Changes in myelinated axons following juggling in diffusion tensor imaging

Scholz et al performed diffusion tensor imaging on 48 adults randomly placed in either a juggling or control group. By the end of the 6-week training each of the adults in the juggling group could perform 2 cycles of the 3 ball cascade, which is somewhat but not overly impressive. As compared to their pre-scanning fractional anisotropy, a somewhat loose measure of myelination, fiber density, and axon diameter, the juggling group had a percent increase of ~ 5.5 +/- 1.5 % immediately following the training, and a percent increase of ~ 4 +/- 1 % four weeks later. The control group had no real increase following training, which makes sense because they didn’t do anything!

The increase four weeks post-training indicates that although the effects of the training diminish somewhat over time, they should last for at least a little while. Perhaps further studies could continue to perform diffusion tensor imaging on the adults to see when the percent increases due to training are extinguished, if ever. A back of the envelope calculation based on a linear trend would suggest that after ~ 15 weeks following training the increased would be gone. But the reality may be wildly different.

Reference

Scholz J et al. 2009 Training induces changes in white-matter architecture. Nature Neuroscience 12 1370-1371. doi:10.1038/nn.2412

Spending our time in worlds that don’t exist

Paul Bloom has a fascinating article in the most recent issue of The Atlantic that touches on libertarian paternalism, behavioral economics, pigeons, Walt Whitman, and gambling addiction. One of the most interesting sections is where he discusses how much time we spend in existences we know to be not real,

… The most common leisure activity is not sex, eating, drinking, drug use, socializing, sports, or being with the ones we love. It is, by a long shot, participating in experiences we know are not real—reading novels, watching movies and TV, daydreaming, and so forth. Enjoying fiction requires a shift in selfhood. You give up your own identity and try on the identities of other people, adopting their perspectives so as to share their experiences.

There are currently more ways to live in unreal worlds than at any other point in history. This doesn’t mean that modern man necessarily spends more time unreal worlds than in previous eras. Indeed, it’d be hard to measure how much people in the 1700s daydreamed without a time machine that doubles as a giant fMRI. But I’d expect that yes, our society currently spends more time in unreal worlds than previous generations.

Should we as a society attempt to reverse this trend? That’s where the libertarian paternalism comes into play. How can we do so without invading people’s private lives? That’s where the behavioral economics comes into play. What a curious world we live in.

Diffusion tensor imaging to reconstruct average brain connectivity

In anisotropic tissue something interferes with the free diffusion of water molecules, for instance cell membranes or microtubles. This means that diffusion will be faster parallel to an axon and slower perpendicular to it. In DTI, the diffusion coefficient will miss these local effects and thus the diffusion coefficient will vary depending upon the orientation in which the tissue is measured. By measuring a given area of tissue (i.e., a voxel) from 6+ directions, you can describe the orientation of average axons in a vector. Following some fancy math, you can determine white matter pathways between voxels as well as connectivity probabilities.

Gong et al recently used diffusion tensor imaging on the whole brains of 80 right-handed young adults in 3-mm slice thickness (no gaps) for 40 overall slices from 6 diffusion directions with a b value of 1000 s/mm^2. Note higher b values lead to greater image contrast. They then interpolated the diffusion-weighted images into 1-mm isotropic dimensions. They partitioned the cerebral cortex into 78 cortical regions and restricted the trajectory of fiber bundles to white matter voxels to evaluate their connectivity to the adjacent cortical region. They then counted number of fiber bundles connecting each pair of regions and focused on the connections consistent across their subjects, to account for the variability in brain anatomy between individuals.

The researchers found 329 statistically significant anatomical connections between cortical regions out of 3003 potential between-region connections, a “sparsity” of 11%. They also identified the nodes and edges in their network that have betweeness values 1 SD above the mean. Betweeness is a measure of how often that vertex occurs on the shortest path between other vertices, and its relative importance to the network. Kind of cool because they can evaluate whether those vertices have also been shown to be important in previous non-DTI studies.

DTI will be a big part of the forthcoming human connectome project of the NIH. Resolution on the individual neuron scale is considered unrealistic by many, as Gong et al noted in their paper, and DTI is a viable alternative. Next steps would be connecting structure more with function, determining changes to the anatomy as a result of neurodegenerative diseases, and fixing methodological snags. DTI in the brain is poised to be very useful in the coming decades.

Reference

Gong G, et al. 2009 Mapping anatomical connectivity patterns of human cerebral cortex using in vivo diffusion tensor imaging tractography. Cerebral Cortex 19:524-536.

Stimulating Kolmer–Agduhr cells elicits swimming in zebrafish

Wyart et al expressed the light-gated ion channel LiGluR as well as the UAS promoter in some zebrafish and crossed these with zebrafish containing the GAL4 transcription factor in various classes of ventral spinal neurons. When the progeny was photo-stimulated in the caudal spinal cord, 94% displayed tail oscillations. The effects were localized down to Kolmer-Agduhr cells and since its action potentials were blocked by the GABA-A antagonist bicuculline, the researchers were able to identify the mechanism as GABAergic, of note because GABA is typically inhibitory. This is a good example of how a given behavior can be elicited in a non-intuitive way by a specific subset of neurons in a particular region of the nervous system.

Reference

Wyart C, et al. Optogenetic dissection of a behavioural module in the vertebrate spinal cord. Nature 461:407-410. doi:10.1038/nature08323.

Neuron classification is essential for circuit models

One assumption of cortical circuitry is that the strength of interactions between neurons and layers should be directly correlated to the amount of overlap between axons and dendrites of pyramidal cells. However, this assumption is cracking under evidence that the cell’s class (as determined by morphology, postsynaptic target, and protein complement) also helps to determine the strength of the synaptic connection between any two neurons. For example, Petreanu et al mapped the dendrite locations of channelrhodopsin-2 expressing pyramidal neurons in the neocortex to determine the axon-dendrite overlap between its various inputs. But when the researchers quantified the strength of inputs in a given column with identical laser powers, one layer of barrel cortex (L5B cells) had 62-fold less input from the posterior medial nucleus than another group (L5A cells), despite the fact that L5B dendrites had more overlap with posterior medial nucleus axons than L5A dendrites. So, the functional connectivity between neurons cannot simply be deduced from the structure and overlap of the axons and dendrites. The actual class and region of the two given neurons has predictive power for synaptic strength as well.

References

Brown SP, Hestrin S. 2009 Cell-type identity: a key to unlocking the function of neocortical circuits. Current Opinion in Neurobiology 19:415-421. doi:10.1016/j.conb.2009.07.011

Petreanu L, et al. 2009 The subcellular organization of neocortical excitatory connections. Nature 457:1142-1145. doi:10.1038/nature07709.

Sleep and Consciousness: Anatomical Help Wanted

A mathematician wakes up and smells smoke. He goes to the hall and sees both a fire and a fire hose. He thinks for a moment and exclaims, “Ah, a solution exists!” and then goes back to bed.

Many researchers studying consciousness are similarly content to theorize over the mere feasibility of explaining phenomenal experiences, instead of examining the anatomical substrates that will allow us to answer the more intermediary questions. As an exercise in demonstrating the utility of these intermediary steps for our theories, this paper will examine how human neural circuit diagrams could immediately improve our understanding of the relationship between various components of sleep and consciousness.

One of the most prominent theories of the function of sleep is metabolic restoration. Benington and Heller (1995) propose that glycogen stores in astrocytes are exhausted during waking and replenished during sleep, and that the need for sleep is dictated by high levels of adenosine, which promotes neural synchronization. The authors advocate that astrocytes are necessary for preserving the proper cellular environment for neurons during and between action potentials. So, in their model the levels of glycogen in astrocytes indirectly affect neuronal responsiveness. But more recent research suggests that astrocytes may be directly involved in synaptic networks via the calcium-dependent release of glutamate to neurons (Fiacco et al, 2009). Additionally, astrocytes have been found to release ATP, which is quickly hydrolyzed to adenosine, and then causes a persistent synaptic suppression at neurons with adenosine-1 receptors (Pascual et al, 2005). Moreover, when mice without adenosine-1 receptors are sleep deprived, they do not display any change in synchronized slow-wave activity, even though synchronized slow wave activity does vary directly with previous waking duration in wildtype mice (Bjorness et al, 2009). This implies a mechanism through which adenosine could regulate sleep homeostasis. Overall, Benington and Heller’s original proposal has been falsified in some respects and confirmed in others since its publication, as a result of anatomical data. Given a working circuit diagram in the cerebral cortex and/or thalamus of either mice or humans, the metabolic theory could be further refined and potentially falsified.

Another major theory for the function of sleep is that sleep allows for memory consolidation. Sejnowski and Destexhe (2000) postulate that bursts of synchronized high-frequency action potentials in thalamic neurons depolarize dendrites but not the soma, allowing calcium ions to enter dendrites and affect gene expression in the nucleus. This allows for plasticity but comes at a cost, as synchronized networks can feedforward into epilepsy if not properly regulated. This inhibitory regulation is difficult during the day because we are constantly in the presence of sensory stimuli. So, since sleep allows synchronized high frequency action potentials without the risk of uncontrolled positive feedback, sleep may be the toll we pay for plasticity. The feasibility of their argument is based in large part on a reconstruction of synaptic connections of individual neurons in the thalamus. But in order to simulate the proper levels of firing in their model they had to assume that specific levels of GABAA-mediated and excitatory inputs would be afferent to their model neurons. This trial-by-error technique ensures that their computational results make sense but it risks biological irrelevance. Given more anatomical data, the researchers could have ensured the use of biologically plausible synaptic inputs.

The study of other sleep exotica could also benefit from additional anatomical data. Blackmore’s chapter on sleep (2003) touches on the evolution of dreams, a question that would be easier to answer if we knew the differences between our physiological substrates for dreams and those of other animals. Leberge’s (1990) article on lucid dreaming includes the speculation that seritonergic neurons form a system that normally inhibits hallucinations but is itself inhibited in REM sleep, a hypothesis that requires anatomical data to answer. Iranzo et al (2009) are able to note that a dopaminergic deficiency is likely not the cause of REM sleep behavior disorder, but are unable to conclude which brainstem deficiencies do cause the disorder. Although lesion and knock-out studies are good at falsifying hypotheses, they are unable to suggest any on their own in the way that a large neural circuit data corpus would be able to. Neural anatomical data might one day explain Cheyne and Girard’s (2007) observation that vestibular motor sleep paralysis experiences are associated with bliss while intruder and incubus experiences are not, beyond the presumption that the amygdala is somehow involved. And although Johns (1991) is correct in asserting that self-reports are currently our only gateway into subjective experience, this may not always be the case. Miyawaki et al (2008) were able to use fMRI to determine what subjects were currently viewing, based on training data where subjects saw random images. As anatomical data interfaces with new imaging techniques, the possibilities are tremendous.

If we want to answer the hard question (i.e., explaining phenomenal experiences) we will be best served iterating towards technical solutions of the easy problems. The ultimate solution to this technical problem is the creation of a “connectome,” perhaps initially via diffusion imaging but eventually via serial section transmission electron microscopy. Now, it is quite possible that the individual neuron level will not contain enough information to recover the computational detail necessary for consciousness. Indeed, it is conceivable that the density of NMDA receptors on a given pyramidal neuron in the hippocampus could correlate to the information content of a given memory. In that case, we would need to image and reconstruct the circuit at a resolution level that could capture membrane protein receptors. Although less likely, it is also conceivable that small cellular molecules such as microRNAs, which are known to be asymmetrically distributed throughout the brain (Olsen et al, 2009), could play a particular role in producing consciousness. In that case an effort to computationally simulate a conscious mind would be futile in the near future. But we cannot know the answers to these questions until we try. Although it may not be a sexy answer, the truth is that we need more anatomical data before we can intelligently discuss the hard question.

References

Blackmore SJ. 2003 Consciousness: An Introduction. Oxford University Press, USA, pp. 338-352.

Pascual O, Casper KB, Kubera C, Zhang J, Revilla-Sanchez R, Sul JY, Takano H, Moss SJ, McCarthy K,2 Haydon PG. 2005 Astrocytic purinergic signaling coordinates synaptic networks. Science 310:113-116.

Olsen L, Klausen M, Helboe L, Nielsen FC, Werge T. 2009 MicroRNAs show mutually exclusive expression patterns in the brain of adult male rats. PLoS ONE 4:e7225.

Miyawaki Y, Uchida H, Yamashita O, Sato MA, Morito Y, Tanabe HC, Sadato N, Kamitani Y. 2008 Visual image reconstruction from human brain activity using a combination of multiscale local image decoders. Neuron 60:915-929.

Fiacco TA, Agulhon C, McCarthy KD. 2009 Sorting out astrocyte physiology from pharmacology. Annual Review of Pharmacology and Toxicology 49:157-171.

Bjorness TE, Kelly CL, Gao T, Poffenberger V, Greene RW. 2009 Control and function of the homeostatic sleep response by adenosine A1 receptors. Journal of Neuroscience 29:1267-1276.

Johns MW. 1991 A new method for measuring daytime sleepiness: the epworth sleepiness scale. Sleep 14:540-545.

Cheyne JA, Girard TA. 2007 Paranoid delusions and threatening hallucinations: a prospective study of sleep paralysis experiences. Consciousness and Cognition 16:959-974.

Iranzo A, Santamaria J, Tolosa E. 2009 The clinical and pathophysiological relevance of REM sleep behavior disorder in neurodegenerative diseases. Sleep Medicine Review xxx:1-17.

LaBerge S. 1990 Lucid dreaming: psychophysiological studies of consciousness during REM sleep. In Sleep and Cognition, pp. 109-126.

Benington JH, Heller HC. 1995 Restoration of brain energy metabolism as the function of sleep. Progress in Neurobiology 45:347-360.

Sejnowsku TJ, Destexhe A. 2000 Why do we sleep? Brain Research 886:208-223.

Sharp-wave ripples in slow wave sleep promote memory consolidation

Hypothesis (first porposed by Buzsaki, 1989): Memory consolidation has two steps. In the initial encoding, the hippocampus integrates spatial info afferent from the neocortex, favored by “exploratory” theta rhythms (~ 8 Hz). Next, during consolidation, recurrent excitation transfers the info back to the neocortex, resulting in hippocampus-independent long-term memory. The consolidation phase relies on sharp wave-ripple events (i.e., population bursts) in CA1 during slow wave sleep.

Evidence: First, enhancing the quality of slow wave sleep improves performance on hippocampus-dependent tasks after at least one night of sleep since learning. Also, after extensive learning, sharp-wave ripple events during sleep increase. And now, Ego-Stengel and Wilson have shown that disrupting sharp-wave ripple events disrupts learning. Rats were alternatively trained on two four-armed radial mazes, one of which was always followed by microstimulation of CA1 during sharp-wave ripple bursts during sleep, and one of which was a control learning maze. The number of errors per trial was significantly higher (p < 0.03, n = 12) for the CA1 microstimulation maze than for the control maze. This indicates that spatial learning is impaired when rats have fewer sharp-wave ripple bursts in the hippocampus during sleep. The learning curve for the microstimulation maze wasn’t completely flat, but it was right shifted as compared to the control curve. One interesting facet of their experiment is that most of the stimulations were during non-REM sleep (84%) or while the rats were awake (15%), and less than 1% were during REM sleep. Although REM sleep is often associated with “mental” rest, spatial memories at least do not appear to be dependent upon neural activity during REM.

References

Buzsaki G. 1989 Two-stage model of memory trace formation: A role for ‘‘noisy’’ brain states. Neuroscience 31:551–570.

Ego-Stengel V, Wilson MA. 2009 Disruption of ripple-associated hippocampal activity during rest. Hippocampus, Early View. DOI 10.1002/hipo.20707

Decline in hippocampal plasticity mediated by deficits in D-serine

Most naturally occuring amino acids in animals are of the L stereoisomerism, but D-serine is an amino acid that does have biological activity. It is known to activate NMDA receptors and induce NDMA receptor-dependent synaptic plasticity. And, there is evidence that deficiencies in D-serine are involved in the decline in hippocampus-dependent memory that occurs during aging.

Serine racemase is the enzyme that converts the naturally occuring L-serine to D-serine. Turpin et al looked at the mRNA and protein levels of D-serine in young and old Wistar rats as well as young and old Lou/C/Jall rats, which represent a model of aging without memory decline. D-serine levels were significantly reduced only in the hippocampus of aged Wistar rats as compared to young ones, −47.8% for mRNA levels and −25.1% for protein levels. When the researchers induced isolated NMDA receptor based field excitatory postsynaptic potentials on transverse hippocampal slices in Wistar rats, the recording was weaker in old animals than young ones. This difference between old and young was not apparent in the recordings from Lou/C/Jall rats. Crucially, when exogenous D-serine was added to the cerebrospinal fluid of Wistar rats, the age-related decrease in isolated NMDA receptor mediated synaptic potentials was rescued and there were no longer any signifcant difference between young and old rats. This strongly suggests that diminished D-serine can be responsible for lowered activity by NMDA receptors in the hippocampus.

Interestingly, the authors note that Lou/C/Jall rats have a reduced oxidative metabolism and less ROS production as compared to other strains (i.e., Wistar), don’t show any age-dependent reductions in the expression of serine racemase, and are generally a model for healthy aging without cognitive decline. Thus, the serine racemase gene may be a common and/or prototypical target of DNA-based oxidative damage in the aging brain.

Reference

Turpin FR, et al. 2009 Reduced serine racemase expression contributes to age-related deficits in hippocampal cognitive function. Neurobiology of Aging, Article in Press. doi:10.1016/j.neurobiolaging.2009.09.001.

Asymmetric distribution of miRNA’s in the rat brain

MicroRNA’s are non-coding strands of ~20 nucleotides that regulate mRNA activity by partially base pairing to certain complementary strands and inhibiting translation. Mutations in miRNA’s have been linked to schizophrenia by Feng et al, who found that 8 out of 193 patients with schizophrenia had a mutation in a miRNA on their X chromosome as opposed to only 1 out of 191 control patients. Now Olsen et al have shown via miRNA isolation in rat brains that the distribution of micro RNA’s is variable throughout the brain and is clustered according to biological activity. The researchers conducted unsupervised hierarchical clustering on their normalized expression data to compute the similarity of miRNA expression data in various regions of the brain. Longer branch lengths correspond to more variability in branch length. Here are the results of their cluster dendogram:

Olsen et al, 2009. doi:10.1371/journal.pone.0007225

“Cb” refers to cerebellum, “hip” to hippocampus, “am” to amygdala, “hyp” to hypothalamus, and “sn” to substantia nigra.

This may be an evolutionary accident but more likely the miRNA’s regulate cerebellar or forebrain-specific activities. The authors suggest particularly that the miRNA’s may have a role in neural development, having some sort of interaction with insulin-like growth factor 1. But if miRNA’s are important in the function of the active adult human brain this seems like the kind of thing whose activity would be difficult to simulate in a computer.

References

Feng J, et al. 2009 Evidence for X-chromosomal schizophrenia associated with microRNA alterations. PLoS One Online. PubMed.

Book notes Neuroengineering

I learned a good amount from reading this collection of articles but it would have been better if the book were collected into a coherent whole instead of being so fragmented. Here are summaries of random parts of the book:

Passing electrical current through tissues can stimulate neurons to produce action potentials. In vitro neuron data indicates that action potential initiation occurs in the axon even when the electrode is placed the cell body or dendrites. The electrode is usually placed around 1 mm away from the cell body. Cathode and anode stimulation have different mechanisms but they both eventually lead the depolarization of axons and thus stimulate action potentials in some form. Passing axons near to the electrode may also be stimulated as well as local neurons–this is especially true of cathodic pulses and needs to be taken into account. Moreover, since the action potential is initiated in the axon and not the cell body, extracellular unit recordings of the cell body’s electrical potential may not accurately reflect the neuron’s action potential output.

Deep brain stimulation uses an electrode (aka, a brain pacemaker) implanted in the subthalamic nucleus to stimulate electrical activity to relieve symptoms of Parkinson’s disease, pain, and other neurological disorders. One explanation for its efficay is that neurons are functionally deafferentated by the electrical stimulation, thus limiting the propagation of tremor signals without disrupting other information pathways in the brain.

Studying DBS has yielded some principles that should be true no matter where in the brain the stimulating electrode is placed. The proximate effect of DBS will be axon and dendrite fiber excitation, and it will depend upon their ability to transmit the signal. Below frequencies of ~ 50 Hz the signal should be transduced with high fidelity, but above ~ 100 Hz axons may fail to conduct the signal properly, and synapses may not be able to recycle neurotransmitters in time. This makes sense given that that’s a lot of chemistry that needs to occur more than 100 times per second!

Another interesting avenue is brain-computer interfaces, which change electrophysiological signals (like an EEG rhythm or neuronal firing rate) into a real-world output. Current signal detection methods include EEG, scalp recordings, ECoG, field potentials measured by electrodes, and single units that measured the action potentials of individual neurons. In order to shift the time scale from 1-2 seconds to 2-400 ms, penetrating electrodes that record field potentials or individuals units must be used. Penentrating electrodes that stimulate individual neurons run into issues of glial scarring and general problems with respect to biocompatability, but bioelectrodes and maybe even carbon nanotubes should overcome these problems eventually. ECoG based systems which place electrodes directly below the skin have a 5 times greater magnitude than EEG as well as a wider frequency range. Because ECoG systems also avoid the biocompatability issues of single-unit microelectrodes, they probably have the greatest clinical potential.

Noninvasive sensors will probably required in order for BCIs to become more mainstream. Unfortunately, amplification and recording or spikes noninvasively is not yet possible. Possibly some subset of neurons could be modified and could then turn electrical signals into light of a given intensity so that an external optical sensor could detect it. One possibility way to accomplish this is fluorescent seminconductor quantum dots, but that possibility is only in a theoretical stage currently.

Another interesting technique that the book reviewed was optical nerve stimulation. This is the transient deposition of energy in the form of light leading to an action potential in neural tissue.  A pulsed low energy laser beam has been shown to elicit neural action potentials indistinguishable from conventional electrical approaches on rat sciatic nerves in vivo, with far superior spatial precision and no nerve trauma due to contact. The wavelength of the photon will determine the penetration depth of the simulation, the lazer spot size can be varied down to several micrometers by changing the optical fiber diameter, and the irradiation that the tissue will experience can be varied too.

Since no specific wavelength has been found that always causes stimulation to occur, a single chromophore could not be possible for the direct photochemical effects of the laser. Also in terms of the mechanism of the laser, nerve temperature has been found to increase linearly with laser radiant exposure. So the neuronal activation may be photothermally mediated. As light energy is converted to heat, it causes a temperature gradient in time and space that is relaxed after ~ 90 ms. The molecular mechanism of action has yet to be identified, but the idea that a heat gradient may cause the action potential is very interesting. Importantly, the amount of radiant exposure needed to stimulate neurons (0.3 – 0.4 J/cm^2) is below the energy level that will deal damage to histological tissue (0.8 – 1.0 J/cm^2), making this a clinically viable technique.

Transcranial magnetic stimulation of deep brain regions is also reviewed. This technique relies on an induced electrical field that depends on a time varying magentic field which is generated by rate of change of current (dI/dt) in a bank of capacitors. The coils need to be oriented to produce an E field tangent to the surface of the skull and in the preferred direction of the neurons or axons under consideration. This is another super cool technique. The limiting factor here is often that the intensity of the magnetic field necessary to stimulate the deepest brain regions might cause facial pain or contractions as well as a risk of epileptic seizures.

Most of the techniques reviewed in the book aren’t as new as some might make you believe, although there has been lots of progress in recent years. DARPA grants may spur research into the field, and the book notes that the governments of Japan and France are making BCI’s a national research priority as well. Generally, neuroengineering is a highly technical pursuit that has enormous implications in the long run.

Reference

Editors: DiLorenzo DJ, Bronzino JD. 2008 Neuroengineering. CRC Press, Boca Raton. Amazon Link here.