Approaching human neuroscience for disease understanding

In psychiatric research, neuroscience knowledge is growing at a record rate, in both the acquisition of facts and the development of mechanistic understanding, at the level of the molecule, the synapse, the cell and the neural system. Whereas, only 20 years ago, we talked about brain function in terms of a “black box”, today we understand many dimensions of brain function mechanistically, especially where molecules and physiology support characteristic behaviors (1). It is not only within genetics and synaptic function where knowledge is growing, but also in identifying postsynaptic signaling pathways, cognition mechanisms, epigenetic modifications and systems neuroscience, to name just a few areas.

Translational scientists are challenged to keep up with relevant new knowledge. Science administrators are thoughtful about motivating the field to use basic knowledge both for the purpose of understanding normal brain function and to identify disease-causing perturbations in disease. There never has been a better time for neuroscience growth or for developing biomarkers and molecular targets for brain diseases. The RDoC system challenges every brain scientist focusing on psychiatric diseases to synthesize and apply relevant brain facts to advantage mechanistic disease understanding (2).

There already exist methodologies to examine in vivo brain function in humans histologically, molecularly and phenotypically, enabling measurements of human brain-based behaviors (3). Cognition is a good example of this, since cognitive capacity can be assessed experimentally and is routinely used to make inferences about functioning of the brain itself. Other approaches, like human brain imaging and evoked potential analyses with electroencephalogram (EEG), all use measures of brain molecular, metabolic or electrical activity to represent neuronal activity regionally. Then, also, some experimental approaches use human postmortem brain tissue for histological or molecular analyses directly, albeit in non-living brain tissue. Regional gene expression, generating region-or cell-specific proteins, could be critical for capturing complex brain function and its regional dysfunction in disease. And animal models, if carefully verified, can contribute improved experimental models.

Then, how do perturbations of these normal human-based systems associate with mental symptoms? Again, here is where the RDoCs system comes in. What the RDoC framework contributes is a system for generating and categorizing brain facts as they relate to putative cross-cutting basic behavioral states or functions of brain, leaving to experimental observation the identification of those perturbed in brain pathology.

It would be incorrect to conceptualize RDoC as a diagnostic system. It is, rather, an approach for systematizing brain knowledge to make it pertinent to functional and dysfunctional systems in the brain as they relate to behavioral outcomes. Nor is RDoCs ready to transform psychiatric diagnosis for all of the practically purposes that ICD and DSM are used for. But, the RDoC system does call attention to the essential need in translational neuroscience to base diagnosis on disease understanding and to tether molecular target development to a detailed and demonstrated disease pathophysiology.

The emphasis in the Cuthbert paper on developing dimensional approaches within mental illness is represented within the domain of psychosis by the Bipolar and Schizophrenia Network for Intermediate Phenotypes (BSNIP) project. References to “psychoses” have been made in the literature for many years, creating an expectation for measurable overlap of biomarkers in brain diseases with prominent psychotic features. Recently, the BSNIP study, using dense biomarkers to characterize psychosis, including schizophrenia, schizoaffective disease and bipolar disorder with psychosis, was launched to explore the dimension of psychosis with modern biomarkers (4).

The study recruited individuals with psychosis and phenotyped them densely, using cognition testing, evoked potential evaluation, eye movement assessment, brain imaging and resting EEG assessment, in addition to a full clinical assessment. The resulting phenotypic characterization of the psychoses diagnoses has created a rich database which can be analyzed for the purpose of creating biological markers for diagnosis.

The BSNIP study showed how biomarkers clustered within and across current DSM diagnoses and, in general, across the psychosis dimension. The high variability and the broad overlap of the biomarkers across diagnoses suggest that our DSM diagnoses are biologically heterogeneous. The additional surprise in these data was the considerable overlap in clinical and diagnostic characteristics. The current BSNIP question is how to move from the present state of partial knowledge in clinical phenomenology and emerging neurobiology, to a state of biological understanding in our psychiatric conditions, a research agenda in the field.

The implication of the BSNIP outcomes and RDoC predictions is that, if we examine current diagnostic groups of psychosis using ideal neural biomarkers, we are still likely to be unsuccessful at defining pathophysiology, because of the gross heterogeneity of the identified groups (5). If we approach disease with a dimension, instead of a single diagnosis, we anticipate, in fact utilize, the marked heterogeneity of the group to recognize biologically similar clusters within the dimension and use the clustering of biomarkers to generate biologically-defined disease groups. The development of validating characteristics for the clusters is the research challenge, namely a common systems understanding or a unifying molecular pathology for these biomarker clusters. The BSNIP approach begins dimensionally, using dense biomarker characterization, to form biologically common clusters, potentially useful as disease identifiers with biological targets.

On the other hand, as Cuthbert suggests, we can also approach disease definitions biologically through identifying the genes, molecules, cells and circuits of normal behaviors, then see which normal functions could be altered when these systems are perverted. The framework of the RDoC system, as it is currently articulated, starts at a detailed level of knowledge of domains for normal behaviors (6). Several of these domains are already relatively well understood. Examples are the constructs of “declarative memory”, “acute threat” and probably also “reward learning”. These normal systems, if abnormally executed, could manifest themselves as “psychosis”, “post-traumatic stress disorder” or “drug abuse”, respectively, if the normal tract is perverted.

In our current state of knowledge which lacks even basic biological clues about the nature of psychiatric illnesses, let alone biological targets, it is not an over-extension to say that we should involve both approaches in discovery and use overlap as concept demonstration.