The Interplay between Calmodulin and Membrane Interactions with the Pleckstrin Homology Domain of Akt*

The Akt protein, a serine/threonine kinase, plays important roles in cell survival, apoptosis, and oncogenes. Akt is translocated to the plasma membrane for activation. Akt-membrane binding is mediated by direct interactions between its pleckstrin homology domain (PHD) and phosphatidylinositol 3,4,5-trisphosphate (PI(3,4,5)P3). It has been shown that Akt activation in breast cancer cells is modulated by calmodulin (CaM). However, the molecular mechanism of the interplay between CaM and membrane binding is not established. Here, we employed nuclear magnetic resonance (NMR) and biochemical and biophysical techniques to characterize how PI(3,4,5)P3, CaM, and membrane mimetics (nanodisc) bind to Akt(PHD). We show that PI(3,4,5)P3 binding to Akt(PHD) displaces the C-terminal lobe of CaM but not the weakly binding N-terminal lobe. However, binding of a PI(3,4,5)P3-embedded membrane nanodisc to Akt(PHD) with a 103-fold tighter affinity than PI(3,4,5)P3 is able to completely displace CaM. We also show that Akt(PHD) binds to both layers of the nanodisc, indicating proper incorporation of PI(3,4,5)P3 on the nanodisc surface. No detectable binding has been observed between Akt(PHD) and PI(3,4,5)P3-free nanodiscs, demonstrating that PI(3,4,5)P3 is required for membrane binding, CaM displacement, and Akt activation. Using pancreatic cancer cells, we demonstrate that inhibition of Akt-CaM binding attenuated Akt activation. Our findings support a model by which CaM binds to Akt to facilitate its translocation to the membrane. Elucidation of the molecular details of the interplay between membrane and CaM binding to Akt may help in the development of potential targets to control the pathophysiological processes of cell survival.

Recent studies demonstrated that membrane targeting and subsequent activation of Akt are also modulated by phosphatidylserine (PS) (23). It has been proposed that PS interacts with specific residues in the PHD and HM of Akt to promote binding to PI(3,4,5)P 3 . The synergy between PS and PI(3,4,5)P 3 is thought to induce Akt inter-domain conformational changes required for phosphorylation of Thr-308 and Ser-473 (23). Mutagenesis studies revealed that disruption of Akt-PS interaction impairs Akt signaling and increases susceptibility to cell death, suggesting that PS is critical for Akt activation and cell survival particularly in conditions with limited PI(3,4,5)P 3 availability (23). Taken together, these studies imply that a synergy between membrane lipids is perhaps needed for Akt localization and activation.
In addition to binding to lipids, an emerging role for Akt-(PHD) as a multifunctional domain that can also bind to proteins is now being discovered. Studies have shown that in cells under resting conditions, the PHD interacts with its KD in a closed conformation (24 -26). In that stage, Thr-308 is blocked and is inaccessible to PDK1 (26). This closed conformation is only disrupted when the PHD binds to PI(3,4,5)P 3 (24,25). This disruption is required for subsequent exposure and phosphorylation of residue Thr-308 in the KD. Mutations in the PHD that disrupt PHD-KD interaction result in constitutive Akt activation (27). Furthermore, the PHD has been shown to bind to calmodulin (CaM) to promote Akt1 (will be called Akt throughout this paper) activation in many breast cancer cells (2,28). CaM is a ubiquitous calcium-binding protein expressed in all eukaryotic cells (29 -34). Upon epidermal growth factor (EGF) stimulation of these cells, CaM co-localizes with Akt at the PM to enable membrane anchoring and activation (2). Perturbations of this targeting mechanism by a CaM antagonist inhibited Akt-CaM translocation to the PM and led to apoptotic cell death in tumorigenic mammary carcinoma cells (2,35). These studies are consistent with the findings that CaM is overexpressed in breast tumors and breast cancer cells lines (36,37).
Until recently, the molecular mechanisms governing the Akt(PHD)-CaM interaction were not well characterized at the biochemical, biophysical, and structural levels. Attempting to understand the mechanism by which CaM and lipids contribute into Akt activation, we have initially characterized CaM interaction with Akt(PHD) (38). We have demonstrated that CaM forms a tight complex with the Akt(PHD) with a K d of 100 nM (38). The CaM-binding interface on Akt(PHD) was mapped to two loops adjacent to the PI(3,4,5)P 3 -binding site, representing a novel CaM-binding motif and suggesting a synergistic relationship between CaM and PI(3,4,5)P 3 for Akt activation (38). This proposed role of CaM is consistent with the recent findings that CaM enhances PI(3,4,5)P 3 binding to the PHD of a homologous protein from the Tec kinase family called interleukin-2 inducible tyrosine kinase (Itk) and vice versa (39). Conversely, these findings are inconsistent with other studies that indicated that CaM competes with PI (3,4,5) 3 for Akt(PHD) thereby decreasing the affinity of PI(3,4,5)P 3 to Akt(PHD) (28). Taken together, we strongly believe that further investigation is required to clarify the interplay between CaM and PM lipids in Akt activation. To understand these mechanisms, it is essential to characterize at the molecular level how CaM and lipids bind to Akt(PHD).
Here, we employed NMR, biophysical, biochemical, and in vivo tools to understand the interplay between CaM and membrane interactions with Akt. We utilized dibutanoyl-phosphatidylinositol 3,4,5-trisphosphate ((diC 4 -PI(3,4,5)P 3 ), a soluble analog of PI(3,4,5)P 3 , with truncated acyl chains. We show the following: (i) diC 4 -PI(3,4,5)P 3 induces major changes in Akt-(PHD); (ii) the glycerol backbone and acyl chains of PI(3,4,5)P 3 may contribute to the interaction; (iii) diC 4 -PI(3,4,5)P 3 binds to Akt(PHD) with a K d of 13 M; and (iv) CaM does not enhance binding of PI(3,4,5)P 3 to Akt(PHD). By using PI(3,4,5)P 3 -and PS-incorporated nanodisc membrane model, we also show that Akt(PHD) binds tightly to membrane with a K d of 19 nM, which is able to displace CaM from Akt(PHD). Finally, we demonstrate that inhibition of Akt-CaM binding attenuated Akt activation in pancreatic cancer cells. Elucidation of the mechanism by which Akt interacts with both CaM and lipids will help in the understanding of the activation mechanism, which may provide insight for new potential targets to control the pathophysiological processes of cell survival.

Results
It is established that membrane binding of Akt is regulated by specific interactions between PI(3,4,5)P 3 and the PHD. However, the molecular mechanisms governing Akt(PHD)-membrane interactions and the interplay between membrane lipids and CaM are not well characterized. This is the first study aimed at characterization of the interplay of CaM and lipid/ membrane binding to Akt(PHD).
NMR chemical shifts are very sensitive to variations in the local molecular environment. Small changes in the 1 H and 15 N resonances obtained by collecting HSQC spectra on protein⅐ protein and protein⅐lipid complexes can be used to map the binding interface. These experiments not only allow for identification of residues involved in the interaction and/or accompanying conformational changes, but they also provide an effective method for examining the stability and folding of the protein and assessing the strength of the interaction. We initially assessed how diC 4 -PI(3,4,5)P 3 binds to Akt(PHD) using NMR chemical shift perturbations (CSPs) as detected in the 1 H-15 N HSQC spectra. Because of the propensity of Akt(PHD) to self-associate and form aggregates in solution at high protein concentrations (39,40,48,49), which often complicates the NMR studies, we conducted the HSQC NMR experiments at low protein concentrations (60 M). Titration of diC 4  Asp-108. To identify the interaction interface of Akt(PHD), the CSPs were mapped on the structure of Akt(PHD). As shown in Fig. 2B, the significantly perturbed residues form a well defined interaction interface (red). The PI(3,4,5)P 3 site is composed of a basic patch formed by ␤1/␤2 strands and the loop that connects them (residues 13-23 and 53). This region defines a binding site that is similar to the previously published Akt(PHD) structure with IP 4 (41).
Next, we conducted similar NMR studies with IP 4 . Titration of IP 4 into a 15 N-labeled Akt(PHD) led to significant CSPs for a number of signals. Titrations were conducted to 8:1 IP 4 /PHD. Further addition of IP 4 led to sample precipitation. Overall, the chemical shift changes are similar to those caused by binding of diC 4 -PI(3,4,5)P 3 (Fig. 2). The chemical shift changes in the HSQC spectra (data not shown) indicate a fast exchange, on the NMR scale, between free and bound forms. Some differences, however, were observed. For example, a subset of signals corresponding to some residues such as Arg-15, Trp-22, and Leu-28 exhibited more significant chemical shift changes upon binding of diC 4 -PI(3,4,5)P 3 than that of IP 4 . NMR chemical shift changes under a fast exchange regime allowed for determination of the dissociation constant (K d ) of IP 4 from Akt-(PHD). Non-linear least squares fits of the titration data afforded a K d of 202 Ϯ 46 M (data not shown). In contrast, determination of the K d value from the NMR titration data of diC 4 -PI(3,4,5)P 3 is more complicated because both fast and slow-to-intermediate exchange regimes are observed. Therefore, we resorted to ITC methods to measure the K d and other thermodynamic parameters.
We obtained ITC data upon titration of diC 4 -PI(3,4,5)P 3 into Akt(PHD). ITC provides values of K d , stoichiometry of binding (n), enthalpy change (⌬H 0 ), and the entropic terms (T⌬S 0 ). As shown in Fig. 3, binding of diC 4 -PI(3,4,5)P 3 to Akt(PHD) is exothermic as indicated by the sign of the enthalpy. The binding data were fit into a one-site binding model and yielded the following thermodynamic parameters: K d ϭ 13 M; n ϭ 0.92 Ϯ 0.06; ⌬H 0 ϭ Ϫ2.7 kcal/mol; and ⌬S 0 ϭ 13.0 cal/mol/degree. diC 4 -PI(3,4,5)P 3 binding to Akt(PHD) is also stabilized by hydrophobic factors as indicated by the value of T⌬S 0 (Table 1). Collectively, our ITC data show that diC 4 -PI(3,4,5)P 3 binds to Akt(PHD) in a 1:1 model and that both ionic and possible hydrophobic interactions contribute to the formation of the Akt(PHD)⅐PI(3,4,5)P 3 complex. Interestingly, the affinity of diC 4 -PI(3,4,5)P 3 binding to Akt(PHD) is 15-fold tighter than that observed for IP 4 , suggesting that the acyl chains and/or glycerol group contribute to the overall lipid binding. This result is also consistent with the NMR data.
Binding of PI (3,4,5)P 3 and CaM to Akt(PHD), Competitive or Cooperative?-Akt(PHD) interacts with CaM to mediate membrane translocation and activation. Confocal microscopy stud- ies have shown that upon EGF stimulation, CaM co-localizes with Akt(PHD) at the PM (2). We have recently demonstrated that Akt(PHD) binds tightly to CaM with a K d of 100 nM and a 1:1 binding mode (38). Using NMR CSPs, we mapped the interaction region to two loops on Akt(PHD) and both lobes of CaM, establishing a novel CaM-binding motif. We found that the C-terminal lobe of CaM (CaM-C) recognizes the loop connecting ␤1 and ␤2 strands, whereas the N-terminal lobe (CaM-N) bind to the loop connecting ␤6 and ␤7 strands (38). Interestingly, this region is adjacent to the diC 4 -PI(3,4,5)P 3 -binding site described above. When compared with the HSQC spectra of the Akt(PHD)⅐CaM complex (38), many of the residues that are significantly affected by CaM binding are different from those affected by diC 4 -PI(3,4,5)P 3 binding. For example, the side chain (⑀1) 1 H-15 N resonances corresponding to Trp-80 and Trp-22 exhibited significant chemical shift changes upon binding of CaM but not diC 4 -PI(3,4,5)P 3 (Fig. 1A). Residues that are involved in lipid binding such as Lys-14, Arg-15, Arg-23, Arg-25, and Asn-53 exhibit significant chemical shift changes when titrated with diC 4 -PI(3,4,5)P 3 but not with CaM (Figs. 1 and 2). The aim here was to assess whether binding of CaM and diC 4 -PI(3,4,5)P 3 to Akt(PHD) is competitive or cooperative and whether both CaM and diC 4 -PI(3,4,5)P 3 are able to bind simultaneously to Akt(PHD).
To do so, we conducted 2D HSQC NMR experiments on a 15 N-labeled Akt(PHD) sample upon binding to unlabeled CaM and diC 4 -PI(3,4,5)P 3 (Fig. 4). First, unlabeled CaM was added to 15 N-labeled samples of Akt(PHD) at 1:1 Akt(PHD)/CaM ratio followed by acquisition of 2D 1 H-15 N HSQC data. As expected, substantial chemical shift changes were observed for most of the 1 H-15 N resonances, indicating direct binding. Second, diC 4 -PI(3,4,5)P 3 was titrated to the Akt(PHD)⅐CaM sample followed by acquisition of 1 H-15 N HSQC data (Fig. 4A). At 0.5:1:1 diC 4 -PI(3,4,5)P 3 /Akt(PHD)/CaM, signal intensity decreased   for several signals (data not shown). Further addition of diC 4 -PI(3,4,5)P 3 led to the appearance of several new peaks. Remarkably, at 1:1:1 diC 4 -PI(3,4,5)P 3 /Akt(PHD)/CaM, two separate resonances (doubling) for most of the signals (green) are observed (Fig. 4). Because the NMR spectra were quite complex to analyze, we picked a few signals corresponding to residues within the lipid-binding site (e.g. Lys-14, Arg-15, Arg-23, and Leu-28) to highlight these spectral changes (Fig. 4B). The chemical shifts of the doubling signals correspond to either CaM-or diC 4 -PI(3,4,5)P 3 -bound forms of Akt(PHD). This result indicates that peak doubling of these residues represents an equilibrium exchange of two conformational states, with one state representing the Akt(PHD)⅐diC 4  The above result was unexpected because CaM affinity to Akt(PHD) is ϳ10 3 times higher than that of PI(3,4,5)P 3 . One possible model to explain the above observation is that CaM binding to Akt(PHD) may induce conformational changes in Akt(PHD) to enhance lipid binding. To test this hypothesis, we have obtained ITC data upon titration of diC 4 -PI(3,4,5)P 3 into the pre-formed Akt(PHD)⅐CaM complex under the same buffer conditions (Fig. 5). The binding data were fit into a one-site binding model and yielded the following thermodynamic parameters: K d ϭ 8 M; n ϭ 1.04 Ϯ 0.08; ⌬H 0 ϭ Ϫ3.05 kcal; and ⌬S 0 ϭ 12.9 cal/mol/degree. Interestingly, as indicated by the K d values, the affinity of lipid binding to the Akt(PHD)⅐CaM complex is only slightly higher than that of Akt(PHD), suggesting that CaM does not significantly enhance lipid affinity to Akt(PHD).

Akt(PHD)-CaM
Another possible scenario for the above observation is that the lipid only displaces CaM-C, which is close to the lipid-binding site, but not CaM-N, which is far removed from it. To examine whether this is the case, we picked a few other signals corresponding to Akt(PHD) residues Thr-81, Thr-82, Ile-84, and Gln-113 that are distant from the PI(3,4,5)P 3 -binding site but are located in the ␤6 and ␤7 loop that binds to CaM-N (38). As shown in Fig. 4C, addition of diC 4 -PI(3,4,5)P 3 did not affect these signals, suggesting that the ␤6 and ␤7 loop is still bound to CaM-N. Altogether, NMR data suggest that diC 4 -PI(3,4,5)P 3 partially displaces CaM by only displacing the CaM-C lobe.
To determine whether CaM-N is still bound to Akt(PHD) upon lipid titration, we collected the 2D HSQC data on a 15 Nlabeled CaM-N sample as a function of added Akt(PHD) and diC 4 -PI(3,4,5)P 3 (Fig. 6). As expected, addition of unlabeled Akt(PHD) to CaM-N led to modest chemical shift changes in the HSQC spectra, because CaM-N binds much weaker (112 M) to Akt(PHD) than full CaM (38). Next, diC 4 -PI(3,4,5)P 3 was titrated into the complex to saturation (Fig. 6). The 1 H and 15 N signals did not shift to positions similar to those observed for free CaM-N upon addition of diC 4 -PI(3,4,5)P 3 , indicating that lipid does not displace CaM-N from Akt(PHD). In a control experiment, no changes were observed upon diC 4 -PI(3,4,5)P 3 titration into CaM. Based on these findings, we conclude that the ternary complex formation is not cooperative and that diC 4 -PI(3,4,5)P 3 is able to compete with CaM, or at least displace the CaM-C lobe adjacent to the lipid-binding site.
Membrane Nanodisc Interactions with Akt(PHD)-The results above confirm that diC 4 -PI(3,4,5)P 3 can potently compete with CaM to bind to Akt(PHD). However, to better under-stand the mechanism of Akt-membrane in the context of a biologically relevant system, we conducted these competition experiments with membrane nanodiscs with biologically relevant lipid composition. In these experiments, we also take into account the recent findings showing that PS enhances Akt-(PHD)-membrane binding (23). Therefore, we wanted to assess the molecular mechanism of Akt(PHD) and Akt(PHD)⅐CaM binding to the membrane mimetic model containing PI (3,4,5)P 3 and PS with native acyl chains. Phosphoinositides containing detergent micelles and liposomes have been previously used to study Akt-membrane interactions (23). However, these systems can have some drawbacks because detergents are not homogeneous and may affect protein stability and folding. Liposomes, however, are large with diameters on the order of microns and contain thousands to millions of phospholipids, which may not allow us to precisely characterize their binding to Akt(PHD). Therefore, we have utilized stable and homogeneous membrane nanodiscs, which have been used successfully in structural and biophysical studies (50 -55).
Binding of nanodiscs to Akt(PHD) and Akt(PHD)⅐CaM has been examined by a gel filtration assay, ITC, and cobalt affinity pulldown assay. Two membrane nanodisc models with or without DPPI(3,4,5)P 3 have been prepared as described previously (51,56). All nanodiscs were run on a size exclusion column (Superdex 200) to ensure homogeneity. As shown in Fig. 7A, a single elution peak of 1:78:20:2 MSP1D1/POPC/POPS/ DPPI(3,4,5)P 3 nanodisc was observed at 45.5 ml. A sample mixture containing equimolar amounts of MSP1D1/POPC/POPS/ DPPI(3,4,5)P 3 nanodisc and Akt(PHD) at 70 M was prepared and loaded onto a gel filtration column (Superdex 200). The nanodisc⅐Akt(PHD) complex yielded a peak at 45.2 ml and a broad peak at ϳ85 ml (Fig. 7A). SDS-PAGE analysis of peak fractions indicate complex formation between Akt(PHD) and nanodisc (Fig. 7B). These results demonstrate direct binding of Akt(PHD) to nanodisc. SDS-PAGE data revealed that the pro-  tein eluting at ϳ85 ml is unbound Akt(PHD). To test whether Akt(PHD) binds to nanodisc in the absence of DPPI(3,4,5)P 3 , a mixture of equimolar amounts of Akt(PHD) and nanodisc made of MSP1D1/POPC/POPS were mixed and run on size exclusion chromatography. Two peaks indicating nanodisc and Akt(PHD) were observed (data not shown). Analysis of the fractions via SDS-PAGE showed no evidence of Akt(PHD)-nanodisc binding. Our data unequivocally validate our approach and demonstrate that DPPI(3,4,5)P 3 is required for Akt(PHD) binding to membranes.
Next, we aimed at obtaining quantitative data on nanodisc binding to Akt(PHD), including binding affinity and stoichiometry. We resorted to ITC methods as it provides such information and other thermodynamic properties. A sample of MSP1D1/POPC/POPS/DPPI(3,4,5)P 3 nanodisc at 60 M was titrated into Akt(PHD) at 10 M at identical buffer conditions (Fig. 8A). This experiment was collected in triplicate and yielded similar values. Data were best fit to one-site binding model and yielded a K d value of 19 nM, an affinity that is ϳ10 3 tighter than that obtained for diC 4 -PI(3,4,5)P 3 titration. As indicated by the ⌬H 0 value, the interaction is exothermic and enthalpically driven (Fig. 8A and Table 1). This result is striking and suggests that Akt(PHD) interaction with the membrane is governed by multiple forces. Of particular note, the stoichiometry of binding (n) value is 0.57, which suggests that two molecules of Akt(PHD) bind to one nanodisc. The n value indicating the binding stoichiometry suggests that two Akt(PHD) molecules bind to one nanodisc, perhaps one molecule per membrane layer. ITC experiments conducted in the absence of DPPI(3,4,5)P 3 showed no detectable binding (Fig. 8A), again demonstrating that PI(3,4,5)P 3 is indispensable for Akt(PHD) binding to membranes.
To further assess the stoichiometry of Akt(PHD)-nanodisc binding, we employed a cobalt-affinity pulldown assay. A DPPI(3,4,5)P 3 -nanodisc was immobilized onto a cobalt resin via a His 6 tag fused to the N terminus of the MSP1D1 protein. It is important to note that a nanodisc contains two copies of the MSP1D1 protein. Akt(PHD) was added to the immobilized nanodisc at a 1:1 or 1:2 ratio of nanodisc/Akt(PHD). Fractions of the flow-through, washes, and elutions were analyzed via SDS-PAGE. As shown in Fig. 8B, two bands representing nanodisc and Akt(PHD) are observed. As expected, at 1:1 nanodisc/ Akt(PHD) the MSP1D1 band representing the nanodisc is more intense than the Akt(PHD) because two molecules of MSP1D1 wrap around the phospholipid bilayer. However, at 1:2 nanodisc/Akt(PHD) the intensity of the MSP1D1 protein is very similar to that of the Akt(PHD) band indicating similar protein ratios. Consistent with the ITC data, the cobalt pulldown data indicate that two molecules of Akt(PHD) are bound to the nanodisc.
Does Membrane Nanodisc Displace CaM?-To elucidate the molecular mechanism of Akt activation, it is essential to understand the interplay of CaM and membrane binding to Akt-(PHD). The NMR experiments performed with the diC 4 -PI(3,4,5)P 3 lipid show that the lipid is able to partially displace CaM by competing with the C-terminal lobe. However, we  wanted to examine this competition by using a nanodisc model, which is more biologically relevant. One aim was to test whether both CaM and nanodisc could simultaneously bind to Akt(PHD) and to determine whether CaM enhances or inhibits lipid binding to Akt(PHD). To do so, nanodisc was added to a pre-formed Akt(PHD)⅐CaM complex at 1:1:1 stoichiometry and loaded onto the size exclusion column (Superdex 200). As shown in the elution profile in Fig. 9A, two peaks at 45.5 and 62.1 ml were observed. SDS-PAGE analysis demonstrates that two different complexes were formed, a Akt(PHD)⅐nanodisc complex at 45.5 ml (fractions 38 -54) and Akt(PHD)⅐CaM complex at 62.1 (fractions 60 -70) (Fig. 9, A and B). No CaM protein is detected in the Akt(PHD)-nanodisc fractions (38 -54), indicating that CaM is not associated with the Akt(PHD)⅐nanodisc complex.
To confirm this result, we recapitulated this finding using a co-affinity pulldown assay. Solution fractions of the Akt (PHD)⅐CaM complex were added onto a nanodisc immobilized on a cobalt resin at 1:1 stoichiometry. SDS-PAGE data show that nanodisc binding displaces CaM (Fig. 9C). However, consistent with our gel filtration data, we observe a fraction of the Akt(PHD) protein in the flow-through fractions with CaM (Fig.  9C) suggesting that both CaM and PI(3,4,5)P 3 may have a similar affinity for Akt(PHD). ITC data show that nanodiscs bind to Akt(PHD) 10-fold tighter than CaM (18 versus 195 nM at 150 mM NaCl). Therefore, we wanted to examine whether CaM binding to Akt(PHD) enhanced or decreased affinity of binding to nanodisc. We obtained ITC data upon titration of nanodisc into the Akt(PHD)⅐CaM complex (Fig. 9D). Data fit best to a one-site model and provided an apparent K d of 200 nM, which suggests that CaM may moderately decrease lipid binding to Akt(PHD). However, the ITC data can also be complicated by the fact the heat generated in the cell is a combined effect of formation of two different complexes, nanodisc⅐Akt(PHD) and CaM⅐Akt(PHD).
Next, we sought to find out whether CaM is able to capture Akt(PHD) when prebound to nanodisc. This experiment is important because it may help in understanding at the cellular level whether CaM can capture Akt when bound to the PM. To do so, we designed a cobalt pulldown completion assay in which a preformed nanodisc containing 2% PI(3,4,5)P 3 was first bound to Akt(PHD). CaM was then added to the complex. The resulting complex was bound to cobalt resin, washed, and eluted, and fractions were analyzed by SDS-PAGE. As shown in Fig. 9E, the CaM protein was washed out, whereas Akt(PHD) remained exclusively captured by nanodisc. This result indicates that CaM is not able to bind and pull down Akt(PHD) when it is bound with membrane. In summary, our results demonstrate that Akt(PHD) has a higher affinity to membranes containing PI(3,4,5)P 3 and that CaM is displaced and cannot bind simultaneously, a result that has potential biological implications on Akt activation.

CaM Inhibition Attenuates Akt Activation in Pancreatic
Cancer Cells-To evaluate the functional role of CaM in Akt activation in cancer cells, we determined the effect of CaM inhibition on the activation of Akt by EGF, a growth factor that has been shown to activate Akt in breast cancer cells (2,28). We found that EGF induced Akt phosphorylation at both Ser-473 and Thr-308 in pancreatic cancer MiaPaCa-2 cells (Fig. 10A). Inhibition of CaM by using trifluoperazine (TFP) markedly attenuated EGF-induced Akt phosphorylation and activation. Using immunofluorescent staining and confocal microscopy, we further confirmed that EGF induced phosphorylation and activation of Akt. Furthermore, we show that CaM co-localizes with Akt on the membrane in MiaPaCa-2 cells (Fig. 10B). Immunoprecipitation analysis further shows a direct interaction of CaM and Akt upon EGF stimulation (Fig. 10C), which is inhibited by TFP. Taken together, these results demonstrate that CaM⅐Akt binding is important for regulating Akt activation in pancreatic cancer cells.

Discussion
Akt plays a key role in many cellular processes, including cell survival, apoptosis, and oncogenesis. Growth and survival signals initiated at the cell surface often involve activation of Akt, which is required for Akt to directly phosphorylate its downstream cellular targets such as the BH3-only protein BAD, glycogen synthase kinase 3, and tuberous sclerosis complex 2 (1-3, 11, 57). The PHD, a conserved structural fold found in organisms ranging from mammals to bacteria (20 -22), binds specifically to PI(3,4,5)P 3 on the inner leaflet of the PM (1,11,19). PS has also been shown to enhance Akt(PHD) binding to membrane (23). The mechanisms of Akt translocation to the PM for phosphorylation and subsequent activation of its effector proteins are critical steps in the activation pathway. Understanding these mechanisms became even more complex after the discovery of a role of CaM in mediating Akt-membrane translocation and activation in breast cancer cells (2,35). Disruption of the Akt-CaM interaction using a CaM antagonist blocked translocation to the membrane and led to apoptotic cell death in tumorigenic mammary carcinoma cells (2,35).
To elucidate the mechanism of Akt activation, it is essential to characterize the molecular requirements for these interactions. In a recent report (38), we have characterized the interactions between CaM and Akt(PHD) by NMR, biochemical, and biophysical methods. We have shown that CaM binds tightly to Akt(PHD) (K d ϭ 100 nM). The interaction between CaM and Akt(PHD) is enthalpically driven and dependent on salt concentration, suggesting that electrostatic interactions contribute to the formation of the complex. By using NMR CSPs, we mapped the CaM-binding interface in Akt(PHD) to two loops, of which one of them (connecting ␤1/␤2 strands) is adjacent to the PI(3,4,5)P 3 -binding site. We have also shown that Akt(PHD) engages both CaM lobes simultaneously.
In this study, we employed a battery of methods and techniques to elucidate the interplay between CaM and membrane interactions with Akt(PHD). Structural characterization of protein⅐lipid complexes is often performed with either the polar head of membrane lipids or other soluble analogs bearing truncated acyl chains. Previous NMR and X-ray crystallography studies of Akt(PHD) have used IP 3 and IP 4 , respectively (40,41). Previous studies have shown that IP 4 and PI(3,4,5)P 3 can play different roles in Akt activation (46) and that IP 4 binds to the PHD of protein kinases with a lesser affinity than soluble PI(3,4,5)P 3 analogs with truncated acyl chains (39,47). Herein, we show that although both bind to the same site, diC 4 -PI(3,4,5)P 3 binds to Akt(PHD) 15-fold tighter than IP 4 , which suggests that the acyl chains and glycerol group contribute to binding. Indeed, several additional residues adjacent to the lipid-binding site exhibited chemical shift changes suggesting a direct contact with the acyl chains and glycerol group.
To be able to understand the role of CaM in Akt activation, it is important to assess whether CaM and membrane binding to FIGURE 10. CaM inhibition attenuates Akt activation. Pancreatic cancer cells, MiaPaCa-2, were exposed to serum-free media with or without TFP (10 M) for 30 min and subsequently exposed to EGF (20 ng/ml) with or without TFP. A, Western blotting analysis of the phosphorylation of Akt at Ser-473 and Thr-308 in response to EGF for up to 15 min. B, confocal imaging analysis of EGF-induced Akt phosphorylation (pAkt, red) and co-localization of pAKT (red) with CaM (green) at 5 min. Enlarged image shows co-localization of pAkt with CaM. Scale bar, 10 m. C, interaction of CaM and Akt, assessed by immunoprecipitation (IP) with CaM antibody followed by Western blotting analysis of CaM and Akt with specific antibodies. Representative results of at least three independent experiments are shown.
Akt(PHD) is competitive or cooperative. Our NMR competition assay demonstrates that the ternary complex formation is not cooperative and that diC 4 -PI(3,4,5)P 3 competes with CaM to bind to Akt(PHD). The finding that diC 4 -PI(3,4,5)P 3 displaces only the C-terminal lobe of CaM was interesting because CaM binds tighter than the lipid. Therefore, it was necessary to establish a competition assay with membranes containing biologically relevant lipid composition, including PI(3,4,5)P 3 and PS. By employing a nanodisc binding assay, we demonstrate that Akt(PHD) binds very tightly to the membrane with a K d of 19 nM. This interaction is exothermic and enthalpically driven, suggesting that multiple forces govern Akt(PHD)-membrane interaction. We provide evidence that CaM is fully displaced upon Akt(PHD) binding to nanodisc.
Our findings have biological implications on Akt activation because we provide the first mechanistic evidence for the interplay of both CaM and membrane binding to Akt(PHD). Furthermore, we demonstrate that inhibition of CaM⅐Akt binding attenuated Akt activation in pancreatic cancer cells, which is also associated with disruption of membrane co-localization of CaM and Akt. Our results are consistent with previous observations in breast cancer cells that a CaM antagonist prevented translocation of both Akt and CaM to the PM (2). These results suggest that CaM binds to Akt(PHD) in the cytoplasm and mediates its transport to the PM. Studies have shown that under no stimulus in the cytoplasm, the PHD interacts with the KD in a closed formation (24,25). The Akt(PHD)-KD interface was mapped to ␤1/␤2 strands and VL1 and VL3 of Akt(PHD) with the PI(3,4,5)P 3 binding pocket partially blocked (26). Interestingly, we mapped the CaM binding domain to the VL1 and VL3 loops of Akt(PHD) (38). We propose a model (Fig. 11) by which CaM binding to Akt(PHD) disrupts its interaction with the KD, leading to enhancement of membrane targeting and localization because studies have shown that mutations that disrupt the Akt(PHD)-KD leads to constitutive Akt activation (27). Indeed, this role is consistent with the finding that membrane binding to Akt(PHD) disrupts the Akt(PHD)-KD interaction (24,25). Confocal microscopy studies have shown that upon EGF stimulation, CaM co-localizes with Akt(PHD) at the PM suggesting that CaM may play a role in shuttling of Akt to the membrane.
Interestingly, our findings are not in agreement with previous studies which indicated that CaM competes with PI(3,4,5) 3 for Akt(PHD) thereby decreasing the affinity of PI(3,4,5)P 3 to Akt(PHD) (28). These studies have been performed by using a pulldown assay, phosphoinositide membrane strip, and GSTtagged Akt(PHD) protein. However, our studies were conducted with native and untagged Akt(PHD) protein and a physiologically relevant membrane mimetic containing PS, which has been shown to enhance membrane binding (23). It is possible that these differences in experimental design can adversely affect the results. In another report (39), it has been shown that CaM act in a positive feedback loop to potentiate Ca 2ϩ signaling downstream of the Tec kinase family member Itk. Similar to our findings, the CaM-binding site was mapped to two loops adjacent to the lipid binding pocket within the Itk(PHD). However, in a drastic difference Itk(PHD) bound synergistically to CaM and PI(3,4,5)P 3 (and not IP 4 ) such that binding to CaM enhanced the binding to PI(3,4,5)P 3 and vice versa. Altogether, it appears that the interplay between CaM and phosphoinositide is key to activation of Akt and analogous PHD-containing proteins such as Itk. Elucidation of the interplay between membrane lipids and CaM-mediated Akt activation in cancer cells will advance our understanding of the mechanisms governing Akt activation in general and may contribute to the development of novel anti-cancer drugs that disrupt these interactions.

Experimental Procedures
Sample Preparation-Expression and purification of the CaM protein were performed as described (58). CaM samples were stored in a buffer containing 50 mM HEPES or Tris (pH 7.0), 150 mM NaCl, and 1 mM CaCl 2 . The CaM-N domain (residues 1-80) was expressed and purified as described (59). Because CaM-N has zero extinction coefficient, protein concentration was measured using the bicinchoninic acid assay (BCA; Thermo Scientific). A plasmid encoding for the Akt-PHD protein fused to a small ubiquitin-like modifier tag was constructed as described (38). The Akt protein sequence is 98% identical to that of human Akt1 (Swiss-Prot code P31749). Expression and purification of Akt(PHD) were conducted as described (38). The Akt(PHD) protein was stored in 50 mM Tris or HEPES (pH 7.0), 150 mM NaCl, and 1 mM CaCl 2 . diC 4 -PI(3,4,5)P 3 and IP 4 were used as received from Echelon Biosci- FIGURE 11. Proposed model highlighting the role of CaM in Akt activation. Based on our data, we propose that upon EGF stimulation CaM interacts with the PHD of Akt and mediates Akt translocation to the PM for phosphorylation and activation. CaM is then released upon binding of Akt to the PM. These results are consistent with previous studies suggesting that CaM may be regulating Akt activity by modulating its subcellular location. and exposed to EGF with or without TFP for 5 min. Cells were washed with ice-cold PBS and fixed with 4% (w/v) paraformaldehyde, blocked with 1% BSA, and incubated with primary antibody, a mouse monoclonal anti-CaM, and a rabbit polyclonal anti-phospho-Akt(Thr-308) antibody (10 g/ml) at room temperature for 1 h and then secondary antibody, Alexa Fluor 488 anti-mouse IgG, or Alexa Fluor 594 anti-rabbit IgG (10 g/ml) for 30 min. Microscopic images were taken at ϫ100 magnification using a confocal microscope (Nikon A1R, Nikon Instruments Inc., Melville, NY).
Author Contributions-C. A., R. H. G., F. X., Y. S., Y. C., and J. S. S. designed and executed the experiments and analyzed the data. C. A., Y. C., and J. S. S. compiled the figures and wrote the manuscript.