Several functions of APP seem to be mediated by its soluble cleavage product APPsα. Selective expression of APPsα in mice with APP^−/− background abolishes most of their deficits, rescuing LTP as well as the typical anatomical and behavioral abnormalities (Ring et al., 2007; Hick et al., 2015). Mice selectively expressing APPsα on APP-KO/APLP2-KO background (which by itself is lethal) survive well into adulthood and show only a mildly altered phenotype, similar to simple APP-KO animals (Zhang et al., 2013). Enhancing APPsα levels by over-expression of ADAM-10 increases cortical synaptogenesis in vivo (Bell et al., 2008). Intraventricular application of APPsα enhances memory in mice (Meziane et al., 1998). Altogether, there is strong evidence that APPsα mediates many of the effects of APP on brain development and supports several cognitive functions. In addition, the APPsα fragment has been shown to mediate a variety of neuroprotective and trophic effects (Hick et al., 2015; Fol et al., 2016; Hefter et al., 2016; Plummer et al., 2016), as discussed in following sections. It is important to note that secretion of APPsα is regulated by neuronal activity (Kirazov et al., 1997; Gakhar-Koppole et al., 2008) and by activated astrocytes (Yang et al., 2015). This may point towards state-dependent functions of the protein, in line with the neuroprotective effects described below. The trophic effects of APPsα are dose-dependent, beginning as low as 100 pM, reaching an optimum at 10 nM and decreasing at higher doses (Demars et al., 2011).

Notably, APPsβ fails to mimic the beneficial effects of APPsα, although there is only a difference of 16 amino acids between both proteins (Hick et al., 2015). In other studies, however, trophic effects of APPsβ were detected, albeit less potent than those of APPsα (Chasseigneaux et al., 2011). Interestingly, APPsβ was found to undergo further proteolytic cleavage and bind to “death receptor 6”, activating caspase-6 and thus contributing to neurodegeneration (Nikolaev et al., 2009).

The intracellular domain of APP, termed AICD, interacts with various cytosolic signaling cascades including glycogen synthase kinase 3 (GSK-3), Ras proteins and MAPK pathways and is able to translocate to the nucleus after forming a complex with the adaptor protein Fe65 (Schettini et al., 2010). There, it is involved in regulation of genes associated with survival and apoptosis (Müller et al., 2008; Multhaup et al., 2015). Indeed, overexpression of AICD was found to induce apoptosis by interaction with the p53-pathway (Ozaki et al., 2006; Nakayama et al., 2008). Moreover, AICD modulates intracellular calcium homeostasis and ATP content (Hamid et al., 2007) and affects synaptic plasticity and hippocampus-dependent learning by increasing LTP (Klevanski et al., 2015).

Resulting from APP cleavage by BACE-1, Aβ peptides can accumulate extracellularly as soluble oligomers or in amyloid plaques, promoting neurodegeneration in AD (Haass, 2010). Interestingly, Aβ can be internalized by neurons and accumulates in the cytosol as well as within endosomes/lysosomes and mitochondria (Chen and Yan, 2006). It exerts neurotoxic effects via a variety of mechanisms, such as disruption of calcium homeostasis (Berridge, 2010), overactivation of mGluR5 (Zhang et al., 2015), impairment of synaptic transmission, plasticity and network function (Kamenetz et al., 2003; Palop and Mucke, 2010), mitochondrial dysfunction (Chen and Zhong, 2013) and apoptosis (Umeda et al., 2011). Remarkably, it is also able to translocate into the nucleus and influence apoptosis-related gene transcription (Barucker et al., 2014; Multhaup et al., 2015). The APP fragment has also been suggested to form channel-like pores in neuronal membranes, but the underlying mechanisms are currently unknown (Barucker et al., 2014).

TBI is a debilitating and life-threatening condition which is the leading cause of disability in people under 35 years in industrial countries (Feigin et al., 2013). Besides acute primary damage, TBI promotes secondary neurodegeneration and increases the risk for developing AD by ~2-fold (Mortimer et al., 1991; Mayeux et al., 1993; Schofield et al., 1997; Guo et al., 2000; Fleminger et al., 2003). Following TBI, diffuse Aβ deposits can be found in the temporal cortex as early as 2 h after the insult (Ikonomovic et al., 2004). Furthermore, post mortem histological analysis shows that deposition of amyloid ß-protein in the brain occurs in approximately one-third of individuals who die shortly after a severe head injury (Roberts et al., 1994). Aβ levels are altered in cerebro-spinal and interstitial cerebral fluid in patients with TBI (Magnoni and Brody, 2010; Tsitsopoulos and Marklund, 2013) and correlate with clinical outcome (Magnoni and Brody, 2010). A history of TBI prior to the onset of dementia correlates with greater amyloid burden in patients with mild cognitive deficits (Mielke et al., 2014) and is associated with faster rates of cognitive decline in AD patients (Moretti et al., 2012; Gilbert et al., 2014). In TBI patients, APP transcription is upregulated and its axonal transport is interrupted due to diffuse axonal injury, which results in deposition of APP and its products in axonal “bulbs” (Hayashi et al., 2015). These results from human studies are in line with a large body of evidence from various models of TBI in mice, rats and sheep, where APP overexpression following TBI has been extensively studied. In models of focal cerebral injury inflicted by stabbing or weight drop local APP immunoreactivity increased in neurons as well as in astrocytes (Otsuka et al., 1991; Lewén et al., 1995, 1996). In the midline fluid percussion model of diffuse brain injury in adult rats APP expression was globally elevated in cortex and hippocampus within hours following the insult (Murakami et al., 1998); in a lateral fluid percussion model APP was overexpressed as early as 1 h after the insult (Pierce et al., 1996). In a weight fall model of brainstem injury in adult rats APP mRNA levels rose as soon as 1 h post-impact, peaked 3 h after the injury at almost twofold baseline level and declined to baseline within 24 h (Yang et al., 2014). Similarly, in an ovine TBI model APP mRNA was up-regulated as soon as 30 min post-impact (Van den Heuvel et al., 1999).

While over-expression of APP following mechanical insults has been observed several decades ago, the functional effects remained unclear until recently. By now, evidence from different animal models points towards an acute neuroprotective effect of APP and APPsα in TBI (Plummer et al., 2016). In diffuse traumatic injury in rats, intraventricular administration of APPsα 30 min after the insult reduced axonal injury and apoptosis and improved motor and cognitive outcome (Thornton et al., 2006). In the same model of TBI, mice lacking APP suffered from greater cognitive and motor impairment in correspondence with larger lesions and increased hippocampal cell loss as compared to WT, again suggesting a protective role of APP in TBI (Corrigan et al., 2012a). Once again, posttraumatic application of exogenous APPsα mitigated these deficits (Corrigan et al., 2012b). These protective effects were found to be mediated by the heparin-binding D1 and D6a domains of APPsα (Corrigan et al., 2011). In an additional study conducted by the same group, the neuroprotective site was pinned down to the APP96-110 sequence in D1, which, applied intraventricularly post-trauma, was enough to significantly improve histological and functional outcome (Corrigan et al., 2014).

At first glance, these findings seem to contradict exacerbation of amyloid pathology and increased risk of AD following TBI. However, there are several possibilities how the two mechanisms may be reconciled. First, APPsα may exert neuroprotective functions independent from the detrimental effects of Aß or amyloid plaques. Second, APPsα could prevent deposition of Aß and further growth of plaques. Third, APPsα may promote clearance of plaques. There is evidence for all three mechanisms. APPsα was shown to modulate BACE activity, possibly inhibiting amyloidogenic cleavage (Obregon et al., 2012). As described in sections “APPsα and APPsβ” and “Mechanisms of Neuroprotection by APP and APPsα Protection in Hypoxia-Ischemia, Excitotoxicity, Degeneration”, APPsα counteracts Aß-mediated excitotoxic damage and delayed degeneration by various trophic and regulatory effects on calcium homeostasis, synaptic function and survival pathways. Recently Fol et al. (2016) discovered that APPsα can ameliorate amyloid pathology by recruitment of microglia, underlining its involvement in clearance of amyloid.

Cardiovascular diseases and ischemic stroke share overlapping genetic and metabolic risk factors such as hypertension, dyslipidemia, glucose intolerance or diabetes and adipositas (Arboix, 2015). Recently these risk factors were established to also increase the odds of developing AD (Orehek, 2012; Wiesmann et al., 2013; Traylor et al., 2016). Moreover, hypoxic-ischemic conditions of the brain such as in ischemic stroke (Honig et al., 2003), heart arrest (de la Torre, 2006), and cerebral small vessel disease (Cai et al., 2015) directly correlate with AD risk, suggesting that cerebrovascular dysfunction is one possible cause of the neurodegenerative disease (Humpel, 2011; Orehek, 2012). Data from a large meta-analysis (Zhou et al., 2015) and a longitudinal study with over 6500 participants (Tosto et al., 2016) show that ischemic stroke increases AD risk by about 1.6 to 2.2-fold, respectively. Several studies indicate that, vice versa, AD patients have an increased risk to develop ischemic (Chi et al., 2013) and hemorrhagic (Chi et al., 2013; Tolppanen et al., 2013; Zhou et al., 2015) stroke and have a higher prevalence of cerebrovascular lesions (Jellinger, 2010). Other studies, however, did not find an increased risk of ischemic stroke in patients with AD (Imfeld et al., 2013; Tolppanen et al., 2013; Zhou et al., 2015). Not surprisingly, cerebrovascular disease and AD contribute additively to cognitive impairment in patients (Hohman et al., 2015) and mouse models (Pimentel-Coelho et al., 2013), possibly forming a vicious cycle of ischemia and neurodegeneration (Pluta et al., 2013).

It has been suggested that cerebrovascular disease, vascular dementia and AD share common pathophysiological cascades such as altered APP processing (Selnes et al., 2010), perturbed energy metabolism (Chen and Zhong, 2013) and pathological immune response (Brod, 2000). These common pathways may then result in overlapping histopathological findings (de la Torre, 2002; Pluta et al., 2009, 2012; Attems and Jellinger, 2014). APP overexpression and Aβ deposition likely play a pivotal role in these processes. In ischemic stroke patients, expression of APP is indeed increased (Pottier et al., 2012) and serum Aβ levels are elevated, correlating with infarct size and clinical outcome (Lee et al., 2005). Likewise, patients who suffered from hypoxia during a cardiac arrest present with increased Aβ levels, which—again—correlate with clinical outcome (Zetterberg et al., 2011). Increased age-related deposition of Aβ was also shown in chronic cerebral vascular disease in rats (Schreiber et al., 2014). However, elevation of Aβ following ischemia is transient. A recent study employing Pittsburgh Compound-B positron emission tomography (¹¹C-PiB-PET; an in vivo imaging method of amyloid), revealed no accumulation of Aβ in patients 18 months after ischemic stroke (Sahathevan et al., 2016).

At the first glance, findings concerning APP in conditions of hypoxia/ischemia seem to be contradictory. On the one hand, the pathological role of APP is supported by multiple animal studies. On the other hand, several studies show beneficial effects of APP in animal models of hypoxia-ischemia. It can be assumed that these opposing effects are mediated by the different cleavage products of APP.

On the one hand, ischemia and oxidative stress enhance BACE-1 and γ-secretase activity, resulting in increased Aβ deposition in rats and mice (Sun et al., 2006; Guglielmotto, 2009; Li et al., 2009). APP accumulates in regions of neurodegeneration following focal cerebral ischemia in the rat (Stephenson et al., 1992). Stroke in rats with Aβ pathology leads to aggravated comorbidity, hippocampal atrophy, and cognitive impairment, similar to the consequences of stroke in AD patients (Amtul et al., 2014).

On the other hand, postischemic intraventricular application of APPsα increases neuronal survival in a model of transient focal ischemia in rats (Smith-Swintosky et al., 1994). APP-KO as well as BACE-KO mice are unable to maintain cerebral blood flow and experience drastically increased acute mortality in a model of global cerebral ischemia (Koike et al., 2012). Overexpression of APP provides neuroprotection following middle cerebral artery occlusion in rats (Clarke et al., 2007). There is compelling evidence that APP acts as a potent anti-thrombotic agent (Van Nostrand, 2016). Moreover, it is required for effective immune and glial cell responses to inflammatory stimuli (Carrano and Das, 2015). With glutamate excitotoxicity being a key pathomechanism of ischemic neuronal damage (Broughton et al., 2009), activation of ADAM10, and thus facilitation of APPsα production, provides neuroprotection against excitotoxic stress in vivo (Clement et al., 2008). These findings are in line with models of AD where expression of APPsα protects against neurodegeneration and rescues synaptic function (Fol et al., 2016).

Taken together, these data support the importance of balance between the beneficial APPsα and the neurotoxic amyloidogenic pathway, thus resolving the initially contradictive results.

Traumatic and ischemic injury is marked by aberrant neuronal activity and excessive glutamate release from neurons and glia, mediating excitotoxicity through enhanced activation of glutamate receptors including the calcium-permeable NMDAR. These processes form a vicious cycle of excessive cation influx, further depolarization, opening of more channels and, eventually, breakdown of the membrane potential, osmotic cell swelling and death (Broughton et al., 2009; McAllister, 2011; see also Figure 1). Application of NMDAR blockers is an established neuroprotective strategy in models of excitotoxicity, hypoxia-ischemia and TBI (Kubo et al., 2001) models. Remarkably, APPsα suppresses NMDAR-mediated currents (Furukawa and Mattson, 1998), potently attenuating calcium responses and thus providing protection against NMDAR-mediated excitotoxicity in hippocampal cell culture (Furukawa et al., 1996; Furukawa and Mattson, 1998; Figure 2). Seemingly contradicting these results, APPsα was shown to enhance LTP in acute hippocampal slices (Ring et al., 2007) as well as in vivo, where intrahippocampal application of the protein increased NMDAR currents, rescued LTP and memory performance (Taylor et al., 2008). This apparent discrepancy may be due to activation of different NMDAR subtypes which, dependent on their subcellular localization (synaptic vs. extrasynaptic) may promote either synaptic potentiation or proapoptotic effects (Hardingham et al., 2002; von Engelhardt et al., 2007). In our experiments slices from wildtype mice showed postischemic potentiation of evoked population responses whereas synaptic transmission in slices from APP-KO mice was drastically reduced (Hefter et al., 2016). As this kind of plasticity also depends on NMDAR (Maggio et al., 2015), it may also be modulated by APP.

However, we could not observe involvement of NMDAR in APP-mediated protection from hypoxia. Taking into account that secretion of APPsα is activity-dependent (Kirazov et al., 1997; Gakhar-Koppole et al., 2008), its effects on NMDAR may provide a negative-feedback loop on extrasynaptic NMDAR in excitotoxicity or a positive feedback loop on subsynaptic NMDAR in LTP and learning.

LTCC belong to the family of VGCC, which are—depending on the membrane potential—almost exclusively conductive for calcium (Zuccotti et al., 2011). They are one of the major sources of extracellular calcium influx in ischemia (Cataldi, 2013) and contribute to neurodegeneration in AD when over-activated by Aβ (Webster et al., 2006). The subtype Ca_v1.2 has been identified as a potential pharmacotherapeutical target (Anekonda and Quinn, 2011). Several studies point towards beneficial effect of LTCC blockers in AD patients (Anekonda and Quinn, 2011; Lovell et al., 2015) as well as in animal models of ischemia and neurodegeneration (Gholamipour-Badie et al., 2013). APP was shown to interact directly with Ca_v1.2 in cultured hippocampal and striatal inhibitory interneurons, with lack of APP resulting in aberrant activity of Ca_v1.2 and altered short-term plasticity (Yang et al., 2009). In primary cultures of rat cortical neurons expression of human APP inhibited calcium oscillations by modulation of LTCC, suggesting a pivotal role in control of neuronal excitability (Santos et al., 2009). In line with these results we recently found an important role of LTCC for APP-mediated neuroprotection in hypoxia (Hefter et al., 2016). These studies suggest that regulation of LTCC function and thereby cytosolic calcium levels by APP may be neuroprotective. However, the molecular mechanisms underlying regulation of LTCC by APP, the function in healthy neurons and the role in ischemia and degeneration remain elusive.

Internal calcium stores, most importantly the ER and mitochondria, play a major role in the regulation of intracellular calcium homeostasis and contribute to elevations of calcium levels under pathological conditions (Mattson et al., 2000). Regulation of store-related calcium homeostasis appears to be mediated by the intracellular domain of APP, i.e., AICD. In cell culture studies, AICD-deficient cells show increased cytosolic calcium concentrations, decreased ability of the ER to buffer calcium and decreased levels of ATP (Hamid et al., 2007). Although a direct binding of AICD to ER receptors such as ryanodine or inositol triphosphate (IP3) receptors has not been described, indirect effects on ER stores and on calcium signaling in general are discussed. One such mechanism is modulation of phosphoinositide-regulated signaling by regulation of the PIKfyve complex, an essential kinase that synthesizes phosphatidylinositol-3,5-bisphosphate. Its loss of function results in neurodegeneration (Balklava et al., 2015; Currinn and Wassmer, 2016). APP was also proposed to modulate IP3 by affecting the transcription of GSK 3b (Hamid et al., 2007). Moreover, AICD may affect calcium levels by binding to X11, BP1, ShcA and other adaptor proteins which might link it to calcium signaling pathways (LaFerla, 2002) and regulate the expression of genes involved in calcium homeostasis such as S100a9 (Leissring et al., 2002; Pardossi-Piquard and Checler, 2012).

In severely compromised tissue, such as the ischemic core in stroke, neurons undergo necrosis due to osmotic swelling, lack of energy metabolites and breakdown of ion gradients (Lo et al., 2003). Under milder and longer-lasting metabolic stress, such as in the ischemic penumbra zone or in chronic cerebral hypoperfusion, the balance between anti- and proapoptotic pathways inclunding NF-κB and p53-pathways may tilt towards apoptotic death (Dirnagl et al., 1999; Broughton et al., 2009). In studies on cultured cells, APPsα was shown to exert anti-apoptotic effects by mechanisms such as upregulation of immediate early gene transcription factors, activation of CREB and NF-κB, genes related to cell survival (Guo et al., 1998; Ryan et al., 2013), phosphorylation of glycogen synthase kinase 3β (GSK-3β; Jimenez et al., 2011) or regulation of expression of cyclin-dependent kinase 5 (CDK-5; Hartl et al., 2013). Effects may be mediated by binding of APPsα to several different receptor proteins which are not yet unambiguously identified (Gustafsen et al., 2013). Potential targets include membrane-bound APP itself (Milosch et al., 2014), and direct inhibition of BACE-1 by APPsα which would counteract Aβ-mediated neurotoxicity (Obregon et al., 2012). Further protection of amyloid toxicity by APPsα was mediated by increased expression of the neuroprotective proteins transthyretin and insulin-like growth factor 2 and subsequent inhibition of the proapoptotic BAD (Stein et al., 2004). AICD was described to interact with more than 20 adaptor proteins including Fe65 proteins members of the Mint/X11 family and members of the JIP family (c-jun-N-terminal kinase interacting protein, JIP1b and JIP2), translocate to the nucleus and interact with survival-related genes (Pardossi-Piquard and Checler, 2012).

Recent years have provided evidence for neurogenesis in several regions of the adult human central nervous system including the dentate gyrus, striatum and olfactory bulb. This mechanism can, in principle, enhance cognitive functions and support recovery from neuronal damage (Inta and Gass, 2015). APPsα was found to stimulate proliferation of neuronal progenitor cells in the subventricular zone (Caillé et al., 2004) and in the hippocampus (Baratchi et al., 2012), whereas AICD was reported to have antiproliferative effects (Zhou et al., 2011). An imbalance between these APP products and thus between neurogenesis and degeneration may contribute to the development of AD (Zhou et al., 2011). Recently APP was shown to control adult hippocampal neurogenesis through GABAergic interneurons, regulating GABAergic synaptic transmission (Wang et al., 2014). APP’s known trophic effects on neuronal viability, cell adhesion, axonogenesis, dendritic arborization and dendritic spines may also contribute to recovery from traumatic and metabolic insults and counteract degeneration (Perez et al., 1997; Lee et al., 2010). APPsα-mediated trophic effects are activity-dependent and stimulated by activation of 5-HT₄ and NMDAR (Gakhar-Koppole et al., 2008; Cochet et al., 2013), suggesting them to be a feasible adaptive strategy in LTP, plasticity and excitotoxicity.

A variety of potential strategies may shift the balance towards non-amyloidogenic cleavage of APP, including modulation of expression, trafficking and regulation of ADAM10 (Postina, 2012). Direct activation of α-secretases by etazolate has been shown to be beneficial in TBI in mice (Siopi et al., 2013). As an indirect mechanism, activation of muscarinic M1 acetylcholine receptors has been reported to increase α-secretase cleavage of APP and decrease Aß levels (Beach et al., 2001). Another activator of α-secretase and inhibitor of β- und γ-secretase is melatonin (Mukda et al., 2016). Increased dimerization of APP via specific compounds such as disulfiram was shown to shift the balance towards non-amyloidogenic cleavage products of APP and thus may present a novel therapeutic approach (Libeu et al., 2012). However, overexpression or activation of ADAM-10 may also have harmful consequences due to effects on other substrates of this enzyme (Clement et al., 2008).

In mice, a selective γ-secretase inhibitor has already been successfully tested, reaching a 33% reduction of Aβ levels within 1 week, without causing severe side effects (Basi et al., 2010). In a mouse model of TBI, pharmacological inhibition of γ-secretase activity reduced post-traumatic tissue loss and improved motor and cognitive recovery (Loane et al., 2009). Both γ- and β-secretase process various other substrates than APP, complicating the use of respective inhibitors (John et al., 2003). However, strategies to specifically inhibit APP cleavage by BACE-1 do exist (Ben Halima et al., 2016) and first BACE-1 inhibitors made it into clinical trials (Vassar et al., 2014).

Regulation of intracellular calcium is a promising neuroprotective strategy (Duncan et al., 2010). As discussed above, APP stabilizes calcium homeostasis by interacting with LTCC, NMDAR and other signaling pathways, offering some feasible pharmacological leverage points. LTCC blockers of the dihydropyridine family such as the common antihypertensive drugs nimodipine and nifedipine attenuate progression of dementia in humans, inhibit Aβ formation in cell culture (Lovell et al., 2015), counteract Aβ-mediated calcium increase and excitotoxicity (Anekonda and Quinn, 2011) and alleviate Aβ-related memory deficits in animal models (Gholamipour-Badie et al., 2013). The NMDAR blocker memantine is not only an established drug for treatment of AD (Danysz and Parsons, 2012), but has also protective effects against excitotoxicity in small doses, being potentially beneficial in patients with high risk of ischemic stroke (Trotman et al., 2015). Other potential calcium-stabilizing approaches target downstream pathways of AICD (Nagase and Nakayama, 2014).

As proven in rodent models of TBI, intraventricular application of exogenous APPsα or its heparin binding domain promote neuronal survival and improve functional outcome (Corrigan et al., 2014). Following TBI or malignant stroke, patients often receive a decompressive craniotomy including ventricular drainage or insertion of an intracranial pressure probe. Application of APPsα through these entries seems feasible. However, these results are highly preliminary and it remains to be proven whether this technique is safe, beneficial and technically feasible in human patients.