fMRI, BOLD and the Beautiful

When we want to examine the brain of a person noninvasively by Computed Tomography (CT) or MRI, we get a ‘snapshot’ of the anatomy (or pathology, if any) of the subject’s brain. We are however clueless as to its functional aspect. fMRI or Functional Magnetic Resonant Imaging allows us to do just that. The difference is not unlike a ‘still picture’ versus a ‘video of a moving train’. PET scans, previously described, also can asses the functional state of the brain.

Whenever we do a task, think, dream, memorize, speak or see things, the brain is not activated as a whole; but only certain portions of it are activated. Activation, here, means increased metabolic activity of neurons in certain areas of the brain. Naturally, these ‘metabolically active’ neurons would demand more energy which would power them. The blood supply to these areas increases as a result of this metabolically driven vasodilation. The arteries then bring in glucose and oxygen with them, with Oxygen being transported in the form of Oxyhemoglobin (oxygenated hemoglobin or HbO2). Neurons on the other hand use up the oxygen contained in the blood, thereby reducing it to de-oxyhemoglobin or simply Hb. However, the alteration in tissue perfusion exceeds the extraction of oxygen by the neurons, so the concentration of deoxyhemoglobin within ‘the areas’ decreases. This causes molecular inhomogeneities in the magnetic field.

Oxyhemoglobin is diamagnetic, meaning that they align perpendicularly to magnetic field lines. On the other hand, deoxyhemoglobin is paramagnetic, i.e. it aligns parallely and proportinately with the intensity of the magnetic field. This causes the inhomogeneity within the magnetic field (magnetic susceptibility) in the tissue sampled. This inhomogeneity is exploited in fMRI in terms of decay of transverse magnetization, T2*, with longer T2* values in HbO2 blood and shorter values in Hb (paramagnetic) blood.Since this stems from the oxygen content in blood, fMRI is also known as the BOLD ((blood oxygenation level dependent) effect.

The machine is essentially the same as the MRI machine with echo planar imaging technology that permits faster imaging due to faster gradient switching, improved algorithm and faster CPU processing power. The patient/subject is placed inside the magnetic chamber and MRI signals are acquired, Fourier transformed and corrected for artifacts. Finally the computer reconstructs a 3D fMRI image out of this.

As is obvious, we can learn about the motor areas of a patient by asking him to grasp an object or giving him any motor task and noticing which area(s) of the brain lights up. A neurosurgeon can then be cautious about not hurting these areas. Similarly, the mapping will help spare motor and other vital areas like auditory, visual and language areas from damage in radiotherapy procedures, in addition to neurosurgery. It can also detect occult Alzheimer’s disease and cognitive deficits including those of the autism spectrum and dyslexia (reading disorder).

fMRI can also be employed to ‘read peoples’ minds’, thoughts, intentions including lie detection. Watch the video below which explains how an fMRI scan is done and interpreted.


Thus the legal and forensic implications are obvious. However, in fMRI, correlation doesn't always mean causation. Whatever it may be, it seems that fMRI is very much here to stay, both in the clinics as well as in cognitive neuroscience research. It may also be combined with tractography, MRI or other diagnostic radiologic modalities.

Hardenbergh et al combined Tractography techniques with fMRI, using a technique capable of rendering multiple color-coded functional activation volumes and fiber tract bundles. Many pharmacologically active drugs have effect on memory impairment, which can be seen in ‘telltale’ fMRI scans. Sperling et al studied the effects of lorazepam (a benzodiazepine) and scopolamine (an anticholinergic drug once used as ‘truth serum’ by the CIA) on healthy volunteers and found that they did impair memory and their functional coordinates could be reproducively mapped on fMRI scans (see figure on the left). I still shudder at the thought of what happened during my PG exam when I took a benzodiazepine.

Last modified: Mar 09, 2014
Reference: Integrated 3D Visualization of fMRI and DTI tractography
Gore, J. (2003). Principles and practice of functional MRI of the human brain Journal of Clinical Investigation, 112 (1), 4-9 DOI: 10.1172/JCI200319010

Mirror Neurons: Resonant Circuitry in Brain?

Back in the time of the “black and white” motion picture days, when “talkies” weren’t even born, we still could make out the essence of what Charlie Chaplin had to “say”. We understood his unspoken words, courtesy a system of neuronal networking, called the mirror neuron system. Another example: you observe a man kissing ‘his’ girlfriend, ‘your’ neuronal network that would otherwise activate when you ‘actually’ kissed her, would fire! Mirror neurons are at work. Seems to me a bit like ‘mechanical resonance’, where the string of a guitar resonates (vibrates at the fundamental or overtone frequency of its chord's natural frequency of vibration) when a second guitar/chord is strummed nearby.

It all began with the experiment led by Giacomo Rizzolatti, a neuroscientist at the University of Parma. His team wanted to locate regions in the brain which controlled hand and mouth actions in monkeys, such as grasping or licking of an object. So, they had placed electrodes in the ventral premotor cortex, a part of the brain, [see fig] of a macaque monkey with the hope that whenever ‘that part’ of the brain were activated, the electrode would activate an electronic circuitry and give an audible beep. But all hell broke loose when a student entered the lab with an ice cream in his hand. Every time he was raising the ice cream to his lips, the system responded with a beep! Thus, although the monkey wasn’t having the ice-cream himself (and not moving his limbs), the mere observation of ‘the act’ fired the neurons that would otherwise be stimulated if the monkey ‘actually’ indulged in ‘the act’. The mirror neuron area, ventral premotor cortex, is also known as ventral premotor area F5.

Mirror neurons are defined as ‘those’ neurons that fire when an animal performs some work and also when the animal observes the ‘same work’ being performed by others. In humans, the activity has been traced down to the ‘premotor cortex’ and ‘inferior parietal cortex’ regions of the brain. When a part of the brain ‘fires’ (discharges), it becomes metabolically active and the areas of this enhanced activity may be mapped by a procedure called fMRI (functional Magnetic Resonance Imaging). In a study by Iacoboni et al, 23 right-handed participants were shown different types of image clips (figure on the left). The pictures consisted of a teapot, a mug, cookie jar and related objects in different contexts, action and intention. At the same time the subjects were shown the pictures, the participants’ brains were also being mapped by fMRI to assess the regions of the brain that lit up during the procedure. The premotor cortex and some other parts of the brain showed a significant signal increase on fMRI scans in the action and intention clips. But the signal increase in the Intention condition was much higher compared to the Action condition, with high activity recorded in visual areas and in the right inferior frontal cortex, they noted. Thus the mirror neuron areas of right inferior frontal cortex were involved in understanding the intentions of others, in addition to action recognition.

This ‘sniffing’ of intention behind action is essential to social animals like humans and a deficit in understanding this is seen in autism, a developmental disorder where there is lack of social smile, aloofness, absent eye to eye contact and marked impairment in interpersonal interaction. Autistic children can see sad or happy faces but they fail to ‘read’ the underlying emotions (sadness or happiness). Normally, children acquire mirror neuron activity by the time they are 1 year old. Exactly how they ‘program’ their neurons into being mirror neurons is not known. Learning by Hebbian association has been proposed. Mirror neurons are also involved in language acquisition, empathy and even possibly mind reading, giving credence to the ‘theory of mind’. Telepathy and clairvoyance now seems plausible (psychologists frequently employ transference and counter-transference, kind of ‘feeling’ a patient by their ‘mirror neuron systems’ and consequently ‘filling’ the patient with his own thoughts to remedy patients, in clinical practice.)

Considering their importance in social communication, our brain would have sufficient number of them. Here, I would like to wonder if pedestrian neurons could spontaneously organize into ‘mirror neuron system’ as a person watched say, an action film. Certainly, this can not happen in real-time, as there will be a delay due to visual processing and synaptic passage within the brain. But, given the plasticity of the brain and the dynamicity of dendritic spines, the idea seems conceivable. Mirror neurons also respond to sound. Breast milk ejection of a mother in response to her baby crying is an example. In cases of postoperative urinary retention, sound of running water has helped the patient to pass urine (1). This may be another example in point. It may also shed light about how ‘suggestion’ works in Hypnosis.

Given the diverse range of inputs, the brain must manage (compress) its database as space within the skull is limited. It certainly can not afford to have different sets of mirror neurons for red oval tea cups or green cylindrical ones and so on. So, what the brain does is pattern matching by some ‘fuzzy logic’ or it may simply analyze the scene; break down the signal by some kind of Fourier analysis into simpler functions and then compare resulting signal with its prior database.

Mirror neurons may explain the elusive LSD Flashback phenomenon. It occurs in LSD abusers who are NOT currently taking the drug, but find themselves in a situation reminiscent of a previous drug spree. The person gets a ‘kick’ even though he may have taken it days ago. Clearly, psychedelic lights may trigger a flashback (and watching violent TV programs has been found to activate mirror neurons in children). We should also ask ourselves if dreams, at least some of them, were the handiwork of some of these neurons.

In her fantastic article ‘Cells That Read Minds’, Sandra Blakeslee ponders and exculpates all men from voyeurism:
“In yet another realm, mirror neurons are powerfully activated by pornography, several scientists said. For example, when a man watches another man have sexual intercourse with a woman, the observer's mirror neurons spring into action. The vicarious thrill of watching sex, it turns out, is not so vicarious after all.”

In a lighter vein it may be said that the search engine Google has developed 'mirror neuron like' properties. Just type, “how can i get my girl” in Google search box and watch: Google would ‘ping’ your intention and come up with some real smart choices.

Last modified: never
Reference: (1) Bailey & Love; A Short Practise of Surgery,18e, page 1230
Mirror neurons and the simulation theory of mind-reading
Cells That Read Minds
THE MOTOR CORTEX
Iacoboni M, Molnar-Szakacs I, Gallese V, Buccino G, Mazziotta JC, & Rizzolatti G (2005). Grasping the intentions of others with one's own mirror neuron system. PLoS biology, 3 (3) PMID: 15736981

An Anatomy of Noise And Its Implications

Noise is something we dislike, because by definition, noise means unwanted sound. But this definition is subjective, for what is music to my ears (say the heavy metal band Metallica) is noise to most people. In fact Iraqi prisoners were forced to listen to Metallica songs as a means of torture (culture shock and noise) by the American soldiers. Perhaps a better definition is, wrong sound at the wrong place at the wrong time.

Apart from acoustic noise; there is visual noise as found in television as ‘snow’, electronic noise (e.g. thermal noise or Johnson noise), cosmic noise and so on. Speaking of acoustic noise, one can’t help but think about the dreaded ‘noise pollution’ that seems to envelop us all. In addition to the nuisance it poses, it also causes anxiety, insomnia, increased blood pressure (hypertension), deafness and a hell lot of other bad things. So, it seems that noise is all bad. It’s not always so!

There is a disease called otosclerosis. In this disease, the footplate of stapes (a small bone in the middle ear) gets fixed to the oval window of the internal ear, producing conductive deafness. The patient can not hear normally as the ossicular (bony) conducting chain is at fault. But surprisingly, such persons hear well in noisy places (market, railway station). This phenomenon called Paracusis Willisii is said to occur due to the fact that one has to speak out real loud (over and above the background noise) in such places; thus making this loud voice cross the patients’ threshold of hearing. However, it may also be possible that the amplitude of the voice (in decibel) might ‘ride’ (summate) on the background noise amplitude, and this combined sound amplitude is heard by the ears. The brain then does some kind of fuzzy logic (or acts as a differential amplifier); and the ‘information’ is decoded. So, it seems that noise isn’t all that bad.

In ‘information theory’ even noise is said to contain information in it. One fine example that illustrates how visual noise might contain information is random dot stereography (and autostereogram). So, noise could be meaningful.

In diabetes mellitus, a very common disease across the globe, the blood glucose level rises. This and other metabolic products causes a condition called diabetic neuropathy, among other things. The person’s sense of touch is diminished and this results in inattention to sustained pressure(causes decreased circulation) or trauma to the affected area. This, along with the increased blood glucose and infection may then cause gangrene of the limb which might require an amputation of that limb. Cloutier et al have resorted to noise in an attempt to address the issue.

They applied mechanical noise directly over sensory neurons and have found that both vibration and tactile perception in these patients improved. This mechanical noise was christened as ‘stochastic resonance’ (stochastic means random or probabilistic; this particular term is coined since the frequencies are not tuned to match any particular frequency), and was applied at an imperceptible level. They applied this noise to the great toe of some of the affected individuals, while the controls received none (i.e. no SR). The effect was studied by measuring the vibration perception threshold (VPT). VPT was significantly lower in patients receiving SR compared to the controls (no SR). As the threshold was low, the patients’ sensitivity to detect vibration and tactile sensation improved. They hoped that a continually vibrating shoe insert could improve nerve function in these cases.

In another instance, Toshio Mori and Shoichi Kai of the University of Kyushu, Japan, showed that noise might improve brain function. They shone periodic signals (of 5 Hz flicker) onto the right eyelids and noisy signals onto the left eyelids of the subjects when they were at rest, and measured the intensity of their brain waves. Brain waves are electrical signals that occur in the brain due to the firing of neurons and are detected by electroencephalography (EEG). They found a sharp peak at 5 Hz, the frequency of the periodic varying signal. As they increased the strength of the noise signal relative to the periodic signal, a ‘harmonic’ peak emerged in the alpha wave band at 10 Hz. As the noise signal gained strength, this peak first increased and then diminished. The researchers believe that this harmonic peak is indicative of stochastic resonance in the cerebral visual cortex. Stochastic because of the non-linear way the brainwave behaves in response to the external stimulus. They argue that naturally occurring background electrical noise in the brain (from electron transport chains, neuronal activities) may play important roles in cognition and behavior.

However, not everything about noise is healthy as researchers from the University of California at San Francisco, USA suggest. They exposed healthy young rats to ‘white noise’, (random audio frequencies covering the full spectrum with randomly assigned amplitudes) and found that the development of their auditory cortex was delayed. They used electrophysiology tools to explore this. They also suspected that everyday environmental noise, also a type of white noise, could harm children by interfering with language acquisition and speech.

The story doesn't end here. Researchers have shown that noise has an important role in eukaryotic gene expression. When messenger RNAs (mRNA) are transcribed in the nucleus of a cell, they do so in a 'quantal' way; meaning that mRNAs are produced in spurt, in a stochastic (random) manner. The transcription process needs energy; as the promoter sequence have to be activated and for other biochemical reactions. This transcriptional noise may have implications in phenotypte diversity and cell differentiation process. Alternatively, bacterial pathogenicity may be increased by this 'noise' in gene expression.

The question is: should we scold our children when they continue with those awful noises? I am confused. But one more thing; it was this noise (in the microwave spectrum) that gave scientists the experimental proof that the Universe was expanding.

Last modified: never
Reference: Prolonged Mechanical Noise Restores Tactile Sense in Diabetic Neuropathic Patients.
Cloutier R, Horr S, Niemi JB, D' Andrea S, Lima C, Harry JD, Veves A.
Int J Low Extrem Wounds. 2009 Jan 6.

Noise in eukaryotic gene expression, doi:10.1038/nature01546

Noisy signals strengthen human brainwaves
T Mori and S Kai 2002 Phys. Rev. Lett. 88 218101

White Noise Delays Auditory Organization in the Brain

Noise, Wikipedia

Mori, T., & Kai, S. (2002). Noise-Induced Entrainment and Stochastic Resonance in Human Brain Waves Physical Review Letters, 88 (21) DOI: 10.1103/PhysRevLett.88.218101

Phase Alignment of Neocortical Gamma Oscillations by Hippocampal Theta Waves

An empty brain is the devil’s workshop, goes the proverb. Actually, the brain is never empty. Even in our deepest slumber, the brain continues to weave waves of electrical rhythms that can be seen with the aid of electroencephalogram or EEG. When we place electrodes on the scalp or on the cortex (inside the skull), and amplify the faint signals via bioinstrumentation amplifier, we can lay our hands on these fluctuating rhythms. (More on the electronics of EEG may be found at the OpenEEG project site).

We have as many as 100 billion neurons in the brain. In the superficial layers of the cortex, the neurons have numerous dendrites branching out from the soma or cell body (shown in grey oval in this picture).These neurons have been compared to a forest of trees where the branches are the dendrites and the trunk the axon. These dendrites make extensive connections among each other. They also get connections from the axon collaterals of neighboring axons (i.e. the 'trunks' of other trees connect to these 'twigs' by offshoot from the trunks). Since there are a lot of axons converging on the dendrites of each neuron, and given the fact that these axons can be excitatory (red) or inhibitory (green) depending on the neurotransmitter, the sum of input may be either negative or positive (with respect to the cell body). Thus an alternating current (cortical dipole) will flow between the shifting dendrites and the soma. This along with thalamocortical oscillations produces the EEG waves.

The brain doesn’t churn out the rhythm just like that. Had the neurons fired randomly the oscillations would have cancelled out.EEG waves occur due to synchronous discharge of neurons producing the alpha, beta, theta, gamma and other telltale waves. Like all other electrical waves, they too have a frequency and amplitude. Alpha waves, for example, have a frequency of 8-12 Hz (cycles per second) and an amplitude ranging from 50-100 microvolt when recorded from the scalp, and it is found when a person is resting comfortably with eyes closed and the mind wandering. On the other hand, gamma rhythm has a frequency of 30-80 Hz, and it is found when a person is deeply engrossed on some work.

It was known for a long time that the hippocampus exerted a role in learning by fostering long term potentiation (LTP) by aligning the neocortex, where memories are stored. The mechanisms behind this are now emerging. Sirota et al and Siapas et al have analyzed rat brains and found out that there were many localized gamma oscillators within the brain that gave rise to neocortical gamma bursts. These oscillators had varying frequencies but they phase aligned themselves with the arrival of hippocampal theta waves. A large fraction of pyramidal cells and interneurons too were phase aligned to the hippocampal theta rhythm.This is similar to a bar magnet aligning iron dust or other ferromagnetic materials by virtue of its magnetic field. Apart from the cerebral cortex, the cerebellar cortex and the hippocampus too can generate brain waves. Such a mechanism may explain the orchestration of many parts of the cortex (and hence the memory engrams they contain); and data synchronization and downloading to the hippocampus for memory retrieval. It also shows how hippocampus does the ‘indexing’ of cortical contents. These experiments throw light on neuronal plasticity and information flow, and may be someday they could help clinicians in fighting memory loss as it occurs in neurodegenerative diseases like Alzheimer’s disease.

Last modified: never
References:

Prefrontal Phase Locking to Hippocampal Theta Oscillations
Athanassios G. Siapas, Evgueniy V. Lubenov and Matthew A. Wilson. doi:10.1016/j.neuron.2005.02.028
A SIROTA, S MONTGOMERY, S FUJISAWA, Y ISOMURA, M ZUGARO, G BUZSAKI (2008). Entrainment of Neocortical Neurons and Gamma Oscillations by the Hippocampal Theta Rhythm Neuron, 60 (4), 683-697 DOI: 10.1016/j.neuron.2008.09.014

Do We Really Forget? Fathoming The Esoteric Realms of Memory

“Smells like teen spirit”, but it could be true that we never really loose any memory in our lifetime. Our memories are stored in the synapses (junction, more specifically, gaps between adjoining neurons) as a function of synaptic strength, in the nerve cells like dendrites as proteins, and some other processes which mostly encompasses a chemical interaction. I am excluding memories such as T cell or B cell memories here; memory, here, will refer to neural ones that occur in the CNS.

We know that in dementias such as global multi infarct dementia, Alzheimer’s disease; there are diffuse losses of neurons and losses of cholinergic neurons in particular, respectively. In surgical cases of epilepsy or brain tumor, there are losses of neurons too. In these cases, memory loss may be irrecoverable, though cases are on record which points to shifting of those memories into some other safe havens. But what about the rest of the population? Does an established long term memory vanish completely?

Let’s consider some facts. The numbers of synapses and their strengths are finite, though both can change in response to stimuli. Even the number of neuron themselves can increase, contrary to the belief held earlier. Neuronal stem cell pool has been identified in the brain. Memories stored in the brain are finite too. Memories are inherently dynamic in nature. Even long term memory stored in the neocortex (medial temporal lobe, on the other hand, stores memories as a buffer, like a D RAM chip, a temporary storage) can change location, as much as transferring itself to the other hemisphere (intercortical transfer), when needed; via the optic chiasm and corpus callosum. So, we see that the number of synapses, though finite, can rise to the demand of an enhanced input from sensory cues which are finite too, leading to memories that can jump across their own allocated territories. A finite brain capacity (say C) can certainly contain a finite memory (say M) as long as C is greater than/ equal to M. Certain computer softwares even trespass this limit; a zip file of 2 MB may deliver 3MB of contents on unzipping! Who knows if the brain isn't using this for the past thousand years.

Synapses, simplistically, may be thought of in binary terms: 1, when it is on; 0, when it is off. Both 1 and 0 is a bit in Boolean terms. We leave aside the synaptic strength part here for the sake of simplicity. In addition, memories may shuttle between synapses in such a way so that it is present in the brain, but not represented by any synapse. I will explain. We all have seen those jugglers juggling those colorful balls too many at a time using only their two hands. A similar thing like dipole dynamics may occur in the brain. Added to this is quantum superposition, which allows the situation of BOTH 1 and 0 state at the same time at the synapse. That the brain can be in a quantum state at the core body temperature and the brain can effectively avoid ‘decoherence’ in the background thermal noise has been discussed by Roger Penrose and Stuart Hameroff. We also know that memories aren’t kept as such, but they are fragmented into individual elements, which are mostly matched to existing elements and are associated. This is economic as it saves space, and useful for indexing and contextual retrieval.

Thus it seems that we ought to have immense memory storage. Haven’t we encountered long forgotten memories in our dreams? Electrical stimulations in some parts of the hippocampus (deep brain stimulation or DBS, figure shown) during routine surgical procedures have given rise to ‘deja vu’ phenomena. The patients remembered things considered long forgotten. We may not be aware of the vast database of memories and are liable to infer that we have “killed ‘em all”, but in reality this may not be the case as Norio Ota et al clearly points it out in their paper. It smells like another chapter from your favorite science fiction novel, but it could be true.

Last modified: never; N.B.There is a substantial amount of speculation in this paper. Please exercise your own judgment and enlighten me about any possible error.
Reference: hyper-links, unless specifically mentioned.

Scientists Simulate Learning In Amoeba Using Memristor

It is surprising how small insects get energy from a wide range of food (not merely petrol or diesel), crawl, fly, reproduce and do so many maneuvers. Now it has been seen that amoeba, a unicellular organism, can learn and memorize too. We are far from creating devices of such versatility, let alone making them as compact as they are.

Amoebae can move, and they do this by changing the physical state they are made of: sol-gel state. The interior of amoebae contains endoplasm, which is in sol state; while the surrounding ectoplasm remains in gel state. The ectoplasm, being in gel state, is more viscous than the inside. When the organism moves, its contractile elements made of actin myofilaments contract, pulling the inside of the amoeba. This causes tension in the endoplasm, creating a change in the sol-gel state. If you squeezed a sponge ball that had been dipped in water, you would notice that water would spurt out from the pores of the sponge. Likewise, the increased tension inside, will create channels through the more viscous ectoplasm, courtesy some parts of ectoplasm (gel state) giving away (to sol state).

We know that reptiles hibernate in winter, when the humidity and temperature is low (we too are no exception 😊). Amoebae too, slow their locomotion in response to these conditions. There are inherent oscillations within the amoeba (alternate sol gel transformation, changes in ionic flux etc) which are continuously adjusted with external signals like temperature and humidity. We, complex multicellular organisms, too have our own master oscillator (circadian clock) in the suprachiasmatic nucleus, which also continuously adjusts by lights falling on the retina.

Yoshiki Kuramoto of Kyoto University and colleagues subjected Physarum polycephalum, an amoeba, to three regularly-spaced dips in temperature and humidity, and found that its locomotive activity decreased. Thereafter, they noticed that a single dip was sufficient to elicit this response. It seems they adjusted their oscillations to the external cue and developed a conditioning later. The study implied that the amoeba anticipated that other such dips might be forthcoming, from the memory it learned. Such response did not occur when the temperature and humidity changes were irregular.

Memory in this case occurs due to the persistence of the channels etched by the organism in the ectoplasm. But this ‘memory’ did not persist for long, if we continued giving them a single dip instead of a regular triplet. This plasticity (change due to reorganization as a function of a stimulus) in amoeba has now been simulated with the aid of electronics by Massimiliano Di Ventra et al.

They used a capacitor, a resistor, an inductor in series and connected a ‘memristor’ in parallel with the capacitor. Memristors (for memory resistors) are devices which consist of two layers of titanium dioxide (often present in medicine coatings and chewing gums). When current is applied to one layer, the resistance of the other changes. Leon Chua, of the University of California at Berkeley, predicted it long time ago; and now R. Stanley Williams and colleagues at Hewlett Packard have developed it. It can store memory like DRAM, but unlike DRAM it doesn’t forget when a current is no longer flowing. They hold promise as energy efficient chip for computers and we can also expect faster ‘booting’ of computers, since memory will already have been stored there. The adjoining figure shows ‘memristors’ in a row, as seen by atomic force microscopy (AFM).

Now when a current, fluctuating (AC) in a non periodical manner or a stable DC, was made to pass through the circuit, the memristor went to a low resistance state, virtually short circuiting and dampening the oscillation. However, with a regularly fluctuating current, whose frequency matched the resonant frequency of the circuit, the memristor went into a high resistance state, strengthening the oscillation. What connects electronics to amoeba is the memory that both the circuit retain. The memory of memristor, called memristance, is due to atomic rearrangement in the device. The high resistance state lingers for quite some time, so that next time one single pulse was necessary to put it into oscillation. This phenomenon is quite akin to the protozoal response.

It seems that those days are certainly not far when we will just need to jack-up a USB device in our head to boost up our memory.

Last modified: Nov 13, 2008
Reference: http://arxiv.org/abs/0810.4179?context=q-bio Tetsu Saigusa, Yoshiki Kuramoto (2008). Amoebae Anticipate Periodic Events Physical Review Letters, 100 (1) DOI: 10.1103/PhysRevLett.100.018101

Unlearning Memories: You Have Been Erased!

In the movie 'Eraser', John Kruger, played by Arnold Schwarzenegger, removed the identities and all relevant information of persons-at-risk and gave them brand new identities, to save their lives. The title of this topic derives its name from the famous dialog: "You have been erased!”.

In reality, memories can be erased too, even long term ones! In humans, diseases like Alzheimer’s dementia, cause defects in short term memory limiting further memory acquisition; which finally progresses to erasure of previously stored memories. In experimental animals, long term memories can be prevented if they are subjected to electrical shock, anesthetics (possibly work by disrupting London forces operating in hydrophobic pockets in dendrites thereby causing 'unbinding' of memory elements), hypothermia and other insults within 5 minutes of learning a specific task.

There are guardians of memories keeping a constant vigil so that ‘memories are forever’. For example, we have seen masons at work standing on those makeshift scaffolds. In our bodies too, a protein called PKM zeta constantly ferries across the synapse, to give its healing touch. A protein called PSD-95, when phosphorylated, takes the role of the scaffold, in this analogy. Inhibitors of PKM zeta cause loss of memory, as demonstrated in rats, when their memories for tastes vanished after a single ‘shot’ in their taste cortex. So, LTP is not the only mechanism that fosters memory and neural plasticity. Long term memory also uses gene activation and expression, and protein synthesis. Any interruption in either of these mechanisms will have its own deleterious effect.

Genes have been identified which produce proteins that are necessary for memory formation and their maintenance. These genes act fast in the central nervous system and are known as IEGs (Immediate Early Genes). IEG Arc, one such gene codes for Arc protein, which is abundant in the hippocampal neurons. Obviously, the mRNA (messengerRNA can be thought of as ‘command’ from a gene, while proteins are like the results of this command) for that gene is formed in the nucleus by transcription, but once transcribed, it goes to the dendritic spines (especially those which are active) along the cytoskeletal rail road formed by microtubules. In the dendritic spines they (mRNA) translate themselves into proteins, Arc proteins, in this example.

Guzowski et al did an interesting experiment. They infused antisense oligodeoxynucleotide (for the inhibition of Arc protein synthesis) straight into the hippocampus of rats. Antisense oligodeoxynucleotides (antisense ODN) are short sequences of deoxynucleic acids which block the translation of mRNA into protein. The rats forgot the tasks they learned, while they could still form new memories. Their spatial memory was badly damaged. They concluded that antisense oligodeoxynucleotides interfered with the maintenance phase of LTP.

Rats can be challenged spatially using the Morris water maze. As we see in the picture, the water pool has two raised platforms, which rats can locate (and remember) spatially after some training. Even the rats already trained successfully, lost this spatial (co-ordinates in space) memory and had severe difficulty floating, when they were challenged with these chemicals. Looks like, You Have Been Erased!

Contradictory (against the motion) Link: Memories like diamonds may be forver

C. K. McIntyre (2005). Memory-influencing intra-basolateral amygdala drug infusions modulate expression of Arc protein in the hippocampus Proceedings of the National Academy of Sciences, 102 (30), 10718-10723 DOI: 10.1073/pnas.0504436102

LTP Ensures That Memories Are Forever

LTP or Long term potentiation is a process that may explain how memory gets stored in the brain for long term use. When you stimulate the presynaptic neuron by giving a brief (of transient duration) but rapid train of stimulus, the post synaptic neuron adjusts its ‘weight of association’ with respects to the presynaptic one, in the form of a chemical reaction. Though LTP occurs throughout the brain, it has been studied mostly in the hippocampus. If we are to understand the underlying molecular mechanism of memory, we can not do without LTP.

Two different types of LTP are known: mossy fiber LTP and Schaffer collateral type LTP. While the basis of mossy fiber LTP is not clearly known; it involves modification of the presynaptic terminal, and is independent of NMDA. A schematic and functional diagram for Schaffer collateral LTP is presented here. But before that, allow me to digress a little bit.

Your computer has a DRAM (Dynamic Random Access Memory) memory chip: memory because it can store and retrieve information; Random access as it allows you to search anywhere within the memory at random (it does not have to reach D via A through B, and then through C, sequentially), and dynamic, since the memory needs to be refreshed from time to time. To store a bit of data in memory, your computer charges devices called ‘capacitors’ within the chip, which retain their charge; and all your computer have to do is to read the data in the form of those stored charges for later retrieval. But these capacitors lose their charge over time and hence dynamic refreshing is necessary to maintain their memory.

You’ll now understand why this digital analogy as we discuss LTP. The picture on the left portrays a presynaptic neuron which discharges glutamate, the main excitatory neurotransmitter of the brain and the spinal cord. Glutamate after being released upon the stimulation of the presynaptic terminal, binds with their ‘receptors’ in the postsynaptic neuron. The post synaptic neuron, downstream, has 2 types of Glutamate receptors: NMDA (N methyl D Aspartate) and AMPA (alpha Amino 3 hydroxy 5 Methyl isoxazole 4 Propionate). Glutamate binds with both NMDA and AMPA receptors. NMDA receptors have a Magnesium ion, guarding at its channel entry. So, for NMDA receptors to act, it needs to be partially depolarized first, so that this magnesium block is removed. This is achieved by the AMPA receptors, which upon binding with glutamate, allows the entry of Sodium ions inside, thereby raising the cell voltage. NMDA receptors now swing into action as it now allows huge amounts of Calcium ions (and Sodium ions) to enter inside.

These Ca++ then bind with Calmodulin present within the cell to form a complex, which then activates calcium-calmodulin kinase 2 (Ca/Cam k2). This newly formed compound then activates (phosphorylates) AMPA receptors, resulting in: 1) increased activity conductance of the already existing AMPA receptors in the cell membrane 2) Recruitment of AMPA receptors from within the cell to the cell membrane. So we can see that the synaptic strength is increased with each firing by both AMPA recruitment and increased AMPA conductance. The synapse stops at not only this, the postsynaptic neuron also discharges a ‘diffusible’ messenger, nitric oxide (NO), which 'tells' the presynaptic neuron to discharge more quantal release of glutamate next time. The phenomenon epitomizes Hebbian learning: Cells that fire together, wire together.

But the memories so formed need to be stabilized as in the case of DRAM. In the central nervous system, dendritic spines are the main postsynaptic sites. These tiny protrusions form and change over a few hours. In hippocampal slice cultures it was shown, by De Roo and colleagues, that application of theta burst remodeled the dendritic spines; unused ones were shed (trimmed) while used ones were stabilized and new spines were formed. LTP was the chemical basis of all these modifications. They used GFP or green fluorescent protein for visualizing these changes of neural plasticity. However, they (physical units of memory) can also be seen by restorative deconvolution microscopy, in the form of flattened synapses (as if the ohmic resistance getting diminished in their electronic cousins) and hence more area for contact between the pre and postsynaptic neurons. So like DRAM chips, our memory chips too need to be constantly refreshed, even long term memories need maintenance.

Last modified: Jun 26, 2010
Reference:Dominique Muller, Morgan Sheng, Mathias De Roo, Paul Klauser, & Morgan Sheng (2008). LTP Promotes a Selective Long-Term Stabilization and Clustering of Dendritic Spines PLoS Biology

Neural Networking, Alzheimer's Disease and Memory Share a Few Things

The hippocampi, as has already been discussed, are endowed with the task of binding elements of memory into a coherent trace. There are extensive interconnections within the hippocampi in the form of a local area network (LAN) or intranet. This LAN like architecture resembles the “Hopfield Network” of artificial neural networks. It is especially abundant in the CA3 neurones of the hippocampi. CA stands for Cornu Ammonis; meaning the horn of Ammon (After the ancient Egyptian God Amun).

In a Hopfield network, neurodes (neuronal equivalent of neurons) are connected to each other in the form of a bidirectional interconnection. However, a Hopfield network assumes that each neurode can have either of the two states: on (1) or off (0). But human neurons, we know, can also have additional states when it is in the summation or subtraction mode (spatial summation, temporal summation etc, whereby one neuron is incapable of eliciting a binary 1 or no action potential; but when suitably combined in a time [temporal scale] or a space scale [spatial], it can ). Reverberatory Hopfield networks reinforce each other, thereby strengthening the associations among themselves.

Earlier, mathematician cum scientist von Neumann tried to find an analogy between the memory architecture of a computer with that of a human brain. In a von Neumann architecture based computer, data can be fetched from their ‘address locations’ within the database when queried. Retrieval of data in a von Neumann model is sequential, that is the information flow is one by one. Human brain, on the other hand, handles data parallely, and there is no ‘bottleneck’ (which is inherent in the von Neumann model due to different locations of CPU and memory). No bottleneck means higher data transfer rate in the human brain. Human brains differ in another aspect. Secondly, human brain can think in terms of abstract terms, while a fixed program computer like von Neumann’s one, can not do so easily. In fact, human brain is far too superior in ‘fuzzy logic’ or abstract thinking compared to their silicon cousins.

The reverberatory circuit in the hippocampi is continuously synchronized with each input stimulus. This is akin to the synchronization of our circadian rhythms with that of daylight, via the retino-hypothalamic tract. The hippocampal anatomic correlate is somewhat like this: entorhinal cortex---> dentate gyrus---> CA1 & CA3 pyramidal neurons---> subiculum--> back to entorrhinal cortex. This circuit is heavily damaged in Alzheimer’s dementia resulting in loss of episodic memories and preventing acquisition of new memories. In artificial neural network model simulations, Traub and colleagues showed that the synchronization was done in the gamma frequency range. They showed that with the arrival of each theta cycle, the attractor-based autoassociative memory process got stronger and the attractor got stable after a few theta cycles. An attractor is somewhat like a binder, which indexes and binds information.

The hippocampi also get input from the basal ganglia and other portions of the brain. These regions have a diverse range of neurotransmitter chemicals. They include acetyl choline, GABA, NMDA, Dopamine, AMPA and many others.

In acetyl choline deprived state, as occurs in Alzheimer’s disease, the frequency of gamma  discharge diminished, leading to diminution of theta frequency thus delaying learning (and promoting unlearning or unbinding too).

Its time to disentangle the intracytoplasmic neurofibrillary tangles; and clear the mystery of the amyloid plaques which are so characteristic of AD. The stage, it seems, has been set. 

Last modified: May 1, 2013
[Picture courtesy: http://wikis.lib.ncsu.edu/]

Reference: Attractor neural network models of spatial maps in hippocampus

Misha Tsodyks

Department of Neurobiology, Weizmann Institute of Science, Rehovot, Israel 

E . Menschik (2003). Neuromodulatory control of hippocampal function: towards a model of Alzheimer''s disease . Artificial Intelligence in Medicine, Volume 13, 99-121

The Organization of Memory And Internet Analogy

You are reading this particular article now on my website (in your computer and browser, of course!). There are billions of such websites and you picked mine. Perhaps you put your queries in the search box, which then showed you a list of results. You clicked my link there and lo, you are here!

The search engines (Google, Yahoo!, MSN and others) parse or read the webpages, index them and rank them. Each search engine has its own search robot, called bot, which crawls the pages. Now, this particular webpage that you are seeing is a part of my website. If you think of a website as a tall building, webpages may be considered as its floors. The search spiders not only crawl the website vertically (that is vertically up and down its individual floors), but they also move horizontally (clicking hyperlinks will lead you to arrive at some floor of another building) to be navigated to another webpage of another website.

Longitudinal (vertical) scanning picks up individual elements of the content, which are indexed (tagged) and remembered with respect to their locations, much like the human episodic memory in the hippocampus.

Now, its time for the bot to leave. It follows a hyperlink, if you have one, to a page of another website. So the search engine spider is directed to some floor in another building. Thus one might think of this search engine analogy to “neurodes and synapses” in artificial neural network. Here neurodes are the floors of the buildings and synapses are the hyperlinks (connections between them).

Next the search engine (Google) looks around in the floor to find similarities and dissimilarities from the floor it came. That is, it compares the webpages for relevance, in much the same way the anterior cingulated cortex (ACC) calculates “error related negativity” (ERN), and learns from it. It then calculates the ‘weight’ of association. It too learns from time to time by constant error related feedback, as happens in the ACC. Thus it assigns its hallowed “pagerank”.

PageRank is the relative importance of a page in Google’s eyes. ‘Page’ could come from Larry Page, co-founder of Google or more likely after ‘webpage’. One can see them in the Google toolbar in IE or Firefox, as a small green bar (Google’s own browser ‘chrome’ is yet to provide with a pagerank though). The lowest rank a page can have is 0, and the highest 10. Though the algorithm of pageranking is kept a secret like that of CocaCola, some insights may be had from “The Anatomy of a Large-Scale Hypertextual Web Search Engine” by Sergey Brin and Lawrence Page, founders of Google. I found a lot of resemblance of this paper with the architecture of learning and memory.


Last modified: never
Reference: hyper-links, unless specifically mentioned

The Binding Of Memory And Hebbian Learning

Imagine this. Your physics professor is taking his regular class, and a boy sneezes. The professor is startled and the chalk falls from his hand. The whole class breaks into laughter. The memory of this particular incident gets stored in your head.

The whole process is somewhat like this. The visual scenes (your teacher’s attire, the chalk and the blackboard etc) are analyzed and processed in the visual association areas; the sounds (sneeze, laughter) get processed in the sound processed in their respective association areas and so on. Thus memory seems to be broken down to its individual elements; and these elements confine themselves in the areas in the neocortex where they were first processed.


Hippocampus (means sea horse in Greek) is a small banana shaped structure inside the brain. It gets its input from these association areas via another structure called the parahippocampal cortex. Hippocampus, as if, queries those areas: ‘what happened’, ‘when did it happen’ and ‘where did it happen’? Then the hippocampus binds all those information in the form of an event. For this particular episodic memory (classroom drama); the hippocampus wires together the respective areas so that the whole event is now bound together into an ‘engram’, the proposed neuro-anatomical representative of a particular memory.

While all these are happening in the medial temporal lobe (MTL), more specifically the hippocampus, the actual memory elements are still in the neocortex. With each recapitulation, voluntary (by thinking about the incident) or involuntary (someone else’s sneezing reminds you of that event), the association gets stronger. Thinking of sneezing (memory stored in neuron A) reminds you of chalk falling (neuron B). This way as A becomes active (fires), B is associated too, and this leads to wiring them. This is known as 'Hebb’s rule', after Donald Hebb, a Canadian scientist. Simply put, it says, ‘cells that fire together, wire together’.

The strengthening of synapses as a basis of learning were later found to have been mediated in part to LTP or long term potentiation, a chemical process. While the memories are being strengthened this way, the neocortical areas become more and more inter-connected. This releases the MTL connection, as the memory finally gets settled in the neocortex. This 'plasticity' is important. The neocortex stores memories effectively, has a large storage space, but it learns slowly. The MTL learns quickly, but has little storage space. One can think of neocortex as the ‘hard disk’ and MTL as the ‘RAM’ of a computer. Freeing the MTL would enable it to acquire memory more efficiently.

No wonder my teacher said, ‘Read once, write twice, think thrice’. Its time we consolidated our memories by recap.  

Last modified: never
Reference: hyper-links, unless specifically mentioned

Learning Memory Lessons From Aplysia

While you are reading this on your computer, there may be many distractions in the background. Your mom may be shouting at someone or your daughter may be receiving her piano lessons. But after you finished reading this, no memory trace of this background remains. They are not registered with your memory. We seem to ignore this 'negative memory' by a process called 'habituation'. We remember by consolidating memories by another process known as 'sensitization'. But what is this that we call memory? Memory helps us to store, retain and retrieve information. To start it all, learning is needed.

Now you might ask why Aplysia out of all animals?

Aplysia californica, called 'sea hare' by the ancient Greeks, is a marine snail which has some resemblance to a rabbit. A hermaphrodite by sexual orientation, Aplysia feeds on marine algae, and often takes the color of the algae that it eats. When threatened, it liberates a colored, irritant compound to blind the attacker, as seen in the left. It was Eric Kandel who used this phenomenon to study how neural transmission occurred through synapses. He was awarded the Nobel Prize in Physiology or Medicine in 2000 for his pioneering work on Aplysia. Aplysia offers a distinct advantage by: 1. eliciting a visible (measurable) response (siphon-mediated gill withdrawal reflex) to a stimulus, that can be studied directly; 2. the response is triggered by several electrical synapses firing simultaneously, exemplifying several output in response to a single input; 3. Smaller number of neurons, about 20,000; 4. BIG neurons. Hence this animal is considered a role model in neurobiology.

The organism can be stimulated by the application of a minor electric shock to one side of the siphon, and the magnitude of response can be measured by a force transducer, a device, usually piezoelectric, which converts mechanical pull in terms of electricity. Applying a more intense shock to the tail, or administration of chemicals into its abdominal ganglion are other ways of stimulating the mollusk. After careful observations, scientists postulated that habituation, sensitization and classical conditioning were responsible for the learning that occurred in Aplysia.

In the picture on the left, you can see a presynaptic neuron on the left side, a postsynaptic neuron on the right and another neuron (facilitator terminal) on top left stimulating the presynaptic neuron. The flow of impulse propagation is from left to right, that is, from presynaptic to postsynaptic neuron. When the presynaptic neuron was stimulated alone, the post-synaptic neuron responded in a ‘what is it’ response. When the stimulus was given repeatedly, then the response of the post-synaptic neuron became less and less. This implies that the post synaptic neuron became as if ‘habituated’. On the other hand, when a noxious stimulus was applied at the ‘facilitator terminal’ at the same time the presynaptic neuron was stimulated, the response became stronger and stronger. They called this ‘sensitization’.

In the case of habituation, it was found that the release of neurotransmitter diminished in the presynaptic neuron, possibly due to progressive inactivation of calcium ion channels. It is the calcium ion channel which allows degranulation and release of neurotransmitters. Thus inactivation means blockage of impulse. Structurally, both the presynaptic vesicle number and size are seen to decrease in habituation.

In sensitization, both the presynaptic vesicle number and size increase. Here the ‘noxious stimulus’ delivered alongside, causes release of serotonin from the facilitator terminal. This chemical, also called 5HT, binds with 5HT receptors in the presynaptic terminal. This is followed by activation of an enzyme known as adenylyl cyclase, which in turn produces cyclic AMP from ATP. cAMP then activates another enzyme, protein kinase A (PKA), which phosphorylates potassium ion (K+) channels. As a result K+ channels get blocked. When an imulsee arrives at the synapse, the membrane gets depolarized, i.e., the inside of the cell becomes positive with respect to the outside. For the cell voltage to return to its normal polarized state, the potassium ions must diffuse out, through K+ channels. But the exit route is blocked now. The action potential remains longer; more Ca++ enters into the presynaptic terminal.

Recent evidence points at the role of post synaptic neurons in habituation, sensitization and classical conditioning. Glanzman et al have proposed that activation of postsynaptic glutamate receptors might play a critical role in mediating long-lasting habituation of gill withdrawal reflex in Aplysia. A whole new range of activity starting at the synaptic knob to the nucleus and thence back to the knob again has been proposed for memory storage.

Short-term memory, which usually lasts for a few minutes, involves covalent bonding of pre-existing proteins leading to alterations in the strength of already existing connections. By contrast, long-term memory requires mitogen activated protein kinase (MAPK), CREB and new mRNA and protein synthesis; in addition to PKA.

MAPK migrates into the nucleus where it phosphorylates cAMP-responsive element binding protein (CREB). This then regulates gene expression, resulting in transcription and translation. Moreover, long-term memory is associated with the growth of new synaptic connections, phosphorylation of post synaptic densities (PSD is like a scaffold on which proteins are assembled) and many other processes. Thus it seems post synaptic mechanisms might have a bigger role than imagined.

So far, we discussed about only one form of memory: implicit or non-associative memory. Declarative or explicit memory consists of semantic (book) and episodic (related with places, memory of events) memory, which are consciously stored. Humans are perhaps, the best or only known subject in this regard. We certainly can't expect Aplysia to speak out.

Last modified: never
References:
Prolonged Habituation of the Gill-Withdrawal Reflex inAplysia Depends on Protein Synthesis, Protein Phosphatase Activity, and Postsynaptic Glutamate Receptors
Youssef Ezzeddine, David L. Glanzman

Molecular Mechanisms of Memory Storage in Aplysia
Robert D. Hawkins, Eric R. Kandel and Craig H. Bailey

C. Bailey, M Chen (1983). Morphological basis of long-term habituation and sensitization in Aplysia Science, 220 (4592), 91-93 DOI: 10.1126/science.6828885

Dendrites and Memories

In my previous post 'Spiny Dendrites', I discussed about dendrites' role in learning and memory as well as its role in evoking consciousness. Exactly how these dendrites learn and more importantly memorize, has not been clearly understood, but they are being discovered at a rapid pace.

On the dendritic cell membrane, nerve endings from various other neurons converge at the synapse, the gap between those neuronal processes and the neurone on which it is converging upon. The synapses are generally chemically coupled (by neurotransmitters), but they may also be electrically coupled (as in the lateral vestibular nucleus) or both electrical and chemical coupling may be found (conjoint). Chemical coupling is the most common. In this mode of signal transmission, whenever a presynaptic nerve gets an electrical signal (in the form of an action potential), it releases a neurotransmitter (the chemical coupler) into the synaptic cleft. The post-synaptic nerve (the nerve it verges upon, downstream) senses these chemicals (acetylcholine, adrenaline, glutamic acid, GABA etc) by virtue of 'receptors', located mostly in the 'post synaptic densities' (thickened portions of post-synaptic dendritic/neuronal cell membranes, adjacent to that part of the presynaptic membranes releasing the neurotransmitters, i.e. the synaptic vesicles).

Recently, it has been discovered that PSD-95 (post synaptic density-95), a key protein constituent (of postsynaptic densities), on which other proteins are assembled, needed to be phosphorylated, for it to act correctly (as a scaffold, for other proteins to act). Researchers suggested that this (phosphorylation) could help improve cognitive function in schizophrenia, autism and depression. I, personally, think that they will be of use as nootropics too.

In another development in neurobiology, it was shown that memories were burned into the neurons (as you burn 'memories' into a CD using Nero or other programs), and these basic units of memories had to be constantly recycled so that the memories remained there. These memory units, in the form of 'receptors' (of the neurotransmitters; which were located in the postsynaptic neuronal membranes), were in constant motion within the 'fluid' cell membranes. This 'recycling' might explain how the dendritic spines maintain their characteristic identity despite constant molecular turnover. This research might help us fight neurological disorders, such as Alzheimer's disease, schizophrenia, or learning disorders like autism.