A PDK-1 allosteric agonist neutralizes insulin signaling derangements and beta-amyloid toxicity in neuronal cells and in vitro

The Alzheimer’s brain is affected by multiple pathophysiological processes, which include a unique, organ-specific form of insulin resistance that begins early in its course. An additional complexity arises from the four-fold risk of Alzheimer’s Disease (AD) in type 2 diabetics, however there is no definitive proof of causation. Several strategies to improve brain insulin signaling have been proposed and some have been clinically tested. We report findings on a small allosteric molecule that reverses several indices of insulin insensitivity in both cell culture and in vitro models of AD that emphasize the intracellular accumulation of β-amyloid (Aβi). PS48, a chlorophenyl pentenoic acid, is an allosteric activator of PDK-1, which is an Akt-kinase in the insulin/PI3K pathway. PS48 was active at 10 nM to 1 μM in restoring normal insulin-dependent Akt activation and in mitigating Aβi peptide toxicity. Synaptic plasticity (LTP) in prefrontal cortical slices from normal rat exposed to Aβ oligomers also benefited from PS48. During these experiments, neither overstimulation of PI3K/Akt signaling nor toxic effects on cells was observed. Another neurotoxicity model producing insulin insensitivity, utilizing palmitic acid, also responded to PS48 treatment, thus validating the target and indicating that its therapeutic potential may extend outside of β-amyloid reliance. The described in vitro and cell based-in vitro coupled enzymatic assay systems proved suitable platforms to screen a preliminary library of new analogs.


Introduction
Clinically-based Alzheimer's Disease (AD) currently affects 5.8 million or 1 in 10 adults (10%) in the U.S.A. over age 65 and 32% in the >85 age group. Several phase III clinical trials of promising agents to prevent AD progression, based primarily on the amyloid hypothesis, have yielded disappointing overall results. These included anti-amyloid agents such as γ-secretase a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 autophagy markers [68,69]. Intra-hippocampal injections of anti-Aβ antibody or immunization normalized the hyper-activation of Akt and mTOR in transgenic AD mice [53,70]. The mechanisms underlying the paradoxical hyperactivity of Akt and mTOR under basal conditions, are not completely understood. In addition to direct activation of mTOR noted above, Aβ can directly inactivate PTEN (phosphatase and tensin homolog), thereby disinhibiting PI3K [69,71]. One seminal study found all IRS-1 -S and -Y sites were hyperphosphorylated in live AD hippocampal and cerebellar tissue, essentially isolating IRS-1 from binding to insulin receptors and p85-PI3K. Intrinsic over-activation of mTOR/S6K and other kinases was held responsible. Importantly, their work proved resistance to insulin/IGF-1 action in AD; a 90% decrease in Akt, IRS-1, IR, and mTOR phospho-activations in response to insulin stimulation [52].
There is opposing evidence gathered from several AD models that basal Akt is deactivated, which is also consistent with IR in AD. Inhibited Akt is further noted in post-mortem tissue from AD [64,72], Huntington's and Parkinson's diseases [67, [73][74][75]. In two AD models, the inhibition of PTEN instead rescued synaptic and cognitive impairments, mediated through the stimulation of PI3K/Akt [76]. Conversely, PTEN over-expression led to synaptic depression. Aβ peptides applied to hippocampal neurons induced the same synaptic defects and dephosphorylation of Akt by recruiting PTEN to dendritic spines [76]. In 2576 AD mice, where cellular Aβ is co-localized to mTOR, it was actually found to have an inhibitory role [72]. Moreover, reduction in mTOR signaling markers and basal phospho-Akt levels/enzymatic activities, were found in 2xAPP/PS1 mice and in AD brain. These were correlated with oxidatively damaged synaptic Akt. Akt enhancement rescued BDNF-induced protein translation [77]. Deactivation of Akt is reported in rat PCNs and N2a cells exposed to oligomeric Aβ, resulting in inhibition of normal BDNF-induced Akt/mTOR activation [78,79].
Due to conflicting reports and the paucity of preclinical and clinical data on direct Akt/ PDK-1 intervention in AD models, we tested the hypothesis that targeting insulin resistance at this step may be beneficial. In a previous study, intraneuronal Aβ42 (Aβi) expression led to a decrease in the levels of p-Akt and activity, causing p-Tau accumulation and apoptosis [80]. Aβi inhibited the association of PDK-1 with Akt, resulting in the loss of normal insulin-stimulated pathway activation [64,81]. This added mechanism for IR presents a novel target for the treatment of AD. We reasoned that an allosteric ligand acting on the Akt/PDK-1/mTORC2 interaction complex could normalize insulin sensitivity and restore the imbalance in Akt activity. Promising results from early clinical trials in MCI and mild AD of insulin sensitizers (metformin, [82]), GLP-1 receptor agonist/ incretin analogs (liraglutide, [83]), intranasal (IN) insulin [84,85] and insulin-sensitizing PPAR-γ agonists that target genes such as IRS-1, GLUT-4 and PI3K [86,87] (Rosiglitazone, [88]; Pioglitazone, [89]), support finding druggable targets in this pathway and several relevant ongoing trials: Metformin (phase 3, NCT04098666), liraglutide (phase2b, NCT 01843075, [90]) and semaglutide (phase 3, NCT 04777396). However, some phase 3 trials have not met their primary endpoints or been terminated for lack of efficacy, e.g. IN insulin [91], Rosiglitazone [92,93] and Pioglitazone [94,95]. no. 462439-3/2013, PI: HWQ. All personnel and collaborators involved in these experiments, performed on mice and rats, were included. All procedures, tissue collection, biohazard uses, recombinant DNA, husbandry, special diets, breeding and euthanasia were covered. Euthanasia was by either sodium pentothal (120-200 mg/Kg IP) or ketamine (80-100 mg/Kg IP) followed by decapitation.

Cell viability
SH-SY5Y Cells were washed twice in warm DPBS and incubated in 1 ml DMEM containing 0.5 mg (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide) (MTT or WST; Molecular Probes, Eugene, OR) for 2-3 h at 37˚C and 5% CO2. The medium was aspirated and the cells were washed twice with pre-warmed DPBS. The formazan salts were dissolved in 1 ml pure ethanol before use. Cells were homogenized by repetitive pipetting and centrifuged for 5 min at 4500 rpm, and the supernatant collected. Absorbance was read against an ethanol blank at 590 nm.

Western blot analysis
Whole-cell extracts were used directly for western blot analysis (20~30 μg). Extracts from cultured cells prepared in lysis buffer, were diluted into Laemmli sample buffer, heated (95˚C, 10 min), cleared by centrifugation, separated on SDS-PAGE and transferred to PVDF membrane (Immobilon-P; Millipore). Membranes were blocked in TBS containing 0.3% Tween-20 and 5% (wt/vol) non-fat dry milk. After incubation with primary antibodies (18 hr at 4˚C in buffer containing 5% BSA and 0.05% NaN3), blots were washed and incubated in HRP-conjugated secondary antibodies (1:2000 dilution; Cell Signaling). Signals were detected using ECL reagents and quantified using a Kodak Image Station 4000R.
An in vitro radio assay (EMD Millipore, KinaseProfiler) was also adapted as follows. PKBα (human, recombinant, inactive, 209 nM) is incubated in 8 mM MOPS pH 7.0, 0.2 mM EDTA with 30 μM GSK3α/β consensus sequence GRPRTSSFAEGKK and PDK1 (human, recombinant, 285 nM). β-amyloid peptide (oligomerized, 5 μM final) and PS48 are added. Final DMSO is 2%. 10 mM Mg Acetate and [γ-33P-ATP] (specific activity approx. 500 cpm/pmol) are prepared. The reaction is initiated by the addition of the Mg ATP mix (200 μM ATP final). After incubation for 40 minutes at 37˚C, the reaction is stopped by adding 3% phosphoric acid solution. 10 μl of the reaction is spotted onto a P30 filtermat and washed three times for 5 minutes in 75 mM phosphoric acid and once in methanol, prior to drying and scintillation counting. immersed in cold (5-7˚C), oxygenated (95% O2/5% CO2) artificial cerebrospinal fluid (ACSF) containing (in mM): 126 NaCl, 3 KCl, 1.25 NaH2PO4, 1 MgSO4, 2 CaCl2, 26 NaHCO3, 10 dextrose. Coronal slices of prefrontal cortex from day 14 rat pups (400 μm) were perfused with Aβ oligomers (40 nM) or Aβ plus PS48 (10 μM) x 60 min. before applying the high frequency stimulation (HFS) protocol. Control treatment is DMSO in ACSF (artificial cerebrospinal fluid, 20 μM bicuculline). Extracellular postsynaptic field potentials were recorded using an AxoClamp2B amplifier (Axon instruments) and EX1 differential amplifier (Dagan), and digitized at 10 kHz. Data was acquired using Igor Pro (Wave Metrics) and Neuromatic (www. neuromatic.thinkrandom.com). The stimulus intensity eliciting 50% of the maximum amplitude (~32 μA) was used for all measurements before and after LTP induction. Baseline amplitudes were recorded for 20-30 minutes using single field stimuli applied every 30 sec (2Hz) to layer lV-V using concentric bipolar electrodes. Following a stable baseline period, LTP was induced by two sets of high-frequency stimulation (HFS) at 100 Hz, 60 μA (twice stimulus intensity), for 1 sec, 20 sec apart. Extracellular postsynaptic field potentials were measured from layer ll-lll using glass micropipettes filled with 0.9% NaCl. The amplitude rather than the slope of evoked FPs was used as a measure of the population excitatory synaptic response because in the neocortex the initial slope is contaminated by antidromic stimulation. The last 10 minutes (20 points) of the stimulated EPSP recordings were normalized to baseline and then averaged. LTP values were expressed as a percentage of mean baseline EPSP ± SEM or % change normalized to baseline. ANOVA with Tukey's multiple comparisons and paired twotailed t-tests were used for statistical analysis.

Drug screening
Two focused libraries of PS48-family compounds were designed based on known and hypothesized structure-activity relationships and tested using an in vitro screen carried out in 48 well plates as follows (in order of rapid additions, final concentrations): 10x Kinase buffer, recombinant PDK-1 (5 μl, pre-immunoprecipitated onto agarose beads using monoclonal IgG), PIP3 50 nM, Aβ42 (as ADDL oligomers, 10μM), recombinant Akt-1 (5 μl, pre-immunoprecipitated onto beads using polyclonal goat IgG, treated with PP2A to dephosphorylate Akt and washed), compound or PS48/ PS47 (solubilized in DMSO then diluted with H2O, 10 μM), ATP (to initiate Akt activation, 200 μM). Incubation proceeded for 15 min. GSK-tide (Cell Signaling) was then added and the reaction allowed to carry for 20 min more before termination in sample buffer and fractionation on SDS gel. Transfers were probed with anti-p-GSK, anti phospho-473 and -308Akt and total Akt. The in vitro results were validated using a cell culture-based assay, as above. Briefly, adenovirus-infected PCNs (2 days) were induced with doxycycline (48 hrs) to express Aβ42. The compound was added for 12 hrs. The cells were stimulated with insulin before harvest.

Statistical
Where quantified, experiments were carried out in triplicate unless otherwise stated. Mean, standard errors and significance levels using students t-test were computed in Excel or Prism. In vitro Akt activation assay data (see above), in which the % inhibitory effects of Aβ42 monomers and ADDLs were tested, was fitted using a 2 site (hyperbolic), non-linear algorithm (Prism) to obtain Imax and K0.5 equilibrium constants. Western signal intensities were all quantified by densitometry. Akt activation (phosphorylation) and activity (GSK phosphorylation) western results (stimulation or inhibition) were for the most part, concordant and equivalent in the fraction of change versus control. Therefore, where both endpoints were evaluated, their normalized values were combined in the quantification as indicated. ANOVA (1 way; between treatment groups or columns and 2 way; between treatments groups and between repeated measures or rows) was carried out on Western, cell viability and LTP experiments. Where indicated, Dunnett's or Tukey's multiple comparisons were applied. Effect sizes for Aβ and drug additions are given as mean differences (± SE difference) and 95% confidence intervals.

Results
We had previously shown that cellular β-amyloid expression inhibits PI3K-PDK1-Akt signaling [80,81,105]. To summarize, in vivo assays of phospho-Akt/total Akt and downstream substrate, phospho-GSK3β levels were carried out on extracts from cultured neurons exposed to an inducible adenoviral vector encoding Aβ42 [64,80]. Cells were pretreated with insulin for 20 min prior to harvest in order to activate Akt, finding that insulin-stimulated p-Akt levels were reduced to baseline in the presence of Aβ42 expression. To confirm this, Akt enzymatic activity was measured in cell lysates using a coupled assay; immuno-precipitated (IP) Akt from insulin-stimulated cells, phosphorylated a synthetic substrate peptide bearing the phosphorylation consensus sequence GSK3β fused to paramyosin (crosstide). An in vitro kinase assay was also developed (see methods), finding that 5 μM Aβ peptide inhibited the Akt-dependent phosphorylation of the target substrate, accompanied by an expected reduction in pSer473Akt. Further data [64] indicated that Aβ inhibits the PDK-1-dependent activation of Akt by disrupting their interaction.
The pathological target identified in these in vivo and in vitro platforms suggested that a small molecule could be found that modulates the insulin-PDK-Akt activation cycle in such a way to relieve the inhibitory amyloid effect. A chemical data base search identified an allosteric activator of PDK-1 (CAS 1180676-32-7, PS48), a chlorophenyl pentenoic acid having a MW of 286.7 [106] (Fig 1 top). Furthermore, it has an inactive 'E' isomer, PS47, for control use [96,97]. It is unique in its action to bind the hydrophobic motif/PIF binding pocket of PDK and not the ATP binding site. The compound has other possible beneficial actions that may translate to improve hippocampal neurogenesis [107].
https://doi.org/10.1371/journal.pone.0261696.g001 introduced to produce insulin resistance. At moderate insulin doses (20 nM), Akt activation is inhibited by Aβ42, whereas high doses of insulin (100 nM) overcame the effect. Thus Aβ42 was shown to desensitize insulin action, raising the insulin concentraton threshold to achieve an equivalent response (S1B Fig in S1 File). In this system, we find that PS48 exposure did not intrinsically activate basal Akt. However, in the presence of low dose insulin (3 nM) where Akt activation is subthreshold, PS48 augmented the response. This is in line with its purported allosteric action to positively modulate PDK activity (S1C Fig in S1 File). At moderate insulin doses, that produce robust Akt phosphorylation, PS48 did not further enhance it. Neither did PS48 intrinsically affect at least one critical downstream factor in this pathway, mTOR(S1D Fig in S1 File). These properties make it ideal to test if it will protect insulin signaling against amyloid peptide toxicity, while not over-regulating the pathway.
Several downstream effectors and substrates of Akt were examined for sensitivity to Aβ toxicity and PS48. CREB, the cAMP response element binding transcription factor, has pleiotropic actions to promote neuronal survival, progenitor proliferation, neurite outgrowth and differentiation. It is also well documented to control the activity-driven and neurotrophin-dependent expression of proteins essential to long term memory formation (LTM) and synaptic plasticity (LTP) (see reviews by [42,108,109]). It is situated in the PI3K/Akt/CREB pathway to transduce effects of Insulin, IGF-1 and BDNF on protein expression critical to neurogenesis and plasticity [40,43,110]. CREB supports LTM by stabilizing synaptic strength, regulating intrinsic neuronal excitability and recruiting subsets of neurons in the hippocampus and amygdala that encode the memory trace [111][112][113][114]. We focused on CREB because it can be directly activated by Akt [115,116], is protective against neuronal apoptosis [117,118] and supports LTP [119]. We found consistent inhibition of insulin-stimulated CREB phosphorylation (pS133) by intracellular Aβ42 and this was also stabilized by PS48 (50 nM) (S2A Fig in S1 File).
Previous work had shown sensitivity of endogenous GSK3α/β (inhibitory S9A phosphorylation) to viral expressed Aβ42 [80], however the current experiments under combined Aβ42 pressure and PS48 proved inconclusive. Nevertheless, PS48 had no effect on resting cellular pGSK levels (S2B Fig in S1 File). Finally, in testing for changes in activating phospho-levels of indirect downstream substrate and metabolic sensor mTOR, we found that neither Aβ expression (also shown in [105]) nor PS48 applications had any effect (S2C Fig in S1 File).
We next tested PS48 in an in vitro assay of both Akt activation (phosphorylation of T308) and enzyme activity. In Fig 4A, recombinant Akt and PDK-1 were added to a reaction mixture containing synthetic Aβ42 peptide oligomers (10 μM) and ATP to start the reaction. Some experiments employed added PI3P and/or pre-dephosphorylation of Akt by treatment using PP2A, with variable improvements in the efficiency of activation. A GSK fusion peptide was added as substrate for the enzymatic readout (Westerns of phospho-S9 GSK3α/β). Initial experiments used high dose PS48 (100 μM) and tested if added before (pre) or after (post) the Aβ peptide made a difference. pAkt-T308 levels were reduced in the presence of Aβ, similar to the cell-based expeiments. PS48 restored Akt phosphorylation. Moreover, it had greater efficacy when added after Aβ peptide equilibration. So, this procedure was followed in Fig 4B, that alternatively used recombinant Akt and immunoprecipitated PDK-1. PS48 was found to be active at 0.1 and 10 μM in reversing the inhibition of GSK phosphorylation. The pooled results from same experiments over a range of PS48 concentrations are quantified in Fig 4C. In the presence of 10 μM Aβ42,~30% of Akt activation is inhibited (expressed as a remaining fraction of the control (absent Aβ, set to 1.0), thus 0.70 ± 0.09. Beginning at 10nM PS48, activation/activity is increasingly normalized until a maximum of 0.95 ± 0.08 of control is reached at �1 μM PS48 (includes 10 and 100 μM data points). When all values � 0.1 μM are combined (n = 8), the trend toward normalization reached significance (p < .02, compared to 0 nM PS48, t-test, n = 6).
To further explore target engagement, we probed the characteristics of Aβ42 interaction with PDK-1 and/or Akt-1. We conducted equilibrium Aβ42 dosage and binding experiments in solution to the respective kinase targets. First, the in vitro pardigm above was used to test whether Aβ42 peptide exhibits a dose-dependant specificity to inhibit the activation (phosphorylation of T308) and enzymatic activity of Akt (to phosphorylate its consensus substrate). The combined Western densitometry data in Fig 5A suggests the possibility that either Aβ monomers or oligomers act to inhibit either one or both of the two kinases, separately or in complex, and in a saturable manner. Using a 2 site, non-linear fit algorithim a Imax value of5 2% inhibition of control activation is obtained for both species, however the oligomers showed greater affinity (ADDLs: K0.5 = 0.08 μM, monomers: K0.5 = 0.31). To independently confirm the reversal of the PDK/Akt activation sequence by Aβ42 peptide, we employed a modified radiolabelled-based assay in Fig 5B. Again, PS48 dose-dependently reduced the inhibition of P 32 -labelled phosphate addition to a consensus peptide. The effect was first noticed at 10 nM (Fig 5B), consistent with Fig 4C. Next, a novel assay was adapted to determine if Aβ42 bound to either PDK-1 or Akt-1 in solution and to discern any effect of PS48 on this,  In vitro assay data are presented as % inhibition of either activation of Akt (phospho-T308) by PDK-1 or of its subsequent activity to phosphorylate a GSK3β consensus peptide fused to paramyosin ('crosstide'). Densitometry results from both western blots were combined. Imax~52% for both Aβ preparations: ADDL, K0.5 = 0.08 μM; monomers, K0.5 = 0.31 μM. n = 2-7 experiments each point, ± 1 SEM. 5B. In vitro radioassay (see methods). PS48 from 10 nM to 0.5 μM, gradually diminishes the inhibitory effect of Aβ (5 μM ADDL alone, bar 2, � p < .05 vs. vehicle) on the phosphorylation of GSK peptide (measured in cpm). The trend is toward control activity, not significant from bar 1 or other vehicle controls. 5C. Quantification of Aβ42 binding to kinase targets by fluorescence polarization (FP). The probe was FAM-tagged-Aβ42. Recombinant kinase titrations shown along bottom. The spectrophotometric polarization signal increases as the probe becomes more restricted by receptor binding. Bmax: Akt 0.32; PDK 0.16. IC50: Akt~2 μM, PDK~1 μM. 5D. Direct binding of recombinant PDK-1 and Akt-1 to tagged Aβ42 in solution is confirmed and exhibits saturation characteristics. Aβ42 was immunoprecipitated (6E10), fractionated by Western and co-precipitates are detected using anti-Akt and -PDK. Aβ42 concentration was fixed at 200 nM and detected using polyclonal R1282. Lane 1 is Aβ peptide control, lane 2 is beads without 6E10 control. https://doi.org/10.1371/journal.pone.0261696.g005 Akt, but fewer binding sites (Fig 5C). PS48 additions did not affect either polarization signals, concluding that Aβ42 is probably not competitively occupying the pocket site (results not shown). The in-solution FP assay was validated by co-immunoprecipitating the bound reaction products. IP of FAM-labeled Aβ42 with anti-6E10 pulls down increasing amounts of Akt and PDK according to the titration until saturation (Fig 5D). Aβ concentrations were constant as shown in the anti-R1282-developed blot shown below. Previous work has shown that Aβ peptide in AD brain co-immunoprecipitated with PDK and Akt and that cellular expression of Aβ42 disrupts Akt-PDK interaction [64].
Lastly, we launched drug screening experiments using a focused library of novel compounds that were synthesized using PS48 as the starting scaffold. Two generations were created based on structure-activity relations (SAR) in an attempt to optimize performance. The quantified results of western-based in vitro assays and signal intensities, compared to PS48, are shown in Fig 6A. The activity of each compound against the Aβ42 effect to inhibit the PDK-Akt activation sequence is shown as two indices: 1) 'fraction of control activity', as remaining in the presence of Aβ with drug [Aβ with drug]/[control], where the control is absent Aβ and normalized to 1.00. The fraction of control activity observed in reactions having just Aβ alone (maximal inhibition, no drug) was 0.62 ± 0.16 (top left), and 2) 'fraction recovery to control level from Aβ42-induced inhibition', where the index is normalized instead to the Aβ inhibitory effect size [Aβ with drug-Aβ]/ [control-Aβ] signals. The fraction recovery with just Aβ alone (or no drug effect) is therefore 0.0 on this scale (top right). Using either index, a value of �1.00 is a complete reversal. Based on this data, a rough SAR is beginning to emerge where it appears that the greatest effect is realized by modifying the acidic portion of PS48, e.g. extension from the linker and aryl groups. An increase in potency over PS48 was achieved with compounds 7, 25, 31 and 68. Examples of in vitro assay performance are shown below (Fig 6B) for compounds no. 25 and 31 (generation 1) and 68 (generation 2). These were then validated as correcting the inhibition of Akt phosphorylation in the cell-based model of Aβ toxicity (as in Fig 1), shown at the bottom for compounds 25 (with quantification) and 68 (Fig 6C). The control stereoisomer PS47 had no effect in these in vitro reactions (table bottom right). The new 'hit' compounds also had no direct or indirect effects on potential downstream off-targets such as mTOR (protein levels and phosphorylation status, results not shown). Moreover, the LD50 of several promising compounds in N2A neural cultures proved even higher than PS48 (e.g. no. 25; 350 μM, WST assay, result not shown).

Discussion
Akt (PKB) is an essential kinase in the insulin/IGF signal cascade having pleotropic influence over many cell survival and metabolic pathways. It's co-crystal structure in complex with a substrate peptide (GSK3β) and an ATP analog, reveals the structural relationship between the C-terminal hydrophobic motif (HM) and the activating phosphorylation of Akt on the Thr 308 residue by PDK-1 [120]. The co-crystal structure of PDK-1 in complex with ATP reveals the HM-binding pocket (PIF domain located on the N terminus) and phosphoSer-binding pocket through which it docks with its many substrates including: Akt, SGK, S6K, PKC and RSK [121]. Interestingly, Akt is the only substrate not requiring docking at the PIF-pocket site to undergo catalytic activation by PDK-1 [122,123]. PS48 and family of it's analogs are small molecule allosteric activators of PDK-1, binding within the PIF pocket [97], thereby facilitating Akt activation by IGF-1 [96,106,124]; for review see Xu et al. [125]. These features are presented in schematic in S3 Fig in S1 File. Additional actions of this allosteric binding include supporting the induction of pluripotent stem cells from somatic cells [107].
Herein, we characterized PS48 action in the context of insulin signaling using neuroblastoma cell lines and then demonstrated its ability to partially or wholly normalize Aβ42 oligomer induced insulin resistance and toxicity using an adenoviral expression system. First, a dose dependency of Akt activation by insulin in PCN and N2a cultures was established. Results are combined phosphorylations of Akt (T308, Akt activation) and pGSK3β consensus peptide substrate (S21/9, Akt enzymatic activity). 6B. Selected Westerns of in vitro reaction mixture by-products, highlighting significant GSK peptide (S21/9) and Akt (T308) phosphorylations in the presence of Aβ by compounds 508-1-25 and -31 (1st generation) and 508-1-68 (2nd generation). These 'hits' restored activations at or better than PS48. 6C. In vivo verfication studies in SH-SY5Y cells. Addition of Doxycycline results in intracellular amyloid accumulation (6E10 blots). Compounds 25 and 68 (1 μM) intrinsically boost Akt phosphorylation to similar degrees, that remain sustained under Aβ pressure (as the case with PS48). Phospho-Akt levels, both T308 (as exampled) and S473 (not shown) were quantified and pooled for the bar graph, featuring 508-1-25. All cultures were stimulated with insulin (40 nM, 240 ng/ml) in the last 20 minutes of the Aβ-expression period. n = 3 experiments. � p < .05 vs. Con, + p < .01 vs. Aβ42. https://doi.org/10.1371/journal.pone.0261696.g006 Intracelluar expression of Aβ42 oligomers resulted in a reduction of sensitivity to insulin, such that higher insulin doses were required to overcome the resistance. At low, subthreshold doses of insulin (3 nM), PS48 pretreatment appeared to sensitize Akt activation. Next, PS48 (0.1 to 1 μM) is shown to significantly counter the inhibitory effect of Aβi expression on submaximal insulin-induced Akt phosphorylation in live cells, similar to the effect of a higher insulin dose (100 nM) alone. The same findings were obtained in another insulin responsive cell line, C2 myotubes. Moreover, PS48 partially overcame insulin resistance in a non-amyloid model of cellular toxicity, to the saturated fatty acid palmitate. A downstream effector and substrate of Akt, CREB, is also hypophosphorylated after Aβ expression, and accordingly corrected by treatment with PS48. On the other hand, another effector, phospho-mTOR, remained unaffected by either treatment. The lead compound furthermore partially prevented Aβ-induced cell death in a neuroblastoma cell line, as did high dose insulin and pioglitazone treatments. Next, it significantly normalized the effect of synthetic Aβ peptide (ADDL oligomers) to inhibit LTP in rat prefrontal cortical slices. To test the purported cellular step involved in this mechanism of insulin resistance, we performed in vitro reactions using recombinant Akt and constitutive active PDK kinases to phosphorylate a GSK3α/β-based consensus peptide substrate. Aβ oligomers (ADDLs) inhibit Akt activation and crosstide phosphorylation and PS48 (10 nM to 1 μM) restored this to~90% of control levels. Mechanistically, tagged Aβ42 is shown to bind to both recombinant Akt and PDK using an in-solution fluorescence polarization assay, the two kinases having different affinities and saturation levels. Finally, in vitro and cell-based assay platforms were employed in a focused medicinal chemistry effort to probe the structure-activity characteristics of the parent molecule. Other analogs were found that reversed the inhibited Akt activity by better than 90%.
The two main physiological readouts of Aβi toxicity in this study that were partially protected by PS48 were cell death and inhibited synaptic plasticity (LTP). We show that PS48 and analogs in development significantly improved Akt activation by insulin from inhibition by Aβi accumulation. Akt is critical to both neuronal survival [126] and LTP as demonstrated in prefrontal cortex, amygdala and hippocampus, [127][128][129]. Hippocampal LTP is particularly sensitive to Aβ oligomers [48,49]. Among the possible effectors of these two outcomes, we find the CREB link plausible because this result mirrored the Akt responses to Aβi and PS48 (Fig 1). CREB is activated by several canonical receptor-activated kinase pathways (i.e. PKA, CaMK, MAPK). In particular, BDNF/TrkB receptor activation has been well studied [130][131][132]. However, insulin and IGF-1 also phosphorylate CREB via PI3K/Akt [43], and this can occur directly (see also [133,134]). Another effector intervening between PI3K/PDK/Akt and CREB is GSK3αβ. Akt stimulation results in GSK3 inactivation, resulting in CREB de-repression (via inhibitory pS129: [135,136]). Moreover, GSK3 activation promotes apoptosis [137,138] and depresses spatial learning and LTP in mice [139,140]. Although our results were inconclusive on cellular GSK, the Akt activity assay data using GSKtide could still be consistent with a partial role for this mechanism in our endpoints.
Several limitations of this work will require clarification in future studies. Practical ones include finding the optimum concentrations of drug to test, which depends on the assay employed. Taking for instance in vitro drug screening, the optimum may actually lie between 100 nM and the 10 μM concentrations reported here (Figs 4C and 6). The same applies to cell viability assays (Fig 2). Another issue that arose during the in vitro reactions was the basal phosphorylation status of commercial, recombinant Akt. It was generally high and only partially decreased by PP2A treatment. However, we believe this does not affect our results because drug efficacy was based on: 1. maximal stimulation with PDK (recomb Akt was always further phosphorylated after PDK (constitutively active) addition, regardless of its basal phospho state), 2. the level to which Aβ inhibited this, and 3. the recovery from that by drug. Moreover, regardless the phospho level of unstimulated recombinant Akt, it had no enzymatic activity (Figs 1A and 6B), which further supports it use for the same reasons. Nevertheless, experiments using more completely dephosphorylated Akt are a consideration to see if an even greater therapeutic effect size can be realized. Next, we found adding PS48 after Aβ peptide slightly more effective in the same in vitro experiments. This interesting result raises the speculation that an Aβ-induced conformational change in the Akt-PDK complex, makes the allosteric PS48 modification more effective thereafter. Preventative and treatment roles for PS48 therefore deserve further study. Finally, the AD model used here is based on cellular Aβ42 and is only one of many others. While our data are limited to this view of pathogenesis, the electrophysiology data on ex vivo slices exposed to soluble oligomeric Aβ species recommend that PS48 could also be tested in models emphasizing extracellular Aβ, or for that matter Tau, accumulations.
The results of published studies have been mixed with respect to the state of Akt activation in AD brain and models, reporting either over or under phosphorylation or activity. Far fewer reports specifically address the 3-phosphoinositide-dependent kinase, PDK-1 in AD or neurodegeneration. Pietri and colleagues [62, 141] found PDK-1 activity increased in neurons infected with prion protein PrPSc or in transgenic mice affected by β-amyloid pathology, as well as in AD brain. In a novel but complicated mechanism, PDK-1 overactivation is held responsible for loss of TACE-mediated APP and PrPc α-secretase cleavages, from accelerated TACE internalization. The result is an over-production of Aβ and TNF-mediated neurotoxicity and memory deficit. Accordingly, PDK-1 silencing or inhibition restored survival and memory and reversed pathology parameters. Both PrPSc and Aβ were hypothesized to stimulate PrPc to recruit Src and PI3K kinases to overactivate PDK-1. The relevance of these changes to the insulin/Akt axis was however not explored. In contradistinction, a PrPSc-like peptide  inactivated Akt and caused death in SHSY5Y and primary granule cells, outcomes confirmed in a PrPSc infected mice model. These were reversed by constitutive activation of Akt or insulin treatment [142].
It remains plausible but clinically untested if a strategy to restore Akt responsiveness to insulin, has value in prevention or treatment of AD. Based on our data targeting the PDK-Akt activation sequence, PS48 or a structurally similar allosteric analog may be a viable candidate. Importantly, PS48 does not itself over-stimulate normal insulin signaling in PCNs, nor overactivate basal Akt, lessening potential oncogenesis concerns [143,144]. Notably, PS48 was not toxic to cells (LD50 = 250 μM). This is possibly due to the purported allosteric modulatory action of this compound, which has also been observed in various other drugs of this class [145,146]. Other bi-aryl, halogenated carboxylic acids have a safe record in humans, for instance, Tolfenamic acid, used for the treatment of migraines [147,148].
In support of efforts to facilitate Akt/PDK signaling, other interventions have had similar action on the insulin /Akt transduction pathwy to mitigate Aβ toxicity. For instance, α7nAcR stimulation (nicotine on PCNs) activates PI3K and pAkt to block Aβ-enhancement of mitochondrial AIF release/nuclear translocation [149], as well as block Aβ-mediated glutamate toxicity and prevent mitochondrial apoptosis [150][151][152]. We note with interest several recent reports that direct pharamacological activation of Akt in Aβ-injected and in 5X FAD AD mice, resolved memory impairments and synaptic LTP deficits and restored inhibited Akt to control levels [153]. Activation of Akt/PI3K in primary mouse neurons also proved protective against transfected mutant APP and improved locomotor activity in an Aβ42-drosophila model [154]. The aforementioned insulin pathway clinical trials were all supported by robust preclinical cell and animal data, e.g. GLP-1 mimetics [155][156][157][158]), IN insulin [159][160][161][162], PPAR agonists [14], as well as by epidemiological data, e.g. metformin [163,164]. Recent reviews of these drug classes in AD prevention endorse continued clinical tials, where supported by basic studies [165][166][167].
Our future studies will focus on an expanded class of modified biphenyl pentanoic acids and optimization of the PS48 pharmacophore with the goal to push potency into nM range and improve cell permeability. PS48 also has a low toxicity profile in preliminary animal testing, and further clinical data will be reported separately.