Coordinate phosphorylation of multiple residues on single AKT1 and AKT2 molecules

Abstract

Aberrant AKT activation is prevalent across multiple human cancer lineages providing an important new target for therapy. Twenty-two independent phosphorylation sites have been identified on specific AKT isoforms likely contributing to differential isoform regulation. However, the mechanisms regulating phosphorylation of individual AKT isoform molecules have not been elucidated due to the lack of robust approaches able to assess phosphorylation of multiple sites on a single AKT molecule. Using a nanofluidic proteomic immunoassay (NIA), consisting of isoelectric focusing followed by sensitive chemiluminescence detection, we demonstrate that under basal and ligand-induced conditions that the pattern of phosphorylation events is markedly different between AKT1 and AKT2. Indeed, there are at least 12 AKT1 peaks and at least 5 AKT2 peaks consistent with complex combinations of phosphorylation of different sites on individual AKT molecules. Following insulin stimulation, AKT1 was phosphorylated at Thr308 in the T-loop and Ser473 in the hydrophobic domain. In contrast, AKT2 was only phosphorylated at the equivalent sites (Thr309 and Ser474) at low levels. Further, Thr308 and Ser473 phosphorylation occurred predominantly on the same AKT1 molecules, whereas Thr309 and Ser474 were phosphorylated primarily on different AKT2 molecules. While basal AKT2 phosphorylation was sensitive to inhibition of PI3K, basal AKT1 phosphorylation was essentially resistant. PI3K inhibition decreased pThr451 on AKT2 but not pThr450 on AKT1. Thus NIA technology provides an ability to characterize coordinate phosphorylation of individual AKT molecules providing important information about AKT isoform-specific phosphorylation, which is required for optimal development and implementation of drugs targeting aberrant AKT activation.

Introduction

The serine/threonine kinase AKT (also known as protein kinase B, PKB), comprising a group of 3 isoforms (AKT1, AKT2, and AKT3), plays critical roles in many aspects of cancer pathophysiology including cell survival, growth, metabolism, and metastasis (1-3). AKT is a bona fide oncogene that is frequently activated in cancer through a variety of mechanisms including amplification of growth factor receptors (i.e., HER2/neu, EGFR), amplification or mutation of phosphatidylinositol 3-kinase (PI3K), amplification or mutation of AKT isoforms, and inactivation of phosphatase and tensin homolog (PTEN) or inositol polyphosphate-phosphatase type II (INPP4B) (3). Different AKT isoforms appear to mediate critical non-redundant functions in cancer pathophysiology (4-6). For example, AKT1 has been implicated as a major contributor to tumor initiation, whereas AKT2 appears to primarily increase tumor metastasis (5, 6). Therefore, elucidation of the mechanisms regulating AKT activation, especially AKT isoform-specific activation, will facilitate therapeutic approaches to targeting AKT signaling.

In the canonical AKT activation model, growth factors or other stimuli activate class I PI3K at the cell membrane to phosphorylate PtdIns(4,5)P2 to form PtdIns(3,4,5)P3 on the inner cell membrane. AKT is then recruited to the cell membrane through interaction between its pleckstrin homology (PH) domain and PtdIns(3,4,5)P3, where AKT is phosphorylated at two critical residues, Thr308/309 in the activation T loop and Ser473/474 in the hydrophobic domain of AKT1/2 (unless designated otherwise, phosphorylation sites are based on the AKT1 amino acid sequence). 3-Phosphoinositide-dependent protein kinase 1 (PDK1) (7) phosphorylates AKT at Thr308, and mTORC2 (8) as well as other potential PDK2 phosphorylate AKT at Ser473 (9). Activated AKT then translocates from the cell membrane to other cell compartments to phosphorylate its downstream substrates transducing membrane signals to appropriate functional outcomes (10). Phosphorylation of Thr308 and Ser473 has been proposed to be required for full activation of AKT kinase activity (11). However, whether Thr308 and Ser473 phosphorylation is sufficient for full activity or the multiple other phosphorylation sites in AKT isoforms are required for processive phosphorylation or modulate the stability, substrate access or activity of AKT has not been elucidated (12-14). Further selective phosphorylation of Thr308 and Ser473 alters the substrate selectivity of AKT. Thus an improved understanding of the role of the multiple phosphorylation sites in AKT is required to fully elicit the functional regulation of AKT.

In addition to Thr308 and Ser473, currently 20 other residues of AKT1 have been experimentally validated as sites for phosphorylation using mass spectrometry or site specific approaches, including 8 serine residues (122, 124, 126, 129, 137, 246, 475, 477), 7 threonine residues (34, 72, 146, 305, 312, 450, 479), and 5 tyrosine residues (176, 315, 326, 437, 474) (http://www.phosphosite.org, (15)(12-14). Similarly, AKT2 and AKT3 have 22 and 18 validated phosphorylation sites, respectively. Additional AKT isoform specific phosphorylation sites may remain to be identified. The regulation and importance of phosphorylation of sites other than Thr308 and Ser473 is only beginning to be elucidated. For example, phosphorylation at Thr305, Thr312, and Tyr474 was shown to contribute to optimal AKT activation. Thr72 and Ser246 have been proposed to be autophosphorylated in trans, whereas Thr34, Thr450, and Tyr176 phosphorylation appears to be mediated by upstream kinases including atypical protein kinase C, c-Jun N-terminal kinases, and Ack1 (12-14). However, the regulation and function of phosphorylation sites other than Thr308 and Ser473 has not been well studied because of the lack of reagents able to assess site-specific phosphorylation. The gap in knowledge is particularly acute in terms of relative roles of coordinate phosphorylation of sites on individual AKT1, AKT2 and AKT3 molecules.

A newly emerging technology, nanofluidic proteomic immunoassay (NIA), has the potential to characterize protein phosphorylation of multiple different sites and, in particular, coordinate phosphorylation of single AKT molecules including segregating effects on different AKT isoforms (16). NIA combines isoelectric focusing of proteins with sensitive chemiluminescence detection with highly specific antibodies, providing a sensitive and quantitative approach to analyzing protein phosphorylation (16). The NIA approach demonstrated marked differences in coordinate phosphorylation of multiple sites in AKT1 and AKT2 under both basal and stimulated conditions.

Results

NIA can characterize coordinate phosphorylation on individual AKT molecules We used wild type (wt) and AKT2−/− and AKT1−/− knockout HCT116 colon cancer cells (17) to characterize AKT isoform-specific phosphorylation. HCT116 cells express AKT1 and AKT2 in the absence of AKT3 (17). We tested several phospho-specific, isoform-specific, and total AKT antibodies and identified a panel of antibodies applicable for characterization of AKT residue- and isoform-specific phosphorylation (see methods). The AKT1 specific antibody used did not cross-react with AKT2 in AKT1−/− cells, and the AKT2 specific antibody used did not cross react with AKT1 in AKT2−/− cells, validating the specificity of the antibodies (Fig. 1A). In contrast a pan (total)-AKT antibody recognized both AKT isoforms with AKT1 being the predominant AKT isoform. AKT2 was present at about 1/3 the level of AKT1 (Fig. 1A), but was readily detectable by the pan AKT and AKT2 specific antibodies.

Fig. 1. Characterization of phosphorylation of AKT isoforms by Western blotting and NIA.

(A) AKT isoforms and phosphorylation characterized by Western blotting. Wild type (wt), AKT2−/−, and AKT1−/− HCT116 colon cancer cells were serum starved overnight and treated (or not treated) with insulin (20 μg/ml) for 30 minutes. Cells were lysed in RIPA buffer with protease inhibitors and phosphatase inhibitors. Lysates (50μg/lane) were resolved in 10% SDS PAGE. Antibodies for each blot are listed to the left. GAPDH immunoblotting shows equivalent loading. Scanning densitometric values for Western blots were obtained using the ImageJ software (version 1.46r; National Institutes of Health, Bethesda, MD). The total AKT, pT308, pS473, and pT450 levels were normalized to the loading control and presented as relative conversion to values in parental cells. (B) AKT isoforms characterized by NIA. Wild type (wt), AKT2−/−, and AKT1−/− HCT116 cells were serum starved overnight and lysed in NIA RIPA buffer. Samples were analyzed with anti-total AKT antibody. Protein isoelectronic point (pI) is shown on x-axis and Chemiluminescence on y-axis. Only major AKT peaks are shown for clarity and simplification of interpretation. Experiments were repeated at least 3 times and representative data are presented.

We employed Thr308, Ser473, and Thr450 phosphospecific antibodies to characterize residue-specific phosphorylation of AKT. In HCT116 wt, AKT2−/−, and AKT1−/− cells, Thr308 or Ser473 phosphorylation was undetectable under serum starvation conditions (Fig. 1A). In contrast, AKT Thr450 was constitutively phosphorylated. Insulin treatment for 30 minutes induced substantial increases in AKT1 pThr308 and pSer473, but not pThr450, in AKT2−/− cells, and only modest increases in AKT2 pThr309 and pSer474 in AKT1−/− cells potentially due to lower levels of total AKT2 in the HCT116 cells (Fig. 1A).

In HCT116 wt cells as assessed with a pan (total) AKT antibody in NIA, AKT migrated as eight major peaks (Fig. 1B). The multiple peaks likely represent both AKT1 and AKT2 as well as differences in post translational modification (PTM) of each molecule. Consistent with this contention, in AKT2−/− cells, AKT1 ran as six peaks. In AKT1−/−cells, AKT2 consisted of two dominant and one minor peak (Fig. 1B). The AKT peaks in HCT116 wt could be reconstituted, both in pattern and relative magnitude, by combining peaks from AKT2−/− and AKT1−/− cells, validating the use of knockout cells to characterize AKT isoform-specific phosphorylation. The AKT1 and AKT2 isoform-specific antibodies also recognized the peaks predicted by the studies of HCT116 knockout cells (Fig. S1A and S1B). The only overlap between the AKT1 and AKT2 pI pattern was a peak located at P5.75.

Based on the knockout cell lines and isoform-specific antibody data, it was possible to unambiguously assign the identity of each pI peak identified with the pan (total) AKT antibody to a specific AKT isoform with the exception of the P5.75 peak, which is present in both AKT1 and AKT2. The ability to assess both AKT isoforms with a single antibody greatly aids in quantification of relative amounts of each AKT1 or AKT2 peak. Therefore, unless otherwise indicated, we utilized the pan (total)-AKT antibody to establish relative changes in AKT1 and AKT2 migration.

Phosphorylation determines AKT pI values As 22 different AKT1 phosphorylation sites have been independently identified by either mass spectrometry (19 out of 22) (http://www.phosphosite.org) or alternative approaches (13 out of 22) (15) and each of first 10 phosphorylation events will change pI with an expected change of 0.04-0.07 pH units (http://web.expasy.org/compute_pi/, (18) due to the charge added by phosphate, migration of specific isoforms of AKT on NIA should reflect the effects of phosphorylation of single and multiple sites on a single AKT molecule. As AKT2 had a simpler pattern of peaks than AKT1, we initially used AKT1−/− cells to characterize AKT2 isoform-specific PTM. Phosphatase treatment shifted AKT2 into a single P6.03 peak, indicating a pI value of unphosphorylated AKT2 of 6.03, which is compatible with the predicted pI value (http://web.expasy.org/compute_pi/) (18). This suggested that under the conditions assessed, phosphorylation is the dominant PTM in AKT2 explaining the P5.88 and P5.75 peaks observed under serum starvation conditions. Insulin treatment for 30 minutes shifted a fraction of AKT2 into 3 previously undetected peaks of lower pI of P5.62, P5.53, and P5.42 indicative of additional modifications of single AKT2 molecules (Fig. 2A). Phosphatase sensitivity once again indicated that the peak shifts induced by insulin were due to phosphorylation (Fig. 2A). The overlap of peaks under basal and insulin stimulated conditions supported the concept of discrete phosphorylation events accounting for the multiple AKT2 peaks observed under both basal and insulin stimulated conditions (Fig. 2B).

Fig. 2. Characterization of AKT2 and AKT1 posttranslational modification (PTM) by NIA.

(A) pI values of unphosphorylated and phosphorylated AKT2. AKT1−/−HCT116 cells were serum starved overnight and treated (or not treated) with insulin (20 μg/ml) for 30 minutes. Cells were lysed in RIPA buffer then treated (or not treated) with λ phosphatase. Samples were analyzed with anti-total AKT antibody. Experiments were repeated at least 3 times and representative data are presented. (B) Overlap of AKT2 peaks at basal and insulin stimulated conditions. (C) pI values of unphosphorylated and phosphorylated AKT1. AKT2−/− HCT116 cells were prepared and analyzed as in (A). Experiments were repeated at least 3 times and representative data are presented. (D) Overlap of AKT1 peaks at basal and insulin stimulated conditions. (E) Phosphorylation of AKT1 determines the discrete peaks in NIA. AKT1−/− AKT2−/− double knockout HCT116 cells were infected to stably express AKT1 wt and indicated mutations. Cells were serum starved overnight with/without insulin stimulation for 30 min and lysed in RIPA. Samples were analyzed with anti-total AKT antibody. Only major AKT peaks are shown for clarity and simplification of interpretation. (F) Effects of AKT1 mutations on cell invasion. AKT1−/− AKT2−/− double knockout HCT116 cells were infected to stably express AKT1 wt and indicated mutations. Cell invasion was measured by BD Matrigel Invasion Chamber according to manufacturer's protocol. Representing images are shown. Image quantifications were obtained using the ImageJ software (version 1.46r; National Institutes of Health, Bethesda, MD). Invasion levels were presented as relative conversion to values in HCT116 double knockout cells. The error bars represent standard deviations of three repeats.

In AKT2−/− cells, under serum starvation conditions AKT1 was present in 4 major peaks, P5.29, P5.40, P5.50, and P5.60 and 5 minor peaks, P5.22, P5.35, P5.44, P5.53, and P5.75, (Figs. 1B and 2C). Phosphatase treatment shifted the majority of AKT1 into a single peak at P5.75 (Fig. 2C), indicating that the P5.75 peak in resting cells represents unphosphorylated AKT1, which is consistent with the predicted pI (http://web.expasy.org/compute_pi/) (18). Two minor peaks P5.66 and P5.60 remained after phosphatase treatment, potentially representing phosphorylation events resistant to the phosphatase conditions used or alternatively other PTM such as ubiquitination (19) or acetylation (20). It may also represent splicing variants or degradation of AKT1. Insulin treatment increased the number of AKT1 peaks from 9 to 15 (Fig. 2C). Again, phosphatase treatment shifted AKT1 into the major P5.75 peak and the two minor peaks noted above, consistent with phosphorylation being the primary cause of the peak shift induced by insulin. Furthermore, there was an overlap of the peaks identified under basal and insulin stimulated conditions (Fig. 2D) compatible with discrete phosphorylation events on single AKT1 molecules explaining the mobility shift.

Formation of protein complexes could alter the migration of AKT on NIA. To assess this possibility, we determined the motility of AKT1 and AKT2 in the presence of 2M urea and 40mM DTT treatment (Fig. S2) to disrupt protein complexes. 2M urea and 40mM DTT did not alter the motility of AKT1 or AKT2 consistent with the contention that the motility shift of AKT is due primarily to post translational modification.

To identify phosphorylation events contributing to the discrete peaks of AKT1 observed in NIA, we created non-phophorylatable mutants of Thr308 (T308A), Ser473 (S473A) Thr450 (T450A) and Ser124 (S124A). Wild type and mutants were stably expressed in AKT1−/− AKT2−/− HCT116 cells and detected with total AKT antibody (Fig. 2E). As predicted, the T308A and S473A mutations did not significantly alter the migration of AKT1 under basal conditions (Fig. 2E). In contrast, the T450A and S124A mutations shifted several low AKT1 pI peaks to a higher pI (Fig. 2E). The shift to higher pI seen with S124A AKT1 was more pronounced than would be predicted by loss of a single phosphorylation site. This could be explained by the S124A mutation causing a structural change that altered phosphorylation of other sites in AKT1 or alternatively by Ser124 phosphorylation being required for processive phosphorylation of additional sites in AKT1.

Under insulin stimulation conditions, there was a significantly reduced peak shift with T308A compared to wild type suggesting that phosphorylation of Thr308 contributes to the peak shift. Unexpectedly, insulin stimulation failed to induce a detectable shift in S473A mobility suggesting that phosphorylation of Thr308 and other sites induced by insulin is dependent, to a major degree, on Ser473 phosphorylation. With T450A and S124A, peaks still shifted to lower pI indicating these sites are not required for phosphorylation on other sites in AKT1.

The low protein level of T450A (Fig. S3C) is consistent with this site being required for AKT stability (21, 22). In contrast, S124A, T308A and S473A were expressed to comparable levels to wild type AKT1 (Fig. S3C). Interestingly, each of the mutant forms of AKT1 was downregulated more efficiently than wild type AKT1 on insulin treatment (Fig. S3C). Under basal conditions, phosphorylation of PRAS40, an AKT substrate, was decreased by S124, T308 and S473 mutations. However, most strikingly, the T308 but not other mutations essentially abrogated insulin induced PRAS40 phosphorylation implicating T308 in the ability of AKT1 to phosphorylate PRAS40 (Fig. S3C). We further examined effects of the mutations on cell invasion as an indication of functional outcomes of altered signaling. Stable expression of AKT1-wt in HCT116 AKT1−/−AKT2−/− cells dramatically increased invasion (Fig. 2F). S124A and T450A were modestly less able to induce cell invasion compared with AKT1-wt (Fig. 2F), consistent with S124 phosphorylation being required for optimal cell invasion. As T450A levels are lower due to decreased stability (Fig. S3C), the decrease in invasion induced by the T450A may be secondary to changes in stability rather than to a functional change in the AKT1 molecule. In contrast, T308A and S473A constructs induced only a modest increase in invasion (Fig. 2F), indicating that phosphorylation of both T308 and S473 are critical for cell invasion. Expression of AKT1-wt or mutants did not alter the proliferation of HCT116 double knockout cells in normal culture conditions indicating that AKT1 is not required for proliferation of HCT116 cells at least under the conditions assessed (Fig. S3D).

AKT1 residue-specific phosphorylation at Thr308, Ser473, and Thr450 We then examined the localization of different phosphorylated residues in particular pI peak using phosphospecific antibodies. Phosphorylation at Thr450 in the turn motif has been proposed to occur as a cotranslational event required for optimal protein folding and protein stability (21, 22) and thus for optimal AKT activity (11). Based on this concept, Thr450 phosphorylation would be expected on the majority if not all AKT molecules. As predicted, the pattern of peaks detected with the pThr450 specific antibody (Fig. 3A Arrow heads designate potential non-specific peaks not seen in other assays) overlapped to a high degree with peaks detected with the total AKT antibody in the presence or absence of insulin (See Figs 2C, 2D). The only exception is the minor P5.75 peak that likely represents unphosphorylated AKT1 that is not recognized by the pThr450 antibody (Fig. 3A see arrow).

Fig. 3. AKT1 phosphorylation at Thr450, Thr308, and S473.

(A, B, and C) AKT1 phosphorylation at Thr450, Thr308, and S473. AKT2−/− HCT116 cells were serum starved overnight and treated (or not treated) with insulin (20 μg/ml) for 30 minutes. Samples were analyzed with anti-pThr450 (A), anti-pThr308 (B), and anti-pSer473 (C) antibodies. (D) Overlap of AKT1 pThr308 and pSer473 peaks with insulin stimulation (E) Overlap of AKT1 pThr308 and pThr450 peaks with insulin stimulation.

Phosphorylation at Thr308 on the T-loop and Ser473 in the hydrophobic motif is required for optimal AKT activation (11). In AKT2−/− cells, AKT1 Thr308 or Ser473 phosphorylation was not detected under basal conditions indicating that the discrete pI peaks under basal conditions are due to phosphorylation of sites other than Thr308 or Ser473 (Fig. 3B and 3C). As expected, insulin induced a marked increase of the AKT1 Thr308 and Ser473 phosphorylation signals (Fig. 3B and 3C). pThr308 and pSer473 phosphorylation demonstrated a high degree of overlap (Fig. 3D) indicating that in most cases the two residues were coordinately phosphorylated, which is consistent with previous findings (11). However, several minor pThr308 peaks did not overlap with pSer473 peaks (Fig. 3D, see arrows) indicating pThr308 and pSer473 phosphorylation was uncoupled in a fraction of AKT1 molecules. In contrast, as predicted by Thr450 being phosphorylated in most AKT1 molecules, pThr308 only identified a subset of the peaks detected with the pThr450 antibody (Fig. 3E). Overlay of peaks identified by the total AKT1 and the pThr308 and pSer473 specific antibodies showed that the majority of pThr308 (∼70%) and pSer473 (∼67%) was in peaks with pI value of 5.20 or lower (Compare Figs. 3B and C with 2C and 2D). Comparing total and phosphospecific peaks, the AKT1 P5.03 peak accounted for 7.94±1.14% of total AKT1, 16.8±2.6% of total pThr308 and 15.4±3.2% of total pSer473 (% data represent the mean ± SD of 3 independent experiments). Assuming that the p5.03 peak is quantitatively phosphorylated at Thr308 and Ser473, AKT1 was extensively phosphorylated at Thr308 (47%) and Ser473 (52%) after insulin treatment (Fig. 3B and 3C).

We further examined the effects of T308A, S473A, T450A, and S124A mutations in AKT1 on insulin-induced phosphorylation of Thr308 and Ser473. Interestingly, while T308A mutation abolished pThr308 and significantly decreased pSer473, S473A mutation abolished both pSer473 and pThr308, indicating Thr308 phosphorylation is completely dependent on pSer473 phosphorylation. Thr308 and Ser473 were phosphorylated in T450A and S124A constructs indicating that Thr308 and Ser473 phosphorylation is not dependent on Thr450 or Ser124 phosphorylation (Fig. S3). However, low pI peaks were shifted to a higher pI indicative the loss of pI shift due to phosphates on Thr450 or Ser124.

AKT2 residue-specific phosphorylation at Thr309 and Ser474 In AKT1−/− cells, AKT2 phosphorylation at Thr309 or Ser474 was not detected under basal conditions. With insulin treatment, pThr309 was detected at low levels in both AKT1−/− and HCT116 wt cells (Fig. 4A). The two minor pThr309 peaks present in AKT1−/− cells were detected in HCT116 wt cells but not AKT2−/− cells indicating that they likely represent AKT2 phosphosites (Fig. 4A). If the phosphospecific antibody has similar affinity for both AKT isoforms, AKT2 pThr309 only accounts for 3.8% while AKT1 pThr308 accounting for 96.2% of total T-loop pThr308/309 phosphorylation in HCT116 wt cells (See Table S1 for the identification and quantification of each peak inFig 4A).

Fig. 4. T-loop and hydrophobic domain phosphorylation on AKT1 and AKT2 with insulin stimulation.

(A and B) Comparison of AKT2 pThr309 and pSer474 with AKT1 pThr308 and pSer473. AKT1−/−, AKT2−/−, and HCT116 wt cells were serum starved overnight and treated with insulin (20 μg/ml) for 30 minutes. Samples were analyzed with anti-pThr308 (309)(A) and anti-pSer473 (474)(B) antibodies. Experiments were repeated at least 3 times and representative data are presented.

Similar to AKT2 Thr309 phosphorylation, insulin only induced modest changes in AKT2 pSer474 in AKT1−/− cells (Fig. 4B). Three pSer474 containing peaks at P5.42, P5.53, and P5.70 detected in AKT1−/− cells were also present in HCT116 wt cells but absent from AKT2−/− cells. AKT2 pSer474 only accounted for 7.5% of total hydrophobic pSer473/474 phosphorylation in HCT116 wt cells (See Table S2 for the identification and quantification of each peak inFig 4B) suggesting that AKT1 accounts for 92.5% of Ser473/474 phosphorylation. In contrast to AKT1, AKT2 phosphorylation at pThr309 and pSer474 were largely uncoupled. For example, the AKT2 P5.77 peak contained pThr309 but not pSer474, and the AKT2 P5.53 and P5.70 peaks contained pSer474 but not pThr309. Only AKT2 in the minor P5.42 peak likely contained both pThr309 and pSer474 phosphorylation (First panels in Fig. 4A and 4B). Whether differential phosphorylation on Thr308 and Ser473 between AKT1 and AKT2 is cell line or stimulus specific or kinetic dependent warrants further studies.

If the pan (total) AKT antibody has similar affinities for both AKT isoforms, AKT2 represents 26% of total AKT (Fig. 1B) (See Table S3 for the identification and quantification of each peak in Fig 1B), but only accounted for much smaller percentage of T-loop or hydrophobic domain phosphorylation (3.8% or 7.5%, respectively). Thus insulin mainly induces AKT1 but not AKT2 phosphorylation and activation in HCT116 wt cells.

Effect of PI3K inhibition on AKT isoform-specific phosphorylation The PI3K pathway is a rich therapeutic target across multiple lineages with several PI3K inhibitors in clinical trials. However, AKT isoform-specific response to these inhibitors has not been elucidated. Pretreatment with GDC0941 (PI3K and mTORC1 dual inhibitor with preference for PI3K) for 3 hours, but not Rapamycin (mTORC1 allosteric inhibitor), abolished insulin-induced AKT phosphorylation on all peaks containing Thr308 and Ser473 in HCT116 wt cells (Fig. 5A and 5B). Thus PI3K activity is necessary for insulin-induced phosphorylation of all peaks containing Thr308 and Ser473 in HCT116 wt cells.

Fig. 5. Effect of PI3K inhibition on AKT isoform-specific phosphorylation.

(A and B) Effect of PI3K inhibition on AKT phosphorylation at Thr308 or Ser473. Wild type HCT116 cells were serum starved overnight and treated with DMSO control or indicated inhibitors (10 μg/ml) for 3 hours, then treated with insulin (20 μg/ml) for 30 minutes. Samples were analyzed with anti-pThr308 (A) and anti-pSer473 (B) antibodies. (C and D) Effect of PI3K inhibition on overall AKT phosphorylation. Wild type HCT116 cells were serum starved overnight and treated with DMSO control or indicated inhibitors (10 μg/ml) for 3 hours, then treated with insulin (20 μg/ml) for 30 minutes (C) or without insulin treatment (D). Samples were analyzed with anti-total AKT antibody. Experiments were repeated at least 3 times and representative data are presented. (E) Effect of PI3K inhibition on AKT phosphorylation characterized by Western blotting. Samples from (A) were analyzed by Western blotting. Antibodies for each blot are listed to the left. GAPDH immunoblotting shows equivalent loading.

Treatment with GDC0941 completely blocked peak shifts of AKT induced by insulin as assessed with pan-AKT antibody (compare Fig. 5C with Fig. 1B), indicating that inhibition of PI3K blocked all insulin-induced phosphorylation independent of which sites were responsible for the mobility shift. Rapamycin did not alter the effect of insulin on mobility of AKT pI peaks (Fig. 5 and Fig. S4B, C).

Intriguingly, GDC0941 did not alter the pattern of basal AKT1 peaks, indicating that basal AKT1 phosphorylation is not dependent on ongoing PI3K activity (Fig. 5D). In contrast, GDC0941 markedly altered the pattern of basal AKT2 phosphorylation resulting in a decrease in the p5.88 peak and an increase in the P6.03 (unphosphorylated AKT2) peak, consistent with basal AKT2 phosphorylation being dependent on ongoing PI3K activity (Fig. 5D). Indeed using a pan AKT antibody to directly compare AKT1 and AKT2, GDC0941 induced a dose-dependent increase in the P6.03 unphosphorylated AKT2 peak without significantly altering the AKT1 peaks present (Fig. S4A). The effect of GDC0941 on basal AKT2 phosphorylation was confirmed using an AKT2 specific antibody (Fig. S4B). Consistent with these results, GDC0941 markedly decreased pThr451 in AKT2 but not pThr450 in AKT1 (Fig. S4B and S4C). Indeed, under basal conditions with GDC0941 treatment for 3 hr, AKT2 pThr451 decreased by at least 50% (Fig. S4C). The mechanism by which GDC0941 selectively decreases Thr451 phosphorylation in AKT2 remains to be determined.

As assessed by Western blotting, GDC0941, but not rapamycin, abolished AKT pThr308 and pSer473 and modestly decreased pThr450 consistent with the NIA results (Fig. 5E).

AKT isoform-specific phosphorylation in other cell lines We examined AKT phosphorylation profiles in multiple cell lines with different genetic backgrounds and cell lineages to determine the generality of phosphorylation patterns. In H358 lung cancer cells, the AKT peaks were similar to those in HCT116 wt cells under both serum starvation and insulin stimulation conditions consistent with similar regulatory mechanisms (Fig. 6A and Fig. S5AB). However, in the A549 lung cancer cell line, although the pattern of peaks was similar, the AKT2 P5.88 peak was significantly higher than the AKT1 peaks consistent with higher AKT2 levels in A549 cells (Fig. 6B and Fig. S5I).

Fig. 6. AKT isoform-specific phosphorylation in other cell lines.

H358 (A) and A549 (B) lung cancer cells, HEK293 embryonic kidney cells (C), and SKOV3 ovarian cancer cells (D) were serum starved overnight and treated (or not treated) with insulin (20 μg/ml) for 30 minutes. Samples were analyzed with anti-total AKT antibody.

In HEK293 cells (Fig. 6C and Fig. S5EF) and SKOV3 cells (Fig. 6D and Fig. S5GH), multiple sites were phosphorylated, as indicated by low pI peaks under serum starvation conditions. Insulin did not alter the location of AKT peaks in SKOV3 cells (Fig. 6D and Fig. S5GH), potentially due to high basal AKT phosphorylation likely as a consequence of HER2 amplification and the presence of a PIK3CA mutation in this line.

The AKT2 P5.70 peak was detected by pSer473/474 but not pThr308/309 antibodies in H358, A549, HEK293 and SKOV3 (Fig. S5A-H), verifying uncoupled AKT2 phosphorylation on Ser474 and Thr309 is generalizable across cell lines and cell lineages.

Discussion

NIA provides a rapid, quantitative, sensitive and specific approach to characterize PTM of proteins in tissue culture and patient samples (16). The approach is fast, with the ability to analyze dozens of samples with multiple antibodies within 3-4 hours. Protein can be detected at picogram levels (16) allowing analysis of PTM changes in samples where limited protein is available. The predicted pI values of 5.75 and 5.96 for unphosphorylated AKT1 and AKT2 (http://web.expasy.org/compute_pi/) (18) are suitable for the isoelectric focusing approach for NIA. In addition to exploring underlying mechanisms regulating AKT phosphorylation, NIA could be used to monitor AKT isoform-specific activation and inhibition in samples from patients undergoing therapies targeting the AKT pathway.

NIA is dependent on both the quality and validation of available antibodies. In order to be fully quantitative, different isoforms and PTM must be identified with an antibody with equal affinity for each isoform analyzed and also where the affinity is not altered by PTM. The identification of peaks representing specific isoforms and PTMs can also prove problematic. The approach herein of using knockout isogenic cells, mutant constructs and multiple isoform and phosphospecific antibodies can alleviate these challenges. Indeed, the validated total, phosphospecific, and isoform-specific antibodies demonstrate a marked concordance in terms of the relative amounts of AKT1 and AKT2 and of the differentially phosphorylated molecules in parental and knockout cells.

In this study, we developed and implemented an approach based on NIA technology to characterize total and residue-specific phosphorylation profiles of specific AKT1 and AKT2 molecules. By combining data from both basal and insulin-treated conditions, we were able to identify at least 12 and as many as 16 independent AKT1 peaks and at least 5 and up to 6 independent AKT2 peaks, indicative of complex patterns of phosphorylation of different residues on individual AKT molecules. To a major degree using knockout cells, mutant constructs and phosphospecific antibodies, we were able to define which AKT isoforms and which phosphorylation sites are found in a particular peak. For example, under serum starvation conditions (Fig. 1B), the P6.03 peak contains AKT2 with no phosphorylation; the P5.88 peak contains AKT2 with pThr451; the P5.75 peak contains AKT2 with pThr451 and AKT1 with no phosphorylation. Under basal conditions, the lower pI peaks consist of AKT1 with pThr450 phosphorylation without pThr308 or pSer473 phosphorylation sites (Table S4). Identification of the phosphorylation sites constituting the peaks present following insulin stimulation is more complex due to the number of peaks, however, the combination of mutant constructs and phosphospecific antibodies indicate that pThr450/451 is present in all peaks and that pThr308 and pSer473 contribute to the mobility shifts following insulin treatment resulting in accumulation of AKT forms with low pI and complex phosphorylation patterns (Table S5).

We identified unexpected isoform-specific regulation of AKT phosphorylation by insulin stimulation as well as by PI3K inhibition, which had not been appreciated using other approaches. In particular, under basal conditions AKT1 molecules were phosphorylated on multiple sites, but not on the two classical Thr308 and Ser473 sites. Insulin treatment markedly increased AKT1 phosphorylation and in particular increased the complexity of phosphorylation on each AKT1 molecule. Phosphorylation of both Thr308 and Ser473 as well as all other possible residues phosphorylated in response to insulin were dependent on PI3K activity. AKT2 phosphorylation was much less responsive to insulin treatment than AKT1. Insulin markedly increased phosphorylation of Thr308 or Ser473 in AKT1 and to a lesser degree AKT2. In AKT1, pThr308 and pSer473, in most cases, occurred on the same molecule as previously reported (11). However, there was a subpopulation of AKT1 molecules with only pThr308 or pSer473. Intriguingly, Thr308 or Ser473 phosphorylation in AKT2 frequently occurred on independent AKT2 molecules. The specific AKT1 and AKT2 phosphorylation profiles may contribute to their different substrate specificities. For example, GSKβ, p21cip, SKP2, and palladin are preferentially phosphorylated by AKT1, whereas GSKα, MDM2, and AS160 appear to be targeted by AKT2 (4, 23-26). The site-specific AKT phosphorylation profiles also contribute to substrate specificities. AKT with Thr308 but not Ser473 phosphorylation effectively phosphorylates GSK3 and TSC2 but not FOXO1/FOXO3a (27). Our data show Thr308 phosphorylation is, to a major degree, dependent on Ser473 phosphorylation. Previous studies (11, 27) suggested that phosphorylation of these two sites could be independent under certain conditions. However, in contrast to transient expression of tagged AKT in these studies (11), we expressed an untagged AKT by stable infection, to levels similar to those in wild type cells, in AKT knockout cells, which is more likely to reflect physiological AKT phosphorylation processes.

Basal AKT2 but not AKT1 phosphorylation was sensitive to PI3K inhibition. Using reagents validated with AKT1 and AKT2 knockouts, PI3K inhibition decreased pThr451 on AKT2 but not pThr450 on AKT1. Whether this represents differential activity of isoform-specific pThr450 kinases or phosphatases needs to be determined.

In summary, we developed and implemented an approach based on NIA technology for AKT isoform-specific phosphorylation analysis and revealed unexpected and important aspects of regulation of AKT isoform phosphorylation, which has been undetectable by other methods, like mass spectrometry and western blotting. The differential AKT1 and AKT2 phosphorylation under basal and stimulated conditions as well as the differential sensitivity to inhibitors suggests different mechanisms regulating phosphorylation and may contribute to the different functions of AKT1 and AKT2. The approach and the results from this study have the potential to advance our understanding of AKT isoform-specific activation, which will aid in implementation of approaches to target aberrant AKT activation in cancer cells.

Materials and Methods

Cell Culture

The HeLa cell line was a Tet-on derivative obtained from BD Clontech (Palo Alto, CA). The wild type, AKT1−/−, and AKT2−/− HCT116 colon cancer cell lines were gifts from Dr. Bert Vogelstein (Johns Hopkins University). The lung cancer cell lines NCI-H358 and A549, the ovarian cancer cell line SKOV3, and HEK293 were obtained from the American Type Culture Collection (Manassas, VA). Cells were cultured in DMEM (Invitrogen, Carlsbad, CA; HeLa and HEK293 cells), McCoy's 5A (Gibco; HCT116 wild type and knockout cells), or RPMI 1640 medium (Invitrogen; NCI-H358, A549, and SKOV3 cells) supplemented with 10% (v/v) fetal calf serum (Gibco). Cell line identify was routinely confirmed by STR profiling in the MDACC CCSG core.

Reagents and Antibodies

Rapamycin was from Cell Signaling Technology (Beverly, MA). The PI3K and mTORC1 inhibitor GDC-0941 was from Axon Medchem (Groningen, The Netherlands). Antibodies against pan (total) AKT (CS-4691), phospho-AKT Thr308 (CS-2965), phospho-AKT Thr450 (CS-9267), phospho-AKT Ser473 (CS-9271), and AKT2 (CS-3063) were from Cell Signaling Technology. AKT1 antibody (BD-610861) was from BD Biosciences (San Jose, CA). An anti-GAPDH antibody (AM4300) was obtained from Ambion (Austin, TX).

Constructs

Full-length cDNA encoding AKT1 (NM_001014431.1) was from Invitrogen. The indicated mutations were introduced using standard site-directed mutagenesis and verified by sequencing. cDNAs were cloned into a Gateway compatible pBabe-puro retroviral vector (Addgene #1764) by Gateway recombination according to the manufacturer's protocol (Invitrogen).

Lysate preparation

To prepare cell lysates for Western blotting, cells were lysed in RIPA buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with protease inhibitor and phosphatase inhibitor cocktails (Pierce, Rockford, IL). Western blotting was performed as described previously (28). To prepare cell lysates for NIA, cells were lysed in NIA specific RIPA lysis buffer (ProteinSimple Inc., CA) (20 mM HEPES, pH 7.5, 150 mM NaCl, 1 % NP40. Alternatively, 0.25 % sodium deoxycholate) supplemented with protease inhibitor and phosphatase inhibitor cocktails (ProteinSimple Inc.). To remove phosphorylation from proteins, lysates were treated with λ phosphatase (Millipore, MA) according to the manufacturer's protocol.

NIA

Lysates were analyzed using a NanoPro 1000 system with an optimized protocol. The ampholyte premix was G2 5-8 (ProteinSimple Inc. #040-973) nested with G2 5-6 (ProteinSimple Inc. #040-971) at a ratio of 1:1. Samples were loaded at a concentration of 200 μg/ml in capillaries. Focusing conditions were 40,000 μW for 40 minutes. UV Immobilization time was 60 seconds with an exposure time of 240 s. Primary antibodies were used at 1:50 dilution for 2 h. Anti-rabbit (ProteinSimple Inc. 040-656) and anti-mouse (ProteinSimple Inc. # 040-655) secondary antibodies were used at 1:100 dilution for 1 h. Tertiary antibody (ProteinSimple Inc. #041-106) was used at 1:100 dilution for 10 min. The pI standards for 4.9, 5.5, 6, 6.4, and 7.0 (ProteinSimple Inc.) were used based on predicted pI of AKT isoforms.

Supplementary Material

Fig. S1. Characterization of phosphorylation of AKT isoforms by NIA with isoform-specific antibodies. Wild type (wt), AKT2−/−, and AKT1−/− HCT116 cells were serum starved overnight and lysed in NIA RIPA buffer. Samples were analyzed with anti-AKT1 isoform-specific (A), and anti-AKT2 isoform-specific (B) antibodies. Protein isoelectronic point (pI) is shown on x-axis and Chemilumnescence on y-axis. Experiments were repeated at least 3 times and representative data are presented.

Fig. S2 Urea and DTT treatments have no effect on the patterns of AKT1 and AKT2 peaks in NIA. AKT2−/− and AKT1−/− HCT116 colon cancer cells were serum starved overnight and lysed in NIA RIPA buffer then treated (or not treated) with urea (1 or 2 M) with 40mM DTT for 30 minutes. Samples were analyzed with total AKT antibody. Experiments were repeated at least 3 times and representative data are presented.

Fig. S3. Effects of AKT1 mutations on Thr308 and Ser473 phosphorylation, AKT activation, and cell proliferation. AKT1−/− AKT2−/− double knockout (DKO) HCT116 cells were infected to stably express AKT1 wt and indicated mutations. Cells were serum starved overnight with/without insulin stimulation for 30 min and lysed in RIPA buffer. Samples were analyzed by NIA with anti-pThr308 (A) or anti-pSer473 (B) antibodies, or by western (C) with indicated antibodies. Scanning densitometric values for Western blots were obtained using the ImageJ software (version 1.46r; National Institutes of Health, Bethesda, MD). The total AKT, pS473, pT308, pGSK3, and pPRAS40 levels were normalized to the loading control and presented as relative conversion to values in AKT1 wt cells with insulin stimulation. Cell proliferation of indicated cell lines was measured by CellTiter-Blue assay (Promega) (D).

Fig. S4. Effect of PI3K inhibition on overall AKT phosphorylation and pThr450 (451). (A) Wild type HCT116 cells were serum starved overnight and treated with DMSO control or GDC0941 (1nM, 100nM, and 1 μM) for 3 hours. Samples were analyzed with anti-total AKT antibody. (B and C) Wild type HCT116 cells were serum starved overnight and treated with DMSO control, GDC0941 (10 μM), or rapamycin (10 μM). Samples were analyzed with anti-AKT2 (B) or anti-pThr450 (451) antibodies (C).

Fig. S5. AKT Thr308 and Ser473 phosphorylation in other cell lines. H358 (A and B) and A549 (C and D) lung cancer cells, HEK293 embryonic kidney cells (E and F), and SKOV3 ovarian cancer cells (G and H) were serum starved overnight and treated (or not treated) with insulin (20 μg/ml) for 30 minutes. Samples were analyzed with anti-pThr308, anti-pSer473, or anti-AKT2 antibodies. (I) Selected samples were analyzed with anti-AKT2 antibody.

Table S1. pThr308 versus pThr309 in HCT116 wt cells

Note: Calculation based on Fig. 4A.

Table S2. pThr473 versus pThr474 in HCT116 wt cells

Note: Calculation based on Fig. 4B.

Table S3. AKT1 versus AKT2 in HCT116 wt cells

Note: Calculation based on Fig. 1B.

Table S4. Composition of major peaks in HCT116 wt cells under serum starvation conditions

Note: Summary based on Fig. 1B.

Table S5. Composition of major peaks in HCT116 wt cells under insulin stimulation conditions

Note: Summary based on Fig. 2A and 2C. The minor AKT2 P5.42 peak under insulin stimulation conditions was considered as overlapped with the major AKT1 P5.40 peak.