Bioenergetics and Metabolism: A Bench to Bedside Perspective

Abstract

“Metabolism” refers to the vast collection of chemical processes that occur within a living organism. Within this broad designation one can identify metabolism events that relate specifically to energy homeostasis, whether they occur at the subcellular, cellular, organ, or whole organism level. This review operationally refers to this type of metabolism as “energy metabolism” or “bioenergetics.” Changes in energy metabolism/bioenergetics have been linked to brain aging and a number of neurodegenerative diseases, and research suggests mitochondria may uniquely contribute to this. Interventions that manipulate energy metabolism/bioenergetic function and mitochondria may have therapeutic potential and efforts intended to accomplish this are playing out at basic, translational, and clinical levels. This review follows evolving views of energy metabolism’s role in neurodegenerative diseases but especially Alzheimer’s disease (AD), with an emphasis on the bench-to-bedside process whose ultimate goal is to develop therapeutic interventions. It further considers challenges encountered during this process, which include linking basic concepts to a medical question at the initial research stage, adapting conceptual knowledge gained to a disease-associated application in the translational stage, extending what has been learned to the clinical arena, and maintaining support for the research at each of these fundamentally linked but functionally distinct stages.

Basic science-oriented biomedical investigators attempt to understand life and disease from a mechanistic perspective. A mechanistic understanding of biologic systems, therefore, increasingly focuses on molecular biology. Molecular biology is influenced by bioenergetics and cell energy metabolism. Cells obviously depend on energy and systems have developed to produce energy, monitor its levels, and respond to different energy demands and different environmental conditions that challenge energy homeostasis.

Living systems take carbon sources and derive energy from them or else use the carbon to produce infrastructure. The movement of carbon through cells is called a flux and a number of distinct fluxes are recognized. Glucose carbon, for example, can pass through different fluxes including glycolysis, the pentose phosphate shunt, and the hexosamine biosynthesis pathway. As part of another example, fatty acids are processed as part of the beta oxidation flux. Carbon from glycolysis and beta oxidation can enter the Krebs cycle and during these fluxes the reduced versus oxidized state of certain molecules such as nicotinamide adenine dinucleotide (NAD) and flavin adenine dinucleotide (FAD) is determined.

While defining an organism or its parts from the perspective of fluxesis obviously a very molecular-based view of life, this collection of reactions is what ultimately defines the difference between living and inanimate objects. The reactions that constitute these fluxes can both consume and create energy. This consumption and production of energy allows for the maintenance and perpetuation of the functional unit. Sometimes this system breaks down, causing pathology and resulting in clinical disease.

Substantial efforts have long been underway to characterize molecular-level changes in multiple disease categories, including the neurodegenerative diseases. One thing that has become increasingly clear over several decadesis that when neurodegenerative disease brain pathology is assessed at the molecular level, changes in energy metabolism or the infrastructure that produces energy (especially the mitochondria) are frequently observed (Swerdlow 2009). This has given rise to debate over whether the observed bioenergetic changes represent a perhaps inevitable downstream consequence or even adaptation to declining cell or organ physiology, or whether bioenergetic changes actually meaningfully contribute to the onset and progression of the neurodegenerative diseases they appear in. If the latter scenario turns out to be the case, then targeting bioenergetics would seem to represent a reasonable therapeutic target. Efforts are currently underway to do just that.

Bioenergetics in Neurodegenerative Diseases

Serious concerns about energy metabolism’s role in neurodegeneration were raised in the 1980s. In Parkinson’s disease (PD), a toxin called MPTP was shown to induce the loss of dopamine neurons, and a case was soon made that MPTP, through a by product called MPP+, inhibited complex I of the mitochondrial respiratory chain and that this was responsible for substantia nigra dopamine neuron loss and a resultant parkinsonism syndrome (Langston et al. 1983, Nicklas et al. 1985). Then, in 1989 it was shown that PD patients themselves had complex I activities lower than those of control subjects (Parker et al. 1989, Schapira et al. 1989). Interestingly, reduced complex I activity in PD patients was not limited to just the dopaminergic neurons and was also seen in platelets, muscle, and fibroblasts (Parker et al. 1989, Bindoff et al. 1989, Mytilineou et al. 1994, Swerdlow 2012a, Parker et al. 2008). Efforts to identify endogenously produced complex I inhibitor molecules revealed some potential candidates but not in a conclusive fashion (Drucker et al. 1990, Matsubara et al. 1995, Niwa et al. 1987). Searches for common exogenous, environmental toxins similarly revealed some potential candidates but again could not prove a causal role (Hubble et al. 1993, Makino et al. 1988, Rajput et al. 1987). In 1996 it was reported that transferring platelet mitochondria to cultured cells gave rise to cells with reduced complex I activity that perpetuated in culture (Swerdlow et al. 1996). This suggested mitochondrial DNA (mtDNA) was contributing at least partly to low complex I activity in PD subjects. The cytoplasmic hybrid (cybrid) cell lines that were used to determine this exhibited a number of functional and structural changes that were reminiscent of molecular changes observed in PD brain including aggregated α-synuclein protein (Swerdlow 2012a, Trimmer et al. 2004).

While this stream of discoveries piqued interest in the potential role of mitochondria in PD, much of the field’s attention shifted elsewhere after a number of nuclear DNA genes were identified in Mendelian PD variants. Subsequent efforts focused on trying to understand why mutations in these genes, and the resulting changes in the proteins those genes encode, caused PD. Ultimately, a common theme emerged that specified some mutant proteins either localized directly to mitochondria or that a major role of these proteins was to regulate mitochondrial function, maintenance, or integrity (Pickrell & Youle 2015). This rekindled interest in the idea that mitochondria were potentially important in, and perhaps even central to, PD.

In Alzheimer’s disease (AD) investigators had long noted that electron microscopy pictures of AD subject brain mitochondria looked abnormal (Johnson & Blum 1970, Wisniewski et al. 1970), but it really wasn’t until the 1980s that a possible major role for energy metabolism and mitochondria was proposed. First, fluoro-deoxyglucose positron emission tomography (FDG PET) studies of AD subjects showed areas of reduced glucose uptake in regionally specific brain areas (de Leon et al. 1983, Foster et al. 1983, Friedland et al. 1983). Many reasons for this were postulated, including loss of brain volume or synaptic loss, but it also noted to be possible that reduced glucose uptake reflected reduced glucose utilization independent of anatomical brain changes (Hoyer 1993). Investigators reported activities of the Krebs cycle enzyme α-ketoglutarate dehydrogenase were low (Gibson et al. 1988). Activities were low in brain, and also fibroblasts. Other enzymes localized to mitochondria were altered such as pyruvate dehydrogenase complex and mitochondria from brains appeared to consume oxygen in patterns different from those of brains from non-AD subjects (Perry et al. 1980, Sorbi et al. 1983, Sims et al. 1987). Some investigators at this time even declared AD might primarily reflect a direct manifestation of a metabolic systems dysfunction (Blass & Zemcov 1984). In 1990 it was reported that activity of complex IV, or cytochrome oxidase, of the respiratory chain was reduced in AD patient platelets and this was subsequently shown in AD brains as well (Parker et al. 1990, Kish et al. 1992). In 1997 it was reported in AD cybrids made through transfer of platelet mitochondria to mtDNA-depleted cells that cytochrome oxidase activity started low and remained low, consistent with the possibility that mtDNA was at least partly contributing to the observed AD cytochrome oxidase activity reduction (Swerdlow et al. 1997, Sheehan et al. 1997).

Interest in energy metabolism’s role in AD remained limited, though, as many investigators in the field decided to focus attention on the production of beta amyloid (Aβ) and its deposition into amyloid plaques. Largely based on considerations from rare Mendelian forms of AD and models designed to reflect those forms, the amyloid cascade hypothesis was proposed (Hardy & Allsop 1991, Hardy & Higgins 1992), and for over two decades this hypothesis has to a large extent guided the AD research agenda. Interest in the potential role of bioenergetics and mitochondrial function never fully went away, though, and a number of investigators proposed mitochondria and bioenergetics might yet play an upstream if not primary role in an age-related disease such as AD (Parker et al. 1990, Wallace 1992, Beal 1995, Hirai et al. 2001). Plus, research subsequently emerged that linked cell bioenergetics to amyloid precursor protein (APP) processing. This work showed that by manipulating energy metabolism APP processing could be directed away from its non-amyloidogenic processing pathway and probably to its amyloidogenic, Aβ-producing pathway (Gabuzda et al. 1994, Webster et al. 1998, Gasparini et al. 1997). In 2004, Swerdlow and Khan proposed the “mitochondrial cascade hypothesis,” which attempted to comprehensively explain mitochondrial and bioenergetic dysfunction, the appearance of AD pathology, and the association of increasing AD risk with advancing age (Swerdlow et al. 2010, Swerdlow et al. 2014, Swerdlow & Khan 2004, Swerdlow & Khan 2009).

Interest in the potential role of mitochondria in AD received a boost when investigators began to report APP and Aβ physically associated with mitochondria (Anandatheerthavarada et al. 2003, Anandatheerthavarada & Devi 2007, Devi et al. 2006, Manczak et al. 2006, Lustbader et al. 2004, Crouch et al. 2005, Hansson Petersen et al. 2008). This was an important observation for some as it offered a possible mechanism that could explain why Aβ might be toxic to neurons (as was proposed in principle by the amyloid cascade hypothesis as well as the results of cell culture studies showing that Aβ or derivatives of A β could be cell-toxic in culture). Indeed, it was shown that A β was toxic to NT2 cells with functional mitochondria, but not to NT2 ρ0 cells depleted of endogenous mtDNA, which lacked functional respiratory chains (Cardoso et al. 2001). This finding suggested A β toxicity could be mediated through its effects on mitochondria.

Overall, there are currently different perspectives on what role mitochondria may play in AD (Swerdlow 2012b). Some envision mitochondrial changes as a downstream, albeit functionally important consequence of Aβ toxicity (Lustbader et al. 2004, Du et al. 2008). This idea is consistent with and falls within the framework of the amyloid cascade hypothesis. Others postulate changes in cell energy metabolism occur upstream to changes in Aβ, may cause AD, and that a mitochondrial cascade hypothesis may be more likely (Figure 1). Those who support this latter view point out the fact that mitochondrial dysfunction is consistently observed outside the brains of AD patients, in places where Aβ is unlikely to represent the primary cause of the mitochondrial dysfunction (Swerdlow 2012b, Swerdlow et al. 2010).

Figure 1. Two potential interactions between mitochondria and Aβ.

(A) Aβ interacts with mitochondria to interfere with mitochondrial function. (B) Mitochondrial dysfunction alters APP processing so that Aβ is generated. Data supporting both scenarios exist.

Changes in mitochondria, mitochondrial function, and brain energy metabolism are observed in a number of other neurodegenerative diseases including amyotrophic lateral sclerosis (ALS) and Huntington’s disease (HD) (Swerdlow et al. 1998, Browne & Beal 2004). Other neurodegenerative diseases, such as Leber’s Hereditary Optic Neuropathy (LHON) have been associated with mtDNA mutations (Wallace et al. 1988). Other reviews discuss in greater detail the changes in mitochondria and energy metabolism that have been demonstrated in other neurodegenerative diseases (Swerdlow 2009, Lin & Beal 2006).

Therapeutic Targeting of Bioenergetic Function

For neurodegenerative diseases with altered bioenergetic function, bioenergetic manipulation represents a reasonable therapeutic target (Swerdlow 2011). What exactly to manipulate, though, is unclear and many potential targets exist. Stressed mitochondria may overproduce free radicals and radical scavenging has been proposed as a possible therapeutic approach (Lin & Beal 2006, Perry et al. 2002). Cell calcium levels may change when mitochondria are stressed and this could represent a viable target (Lemasters et al. 2009). Stressed mitochondria may lose membrane integrity, with egress of contents to the cytoplasm. Some have considered stabilizing mitochondrial membranes, or blocking the mitochondrial permeability transition (Mattson 2000, Crompton et al. 1999, Saelens et al. 2004, Wang & Youle 2009). Removing impaired mitochondria by enhancing mitochondrial autophagy has been proposed (Rabinowitz & White 2010, Youle & Narendra 2011, Banerjee et al. 2010). Mitochondrial fission and fusion balance has been shown to be perturbed in some diseases and addressing this could have therapeutic potential (Chan 2006, Chen & Chan 2009, Manczak et al. 2011, Wang et al. 2008, Wang et al. 2009, Zhu et al. 2013).

An obvious pathologic consequence of mitochondrial dysfunction could be decreased energy production. This would impart an overall stress to the entire cell, which might in turn alter signaling pathways and protein post-translational modifications. Redox balances (such as NAD(P)+/NAD(P)H and ratios of other redox paired molecules could also represent a consequence of decreased energy production. Bioenergetic fluxes would also predictably change, thereby changing cell levels of intermediates needed to facilitate the synthesis of required molecules, such as nucleic acids used for DNA and RNA production, and fatty acids used for membrane and cholesterol synthesis.

To address neurodegenerative disease-associated alterations in energy metabolism and mitochondrial dysfunction, a number of proposed therapeutic approaches have so far been tested in clinical trials. Antioxidant trials have shown at most very limited benefits (Sano et al. 1997, Dysken et al. 2014). Drugs claimed to stabilize mitochondrial membranes have not shown clear benefits (Gordon et al. 2007). Providing compounds with the ability to store energy has been tried. For example, supplementing creatine with the intent of increasing levels of creatine phosphate, a molecule that cells can use to store high energy phosphate bonds and that could theoretically be used to support ATP production, have not shown success (Bender et al. 2006, Verbessem et al. 2003). Sometimes it is hard to know exactly what mechanism an intervention of interest might work through. For example, coenzyme Q and coenzyme Q analogs have been studied in clinical trials (Mancuso et al. 2010, Di Prospero et al. 2007, Gutzmann & Hadler 1998, Thal et al. 2003). For coenzyme Q, would the presumed mechanism include free radical scavenging, or electron transfer to downstream electron transport chain holoenzymes?

Regardless, in neurodegenerative diseases that show altered bioenergetics the observed alteration is consistently one of reduced capacity. Fluxes are less than normal which suggests increasing fluxes may prove therapeutically useful. This might be accomplished through several approaches. One includes increasing mitochondrial mass through the induction of mitochondrial biogenesis (Swerdlow 2007b, Swerdlow 2011). Increasing levels of bioenergetic intermediates may impact upstream and downstream fluxes through mass action or allosteric modifications of enzymes in the pathway (Swerdlow 2014). Changing redox balance within cells, such as changing NAD(P)+/NAD(P)H ratios could in this respect conceivably have an effect.

Overall, though, when considering potential mitochondria-directed therapeutic interventions it is important to note experimental data reveal manipulating mitochondria can have variable, sometimes even unanticipated consequences. For instance, genetically engineered reductions (as opposed to enhancement) of aerobic capacity have been associated with extended longevity in drosophila and C. elegans (Lee et al. 2003, Copeland et al. 2009). In mice, genetically engineered reductions in mitochondrial function also reportedly increase insulin sensitivity, counter obesity, and raise the diabetes threshold (Pospisilik et al. 2007, Quintens et al. 2013, Vernochet et al. 2012, Wredenberg et al. 2006). Perhaps consistent with this finding a drug commonly used to treat type 2 diabetes, metformin, inhibits complex I (Leverve et al. 2003). In humans, complex I gene variations believed to reduce complex I function are also reported to associate with a longer lifespan (Raule et al. 2014). Conversely, specifically impairing mitochondrial and respiratory chain function can undoubtedly drive neurodegeneration (Szabados et al. 2004, Duty & Jenner 2011, Zhang et al. 2015), and accelerating mtDNA accumulation drives aging phenotypes (Trifunovic et al. 2004, Kujoth et al. 2005). Examples such as these emphasize mitochondrial manipulations can have protean consequences, with subtle but distinct manipulations yielding very different clinical outcomes. Such distinctions should also predictably influence the clinical efficacy (or lack) of a particular intervention.

An Example of a Bench-to-Bedside Therapeutic Program

An example of a bench-to-bedside program now being pursued for the treatment of a neurodegenerative disease is one in which we have sought to use both a shift in redox balance and anaplerosis (the provision of carbon to the Krebs cycle) to increase bioenergetic fluxes and promote mitochondrial biogenesis (Swerdlow 2014). In developing this program we initially considered data showing that two non-pharmacologic interventions that show health benefits and which should conceptually alter cell bioenergetics (at least in some tissues), caloric restriction and physical exercise, increase NAD+/NADH ratios or promote the ability to increase this ratio (Lin et al. 2004, Lin & Guarente 2003, Haigis & Guarente 2006, Yang et al. 2007, Costford et al. 2010). We therefore considered supplementing cells with molecules that are reduced by NADH, with the idea that increasing their levels would convert cytosolic NADH to NAD+. We considered molecules such as pyruvate (which is reduced by NADH to form lactate) and oxaloacetate (OAA, which is reduced by NADH to form malate). We ultimately settled on OAA because we wanted to ensure we did not inadvertently minimize glycolysis flux by loading cells with pyruvate, a downstream glycolysis pathway intermediate (Williamson & Jones 1964). By increasing the cytosolic NAD+/NADH balance we hoped to increase the glycolysis flux, while at the same time generating an intermediate (malate) that could access the mitochondrial matrix and act in an anaplerotic fashion (Figure 2).

Figure 2. Strategy for enhancing bioenergetic fluxes.

The intent of administering OAA was to induce reduction of OAA to malate by the cytosolic malate dehydrogenase. Increasing the NAD+/NADH ratio in this way was intended to increase the glycolysis flux, and provide carbon to the mitochondria in the form of malate. Malate’s entry into the Krebs cycle, in turn, was intended to provide reducing equivalents to the respiratory chain and stimulate respiration. The blue structure represents the mitochondrial matrix, and outside the blue is the cytoplasm.

As part of our preclinical development efforts we administered OAA to mice via intraperitoneal (IP) injection (Wilkins et al. 2014). The brains of mice subjected to this intervention showed activated mitochondrial biogenesis pathway signaling, enhanced insulin pathway signaling, reduced neuroinflammation signaling, and increased hippocampal neurogenesis. Magnetic resonance spectroscopy (MRS) was performed on mice shortly after receiving an OAA IP injection, which showed the intervention increased levels of some brain intermediates including lactate and glutathione. While these MRS data do not tell us why lactate and glutathione levels increased, a potential explanation is that passage of glucose into glycolysis and possibly also the pentose phosphate shunt increased.

These preclinical data were used to justify moving forward into human studies. Major questions at this time included the safety of OAA, as well as how much to administer to subjects. To begin to address this, we designed a safety and pharmacokinetic study of OAA in a small number of human AD subjects. Compared to what the mice were administered, the human subjects received a much lower dose, which was given orally. These subjects took OAA 100 mg twice a day for one month and were followed for adverse events, none of which were observed. Our pharmacokinetic testing found that 100 mg of oxaloacetate had very little effect on serum OAA levels, which suggested that to test for efficacy, higher doses would be required.

We therefore had to test higher doses, but as the safety of higher doses was unknown we needed to acquire additional safety information. A phase Ib ascending dose safety study in AD subjects was therefore planned. This study was also designed to acquire pharmacokinetic data, and in order to provide insight into target engagement MRS and FDG PET biomarker data will be acquired. The MRS component was included based on the preclinical data that showed systemically administered OAA increased brain lactate and glutathione levels (Wilkins et al. 2014). FDG PET is included to assess whether these changes are indeed likely to reflect increased glycolysis fluxes.

It is important to point out that while at the conceptual level data justify pursuing mitochondria biogenesis for the treatment of AD, cautionary data also exist. Supportive findings include the fact that peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α), numbers of normal-appearing mitochondria, and expression levels of respiratory chain transcripts are all reduced in the AD brain (Sheng et al. 2012, Qin et al. 2009, Hirai et al. 2001, Liang et al. 2008). On the other hand, concomitant PGC1 α overexpression in APP-overexpressing transgenic mice actually increases fibrillary Aβ (Dumont et al. 2014). Concern over this point is perhaps mitigated by the fact that human subjects tend to tolerate the acquisition of fibrillary Aβ for years and possibly even decades without showing clinical symptoms (Jack et al. 2010, Sperling et al. 2011). Indeed, by the time clinical symptoms manifest further accumulation of fibrillary Aβ appears to occur at a very slow rate, certainly slower than the rate at which it accumulates during the asymptomatic period (Jack et al. 2013, Burns & Swerdlow 2013). This suggests our current understanding of the role of Aβ in this disease, and especially the factors that underlie its accumulation, is incomplete.

Bench-to-Bedside Research : General Considerations

Bench-to-bedside biomedical research encompasses a spectrum of research disciplines – basic science, translational science, and clinical science. While there certainly is overlap between these categories and it may be difficult sometimes to define what stage of research a project is actually in at a given time, heuristic distinctions between these three categories are often emphasized.

The goal of basic research is to help us better understand a particular phenomenon. In the case of the biological and chemical sciences, basic research advances can be leveraged to improve the human condition, often by improving health. Attempts to adapt and extend basic research advances for this purpose defines a process currently refer red to as translational research. The active translation of a basic, fundamental observation, though, does not guarantee the knowledge gained will ultimately impact human health at public or disease levels. How to implement a translational advance, and the utility of implementing such an advance at the human level, requires further clinical research.

While this approach makes logical and logistical sense, in practice biomedical research often deviates from this sequence. For example, the serendipitous observation that a particular intervention alters the course of a disease may prompt research into why the intervention had such an effect in the first place. Also, perhaps due to the fact that in the United States the largest sponsor of life science research is the National Institutes of Health (NIH) (Rao & Collins 2012), investigators may feel it is necessary to define the translational value of fundamental research before that value can be accurately characterized.

Bench-to-bedside programs need to consider a number of questions, such as how well we truly understand the disease of interest, how well do our models model the disease of interest, and how can one extend what has been learned from preclinical studies to the clinical arena. In the example provided above, a case can be made that our understanding of the cause or causes of AD is incomplete. The AD field is arguably much better at identifying molecular deviations from the norm than it is in explaining how those deviations arose and to what extent those deviations contribute to the disease phenotype. So, while a reasonable argument that addresses the question of why a particular pathology or pathophysiologic change represents a valid therapeutic target can often be made, the decision to pursue that target requires the assumption of risk.

In general, current mechanistic disease research relies to a great extent on disease models. It is important to note the limitations of models. Some disease models address some aspects of a disease better than they address others, and can downplay or overplay the contribution of a particular disease-associated parameter. For reasons such as this, findings from animal models may imperfectly extrapolate to the human condition. Experience tells us that therapeutic interventions that show efficacy when tested in disease models frequently do not show similar effects when tested in human subjects (Swerdlow 2007a).

Beyond the pitfalls of trying to extrapolate data from animal models that imperfectly recapitulate diseases we incompletely understand to the clinical arena, issues of how to translate our preclinical treatment experience to human subjects present multiple practical issues. This includes technical limitations in our ability to deliver interventions. Even with drug interventions that rely upon traditional delivery approaches, routine questions pertaining to absorption, distribution, metabolism, excretion, and toxicity (the classic “ADMET” parameters) need to be addressed. Indeed, ADMET issues typically need to be answered twice, at the pre-clinical stage in animals, and at the clinical research stage in humans. For drugs that have not previously been approved for human use by the Food and Drug Administration (FDA), in addition to procuring approval from the appropriate Human Subjects Committee Institutional Review Board, ADMET testing may require an FDA review of the drug development plan.

In the preclinical development stage there is more leeway to test drugs for target engagement at doses that may ultimately exceed those that are truly necessary. This lets investigators establish proof of concept at a fairly early stage. In other words, if it can be demonstrated that a drug has an intended effect, one can then justify trying to find an optimum dose. Demonstrating an effect in cell culture and even in animals is less constrained by safety concerns. In human clinical studies, though, ensuring the safety of study subjects takes precedence over demonstrating an effect.

Sequentially accomplishing each preclinical and early clinical stage outcome, though, in no way guarantees late stage clinical trials will show a positive outcome or evidence of efficacy. As discussed above bench-to-bedside drug development efforts encompass multiple steps and unexpected or unforeseen factors at any point can subvert the developmental process. The conceptual rationale for a particular approach can misfire, for example, if a proposed target turns out to represent a compensatory as opposed to disease-driving change. At the basic development stage effects observed in particular models or with particular delivery approaches may yield effects that simply do not or cannot extrapolate to humans. Similarly, at the preclinical translational stage animal disease models may not accurately reflect the human condition. Early clinical development may be handicapped by an inability to find a dose that is both safe and engages the biologic target, or that engages the biologic target to an adequate extent. These pitfalls no doubt collectively account for why different strategies, including mitochondrial manipulations or manipulations of bioenergetic-related phenomena (such as oxidative stress), have been attempted in several neurodegenerative diseases with negative or at best unimpressive results. Many of these trials have been previously reviewed elsewhere (Swerdlow 2011). In most cases, it is difficult to know with certainty the exact reason (or reasons) why a particular intervention ultimately failed.

Biomarkers

In the biomedical research setting, biomarkers are frequently thought of as measurable parameters that serve as surrogate endpoints of a disease. This, however, is not the only type or use of biomarkers. Biomarkers may serve as “diagnostic biomarkers” and as such assist in diagnosis. They may provide insight into disease etiology or pathology, or predict the course of a disease. Changes in a biomarker endpoint may predict a response to treatment. Biomarkers can also be used to indicate “target engagement.”

In cases where there is a limited understanding of a disease’s mechanisms, the significance of a diagnostic biomarker may be unclear. For example, for individuals with late-life progressive dementia an accumulation of brain febrile Aβ has long constituted a well-recognized biomarker of AD. The antemortem detection of brain fibrillar Aβ, though, reveals the brains of many cognitively intact older individuals also contain fibrillar A β (Swerdlow 2007a). In an effort to clarify this potentially confusing point, the field has defined a stage of AD called “preclinical” AD, in which the Aβ biomarker in the absence of clinical symptoms is sufficient to indicate the disease is present (Sperling et al. 2011). This convention tells clinicians how they should interpret the presence of a positive amyloid scan but raises other questions, including the question of what a diagnosis of AD truly implies. Is it simply the presence of brain fibrillar amyloid? If so, if the presence or absence of brain fibrillar amyloid itself defines the presence or absence of AD, can Aβ still also be considered a biomarker of the disease since it is implied that Aβ itself is the disease? Independent of these considerations, the presence of fibrillar amyloid does not address why fibrillar amyloid accumulated in the first place.

The less understanding we have about what drives or controls the appearance of a biomarker, the harder it is to know whether treatment-induced changes in the biomarker predicts a positive clinical response, a negative clinical response, or no clinical response. The most straightforward use of a biomarker, therefore, may be to help investigators determine whether a drug is engaging its biochemical or molecular target (Macchi et al. 2015). For the OAA bench-to-bedsided evelopment program discussed above, target engagement biomarkers are being used to address whether doses evaluated in a safety study are able to alter brain bioenergetics. This target engagement data will help indicate when a potentially adequate dose has possibly been identified. Changes in the target engagement biomarkers would justify investing in a larger, longer, more expensive therapeutic trial. A lack of change would suggest higher doses should first be vetted before advancing to a therapeutic trial.

In the particular case of targeting brain bioenergetics, an inability to physically access the brain itself complicates and limits biomarker development. Functional imaging approaches such as FDG PET, MRS, and functional magnetic resonance imaging (fMRI) may provide target engagement insight but observed responses may incompletely reflect or explain how or why an intervention truly affects the brain. Additional mechanistic information may be ascertained through analyses of procurable tissues, including cerebrospinal fluid or blood, but it must be considered that analyzing surrogate tissues may not rigorously reflect what occurs in the brain itself. It is nevertheless possible that using neuroimaging approaches in conjunction with procurable surrogate tissues may yield a fuller picture than either strategy can when used alone. Of course, including additional clinical endpoints also requires additional resources.

Conclusions

The world’s largest supporter of neurochemistry research is currently the NIH. Their mandate is to improve public health at both population and individual levels. Perhaps for this reason, many neurochemistry research projects are now conceptualized, presented, and pursued as “biomedical” research projects. This infers the ultimate goal of the research is to advance our understanding of a particular neurologic disease or to identify disease treatments. Accordingly, neurochemistry investigators may feel compelled to identify a bench-to-bedside timeline or pathway that emphasizes the potential clinical impact of the area of study. Each step of this bench-to-bedside process has its own set of challenges. A careful consideration of these challenges can affect whether a basic research project ultimately evolves to the level of a clinical application.

For this anniversary issue, this review has sought to present an example of a neurochemistry-based, bench-to-bedside research program. This program focuses on AD, a neurodegenerative disease we do not fully understand at the mechanistic level, but which shows clear-cut bioenergetic dysfunction and mitochondrial impairment (Swerdlow 2012b). The presence of bioenergetic dysfunction and mitochondrial impairment in human AD subjects implicates bioenergetics and mitochondria as justifiable therapeutic targets. While different approaches may be used to achieve these ends, we have used a metabolism intermediate, OAA, to manipulate glycolysis, Krebs cycle, and oxidative phosphorylation fluxes and activate mitochondrial biogenesis pathways. In pursuing this goal, the challenges we have met along the way are fairly typical of bench-to-bedside research programs, and illustrate the process of moving from the preclinical to early-stage clinical setting.

Abbreviations

Aβ

beta amyloid

AD

Alzheimer’s disease

ADMET

absorption, distribution, metabolism, excretion, and toxicity

ALS

amyotrophic lateral sclerosis

APP

amyloid precursor protein

Cybrid

cytoplasmic hybrid

FAD

flavin adenine dinucleotide

FDA

Food and Drug Administration

FDG PET

fluoro-deoxyglucose positron emission tomography

fMRI

functional magnetic resonance imaging

HD

Huntington’s disease

IP

intraperitoneal

LHON

Leber’s Hereditary optic neuropathy

MRS

magnetic resonance spectroscopy

NAD

nicotinamide adenine dinucleotide

MPP+

1-methyl-4-phenylpyridinium

MPTP

1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine

mtDNA

mitochondrial DNA

NIH

National Institutes of Health

OAA

oxaloacetate

PD

Parkinson’s disease

PGC1α

peroxisome proliferator-activated receptor gamma coactivator 1-alpha