An exploration of the complexities of defining neuronal identity and why function is its final arbiter
These are phenomenal times for neuroscience. Fueled by revolutionary technologies, neuroscience has caught the public’s attention like never before. From Barack Obama’s BRAIN initiative, to the investor buzz on Sand Hill Road, to this February’s issue of National Geographic, it is literally neuroscience EVERYWHERE. As a researcher of brain architecture, the permeabilizing effect of these technologies on molecular, systems, computational, and behavioral neuroscience has been truly remarkable. Just think about it. Today, we can illuminate the morphology of specific neurons with BAC transgenics; trace their precise circuitry with viruses; watch them fire action potentials during actual behavior with fingertip microscopes; and silence their activity through optogenetics. This was science fiction when I first entered a laboratory thirteen years ago.
But it is easy to be seduced by technology. Real progress will arise not from innovative technologies alone, but from the caliber and magnitude of questions that they empower us to answer. In all the excitement of frontier neuroscience, it is imperative that we step out of the picture and contemplate: Are we truly closer to elucidating the fundamental principles of brain structure, function, and disorder?
I have two goals for my blogposts. First, I look forward to sharing my insights and opinions about recent articles and discoveries in the neuroscience arena. Second, I intend to make these blogposts an expository space that provides context and big picture analysis, and that brings to the forefront questions that do not have crisp answers. Yet.
The Complexities of Neuronal Identity
One such question that has consumed my mind-space and occasionally complicated research efforts is how to define the identity of a neuronal cell type? It is of utmost importance to get a handle on this question because a central dogma of neuroscience is that specific cell types control specific functions in complex nervous systems. However, conceptualizing standardized frameworks that classify neurons into distinct cell types and evaluate their identity (distinct but not mutually exclusive tasks) is difficult. Neuronal identity is extremely contextual, and thus, different facets of identity become more relevant or prominent in different settings. For example, olfactory sensory neurons (OSNs) of the olfactory epithelium are best defined by their choice of odorant receptor expression. On the other hand, projection neurons of the six-layered neocortex are routinely classified into distinct cell types by their laminar location and target choice of their long distance axons (think layer 6 corticothalamic, or layer 2/3 callosal). Additionally, in the absence of a consensus definition of neuronal identity, investigators subjectively apply different molecular, morphological, and electrophysiological parameters based on personal preferences, biases, and expertise. This further complicates standardization efforts.
Classifying Neurons by Molecular Ground State
In a recent issue of Neuron, Gord Fishell (NYU) and Nathaniel Heintz (Rockefeller) make a brave attempt to standardize neuronal classification. Their Perspective titled, “The Neuron Identity Problem: Form meets Function”, robustly argues that “molecular ground state” is the optimal parameter to classify neurons into distinct cell types. The authors suggest that molecular ground states can be identified from gene expression profiling (eg. microarrays, RNA sequencing) of randomly chosen individual neurons or genetically targeted populations. They define the molecular ground state of a neuronal cell type as a suite of genes (potentially, enriched transcription factor mRNAs) that is: 1) Stably expressed throughout the life of the cell; 2) Universally expressed by every cell of that population; 3)Uniquely expressed i.e. not expressed by any other cell types; 4) not activity-dependent or stochastically expressed to diversify finer functional capacities of individual cells within that type. The authors emphasize that these genes should also broadly determine functional capacities, and should be such that “it is not possible to identify subprofiles that further divide the population into stable, defined subtypes of cells”.
Classification of neurons by molecular ground states is indeed logical and intuitive. However, given the stringent criteria, identification of different ground states may not be equally feasible for every context. For example, classification of OSNs of the murine olfactory epithelium by odorant receptor (OR) expression into ~ 1500 cell types is straightforward; each OSN expresses only 1 of the ~1500 OR genes throughout its life, which also dictates its connectivity and functionality. On the other hand, identification of stable and unique molecular ground states to classify neocortical projection neurons will be difficult; gene expression, especially of “fate” specifying genes is highly dynamic and overlapping between neocortical cell types, and may not predict function. At some point, differential gene expression will have to give way to differential levels of gene expression, and this can affect reproducible identification of the same neocortical cell type. Additionally, while molecular ground states might simplify neuronal classification, they have limited utility in neuronal identity evaluation.
Function as the Final Arbiter of Neuronal Identity
Neuronal identity evaluation is indeed a different beast. As the authors write, “Although it is apparent that cell types can be defined molecularly, an understanding of the nervous system cannot be reached without comprehensive data regarding the circuits in which they are embedded, their connectivity, and their activity patterns in response to appropriate external stimuli”. Thus, identity evaluation of a cell type requires an “all-inclusive accounting” of its myriad properties that also draws out its “uniqueness” in relation to others cell types. Perhaps, evaluating neuronal identity requires an exhaustive list.
Here’s what such a list might look like:
1) Developmental origin
2) Birth date
3) Final location
4) Gene expression profile
5) Epigenetic landscape
6) Neurotransmitter/neuropeptide phenotype.
7) Specialized gene expression; e.g. odorant receptor choice
8) Cell body dimensions
9) Dendritic Architecture
10) Long distance axon targets
11) Local axonal branching pattern
12) Synaptic connectivity (e.g. direct monosynaptic inputs)
13) Modality: Cognitive, Motor, Sensory
14) Electrophysiological properties
15) Functional connectivity
16) Specialized features e.g. ocular dominance.
Now of course, not every element of this exhaustive list is equally applicable to every context and experimental paradigm. They must be selectively applied depending on the question being investigated. And this in turn makes standardizing identity evaluation a greater challenge than standardizing neuronal classification. Nonetheless, for most experimental paradigms, this limitation can be substantially overcome by performing a comprehensive and multidimensional evaluation at a molecular, structural and functional level. It is indeed worrying that today most studies (especially of the “before-and-after”, and “fate-switch” genre) begin and end with a molecular and (perfunctory) structural evaluation. While these are important first steps, function remains the most veritable indicator of identity. Two neurons with completely different ion channel profiles can exhibit the same firing pattern. Two neurons with vastly different morphologies can perform the same computation. Countless studies have shown that an altered molecular identity may not manifest as a change in circuitry. That an altered circuitry may not manifest as a change in function. At the end of the day, function is and remains the final arbiter of neuronal identity. To quote Deng Xiaoping, father of China’s economic reforms: “It doesn’t matter if a cat is black or white, so long as it catches mice”.
The author’s views are entirely his or her own and may not reflect the views of Inscopix.