Astrocyte Computing

The term is supposed to parallel 'Neural Computing', and is intended to work out how astrocytes in the brain are capable of doing computations. 

In the past, I've wondered about astrocytes in relation to hemodynamics in the brain. The standard (T2*) BOLD response in fMRI is primarily caused by local changes in the OxyHb:DeoxyHb ratio. The assumption has been that this ratio reflects local neural activity, through a coupling between neural activity and blood flow. Other studies have shown that the coupling is largely mediated through astrocytes. 

But do astrocytes actually perform any computations beyond just recruiting blood flow? Not entirely clear; but a new paper from the Mriganka Sur group shows that astrocytes in the ferret visual cortex have properties similar to neurons.

Here are stimulus-dependent fluorescence changes in neurons and astrocytes: Notice that the time scale is comparable- astrocytes are off by just a couple of seconds. Turns out that the astrocytes are also more sharply tuned (e.g., to orientation, in these studies).

And another really cool result: the authors add a glutamate transporter antagonist (TBOA). This has the effect of not clearing glutamate from the synaptic cleft. This causes (a) a reduction in the activity of the astrocytes and (b) an increase and prolongation in the activity of neurons due to extra glutamate in the cleft. But, look at what happens to the blood volume change (measured as an intrinsic optical signal at 546-nm): it decreases! 

I.e., although neuronal firing increases, blood volume change decreases. Of course, this is in the presence of TBOA, but it makes you wonder if there are physiological conditions under which the glu-transporters are downregulated (perhaps at a different timescale), leading to a decrease in blood-flow changes.

Of course, blood-flow and BOLD are themselves not super-directly related. But what these studies show is that hemodynamic measures might be actually capturing astrocyte activity, rather than neural activity.

It remains to be seen how much of the computation in the brain is attributable to the astrocytes. And also what 'neural network' models might really mean ;)

The Mystery of the BOLD Signal

I've been pretty much bugged about trying to understand the physiological basis of the BOLD signal in fMRI. In a previous post, I sketched an initial attempt to undertsand the vascular changes, mainly in relation to Optical Topography. But then, I realise I don't understand the BOLD response very well, after all...
So here's an update.

MR Basics¹

• Inside the magnet, anything with a magnetic moment will align itself with the big stationary magnetic field (B₀), along the z-axis, and all spins will be in phase.
  
• A radio-frequency pulse makes a second magnetic axis (e.g.) perpendicular to the main magnet axis, and several molecules/atoms use this as an excuse to spin away from the main axis and into the transverse (x-y) axis.
  
• When the RF pulse is turned off, {a} the spins start coming back to their equilibrum position along B₀ with a time constant T1, and {b} in the absence of an external driving force, the spins in the x-y plane lose coherence (and so lose a net magnetic moment) with a time constant T2*. (There's also another decay time constant in this transverse field, the T2, which is the decay constant during repeatedly refocusing the phase of the transverse spins).
  
• The presence of a paramagnetic substance like deoxy-Hemoglobin (HbR) creates magnetic inhomogeneities, such that the going-out-of-phase in the x-y plane happens much quicker. So, a much smaller signal is produced. Thus, the more the HbR, the smaller the signal. In contrast (heh!) HbO is diamagnetic, and for all magnetic purposes behaves like water, so doesn't contribute much to the (T2*-dependent) BOLD signal.

So far so good.

The basic reason why one would use the BOLD signal is because of certain findings that indicated to Messrs. Roy and Sherrington (1980):

...the existence of an automatic mechanism by which the blood supply of any part of the cerebral tissue is varied in accordance with the activity of the chemical changes which underlie the functional action of that part. Bearing in mind that strong evidence exists of localisation of function in the brain, we are of opinion that an automatic mechanism, of the kind just referred to, is well fitted to provide for a local variation of the blood supply in accordance with local variations of the functional activity.

In short, a neurovascular coupling.
Naively, one would expect the following sequence of events²: a stimulus comes in, the relevant neurons hike up their activity, they require more energy, and so start burning more glucose and turning more HbO into HbR. Therefore, there should be an increase in HbR in the area close to the activity. This means that the BOLD signal should be smaller following stimulation.

Now, other things seem to happen when the stimulus arrives: an increase in the blood-flow rate, and an increase in the blood volume. Again, both of these increase the amount of HbR, and so should decrease the BOLD signal.

BUT, as it happens, over and over again (something like 90% of the studies, according to one source), one finds an increase in the BOLD signal!

Actually, that's not strictly true. There is evidence for an initial "dip" in the BOLD signal, as one would expect³. One possible sequence reconstructed by Malonek et al goes something like this:

• sensory stimulus
• blood flow increase
• as soon as blood flow increases, HbO increases
• simultaneously with blood flow change, HbR starts to decrease
  

But then what causes the much larger BOLD signal? Simultaneous recordings of spikes, local field potentials, blood oxymetry and so on favour a view in which the BOLD somehow reflects

"...population synaptic activity (including inhibitory and excitatory activity), with a secondary and potentially more variable connection with cellular action potentials."

In short, not the spikes, but the more widespread synaptic signalling, whether or not it leads to spiking⁴.

So, one might wonder, how does the big BOLD signal come about?? Clearly, there is much less HbR per unit volume post-stimulation that there was before. How can THAT be? One possibility is that, somehow, the incoming blood is (a) plentiful = large volumes and (b) progressively enriched in HbO.

So, larger quantities of greater HbO-containing blood can both satisfy both the observations - that volume increases and that HbR goes down.

Luckily (for me, at any rate), last year there was a paper in Nature Neuroscience: "Astrocyte-mediated control of cerebral blood flow" by Takano et al. The abstract reads:

Local increase in blood flow during neural activity forms the basis for functional brain imaging, but its mechanism remains poorly defined. Here we show that cortical astrocytes in vivo possess a powerful mechanism for rapid vasodilation. We imaged the activity of astrocytes labeled with the calcium (Ca2+)-sensitive indicator rhod-2 in somatosensory cortex of adult mice. Photolysis of caged Ca2+ in astrocytic endfeet ensheathing the vessel wall was associated with an 18% increase in arterial cross-section area that corresponded to a 37% increase in blood flow. Vasodilation occurred with a latency of only 1–2 s, and both indomethacin and the cyclooxygenase-1 inhibitor SC-560 blocked the photolysis-induced hyperemia. These observations implicate astrocytes in the control of local microcirculation and suggest that one of their physiological roles is to mediate vasodilation in response to increased neural activity.

The crème de la crème is the finding that the astrocytes selectively open up the arterial flow! Here's the graph from their paper, showing the relative changes for arteries, veins and capillaries.








This explains both the increased oxygenation of the blood (and so the lower [HbR]) and the larger volume. Add to this the finding from last year that pericytes - cells sitting around the capillaries - can bidirectionally squeeze capillaries, and you might also understand why local blood velocity goes up.

---

References

1. For an excellent online source, look at http://www.e-radiography.net/mrict/mri%20ct.htm.
2. See this previous post.
3. Malonek, D. & Grinvald, A. (1996) Science 272, 551–554
4. Arthurs, OJ & Boniface, S. (2002) TINS 25(1), 27-31
   

The demise of neural nets?

I don't really like neural nets. Rather, I don't like neural nets when it comes to the human mind. The only reason that neural nets seem to exist is because someone saw their potential for computation. Of course, that potential itself has not been fuly realized; most "neurons" in typical neural nets are little points with rather simple I/O functions. Of course, most neural netters will tell you that the brain is full of neurons. The whole central nervous system is full of neurons. But really, there are ten times fewer neurons than there are glial cells, and they too are busy, nourishing, maintaining, organizing neurons and even participating in neural transmissions. And a recent paper in Trends in Neuroscience now says, "Astrocytic complexity distinguishes the human brain". I'm just waiting for the discovery of the role of astrocytes in setting up neural circuits, and participating in neural function! After all, the radial glia are busy directing neurons to the cortex.. how much role do they play in setting up the initial cortical specificities? Could be LOTS! And hopefully we'll be rid of the "neural nets do everything" culture. I just hope we won't usher in an "astrocytes do everything" era...