected sample volume. When needed, the background GDP concentration present in the GTP standard was determined in control experiments for correction of the value measured in the analysis.
Determination of kinetic parameters In order to monitor the reaction rate and calculate the initial velocity, the concentration of GDP produced was plotted against the reaction time. The initial reaction velocity, v0, was determined from the slope of a straight line adjusted to the linear portion of the experimental data. The kinetic parameters vmax and km were obtained from a Michaelis–Menten plot, v0 versus [GTP], by fitting a rectangular hyperbole (R2 > 0.985) to the experimental points (Eq. 1). Both linear and non-linear fits were performed using the software Microcal Origin 6.0 (Microcal Software, Northampton, MA, USA). v0 ¼ vmax
½S km þ ½S
ð1Þ
Results and discussion Methodology A simple and robust method for GTP and GDP analysis by CE is reported. The method was optimized for enzyme kinetic studies achieving baseline resolution separations in times as short as 1.7 min (Fig. 1). The RSD for migration time and peak area were determined for a set of five replicates. The RSD values were lower than 2.5% for migration time and lower than 3.9% for peak area. This performance was achieved for a 32.5-cm-long PVA-coated capillary (24.2 cm to the detector) under 770 V cm1. This method based on the CE technique could readily replace the classical methodology used to probe GTPase activity (radioisotope labeling combined with thin layer chromatography) as it presents several advantages. In addition, we exemplify the application of this approach in the characterization of the enzymatic activity of a protein construct containing the catalytic
Fig. 1 Enzyme-catalyzed GTP hydrolysis monitored by means of capillary electrophoresis. GST-rDGTPase (0.25 mg mL1) was incubated with 500 lM GTP and 1 mM MgCl2. Aliquots where collected after different time intervals and analyzed as described in the ‘‘Experimental’’ using a PVA-coated capillary
domain of the recombinant Human Septin SEPT4 (Bradeion b). In order to accelerate the whole analytical process in dedicated procedures for kinetic characterizations, some simplifications were made without any significant loss in precision. This simplified approach does not require the construction of calibration curves or the use of internal standards to quantify the GDP produced by enzymatic hydrolysis. The percentage of converted GTP was calculated by normalizing the peak area of the remaining GTP by the total area (GDP + GTP). Therefore, a common error arising from differences in the sample volume is avoided. Hydrolysis can then be expressed either as a percentage of metabolized substrate or as a product concentration by relating that percentage to the initial substrate concentration. This simplification is based on the assumption that GDP is the only possible reaction product and that the amount of GTP or GDP bound to the protein is negligible. We have compared this approach with the classical internal standard method (data not shown) and found both methods to yield the same precision in determining the percentage of converted GTP. However, this procedure may not be suitable for absolute GTP quantification for which the use of the internal standard method is advised. In the range of 0.5–4 mM initial GTP concentration, we were able to detect GDP formation of the order of 105 M, which was quite satisfactory for the purposes of enzyme kinetics. Within the range studied in this work (substrate from 0.5 to 4 mM and enzyme equal to 0.184 mg mL1) the lowest detectable activity was observed after 30 min. In this time interval, the concentration of GDP varied from 7.36 · 103 mM to 2.15 · 102 mM. It means that the protein activity (the enzyme amount required to convert 1 lmol of GTP in 1 min) detectable under the experimental conditions was 6.6 lM.
95
The GTP used in this work was analytical grade and no further purification was carried out prior to the assays. All commercial GTP tested presented a small percentage of GDP, likely the product of spontaneous hydrolysis. Therefore, when running a Michaelis–Menten sequence with varying initial substrate concentrations, the GDP concentration at t0 was neither zero nor constant for the different samples (Fig. 2a). The average background GDP concentration in commercial GTP standards was ca. 5%. We assume that in this range it does not affect the reaction kinetics given that the GDP concentration is much smaller than the initial GTP concentration. When determining the absolute amount of GDP produced by a GTPase, the background GDP
concentration should be carefully measured in a control experiment and then subtracted from the reaction analysis. However, in a Michaelis–Menten experiment, the initial reaction rate is the relevant measurement and it is determined from the slope of the linear region of the reaction-monitoring graph (Fig. 2a). For this application there is no need to measure the background GDP, since the starting point and the slope are independent parameters. In spite of the fast separation achieved with the method described here, construction of a Michaelis– Menten curve is a quite laborious task. A five-point Michaelis–Menten plot required the same number of reaction monitoring graphs with at least six points each (30 runs). In addition, every reaction required a control mixture (without enzyme) from which three aliquots were collected (15 runs). Summing up, this gives a total of 45 analyses. Since each analysis takes about 6 min, the whole sequence can be processed in about 5 h. However, by taking advantage of the sequence programing tools available in the CE instrument software, the whole routine, including running buffer replenishment can be run without supervision. Compared to the classical radioisotope labeling methods, the CE technique and the method reported here represent a robust, reliable, and efficient alternative for probing GTPase activity or for investigating mechanistic questions by means of enzyme kinetic studies. Case study
Fig. 2 Example of the graphical approach used for the enzyme kinetics study. The GDP concentration in the reaction mixture was determined by CE and plotted as a function of reaction time (A) for 1 mM (squares) and 4 mM (circles) GTP both in the presence of 1 mM MgCl2. Protein concentration in the assay was 0.184 mg mL1 (ca. 3.1 lM). A straight-line function was fitted to the points within the linear region to yield the initial reaction velocity, v0. (B) Michaelis–Menten plot of the initial reaction velocity as a function of the substrate concentration. The parameters km and vmax were obtained by non-linear fitting of a rectangular hyperbole to the experimental points. Two experiments carried out on different days using different lots of protein are presented. The Michaelis–Menten parameters calculated are the following: (lozenges and solid line) vmax = 1.7 lM min1 ± 0.1 and km = 1.1 mM ± 0.3; (circles and dashed line) vmax = 1.7 lM min1 ± 0.1 and km = 1.0 mM ± 0.2
In this work we have focused on the enzyme kinetics of the catalytic domain expressed as a fusion protein with GST. This construct presents chemical properties and behavior similar to the whole protein and is more stable. For example, the GTPase activity of the GTPase domain construct is detectable even after 48 h post-purification if the protein is kept in ice. On the other hand, the GTPase activity of the whole protein is significantly reduced after 24 h. Therefore, the catalytic domain presents a significant advantage as a model system for the whole protein particularly for the types of kinetic studies described here. In the absence of a three-dimensional structure for the protein, kinetics studies may provide important information to aid in clarifying the catalytic mechanism and thereby help to understand the biological function of the protein. The Michaelis–Menten plot is shown in Fig. 2b where data from two distinct experimental datasets are overlaid. The excellent reproducibility is due to both the precision of the CE method and the protein expression and purification procedure, which have been previously optimized. As we have observed in our studies, GTP hydrolysis was readily measurable suggesting that both catalysis and product liberation are able to occur in the absence of additional auxiliary proteins. Other GTPases, such as FtsZ and Tubulin [14], which also form filaments,
96
Fig. 3 Percentage of GDP after incubation with GST-rDGTPase at different time intervals with 0.5 mM MgCl2 or MnCl2 in comparison to a control mixture. Reaction mixtures contained 3 mM GTP and 0.18 mg mL1 protein
undergo nucleotide exchange induced by polymerization in the absence of exchange factors. This is consistent with the fact that no guanine nucleotide exchange factors have thus far been described in the case of septins [15]. However, the possible existence of such factors may significantly affect the kinetics of GTP hydrolysis by Human SEPT4 and other septins in vivo. Therefore, we have adopted the term ‘‘apparent Kcat’’ (apKcat) as the observed Kcat in vitro under the described experimental conditions. The average values for the kinetic parameters determined in these experiments were as follows: vmax = 1.7 lM min1 ± 0.1, km = 1.0 mM ± 0.3, and apKcat = 9 · 103 s1. In principle, these parameters were expected to be comparable to those from the recombinant Bradeion b: km = 2.0 mM and apKcat = 0.25 s1 (unpublished results); however significant differences are observed. The apKcat for the recombinant SEPT4 Bradeion b is about 25 times greater than for GSTrDGTPase, whilst the apKcat/km values for Bradeion b and the GST-rDGTPase are 0.12 mM1 s1 and 9 · 103 mM1 s1, respectively. The specificity constant of the recombinant Bradeion b is therefore about 13 times greater than that for GST-rDGTPase. These results clearly indicate that the integral protein is more active than the isolated domain. Changes in specificity constants of GTPases due, for example, to mutation of a single amino acid in the active site are quite high [16]. Differences reported are commonly in the range of 102 – 104 times. Furthermore, a difference of 13 times in the specificity constant of a GTPase for GTP in comparison to non-cognate substrates like XTP was considered to be quite low [17]. The reason for the observed difference in the catalytic efficiency is at present unknown and requires further investigation. Bradeion b may contain additional structural domains and biochemical activities such as membrane binding and filament formation which may be
related to GTP binding and/or hydrolysis [18]. The efficient binding of substrate and release of product might be facilitated by conformational changes in these domains, which are lacking in the GST-rDGTPase. Furthermore, there is evidence that the Bradeion b may oligomerize in solution, in a manner similar to other septins and it is still unknown if this affects the enzyme activity or not. Finally the possibility that the GST present in the fusion protein may affect the activity of the GTPase domain cannot be ruled out. Additional structural and biochemical studies are necessary to elucidate these details. In agreement with the chemical behavior of other GTPases, the activity of the GST-rDGTPase is also dependent on the presence of the divalent cation magnesium [19, 20]. No enzymatic activity was observed in the absence of Mg2+. Furthermore, the cation Mn2+
Fig. 4 Dependence of enzyme catalysis on the magnesium concentration. (A) Initial reaction velocity as a function of the Mg2+ concentration. The reaction mixtures containing 3 mM GTP, 0.18 mg mL1 protein and the indicated magnesium concentrations were incubated and the v0 values were determined as described for the Michaelis–Menten experiments. The x-axis in this case is on a logarithmic scale. (B) The same data presented in Fig. 4a plotted as a function of GTP-to-Mg2+ concentration ratio
97
Conclusions A general technique for the quantification of triphosphatase activity has been presented, which although poor regarding the detection limits is readily automated, robust, and reproducible. Compared with traditional assays it has the considerable advantage of not requiring the handling of radioactive isotopes. It is expected to represent a useful tool for routine application in the biochemical investigation of GTPases and related molecules. In the case of septins in particular, it should be valuable in aiding the elucidation of the relationship between GTP hydrolysis and the diverse physiological roles attributed to these molecules, which include essential cellular functions such as cytokinesis and exocytosis. Fig. 5 Separation of (A) cSGTP/GDP, (B) ATP/ADP, and (C ) GTP/GDP under conditions described in the ‘‘Experimental’’ using a PVA-coated capillary. These samples were originally 5 mM solutions from commercial triphosphate nucleotide standards. Prior to the CE analysis, cSGTP had been stored at 20C for more than 6 months; GTP was freshly prepared and ATP had been stored for 2 weeks at 7C. Under these conditions, a peak corresponding to the appropriate diphosphate nucleotide was clearly visible in all cases
Acknowledgments We thank Agilent Technologies for the kind donation of the PVA-coated capillaries. This work was supported by FAPESP via a grant to the Centro de Biotecnologia Molecular Estrutural, and through fellowships to SH, WG, and MDC. The present work was also partially supported by a grant from Azwell Inc. to Tokai University and AIST.
References 2+
was able to substitute Mg in the activation of hydrolysis at a lower extent (Fig. 3). It is therefore evident that these ions are co-factors, which play an essential role in the reaction mechanism. The role of magnesium in the reaction may not be limited to activation alone but it may also have a regulatory function. In the presence of an excess of the metal ion, the enzyme activity falls off. Figure 4a shows that v0 decreases for Mg2+ concentrations greater than 2.5 mM (in reactions containing 3 mM GTP). Interestingly, the effect of Mg2+ was found to be related to the GTP concentration, rather than to its absolute amount. This is more clearly seen in the results presented in Fig. 4b, where v0 from different reaction mixtures is plotted against the concentration ratio [GTP]/[Mg2+ ]. This indicates that optimal activation is achieved for [GTP]/ [Mg2+ ] between 1 and 4 and that outside this range the reaction rate decreases progressively. Maximum activity therefore roughly corresponds to the situation where all GTP molecules are bound to Mg2+ but for which the metal ion is not present in great excess. It is important to point out that the method described here for GTP and GDP analysis may also be applied to the analysis of other nucleotide pairs. Although in our method, GTP and ATP cannot be readily analyzed simultaneously, diphosphated and triphosphated forms of all nucleotides are usually fairly well separated under the same electrophoretic conditions so that the methodology described here can be readily adapted to studying the specificity of triphosphatases in general. Figure 5, for example, shows the separation of cSGTP/ GDP and ATP/ADP in comparison with GTP/GDP. As can be seen, baseline resolution is achieved in all cases.
1. Macara IG, Baldarelli R, Field CM, Glotzer M, Hayashi Y, Hsu SC, Kennedy MB, Kinoshita M, Longtine M, Low C, Maltais LJ, McKenzie L, Mitchison TJ, Nishikawa T, Noda M, Petty EM, Peifer M, Pringle JR, Robinson PJ, Roth D, Russell SE, Stuhlmann H, Tanaka M, Tanaka T, Trimble WS, Ware J, Zeleznik-Le NJ, Zieger B (2002) Mol Biol Cell 13:4111–4113 2. Tanaka M, Tanaka T, Kijima H, Itoh J, Matsuda T, Hori S, Yamamoto M (2001) Biochem Biophys Res Commun 286:547– 553 3. Tanaka T, Kijima H, Itoh J, Matsuda T, Tanaka T (2002) Cancer Gene Ther 9:483–488 4. Mejilano MR, Shivanna BD, Himes RH (1996) Arch Biochem Biophys 33:130–138 5. Bhullar RP, Seneviratne HD (1996) Biochim Biophys Acta 1311:181–188 6. Okada D, Tsukida K, Ikai A (1985) Phochem Photobiol 42:405–411 7. Binns DD, Helms MK, Barylko B, Davis CT, Jameson DM, Albanesi JP, Eccleston JF (2000) Biochemistry 39:7188–7196 8. Tsuda T, Nakagawa G, Sato M, Yagi K (1983) J Appl Biochem 59:330–336 9. Uhrova M, Deyl Z, Suchanek M (1996) J Chrom B 681:99–105 10. Geldart SE, Brown PR (1998) J Chrom A 828:317–336 11. Qurishi R, Kaulich M, Muller CE (2002) J Chrom A 952:275– 281 12. Kawata H, Kuroda K, Endo Y, Inoue Y, Endo G (2000) Tohoku J Exp Med 192:67–79 13. Gill SC, von Hippel PH (1989) Anal Biochem 182:319–326 14. Huecas S, Andreu JM (2004) FEBS 569:43–48 15. Field CM, Kellogg D (1999) Trends Cell Biol 9:387–394 16. Bishop A, Buzko O, Heyeck-Dumas S, Jung I, Kraybill B, Liu Y, Shah K, Ulrich S, Witucki L, Yang F, Zhang C, Shokat KM (2000) Annu Rev Biophys Biomol Struct 29:577–606 17. Shan S, Walter P (2003) PNAS 100:4480–4485 18. Zhang J, Kong C, Xie H, McPherson PS, Grinstein S, Timble WS (1999) Cur Biol 9:1458–146 19. Zhang BL, Zhang YQ, Wang ZX, Zheng Y (2000) J Biol 275:25299–25307 20. Zhao J, Wang WN, Tan YC, Zheng Y, Wang ZX (2002) Biochem Biophys Res Commun 297:653–658
