Beta-amyloid interacts with and activates the long-form phosphodiesterase PDE4D5 in neuronal cells to reduce cAMP availability

Inhibition of the cyclic-AMP degrading enzyme phosphodiesterase type 4 (PDE4) in the brains of animal models is protective in Alzheimer’s disease (AD). We show for the ﬁrst time that enzymes from the subfamily PDE4D not only colocalize with beta-amyloid (A b ) plaques in a mouse model of AD but that A b directly associates with the catalytic machinery of the enzyme. Peptide mapping suggests that PDE4D is the preferential PDE4 subfamily for A b as it possesses a unique binding site. Intriguingly, exogenous addition of A b to cells overexpressing the PDE4D5 longform caused PDE4 activation and a decrease in cAMP. We suggest a novel mechanism where PDE4 long-forms can be activated by A b , resulting in the attenuation of cAMP signalling to promote loss of cognitive function in AD

Cyclic-AMP (cAMP) signalling is a crucial pathway for memory formation/cognition and the down-regulation of this second messenger in the brain during Alzheimer's disease (AD) is thought to play a part in the cognitive deficits that are a characteristic of the disease [1,2].cAMP-specific enzymes from the phosphodiesterase type 4 (PDE4) family have been identified as key players in shaping cerebral cAMP gradients that activate cAMP response element-binding protein (CREB) via PKA phosphorylation (reviewed in [3]).Several studies involving pharmacological inhibition [4][5][6], RNA interference [7,8], dominant-negative PDE4 transfection [9], PDE4 knock-out mice [7] and CRISPR-Cas9 [10] have indicated that isoforms from the PDE4D subfamily are most influential as targets for therapeutic strategies.Recent evidence also supports the role of PDE4B in this regard [11].Said strategies counteract the aberrant cAMP signalling that results in a down-regulation of CREB transcription factor activity and subsequent loss of synaptic plasticity.These observations have been supported by recent evidence from diseased human brains that show increased PDE4D expression when Abbreviations AD, Alzheimer's disease; APP/PS1, amyloid precursor protein/presenilin1; Ab, b-amyloid; cAMP, cyclic AMP; CREB, cAMP response element-binding protein; FRET, fluorescence resonance energy transfer; GST, glutathione S-transferase; IBMX, 3-isobutyl-1-methylxanthine; mRNA, messenger RNA; PDE, phosphodiesterase; PDE4, phosphodiesterase type 4; PKA, protein kinase A; UCR, upstream conserved regions.
compared with controls [12].Although mRNA transcripts [12] and western blotting [13] have been used to demonstrate elevated levels of PDE4D expression in AD models, there have been no attempts to look at the activation state of PDE4D enzymes during neurodegenerative disease.PDE4 enzymes, especially those designated as longforms, have an intrinsic activity that can be enhanced or inhibited by post-translational modification [14][15][16] or association with peptides [17] or lipids [18].Here, we report that PDE4D enzymes colocalize with beta-amyloid (Ab) plaques in brains of amyloid precursor protein/presenilin1 (APP/PS1) mice and that a direct association takes place between Ab and the PDE4 enzyme.We provide peptide mapping evidence to show why PDE4D isoforms may have more relevance to AD than the other sub-families (PDE4A, B and C) and suggest that the low cAMP concentrations observed in AD brains may be a result of the activation of PDE4D longforms caused by direct Ab association.

Co-localization of PDE4D and Ab
To investigate the potential formation of complexes between PDE4D and Ab, mouse brain sections obtained from 7-month-old APP/PS1 Alzheimer's mice (MMRRC strain #034832-JAX) were subjected to dual staining for both markers.Initially, an antigen retrieval step was carried out by incubating the sections with 70% formic acid for 15 min, followed by thorough washing in tris-buffered saline (TBS).Subsequently, the sections were incubated overnight at 4 °C with a concentration of 2 lgÁmL À1 rabbit anti-PDE4D (Abcam, Ab14613, Cambridge, UK) diluted in 0.3% TBS-T.Following the primary antibody incubation, sections were treated with a donkey anti-rabbit biotin secondary antibody (1/400 dilution in 0.3% TBS-T) (Thermo-Fisher, Loughborough, UK) for 1 h at room temperature.After completing TBS washing steps, sections were subjected to incubation with streptavidin-647 (1/500 in 0.3% TBS-T) (ThermoFisher) for an additional 1 h at room temperature to visualize PDE4D.Subsequent to the PDE4D visualization, the sections were further incubated overnight at 4 °C with a primary mouse anti-human Ab, 17-24 antibody (clone 4G8) (BioLegend, San Diego, CA, USA, Ab800712) at a dilution of 1/500 in 0.3% TBST-T.The following day, sections were washed and subsequently incubated with a secondary anti-mouse AlexaFluor 488 antibody (1/250 in 0.3% TBS-T) (Invitrogen, Inchinnan, UK) for 1 h at room temperature.A Hoechst counterstain was performed to visualize cell nuclei.
To further validate PDE4D and Ab colocalization within neurons, SHSY5Y cells (sourced from ATCC within last 3 years) (RRID:CVCL_0019) were plated at 50 000 cells on glass coverslips and cultured in DMEM/F-12 (Gibco, Grand Island, NY, USA) supplemented with 10% foetal bovine serum, 1% L-glutamine, 1% nonessential amino acids and 1% penicillin/streptomycin mixture until ~80% confluency.Cells were routinely checked for mycoplasma.Cells were then treated with 5 lM FAM-Ab (Anaspec, Kidderminster, UK) for 1 or 24 h and fixed in 4% paraformaldehyde.Blocking with 1% BSA and 0.1% Tween80 was performed, followed by overnight incubation with a primary PDE4D antibody (Abcam, ab14613) and a 1 h incubation with an Alexa fluor 555 secondary antibody (Invitrogen, A31572).Counterstaining was done using DAPI dye and coverslips were mounted using fluoromount (Invitrogen).Samples were imaged using the Zeiss LSM900 confocal microscope using a 639 oil objective.Z-stacks were obtained and processed in ZEN 3.4 LITE (Zeiss, Cambridge, UK) and FIJI IMAGE J software (Fuji, Madison, WI, USA).Colocalization of PDE4D and FAM was determined using the IMAGE J colocalization threshold plugin.

Fluorescence polarization
Fluorescence polarization measurements were performed on a Mithras LB 940 plate reader (Berthold Technologies, Bad Wildbad, Germany) with the excitation and emission wavelengths 485 and 535 nm respectively.All FAM-Ab peptides  were synthesized by AnaSpec and dissolved in DMSO to a stock concentration of 10 mM.The assay was formatted using 10 lL reaction volume per well in non-binding, black 384-well plates (Cat no.# 262260) and Thermo Fisher Scientific Assay buffer (PBS, 1 mM DTT and 0.25% Tween-20) was used to dilute all ingredients.All polarization values are expressed in millipolarization units (mP).A fixed concentration of 10 lM FAM-Ab (1-40) was used with increasing amounts of GST and GST-PDE4D5.Reactions were incubated for 30 min at room temperature in the dark.In order to examine the existence of non-specific binding, GST protein was used as a negative control.All the experiments were independently repeated at least thrice.

Peptide array
The PDE4 isoform sequence peptide arrays were synthesized as sequential 25mers shifted by 5 amino acids via SPOT synthesis [19] on continuous cellulose membranes using Fmoc-chemistry with a MultiPep 2 instrument (CEM Corporation, Matthews, NC, USA).For the alanine scanning arrays, versions of arrays were synthesized to incorporate alanine residues in place of the endogenous amino acid.In the event of alanine being the original residue, an aspartic acid or glycine was incorporated.The membranes were blocked with 5% milk/TBST (w/v) for 1 h.The PDE4 arrays were then overlaid with either Ab 1-42 or Ab scr (Anaspec) overnight at 4 °C.The arrays were then analysed utilizing a far-western immunoblotting approach.Analogous methods were used to probe overlapping Ab 1-42 arrays with GST-fusion PDE4D5 protein in order to determine which domains within Ab 1-42 are responsible for binding PDE4s.Specifically, Ab 1-49 arrays were overlain with PDE4D5-GST and a mouse monoclonal GST-HRP (Sigma, A7340, Lee on the Solent, UK) was used at 1 : 5000 for 2 h at room temperature to detect binding.

Co-immunoprecipitation
SHSY-5Y cells cultured in six-well plates were treated with either 10 lM Ab 1-42 or Ab scr for 24 h.The neurotoxin 10 lM Ab 1-42 derivatives were created as previously described [20].Cellular lysates were prepared in lysis buffer [25 mM HEPES, 2.5 mM EDTA, 50 mM NaC1, 50 mM NaF, 30 mM sodium pyrophosphate, 10% (v/v) glycerol and 1% (v/v) Triton X-100, pH 7.5, containing Complete TM EDTA-free protease inhibitor cocktail tablets (Roche, Welwyn Garden City, UK)] after the treatment.Protein concentration of lysates was determined using the Bradford assay and all samples were equalized for protein concentration (400 lg protein per IP reaction).Goat anti-Pan-PDE4D antibody (in-house) was used to immunoprecipitate endogenous Ab.The resulting immunocomplexes were captured using 25 lL of Protein G magnetic beads per sample (Pierce #88847, Huddersfield, UK) at 4 °C overnight with mixing.Normal goat IgG (Bio-Techne Ltd, #AB-108-C, Minneapolis, MN, USA) was used as a mock IP control.The beads were washed three times using lysis buffer.Bound proteins were then eluted in SDS/PAGE sample buffer and subjected to SDS/PAGE for immunoblotting using mouse anti-Ab antibody (Sigma #A8354, 1 : 5000), followed by goat anti-Pan-PDE4D antibody (in-house, 1 : 5000) to confirm protein input.Immunoreactive proteins were detected by IRDye 680RD donkey anti-mouse IgG (Li-COR, #926-68072, Lincoln, NE, USA) and Alexa Fluor 790 donkey anti-goat IgG (Abcam, #ab175784) respectively.Images were acquired using Li-Cor Odyssey CLx Imaging System and signals were detected at 700 and 800 nm channels.

Proximity ligation assay
SH-SY5Y cells cultured in eight-well chamber slides (Falcon #354118; Fisher Scientific, Waltham, MA, USA) were treated with either 10 lM Ab 1-42 or Ab scr (Anaspec) for 24 h.The cells were fixed with 4% (v/v) paraformaldehyde in PBS for 15 min at room temperature.Cells were counterstained with cell surface marker wheat germ agglutinin (WGA) conjugated to AlexaFluor 488 (Invitrogen #W11261) for 5 min, followed by permeabilization using 0.1% Triton X-100 (Sigma-Aldrich) in PBS for 10 min at room temperature.In situ detection of the exogenous Ab and PDE4D protein-protein interaction was carried out utilizing Duolink Ò proximity ligation assay [21] as per manufacturer's instructions (Duolink Ò , Merck, Irvine, UK).Equal concentrations (1 : 4000) of immunocytochemistryvalidated PDE4D (goat) and Ab (mouse) primary antibodies were used in combination with respective Duolink Ò PLA anti-goat (PLUS) and anti-mouse (MINUS) probes.Slides were finally mounted under coverslips with Prolong Gold Antifade reagent with DAPI (Invitrogen, P36935) and visualized.Images were acquired using an upright Zeiss LSM 880 confocal laser scanning microscope under a 639 oil immersion objective (excitation 594 nm and emission 624 nm).In order to detect all PLA signals, a series of Z-stack images were collected and analysed by IMAGEJ software.

PDE4 activity assays
An expression PCDNA3 plasmid encoding human PDE4D5-VSV was used before by us [22], which was prepared using the Maxi-prep system (Qiagen, Dusseldorf, Germany).For transient transfections, SH-SY5Y cells were transfected using PolyFect Ò transfection reagent (Qiagen) in accordance with manufacturer's instructions.Cells (~90-100% confluent) were transfected for 48 h with cDNA-encoding PDE4D5 and treated for 6 h with Ab 1-42 or Ab scr .Cells were then washed with PBS and harvested by using a cell scraper in KHEM buffer (50 mM KCl, 50 mM HEPES; pH 7.2, 10 mM EGTA, 1.92 mM MgCl 2 and 1 mM dithiothreitol (DTT)) supplemented with protease inhibitor Mini-Complete (Roche).Samples were then frozen on solid CO 2 , thawed and then manually homogenized, followed by passage through a 26-gauge needle several times to ensure complete cell lysis.Cells were centrifuged at 20 784 g for 10 min to remove any unbroken cells, and the resulting supernatant was frozen in solid CO 2 and stored at À80 °C until required.For experimentation, the protein concentration of whole-cell lysate from transfected and mock-transfected (vector only) cells was equalized (typically to 1 lgÁlL À1 ).Protein concentration was determined through Bradford assay using bovine serum albumin as standard.PDE activity was determined using a two-step radioassay procedure as described previously [23].Activities were related to a non-treated sample (100% control) over an increasing dose of the Ab 1-42 or Ab scr .In all cases, the transfected PDE accounted for over 97% of the total PDE activity when compared with the untransfected control lysates.

cAMP FRET reporter assay
Assays were conducted as previously described [24].HEK293 cells were seeded onto sterile glass coverslips and incubated for 24 h.The cells were transiently transfected with a cAMP monitoring FRET sensor based on the structure of EPAC1 (EPAC1-cAMPs) [25] using Lipofectamine LTX (Invitrogen).FRET imaging was performed 24 h following transfection and 60 min following pre-incubation with DMSO, Ab 1-42 or Ab scr .For imaging, the cells were buffered in a solution of 125 mM NaCl, 5 mM KCl, 20 mM HEPES, 1 mM Na 3 PO 4 , 1 mM MgSO 4 , 1 mM CaCl 2 and 5.5 mM glucose, pH 7.4, and stimulated with 1 lM forskolin followed by a cAMP saturating concentration of 25 lM forskolin and the non-specific PDE inhibitor IBMX (100 lM).Analysis was undertaken using an Olympus IX71 inverted microscope with a 609 oil immersion objective (Zeiss, Cambridge, UK) and an optical beam splitter (Photometrics, Tucson, AZ, USA).META- FLUOR software (Molecular Devices, San Jose, CA, USA) allowed image acquisition and real-time monitoring.FRET changes were measured by excitation at 440 nm and obtaining a ratio of the intensity of emissions at 480 and 545 nm.Data are expressed as the % FRET change normalized to the baseline FRET ratio at t = 0.

Mice
The postmortem brain tissue used was derived from animals used in a previous study [10].All animal experiments were approved by the local ethical committee of Hasselt University for animal experiments (matrix ID 201838) and met governmental and institutional guidelines.Female transgenic Alzheimer mice (APPswe/PS1dE9 on C57bl/6 background) were generated in an in-house breeding (matrix ID 201759).Mice were genotyped by PCR analysis of ear biopsies.At the age of 7 months, animals were housed individually in standard cages.They were kept under a reversed 12/12 h light/dark cycle and food and water were provided ad libitum.At 8 months of age, animals were perfused with PBS-heparin, and brain sections were made by cryostat.

Results
As previous work has indicated that increases in expression of PDE4B and D enzymes are associated with cognitive impairment in AD [12,13,26,27], we decided to stain hippocampal slices taken from APP/PS1 mice with antibodies raised against PDE4D and an antibody that detects Ab.Magnification at 109 and 209 allowed clear observation of co-localization between the phosphodiesterase and Ab in plaques that were surrounded by cells expressing PDE4D at a higher level than those positioned more remotely from plaques (Fig. 1A).This was not an artefact of non-specific staining, as plaques formed on coverslips by FAM-Ab 1-42 were not recognized by the PDE4D antibody (Fig. S1).We also recreated this experiment in vitro using exogenous addition of Ab to cultured SH-SY5Y cells (Fig. S2) and again we detected an intracellular colocalization of PDE4D and Ab.Co-localization was seen both at 1 h when FAM-Ab was intracellular but diffuse and at 24 h when intracellular aggregates had formed (Fig. S2).To further support this notion, we utilized proximity ligation (PLA), a technique which we have used before to visualize co-segregation of PDE4 and binding partners [28].PDE4D and Ab 1-42 could be detected in close proximity in the cytoplasm, but not in nucleus of SH-SY5Y cells (Fig. 1B) and there was significantly more PLA signal when cells were treated with Ab 1-42 compared with Ab 1-42 scrambled (Fig. 1C).As this is the first indication that PDE4D and Ab exist in close proximity inside brain cells, we used a biophysical assay to assess the likelihood that the PDE4 and Ab could associate directly.Fluorescence polarization experiments using fluorescently labelled 1-42 Ab peptide indicated a dose-dependent association with GSTtagged PDE4D5 but not with GST alone (Fig. 1D).Additionally, using a technique previously used to investigate the interaction of Ab with VDAC1 [29] we were able to co-immunoprecipitate Ab with PDE4D from SH-SY5Y cells pre-treated with the peptide (Fig. S3).The co-immunoprecipitating Ab bands were at molecular weights observed in the prior study [29].
The direct interaction between PDE4 proteins and Ab detected in Fig. 1 allowed us to map the binding sites by peptide array, a technique that we have used on multiple occasions to discover PDE-binding partner docking domains [24,30,31].Immobilized libraries of PDE4 sequences corresponding to PDE4A4 (Fig. 2B),  PDE4B1 (Fig. 2C), PDE4D5 (Fig. 2D) and PDE4D7 (Fig. 2E) were constructed as 25mers sequentially shifted by 5 amino acids.Arrays encompassing the whole sequence of each PDE4 were overlain with 1-42 Ab peptide or a scrambled control version.Arrays were then blotted for Ab, with dark spots signifying a direct interaction between the immobilized PDE4 sequence 25mer and Ab. Figure 2A depicts the modular structure of PDE4 enzymes, with the conserved catalytic region following on from the two regulatory regions (UCR1 and 2) and the N-terminal targeting domain that is unique to each isoform.Interestingly, Ab bound to all four isoforms of PDE4 in a region close to the start of the catalytic unit.The Ab binding region on PDE4s has three putative binding motifs, one of which is unique to PDE4D isoforms (Site 3 outlined in Fig. 2A).Firstly, there is a double-lysine ('KK', Site 1) motif that appears in all PDE4s (except PDE4C) and starts only 8 amino acids from the start of the catalytic unit (Fig. 2A, lower panel).Point alanine substitution of either of the lysines dramatically reduced Ab binding to PDE4B sequences, whereas a double-alanine substitution was required to decrease binding to PDE4D and ablate PDE4B and PDE4A association (Fig. 3A).Secondly, the arginine-phenylalanine ('RF', Site 2) motif appears in all PDE4 enzymes and alanine substitution of the 'R' ablates Ab binding to PDE4B and dramatically reduces binding to PDE4A and D (Fig. 3A).Substitution of the 'F' with alanine slightly reduced PDE4A and D binding and dramatically reduced PDE4B association.Double substitution of the 'RF' motif to alanine ablated PDE4B binding and attenuated that of PDE4A and D. Quadruple substitution of all four residues in both motifs resulted in a complete loss of Ab association with the PDE4 sequences (Fig. 3A).Site 3 is unique to PDE4D isoforms and appears in a rare region of the catalytic site that is less well conserved than the other parts (Fig. 3A, lower panel).Alanine scanning analysis shows that each of the residues depicted in bold at the top of Fig. 3B is essential for Ab binding to PDE4D.
The Ab-binding sequence was superimposed onto an existing crystal structure of a PDE4D catalytic domain dimer (PDB: 7XAB) [32].The binding site is external to the dimerization interface and is composed primarily of a-helical secondary structure (Fig. 3C).
As we have discovered Ab binding sites on PDE4 sub-family members, we investigated potential complementary PDE4 binding sites on the Ab peptide sequence using peptide array.Using APP transmembrane domain (D672 -L720) which corresponds (following sequential proteolytical cleavage) to the Ab peptide sequence, we constructed arrays consisting of 20mers, sequentially shifted by 2 amino acids and overlaid these peptide spots with purified PDE4D5-GST or GST alone as a negative control (Fig. 4).In doing so, we identified one 20mer sequence (G696-V715) that bound strongly to PDE4D5-GST but not GST alone (Fig. 4A).Point alanine (Fig. 4B) and truncation (Fig. 4C,D) analysis of the Ab 20mer identified that K699 was essential for the association of PDE4D5 and Ab.The Ab mutants K699D and N698E also ablated binding (Fig. 4B, third and second last spots), suggesting that a salt bridge may form between Ab and negatively charged residues on PDE4D5.Visualizing the PDE4 binding sequence on an existing 3D structure (solution NMR) of the APP Q686-K726 dimer (PDB: 2LOH) [33] reveals PDE4D5 binding to the dimerization interface, suggesting PDE4D5 binding may influence the dimerization/oligomerization of Ab (Fig. 4E).
In light of the fact that FP and peptide array have indicated that Ab may form a complex with PDE4D proteins in the catalytic region, and in cognizance of recent reports that there is an insignificant change in PDE4D expression in the hippocampus of the APP/PS1 mice compared with WT but a highly significant threefold change in PDE4 activity [10], we next wanted to check whether PDE4 activity is altered following Ab binding.Addition of increasing concentrations of Ab 1-42 and Ab 1-42 scrambled to cells overexpressing PDE4D5 isoforms followed by    assessment of PDE4 activity in cell lysates showed that Ab could activate PDE4D5 whereas scrambled Ab could not (Fig. 5A).Using a transfected cytosolic FRET-based cAMP-reporter (Fig. 5B), we observed a reduction in cellular levels of cAMP following treatment with a sub-optimal concentration of the adenylate cyclase activator Forskolin (1 lM), where cells were pre-treated with Ab 1-42 for 2 h (Fig. 5D).This was not observed with Ab 1-42 scrambled control peptide (Fig. 5E) or DMSO control (Fig. 5C).Statistics is shown in Fig. 5F.Although small but significant changes in cAMP were detected, these data once again suggest that PDE4 activity was increased when exposed to Ab.There was no difference between treatments when the probes were saturated following treatment with IBMX and forskolin (Fig. 5F) with the maximal FRET change being approximately 20% (Fig. 5F).

Discussion
Many review articles [34,35] have catalogued the variety of benefits afforded by PDE4 and PDE4D selective inhibitors [6] in models of AD.Indeed, other ways of attenuating PDE4D activity (e.g.siRNA [36], dominant negatives [9] and genetic silencing [7]) have also provided rescue from maladaptations conferred by a down-regulation of cAMP signalling associated with loss of synaptic plasticity.Recent reports of increased PDE4D expression in human AD brains [12] and animal AD model brains [37] are also consistent with the studies describing decreased phospho-CREB in AD models and may represent one reason why PDE4 inhibition is so effective in this disease context.Robust evidence supporting a similar role for PDE4B has also recently been published [11,26,27].One point that has not been addressed to date is the possible activation of PDE4 longforms during AD progression.Previous work using human protein microarrays had shown that Ab and PDE4 can physically interact [38] and in line with our data, only the PDE4D subfamily showed a robust interaction.We provide evidence that Ab binds directly to PDE4D in a region at the start of the catalytic core and that this event activates the longform enzyme PDE4D5 (Fig. 5A).Such an action could facilitate a reduction in cAMP (Fig. 5F) that results in a loss of phospho-CREB.It is possible that the Ab peptide in some way relieves the UCR2 'transcapping' of active sites only observed in PDE4 dimers [39].A similar mode of activation has recently been described for allosteric PDE4 activator compounds that phenocopy the actions of PKA phosphorylation of UCR1 [40].Similar activations of PDE4 longforms can also be triggered by antibodies against UCR2, peptide fragments of the regulatory regions and phosphatidic acid (summarized in [41,42]).All of these are thought to bind to the PDE4 enzyme to confer stabilization of PDE4 dimers in conformations that relieve the auto-inhibition by UCR2 transcapping.Hence, it is possible that the Abactivation mechanism of PDE4 longforms is similar, although structural work and experiments with PDE4 phospho-mimic mutants would be required to confirm this.
In converse experiments, PDE4D5 was shown to bind to an Ab-site containing K699 (Fig. 4).This region is known to be crucial for self-assembly of toxic oligomers and fibrils [43,44] suggesting that PDE4D5 may prevent or reverse Ab oligomerization.We have already shown that a PDE4D-binding partner, HSP20, can bind to the oligomerization domain of Ab when HSP20 has been phosphorylated by PKA [20].This action is facilitated by PDE4 inhibition which promotes HSP20 serine 16 phosphorylation by PKA [24] and promotes HSP20-Ab interaction in order to prevent Ab oligomerization [20].It is possible that HSP20 is maintained in its inactive (non-phospho) form in a three-way PDE4-HSP20-Ab complex where the activating action of Ab on PDE4 keeps local cAMP concentrations low.
Finally, we show that PDE4D intimately localizes with Ab in hippocampal cells from APP/PS1 mice.This is to our knowledge the first report of such a relationship where PDE4D expression seems to be enhanced around the areas where plaques have  formed.This could be related to possible antiaggregation effect of PDE4D5 that sequesters monomeric Ab.Additionally, as we have also shown the intracellular colocalization of PDE4D and exogenously applied Ab, albeit in a cultured cell line, we speculate that cells located around plaques may have increased cytoplasmic Ab and therefore enhanced PDE4 activity.We appreciate that this mechanism is yet to be proven in human brains, however, as it is known that PDE4D selective inhibitors reverse learning and memory deficits in this mouse model via PKA and phospho-CREB (summarized in [3]), the robustness of that effect may be down to the fact that concentrated areas of Abactivated PDE4D enzymes are depressing local cAMP concentrations.
In summation, our data suggest a new molecular mechanism by which Ab can down-regulate cAMP in order to promote cognitive deficits associated with AD.

Fig. 2 . 6 FEBS
Fig. 2. Peptide mapping of Ab binding domains on PDE4.(A) Upper panel depicts the modular structure of PDE4 long-form enzymes showing the N-terminal unique region, upstream conserved region 1 (UCR1) and upstream conserved region 2 (UCR2), core catalytic region and sub-family-specific C-terminal region.Lower panel depicts the amino acid sequences of three Ab binding sites at the start of the catalytic unit for each of the different sub-families (PDE4A, PDE4B, PDE4C and PDE4D).(B) Sequential 5-amino shift of PDE4A4 sequence with site 1 and site 2 depicted in bold/underlined.Control is overlain with scrambled Ab. (C) Sequential 5-amino shift of PDE4B1 sequence with site 1 and site 2 depicted in bold/underlined.Control is overlain with scrambled Ab. (D) Sequential 5-amino shift of PDE4D5 sequence with site 1, site 2 and site 3 depicted in bold/underlined.Control is overlain with scrambled Ab. (E) Sequential 5-amino shift of PDE4D7sequence with site 1, site 2 and site 3 depicted in bold/underlined.Control is overlain with scrambled Ab.All peptide array experiments were repeated n = 2.

Fig. 3 .
Fig. 3. Delineation of essential binding residues in Ab docking motif.(A) Amino acids implicated in Ab binding sites 1 and 2 (bold font) from PDE4A, B and D were substituted to alanine (red font) and the binding of Ab assessed.Control is overlain with scrambled Ab.Results typical of n = 2. (B) Amino acids implicated in Ab binding site 3 (red bold font) from PDE4D were substituted to alanine (red font) and the binding of Ab assessed.Control is overlain with scrambled Ab.Results typical of n = 2 (C) structural representation of binding site on PDE4 long-form dimer.Protein sequence as per Uniprot Q08499-1.Light grey, PDE4D catalytic domain monomer 1. Dark grey, PDE4D catalytic domain monomer 2. Red line, dimer interface.Green, compound 22d.Blue, zinc molecule.Red, magnesium molecule.Orange, Q390-V422.Magenta, PDE4D-specific Ab binding site residues H407, F409, R410, A412 and R418.

Fig. 4 .
Fig. 4. Mapping the PDE4 binding sites on Ab. (A) Sequences from Ab were sequentially shifted by 2 amino acids and overlain with either GST or GST-PDE4D5.(B) The binding region detected in (A) was alanine scanned and overlain with either GST or GST-PDE4D.(C) The binding region detected in (A) was sequentially truncated from the C-terminal and overlain with either GST or GST-PDE4D.(D) The binding region detected in (A) was sequentially truncated from the N-terminal and overlain with either GST or GST-PDE4D.Results typical of n = 3. (E) Structural visualization of PDE4D5 binding site on Ab.

Fig. 5 . 10 FEBS
Fig. 5. PDE4 longforms are activated by Ab. (A) SH-SY5Y cells transfected with PDE4D5 were treated for 6 h with indicated concentrations of Ab 1-42 or Ab scr .before lysates were harvested and evaluated for rolipram-inhibited PDE4 activity.Activity is normalized to lysate from non-treated transfected cells.n = 3 Error bars represent SEM.(B) HEK293 cells were transfected with the cAMP reporter EPAC1-cAMPs and treated as indicated.The images show representative images that allow visualization of reporter.Scale bar = 10 lm (C) Transfected HEK293 cells were pre-treated with DMSO and the reaction to 1 lM forskolin was monitored.(D) Transfected HEK293 cells were pretreated with 10 lM Ab 1-42 and the reaction to 1 lM forskolin was monitored.n = 3, 15 cells per treatment.(E) Transfected HEK293 cells were pre-treated with 10 lM Ab scr and the reaction to 1 lM forskolin was monitored.n = 3, 15 cells per treatment.(F) Quantification of relative FRET changes produced in response to 1 lM Forskolin (left) or 25 lM Forskolin plus 100 lM IBMX.n = 3, 15 cells per treatment.Students t-test *P < 0.05.Error bars represent SEM.

2
FEBS Letters (2024) ª 2024 The Authors.FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

4
FEBS Letters (2024) ª 2024 The Authors.FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.