Cyclic cis -Locked Phospho-Dipeptides Reduce Entry of AβPP into Amyloidogenic Processing Pathway

Materials

The human H4 neuroglioma cell line was purchased from ATCC (Manassas, VA). Dulbecco’s Modified Eagle Medium (DMEM), Penicillin-Streptomycin and culture dishes were from Corning Life Sciences (Tewksbury, MA); Fetal bovine serum (Premium Select) was purchased from Atlanta Biologicals (Flowery Branch, GA); DC Protein Assay Kit, Pre-stained Dual Color Protein Standards, Clarity Western ECL, Chemidoc MP System, Image Lab Software, Chemi Hi Sensitivity blot application, Immuno-Blot LF PVDF membrane, 4–20% Mini Protean TGX Stain Free precast gels, and 4–15% Mini Protean TGX Stain Free precast gels were obtained from BioRad (Hercules, CA); polyvinylidene fluoride (PVDF) transfer membrane (0.45 mm) was purchased from Perkin Elmer (Waltham, MA); Whatman nitro-cellulose transfer membrane (0.2 μm) was purchased from GE Healthcare Life Sciences; Lipofectamine 2000, Pierce ECL Western Blotting, Pierce SuperSignal West Pico substrates and G418 sulfate (Geneticin) were all from ThermoFisher Scientific (Waltham, MA); Recombinant Human BACE1 protein, CF (931-AS) and Mca-SEVNLDAEFRK(Dnp)RR-NH2 Fluorogenic Peptide Substrate (ES004) were acquired from R&D Systems (Minneapolis, MN); the BACE1 inhibitor LY2811376 was purchased from Selleckchem (Houston, TX).

Plasmids

pCAX-AβPP⁶⁹⁵ expressing full-length human AβPP695 was a gift from Dennis Selkoe and Tracy Young-Pearse (Addgene plasmid # 30137). pCMV6-XL5-BACE1 expressing full-length human BACE1, transcript variant a, NM_012104.3 (# SC115547) was purchased from Origene, Rockville, MD. pSV2neo [33] which provides a selectable marker for resistance to the antibiotic G418 in mammalian cell lines was a gift from David Shalloway, CornellUniversity.

Antibodies

Rabbit anti-sAβPPβ (poly8134, 813401) and the mouse monoclonal antibody 6E10 (803001) were purchased from BioLegend (San Diego, CA) and the anti-β-tubulin mouse monoclonal antibody (2G7D4, A01717-40) was acquired from GenScript (Piscataway, NJ). The rabbit monoclonal antibodies anti-AβPP (EPR5119(2), ab133588), anti-C-terminal AβPP antibody Y188 (ab32136) and anti-BACE1 (EPR3956, ab108394) were all purchased from Abcam (Cambridge, MA). Goat anti-rabbit and goat anti-mouse horseradish peroxidase (HRP)-conjugated secondary antibodies were from Jackson ImmunoResearch Laboratories (West Grove, PA). The anti-pan-Aβ rabbit monoclonal antibody (D54D2) was purchased from Cell Signaling Technology (Danvers MA).

Cell culture

All H4 neuroglioma cell lines were routinely grown in monolayer culture in growth medium consisting of DMEM (4.5 g/L glucose, 3.7 g/L sodium bicarbonate) supplemented with 10% fetal bovine serum and 100 IU /100 μg/mL penicillin/streptomycin at 37°C in a humidified atmosphere (90%) containing 10% CO₂. Cells were isolated by trypsinization and routinely passaged to maintain stocks or plated for experiments as described below or in the figure legends.

Generation of H4 neuroglioma cell lines overexpressing AβPP⁶⁹⁵ or BACE1

Cell lines constitutively over-expressing human AβPP⁶⁹⁵ (H4-AβPP695) or human BACE1 (H4-BACE1) were created by co-transfecting either pCAX-AβPP⁶⁹⁵ or pCMV6-XL5-BACE1 (2–4 μg) with the G418 resistance plasmid pSV2neo (0.2–0.4 μg) into H4 neuroglioma cells using Lipofectamine 2000 according to the manufacturer’s instructions. Forty-eight hours after transfection, cells were split into growth medium supplemented with 500 μg/mL of G418 and after 2-3 weeks G418 resistant colonies were screened by immunoblotting for AβPP or BACE1 over-expression as compared to the parental H4 cell line. Positive clones were expanded and maintained in growth medium containing 200 μg/mL G418 until frozen.

Analysis of custom-synthesized compounds

Nuclear magnetic resonance (NMR) analysis: All pCDPs used in this study (pCDP, pCDP-Bzl, pCDP-diBzl, pCDP-diPOM) were custom synthesized using green chemistry [34] and phosphorylation of Threonine completed [35, 36] by Viva Biotech Ltd. (Shanghai, China) and purchased through Trillience (Toronto, Ontario). Deuterated-methanol NMR solvent (99.8%, CD3OD) was purchased from Cambridge Isotope Laboratories and used for each sample preparation. NMR spectra (1H and 13C) were recorded at room temperature (RT) with a Bruker Avance^III HD 800 MHz instrument (SUNY-ESF) or a Varian Inova 600 MHz instrument (Cornell University). Chemical shifts are reported in δ (ppm) units relative to residual solvent peaks CD3OD (3.31 ppm for 1H, 49.0 ppm for 13C). Splitting patterns are assigned as s (singlet), d (doublet), t (triplet), q (quartet), p (pentet), m (multiplet), dd (doublet of doublets), ddd (doublet of doublet of doublets), dddd (doublet of doublet of doublet of doublets), dq (doublet of quartets), dt (doublet of triplets), td (triplet of doublets), qd (quartet of doublets), and pd (pentet of doublets). All NMR spectra were analyzed with MNOVA software (v.10.0). Supplementary Figures 1–4 show 1H NMR spectra for the pCDP compounds.

Liquid chromatography mass spectrometry (LCMS) analysis: Each pCDP standard sample was diluted from 100 mM of stock pCDP in dimethyl sulfoxide (DMSO) down to 1 μM pCDP in 1:1 methanol/water and 2 μL was separated using reverse-phase high resolution UHPLC-MS with an Agilent Zorbax RRHD Eclipse XDB-C18 column (2.1×100 mm, 1.8 μm particle diameter), 0.1% formic acid in acetonitrile (organic phase), and 0.1% formic acid in water (aqueous phase) at a rate of 0.5 mL/min for 15 min at 40°C (through the Chromelon Xpress software system) on a Thermo Scientific Dionex Ultimate3000 UHPLC system, equipped with a diode array detector and connected to a Thermo-Scientific Q Exactive hybrid quadupole-Orbitrap mass spectrometer (Boyce Thompson Institute Mass Spectrometry Center, Cornell University). A solvent gradient scheme was used: 5% organic for 1.5 min, a linear increase to 100% organic over 11 min, and then a 2-min hold at 100% organic before decreasing back to 5% organic over 0.1 min with a final hold at 5% organic for the last 0.4 min, for a total of 15 min. The Thermo-Scientific Xcalibur software package was used to visualize, analyze, and depict the LCMS data shown in SupplementaryFigure 5.

pCDP (Supplementary Figures 1 and 5A)

1H NMR (800 MHz, CD3OD, 25°C): δ 1.53 (d, J = 6.5 Hz, 3H), 1.90–1.96 (m, 1H) 1.98–2.04 (m, 2H), 2.28–2.34 (m, 1H), 3.45 (ddd, J = 11.9, 9.1, 3.0Hz, 1H), 3.65 (dt, J = 11.6, 7.9, 1H), 4.12 (dt, J = 3.4, 1.6Hz, 1H), 4.22 (ddd, J = 9.8, 6.8, 1.7Hz, 1H). 13C NMR (800 MHz, CD3OD, 25°C): δ 19.01, 22.85, 29.10, 46.02, 60.64, 59.89, 72.91, 165.82, 171.71. HR-LCMS (ESI+): Calculated C9H15N2O6P + [M + H]+ = 279.07471; found [M + H]+ = 279.07360, mass tolerance 0.1 mmu, retention time = 1.12 min.

pCDP-Bzl (Supplementary Figures 2 and 5B)

1H NMR (600 MHz, CD3OD, 25°C): δ 1.52 (d, J = 6.3 Hz, 3H), 1.84–1.96 (m, 2H), 1.97–2.07 (m, 1H), 2.22–2.30 (m, 1H), 3.39 (ddd, J = 11.9, 8.7, 3.7 Hz, 1H), 3.49 (dt, J = 11.7, 8.2 Hz, 1H), 4.02 (d, J = 5.9 Hz, 1H), 4.18 (dd, J = 8.1, 8.3 Hz, 1H), 4.73 (dt, J = 7.4, 6.1 Hz, 1H), 4.92 (d, J = 6.2 Hz, 2H), 7.24 (t, J = 7.3 Hz, 1H), 7.31 (t, J = 7.6 Hz, 2H), 7.40 (d, J = 7.4 Hz, 2H). 13C NMR (800 MHz, CD3OD, 25°C): δ 19.41, 22.98, 28.95, 45.96, 59.91, 60.73, 68.55, 71.95, 128.12 (2C), 128.49, 129.09 (2C), 139.15, 166.04, 171.84. HR-LCMS (ESI+): Calculated C16H21N2O6P + [M + H]+ = 369.12166; found [M + H]+ = 369.12050, mass tolerance 0.5 mmu, retention time = 4.43 min.

pCDP-diBzl (Supplementary Figures 3 and 5C)

1H NMR (800 MHz, CD3OD, 25°C): δ 1.51 (d, J = 6.8 Hz, 3H), 1.82 (m, 3H), 2.26 (m, 1H), 3.38 (m, 2H), 4.17 (m, 2H), 5.03 (m, 4H), 5.27 (pd, J = 1.8, 6.8 Hz, 1H), 7.35 (m, 10H). 13C NMR (800 MHz, CD3OD, 25°C): δ 18.57, 22.98, 29,58, 46.24, 60.08, 60.95, 70.60, 70.81, 75.08, 128.91 (3C), 129.11 (3C), 129.60, 129.65, 129.66, 129.68, 137.32 (2C), 164.96, 171.66. HR-LCMS (ESI+): Calculated C23H27N2O6P + [M + H]+ = 459.16861; found [M + H]+ = 459.16706, mass tolerance 0.1 mmu, retention time = 7.49 min.

pCDP-diPOM (Supplementary Figures 4 and 5D)

1H NMR (800 MHz, CD3OD, 25°C): δ 1.25 (d, J = 2.5 Hz, 18H), 1.58 (d, J = 6.8 Hz, 3H), 1.93–1.03 (m, 2H), 2.03–2.08 (m, 1H), 2.34 (dddd, J = 9.7, 6.7, 4.4, 1.7 Hz, 1H), 3.50 (ddd, J = 12.0, 8.8, 3.4 Hz, 1H), 3.69 (dt, J = 12.0, 8.2 Hz, 1H), 4.25–4.22 (m, 2H), 5.29 (pd, J = 6.8, 1.9 Hz, 1H), 5.62–5.68 (m, 4H). 13C NMR (800 MHz, CD3OD, 25°C): 18.61, 27.23 (6C), 23.12, 29.58, 39.74 (2C), 46.38, 60.12, 60.80, 75.87, 84.19, 84.31, 164.86, 171.65, 177.86, 177.91. HR-LCMS (ESI+): Calculated C21H35N2O10P + [M + H]+ = 507.21087; found [M + H]+ = 507.20940, mass tolerance 0.1 mmu, retention time = 7.96 min.

Preparation of pCDP and inhibitor treatment solutions

All pCDPs were solubilized in DMSO to a stock concentration of 100 mM and stored at 4°C. The BACE1 inhibitor LY2811376 was solubilized in DMSO to a stock concentration of 10 mM and stored at –80°C. Further dilutions of these compounds were routinely prepared in DMSO (unless indicated otherwise) and the final concentration of DMSO in media during all cell treatments or in all enzymatic assay solutions did not exceed 0.4%.

Cell treatments for analysis of AβPP processing

Cells were routinely plated and treated as described below unless otherwise indicated in the figure legends. Cells were plated at a density of 2×10⁵ cells per well in six-well cluster dishes in 2 mL of complete media and grown for 16–24 h. Cells were washed twice with DMEM only and re-fed with 1 mL of complete media containing pCDPs, inhibitors or DMSO and incubated for 24 h. Cells were washed, re-treated for an additional 24 h and the conditioned media (CM) was collected and cleared by centrifugation at 12,000× g for 20 min at 4°C to remove intact cells and cellular debris. An aliquot of the cleared CM was combined with 4X Laemmli SDS-sample buffer containing 25 mM dithiothreitol (DTT) and used for the analysis of secreted AβPP fragments by immunoblotting. This same 48-h cell treatment procedure was performed to assess cell viability using a TC20 automated cell counter (BioRad), where live cells were distinguished by selective Trypan Blue staining of dead cells.

Preparation of cell lysates

Following the removal of the CM as described above, cell monolayers were washed twice with icecold phosphate buffered saline (PBS) solution to remove residual media and lysed in 200–250 μL of lysis buffer (LB; 50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), pH 7.2, 150 mM NaCl, 2 mM Ethylenediaminetetraacetic acid (EDTA), 1% Triton X-100, 1 mM sodium orthovanadate, 50 mM sodium fluoride, 10 mM β-glycerophosphate, 500 μM 4-(2-Aminoethyl)benzenesulfonyl fluoride hydrochloride (AEBSF), 10 μg /mL Aprotinin, 10 μg/mL Leupeptin, and 5 μg/mL Pepstatin) for 30 min at 4°C with rocking. Crude lysates were collected and centrifuged at 28,000× g for 20 min at 4°C to remove cellular debris and insoluble material. The clarified whole cell lysates were removed, total protein was quantified using the DC Protein Assay with Bovine Serum Albumin (BSA) as a standard and then combined with 4X Laemmli SDS-sample buffer containing 25 mM DTT for analysis by immunoblotting.

Time course of pCDP-diBzl treatment

To obtain roughly the same protein levels and to harvest CM for all time points simultaneously, all cells were plated at the same time at a density of 2×10⁵ cells per well in six-well cluster dishes in 2 mL of complete media and grown for 16–24 h. All wells were then washed twice with DMEM only and two wells were re-fed with 1 mL of complete media containing 400 μM of pCDP-diBzl, two wells were re-fed with 1 mL of complete media containing DMSO only, and the rest of the wells were re-fed with complete media only (48-h time point). After 24 h, media was removed from four currently untreated wells, washed with DMEM only, and then two wells were re-fed with 1 mL of complete media plus 400 μM of pCDP-diBzl while the other two wells were re-fed with 1 mL of complete media plus DMSO only (24-h time point). At approximately the same time, the cells for the 48-h time point were washed and re-fed with media containing DMSO or 400 μM of pCDP-diBzl for another 24 h (it was from this final 24-h collection period that the 48-h time point was collected). The same process was followed after an additional 12, 16, and 20 h to obtain the 12-, 8-, and 4-h treatment time points, respectively. At the end of the 48-h experiment, all CM was collected and cleared by centrifugation at 12,000× g for 20 min at 4°C to remove intact cells and cellular debris. An aliquot of the cleared CM was combined with 4X Laemmli SDS-sample buffer containing 25 mM DTT and used for the analysis of secreted AβPP fragments by immunoblotting. Cells were washed and lysed as described above.

Sodium dodecyl sulfate polyacrylamide gel electrophloresis (SDS-PAGE) and immunoblotting

Whole cell lysates (WCLs) and CM were prepared as described above and various amounts were analyzed by immunoblotting as indicated in the figure legends. Samples and protein standards were adjusted as needed to the desired total protein amount using LB (for WCL samples) or to equal volume using growth media (for CM samples). Housekeeping protein β–tubulin was used as a loading control for all WCL blots except those corresponding to treatments of H4-BACE1 cells, where Stain Free total protein provided a more reliable loading control at the high total protein loading level that was needed to detect endogenous AβPP [37].

All CM samples (except for sAβ, described below) were separated on 10% SDS-PAGE gels [38] and transferred to PVDF membrane without methanol [39] with constant cooling. Membranes were blocked with Tris-Buffered Saline with Tween (TBST; 25 mM Tris-HCl pH 7.2, 150 mM NaCl, and 0.1% Tween 20) containing 5% milk for 1-2 h at RT and then incubated with primary antibodies (diluted in TBST containing 0.5–1% BSA) at the following dilutions: anti-sAβPPα (EPR5119(2), 1:15,000) or anti-sAβPPβ (25 ng/mL). Anti-sAβPPα was incubated for 2-3 h at RT and anti-sAβPPβ antibody was incubated overnight at 4°C. Membranes were incubated for 2-3 h at RT with the appropriate HRP-conjugated secondary antibody (diluted 1:10,000 in TBST containing 1% milk). Pierce ECL Western Blotting substrate (for sAβPPα) or Pierce SuperSignal West Pico substrate (for sAβPPβ) were used to visualize the results on X-ray film. Band intensities were quantified by ImageJ and a background area of the same size was taken above each band and subtracted. These resulting values were first normalized to the averaged DMSO control for each protein and then further normalized for the total protein of the corresponding WCL for each sample (as determined by the DC Protein Assay).

All WCL samples were separated on 4–15% or 4–20% Mini Protean TGX Stain Free precast gels (Bio-Rad). Covalent coupling of the chromophore in the gel to the separated proteins was achieved using an activation time of 1 min. Proteins were transferred to Immun-Blot LF PVDF membrane with 10% methanol for 2 h at constant current of 250 mA [39]. Membranes were blocked with TBST containing 5% milk for 2 h at RT or overnight at 4°C. Each membrane was washed four times with TBST (5 min per wash). Membranes were incubated with anti-AβPP (Y188, 1:40,000) overnight at 4°C or for 2.5 h at RT, with anti-β–tubulin (1:30,000) for 2 h at RT, or with anti-BACE1 (1:5,000) for 4 h at RT. Membranes were incubated for 1.5–3 h at RT with the appropriate HRP-conjugated secondary antibody as described above.

For WCL blots of treated H4-BACE1 cells, proteins transferred membranes were first imaged using a stain free blot application on a Chemidoc MP system (Bio-Rad), then were incubated with Pierce SuperSignal West Pico Chemiluminescent Substrate for visualization. For all other WCL blots, Pierce SuperSignal West Pico Chemiluminescent Substrate (for comparison of AβPP levels across different cell models), Clarity Western ECL (for comparison of BACE1 levels across different cell models), or Pierce ECL Western Blotting Substrate (for AβPP and β-tubulin in H4-APP695 cells) were used for visualization. All blots were visualized using Chemi Hi Sensitivity blot application in the Chemidoc MP System (Bio-Rad). Band volumes were quantified using Image Lab (Bio-Rad). For use of β-tubulin as a loading control, desired protein band volumes in a given lane were normalized to the β-tubulin signal volume in that lane. For use of stain free total protein as a loading control, desired protein band volume(s) in each lane were normalized to total protein as quantified using Image Lab (Bio-Rad). For all WCL blots of FL-AβPP where mature and immature FL-AβPP were adequately resolved to allow quantification of each, previously normalized band volumes (for β-tubulin signal or stain free total protein) were further normalized to the average control immature AβPP band volume, as shown in the correspondinggraphs.

Total secreted Aβ levels were determined by immunoblotting using the antigen epitope retrieval method essentially as described [40]. Briefly, 20 μL of conditioned media was separated in a 12.5% Bis-Tris-Mes gel and transferred to 0.2 μM nitrocellulose membranes for 90 min at 40 V using the BioRad Trans blotting system containing 20% methanol [39]. Membranes were steamed for 15 min to expose latent epitopes, blocked with TBST-milk and incubated with anti-Aβ antibody (1:4000) for 1 h at RT followed by overnight incubation at 4°C. Membranes were incubated with HRP-conjugated secondary antibody and results were visualized using the Pierce ECL Western Blotting substrate. Band intensities were quantified by ImageJ as describedabove.

Observation of pCDP uptake by LCMS

The following LCMS procedure for cell lysates was adapted from [41]. H4-AβPP695 cells were plated at a density of 1.5×10⁶ cells per 10-cm dish (x20) in complete media and grown at 37°C in a humidified atmosphere (90%) containing 10% CO₂. After ∼24 h, cells were re-fed with 5 mL of fresh media containing various pCDPs (400 μM) or DMSO as indicated in the figure legend (two plates per condition). The remaining ten plates were re-fed with 5 mL of fresh media and were used as a “spiked” positive control set. After 24 h, the treated cells were scraped into their conditioned media (duplicates were combined) and collected by centrifugation for 5 min at 2000× g at RT. The remaining 10 untreated dishes (“spiked” positive control set) were scraped, divided into five equal aliquots and cells were collected by centrifugation. Cell pellets were washed three times with cold (4°C) PBS to remove all residual conditioned media and pCDPs. Cell pellets were quick frozen in liquid nitrogen and stored at –80°C until lysis.

Frozen pellets were thawed on ice for 30 min, re-suspended in 500 μL of water and 500 μL of methanol, vortexed, and incubated on ice for 1 h to facilitate complete lysis. The five “spiked” positive control samples were prepared by adding 1 μL of 1 mM for each pCDP or DMSO (1:100 dilution of the stocks in water) to individual pellets, vortexed and incubated on ice for 1 h. Lysates were centrifuged at 15,000 RPM for 30 min at 4°C and the supernatants were collected and lyophilized to dryness. Each dried sample was dissolved in 300 μL of 1:1 methanol/water and 2 μL was separated using reverse-phase high resolution UHPLC-MS exactly as described above (see LCMS Analysis section of Analysis of custom-synthesized compounds). The Thermo-Scientific Xcalibur software package was used to visualize, analyze, and depict the LCMS data presented here.

BACE1 activity assay

BACE1 activity was determined by incubating 400 μM of individual pCDPs, the BACE1 inhibitor LY2811376 (positive control) or DMSO (negative control) in a reaction mixture containing 25 mM sodium acetate assay buffer, pH 4.42, 1% BSA, 0.2 μg of the catalytic domain of recombinant human BACE1 and 10 μM of Methyl cumaryl amide (Mca) fluorogenic substrate in a final volume of 110 μL. Fluorescence intensity was measured with a Synergy H1 hybrid reader (BioTek, Winooski, VT) (excitation 320 nm, emission 405 nm) using a black microplate with half-area wells and an opaque bottom with continuous gentle shaking at 37°C for 1 h with readings acquired every 5 min. Results were analyzed in Excel 2010 by averaging the DMSO control 1 h fluorescence intensity values and then normalizing all experimental results for pCDPs and BACE1 inhibitor to this value. The normalized data was then averaged over repeated points (n = 6 for all pCDP treatments and DMSO negative control, n = 5 for BACE1 inhibitor data), the standard deviations were determined, and the data were graphed using Excel 2010 software.

Statistical analysis

Averages, 20% Coefficient of Variation (CV), standard deviation (S.D.), and statistical analysis for significance via the unpaired, independent Student’s T test for data were determined using Excel 2010. CM and WCL data were expressed as the mean±20% CV. CM data were deemed significant at p < 0.05. 20% CV was used as an estimate for western blot variability to yield more conservative assignments of statistical significance [37]. Cell viability data were expressed as mean±S.D. and were deemed significant at p < 0.05. All graphs were generated in Excel 2010. Approximate EC50 values were determined by fitting the sAβPPβ level versus dose curves to the 4-parameter logistic model Y = (d - a)/{1 + (X/c) ^b}, where a is the lower asymptote, d is the upper asymptote, X is the dose concentration, c is the concentration at which the sAβPPβ level is midway between a and d, and b is the slope factor that describes the steepness of the central linear portion of the curve [42]. Parameters a, b, c, and d were optimized by minimizing the sum of the squared differences between experimental and model-predicted values using the Solver add-in within Excel 2010 (Microsoft).

H4 neuroglioma cells stably overexpressing AβPP695 or BACE1 enable reliable detection of sAβPPβ

BACE1 cleavage of AβPP produces sAβPPβ and C99 fragments (Fig. 1A) and is essential for the production of Aβ. The C-terminal fragment, C99, is a transient intermediate in the amyloidogenic pathway that reflects the relative balance of multiple pathways. C99 is processed by γ-secretase [43], is also a substrate for α-secretase [44], and is turned over by both ERAD and ubiquitin-independent lysosomal degradation pathways [45]. Moreover, C99 can be additionally cleaved to C89 by BACE1 and BACE2 [46], and the proportion of C89 relative to C99 increases with increasing BACE1 expression [47]. In contrast, sAβPPβ (∼100 kDa) is secreted to the medium and can be detected (and distinguished from sAβPPα) by antibodies specific for the C-terminus of sAβPPβ [48]. The sAβPPβ fragment therefore serves as an effective proxy for the maximum possible amyloidogenic processing; for every Aβ peptide molecule produced, a corresponding sAβPPβ fragment must be generated. Importantly, sAβPPβ reflects the amount of AβPP that undergoes the entry step (i.e., β-secretase cleavage) into the amyloidogenic pathway (Fig. 1). Our goal was to generate two distinct cell lines, via overexpression of AβPP695 or BACE1 that recapitulate two distinct causative mechanisms of AD [16, 49], to enable reliable detection of sAβPPβ for evaluating the effects of pCDP treatment on AβPP amyloidogenic processing.

The human H4 neuroglioma cell line, derived from a neuroglioma tumor and adapted for growth in culture [50, 51], has been broadly used in AD research [52–55]. In H4 cells, the endogenous AβPP751 isoform (Fig. 3A) predominantly undergoes nonamyloidogenic processing, as shown by the absence of sAβPPβ in 15 μL of CM (Fig. 3B) and the abundance of sAβPPα detected in just 2 μL of the same CM (Fig. 3C). To generate detectable levels of sAβPPβ, we chose to take two separate approaches: (1) increase the level of the AβPP695 isoform, since this is the predominant form found in neurons [56] and overexpression of AβPP695 is known to increase amyloidogenic processing [57], and (2) increase the level of BACE1 to shift the balance toward amyloidogenic processing [49].

To this end, we generated two distinct H4 cell lines that serve as AD models: the H4-AβPP695 cell line that overexpresses AβPP695 (wild-type) and the H4-BACE1 cell line that overexpresses BACE1 (wild-type) (Fig. 3A). The quantity of AβPP695 is increased approximately 15-fold in H4-AβPP695 lysates compared to endogenous AβPP751 in H4 and H4-BACE1 cell lysates, using β-tubulin for normalization across lanes (Fig. 3A). BACE1 is robustly detected in 6.8 μg of H4-BACE1 cell lysate, whereas endogenous BACE1 was not detected in 6.8 μg of H4 cell lysate (Fig. 3A). Additionally, the level of mature AβPP751 is reduced in the H4-BACE1 cells (Fig. 3A), which is consistent with previous results that have shown that most cleavage of AβPP via BACE1 happens after O-glycosylation [58].

As anticipated, both H4-AβPP695 and H4-BACE1 cell lines enable sensitive detection of sAβPPβ (Fig. 3B) and produce different relative levels of sAβPPα (Fig. 3C). With only 2 μL of CM, sAβPPα is detected in H4-AβPP695 cells after a 3-s exposure time (Fig. 3C, top blot). After a 2-min exposure of the same blot, the less abundant sAβPPα from both H4 and H4-BACE1 cells is detectable (Fig. 3C, bottom panel) at the expected (different) molecular weights. Due to robust overexpression of AβPP695 in the H4-AβPP695 cell line, the highly abundant sAβPPα secreted into CM (Fig. 3C) is not conducive to observing small changes in nonamyloidogenic processing. However, since AβPP expression remains at an endogenous level in the H4-BACE1 cell line, changes in both sAβPPβ (Fig. 3B) and sAβPPα (Fig. 3C) can be detected in CM, thereby allowing amyloidogenic and non-amyloidogenic processing of AβPP to be simultaneously monitored in this cell line. Collectively, these data demonstrate the utility of H4-AβPP695 and H4-BACE1 cells for monitoring sAβPPβ as a proxy for the amyloidogenic processing of AβPP.

The cis-locked pCDP mimic was derivatized to aid in cellular uptake

Since the cis-locked pCDP (Fig. 1C) is charged and is not expected to be taken up by cells, we prepared three derivatives that block the phosphate group and are more hydrophobic (Fig. 2). The phosphate group in pCDP (Fig. 2A) was protected by a single benzyl group (Fig. 2B), two benzyl groups (Fig. 2C), or two pivaloyloxymethyl (POM) groups (Fig. 2D). Notably, since the POM groups are easily removed by cellular esterases [59], the pCDP-diPOM molecule is a “pro-drug” version of the pCDP (Fig. 2A). All pCDPs were analyzed by ¹H NMR and LCMS to verify their structure and MW (Supplementary Figures 1–5).

Only pCDP-diBzl reduces sAβPPβ in the conditioned media of H4-AβPP695 cells

Using sAβPPβ as a proxy for monitoring amyloidogenic processing of AβPP, the effects of the pCDP compounds (Fig. 2) were investigated in H4-AβPP695 cells. H4-AβPP695 cells were treated with 400 μM of pCDP, pCDP-Bzl, pCDP-diBzl, and pCDP-diPOM for a total of 48 h and 15 μL of CM from the final 24 h of treatment was analyzed by western blot for changes in sAβPPβ (Fig. 4A). Only pCDP-diBzl had a potent effect on the quantity of sAβPPβ present in the media, severely reducing the amount of detectable sAβPPβ to less than 1% of the control without a notable change in the levels of mature or immature FL-AβPP (Fig. 4A). The well-characterized BACE1 inhibitor (LY2811376) was also used to treat cells at 2.5 μM (Fig. 4A) and exhibited a similar effect on H4-AβPP695 cell line as the pCDP-diBzl. The other CDP variants (pCDP, pCDP-Bzl, pCDP-diPOM) had little to no effect on the amount of sAβPPβ, despite their related structure. To assess the impact of pCDP-diBzl treatment on cell viability, H4-AβPP695 cells were treated with 200 μM and 400 μM of pCDP-diBzl for 48 h, trypsinized, stained with Trypan Blue, and counted. Although growth seemed to be slowed in the presence of 200 μM and 400 μM pCDP-diBzl, no significant toxicity was observed (>96% of all cells counted were alive) (Supplementary Figure 6).

A dose-dependence experiment using 25, 50, 100, 200, and 400 μM of pCDP-diBzl was performed to monitor changes in sAβPPβ in H4-AβPP695 cells (Fig. 4B). As shown, 48 h of treatment with 400 μM of pCDP-diBzl has the most potent impact on sAβPPβ levels (Fig. 4B) with an approximate EC50 value of 126 μM (Fig. 4B). As described above, the robust production of sAβPPα in these cells (Fig. 3C) prevents reliable evaluation of small changes in the non-amyloidogenic pathway. The mature and immature AβPP signal for the DMSO and 400 μM pCDP-diBzl samples from Fig. 4A and B were combined (n = 4) and, for this highest level of treatment with pCDP-diBzl, no significant impact on full length AβPP (FL-AβPP) levels was observed (Fig. 4C).A same dose-dependent response was also observed on sAβ levels in H4-AβPP695 cells (Fig. 4D) with no significant impact on FL-AβPP signal. Additionally, a time-dependence study was performed to see if 48 h of treatment was crucial for the observed effect on sAβPPβ reduction (Fig. 4E). A short exposure time (30 s) shows clear reduction of sAβPPβ in CM of pCDP-diBzl treated cells for 12, 24, and 48 h total time periods (Fig. 4E, left). A much longer exposure (10 min) reveals additional observable reduction of sAβPPβ at 4 h (Fig. 4E, right), whereas sAβPPβ was not yet detectable even in untreated cells at 2 h (data not shown). These data demonstrate that pCDP-diBzl treatment of H4-AβPP695 cells reduces secreted sAβPPβ levels without significantly changing FL-AβPP levels, and that this reduction is evident at the earliest time point at which sAβPPβ can be detected.

pCDP-diBzl and pCDP-diPOM reduce sAβPPβ in the conditioned media of H4-BACE1 cells

As was observed in the H4-AβPP695 cells, treatment of H4-BACE1 cells with 400 μM of pCDP-diBzl yielded a similar decrease in sAβPPβ without a significant impact on FL-AβPP levels (Fig. 5A). Additionally, sAβPPα (derived from endogenous AβPP751 in these cells) does not significantly change upon treatment with any of the pCDP compounds, and shows no significant dose-dependence with pCDP-diBzl treatment (Fig. 5). The dose-dependent reduction of sAβPPβ induced by pCDP-diBzl treatment of H4-BACE1 cells (Fig. 5B) is similar to what was observed in H4-AβPP695 cells, with an EC50 value of approximately 67 μM (Fig. 5B). Again, mature and immature FL-AβPP signal for the DMSO and 400 μM pCDP-diBzl samples from Fig. 5A and B were combined (n = 4) to show no significant impact on FL-AβPP levels (Fig. 5C). The similar dose dependence and EC50 values in these distinct cell lines, where either AβPP or BACE1 are significantly overexpressed, indicate that pCDP-diBzl does not act through direct competition for a binding partner of either AβPP or BACE1. These data further suggest that pCDP-diBzl might act in both distinct cell lines via a common mechanism to reduce sAβPPβ. In contrast, while 400 μM pCDP-diPOM treatment of H4-AβPP695 cells showed no significant effect, treatment of H4-BACE1 cells with 400 μM pCDP-diPOM significantly reduced the amount of sAβPPβ without changing FL-AβPP signal (Fig. 5A, B). Although this 5-fold reduction is approximately the same as the effect of pCDP-diBzl treatment in these cells, the observation that pCDP-diPOM is ineffective in cells overexpressing AβPP is consistent with a mechanism of action in which pCDP-diPOM blocks AβPP interaction with an amyloidogenic binding partner.

As expected, treatment with 2.5 μM BACE1 inhibitor LY2811376 again inhibits sAβPPβ production in H4-BACE1 cells. Interestingly, a clear, significant increase in both sAβPPα and mature (glycosylated) full length AβPP is observed in this cell model, where only endogenous AβPP751 is expressed (Fig. 5A). This suggests that direct inhibition of BACE1 catalytic cleavage of AβPP might increase recycling of AβPP back to the plasma membrane via trafficking of endocytosed AβPP back to the trans-Golgi network and subsequent secretory pathway, where glycosylation and β-secretase activities are active. This effect is not observed in treatments with 400 μM pCDP-diBzl or 400 μM pCDP-diPOM and suggests a different mechanism for these molecules.

Association of pCDPs with and their conversion by H4-AβPP695 cells

To gain insight regarding the fate of pCDPs inside cells, LCMS was used to detect the presence of each compound in lysates from H4-AβPP695 cells treated with the various pCDPs. Importantly, these lysates were extracted using 1:1 methanol/water by volume, which is expected to at least partially solubilize most polar lipids and to denature some proteins. Although the pCDP variants are chemically very similar, LCMS was highly effective for detecting the pCDP variant and its conversion products. The neutral-charge MW of each pCDP molecule is shown in Fig. 2 and the electrospray ionization in positive mode (ESI+) MW for each pCDP is the following: 279 for pCDP, 369 for pCDP-Bzl, 459 for pCDP-diBzl, and 507 for pCDP-diPOM.

For comparison, lysates from cells treated with 400 μM of each pCDP variant (Fig. 6A) and lysates from untreated cells that were spiked with 100 μM of respective pCDPs (Fig. 6B) were analyzed by LCMS. Lysate from cells treated with 400 μM of pCDP-diBzl displays not only the expected MW 459 but also a substantial amount of MW 369, corresponding to pCDP-Bzl. In contrast, lysate from untreated cells spiked with pCDP-diBzl shows predominantly MW 459 and significantly lower MW 369. This suggests that pCDP-diBzl is converted to pCDP-Bzl inside the cell. Similarly, treatment with 400 μM of pCDP-diPOM shows an elevated amount of MW 279 and no detectable MW 507, while the untreated cell lysate spiked with pCDP-diPOM shows a substantial peak only for MW 507 (Fig. 6). This supports successful entry of pCDP-diPOM into cells where active esterases efficiently cleave the POM groups in vivo [59], whereas esterase activity in the spiked case could be inhibited in 1:1 methanol/water. Lysate from pCDP-Bzl treated cells does show MW 369, demonstrating that the pCDP-Bzl molecule does associate with cells, although as shown above it does not significantly reduce sAβPPβ (Fig. 4A). Cells treated with pCDP do not show MW 279 significantly above the DMSO control, but spiked cells with pCDP do, indicating that although pCDP itself is detectable it does not associate with treated cells sufficiently enough to be detected. Together, these results show that pCDP-diBzl, pCDP-Bzl, and pCDP-diPOM associate with H4-AβPP695 cells, and that pCDP-diBzl and pCDP-diPOM are modified bythe cell.

pCDPs do not specifically inhibit the catalytic activity of BACE1

Since a reproducible effect on reduction of sAβPPβ by pCDP-diBzl and pCDP-diPOM was observed, we sought to determine whether these compounds directly inhibit the activity of the BACE1 catalytic domain. Using recombinant human BACE1 catalytic domain and an intra-molecularly quenched Mca fluorogenic substrate as part of a FRET-based assay [60], we tested 400 μM of each pCDP variant for inhibition of BACE1catalytic activity. Additionally, we tested the known BACE1 inhibitor LY2811376 at 2.5 μM as a positive control for reduced BACE1 catalytic activity. As shown in Fig. 7, while LY2811376 showed a potent decrease as expected, we saw no significant decrease in the catalytic activity of recombinant, purified BACE1 catalytic domain with any of pCDP variants at 400 μM (as measured by fluorescence intensity). These results demonstrate that pCDP-diBzl and pCDP-diPOM do not act as a direct inhibitor of BACE1 catalytic activity, and point to an alternate mechanism by which they inhibit amyloidogenic AβPP processing in H4 cell models of AD.