Alzheimer’s disease is an incurable neurodegenerative disorder that is the most common cause of dementia. The late-onset form of AD is a slowly developing, progressive disorder, with age as the most significant single risk factor. Major pathological hallmarks of the disease include the accumulation of extracellular Aβ protein and intracellular tau protein, accompanied by prominent, widespread neuroinflammation (Itagaki et al., 1988; McGeer et al., 1988; Rogers et al., 1988; Eikelenboom et al., 1989). With disease progression, numerous neuroinflammatory molecules, including prostaglandins and cytokines, become dramatically upregulated within cerebrospinal fluid and throughout the brain parenchyma, and are associated with cognitive impairment (Casolini et al., 2002; Brüünsgaard and Pedersen, 2003; Starr et al., 2009; Cribbs et al., 2012). These findings suggest that persistent inflammation may be a driver of neurodegenerative disease (Holmes et al., 2009; Noble et al., 2010; Simen et al., 2011). The prostanoid family of lipid metabolites are the main mediators of acute inflammatory responses; however, their role in chronic inflammatory disease states remains controversial due to the dual role of many prostanoid receptors (Ricciotti and FitzGerald, 2011). For example, prostaglandin D2 (PGD₂) bound to the PGD receptor (DP) stimulates mast cell secretion of type 2 cytokines in the lung leading to eosinophil infiltration and bronchoconstriction, clinically known as allergic asthma (Matsuoka et al., 2000). Conversely, in the CNS, DP1 receptor activation attenuated glutamate toxicity in organotypic hippocampal cultures indicating a neuroprotective role of prostanoid receptors (Liang et al., 2005b).

Prostanoids represent a broad class of arachidonic-acid-derived molecules, including thromboxane, prostaglandins, and prostacyclin, that can exhibit paracrine signaling activity. Prostanoid biosynthesis is carried out by cyclooxygenase (COX) enzymes. Typically, these prostanoids fill a variety of diverse physiological processes that are important for tissue homeostasis (Edelman et al., 2013; Liu et al., 2013; Vitale et al., 2013). Inflammatory events and neurotoxic insults can significantly upregulate prostanoid expression, a process that significantly contributes to the AD-associated neuroinflammatory response (Yermakova et al., 1999; Hoozemans et al., 2001; Choi and Bosetti, 2009; Coma et al., 2010). In Alzheimer’s disease, one of the crucial triggers associated with prostanoid up-regulation is the accumulation of Aβ protein, which both stimulates pro-inflammatory cytokine production and increases cellular phospholipase activity, the first step in prostanoid biosynthesis (Desbène et al., 2012; Shi et al., 2012; Sagy-Bross et al., 2013).

The liberation of fatty acid substrates, such as arachidonic acid, is the initiating step of the prostanoid signaling cascade. Free arachidonic acid undergoes conversion, in a two-step reaction, to prostaglandin H2 (PGH2) by COX enzymes. Two major COX enzymes exist, COX-1 and COX-2, which have distinct but overlapping tissue profiles and activities. COX-1 is expressed in the periphery and is a consistently active enzyme that is important for the maintenance of tissue homeostasis. In contrast, COX-2 represents an inducible form of the enzyme. In the brain, COX-2 expression usually is low but becomes dramatically upregulated in Alzheimer’s disease (Kitamura et al., 1999; Hoozemans et al., 2008). Prostaglandin E2 (PGE₂), prostaglandin D2 (PGD₂), prostacyclin (PGI₂), prostaglandin F2α (PGF_2α), and thromboxane A2 (TXA₂) become upregulated with COX-2 expression; several of these prostanoids, such as PGE₂ and PGD_2, can stimulate Aβ production and can play roles in other amyloid associated pathologies (Liang et al., 2005a; Hoshino et al., 2009; Keene et al., 2010; Zhen et al., 2012). However, the impact of increased PGI₂ in AD is not well understood.

Prostacyclin is synthesized from arachidonic acid through a two-step reaction of the COX and PGI₂ synthase (PGIS) enzymes. PGI₂ acts via the G-protein coupled IP receptor to activate adenylyl cyclase and protein kinase A, thereby increasing intracellular cAMP to produce vasodilatory and anti-inflammatory effects (Cheng et al., 2002). While prostacyclin signaling is largely associated with peripheral vasoregulatory activity, multiple cell types in the brain express both PGIS and the IP receptor, including neurons, glia, endothelial cells, and smooth muscle cells (Oida et al., 1995; Siegle et al., 2000; Fang et al., 2006). These findings suggest that prostacyclin may act as a modulator of CNS activity.

Experimental applications of stable analogs of PGI₂ within the brain have shown improvements in vascular functions and recovery from neuronal damage. For example, the application of iloprost, a stable analog of PGI_2, was able to significantly reduce infarct size after 6 h of middle cerebral artery occlusion in rabbits (Dogan et al., 1996). While the infusion of TEI-7165, another analog of PGI₂, was able to rescue hippocampal neurons and improve response latencies in a step-down passive avoidance test in gerbils subjected to forebrain ischemia (Matsuda et al., 1997). Additionally, PGI₂ appears to impact influence behavior and cognitive processing positively; CP-Tg mice contain a modified Cyclooxygenase-1 enzyme-linked to Prostacyclin-I synthase, which elevates levels of PGI₂ production (Ruan et al., 2006, 2008; Ling et al., 2018) by 60% in the CNS (Vollert et al., 2014). In our previous work, we found that increased levels of PGI₂ improved short-term memory as the CP-Tg mice learned significantly faster in training compared to controls in a contextual fear conditioning test (Vollert et al., 2014).

Several studies have suggested that PGI₂ may have some impact on neurodegenerative disorders such as Alzheimer’s disease (Bush et al., 1990; Smith et al., 2004). He et al. (2017) was able to show that agonist-induced activation of the IP receptor stimulated production of soluble amyloid precursor protein α (sAβPPα), a neuroprotective isoform of the Aβ precursor protein (AβPP), in isolated human microvascular endothelial cells. Wang et al. (2016) demonstrated that injections of PGI₂ into an AβPP/presenilin 1 transgenic mouse model increased Aβ levels and proposed this was due to upregulation of the anterior pharynx defective 1 homolog α (APH-1α) subunit of the AβPP cleaving protein, γ-secretase, via the PKA/CREB and JNK/c-Jun pathways. γ-secretases are responsible for cleavage of the AβPP to produce cytotoxic Aβ peptides, suggesting that PGI₂ may regulate Aβ-associated pathology. These data suggest that increased PGI₂ is likely to play a modulatory role in cognitive function associated with AβPP metabolism.

For this work, we evaluated the effect of PGI₂-overexpression in a model of neurodegenerative disease. CP-Tg mice were crossed to APdE9 mice, a model of Alzheimer’s disease that develops prominent Aβ pathology and develops spatial memory impairments by 12 months of age (Lalonde et al., 2005). APdE9/CP-Tg mice, along with age-matched controls, were subjected to behavioral tests to assess possible changes in cognitive and anxiety-like behaviors. To investigate the impact of prostacyclin overexpression on the Aβ phenotype, we performed Aβ ELISAs on whole brain homogenates. We also investigated prostacyclin-mediated changes to the neurovasculature using immunohistochemical imaging.

In this study, we characterized the effect of upregulated PGI₂ production on behavior, Aβ pathology, and vessel morphology in a mouse model of AD. Previous work has implicated that exogenous application of PGI₂ recovers neurological activity in animal models of ischemia or stroke, improves cerebral blood flow, and prevents pericyte loss and vascular leakage after lipopolysaccharide-induced spinal cord injury in mice (Dogan et al., 1996; Matsuda et al., 1997; Muramatsu et al., 2015; Yang et al., 2017).

We first addressed the possible neuroprotective effects of PGI₂ overexpression on cognitive and non-cognitive behaviors. In an open field assessment of locomotor activity, both Aβ-expressing lines (APdE9 and APdE9/CP-Tg), displayed hyperactivity when compared to the CP-Tg or NTg controls at 12 months of age. CP-Tg mice exhibited decreased activity and spent a larger amount of time along the margin of the field, similar to findings from our previous work (Vollert et al., 2014). PGI₂ overexpression, alone and with the APdE9 phenotype, significantly improved coordination by the last two rotarod trials; we speculate that improvements in motor coordination may be due to increased IP receptor signal in regions involved in motor learning (i.e., striatopallidal and cerebellum) as elevated cAMP is associated with improved motor function (Augustin et al., 2014). Elevated plus-maze and light-dark tests were used to assess anxiety-like behavior. Both measures of anxiety-like behavior revealed increased anxiety in all transgenic models (CP-Tg, APdE9, and APdE9/CP-Tg) measured by time spent in the lit compartment and open arms. Previous reports of the APdE9 mouse model observed similar results as the APPswe/PS1dE9 mice displayed increased exploration in an open field and open arms of an elevated plus-maze (Lalonde et al., 2005). Notably, PGI₂ overexpression did not affect anxiety-like behavior in the APdE9 model as the APdE9/CP-Tg mice performed similarly to APdE9 mice. However, PGI₂ overexpression was enough to produce an anxiety phenotype.

Recent work has suggested that activation of the IP receptor, a cyclic adenosine monophosphate (cAMP)-dependent PKA pathway, may be sufficient to drive stress responses in vivo (Keil et al., 2016). Stimulation of this pathway leads to activation of cAMP-responsive elements binding proteins in the nucleus to synthesize new proteins that alter fear learning and memory formation (Keil et al., 2016). Increased expression of prostacyclin also appeared to exert detrimental effects on other cognitive tests. We observed impaired learning in APdE9/CP-Tg mice during an associative fear conditioning test. Mice were conditioned to a cue and then assessed for contextual and cued memory to the environment or a tone, respectively. During training, APdE9/CP-Tg mice exhibited a significantly reduced freezing response compared to controls indicating a delay in memory acquisition. Additionally, CP-Tg mice also showed delayed acquisition when compared to non-transgenic controls at the third and fifth minutes following the 2-min and 4-min tone-shock pairings. A hippocampal-dependent contextual test was performed 1 h and 24 h after training. APdE9/CP-Tg mice again showed reduced freezing response compared to all controls at 1 h, and 24 h were comparable with CP-Tg and non-transgenic controls.

Conversely, previous studies using an exogenous application of a prostacyclin analog, MRE-269, after stroke in aged rats, were able to recover long-term locomotor and somatosensory functions (Yang et al., 2017). Although IP receptor deletion in mice can be neuroprotective with acute insults such as ischemic damage (Wei et al., 2008), chronic PGI₂ overexpression may exert adverse effects. Our work demonstrates that elevated levels of prostacyclin impair hippocampal memory function in the presence of Aβ insults. Fear behavior during a 24-h auditory cue test was comparable between all groups. As cued conditioning is hippocampal-independent as it requires the use of the amygdala, our results indicate the combination of Aβ, and prostacyclin expression has little to no effect on amygdala-mediated memory retention.

We also find that PGI₂ overexpression significantly increases Aβ production in APdE9/CP-Tg mice. APdE9 mice begin to develop plaques around 6 months of age, with abundant plaques around 12 months of age, and show behavioral impairments in spatial learning by 12 months of age (Jankowsky et al., 2004). In this study, ELIZAs of whole-brain homogenates show a more than 1.5-fold increase in insoluble Aβ40 and Aβ42 in the double transgenic line compared to the APdE9 mice suggesting Aβ burden is directly affected by prostacyclin expression. Intriguingly, these increases are associated with a selective increase of Aβ42 in the soluble pool, suggesting that prostacyclin may selectively bias γ-secretase processing. Another interesting observation in this study was the finding that prostacyclin caused an average 40% increase in plaque diameter, but without a detectible increase in the overall number, perhaps suggesting that factors involved in Aβ metabolism may be modulated through the IP receptor. These findings are supported by recent work on prostacyclin signaling influencing APP processing mechanisms. Agonist activation of the IP receptor was found to increase the expression of the AβPP, resulting in the increased production of Aβ (He et al., 2017). Another report suggests that downstream activation of the PKA/CREB pathway after prostacyclin injections induces upregulation of γ-secretase cleaving enzymes and leads to an accumulation of Aβ c-terminal fragments in an APP/PS1 transgenic mouse model (Wang et al., 2016).

Substantial evidence has shown that disruptions to the neurovasculature are evident in humans and mouse models of Alzheimer’s disease (AD) that directly coincide with an earlier onset and accelerated progression of AD-related pathologies (Bell and Zlokovic, 2009; Zlokovic, 2011; Snyder et al., 2015). These disruptions include altered cerebrovascular functions such as reduced cerebral blood flow velocities and higher resistance indexes (Jin et al., 2017) that have been correlated with impaired cognition (Roher et al., 2012; Ezzati et al., 2017). Structural abnormalities include changes such as capillary atrophy, cell degeneration, loss of tight junction proteins within the blood-brain barrier, and basement membrane thickening due to cerebral amyloid angiopathy (CAA) in both mouse models and individuals with AD (Thal et al., 2008; Bell, 2012; Kelleher and Soiza, 2013; Sengillo et al., 2013; Lai et al., 2015). However, a complete description of how vascular pathology can affect AD progression, and the role of the inflammatory pathway in vascular and neuronal injuries, is still being investigated.

Using a volumetric analysis, we measured the impact of PGI₂ overexpression on vascular structure within the brain, quantifying measures of vessels such as length, branching, diameter, and volume. Our results show that the combination of PGI₂ overexpression with the Aβ phenotype was quite detrimental and worsened multiple measures of vascular structure. Compared with the control group, APdE9/CP-Tg mice had fewer vessels with shorter, smaller diameters. Compared to NTg mice, APdE9 mice had significantly smaller branch lengths, cross-sections, and diameters, as well as a reduction in total volume. Reductions in microvascular density and vessel constriction have been reported in individuals with AD and mouse models of AD. Other studies have suggested that Aβ insults can increase microvascular density in mice (Biron et al., 2013), and there is evidence of angiogenic vessels, accompanied by higher microvessel density, in the hippocampus of post-mortem tissues of individuals diagnosed with AD (Desai et al., 2009). Moreover, some studies have found no changes in vessel volume in both humans and APP/PS1 mice (Hooijmans et al., 2007; Hunter et al., 2012; Burke et al., 2014). In our studies, we also found an increase in total vessel length and degree of branching in the CP-Tg mice compared to their non-transgenic controls or both Aβ models. However, the CP-Tg mice exhibited significantly smaller diameters and cross-sections and had a lower volume than that of NTg mice, indicating that the overproduction of PGI₂ is detrimental to the neurovasculature.

The regulatory function of the IP receptor in the vasculature of the brain is poorly understood. Consequently, alterations in vasculature led us to examine perivascular cells or pericytes of the BBB. These cells maintain the integrity of the cerebral vasculature by promoting tight junction protein expression, facilitate cell-to-cell alignment, and are integral to preventing leakage of neurotoxic macromolecules that lead to neuronal damage (Bell et al., 2010). Application of PGI₂ after lipopolysaccharide-induced vascular damage in a mouse model of spinal cord injury was able to rescue pericyte loss and concurrent leakage of blood components into the central nervous system (Muramatsu et al., 2015). Previous work has reported that accumulation of Aβ in the neurovasculature of AD individuals was correlated with reduced coverage of pericytes on capillaries in AD brains compared to controls (Sengillo et al., 2013). Furthermore, studies of pericyte-deficient, APP^sw/0 (Swedish mutation) mice showed that they had accelerated levels of Aβ deposition, increased tau pathology, and increased neuronal degeneration by 9 months of age when compared to AD control mice (Sagare et al., 2013). In our studies, we evaluated the percentage of CD-13-positive pericytes that co-located with the vessels’ basement membrane as a measure of coverage. CP-Tg mice exhibited a significant reduction in pericyte coverage compared to non-transgenic controls, demonstrating that PGI₂ plays a modulatory role in pericyte function. We identified significant reductions in pericyte coverage of APdE9 and APdE9/CP-Tg mice compared to NTg and CP-Tg mice. However, no difference was observed between APdE9 and APdE9/CP-Tg mice, indicating that Aβ-dependent reductions in pericyte coverage are unaffected by prostacyclin overproduction.

We conclude that PGI₂ expression worsened multiple measures associated with degenerative pathology. We observed that PGI₂ overexpression influences anxiety, possibly due to increases in cAMP activity that has been reported to alter gene expression of proteins involved with the fear response. APdE9 and APdE9/CP-Tg mice exhibited the same level of anxiety as the CP-Tg mice in the light-dark and elevated-plus maze behavioral tests suggesting PGI₂ does not increase the anxiety effect observed in the AD model. In a fear conditioning test, the APdE9/CP-Tg mice exhibited significantly worse associative memory acquisition and consolidation than APdE9 controls. Although earlier reports using PGI₂ analogs suggest activation of the IP receptor could be neuroprotective, CP-Tg mice presented with memory deficits when compared to non-transgenic controls; degenerative pathology was not associated with neuronal loss. In our studies, increased production of PGI₂ accelerated amyloidogenesis by more than 50%, significantly increasing the production of soluble and insoluble Aβ peptides. Overexpression of PGI₂ severely impacted multiple measures of vascular health, a process that was exacerbated with Aβ pathology. Given that prostacyclin is largely protective when expressed in peripheral tissues, our observations suggest that IP signaling is fundamentally different in the cerebrovasculature. Neuroinflammation was also synergistically elevated in the double transgenic model that was not seen with prostacyclin or amyloid overexpression alone. This suggests an immunomodulatory role for prostacyclin signaling in AD. Future studies should explore the mechanism of PGI₂ signaling through its respective receptor within specific cell types of the CNS to better understand the role PGI₂ plays in influencing the progression of neurodegenerative disease.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by Institutional Animal Care and Use Committee, University of Houston.

TW, OO-N, MM, DM, and JE designed research. TW, OO-N, CV, SM, MS, and TB performed research. TW, CV, OO-N, TB, and JE analyzed data. TW and JE wrote the article. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.