For decades, two complementary but often siloed approaches have guided neuroscience: cellular neuroscience, which seeks to understand how individual neurons work; and systems neuroscience, which aims to uncover how networks of neurons coordinate to produce thoughts, movements and behaviors. One studies the tree; the other studies the forest.
Each approach has produced tremendous advances. For instance, cellular neuroscientists have revealed how ion channels shape the electrical language of the brain, how synapses strengthen or weaken with experience and how gene expression governs neuronal function. Meanwhile, systems neuroscientists have mapped entire circuits, recorded the activity of tens of thousands of neurons during behavior and identified patterns of activity that correlate with memory, decision-making and emotion.
But for all these advances, a question lingers: Are we actually any closer to understanding how the brain works? The jaw-dropping datasets produced by systems-level studies are seldom reconciled with biology, and the exquisite detail uncovered by cellular-level studies is rarely extrapolated from circuits to behavior. These disconnects don’t reflect failures of either approach. Rather, they reflect the vast intellectual and material resources that each requires.
Nevertheless, the brain is a multiscale organ. It is organized across multiple hierarchical levels operating in concert, not in parallel. To unravel the brain’s deepest complexities, we need to bridge cellular and systems neuroscience. Because of recent technological advances in high-density electrical probes, genetically encoded fluorescent sensors, multiphoton imaging and high-performance computing, we are better suited to do this now than ever before.
So where do we start?
Let’s return to the tree and forest metaphor. If cellular neuroscientists study trees, and systems neuroscientists study forests, what connects the two? Branches. A tree’s branches fan out to sample its environment and interact with those of neighboring trees, effectively connecting it to the forest. The leap back to the brain is a small one. A neuron’s dendritic arbor allows it to sample the activity of other neurons, both near and far. In this light, a single neuron’s dendritic arbor endows it with a systems-level view of brain activity. To understand how the brain computes, we must focus on how individual neurons reciprocally interact with their broader circuitry. Dendritic arbors provide us with an ideal opportunity to bridge systems and cellular neuroscience.
T
The input-output function at the top states that the neuron’s action potential firing is a function of its synaptic inputs from other neurons. A tad staggering considering that a single neuron has more than 10,000 inputs and that the human brain has tens of billions of neurons—but simple enough, no?
Let’s dial things up a notch. This CA1 neuron receives most of its excitatory synaptic signals from three different brain circuits: CA2, CA3 and Layer II/III of the entorhinal cortex. These circuits send different types of information to CA1, and each information stream is important and complementary to the others.
Now, this single neuron faces a considerably more complicated task: to combine three different types of information across many thousands of synapses into a train of action potentials.
But wait, there’s more. You may have picked up on the fact that those three presynaptic circuits contact different parts of the CA1 neuron’s dendritic arbor—and this alone is important. Dendrites act like any other cable or water pipe: They are resistive and therefore gradually attenuate the signals traveling through them. Inputs from the entorhinal cortex, way down at the farthest reaches of the neuron’s dendrites, often exert less influence on somatic action potential firing than do more proximal inputs from CA3. As a result, entorhinal cortical signals tend to be weaker and stretched out in time when they arrive at the soma.
OK, fine, you may think—inputs from different circuits influence a neuron’s firing more or less, depending on where they contact its dendrites. A few coefficients on those input terms and job done, right? Not quite.
Since Wilfrid Rall’s seminal work was published in the 1960s, the field has increasingly appreciated that individual dendritic branches can act like neurons within a neuron. Dendrites can fire their own version of action potentials, known as dendritic spikes. Dendritic spikes can actively regenerate as they spread through the arbor, overcoming distance to more strongly influence a neuron’s action potential output. These dendritic regions are, in fact, so distinct that they have their own names: basal, radial oblique and distal tuft dendrites.
Each dendritic compartment has its own biophysical profile that dictates how it sums up synaptic inputs across time and space, allowing it to interpret different types of information according to customary rules. Considering that this process plays out across tens of thousands of synapses, dendrites render a single neuron into a biological supercomputer.
R
For example, in the CA1 region of the hippocampus, intracellular pools of calcium shape synaptic plasticity mechanisms in radial oblique dendrites of place cells. In the retrosplenial cortex, compartment-specific synaptic specializations in pyramidal neurons align with distinct input pathways. In the auditory cortex, those pyramidal neurons display stronger activity in distal tuft dendrites after mice learn to fear specific tones. And in the motor cortex, two completely different plasticity mechanisms are at play in apical versus basal dendrites. In all of these discoveries, researchers found new ways in which specific dendritic compartments change the strength of their synaptic connections with other neurons and circuits.
By observing and probing these interactions in the brains of behaving organisms, and by feeding our experimental data to biologically realistic computational models, we may witness the next transformative moment in neuroscience. The time is now. We as a field have become incredibly interdisciplinary and, as a result, finally find ourselves with the tools to work at this level in behaving animals through collaborative efforts.
Systems neuroscientists possess the optical tools to monitor dendritic dynamics in vivo and the computational approaches to understand them. For instance, volumetric multiphoton imaging can provide simultaneous readouts of the synaptic signals arriving at dendrites and how dendrites respond, whereas detailed biophysical models provide insights into how these complex dynamics govern action potential firing. At the same time, cellular neuroscientists possess the molecular tools to shed light on dynamics within a single cell and to target, label and manipulate its various input types. For instance, we now have fluorescent reporters of nearly every neurotransmitter and numerous photoactivatable tools to control the activity of specific molecules involved in dendritic function.
Much of the brain’s complexity likely lies in dendrites. By combining systems and molecular neuroscience, we are ready to unlock it in earnest as a unified scientific field.
