Glial Biology in Learning and Cognition

Connecting All the Parts: Glia in Spatial Integration

All types of glia can respond to and influence neurotransmission in several ways. The three major glial cell types in the brain, astrocytes, oligodendrocytes and microglia, communicate with each other and with neurons by using neurotransmitters, other small molecules, and gap junctions. Oligodendrocytes greatly increase the speed of electrical transmission through nerve axons by forming the axonal myelin sheath and clustering ion channels at nodes of Ranvier where action potentials are generated (Nave 2010). Microglia prune synapses in part by monitoring synaptic transmission (Wake and others 2013), thus they rewire neural connections in accordance with neuronal function (Schafer and others 2012). Astrocytes can regulate synaptic transmission between neurons by modifying the concentration of extracellular potassium, controlling local blood flow, releasing and taking up neurotransmitters and other neuromodulatory substances, delivering nutrients to neurons, and altering the geometry and volume of extracellular space between brain cells; all of these could coordinately couple neurons into functional assemblages.

The large spatial domain over which astrocytes and oligodendrocytes can couple and regulate neuronal activity is at odds with traditional thinking of learning where synapse-specific modification is paramount. Restricting plasticity mechanisms to individual synaptic connections is thought to increase the capacity for information storage and learning, but global changes may help organize the memory, and in a combinatorial manner increase storage capacity. If astrocytes modulate synaptic strength and couple domains of synapses into functional assemblies, the degrees of freedom for information storage are multiplied beyond synapse-specific coding (Fig. 1). Astrocytes do have anatomical and physiological properties that may impose a higher order organization on information processing in the brain. The volume of human astrocytes is almost 20-fold larger than their rodent counterparts, enabling human astrocytes to integrate input from a comparably staggering number of synapses (~2M compared with 0.1M in rodent brain) (Oberheim and others 2006) (Fig. 2). Curiously in the gray matter regions of cerebral cortex and hippocampus, astrocytes are organized in non-overlapping domains (Bushong and others 2002); the significance of this tile-like organization is completely unknown.

Similar large-scale changes, as opposed to point-to-point changes in neuronal synaptic connections, may also participate in other aspects of information processing and cognition that are not currently understood from a strictly neuronal perspective. For example, when we see an object against a background, our visual system assigns the inside and outside of the bounding edges separating the figure from the ground (Fig. 3). This is a difficult computational problem in computer vision because it involves global information that cannot be decided locally (Zucker 2012). How does the brain do this? Although the importance of astrocytic domains is far from being fully understood, astrocytes in a receptor-mediated pathway can modulate extracellular K⁺ and regulate neuronal spiking activity (Wang and others 2012). A single astrocyte, acting via purines, produce an UP state (persistent depolarization resulting in high-frequency firing) in a sizable population of neighboring neurons (Poskanzer and Yuste 2011). Effects of this kind working in conjunction with synaptic plasticity could create neuronal clusters that become associated with particular input signals. Computer simulations of molecular diffusion in the neuropil show that astrocytes are highly effective in limiting the spread of glutamate released at synapses. This same approach could be used to explore the modulation of synaptic activity by glial regulation of extracellular K⁺ (Kinney and others 2013).

Glia in Learning and Memory

Glia can affect both the encoding and consolidation of memory. Activation of intracellular Ca²⁺ signaling in astrocytes can augment or suppress both excitatory and inhibitory synaptic transmission and influence state dependent changes in cortical activity during, for example, sleep (Halassa and others 2009) and working memory. Recent evidence obtained in situ and in vivo has shown that astrocytic signaling participates in some forms of synaptic plasticity (i.e., cholinergic-induced long-term potentiation (LTP) in cortex and hippocampus (Navarrete and others 2012); spike-timing dependent cortical plasticity, hippocampal long-term depression (LTD) and working memory (Han and others 2012)). Transient disruption of the flow of energy substrates from astrocytes to neurons has little effect on learning but severely impairs the later formation of long term memory (Suzuki and others 2011). Astrocytes associated with blood vessels shuttle ions and energy substrates between neurons and the blood stream and regulate local blood flow, which in turn modulates neuronal function (Moore and Cao 2008). Moreover, microglia, oligodendrocytes, and astroglia are primary sources for extracellular matrix components now thought to play critical roles in late stages of LTP and memory consolidation (Babayan and others 2012). Recently, humanized chimeric mice, in which a large proportion of mouse astrocytes were replaced with human astrocytes, were found to exhibit increased synaptic plasticity and faster learning (Han and others 2013).

The Importance of Timing

The temporal coincidence in firing of multiple inputs converging onto a neuron is fundamental to most rules of learning and synaptic plasticity. Thus the physiological mechanisms underlying learning will be strongly influenced by conduction delays. Similarly, oscillations and waves of neuronal activity link neuronal firing in phase-locked mode with other neurons and such oscillatory activity during sleep is important in memory consolidation (Buzsáki 2006). Astrocytes influence neuronal membrane potential and excitatory and inhibitory synaptic transmission underlying neural oscillations locally, and oligodendrocytes provide rapid impulse propagation for long-range oscillations and synchrony of spike time arrival between distant populations of neurons.

Myelination of axons by oligodendrocytes (Fig. 4) enables saltatory conduction between nodes of Ranvier, which greatly increases the speed of impulse propagation. Accumulating evidence indicates several forms of activity-dependent signaling between axons and myelinating glia that control myelination of specific axons (Wake and others 2011; Fields 2013), suggesting a hitherto unexplored mechanism of experience-dependent plasticity and learning (Fields 2008). The consequences of altering conduction velocity by myelin plasticity has been modeled recently and the results indicate profound effects on neural network function would result in terms of spike time arrival, oscillation frequency, phase coupling, and brain wave propagation (Pajevic and others 2013). Recent data from human brain imaging by MRI, notably using diffusion tensor imaging, has revealed structural changes in myelinated tracts after learning a wide range of tasks (Zatorre and others 2012). The cellular basis of these changes in white matter is being investigated, but present evidence supports some contribution by glia (astrocytes and myelin).

Adult Neurogenesis

Adult neurogenesis is influenced by environmental experience (Glasper and others 2012), and contributes to memory formation (Sahay and others 2011), but adult gliogenesis has received much less attention. Astrocytes produce growth factors that modulate neural stem cell differentiation (Barkho and others 2006), and thus regulate neurogenesis. Oligodendrocyte progenitor cells (NG2 cells) are the largest population of dividing cells in the adult brain, suggesting a possible role in myelination in the adult brain (Young and others 2013) that could contribute to optimal information processing.

Future Challenges

Current knowledge about astrocytes, oligodendrocytes, and microglia and their dynamic changes is rudimentary in comparison to neurons, and little effort has been made to include glia into realistic computational modeling. Computational models that consider activity-dependent changes in the timing of intercellular communication (such as provided by myelination) need to be developed and incorporated together with the broad spatial and temporal dynamic interactions between networks of astrocytes and neurons. Because neuroscience developed from a focus on electrical signaling in neurons, both conceptual and technical advances are necessary to understand the function of glia in learning and cognition. A potential route toward accelerating progress in this field emerged from the meeting: rather than relying on indirect measurements performed in neurons to ascertain glial function, cell biological approaches to directly assess the chemical signals of glia will likely benefit research.

Viewing the cellular basis of brain function as bicellular rather than monocellular may bring a new understanding of brain function, broadening without undermining current understanding. In many respects glia are far more complex than neurons and difficult to study. Adult neurons do not divide, differentiate, migrate, or engage in such a wide range of multitasking roles as glia. Microglial involvement in activity-dependent synaptic pruning (Tremblay and others 2010) illustrates this difficulty, because their immune system function provokes numerous cellular changes that engage these cells in circuit remodeling by releasing cytokines, reactive oxygen species, growth factors, enzymes and other substances when the brain is exposed for an experiment (Fig. 5). Such multitasking responses in glia can perturb the processes being measured. It is this complexity of glial biology, however, that likely contributes much of the intriguing complexity evident in learning and cognition in the human brain.