Larkum 2018 - Cortical Layering

A Perspective on Cortical Layering and Layer-Spanning Neuronal Elements

Larkum Matthew E., Petro Lucy S., Sachdev Robert N. S., Muckli Lars

Frontiers in Neuroanatomy, 2018

https://www.frontiersin.org/articles/10.3389/fnana.2018.00056
10.3389/fnana.2018.00056

 

Key takeaways 

• Reevaluate understanding of layered structures in cortex. Move from viewing it as simply an arrangement of neurons and synaptic connections, to one which considers its functional anatomy. 
• Consider not only where axons terminate but also biophysical properties such as dendritic integration processes spanning multiple layers simultaneously when interpreting any given recordings or explaining behavior observed using fMRI techniques. 
• Develop models that can accurately describe how inputs between and within cortical areas are transformed into laminar specific output signals throughout the system by examining all cell types of cortex, capturing their properties, understanding how they integrate synaptic inputs etc. 
• Application: The interpretation of BOLD signals from high-res fMRI recordings, for instance, cannot be attributed simply to the spiking neuronal activity occurring in the same layers as the BOLD signal.
With the increased spatial resolution of fMRI in recent years, BOLD is now measured at different cortical depths and can therefore be used to characterize the summed energy consumption in different layers of cortex. That is, fMRI, which measures BOLD contrast, is better linked to neuronal activity by summed energy consumption than by spiking neuron output.

• 
  

Conceptual Shift:  

This perspective requires us to look beyond simple point-to-point connectivity between cells and instead consider complex interactions occurring across various compartments throughout layer structures when interpreting any given recordings or explaining behavior observed in humans subjects using high resolution fMRI techniques

What Defines a layer as a subunit of function? Need to understand cortical layering in terms of its functional anatomy. It suggests that we should consider not only where axons terminate but also take into account biophysical properties such as dendritic integration processes spanning multiple layers simultaneously, or local field potentials measuring energy related synaptic currents located directly abutting terminal sites rather than spiking neuron outputs from cell bodies.

This means looking beyond simple point-to-point connectivity between cells and instead considering complex interactions occurring across various compartments throughout layer structures when interpreting any given recordings or explaining behavior observed in humans subjects using high resolution fMRI techniques.

Figure 1 talks about the shifting perspectives with respect to Cortical Layering. Fig 1A looks at the 19th-20th century perspectives which focused on cytoarchitecture and neuronal projections. Fig 1B looks at the advances in  functional anatomy or connectivity, but this came at the expense of complexity of components, which were treated as simple point neurons. Fig 1C is a combined perspective that takes into account the broad influences of long- and short-range connectivity and the functional components that span multiple layers simultaneously. Neurons can be considered to be the functional components of the cortex. Neurons have dendritic compartments, which span across the cortical layers, meaning that they can receive inputs from neurons in different layers. Small boxes represent neurons that can still be described as point neurons, meaning that they are relatively simple and can be described with a single point. The functional components (compartmental neurons) that span multiple layers simultaneously should also be taken into account. This means that the neurons should be able to receive inputs from neurons in different layers and process them in order to produce an output.The proposed combined perspective should take into account the broad influences of long-range (Feedback, red and Feed forward, blue) and short-range (Recurrent, green) connectivity.

The Somato-Centric-Perspective 

(Box1)  in layered structures was first proposed by Cajal in 1894 and states that information flows unidirectionally across neurons in the nervous system. This has led to a reductive description of how neurons operate, which is apparent in the way neuroscientists talk about "activity" in the brain and the way the operation of neurons is formalized.

The somato-centric perspective is based on the idea that axon terminals are the "input" to a neuron, and the neuron's "output" is emitted from its "conceptual cell body". This perspective has been used for more than a century because it fits the intuition that cell bodies are both physically prominent and provide a convenient locus for recording action potentials. However, recent advances in optical methods have revealed that the output of a neuron almost always manifests as the release of transmitter at the axon terminals, while the input is best described as synaptic currents located directly abutting the terminals.

Brings us the last key takeaway around fMRI interpretation. 

With the advent of methodologies that can more precisely resolve the layering of the cortex, it becomes necessary to shift perspectives in order to interpret the signals. Because of the close apposition of input and output, the choice of label for synapses reduces quickly to semantics. It is increasingly well understood that the "function" of a neuron (i.e., the transformation from input to output) occurs via the process of dendritic integration that in the cortex frequently occurs in active and layer-spanning dendritic trees. Thus, the computation of a cortical neuron is actually a complex spatio-temporal phenomenon that transforms inputs arriving over various layers to output delivered to various other layers. In this transformation process, the cell body specifies neither the location of the processing nor the output of the neuron and could in principle be collapsed to a dimensionless node in any specific layer without substantially changing the input/output function of the neuron. It is therefore not correct from a functional perspective to attribute the "activity" of a cell to the layer in which the cell body is located. 

The suggestion is therefore that we should move away from viewing it as simply an arrangement of neurons and synaptic connections, to one which takes into account its functional anatomy - i.e., how inputs arriving at different layers affect activity within cortex. This means considering not only where axons terminate but also taking into account biophysical properties such as dendritic integration processes spanning multiple layers simultaneously, or local field potentials measuring energy related synaptic currents located directly abutting terminal sites rather than spiking neuron outputs from cell bodies . The Box 1 diagram further highlights both potential gains that can be made by informed interpretations as well as pitfalls when proceeding without such understanding; it is ideal for a case study because unlike traditional fMRI this approach calls for an appreciation of cortical operation from laminar perspective. 

Combining Components & Connectivity in Describing Cortex. 
 

The suggestion is that we should consider not only where axons terminate but also take into account biophysical properties such as dendritic integration processes spanning multiple layers simultaneously, or local field potentials measuring energy related synaptic currents located directly abutting terminal sites rather than spiking neuron outputs from cell bodies. 

This means looking beyond simple point-to-point connectivity between cells and instead considering complex interactions occurring across various compartments throughout layer structures when interpreting any given recordings or explaining behavior observed in humans subjects using high resolution fMRI techniques. The goal is to develop a model which can accurately describe how inputs between and within cortical areas are transformed into laminar specific output signals throughout the system.

Figure 2  looks at approaches for combining components with cortical layering.  It is based on a hypothesis for the possible ramifications of the associative properties of cortical pyramidal neurons with dendritic calcium spikes at the network level. This hypothesis suggests that the active properties of the apical dendrites associate feed-forward and feedback information streams arriving at different layers.  Figure 2A shows this hypothesis with blue arrows indicating feed forward information streams and red arrows indicating feedback. Figure 2B shows the missing components (gray) needed for an expanded theory of the one shown in part A. This expanded theory should include the intrinsic properties of neurons, dendrites and synaptic inputs. Feedback and feed forward axonal input are indicated with red and blue lines, respectively. Figure 2C provides an example of abstractions of neurons needed for new theories within the new perspective. Here, A = dendrites and B = Somata. 

A Case Study using Ultra-High Resolution fMRI

Ultra-High res fMRI can be used to separate different depth layers in cortex can lead to appreciation of cortical operation from laminar perspective. 

Underpinning these functional cognitive brain imaging studies, there exists a broad field advancing laminar differences in cerebral blood flow, neurovascular coupling, vascularity etc which will all benefit from interpreting layer specific data taking into account synaptic inputs distribution across layers as well as properties of neurons spanning multiple layers. The promise here lies in allowing us insight into how feed forward and feedback pathways interact during complex cognitive states like emotion or mental time travel using non invasive approaches since we are not yet able to image at cellular/subcellular level nor manipulate specific pathways optogenetically.

Figure 3 discusses use of high-res fMRI   to measure cortical depth levels in the human brain. It explains that fMRI can be used to label voxels (small 3D cubes) according to their cortical depth, and that this can be used to separate the layers of the cortex into upper, middle, and lower layers. This is done by separating feedforward and feedback processing, which are two different types of neural pathways.

The figure also says that while the bands of cortical depth measured with fMRI are still insufficient to separate all six anatomical layers of the cortex, there are important gradients that are functionally different in their processing of internal mental states. This means that, even though the bands of cortical depth measured with fMRI are not enough to separate all six layers of the cortex, they can still be used to measure important gradients in the processing of internal mental states.

Also, the fMRI data used in the example is at 0.8 mm3. very small voxel size which allows for very precise measurements of cortical depth levels in the human brain.

So what still needs to be done? 

Need to develop a model which can accurately describe how inputs between and within cortical areas are transformed into laminar specific output signals throughout the system.  

1. examine all cell types of cortex and capture their properties as well as understand how they integrate synaptic inputs (Figures 2B,C). This task should be started in vitro but eventually needs validation under awake behaving conditions. 
2. Once you get biophysical facts about components of cortex, then necessary to create abstractions so that functionality is captured by a model.
3. Combining these two pieces with connectivity information allows us interpret layer-specific data collected from brain whether electrical recordings or imaging activity at various levels/depths more effectively than before since now we have an understanding not only where axons terminate but also take into account biophysical processes such as dendritic integration spanning multiple layers simultaneously etc when interpreting any given recording or explaining behavior observed using fMRI techniques