Saturday, December 24, 2011

Brains a-building

Just how does the self-constructing computer self-construct?

The brain is probably the most exciting and complex frontier of biology. How does it work? How does the mind happen in amongst those 100 billion neurons? However, before we get to all those questions, the brain has to develop, all by itself in the fetal and infant body, from the most minimal ingredients and from an extremely spare blueprint comprising some fraction of the 25,000 genes of our genome.

One is tempted to call it a miracle, except that it happens all the time, all over the world, more or less dependably and following, as far as our incomplete knowlege reaches, no script but those of the physical / biological world. A paper came out recently highlighting the varying role of gamma oscillations in brain development, which seemed worth reviewing.

Many people may not be aware that brain development involves vast migrations of cells, from one place to another in the developing brain. "So, most neurons migrate from the site of their last mitotic division, near the ventricle, towards the outer surface of the CNS, where they integrate into specific brain circuits." (From a very nice review of the field.) Doubtless this inefficiency reflects a long evolutionary history, as do numerous other weird anatomical paradoxes of the body. An idea that didn't seem so bad in the infinitesimal brain of a gnat becomes a bizarre marathon of long-distance cell treks in our own. Which is a little ironic, because after they get to their final position, these neurons spend the rest of their lives in one position, with their plasticity confined to forming or deleting synapses among their far-flung axonal and dendritic branches.
"Vertebrates show far more widespread neural migrations than previously realized. In general, these migrations can be seen as DV [dorsal-ventral] or AP [anterior-posterior] migrations, pathways thought to be prominent in lower organisms but not in vertebrates. Indeed, genes discovered in C. elegans and Drosophila provide molecular mechanisms for the DV and AP migrations in higher vertebrates."
Schematic of a few major pathways of cell migration in early brain development. Neurons in the cortex all come from stem areas near the core of the brain. MGE is the medial ganglionic eminence.

Indeed, inhibitory interneurons and excitatory neurons are distinct cell classes, and orginate from different stem locations and migrate by separate pathways, but migrate into close proximity to make up the final brain network.

Later on during development, the micro-architecture of the cortical layer (the side-by-side columns of cells with related functions, occurring all over the sheet-like cortex) refines itself through feedback from connected areas. For instance, the columnar arrangement of our visual cortex maps strikingly to our visual fields and other salient properties of vision- a mapping which forms soon after the eyes first open (the "critical period").

The current researchers looked at whisker sensing areas in the rat brain, which organize themselves similarly as the visual system, into columns of discrete function. "In the rodent 'barrel' cortex, each cortical barrel column receives a specific input, conveyed via the thalamus, from a corresponding whisker." In the critical period for this region, (days 2-7 after birth), this area doesn't communicate much with other areas of the brain, but only with its whisker inputs and perhaps with local neighbors.

These columns are further divisible vertically into layers that extend over most of the cortex. The cells of these layers are somewhat distinct at each level, and closely connected to each other up and down the column, while inputs and outputs to other columns and other brain regions are typically differentiated by layer. Which is to say that inputs to the column from one area of the brain may typically come to a subset of layers, while outputs to some other brain region may emerge from another subset.

The key topic is the "early gamma oscillation", or EGO. Gamma waves are famous as the highest-frequency brain waves, which are the leading candidate for "binding" mental contents over long distances across the mature brain. The interesting finding here is that, unexpectedly, in early development, gamma oscillations happen but seem to have a quite different and simpler function- that of binding a developing neural zone to its sensory inputs, and thus helping it self-organize.

Layers:SG- supragranular, G- granular, IG- infragranular, Pia- the pia matter, or innermost membrane surrounding the brain. LFP- localized field potential, MUA- multiunit activity (individual spikes), CSD- current-source density (overall conductance/resistance).
In this figure, a needle electrode is shown as it is stuck vertically into one column of a rat brain, with cell bodies stained in green and electrode points shown as dots along the electrode depth. The graph shows the associated recording, with the rat's whisker touched by the experimenter at time 0. The 50 Hz gamma oscillation is obvious over several layers. The researchers claim that these gamma oscillations had been missed previously because typical surface recordings wouldn't catch them.

What are they doing? Normal gamma oscillations coordinate large regions of brain activity, but these have only localized coherence- with the whisker input. The next figures compare gamma intensity from different stimuli and at different ages:

PW means principle whisker (the one directly innervating the probed column), while AW means adjacent whisker, which innervates nearby columns. LFP is "local field potential", i.e. the Y-axis, while GR is granular layer, the middle one shown above, and SG means supra-granular layer, the one over it, which the researchers find ties up to nearby columns during development. The next figure shows graphic summaries of the same data by electrode depth, at postnatal day 5 (P5) and day 33 (P33). The sequence is clear- that low-level, exclusively local and whisker input-driven gamma oscillations at early times are followed by more powerful oscillations located in higher cortical layers and driven not only by the innervating whisker, but by nearby ones to a high degree. One can truly see the knitting together of neural networks over time.

The experimenters also probed the thalamus, which conducts the signals from the whiskers to the cortex, showing that the signal timing is appropriate. The gamma peaks in the thalamus ("VPM") lead those in the developing cortex by about nine milliseconds.

Indeed, if they remove brains entirely and probe slices that retain the thalamus-cortical connection (shown below, H), they can simulate whisker stimulation by electrical stimulation on the thalamic area (VPM) which creates artificial early gamma oscillations (aEGO's) in the cortical region. If these are carefully timed to sync with endogenous cortical neural firing, they strengthen their neural connections, which can be assayed by downstream currents out of the cortical layer (excitatory postsynaptic currents; EPSCs- the red vs the black graphs below). By this method of artificial "learning", evoked EPSCs are significantly stronger after a bout of thirty induced aEGOs than they were before, using low levels of aEGO stimulation to test with.

OK- that was seriously technical. But the lesson is that by enough poking, prodding, and taking things apart, we are beginning, in baby steps, to understand that most intricate and delicate mechanism- the wetware of our minds.

"What is to be done? That demands a huge agenda. It must cover employment, education, corporate governance and financial reform and, however difficult, also elements of redistribution. It will be unavoidably divisive. So be it. This debate cannot be avoided if western democracies are to stay legitimate in the eyes of their peoples. That may not be true in the US. It is surely true in the UK. Warren Buffett has argued that 'there’s been class warfare going on for the last 20 years and my class has won.' The remark has not made him popular with his peers. But he was surely right."

No comments:

Post a Comment