Failure of monotherapeutics
Neurodegenerative disease therapeutics has been, arguably, the field of greatest failure of biomedical therapeutics development. Patients with acute illnesses such as infectious diseases, or with other chronic illnesses, such as cardiovascular disease, osteoporosis, human immunodeficiency virus infection, and even cancer, have access to more effective therapeutic options than do patients with AD or other neurodegenerative diseases such as Lewy body dementia, frontotemporal lobar degeneration, and amyotrophic lateral sclerosis. In the case of Alzheimer's disease, there is not a single therapeutic that exerts anything beyond a marginal, unsustained symptomatic effect, with little or no effect on disease progression. Furthermore, in the past decade alone, hundreds of clinical trials have been conducted for AD, at an aggregate cost of billions of dollars, without success. This has led some to question whether the approach taken to drug development for AD is an optimal one.
Therapeutic success for other chronic illnesses such as cardiovascular disease, cancer, and HIV, has been improved through the use of combination therapies [5]. In the case of AD and its predecessors, mild cognitive impairment (MCI) and subjective cognitive impairment (SCI), comprehensive combination therapies have not been explored. However, the past few decades of genetic and biochemical research have revealed an extensive network of molecular interactions involved in AD pathogenesis, suggesting that a network-based therapeutics approach, rather than a single target-based approach, may be feasible and potentially more effective for the treatment of cognitive decline due to Alzheimer's disease.
Preclinical studies
Extensive preclinical studies from numerous laboratories have identified multiple pathogenetic targets for potential intervention. These include, in addition to amyloid-β (Aβ) oligomers and tau, inflammatory mediators, apolipoproteins and lipid metabolism factors, hormonal mediators, trophic factors and their receptors, calcium regulatory pathways, axoplasmic transport machinery, neurotransmitters and their receptors, prion protein, and a host of other potential targets. However, one of the drawbacks of these preclinical studies is that many have implicated single pathways, and shown large effects of targeting one pathway, whereas in human studies, such approaches have not been borne out. There are several possible inferences from such discrepant results: first, it is possible that it will be necessary to target multiple pathways simultaneously in order to effect an improvement in symptoms and pathophysiology. Second, it is possible that targeting a single pathway will be sufficient, but that earlier intervention will be required. Third, it is possible that all of these seemingly disparate pathways will converge on a single critical pathway, so that either a single targeted therapy or a multi-component, multi-targeted approach may be effective. And fourth, of course it is possible that neither of these two types of approaches will be sufficient. It is worth noting, however, that it is possible that addressing multiple targets within the network underlying AD pathophysiology may be successful even when each target is affected in a relatively modest way; in other words, the effects of the various targets may be additive, multiplicative, or otherwise synergistic.
Based on a combination of in vitro and in vivo studies, we have advanced a model in which AD results from an imbalance in endogenous plasticity signaling (Fig. 1), 5-9, and in which the β-amyloid precursor protein (APP) is a mediator of such plasticity-related signaling. Thus the model suggests that AD is analogous to other chronic illnesses such as cancer, osteoporosis, and atherosclerosis. In the case of osteoporosis, osteoblastic signaling is chronically exceeded by osteoclastic signaling, resulting in an age-associated chronic illness featuring loss of bone. By analogy, in Alzheimer's disease, there is a fundamental, age-associated imbalance between the dynamically opposed physiological processes that mediate plasticity, i.e., between synaptoblastic and synaptoclastic activity. This signaling involves physiological mediators of synaptic development, maintenance, repair, and remodeling, including APP, its derivative peptides, ApoE, and tau, and is modulated by all of the many disparate factors associated with Alzheimer's disease. Furthermore, just as for neoplasia, positive feedback selects and amplifies the disease process; however, whereas in oncogenesis, the positive feedback occurs at the cellular level, in Alzheimer's disease, the positive feedback occurs at the molecular species level, in the form of prionic loops [5, 8, 9].
Figure 1. Alternative processing of, and signaling by, APP. [5].
In support of this model, the four peptides derived from the amyloidogenic processing of β-amyloid precursor protein (APP)—sAPPβ, Aβ, Jcasp, and C31—have been shown to mediate neurite retraction, synaptic inhibition, caspase activation, and programmed cell death [6, 10-12]; whereas, in contrast, the two peptides derived from the non-amyloidogenic processing of APP—sAPPα and αCTF—mediate neurite extension, and inhibit Aβ production, caspase activation, and programmed cell death 13-15. Thus APP appears to function as a molecular switch, mediating plasticity-related processes, and AD is associated, whether causally or incidentally, with an increase in the ratio of the neurite-retractive peptides to the neurite-extending peptides. Reducing this ratio, whether by affecting BACE (β-site APP cleaving enzyme) or other cleavage of APP, appears to mitigate the AD severity [7, 16, 17].
Of particular interest for the development of a therapeutic program whose goal is to correct the hypothesized chronic synaptoblastic:synaptoclastic imbalance is the feedback mechanism: whereas homeostatic (negative) feedback is utilized by biological systems with single goal outcomes (e.g., serum pH) and no requirement for amplification, prionic loop (positive) feedback is utilized by biological systems with multi-goal outcomes and a requirement for rapid amplification (e.g., thrombus formation or, potentially, synapse modulation), and such systems therefore function as molecular switches [9]. In these latter systems, the positive feedback feature of the systems dictates that the molecular mediators involved, or a subset thereof, beget more of themselves, or enhance their own activities. Thus such amplifying systems are prionic, with the degree of infectivity depending on the stability of the molecular species involved. In the case of APP signaling, binding of a trophic ligand such as netrin-1 increases the production of sAPPα [18], which inhibits BACE cleavage [19], with the complementary fragment, αCTF, inhibiting γ-secretase cleavage [14]; thus cleavage at the α-site produces fragments that inhibit cleavage at the β-site and γ-site rather than feeding back to reduce α-site cleavage. Similarly, cleavage at the β-site and γ-site to produce Aβ feeds back positively to increase APP-C31 production [20], thus favoring the pro-AD, anti-trophic processing of APP. Moreover, Aβ itself has been shown to exhibit prionic properties [21], although the mechanism by which it does so has not been clarified.
Thus APP processing displays positive feedback, and therefore APP and its derivative peptides function as a molecular switch. This has critical implications for therapeutic development, since it offers a mechanism by which a threshold effect occurs. We have taken advantage of this phenomenon to develop drug candidates that increase the anti-AD, trophic APP signaling, while reducing the pro-AD, anti-trophic APP signaling [22] and enhancing cognition [23].
We have found that the manipulation of the plasticity balance that is mediated or reflected by the APP-derivative peptide balance (Fig. 1), whether genetically or pharmacologically, leads to predictable effects on learning and memory. Mutation of the caspase site at Asp664 inhibits the synaptic loss, memory deficits, and dentate gyral atrophy that otherwise occurs in the PDAPP transgenic mouse model of AD [7, 17, 24-26]. Furthermore, knock-in studies of a wild type mouse D664A support the notion that APP is indeed involved fundamentally in plasticity. (Kane, et al, unpublished data, 2014)
Systems biology and systems therapeutics of AD
The transgenic mouse studies suggest that APP signaling can be manipulated to inhibit AD pathophysiology. However, the mouse models feature mutations in APP or other familial AD-related genes such as presenilin-1, whereas the large majority of patients with AD suffer from sporadic AD, without an APP or PS1 mutation (although the majority do express the ε4 allele of ApoE). Given the many inputs to the APP signaling balance in humans (e.g., estrogen, netrin-1, Aβ, etc.), and the minimal success with each of many potentially therapeutic agents (e.g., estrogen, melatonin, exercise, vitamin D, curcumin, Ashwagandha, etc.), the pathobiology of AD dictates a system or program rather than a single targeted agent. Successes with other chronic illnesses such as cardiovascular disease, neoplasia, and HIV support the efficacy of multiple-component systems. My colleague and I have recently described such a system for AD [5]. The basic tenets for such a comprehensive therapeutic system are the following:
Just as for other chronic illnesses such as atherosclerotic cardiovascular disease, the goal is not simply to normalize metabolic parameters, but rather to optimize them. As an example, a serum homocysteine level of 12 μmol/l is considered to be within normal limits, but is well documented to be suboptimal [27]. Similar arguments can be made for many other metabolic parameters.
Based on the hypothesis that AD results from an imbalance in an extensive plasticity network, the therapy should address as many of the network components as possible, with the idea that a combination may create an effect that is more than the sum of the effects of many monotherapeutics [5].
Just as for other chronic illnesses such as osteoporosis, cancer, and cardiovascular disease, the underlying network features a threshold effect, such that, once enough of the network components have been impacted, the pathogenetic process would be halted or reversed. Therefore, even though it is not expected that most patients will be able to follow every single step of the protocol, as long as enough steps are followed to exceed the threshold, that should be sufficient.
The approach is personalized, based on the contributory laboratory values affecting the plasticity network; and is computationally intensive, since many physiological data points are analyzed, interdependent network-component status is assessed, and many interventions are prioritized to determine the therapeutic program.
The program is iterative, so that there is continued optimization over time.
For each network component, the goal is to address it in as physiological a way, and as far upstream, as possible.