Phosphodiesterase III inhibitor promotes drainage of cerebrovascular β-amyloid

Objective Brain amyloidosis is a key feature of Alzheimer's disease (AD). It also incorporates cerebrovascular amyloid β (Aβ) in the form of cerebral amyloid angiopathy (CAA) involving neurovascular dysfunction. We have recently shown by retrospective analysis that patients with mild cognitive impairment receiving a vasoactive drug cilostazol, a selective inhibitor of phosphodiesterase (PDE) III, exhibit significantly reduced cognitive decline. Here, we tested whether cilostazol protects against the disruption of the neurovascular unit and facilitates the arterial pulsation-driven perivascular drainage of Aβ in AD/CAA. Methods We explored the expression of PDE III in postmortem human brain tissue followed by a series of experiments examining the effects of cilostazol on Aβ metabolism in transgenic mice (Tg-SwDI mice) as a model of cerebrovascular β-amyloidosis, as well as cultured neurons. Results We established that PDE III is abnormally upregulated in cerebral blood vessels of AD and CAA subjects and closely correlates with vascular amyloid burden. Furthermore, we demonstrated that cilostazol treatment maintained cerebral hyperemic and vasodilative responses to hypercapnia and acetylcholine, suppressed degeneration of pericytes and vascular smooth muscle cells, promoted perivascular drainage of soluble fluorescent Aβ1-40, and rescued cognitive deficits in Tg-SwDI mice. Although cilostazol decreased endogenous Aβ production in cultured neurons, C-terminal fragment of amyloid precursor protein expression was not altered in cilostazol-treated Tg-SwDI mice. Interpretation The predominant action of cilostazol on Aβ metabolism is likely to facilitate Aβ clearance due to the sustained cerebrovascular function in vivo. Our findings mechanistically demonstrate that cilostazol is a promising therapeutic approach for AD and CAA.


Introduction
Alzheimer's disease (AD) is the most common form of dementia. In addition to neurofibrillary tangles (NFTs), the major pathological features in AD comprise brain accumulation of amyloid b peptide (Ab) as senile plaques and in the form of cerebral amyloid angiopathy (CAA). 1 The deposition of Ab in cerebral blood vessels influences vascular function and worsens the pathology contributing to cognitive decline. [2][3][4][5][6] Recent studies suggest reduced Ab clearance from the brain rather than increased Ab production is responsible for Ab accumulation in the brain in the common late-onset form of AD. 7 Ab may be degraded by glial cells, 8,9 and proteases (e.g., by neprilysin or insulin degrading enzyme) 10 and is transported across the endothelium by the low-density lipoprotein receptor-related protein 1 (LRP-1) or the receptor for advanced glycation end products (RAGE). 11,12 The drainage of extracellular Ab along the basement membranes of capillaries and arteries appears as the most important mechanism for removal of Ab from the brain. 3,13,14 Besides impacting on the structural and functional components of the neurovascular unit, deposition of Ab within the perivascular drainage pathway lowers the motive force of Ab clearance, thereby contributing to increased parenchymal Ab deposition. 15,16 Indeed, a recent study using real-time imaging showed direct evidence that tracers injected into the living mouse brain were cleared along arteries and capillaries but not veins. 17 Solute clearance along the perivascular route was impaired in mice with hemodynamic insufficiency as well as in APPswe/PS1dE9 transgenic mice, which develop parenchymal and microvascular amyloid deposits. 17 Consistent with such findings, immunotherapy intended to reverse the accumulation of Ab plaques in the brain has been shown to concomitantly increase cerebrovascular Ab and exacerbate microvascular lesions. 18,19 Unusually high levels of soluble Ab have been also detected in the brains of immunized AD patients. 20 These results suggest that Ab immunization may result in the failure of efficient perivascular drainage of Ab solubilized from plaques resulting in an increase in the severity of CAA. 18,21 Thus, approaches that enhance Ab clearance via perivascular drainage and protect the neurovascular unit hold promise for the development of novel therapies for AD.
Cyclic nucleotide phosphodiesterases (PDEs) play critical roles in regulating intracellular cyclic nucleotides (cyclic adenosine monophosphate (cAMP) and cyclic guanosine monophosphate), which are important second messengers involved in intracellular signal transduction in all tissues. Thus far, 11 PDE families have been described, most of them are expressed in the brain, attracting attention as a source of new targets for the treatment of psychiatric and neurodegenerative disorders. 22,23 PDE III, PDE IV, or PDE V inhibitors have been shown to have positive effects on the pathology and behavior in several animal models of AD [22][23][24][25] and human studies of AD patients. 26,27 Most of these studies were based on the perspective of neuronal/synaptic and glial dysfunction, and impairment of neurogenesis and synaptic resilience in AD models and patients. However, considering that vascular impairment in the context of neurovascular unit is also closely associated with the pathogenesis of AD/CAA 28,29 and the failure of Ab immunotherapy, 18,20 restoring vascular integrity in addition to neuronal/synaptic and glial function should be crucial for the treatment of AD/CAA. PDE III is the major cAMP-hydrolyzing PDE (a negative regulator of cAMP) uniquely expressed in vascular smooth muscle cells, and PDE IIIA isoforms are also involved in cardiovascular function by regulating vascular smooth muscle growth regulation and phenotypic changes. Cilostazol, a selective inhibitor of PDE III, increases cAMP in vascular cells and has multiple effects on the vasculature such as vasodilatation, anti-oxidation, anti-inflammation, regulation of smooth muscle cell, 30 increase in cerebral hemodynamics, 31 pulse duration time 32 and arterial elasticity 33 with maintenance of microvascular integrity. 34 We have recently shown that patients with mild cognitive impairment receiving cilostazol exhibit significantly reduced cognitive decline, 35,36 suggesting that cilostazol, originally a drug for stroke and peripheral arterial disease, is effective for dementing disorders through its action on the vasculature and providing rationale of experimental study using an animal model of cerebrovascular b-amyloidosis.
Given the rationale and background outlined above, we first explored the expression of PDE III in postmortem human brains followed by a series of experiments using transgenic mice (Tg-SwDI mice) expressing the Swedish (K670/M671L) and vasculotropic Dutch/Iowa (E693Q/D694N) mutant of the human amyloid precursor protein (APP) to examine whether cilostazol protected against the disruption of the neurovascular unit and rescued the perivascular drainage of Ab in AD/CAA. We further investigated whether cilostazol is beneficial or improves vascular reactivity, cerebral blood flow (CBF), perivascular Ab drainage, vascular morphology, and behavioral/cognitive function in the transgenic mice, which exhibit deficient cerebral clearance of Ab. 37 Furthermore, we examined whether cilostazol can alter the Ab biogenesis in cultured neurons as well as the in vivo system. These experimental approaches provide strong rationale for placebo-controlled pharmacological trials to establish the efficacy of cilostazol in patients with AD/CAA.

Subjects/Materials and Methods
Postmortem human brain material Autopsied brains were obtained from Kyoto University Hospital through a process approved by an institutional research committee. Methods and relevant references are available in the Data S1.

Experimental study
We used transgenic mice, C57BL/6-Tg(Thy1-APPSwDutIowa) BWevn/J (Jackson Laboratory, Bar Harbor, ME) as a CAA mouse model. Experimental details are available in the Data S1.

Results
PDE III, a negative regulator of cAMP, was abnormally expressed in the cerebral vessels of AD/CAA patients and correlated with the vascular amyloid burden On the basis that PDE III is a vasoactive enzyme uniquely expressed in vascular smooth muscle cells and plays a pivotal role in vascular function, we evaluated PDE IIIA immunoreactivity in postmortem human brains. The demographic details and pathological findings are summarized in Table S1. PDE IIIA immunostaining was negligible in the brain without CAA (Fig. 1A-C). In contrast, PDE IIIA-immunoreactivity was upregulated mostly in the vascular smooth muscle cells, of both the leptomeningeal and cortical arteries as CAA severity increased ( coefficient, Fig. S1). Two relatively pure CAA cases (Case No. 9 and 15; moderate and severe, respectively) showed increased PDEIII expression, while two controls (Case No. 16 and 17) without brain pathology showed lower PDEIII expression score (Table S1). PDE IIIA was also detected in the parenchyma ( Fig. S2A-C). Some of the cortical capillaries exhibited PDE IIIA immunostaining in the vascular walls suggestive of pericytes ( Fig. S2D and E). Some of neurons including dystrophic neurites around the amyloid core showed PDE IIIA expression ( Fig. S3E-L). However, neuropil threads and amyloid cores did not express PDE IIIA (Fig. S3E-L). The aged Tg-SwDI mice also showed the increased expression of PDE IIIA in the vascular wall of the leptomeningeal and intracortical arteries (Fig. S4). These results suggested that PDE IIIA expression is abnormally upregulated especially in the vascular smooth muscle cells in close correlation with the vascular amyloid burden.

Cilostazol restored vasoreactivity in Tg-SwDI mice
Impairments in cerebral circulation and vascular reactivity have substantial roles in the onset and progression of AD/ CAA in human 38 and mouse models of AD/CAA including Tg-SwDI mice. 5,39 Consistent with the previous reports, which demonstrated that the Tg-SwDI mice showed microvascular Ab deposition throughout the forebrain by 12 months of age, 37 most of the vessels were affected by microvascular amyloid in vehicle-treated Tg-SwDI mice aged 23 months (Fig. S5). We first investigated whether cilostazol ameliorates vascular dysfunction in Tg-SwDI mice. Resting CBF was marginally increased in cilostazol-treated Tg-SwDI mice compared with vehicle-treated Tg-SwDI mice both aged 12 months ( (Fig. 2H). This was consistent with the results of CBF response. These findings indicated that cilostazol restores cerebral hemodynamic reserve in Tg-SwDI mice.
Cilostazol facilitated perivascular drainage of Ab in Tg-SwDI mice Cerebrovascular dysfunction and reduced arterial pulsations may result in the failure of perivascular drainage of soluble Ab, leading to the accumulation and aggregation of Ab in the vessel walls and in the brain parenchyma. 17 We investigated whether the restoration of vascular reactivity by cilostazol led to an improvement of perivascular drainage of soluble Ab in Tg-SwDI mice. Intracerebral injections of soluble fluorescent HiLyte Ab tracers were performed, following the methodology described before 13 (Fig. 3A). Thirty minutes after the injection into the striatum, soluble fluorescent Ab 1-40 had spread diffusely throughout the brain parenchyma and was present within the basement membranes of parenchymal and leptomeningeal arteries. At 30 min post-injection, most of the Ab had drained out of the brain from the site of injection and Ab was detected only in the leptomeningeal arteries ( Fig. 3B), and within Iba-1-positive microglia as well as NeuN-positive neurons (data not shown). Within the leptomeningeal arteries, Ab was located outside the platelet endothelial cell adhesion molecule-1-positive endothelial cells and colocalized with smooth muscle actin-positive vascular smooth muscle cells (data not shown) and the laminin present in the basement membranes surrounding smooth muscle cells (Fig. 3C). The radial spread of Ab localized in the leptomeningeal arteries from the injection site was calculated. We did not quantify the fluorescent Ab -positive vessels around the needle tract as they could be technical artifacts. The number of vessels analyzed was similar for vehicle-and cilostazol-treated Tg-SwDI mice (cilostazol vs. vehicle; 31 AE 3.7 vs. 29 AE 7.0). There were marginally significant and significant differences in the averaged and maximal radial spread of fluorescent Ab 1-40 , respectively, between vehicle-and cilostazol-treated Tg-SwDI mice at 15 months of age (13.5-month cilostazol treatment vs. vehicle; 2783 AE 234 lm vs. 2395 AE 186 lm for average spread, P = 0.057; 3678 AE 238 lm vs. 2820 AE 386 lm for maximum spread, P = 0.028) ( Fig. 3D and E). These results suggest that long-term cilostazol treatment helps to maintain efficient perivascular drainage of Ab in Tg-SwDI mice.

Cilostazol reduced Ab deposits in the Tg-SwDI mice brain
Tg-SwDI mice develop robust accumulation of both vascular and parenchymal Ab deposits with predominance of Ab 1-40 over Ab  in the perivascular/vascular areas starting at 3 to 5 months of age ( Fig. S7A-D). 37 We hypothesized that facilitation of perivascular Ab drainage with cilostazol reduces Ab deposits. Thus, we assessed the effects of cilostazol on Ab accumulation in the Tg-SwDI mice brain. Cilostazol-treated Tg-SwDI mice showed significantly or marginally significantly decreased immunore-activity of Ab in the frontal cortex (  4D). Both vehicle-and cilostazol-treated Tg-SwDI mice did not show apparent microhemorrhage aged 15 months ( Fig. S7E and F). The evaluation of the blood brain barrier (BBB) integrity using Evans blue also showed that the Tg-SwDI mouse aged 10 months did not exhibit apparent Evans blue extravasation into the central nervous system (CNS) parenchyma; however, Evans blue did diffuse throughout the extracellular space in nonneural tissue (e.g., kidney, heart, and liver) (Fig. S8). Taken together, the above results indicated that cilostazol reduces Ab accumulation in the Tg-SwDI mice brain possibly by promoting perivascular drainage of Ab associated with preserved hemodynamic reserve.

Cilostazol attenuated the degradation of vascular walls with Ab deposit
To examine the nature of Ab accumulation and morphological changes in detail, we performed electron microscopy of Tg-SwDI mice aged 21 months that had been treated with cilostazol or vehicle for 17 months. Whereas 25% (49 per 195) of the vessels in the vehicle-treated Tg-SwDI mice showed some type of abnormality in the vascular wall vis-a-vis degeneration of pericytes or vascular smooth muscle cells or damaged intracellular organelles and alterations in the basement membranes ( These were apparent as numerous large vesicles, membranes around portions of the cytoplasm (called isolation membrane), mitochondrial injury, lysosomal inclusions, and large lysosomes were found in the cytoplasm of pericytes, mostly markers of autophagy (Fig. 5A, and Fig.  S9A and B). The arterioles of vehicle-treated Tg-SwDI mice showed a similar degeneration of vascular smooth muscle cells with vacuolization and destroyed organelles accompanied by basement membrane thickening/alterations (Fig. 5B, C, and Fig. S9C, D). Conversely, cilostazol-treated Tg-SwDI mice only rarely showed the above-abnormal findings (Fig. 5D-F, and Fig. S9E and F). compared with vehicle-treated mice (54.9 AE 3.3%) aged 12 months (P < 0.001; Fig. 6A, right panel). Spontaneous activity was not significantly different between the two groups of mice (Fig. 6A, left panel). At 15 months of age, although there was a trend for increase in alternations of entries in cilostazol-treated mice, no significant difference in both alternations of entries in the arms and spontaneous activity between vehicle-and cilostazol-treated Tg-SwDI mice (Fig. 6B). There were no significant differences in the number of total entries and alternations of entries between vehicle-and cilostazol-treated wild type mice aged 12 months (Fig. S10). Cilostazol-treated Tg-SwDI mice showed significantly better hair condition compared with vehicle-treated Tg-SwDI mice aged 15 months, also suggesting that cilostazol prevents cognitive decline in Tg-SwDI mice (Data S1 and Fig. S11).

Cilostazol suppressed endogenous Ab production in the primary cultured neurons
There is a possibility that the observed reduction in Ab deposition is a consequence of decreased Ab production, in addition to increased Ab clearance. Indeed, a previous study showed that cilostazol decreased extracellular and intracellular Ab levels in mouse neuroblastoma N2a cells expressing human APP with Swedish mutation. 25 We examined the effect of cilostazol on endogenous Ab production from neurons in vitro using the primary cultured rat neurons. This showed that cilostazol decreased the amount of Ab40 and Ab42 in the media released from neurons in a dose-dependent manner ( Fig. 7A and B). A WST assay showed no significant differences among these groups (Fig. 7C). These results suggest that cilostazol can suppress the endogenous Ab production from neurons without affecting cell viability.

Cilostazol did not affect APP processing in Tg-SwDI mice
Although cilostazol does reduce neuronal Ab production, it has been established that only a very minor fraction of cilostazol can cross BBB and enter the brain after the peroral administration. 40 Therefore, the direct effects of cilostazol on neurons were expected to be less significant in vivo than in vitro. Based on the knowledge that b-secretase activity is augmented by hypoperfusion, 41 which may be ameliorated with cilostazol administration, we performed quantitative immunoblotting using antibody against APP C-terminal domain, as a marker of APP processing. There were no significant differences in the levels of APP and APP C-terminal fragments (CTFs) between vehicle-and cilostazol-treated Tg-SwDI mice aged 15 months, which had received vehicle or cilostazol for 13.5 months (Fig. 8).

Discussion
The complex expression patterns and activity of the different PDEs throughout the CNS under normal and pathological conditions have been the subject of intense interest during the past decade, especially in regard to the therapeutic potential of PDE inhibitors for CNS disorders. 22,23 We found that PDE IIIA expression in the microvessels, predominantly in the smooth muscle cell layers, was markedly upregulated as CAA severity increased. To our knowledge, this is a first report showing the upregulation of PDE IIIA expression in the microvessels in close correlation with vascular amyloid burden. Further studies, including in vitro studies, are needed to clarify whether the increased PDE IIIA expression is a cause or consequence of amyloid deposit. Our experimental study demonstrated that cilostazol, PDE III inhibitor, restores vascular reactivity and hemodynamic reserve, promotes perivascular drainage of Ab, reduces degenerative changes of vascular walls with Ab deposits and prevents the decline of cognitive performance in the Tg-SwDI mice, a model for cerebrovascular b-amyloidosis. These protective roles of cilostazol against Ab-induced neurodegeneration seem to be mediated by its vasculotropic, but not antiplatelet, effects, as long-term  aspirin treatment did not reverse vascular Ab deposition, hemodynamic derangements, or cognitive decline in Tg-SwDI mice (unpublished data). Thus, besides its antiplatelet actions, the maintenance of neurovascular integrity with cilostazol could represent a promising approach in the decelerating cognitive decline associated with Ab deposition. Since the motive force for perivascular Ab drainage appears to be generated by arterial pulsations, 42,43 the direct action of cilostazol on the vascular smooth muscle cells to increase pulse duration time 32 and arterial elasticity 33 may have contributed to the facilitated perivascular drainage of Ab and the above positive effects of cilostazol in the Tg-SwDI mice. Furthermore, as electron microscopic examinations have revealed, the protection of pericytes and other vascular components by cilostazol may have led to the preservation of vasodilative and hemodynamic reactivity and BBB integrity along the perivascular drainage pathway. The CNS is devoid of conventional lymphatic vessels, unlike other organs that contain networks of lymphatic vessels, which process various elements such as wastes, fluid, proteins, and cells from tissues to lymph nodes. 15,42 However, the perivascular drainage system in the brain performs the main function assigned to systemic lymphatic vessels. [13][14][15]42,44 Animal studies using various tracers have demonstrated that interstitial fluid and solutes drain rapidly via perivascular "lymphatic" pathways from brain parenchyma along basement membranes in the walls of capillaries and arteries in the opposite direction of the arterial blood flow to cervical lymph nodes. 15,17,42 This drainage route corresponds very closely with the distribution of Ab in the basement membranes of capillary and artery walls in CAA. 45 The failure of this drainage in the ageing brain and in the presence of CAA results in the accumulation of insoluble and soluble Ab and probably other metabolites that would lead to loss of homeostasis of the neuronal environment. 15,42 This notion is also supported by the experimental data that the vascular Ab deposition is increased following bilateral common carotid artery stenosis of CAA model mice 46 or middle cerebral artery occlusion model mice. 47 Such "lymphatic" congestion of the brain may be improved by vasoactive cilostazol.
Nevertheless, there is a possibility that the observed reduction in Ab deposition could largely be a consequence of decreased Ab production, rather than increased Ab clearance. According to previous studies, cilostazol could regulate APP processing via the cAMP-PKA pathway or activation of CREB cascade, 26 and PKA has been implicated in the regulation of APP processing. 48 In addition, the PDE V inhibitor sildenafil has been shown to reduce brain Ab through the cGMP-CREB pathway. 49 Given that the activation of the 5-HT4 receptor could regulate a-secretase activity, which is associated with cAMP functional response, 50 cilostazol-mediated cAMP might enhance a-secretase activity, leading to decreased Ab production. Indeed, the current study showed that cilostazol could decrease endogenous Ab generation from the cultured neurons in vitro. There is also a possibility that the improvement of cerebral hemodynamics by cilostazol could reduce in Ab synthesis as hypoxia/hypoperfusion has been shown to upregulate b-secretase BACE1. 41 However, cilostazol did not alter the levels of APP CTFs in Tg-SwDI mice in vivo. This discrepancy could be attributed to the poor ability of cilostazol to cross the BBB, 40 as well as the finding that BBB was not apparently disrupted in middle-aged Tg-SwDI mice in the current study. It is nonetheless possible that cilostazol, even at a low level, activates cAMP-PKA signaling with a similar mechanism to enriched environment, leading to the activation of b2-adrenergic receptors and inhibition of synaptotoxicity of human Ab oligomers. 51 Taken together, the positive effects of cilostazol in vivo observed in the current study may predominantly result from facilitated Ab clearance due to the sustained cerebrovascular function; however, the previously unidentified mechanisms leading to Ab reduction, could offer a potential source of future investigation. The use of Tg-SwDI mice has two distinctive advantages, when compared to other types of Tg mice harboring mutant APP. Firstly, Tg-SwDI mice have been shown to express mutant human APP at low levels, even below those of endogenous mouse APP in the brain. Despite the low expression levels of mutant APP, Tg-SwDI mice develop early-onset robust accumulation of Ab in the brain. This could result from the deficient clearance of Dutch/Iowa mutant Ab from the CNS at the cerebral vasculature, as shown in the previous studies. 37,52 Since decreased elimination, rather than increased production, of Ab is likely to be a major cause of sporadic late-onset AD, this model would be useful for investigating the mechanisms and therapeutic approaches for sporadic AD. Second, Dutch/Iowa mutant Ab peptides have highly vasculotropic nature and Tg-SwDI mice exhibit extensive microvascular amyloid deposition. A recent report showed that Tg-SwDI mice, which dominantly develop microvascular amyloid exhibit cognitive dysfunction at 3 months of age. However, Tg-5 9 FAD mice that dominantly develop parenchymal amyloid deposits did not exhibit cognitive dysfunction at this age. These findings suggest that cerebral microvascular pathology might contribute to the early stages of cognitive impairment in AD and related disorders, 53 consistent with the notion that cerebral vascular dysfunction is an early contributor for AD pathophysiology. 2 The positive effects of cilostazol on Tg-SwDI mice in the current study suggest that cilostazol is a promising therapeutic target for early stages of cognitive impairment of neurodegenative etiology.
There are several issues to be addressed in future studies. First, we focused on only middle-aged and old Tg-SwDI mice, although emerging evidence shows that early vascular changes may precede Ab accumulation and progressive disease cascades. Further studies are needed to elucidate whether cilostazol can improve early vascular alterations in young Tg-SwDI mice. Second, we specifically focused on perivascular drainage of Ab. However, other clearance mechanisms, such as transcytotic delivery across BBB or degradation of Ab by microglial/astroglial cells or enzymes, might contribute to reduction of Ab overload. In addition, clearance pathways other than perivascular lymphatic routes, such as paravenous routes and bulk fluid flow through astrocytic endfeet 54 or paraaxonal/neuronal pathways, should be considered in future studies. Finally, the current study shows that cilostazol reduces Ab overload by hemodynamic enhancement. However, the accumulation of NFTs within neurons, composed largely of ubiquitin and the microtubule-associated protein tau, is another important hallmark of AD. 1 While the accumulation of Ab may derive mainly from the failure of the clearance system as mentioned above, that of NFTs appears to be associated at least partially with the failure of the ubiquitin-proteasome system. 16 Additionally, hemodynamic insufficiency has been shown to enhance phosphorylation of tau through upregulation of tau phosphorylating enzymes. 55 Further studies are warranted to examine whether cilostazol have positive effects on the neuropathology of NFTs as well as Ab.
In conclusion, our study showed that the vasoactive drug cilostazol, a PDE III inhibitor, prevented cognitive decline triggered with Ab deposition by facilitating Ab clearance out of the brain and preserving neurovascular unit. Cilostazol, which can resolve the failure of the Ab drainage pathway may provide a novel promising therapeutic target for AD/CAA, potentially in combination with early Ab immunization therapy and pharmacological intervention to enhance enzymatic degradation of Ab and absorption of Ab into the blood. A prospective trial is now needed to determine the effects of cilostazol on AD/ CAA.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Data S1. Supplemental materials. Data S1. Supplemental data. Figure S1. The correlation between the CAA severity and the PDE IIIA expression score demonstrates significant linear correlation (Rs = 0.8272, P < 0.001). PDE IIIA expression score was calculated from the average of qualified grading of PDE IIIA expression (0 = none, 1 = low, 2 = medium, 3 = high). Figure S2. (A-E) Representative images of PDE IIIA immunostaining in patients with AD and CAA. PDE IIIA expression is only slightly observed in neurites in the cerebral cortex (A) and pyramidal neurons in the hippocampal CA3 regions (B) and dentate gyrus (C), whereas it is strong in the cortical vessels in patients with AD/CAA (D, E). Scale bar indicates 50 lm. Figure S3. Representative images of the leptomeningeal arteries (A-D), the cerebral cortex (E-H), and the senile plaques (I-L), which were stained for Ab (A, E, I), paired helical filament-tau (PHF-tau) (B, F, J), PDE IIIA (C, G, K), and Congo red (D, H, L) in patients with AD and CAA. Of note, some neurons with dystrophic neurites (J, dashed arrow) showed PDE IIIA expression (G, K), but neuropil threads (F, arrows) and amyloid cores (I and L, arrows) were not positive for PDE IIIA. Scale bar indicates 50 lm. Figure S4. PDE IIIA expression is increased in the vascular wall in aged Tg-SwDI mouse. Representative images of PDE IIIA immnostaining in the leptomeningeal (A-C) and intracortical (D-F) arteries of wild type mouse aged 12 months (A and D), Tg-SwDI mouse aged 4 months (B, E), and Tg-SwDI mouse aged 23 months (C and F). Scale bar indicates 20 lm. Figure S5. Thioflavin-S staining in the Tg-SwDI mouse aged 23 months. Scale bars indicate 100 lm in A-F, 1 mm in G, and 500 lm in H and I. Figure S6. Representative images of vehicle-(A) and cilostazol-treated (B) Tg-SwDI mice showing where vessel diameter was quantified. An averaged vessel diameter across a 25 lm longitudinal segment (8 consecutive segments per mouse) was analyzed before (baseline) and after inhalation of 5% CO 2 or treatment with acetylcholine. Scale bars indicate 50 lm. Figure S7. Tg-SwDI mice exhibit robust accumulation of both vascular and parenchymal Ab with predominance of Ab40 over Ab42 in the perivascular/vascular areas without apparent microhemorrhage or BBB leakage at 15 months of age. (A-D) Representative images of Ab40 (A), Ab42 (B), and Ab (6E10) immunostaining (C and D), Perls-Stieda's iron staining (E), and IgG staining (F) in Tg-SwDI mice aged 15 months. Scale bars indicate 50 lm in A and B, 20 lm in C and D, and 100 lm in E and F. Figure S8. Representative macroscopic (A and B) and microscopic fluorescent (C and D) images of the brain (A, C) and kidney (B and D) in the Tg-SwDI mouse aged 10 months after Evans blue, followed by FITC dextran (green), injection. There were clear differences in the extravascular extravasation of Evans blue (shown in blue in A, B, and red in C, D) between the brain and kidney. Scale bars indicate 50 lm. Figure S9. Electron microscopic images of vehicle-treated (A-D) and cilostazol-treated (E, F) Tg-SwDI mice aged 21 months. (A) Multiple vesicles in the cytoplasm of pericytes. (B) Multiple vesicles (+) and damaged intracellular organelles including mitochondria replaced by dark material (upper inset) in the pericytic cytoplasm. Multiple vacuoles (++) and thickening/alterations of basement membrane (lower inset). The regions marked by dashed squares are enlarged in the right upper and lower inset, respectively. (C) Thickening/ alterations of basement membrane. The region marked by a dashed square is enlarged in the inset. (D) Two transversely sectioned microvessels were seen connected by a dysmorphic process. Multiple stages of autosomes, lysosomes, and damaged intracellular organelles with disruption of mitochondrial membrane and reduced cristae (arrows, right lower inset) were present (++). The regions marked by dashed squares are enlarged in the right upper and lower inset, respectively. (E and F) Almost intact capillaries were seen in cilostazol-treated Tg-SwDI mice. Arrows indicate tight junction of endothelial cells. Pericytic mitochondria were normal. The region marked by a dashed square is enlarged (F, inset). Scale bars indicate 1 lm. As, astrocyte; Ax, axon; BM, basement membrane; E, endothelial cell; En, nucleus of endothelial cell; Ly, lysosome; m, mitochondrion; P, pericyte; Pn, nucleus of pericyte. Figure S10. (A) Histograms showing number of arm entries (left) and alternation behavior (right) in the Y maze test of vehicle-treated (n = 10) and cilostazol-treated (n = 10) wild type mice aged 12 months treated from 11 months of age. Error bars indicate SD. Figure S11. Grooming performance in vehicle-treated and cilostazol-treated Tg-SwDI mice. (A) Images showing the hair (ruffled coat) conditions graded as four categories 0-3. (B) Histogram showing grading score of the hair (ruffled coat) condition of vehicle-treated (n = 10) and cilostazol-treated (n = 12) Tg-SwDI mice aged 15 months. Error bars indicate SEM. *P < 0.05 in vehicle-treated Tg-SwDI mice versus cilostazol-treated Tg-SwDI mice. Table S1. Detailed neuropathological features and analyses of 17 patients with AD or non-AD. These patients were also divided into 4 groups according to the Vonsattel grade of CAA (4 patients without CAA, 4 patients each with mild or moderate, and 5 patients with severe CAA). The grading score of PDE IIIA expression is calculated from the average of qualified grade (0 = none, 1 = low, 2 = medium, 3 = high). In total, 1700 leptomeningeal and cortical arteries (100 per patient) were analyzed in the randomly selected 5 regions (20 vessels per one region) from the temporal and occipital lobes. AD, Alzheimer's disease; ALS, amyotrophic lateral sclerosis; CAA, cerebral amyloid angiopathy; CERAD, Consortium to Establish a Registry for Alzheimer's Disease; CHF, congestive heart failure; CLL, Chronic lymphocytic leukemia; NFT, neurofibrillary tangle; SIVD, subcortical ischemic vascular dementia; VaD, vascular dementia; ND, not determined.