Can We Prevent Dementia and Not Prevent Neurons from Dying?

Many would agree that “the amyloid hypothesis is in deep trouble,” [1] but why? Recently we joined others lamenting the failure of the Amyloid Hypothesis to yield any successes among trials of 300–400 Alzheimer’s disease (AD) drug candidates [2]. The Amyloid Hypothesis’ neuropathologies—most specifically but not exclusively, amyloid-β peptide (Aβ), its oligomers, amyloid deposits, and hyperphosphorylated tau tangles in neuronal bodies and axons—have long been recognized as probable causes of neuronal cytotoxicity and dementia in AD. Extensive studies have sought to identify the roles of each of these proteins, their genetics, and relations to the neuropathological cascade into clinical AD. Yet across more than two decades, in numerous clinical trials, AD investigators have failed to associate the range of Amyloid Hypothesis genotypic pathologies with progression-free AD survival. Presymptomatic interventions with subjects at risk for AD have also failed. These consistent failures of the Amyloid Hypothesis leave academia and industry today demoralized, engaged in fraught efforts to address human efficacy with old targets, and seeking new and promising AD targets to pursue, yet with the risk that these targets also may be drug modifiable but not able to prevent neuronal dysfunction, death, and resulting clinical dementia [1]. Given the extensive history of “troubling method and practice lapses in neuropsychiatric drug developments”, we decided to reevaluate the Amyloid Hypothesis as a strategic source for structuring future AD drug developments [3].

Seeking a new approach to disease modification in AD, we reexamined the assumptions of the Amyloid Hypothesis to identify any critical steps in the progression into AD that must be averted for any therapy to be successful. As a result, we found guidance in two medical traditions to condition our studies. First, to inform our efforts, we returned to medicine’s scientific and humanistic roots in the Hippocratic tradition. Scientifically, leaders in medicine’s historical tradition, such as the Canadian Hippocrates Sir William Osler, teach us that we learn medicine from our patients [4]. Second, the primacy of the best interests of the patient govern clinical pharmacology practices by distinguishing between experimental and investigational administrations of drugs to persons. An experimental administration involves a compound that lacks justifying preclinical animal studies and violates the restrictions on human experimentation inspired by the Nuremberg Trials [5]. An investigational administration of a drug candidate follows as full as possible qualification as a candidate in preclinical studies. An experimental administration occurs when a human subject receives a drug candidate that has not previously been as fully as possible studied to anticipate human safety and efficacy for an intended use. Turning to our clinical experience with AD patients, we noted that in their behaviors each patient taught us that she was experiencing the onset of a clinical dementia. Similarly, in their autopsies, each individual taught us that she had suffered severe losses of neurons at some time in the course of her disease. As others have noted, the presence of pathognomonic AD pathologies at autopsy are not invariably tied to clinical AD [6]. This close association of dementia with neuronal losses but not with inflammatory, amyloid, oxidative stress, and other pathologies is far from news. Alois Alzheimer, in his 1906 publication, describes the prominent amyloid and tangle lesions in the AD brain that investigators have been unable to associate with dementia [7]. Alzheimer also observes how losses of neurons confirmed at autopsy invariably accompany the dementia. The immediate causes of neuronal losses are still not known, neither are their dysfunction and death clearly correlated with the states of cognitive dysfunction experienced in the course of AD. Recent work has shown the mosaic incorporation of APP variants into the neuronal genome risk productions of toxic proteins that could account for AD neuronal dysfunction and deaths [8]. These wide-ranging observations of an anatomical but not neurochemical relationship to dementia point to a worrying potential weakness in the assumptions that contextualize the Amyloid Hypothesis. The hypothesis does not require that the pathological intervention being addressed by a drug candidate prevent the neuronal death that is present at autopsy in clinical AD. This could explain the recent widespread lack among AD drug candidates of efficacy against dementia.

In response, to identify a critical target we turned to a clinical or phenotypic search for a functional impairment that would inevitably lead to clinical dementia. We chose the neuronal dysfunction that would occur in dying neurons and the autopsy evidenced death of neurons as a critical target to be addressed in AD drug development. Because techniques were not available to assay biomarkers able to identify the neuronal programmed cell death (PCD) cascade in humans, we also turned to development of assays of brain marked exosomes as essential to the success of our project. The range of pathologies encompassed by the Amyloid Hypothesis and shared with traumatic brain injury (TBI) [9] led us to employ various animal models; first for study of shared biomarker characteristics indicating the presence of neuronal cell death and then, using these models, as potential screens of drugs for abilities to protect AD affected neurons from PCD [10].

Neuronal cell death in AD is not necrotic and occurs from processes of PCD [10]. In a weight drop TBI mouse animal model, a controlled cortical impact TBI mouse model, a soman stress model, in treating AD transgenic (tg) and wild mice, and in a rat anoxia model, we demonstrated overlapping evidence of neuronal death and of PCD being present [9–13]. To extend study of the mechanism of PCD and drug effects on PCD into humans, colleagues and we are developing the use of exosomes as sources of assays of brain cell biochemistry. Exosomes are plasma sourced extracellular vesicles released from brain neurons and astrocytes, and are marked by proteins that index their cellular origins [14]. In a screen for a pharmacological probe of PCD, we identified (-)-phenserine as sharing preservation of behavioral functions, brain neurons, reductions of lesion sizes, and alterations in markers of PCD pathways independent of activities affecting previously acknowledged Amyloid Hypothesis identified critical AD drug targets [9–13].

The inability of existing tg AD animal models to predict reliable protection against neuronal death calls for modification of the strategies developed to implement the Amyloid Hypothesis and possibly independent sources for choosing new approaches to AD drug development. Neuronal losses are variable across currently used tg AD animal models, not regularly associated with Aβ₄₂ peptide concentrations, never devastating for the animal, and human neuronal-tg mouse hybrids are impractical for drug screening [15–17]. Preclinical studies limited to tg mouse AD model subjects, where neuroprotection does not receive stringent preclinical testing, have not proven adequate to support efficacy of drug candidates in the clinic. Because studies of any AD pathologies will be limited by the inability of present tg animal models to assess neuroprotection adequately, we see no alternative for any new AD drug other than testing with supplemental or alternative models.

We propose for AD drug candidates preclinical testing with weight drop or other PCD evoking animal models or, where this is not pharmacologically indicated, an initial translational probing in humans with the drug candidate to confirm drug candidate protection of neurons against PCD. We suggest this human translational study or a demonstration in a head injury model of activity against PCD or equivalent testing any other function regarded as critical to dementia. We consider this expansion of testing of drug candidates as required in AD drug development before proceeding into extended Phase I safety testing or human efficacy-oriented Phase II and III drug candidate testing. This investigative drug development methodology complies with the declarations of human rights that followed the Nuremberg trials and that requires all possible preclinical evaluation prior to any human drug exposure [5]. These considerations, of the clinical responses to disease as seen in individual patients and compliance with all possible preclinical testing of conditions used in human research, led us to adopt two criteria for preclinical AD course modifying drug developments. We propose first that animal models or, as necessary, small as possible human translational studies, test for predicted efficacy against clinically critical features of the disease. An AD drug candidate would not enter human development without evidence that any known effects on specific molecular targets or neuropathologies results in prevention of neuronal death. Second, we propose that such stringent testing become a precondition to AD drug candidates entering human Phase II and III clinical trials. Critical testing of AD drug targets should, as we intend with our testing for effects from inhibition of neuronal PCD, provide confirmation of why the drug has been successful and, in the case where the drug fails, whether it is the drug or target that has not been confirmed.

Perhaps the Hippocratic tradition also implies the use of clinical or phenotypic drug targets. These can become, as has PCD for us, subjects of study programs in a drug development. Phenotypic assays are agnostic to hypotheses underpinning disease molecular mechanisms and drug targets. Identification of phenotypic targets has previously resulted in the greatest number of successful first-in-class drugs [18]. The physician drug developer may, in important ways, still learn medicine and its clinical pharmacology from her patients [19].