Class IA PI3K isoforms lead to differential signalling downstream of PKB/Akt

Maintenance and generation of cell lines

Generation of p110α/p110β floxed MEFs and retroviral transductions were performed as described before [4]. MEFs are transduced with retroviral pBABE constructs with puromycin, neomycin and blasticidin resistance. Transduced cells were treated with adenoviruses carrying Cre-recombinase to generate PIK3CA and PIK3CB knock-outs. MEFs depicted as wild-type (WT MEFs) are untreated p110α/p110β floxed cells. All of the MEFs were grown in 8 % FBS 1 % Pen/Strep high glucose DMEM.

2D growth assay

A total of 4,000 cells were seeded to each well of 12-well plates in triplicates. When DMSO control wells are at least 80 % confluent, the plates were fixed with 10 % acetic acid 10 % ethanol in dH₂O. The wells are stained with crystal violet stain. The staining is quantified after destaining them with 100 mL acetic acid, 900 mL dH₂O and transferred to a 96-well plate for measurements. The quantification is performed with a microplate reader at 592 nm.

Inhibitors BYL719, KIN193, CAL101, Rapamycin, Dasatinib and EHT1864 were all from Selleckchem. Cisplatin and doxorubicin were purchased from Koçak Farma and Deva Holding.

Western blot

The cells were scraped on ice, and pellets were lysed with RIPA buffer treated with phosphatase and protease inhibitor cocktails. Nitrocellulose membranes (GE Healthcare Life Sciences) were used. To obtain total protein levels, membranes previously probed with phosphoprotein antibodies were stripped with Millipore Membrane Stripping Solution, and reprobed with total protein antibodies. All antibodies used in the study were purchased from CST.

Cell cycle analysis with PI staining

Tyrpsinized cells were fixed with 70 % ethanol in dH₂O on ice for 3 h. Ethanol was removed with subsequent ice-cold PBS washing steps. The cells were resuspended in 50 µL 100 μg/mL RNaseA (Sigma-Aldrich), and incubated at 37 °C for 15 min. The cells were resuspended in 200 µL 50 μg/mL propidium iodide (Sigma-Aldrich), and incubated in dark for 1 h on ice. The cell clumps were removed by using a strainer and the cells were counted and gated with CytoFlex (Beckman Coulter).

Docking experiments with HADDOCK

The following protein models were used for docking simulations obtained from Protein Data Bank (PDB):

The unknown residues were filled with FASTA sequences from UniProt. Default HADDOCK parameters were used. Active residues were selected using existing literature [13, 14].

For mTORC1 docking experiments, the models were aligned to the crystal structure (6BCU) from reference [15] on PyMol and possible residues that may partake in mTORC1-Rac1-p110α and p110β complexes were identified.

Statistical analysis and band intensity quantifications

The experiments were performed as triplicates and the graphs show mean±standard deviation. The Student’s t-test was used to analyze the significance, and p value less than 0.05 was considered significant. Quantification of band intensities was performed with ImageJ (NIH, Bethesda, MD).

MEFs expressing unique p110 isoforms display distinct levels of Akt1 S473 phosphorylation

To investigate isoform-specific roles of class IA p110 catalytic subunits, we used MEFs whose first exons of p110α and p110β were floxed [4]. We ectopically expressed myristoylation (Myr)-tagged p110α, p110β and p110δ in these MEFs, and using an adenoviral Cre-recombinase; we deleted the endogenous PIK3CA and PIK3CB so that these MEFs would be expressing only the activated exogenous constructs. Wild-type p110α/β floxed MEFs were used as controls. p110α/β flox background is depicted as α, β +/+ throughout the manuscript.

First we checked the levels of the exogenously expressed p110 proteins. p110α, p110β and p110δ (a minor PI3K isoform in MEFs) were all expressed at comparable levels (Figure 1A). To demonstrate the validity of our cellular models, we determined the relative sensitivity of our cell lines to isoform-specific PI3K inhibitors in 2D-growth assays. BYL719 specifically targets p110α, whereas KIN193 and CAL101 inhibit p110β, and p110δ respectively [16, 17]. BYL719 almost completely blocks the growth of myr-p110α MEFs, and KIN193 inhibits myr-p110β MEFs significantly (Figure 1B). In line with previous findings [18], BYL719 at the doses used can also block the growth of myr-p110δ MEFs as well as CAL101. To investigate whether activated p110s coordinate isoform-specific roles in MEFs, we examined the downstream consequences of activated PI3K signalling. PI3K activity leads to Akt phosphorylation on Thr308, and Ser473, leading to its full activation [8]. Thus, Akt Ser473 phosphorylation indicates the activated Akt. We employed WT and myr-p110 MEFs and used a phospho-specific antibody for Akt1 Ser473 to highlight the activation profile of the major AKT isoform in MEFs. We observed different Akt1 Ser473 phosphorylation levels among these cell lines, the highest among them being myr-p110β MEFs (Figure 1C). Interestingly, the phospho-Akt1 level of myr-p110δ MEFs was significantly low, comparable to the WT controls. Then, we asked whether phosphorylated Akt1 effectively relays the signal to the downstream components in these MEFs with antibodies detecting phosphorylated Akt substrates (Figure 1D). Of note, the most abundant phosphorylations were observed on proteins around 250 kDa, 75 kDa, and 37–50 kDa sizes in myr-p110β MEFs. As expected, the phospho-AKT substrate signatures of myr-p110δ and myr-p110α MEFs were not as prominent as their myr-p110β counterparts. Overall, MEFs that express various activated p110s exhibit differential activation of Akt1 and Akt substrate phosphorylation, suggestive of isoform-specific functions.

Figure 1:

Ectopic expression of activated p110β leads to a higher level of Akt activation in comparison to other Class IA PI3Ks in MEFs. (A) Relative protein levels of myristoylated p110s and WT cell lines. Each well contains 40 µg of protein lysate. Samples were probed with the indicated antibodies in an immunoblot. GAPDH is used as loading control. (B) Crystal violet stained plates were read at OD 592 nm to infer cellular growth (n=3). The DMSO controls of each cell line is normalized to 100 %, and the remaining reads adjusted to that. Averages were shown with SEM. Statistical significance was checked with Student’s t-test. *: p≤0.05, **: p≤0.01. (C) Proteins were harvested from cells grown to subconfluency. Each well of SDS – PAGE contains 30 µg protein lysate. Indicated antibodies were used in an immunoblot, GAPDH serves as loading control. Band intensity quantifications of the GAPDH normalized signals are indicated. (D) Proteins were harvested from exponentially growing cells. Each well contains 30 µg protein lysate. The approximated molecular weight values are noted at the right hand side of the blot. In the upper panel, phospho-AKT substrate antibody was used to assess level of AKT activity and GAPDH serves as loading control.

Activated p110δ relays phospho-tyrosine signals through JAK/STAT pathway

As myr-p110δ MEFs did not activate the Akt arm of the signaling as much as their p110β counterparts, we hypothesized that p110δ may be transmitting the signals via the tyrosine phosphorylation cascades [19, 20]. We analyzed the tyrosine phosphorylation signatures of myr-p110 MEFs. We observed two distinctive increments in tyrosine phosphorylation: at 130 kDa and at 60 kDa, in the myr-p110δ and the myr-p110β lanes, respectively (Figure 2A). As p110δ is functional mostly in immunological cells and JAK1/2/3 are around 130 kDa, we suspected that pattern might represent activated JAK/STAT signaling [6, 21]. Thus we investigated the phosphorylation of an essential downstream regulator of JAK/STAT signaling, STAT3 [22] (Figure 2B). STAT3 phosphorylation was highly detectable in myr-p110δ MEFs. Previously, it has been stated that Src becomes phosphorylated on a scaffold in the presence of p110β [13]. We wanted to see if the 60 kDa phosphoprotein could indeed be Src because it is prominent in fibroblasts. To corroborate this, we analyzed phospho-Src amongst control, myr-p110α, and myr-p110β expressing MEFs (Figure 2C). Phospho-Src levels were found to be high in p110β but not in p110α-expressing cells. Since the phosphorylated Src is expected to be activated, we investigated whether these results translated to varying levels of Src dependency in myr-p110α and myr-p110β MEFs in 2D-growth assays with a selective Src inhibitor, Dasatinib (Figure 2D). Almost in all inhibitor concentrations tested, we determined that myr-p110β MEFs are significantly more dependent on Src in comparison to their p110α counterparts. In short, our data thus far supports the notion that active p110β and p110δ induce differential activation of Src and STAT3 respectively.

Figure 2:

p110β and p110δ activate differential tyrosine phosphorylation cascades. (A) Phospho-tyrosine blot of myristoylation constructs. The approximated molecular weight values are noted at the right hand side of the blot. GAPDH is used as loading control. (B) Myr-PI3K MEFs were exponentially grown, 30 µg lysate was used per cell line. Immunoblots were performed with phospho-STAT3, STAT3 and ACTIN antibodies. Band intensity quantifications of p-STAT3 were indicated for each sample. (C) Myr-p110α and myr-p110β MEFs were grown exponentially, 30 µg lysate was used per cell line. Indicated antibodies were used in an immunoblot, GAPDH serves as loading control. (D) 10,000 cells were seeded to six well plate. Indicated doses of Dasatinib treatment were conducted for 5 days. Cells were then fixed and crystal violet stainings were quantified in comparison to DMSO controls (n=3). Percents of growth inhibition were displayed with standard deviation, statistical significance was checked with Student’s t-test. Ns: p>0.05, *: p≤0.05, ***: p≤0.001.

Activated p110β has a scaffold function that involves mTORC1

As the greatest levels of Akt1 Ser473 phosphorylation in MEFs was found in myr-p110β MEFs, we decided to study its potential causes and consequences. To do that, we used myr-p110α and myr-p110β MEFs and inhibited p110s by treating them with their respective inhibitors. Although in both cell lines phospho-S6 is totally eliminated upon inhibitor treatment, Pras40 phosphorylation was more persistently decreased in myr-p110α MEFs upon BYL719 treatment. On the other hand, myr-p110β MEFs treated with KIN193 exhibit sustained mTOR auto-phosphorylation despite the catalytic inhibition (Figure 3A) [23]. These results indicate that activated p110β might induce a differential signaling module that partially spares mTORC1 activity. To investigate this possibility further, we used short term Rapamycin treatment to target mTORC1, and EHT1864 to catalytically inhibit Rac1 (Figure 3B), as Rac1 is involved in activation of p110β [4]. We showed that catalytic inhibition of Rac1 decreases mTOR levels as well as its auto-phosphorylation, and these effects are further exacerbated when EHT1864 is combined with Rapamycin in myr-p110β MEFs. Interestingly, myr-p110α MEFs did not exhibit a similar response under identical circumstances. These results support the notion that inhibition of Rac1 catalytic activity inhibits phosphorylation of mTORC1 in myr-p110β MEFs possibly through a previously unidentified, enzyme independent function of p110β.

Figure 3:

Activated p110α and p110β produce distinct marks for mTOR auto-phosphorylation. (A) Indicated cells were grown to subconfluency and treated with indicated inhibitors at 2 μM concentrations for 2 h before the harvest. Each well contains 40 µg of protein lysate. Immunoblots were performed with the indicated antibodies, GAPDH serves as loading control. (B) Indicated cells were grown to subconfluency and treated with indicated inhibitors for 2 h before the harvest. 100 nM Rapamycin and 5 μM EHT1844 were used either alone or in combination. Each well contains 40 µg of protein lysate. Immunoblots were performed with the indicated antibodies, GAPDH serves as loading control.

Previously, it has been shown that mTORC1 becomes activated by RHEB [15]. As we observed a similar pattern where Rac1 catalytic activity somehow activated mTORC1 where p110β is present, we wanted to investigate the possibility of mTORC1, Rac1, p110β, and p85α icSH2 to form a complex in which mTORC1 becomes activated. Since Rheb and Rac1 are both GTPases, we modeled the interaction based on the crystal structure of mTORC1, whose PDB ID is 6BCU [15, 24]. We discovered putative contact sites by aligning Rac1 to Rheb on 6BCU in PyMOL and utilized them as input for HADDOCK, an online data-driven docking program. The catalytic sites of mTORC1 were inserted as passive residues to prevent proteins from docking to those areas [25, 26]. mTORC1, p110β, and Rac1 formed the best cluster (Figure 4A). We explored if p110α could form a comparable complex. HADDOCK simulations failed to produce an mTORC1–p110α complex (Figure 4B). We conducted a competitive docking experiment where p110α and p110β were instructed to engage with mTORC1 at the same residues to validate our claim. Intriguingly, p110α was not docked to mTORC1, but p110β was successfully shown to interact with mTORC1, where Rac1 plays a bridging role between mTOR and p110β (Figure 4C). The scores of the best clusters of the docking experiments could be found in Table 1. The HADDOCK score is calculated as sum of various molecular interactions, where the negative values indicate the high probability of a given cluster of models [27]. The Z-Score indicates the deviation of the cluster from the average, and better results have more negative Z-Scores. Overall, our findings suggest that Rac1 may also allosterically activate mTORC1 similar to RHEB, in a p110β-containing complex [15].

Figure 4:

p110β but not p110α emerges as an interactor of mTOR in molecular docking simulations. (A) The best docked model generated by HADDOCK. p110β crystalized with p85 α IC SH2 domains, Rac1, and mTORC1 (mTOR, mLST8, RAPTOR) were used as input models for HADDOCK. Blue: mTORC1, green: p110β with p85α, magenta: Rac1. (B) The best docked model generated by HADDOCK. p110α crystalized with p85 α NI SH2 domains, Rac1, and mTORC1 (mTOR, mLST8, RAPTOR) were used as input models for HADDOCK. Blue: mTORC1, green: p110α with p85α, magenta: Rac1. (C) The best docked model generated by HADDOCK. p110β crystalized with p85 α IC SH2 domains, Rac1, mTORC1 (mTOR, mLST8, RAPTOR), and p110α with p85α were used as input models for HADDOCK. Blue: mTORC1, green: p110β with p85 α IC SH2, magenta: Rac1, yellow: p110α with p85α.

p110β expressing MEFs exhibit higher G2/M activity than p110α MEFs

We have shown that myr-p110β expressing MEFs have higher levels of phosphorylated and activated Akt1 (Figure 1C and D). As one of the roles of Akt is to promote cell cycle [28], we wanted to investigate whether cell cycle kinetics differ between p110α and p110β expressing MEFs via FACS analysis. For these sets of experiments, we used α, β +/+ Cre MEFs with p110α and p110β ectopic expression without the myr tag, to investigate the basal cell cycle kinetics. We observed that the p110α and p110β expressing MEFs have distinct cell cycle distributions (Figure 5A and B), with more p110β MEFs collected at the G2/M phase than the p110α MEFs (%42.19 vs. %21.47, respectively). Furthermore, fewer p110β MEFs are found in G1 and S phase, than p110α MEFs (%37.22 and %12.84 vs. %53.97 and %19.68, respectively, Figure 5C). These results indicate that p110β expression promotes a higher mitotic index.

Figure 5:

Cell cycle analyses of double knock-out addback p110α and p110β MEFs. Cells were grown and harvested at 50 % confluency. The same gates were used for both of the cell lines. (A) FACS analysis of cycling DKO addback p110α MEFs. (B) FACS analysis of cycling DKO addback p110β MEFs. (C) Schematic representation of the cell cycle distribution of addback cell lines with mean and SEM (n=2).

Activated p110β expressing MEFs are more sensitive to genotoxic stress

We observed that Akt1 Ser473 and Akt substrate phosphorylation (Figure 1C and D) levels were the highest in myr-p110β MEFs. We wished to study the consequences of the differential in Akt activation between p110 isoforms. As Akt has many effectors responsible for activating the cell cycle [29], we reasoned that myr-p110α and β MEFs may have different growth rates and sensitivities for cytotoxic agents as was suggested by our FACS results (Figure 5A and B). We performed 2D growth assays to understand whether growth rates differ among MEFs expressing activated p110s in comparison to controls (Figure 6A). According to our results, MEFs that express myr-p110β exhibit the fastest proliferation amongst the other lines. Cancer is characterized by uncontrolled cell proliferation as a result of cell cycle dysregulation. In response to genotoxic stress inducers, fast-cycling cells tend to accumulate more damage. We therefore used increasing doses of doxorubicin and cisplatin to examine the sensitivity of activated p110 MEFs in response to genotoxic stress. Myr-p110β MEFs are found to be the most sensitive cell line for genotoxic stress induced by doxorubicin and cisplatin (Figure 6B and C), having the lowest IC₅₀ value among the other activated p110 MEF lines. Due to the fact that myr-p110β MEFs exhibited the highest phosphorylation level of Akt1 Ser473 among the activated p110 MEFs, we hypothesized that in these MEFs, Akt effectors that regulate cell cycle are upregulated, and as these cells cycle more frequently, they are more susceptible to genotoxic stress.

Figure 6:

Myr-p110β expression sensitizes MEF more to genotoxic stress inducers in comparison to the other Class IA PI3Ks. (A) Indicated cells were seeded (10⁴) on 6-well plates, cells were fixed after 3 days and crystal violet stainings were performed. Absolute quantifications of stained cells are displayed (n=3) with standard deviation. Statistical significance was checked with Student’s t-test. Ns: p>0.05, *: p≤0.05. (B) Depiction of IC₅₀ values of doxorubicin of each depicted cell line, calculated from dose – response curves (0–200 μM) in three independent biological replicates. WT MEFs are used as baseline control. (C) Depiction of IC₅₀ values of Cisplatin of each depicted cell line, calculated from dose – response curves (0–5 μM) in three independent biological replicates. WT MEFs are used as baseline control. Circles represent mean IC₅₀ values for WT MEFs, whereas square, triangles and reverse triangles display the IC₅₀ values for myr-p110α, myr-p110β and myr-p110δ MEFs respectively.