【拔丝学堂】22: BRAIN, INFORMATION AND BEHAVIORS

罗新

拔丝

生物系

学堂

马明

蔡翘

讲师团

神经学

22

BRAIN

INFORMATION

北大84级微信群的“拔丝学堂”，这一次是由生物系组成了豪华讲师团。庄晓曦、马明红和郑劼三位84生物系的校友，目前都是美国知名大学里的神经学终身教授。他们虽然分处美国不同的城市，但是通过网络共同备课几个月，反复修改磨合，终于形成了这个全面、严谨、深入浅出的讲座。现将英文版本分享给大家。

BRAIN, INFORMATION AND BEHAVIORS

Xiaoxi Zhuang, Minghong Ma & Jie Zheng Part I) How does the nervous system meet the computational level challenges in adaptive behaviors?

Before talking about the brain, let’s first think about a single-cell organism. A bacterium of course does not have a nervous system. Nevertheless, it can sense light although we don’t call it seeing. It can sense mechanical forces although we don’t call it touch or hearing. And most importantly it can sense chemicals in its environment although we don’t call it smell or taste. 

When sensor proteins on bacterial surface sense nutrients in the environment, they activate signaling pathways inside the bacterial cell. The cascade of biochemical events eventually directs the movements of motor proteins in flagella, therefore control movements of the bacterium. The bacterium can not only swim towards nutrients and swim away from harmful chemicals, it can even compare concentrations of nutrients in the environment as it moves along and swim towards more concentrated nutrients from less concentrated nutrients (chemotaxis in a chemical gradient). This is made possible by using temporal sensing to decide whether its situation is improving or not. So even a bacterium has “memory” and can make “decisions”. It is quite amazing to think about bacteria’s adaptive behaviors. “Adaptive” here means that such behaviors increase the chance to survive and to reproduce. 

Now let’s think about organisms that have trillions of cells, have far larger sizes, and much more complex adaptive behaviors. The human body has so many organs, tissue types and cell types. They form systems that specialize in sensing a specific aspect of our constantly changing environment: visual, auditory, olfactory, gustatory, and somatosensory systems. Each cell only does a small part of a special task, and they have to work together. Communication between different cells, i.e., coordination and integration of information, is the first challenge. Information from different cells and then from different senses has to be processed, synthesized and integrated. Similarly, in motor control, different muscles have to be coordinated in specific sequences (motor programs) for any specific task in response to the rapidly changing environment. 

For coordination and integration, the challenge is at least partially met with a centralized control system: the nervous system which receives all sensory s, sends out all motor outputs, and most importantly does all the computations between s and outputs (sensorimotor integration), which will be discussed more below. 

For communication between cells, organs and systems, another challenge is size and speed. The human body is made of cells which are tiny compared to the human body. A typical cell is only 5-20 micrometers large. How can information from sensory organs (e.g. skin on your toe) travel the distance to reach the brain, and then motor control information from the brain travels to muscles? And it has to be quick (think about your toe is on fire). This challenge is met by at least two unique features of neurons, which are a special type of cells that are the basic building blocks of the nervous system. 

Feature #1: Although cell bodies of neurons are not large, about 10-20 micrometers, they can send out thin axons that can be very long (Fig.1-1). For example, the axons of motor neurons in the spinal cord that control the toes can be more than 1 meter long. 

 Feature #2: Within a neuron and axon, information is transmitted via electrical signals. Although such signals do not have the same speed as that of electromagnetic field, they can still travel at 1 to 120 m/s depending on the axon diameter and myelination. 

 Now let’s try to understand the simplest sensorimotor integration, a reflex (Fig. 1-2). In a reflex, motor neurons are almost directly controlled by sensory neurons; it only involves a few neurons. It usually does not go through the brain; it only goes through the spinal cord. It is innate and does not require learning. It does not require attention either; you can do it automatically.

Some complex movements are innate too, e.g. crying. It does not require learning; and you can do it automatically without paying attention.

Another type of movements that you can do automatically without paying attention are habits. However, habits have to be learned. Before habits are formed, you have to spend a lot of attention on what you are doing (e.g. driving, riding a bike, playing piano, playing tennis, etc). However, once habits are formed, you can do them automatically without paying attention. Habits are extremely important to us. They make us very skilled at certain complex movements, make our responses much faster, and free up the brain’s limited on-line computational capacity. However, habits have their downside, i.e., they are not flexible. Think about an experienced mainland driver is now driving a car on the left side in Hong Kong. Think about “bad habits”. 

In our daily lives, our behaviors are a mixture of reflexes, innate complex and stereotyped behaviors, learned habits and also flexible goal directed behaviors. Flexible goal directed behaviors are the opposite of habits. It takes time and computational power to have different options uated, and decisions are made. Sometimes even long-term plans are made (e.g. delayed gratification). For example, if you are a very hungry hunter, you have to calculate your chances of success before chasing a buffalo. If you fail, you will be wasting your precious remaining energy. If you are a hungry forager, you have to decide whether you want to go to a nearby patch to collect a tiny amount of fruits which is a sure thing or go far away to search for a patch with a lot of fruits which is uncertain, expends a lot more energy and time, and also carries the risk of predators. In most cases, we often implicitly calculate the predicted value of each option and choose the option with the highest expected value (with probability of success and risks factored in). 

Another scenario is high expected value now but a lot higher expected value in the future. Which one will you choose? Being able to tolerate delayed gratification is often believed to be important for career success.
All those decisions discussed above require a lot of computation and predictions of the future, i.e., imaginations based on previous knowledges (memory), current external environmental condition and your internal state (e.g. hungry or not, specific nutrition needs, etc.). 

These decisions and actions can even be affected by emotions. Emotions may cloud your judgement. However, the benefits of emotions include courage and energy etc (more discussions are in Part III). In addition to coordinating different muscles, the brain also controls the heart, lung, blood vessels, hormone secretions etc. These are involuntary motor functions, but they are also important in raising or decreasing energy level related to our actions among many other functions.

All these flexible behaviors are far beyond what AI can do. And it comes down to one question: how does the brain solve such a complex problem? We have to start with information processing within the basic unit of the brain: a neuron.

Some sections may be tedious in Part II-V. You may skip the tedious sections as each part is relatively independent. 

Part II) The physical components of the nervous system and information flow. 

Like a computer, the brain receives s and sends out outputs. Like a computer, the brain is wired together with many elements in a complex pattern. However, very different from a computer, the brain is wired together with neurons as its basic units and neurons are connected together via synapses. 

A typical neuron (see Fig. 2-1) has 1) the soma (body of the neuron), 2) one long axon which sends signals from the soma to the end of the axon (the axon terminal) and reach the next neuron via a synapse, and 3) many dendrites which receive signals from axon terminals of many other neurons via synapses. Each neuron usually receives signals from hundreds to thousands of other neurons; and each neuron usually sends signals to another neuron or a few other neurons (in this case the axon will form branches). 

Within a neuron and axon, signals are usually transmitted from dendritic spines, to dendrites, to soma, to axons and eventually to axon terminals. These are electrical signals. Where does the electricity come from? Electrical currents are generated by “ion channels”. Just like transistors used in classic man-made electrical devices, ion channels in neurons generate transient (in millisecond scale) and tiny (10-12 ampere in magnitude) currents. Ion channels are proteins in the surface membrane of neurons (and all other cells in our body!). They form an aqueous pore that can be open and close. When an ion channel opens, the aqueous pore connects the inside and outside of a cell across the greasy non-conducting cell membrane, allowing ions (charge-carriers) to move across. Ion movement produces electrical signals that support calculation and communication operations in our brain (and many other things like muscle contraction, insulin release and fertilization).

Basic operations of a computer are carried out with current pulses. The same is true for a neuron (and brain). To generate a simple electrical pulse in neuron, three types of ion channels are needed: one to hold the neuronal membrane at a base level (the resting membrane potential, at about -70 mV), another to depolarize and a third to repolarize. The first type of ion channels play a house-keeping role and are open all the time. The second and third types of ion channels actively open and close (in a stochastic manner but as a population follow a precise temporal sequence). Their openings lead to depolarization or hyperpolarization because they conduct different ions—the depolarizing ones conduct sodium ions (Na+) and the hyperpolarizing ones conducts potassium ions (K+). In Dr. Feng’s wonderful introduction to stroke, he already mentioned that Na+ and K+ are distributed differently. The inside of a neuron is high in K+; the outside (equivalent to the blood) is high in Na+. Na+ moves into the neuron to depolarize, and K+ moves out of the neuron to hyperpolarize.

Alan Hodgkin and Andrew Huxley showed in the 1950s that a combination of these three types of ion channels was sufficient to produce an electrical pulse that looked qualitatively like one recorded directly from a neuron, which is called an “action potential” (Fig. 2-2). The simple system (model) they built could be represented by an electrical circuitry (Fig. 2-2), whose behavior could be simulated by a computer. At the time, H&H had only a manual computer. It took them days to generate their famous “action potential”. Nowadays, an electrical computer can do it in a flash. (Search online for free programs using the term “action potential simulation”.) 

 H&H’s work was quite impressive because in 1950s no one had seen an ion channel or shown its existence. For their landmark contribution, they were awarded a Noble Prize in Physiology or Medicine in 1963. Erwin Neher and Bert Sakmann later recorded the 10-12 ampere current directly from a single ion channel (Fig. 2-3); they were awarded a Noble Prize in Physiology or Medicine in 1991. Rod MacKinnon solved the structure of a K+ channel at the atomic resolution, from which we can see why it conducts only K+ but not Na+ (Fig. 2-4). He shared a Noble Prize in Chemistry in 2003. Just like in many areas of biomedical research, our understanding of neuronal signaling has advanced at an amazing pace. And the pace is getting even faster right in front of our eyes!

 The above discussion es on information transmission within neurons via electrical signals. Different neurons are connected by synapses (Fig. 2-5). Between two neurons and within a synapse, signals are transmitted from the axon terminal of the first neuron (pre-synaptic neuron) to the dendritic spine of the second neuron (post-synaptic neuron). This is usually achieved by chemicals, called neurotransmitters, released by axon terminals of the pre-synaptic neuron and received by neurotransmitter receptors localized on dendritic spines of the post-synaptic neuron. 

Depending on the neurotransmitter released from the presynaptic neuron and the specific receptor expressed on the postsynaptic cells, firing of the presynaptic cells can either excite or inhibit its postsynaptic cells, depolarizing or hyperpolarizing the cell membrane, respectively (in other words, making the postsynaptic cells easier or harder to fire action potentials).

The predominant excitatory neurotransmitter in the central nervous system is glutamate (Glu) and neurons that release Glu are called glutamatergic neurons, including many long-range projection neurons (e.g. those in the cortex and thalamus; see below). Some Glu receptors (AMPA and NMDA) are fast action ligand-gated cation channels (also called ionotropic receptors) – Glu binding causes Na+ and Ca2+ flowing into the cells and membrane depolarization. Other Glu receptors are G protein coupled receptors (GPCRs) (also called metabotropic receptors). Glu binding activates distinct second messenger signaling cascades that either indirectly affect ion channels or have other functions (e.g. gene expression). These are slow action receptors.

The predominant inhibitory neurotransmitter is GABA. GABAergic neurons include local interneurons in every brain region and projection neurons in some brain regions (e.g. cerebellum and basal ganglia; see below). Some GABA receptors are ligand-gated Cl- channel (GABAA receptor – GABA binding causes Cl- current which typically hyperpolarizes the cell membrane; faster action). Others are GPCRs (GABAB receptor ; slower action). 

Some neurotransmitters (e.g. dopamine, serotonin, norepinephrine) are sometimes called neuromodulators because they almost exclusively activate GPCRs. Such a function is often called neuromodulation as oppose to neurotransmission. Neuromodulation via GPCRs does not cause direct excitation or inhibition, it modulates neurotransmission instead. One important implication of neuromodulation is that a neuromodulator (e.g. serotonin) can send its signal from one place to almost the entire brain to modulate information processing in the entire brain depending on the “state” of those serotonin neurons (e.g. sleep, awake, hunger, stress, or mood)

Many neurons corelease multiple neurotransmitters (Glu or GABA with neuropeptides, dopamine, etc). A postsynaptic neuron sums all its synaptic s spatially and temporally, which shapes its firing pattern at any given time. 
Chemical signals between neurons are much slower, at the speed of a few milliseconds. Chemical synapses dramatically slow down the speed of information transmission in the brain and across the body. In reflexes, there are usually few neurons and few synapses involved so as not to reduce the speed of information transmission too much. In some special but very rare cases, electrical synapses rather than chemical synapses are used to ensure rapid singal transmission.

One interesting note is that many of those knowledges we have accumulated about ion channels and neurotransmitters are helped by neurotoxins or psychoactive drugs found in plants or animals. For example, tetrodotoxin in the pufferfish blocks the Na+ channel responsible for action potentials. Nicotine from tobacco activates one important type of neurotransmitter receptors. Plants and animals have evolved to make special chemicals to fend off their predators or to attract their friends.

It is important to emphasize that neurons do not simply relay information. Instead, each neuron could receive synaptic s from hundreds to thousands of other neurons; the information has to be integrated. Similarly, synapses do not simply relay information. Instead, synapses are the most important places that information is integrated and modulated. For example, one can influence another . As discussed above, this is often called neuromodulation. Therefore, each synapse is very complex on its own. Each synapse has some level of autonomy, i.e., different synapses on the same neuron may behave differently.
Another very important property of synapses is activity dependent changes in synaptic strength (see Part III on memory). This is the most important mechanism underlying learning, i.e., the same can lead to different output next time due to experience. Although each synapse has to be plastic, it also has to be relatively stable as well. Otherwise, memory will not be possible.

To summarize, at microscopic level, information from one neuron is transmitted to another neuron at synapses via chemical signals. Within a neuron, information is transmitted from post-synaptic dendritic spines, to dendrites, to soma, to axons, and to axon terminals via electrical signals. Neurons and their axons and dendrites form a very complex network. Each neuron usually has a dendritic tree that receives signals from hundreds to thousands of other neurons; and each neuron usually sends signals to another neuron or a few other neurons through its long axon(s).  One may suggest that, if a neuron can be simplified into an electrical circuitry, one should be able to build an electrical brain just by scaling up. While there are small successes in this effort, we haven’t seen a computer that thinks like us, remembers (and forgets) like us, or does other manly things. The challenge is a multi-facet one.

The first difficulty has something to do with the scale. While one can program a small neuronal network to perform certain specific tasks, simulating a brain that contains 100 billion neurons that each send out thousands of connections to others is not a simple task. At some point, changes in quantity might lead to a change in quality. 

The second difficulty one has to deal with is the massive complexity of the brain. Unlike a man-made electronic device built with standardized elements such as transistors, neurons are all different in shape, size, connections, and the collection of ion channels in them. To make it even harder to deal with, all these properties are dynamically changing at the time scales of millisecond to lifetime. Dynamic changes in synaptic strength mentioned above is one example.

The third difficulty is our limited knowledge about the brain and its building blocks. Take ion channels as an example. The international collaborative efforts of the human genome project seemed to have already given us the genetic blueprint of human life (the biological life). Yet, the ion channels responsible for sour taste were just discovered last year! This new finding expanded the size of 100+ human genes that we thought to encode all ion channels. With different gene products come together in a mix-and-match way to assemble various ion channels, we have a large number of ion channels. Many of them remain rather foreign to us. A lot needs to be done to even understand these building blocks of an excitable cell like a neuron.

Fourth, chemical synapses and their plasticity are an additional great challenge. As elucidated above, there are many types of neurotransmitters (small molecules vs peptides) and there are multiple types of receptors for each transmitter (ionotropic vs metabotropic). The brain also contains specific neurons that release neuromodulators (such as acetylcholine, serotonin, dopamine, histamine, norepinephrine, etc). These neurons are typically clustered in specific areas but project widely to many brain regions and thus influence neural activity globally (ness, attention, etc). A single neuron receives hundreds to thousands of s at its synapses and each synapse (its efficacy and strength) may be modulated differentially depending on the internal and external state and prior experience. Accurate modeling of even a single neuron can be a daunting task!  

And then there comes the philosophical debate of whether the final goal of fully understanding the brain is even achievable. Can we trust the brain to understand itself?! One may say that there must be some fundamental principles that govern the operation of a brain, while the above are all just trivial details. Indeed, reductionist approaches have dominated biology in every branch and has been fabulously successful. Many of the discussions that we have here are due to findings made with reductionist approaches. However, can such an approach solve the most fundamental problem such as “what is life” “what is consciousness”? After all, we rely on our brain and our senses to try to figure it all out.

Philosophical struggles and amusements aside, dramatic improvements in the quality of human life do come hand-in-hand with advances in biomedical research; and biomedical research is advancing at a rapid pace. We at least start to understand many important components of the brain and how it works. For example, we have heard from Dr. Feng about how we can now better manage the risks of stroke, and how to treat and cope with it when it does happen. A number of diseases or symptoms are caused mainly by ion channel defects, e.g. epilepsy, pain, and some heart diseases. There are many effective Western/modern drugs targeting specific ion channels that bring miracles to our ability to manage these and other diseases. There are also diseases can be effectively controlled by neurotransmitter based therapies. Some ingredients in traditional Chinese medicine might be able to do the same in some cases. We can bet that as we age even better therapeutic startegies and treatments will become available due to advancements in neuroscience research.

In this part, we emphasized that the nervous system uses both electrical (e.g. firing rate) and chemical information (e.g. the type of neurotransmitter). While both are necessary for the brain to send information from one place to another, the exact content of the information is not all encoded in electrical or chemical signals. We argue that the exact content of the information is often based on anatomical coding, i.e., how different parts of the brain are wired and how they are connected to different parts of the body. Neurons that process information with similar electrical and chemical signals at the cellular level can have completely different functions if they are localized to different brain regions or wired differently. This will be the central topic in Part III.  

Part III) Division of labor and localization of functions in brain regions and pathways. 

 
Proper assessments and appropriate responses to the ever-changing world are essential for the survival of any organism. To simply put, the brain does one thing and does it really well: it receives sensory s and sends out motor output. During this process, the brain takes into account of the external and internal state as well as prior experience.

Major divisions of the human brain
For centuries, people (from phrenologists to neuroscientists) have aimed to find out what different parts of the brain do. Based on what we know currently, the brain can be divided into hundreds of distinct regions based on anatomy, molecular and cellular properties, connectivity, and functions. Fig.3-1 shows a simplified functional division of the cerebral cortex. There are dedicated regions for receiving sensory s (primary visual, auditory, olfactory, gustatory and somatosensory cortices as well as other higher-order integrated areas) and for sending out motor outputs (primary motor cortex and higher-order motor-related areas). In stroke patients (refer to Dr. Feng’s lectures), neurologists can often judge which blood vessels in the brain or spinal cord are blocked or bleeding depending on the specific functional loss. What is amazing about the brain is that these distinct functional regions are interconnected allowing information processing in a parallel and distributed manner. (This is kind of analogous to the internet: if you think about how our lives have been changed so dramatically in the last 20 years, it is not only because individual devices become more powerful but mainly because these devices are all connected).    

Penfield sensory and motor homunculus
Dr. Wilder Penfield and his colleagues were among the first neuroscientists to map and differentiate sensory and motor functions with the concept of homunculi (meaning small humans). They used electrical stimulation of different brain regions of patients undergoing open brain surgery and mapped the somatotopic sensory and motor function (Fig. 3-2). Modern non-invasive methods such as functional MRI allow mapping of brain functions in normal subjects. Note that the brain representation is not proportional to the size of each body part; e.g., compared to trunk and arms, fingers have much larger representations which correspond to high densities of sensory receptors and high sensory acuity.

Sensory systems

All species have evolved apparatuses for detecting various sensory cues that signal food, mate, and danger. These apparatuses can be as simple as transmembrane molecules in unicellular organism or as sophisticated as sensory organs in multicellular animals. Humans use the eye, ear, nose, tongue and skin/muscle/joint to inform the brain about light, sound/balance, smell, taste, and somatic sensation, respectively. All sensory systems follow a similar design and the information detected by each peripheral sensory organ is transmitted into the primary sensory cortex through dedicated pathways (labeled lines). For examples, the light detected by the photoreceptors in the retina is carried to the primary visual cortex through four synapses: 

photoreceptorsbipolar cellsganglion cells (the first two synapses within the retina; axons from ganglion cells form the optical nerve)lateral geniculate nucleus in the thalamusprimary visual cortex (“” represents synaptic connection) (Fig. 3-3). Similarly, the auditory system carries sound information from the ear (hair cells detect the sound) through a pair of cranial nerves (there are 12 pairs of cranial nerves serving different sensory and/or motor functions) to the brain stem and then to the thalamus relay stations before reaching the primary auditory cortex. The gustatory system carries the taste information through several cranial nerves to the brain stem and then to the thalamus relay station before reaching the primary taste cortex. The olfactory system carries the smell information from the nose to the olfactory bulb and then to the olfactory cortex. The most complex sensory systems are the somatosensory pathways that carry the touch/proprioception/vibration and temperature/pain/itch from the skin/muscle/joint to the primary sensory cortex through three synapses (Fig. 3-4). Pain is a special type of sensation when the sensory stimulus is too intense and causes tissue damages. Pain research was especially popular in China in the 1970s due to the government’s interest in acupuncture. In a way it saved neurophysiology research in China when most research labs were shut down at that time. 

We would like to emphasize a few features about our sensory systems. 

(1) Although distinct pathways carry different sensory information to the brain, the information is integrated in the brain, typically in higher-order cortical regions. For example, when it comes to cooking, people pay attention to color, odor, taste, and temperature/texture -- the brain simultaneously integrates visual, olfactory, gustatory and somatosensory s. You can add music (auditory system) to enhance your experience!

(2) These pathways do not simply relay the sensory information to the brain. In every step along the pathway, the signal is highly modulated depending on the brain state (e.g. ness and attention) and prior experience (whether the signal carries positive or negative value). To certain extent, the brain constantly “screens” and “selects” the sensory cues that matter the most. Learning and memory play a critical role in sensory perception – think about how easily you can recognize a close friend or family member with a simple glimpse from afar.

(3) Our sensory organs are not passive receivers of the external world. Instead, assessment of the environment often involves active sampling and exploration via motions. For instance, sniffing is an active process to bring odor molecules into the nose. Saccade, a rapid movement of the eye between fixation points, actively positions the most salient object to the part of retina with the highest acuity. In the somatosensory system, perception is often achieved via active exploration of surfaces of objects by moving our fingertips. Generally speaking, active sampling of the external stimuli is an essential, integral part of sensory perception.

(4) In the sensory systems, information flow mostly goes bottom up from peripheral sensory organ to the relay station (typically in the thalamus) and to the sensory cortex (ascending hierarchical sensory pathways); at each stage more abstract information is derived from the earlier stages. For example, the retina has information about individual dots, light intensity, color and motion direction of dots. The thalamus encodes binocular information as well as line orientation etc. The visual cortex encodes angles, shapes etc. Higher order visual cortex encodes 3D objects and movement of objects. Even higher order visual cortex encodes information about object categories etc. It is amazing how the brain is able to extract abstract sensory information. Think about our ability to quickly detect one Chinese character out of hundreds of other letters and characters in one visual field. Or think about a moving object. It is seen from different angles and different distances; completely different cells in the retina and in the brain are stimulated or inhibited. However, we still perceive the same object. Equally amazing, a few dots inside a circle, that only represent a small change in early stages of visual information processing, can change the perception of a circle to a face.

The voluntary motor system

(1) In contrast to sensory systems, information flow in the voluntary motor system is mostly top-down in the “premotor cortex  motor cortex  spinal cord  muscles” descending hierarchical motor pathway. At each stage less abstract information is derived from the earlier stages. The premotor cortex encodes motor plans and goals. They are especially important for internally generated movements. A motor plan is represented in an abstract form rather than as a series of joint motions or muscle contractions. Just like the amazing ability of the sensory systems in extracting abstract sensory information, the premotor cortex is able to send out abstract motor commands. For example, if you have learned to type on a big computer keyboard using both hands, you will find that some of that skills can be transferred to typing on a small cell phone using only your two thumbs. 

The primary motor cortex is closely connected to sensory s and it encodes spatial and temporal patterns of joint angles and muscle activations. For example, the direction of movement of your hand during reaching is encoded in the primary motor cortex by the pattern of activity in a population of cells. Such studies are central in applications of brain machine interface (e.g. to control prosthetic arm movement). 

 
The motor neurons in the spinal cord receive direct sensory s and directly control individual muscles. Fig. 3-5 shows such descending motor pathways. An example of how muscles are controlled by spinal cord motor neurons are shown in a reflex in Fig 1-2. Similar to motor neurons in the spinal cord, motor neurons in the brainstem mostly control muscles in the head and neck.

One important advantage of such a hierarchical system is that the variety of reflex circuits and rhythmic movement circuits in the spinal cord or brainstem can simplify the computations of the motor cortex. By facilitating some spinal cord (or brainstem) circuits and inhibiting others, neurons in the motor cortex can let sensory s at lower levels govern the temporal details of an evolving movement without duplicating those computations carried out by spinal cord (or brainstem) circuits.  

(2) As shown in above Fig. 3-2, one important feature of the primary motor cortex is its somatotopic organization so that different body parts are represented by different primary motor cortical areas. Note that body parts that have better fine motor control occupy more brain areas. Similar somatotopic organization can also be found in the next stage, the spinal cord. In addition to limb movement, body posture is also controlled by the voluntary motor system although it uses a different descending motor pathway. The lateral corticospinal tract shown in Fig. 3-5 is the main pathway for limb movement control while the ventral corticospinal tract is the main one for postural control.

(3) In contrast to reflex, voluntary movements are highly flexible and adaptable—they improve in speed and accuracy with repeated practice. This adaptability may reflect an optimization process in which specific circuits needed to accomplish a behavior are, with training, selected from redundant sensorimotor connections. Such motor learning is made possible by synaptic strength changes in every stage of the descending motor pathway. In addition, the cerebellum and the basal ganglia play special roles in motor learning.

The cerebellum and supervised learning

The cerebellum is outside of the main “premotor cortex - motor cortex - spinal cord - muscles” descending motor pathway. The cerebellum receives from 1) regions of the cerebral cortex that plan and initiate complex movements; 2) somatosensory and motor s that carry information about the actual movements. This arrangement enables an instant comparison of an intended movement with the actual movement and a reduction in the difference (“motor error” correction). Therefore, the cerebellum is important for supervised learning that modulates voluntary motor responses in real time. Impaired cerebellum function affects balance and reaching precision etc. (e.g. in ataxia). However, the cerebellum is not essential for survival. 

The cerebellum generates intense interests from neuroscientists partly because of its beautiful anatomy and special computational properties for supervised learning. Fig. 3-6 shows the organization of different neurons in the cerebellum. Purkinje cells are the sole output of the cerebellum for modulating movement (error correction). Climbing fibers carry highly processes information about the actual movement and contact the Purkinje cells. Each Purkinje cell receive only one climbing fiber . The second major s to the Purkinje cells are parallel fibers that carry information about the intended movement. Each Purkinje cell receive s from thousands of parallel fibers. Information provided by the climbing fibers modulates the strength of parallel fiber-Purkinje cell synapses, therefore corrects errors in intended movement.

The basal ganglia and reinforcement learning 

Like the cerebellum, the basal ganglia (BG) is not part of the main “premotor cortex - motor cortex - spinal cord - muscles” descending motor pathway. Instead, it forms “cortex - BG - thalamus - cortex” feedback loops to modulate cortex activity, which will select the appropriate responses and inhibit the inappropriate responses through reinforcement learning. Reinforcement learning is dependent on the special neurotransmitter dopamine. Dopamine modulates the activities of neurons in the striatum (part of the BG) and synaptic strength of the synapses formed between cortical neurons and striatal neurons (corticostriatal synapses). Here is a simple way to think about reinforcement learning: if you do something under a specific environmental condition and is unexpectedly rewarded, then there will be increased dopamine release, and what you have done will be reinforced, i.e., you will do more of the same in the future. Those actions can even become habits if repeated many times. Habits are important in freeing up the brain’s limited online computational capacity. On the other hand, if you do something and there is no expected reward, then there will be decreased dopamine release, and what you have done will be inhibited in the future. 

A more computational view of dopamine is that it represents positive “prediction error”. When the reward is more than you expected (positive “prediction error”), then there will be increased dopamine release to teach the neural network to adjust future predictions of reward. When the reward is less than you expected (negative “prediction error”), then there will be decreased dopamine release to teach the neural network to adjust future predictions of reward. Another important feature of reinforcement learning is exploration (exploring the unknown) versus exploitation (exploiting what you have learned). Exploitation ensures that what you have learned is used to improve your performance. Exploration ensures that you will always discover new opportunities and to maximize your reward. Genetics may make some people more biased towards exploration while others towards exploitation. 

Parkinson’s disease (PD) is caused by midbrain substantia nigra dopamine neuron degeneration (Fig. 3-7). Without dopamine, there will be excessive inhibition of almost all responses. Patients will experience difficulty to initiate movements. PD patients also develop resting tremor which is thought to be caused by oscillations of activities in the feedback loops mentioned above when there’s not enough dopamine modulation. In contrast to PD, Huntington’s disease (HD, also called Huntington’s chorea) is caused by degeneration of the striatum. HD patients develop jerky, random, and uncontrollable movements, further suggesting that the striatum is very important in inhibiting inappropriate responses.

Dopamine, ventral basal ganglia loop, reward and motivation

The “cortex - BG - thalamus - cortex” loops are not only important in motor functions; it is also important in motivation and emotion. There are multiple parallel BG loops that perform similar computations and modulated by reinforcement learning signal from dopamine. However, they have different functions depending on which cortical areas they are connected to. The dorsal loop that we discussed above is connected to the motor cortex and is important in motor learning; another loop is the ventral loop that is connected to part of the prefrontal cortex important for processing value and emotional information etc. The ventral loop therefore is important for motivation and emotion. Here is a simple way to think about the ventral loop in the BG, reinforcement learning and motivation: if you do something under a specific environmental condition and is unexpectedly rewarded (positive prediction error), then there will be increased dopamine release, and you will be more motivated/energized to do the same in the future. On the other hand, if you do something and there is no expected reward (negative prediction error), then you will be less motivated to do the same in the future. 

Too much negative prediction error signal is thought to be a mechanism for depression (only one of the many theories for depression). If whatever you do does not lead to the expected reward or removal of a stressor, then you will not be motivated to do anything (called learned helplessness). 

Anatomically, the part of the BG in the ventral loop is the ventral striatum which is different from the dorsal striatum discussed above. While the dorsal striatum is modulated by dopamine from the substantia nigra, the ventral striatum is modulated by dopamine from the ventral tegmental area of Tsai (VTA). The VTA sits next to the substantia nigra in the midbrain. It was named after Chiao Tsai (蔡翘) because it was discovered by him (published in 1925) at the University of Chicago. Tsai received his Ph.D. in Psychology from the University of Chicago in 1924 (with Harvey Carr). Discovering VTA was his side project. 

The VTA dopamine neuron - ventral striatum pathway is part of the “reward center”. The reward center was discovered by brain self-stimulation studies. In those studies, an electrode was placed into the brain of a rat and the rat was trained to press a lever to receive self-stimulation. Some parts of the brain including the VTA  ventral striatum pathway are called the reward center because rats would prefer to self-stimulate those brain areas rather than eating, drinking etc. They could even self-stimulate the reward center until dying from starvation. This is a typical example of addiction, i.e., artificial activation of the reward center hijacks the system evolved to reinforce behaviors that are supposed to be adaptive and increase fitness. What’s interesting is that all addictive drugs either directly or indirectly activate the VTA  ventral striatum reward pathway. For example, cocaine blocks clearance of released dopamine so that more dopamine will stay in the synapse. Interestingly, animals can also be trained to self-administer cocaine (or other addictive drugs) just like in brain self-stimulation.   

Sensing and maintaining internal environment (homeostasis)

The kind of motivation discussed above is often called incentive motivation because it is motivation acquired through reinforcement learning. Incentive motivation is different from “drive”. “Drive” is motivation caused by imbalance of your internal environment and you are motivated to do things to fix that imbalance (e.g. hunger makes you motivated to eat; lack of sleep makes you motivated to sleep; coldness makes you motivated to put on clothes etc.). The part of the brain that senses our internal environment and underlies the drive to fix imbalance is the hypothalamus (Fig. 3-8). Different areas of the hypothalamus sense our body temperature, energy balance, etc.; and it controls thermogenesis, feeding (therefore obesity is often a problem in the brain), circadian rhythm, stress responses, reproduction among others. The hypothalamus is closely connected to the ventral striatum discussed above. 

The hypothalamus also controls a lot of the endocrine functions through its projection to the pituitary. Therefore, many of the hormones (e.g. stress hormones, sex hormones, growth hormone etc.) are controlled by the hypothalamus. The endocrine system is not part of the nervous system, but it is largely controlled by the nervous system and it contributes significantly to the balance of our internal environment as well.

Another part of the nervous system that controls internal balance is the autonomic motor system. It is so called because we have no volitional control over our autonomic motor system (although some people claim that they can through practicing meditation or “qi”). The voluntary motor system that we discussed above controls skeletal muscles while the autonomic motor system controls the cardiac muscle and smooth muscles. Smooth muscles control the movement of our internal organs etc. such as gut movement, blood vessel dilation and constriction, perspiration, hormone secretion etc. 
 

There are two divisions in the autonomic motor system. The sympathetic division generally prepares the individual for “flight or fight” (largely true but not always). It increases heart rate, respiration, secretion of adrenaline. It dilates skeletal muscle blood vessels. It inhibits stomach and gut activities. The parasympathetic division does the opposite and it generally facilitates “rest and digest” (largely true but not always).
One special type of internal homeostasis is sleep. Sleep is controlled by both the biological clock (circadian rhythm control) and the drive to sleep when one is awake for too long (homeostatic control). 

In summary, the nervous system maintains relatively stable internal environment through the voluntary motor system (e.g. putting on more clothes to keep warm), autonomic motor system (e.g. blood vessel constriction to prevent losing heat) and its control over the endocrine system (e.g. promoting the secretion of the thyroid hormone to increase thermogenesis).

Emotions are adaptive responses, emotions are for communications, emotions are conscious emotional feelings

Now let’s come back to the external environment again. The above discussions on the reward pathway mostly on appetitive learning and behaviors. What about aversive learning and behaviors? The brain area that is most important for aversive learning and behaviors is the amygdala which is closely connected to the ventral striatum. One example of aversive learning in animals is “fear conditioning”: if a mouse receives an electric shock (the unconditioned stimulus) every time it hears a tone (the conditioned stimulus), then next time a tone can cause the mouse to freeze (an aversive response). In humans, PTSD is believed to be caused by the generalization of conditioning to stimuli that are somewhat similar to the specific conditioned stimulus such that the somewhat similar stimuli can also cause aversive responses. General anxiety is thought to be caused by the generalization of conditioning to the environmental context somewhat similar to the environmental context associated with a stressor. 

 
Appetitive or aversive responses are part of our emotional responses. Emotional responses are essential for our survival. For example, babies cry to get the cares they need. We gather all the strength we have to fight or escape when we are in danger. The systems that we discussed above, voluntary motor system, autonomic motor system and the endocrine system are all important components of emotional responses. The autonomic motor system and the endocrine system are important in causing physiological responses (e.g. increased heartbeat). In fact, an important component of emotion is our conscious awareness of the physiological responses (e.g. increased heartbeat) during emotional responses. There are many experimental data indicating that pure cognitive assessment of our environment without the physiological responses is not sufficient to cause strong emotions. 

Emotional responses not only give us the courage and energy to cope with stressful situations, they are also important components of our communications with others. Baby crying is one example. Happy emotions help to make friends, etc. Angry emotions may also scare off enemies (but it could be bad too; we all know the story of 呆若木鸡). Emotions also help us with our judgements. We all have heard such expressions: “Do you follow your head or your heart?”, “My gut feeling”, etc. Plato famously called emotion as our lower passions (as oppose to reasoning). He was obviously wrong. Overall emotions are good for us and are essential for our survival. However, emotional overreactions are indeed bad for communications and bad for our judgements (either in normal people or in certain diseases prone to emotional overreactions). Maybe that’s how emotion gets the bad name since Plato’s time.

We not only have emotional responses; we are also capable of imagining our future emotions when we make plans for future events. Moreover, we are capable of understanding and imagining other people’s emotions when we interact with others. This is a very important part of our social skills (or E.Q.), which is often impaired in autism. 

The hippocampus, learning and memory

We will devote the last section of Part III to memory. It is probably the topic that neuroscientists are most obsessed with. Perception, motor skills, reward, emotions, etc., all those functions that we discussed above require memory. However, most of those are called implicit memories. For example, after we have learned a motor skill, we cannot “recall” it, yet we can perform it. Here we will mostly discuss explicit memory.

We will start with the famous patient Henry Molaison (mostly known as H.M.;1926 – 2008). It’s fair to say that H.M. contributed to neuroscience more than most neuroscientists. H.M. had severe epilepsy as a young man and the neurosurgeon removed the brain region called hippocampus (Fig. 3-9). The surgery cured his epilepsy. However, H.M. was no longer able to transfer any new short-term memory (which only lasts some seconds) into long-term memory (anterograde amnesia). That means he did not remember anything he did or learned even a few minutes earlier. He only remembered his past before the surgery. He forever remembered himself as a young man until his death. Despite complete loss of his ability to consolidate long-term explicit memory, H.M. was perfect in forming new long-term motor memory. He could learn new motor skills just like us yet he could not recall his training sessions. 

In another case with hippocampus lesion and similar amnesia, one day the doctor put a tack in his hand, walked in, shook the patient’s hand and pricked it. The doctor then walked out of the room, walked back in, tried to shake the patient’s hand again. The patient refused although the patient couldn’t tell why. This is another example that the patient had implicit memory but didn’t have the explicit memory of the experience/episode. That means that the hippocampus is required for the consolidation of new short-term explicit memory into long-term explicit memory, but it is not required in implicit memory. H.M.’s ability to recall long-term memories that existed before his surgery also suggests that the hippocampus is not required for storage or retri of long-term explicit memory. We don’t have a video of H.M. but here is a patient who has very similar symptoms.

 https://www.youtube.com/watch?v=Vwigmktix2Y

So where are memory traces stored? The predominant hypothesis is that information stored as long-term explicit memory is first processed by various cortical areas. It is then processed by the hippocampus. The hippocampus mediates the initial steps of long-term storage. It slowly transfers information back into various cortical areas. Long-term explicit memory is eventually stored in those cortical areas.

In animal studies, the best evidence for the role of hippocampus in explicit memory are: 1) Lesioning the hippocampus impairs spatial memory. 2) In electrophysiological recoding studies, many neurons in the hippocampus of running rats or mice are place cells. Place cells are neurons that exhibit a high rate of firing whenever an animal is in a specific location in an environment. Ensembles of place cells are thought to form a “cognitive map”. This is probably the best evidence that animals have representations of the external world in their brains. 3) The strengthening (synaptic long-term potentiation, LTP) and weakening of specific synapses (long-term depression, LTD) in the hippocampus are correlated with learning and memory. LTP and LTD are often viewed as the cellular level mechanisms of learning. 

Evidence for the role of hippocampus in explicit memory from humans include H.M. and many other patients. In Alzheimer’s disease, the loss of memory is caused by the degeneration of the hippocampus and nearby brain regions. 

 
Even before the discovery of LTP and LTD, Donald Hebb (1904–1985) proposed the Hebbian learning rule as the cellular basis of learning and memory: “When an axon of cell A is near enough to excite cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A's efficiency, as one of the cells firing B, is increased.” The discovery of LTP in the hippocampus was the first demonstration by experimental data of Hebb’s theory on the cellular basis of memory. One neurotransmitter receptor that plays a central role in Hebbian learning is the NMDA receptor (for the excitatory neurotransmitter glutamate). The NMDA receptor is regarded as the "molecular coincidence detector" because it is only activated when “cell A” and “cell B” fire together. NMDA receptor activation in turn facilitates the downstream processes that make “cell A” and “cell B” wire together.

One brain area that we didn’t discuss is the prefrontal cortex. The prefrontal cortex is believed to be most responsible for the differences between a human brain and an animal brain. It is important for almost all high cognitive functions: working memory, attention, decision-making, planning, impulse control, personality and many others. Because of that, it’s specific pathways and mechanisms underlying specific functions are poorly understood. Studies of the prefrontal cortex also often suffer from the implicit assumption that there is a little human in the prefrontal cortex to perform all these complex functions for the prefrontal cortex without explaining the real neural mechanisms underlying the specific functions.

Information flow at the systems level

When we talk about the central nervous system, we often mean the brain. However, the central nervous system includes both the brain and the spinal cord (Fig. 3-10). The spinal cord can be viewed as an extension of the brain stem. The central nervous system is connected to sensory organs and muscles through the peripheral nervous system. The peripheral nervous system is made of nerve bundles formed by groups of axons extended from motor neurons in the spinal cord or brain stem into muscles as well as nerve bundles formed by groups of dendrites (these are specialized dendrites that look like axons) extended from sensory neurons in the spinal cord or brain stem into sensory organs. 

Therefore at the systems level, sensory information from sensory organs (e.g. the skin) enters the brain and the spinal cord through sensory nerves of the peripheral nervous system. This is how sensory organs are connected to the brain and the spinal cord (for vision, audition, olfaction etc., special nerve bundles rather than sensory nerves of the peripheral nervous system are used). Such information is processed by the brain and the spinal cord. After sensorimotor integration, the brain and the spinal cord send motor control information to muscles through motor nerves of the peripheral nervous system that connects the brain stem and the spinal cord to muscles.

To conclude Part III, we can revisit Part I in which we discussed decision making and flexible goal-directed behaviors. After Part II and Part III discussion, we can think about such complex behaviors again. It is obvious that our sensory systems, motor systems, memory systems, and even emotions are all involved. Different parts of the brain have to work together seamlessly for such complex tasks. We know a lot now how each part of the brain works, but we still know very little how different parts of the brain work together. A good place to start with is probably sensorimotor integration but neuroscience still has a long long way to go.

In Part III, we have discussed some diseases. The specific symptoms of various diseases are closely related to the affected brain regions/pathways and their associated functions discussed in Part III. In fact, damages to many different brain regions caused by developmental impairments, neurodegeneration, strokes, infections, injuries or surgeries are often the most important experiments nature did to help us understand functions of specific brain regions. Some of the symptoms could be extremely unique and interesting (e.g. impaired ability to recognize faces) and give us important insights about nervous system functions and mechanisms. We will talk more about nervous system diseases in Part IV.
Part IV) Diseases of the nervous system 

Nervous system diseases are caused by many different factors. The following are the most important causes and they also help us to classify nervous system diseases. However, they have one thing in common, they all indicate that nervous system diseases are especially difficult to treat.

1) Developmental diseases. There are many cell types in the nervous system, and their differentiation into the final cell type from stem cells is a complex and orchestrated process during development. Neurons have to migrate a distance to their final destination; axons have to travel a distance to synapse onto very specific targets. Therefore, the correct wiring of the nervous system during development is almost like a miracle; and mistakes do happen that could lead to diseases. Some diseases are caused by severely impaired development of the brain (e.g. Rett syndrome); some are caused by miswiring of the brain (e.g. autism, schizophrenia). These mistakes cannot be corrected after development; treatment can only on symptoms. 

2) Neurodegeneration. Almost all neurons cannot regenerate (not like the liver or skin). This may be necessary for the brain because memories are stored in the brain; regeneration and new connections would erase old memories. On the other hand, degeneration (e.g. in Alzheimer’s, Parkinson’s) (see Fig. 4-1). The brain is the most active organ metabolically (~3% of the total weight but costs 20% of the total energy) does happen for multiple reasons. Neurons are very sensitive to oxidative stress and mitochondria impairments. 

Neurodegeneration can be caused by genetic defects in mitochondria function or by toxins that impair mitochondria function. Many neurons work very hard almost all the time, and they have to make a lot of new proteins. Mistakes happen very often in making new proteins. If bad proteins cannot be degraded quickly, they can kill neurons; which is part of the reason in neurodegeneration. When degeneration happens, those neurons cannot be replaced. Even if we can transplant new neurons, they will not be able to wire correctly because wiring is a complex and orchestrated process during development. Therefore, cell death cannot be reversed in the nervous system; treatment can only on symptoms. Age is the biggest risk factor for neurodegeneration for at least two reasons. First, neurons in old people are more prone to degeneration simply because their mitochondria function, protein degradation pathway etc. are not as good as young neurons. Second, cell loss in neurodegeneration accumulates with age. For example, every one of us are losing dopamine neurons continuously. We don’t have Parkinson’s disease only because cell loss accumulation has not reached the threshold to cause symptoms (See Fig. 4-2) yet. Hypothetically, if we could live long enough, everyone would develop Parkinson’s disease sooner or later. 

3) Stroke and injury. Almost all neurons cannot regenerate. Therefore, neuron death due to stroke (the brain is very sensitive to the lack of oxygen when there is no blood supply for some time), physical trauma injury or infection cannot be replaced. The specific symptoms are closely related to the affected brain regions/pathways.

4) Functional impairments in neuronal activity. Activities of neurons (both electrical and chemical activities) have to be very precise to carry on their precise roles in transmitting specific information. Any defects in ion channels or neurotransmitter signaling pathways may cause major problems. Some nervous system diseases are caused by malfunction of ion channels (e.g., some types of epilepsy, ataxia, migraines etc.); some are caused by malfunction of neurotransmitter synthesis, neurotransmitter clearance, neurotransmitter receptors, or receptor signaling (e.g. some types of epilepsy, addiction, some types of dystonia, etc.). Using epilepsy as an example. The brain is very vulnerable to epilepsy. 
Close to 1% of the population is affected by epilepsy. Even a normal brain has synchronized activities from time to time; and there are positive feedback loops in the brain as well. Therefore, a normal brain is not that far from developing seizures. Most of us don’t have seizures because the brain is wired in such a way that there are inhibitory neurons everywhere. Where, when and how much excitation versus inhibition is critical to normal brain function. Any mistake due to miswiring in development, in ion channel function, in excitatory neurotransmitter or inhibitory neurotransmitter function can potentially lead to seizures. On the other hand, the electrical and chemical nature of the nervous system also offers opportunities for treatment (e.g. Ion channel blocking drugs for epilepsy; deep brain stimulation to reduce motor inhibition in treating Parkinson’s disease (Fig. 3-7); L-DOPA, a dopamine precursor, to boost dopamine production in treating Parkinson’s disease).

5) Unknown causes, but environmental factors are important. The brain is the most complex organ and our knowledge is still very limited. Many nervous system diseases are associated with functional changes in the brain, but the exact functional changes and causes are not well understood (e.g. depression, anxiety etc.). 

Environmental factors are especially prominent in nervous system diseases simply because the nervous system is all about our body interacts with the environment and responds to environmental changes. This makes the underlying causes of many psychiatric disorders very difficult to understand and therefore making treatment hard. Even during treatment, we interact with the same environment on a daily basis, which also makes effective treatment hard. On the other hand, the environmental factor offers opportunities for treatment because some of the disorders may be treated without medication (e.g. using behavioral therapy).

Adding to the difficulties discussed above, drugs that treat nervous system diseases are also much harder to develop than drugs that treat other diseases. Screening drugs on cultured cells is often a crucial step in drug development. However, you cannot screen depression drugs using cultured cells, at least not until we understand the cellular mechanisms of depression. It is even difficult to test drugs in animal models; you cannot ask the animal if it is depressed or not.

Clinically, nervous system diseases are often classified into either neurological disorders or psychiatric disorders. Therefore patients are treated by different physicians (neurologists versus psychiatrists). Neurological disorders usually have known structural damages (including neurovascular system) and/or molecular defects. Their symptoms are usually well defined such as impaired specific sensory function (e.g. blindness caused by visual cortex, migraines), specific motor function (e.g. Parkinson’s disease), language (some types of stroke), memory (amnesia), or consciousness (coma).

Psychiatric disorders (sometimes commonly known as mental disorders) usually have no obvious structural damages. Their symptoms are usually more related to emotions (e.g. depression), reasoning (e.g. schizophrenia) and personality. Because of the lack of obvious structural damages, for a long time (even now in some parts of the world) psychiatric disorders are considered to be mental weakness rather than real diseases; patients are often discriminated against. Therefore, it is even more important to emphasize that psychiatric patients do have impaired brain functions, and that most are caused by the combination of genetic and environmental factors (see Part V below). Their illnesses are not caused by “weak mind”. 

The boundary between neurological and psychiatric disorders is not clear though. Many neurological disorders have severe psychiatric symptoms, e.g. Parkinson’s disease patients and Alzheimer’s disease patients often have mood problems. On the other hand, sometimes stress can cause movement disorders (psychogenic movement disorders). Many diseases have both neurological and psychiatric symptoms (e.g. autism, ADHD, OCD etc.).

It’s important to point out that anxiety and depression are extremely common in a modern society. Drugs for treating these disorders (e.g. Prozac) are the second or third most prescribed drugs in the developed countries. Because drugs like Prozac are selective serotonin reuptake inhibitors (SSRIs), the serotonin hypothesis for depression and anxiety is very popular. However the hypothesis may not be correct. People respond to a serotonin drug treatment does not mean that their disease is caused by impairments in the serotonin system. Overall psychiatric disorders are very difficult to understand. Hopefully, some of the clues from genetic studies will help.

Most nervous system diseases (whether due to development, degeneration, functional changes or unknown causes) can be attributed to both genetic and environmental factors. This is another reason that we should not discriminate against psychiatric patients. Their illnesses are not caused by “weak mind”; they really have no choice in getting or not getting the disease. However, a healthy life style and low stress always helps in taking care of the environmental factors. The genetic contribution to nervous system diseases is discussed in Part V.

Part V) Genetic diseases and personality traits

The genetic contribution to diseases is increasingly recognized and studied with precision since the completion of the Human Genome Project. 

A gene is the basic physical and functional unit of heredity. Usually it refers to a length of DNA that codes for a specific protein. The human genome has about 30,000 genes and 3 billion base pairs. However, only ~25% of the DNA is related to genes (exon sequences as well as intron and regulatory sequences). Only ~10% of the this 25% (so ~2.5% of the genome) is used in encoding proteins (exon sequences). Protein encoding information is within exon sequences in which the bases (A, T, G, and C) are arranged. The code is written in triplets. For example, GCT codes for the amino acid alanine. Thus, the sequence of amino acids in a protein is determined by the order of triplet base pairs in the gene for that protein.

DNA is packaged into chromosomes. Humans have 46 chromosomes (23 pairs) in every cell (sperm and egg are exceptions). In each pair, one is from father, another from mother. There are 22 pairs of nonsex (autosomal) chromosomes and one pair of sex chromosomes. The sex chromosomes determine whether a person is a boy (XY) or a girl (XX). Sperm or egg cells only have 23 chromosomes due to meiosis. Meiosis is a special type of cell division that reduces the chromosome number by half. In a fertilized egg, the 23 chromosomes from the sperm joins the other 23 in the egg so that an embryo has 23 pairs of chromosomes. 

Therefore, every person has two copies of each gene in each cell, one inherited from each parent (with some exceptions, e.g. genes on the X chromosome in males). Each gene is almost identical in all people. However, at some places the two copies of the same gene in one person could be different (different variants of the same gene). This is important because the chance of a person having two copies of the same abnormal variant (and hence a disorder) is usually small (unless the parents are close blood relatives). Moreover, different people could have different variants of the same gene. These small differences in gene variants contribute to each person’s unique physical features, personality, as well as susceptibility to diseases.

One general view is that some genetic variants alone can cause diseases (purely genetic and single gene mutation); other variants may cause diseases under certain environmental (gene-environment interaction, GxE) or genetic conditions (gene-gene interaction, GxG); yet most genetic variants represent a spectrum of genetic diversity in the human population. These variants may make us susceptible to certain diseases under certain environmental or genetic conditions, but they are also the underlying mechanism for the diversity in personality traits. That means some nervous system diseases can even be regarded as extreme examples in a continuous spectrum of personality traits. Sometimes too much of an adaptive trait (either due to genetic or environmental reasons) becomes maladaptive. 

 
Fig 5-1 illustrate the above in a semiquantitative way. On the leftmost are variants that can absolutely cause diseases. For example, the Huntington’s disease gene and mutation were discovered by Nancy Wexler and colleagues after following the genetics of a large family in Venezuela that is affected by Huntington’s disease. Wexler herself has a family history of Huntington’s disease (including her mother). In genetic diseases such as Huntington’s disease, the disease-causing variants are very rare in the population because they negatively affect fitness and are under negative selection pressure. Nevertheless, because many patients survive well past reproduction age, these diseases may still run in certain families. These variants can be discovered by linkage studies on affected families. Such studies are based on co-occurrence of the disease and known genetic markers; the co-occurrence is due to close linkage on chromosome between the known genetic markers and the to be discovered disease variant (low recombination probability during meiosis due to short distance on a chromosome). 

While Huntington’s disease is always caused by one type of mutation in this one gene and this mutation in this one gene can 100% cause Huntington’s disease (necessary and sufficient), it is a rare exception. Almost all nervous system diseases are usually caused by a collection of variants in many genes plus environmental factors (GxG plus GxE)(some exceptions will be discussed in the next paragraph). As shown on the rightmost side of Fig. 5-1, each variant only makes a very small contribution to disease susceptibility. These are called variants rather than mutations because they usually do not cause diseases by themselves. Whether a specific variant is good or bad for us may depend on the environmental condition. The same variant (e.g. a variant associated with aggression) may cause disease susceptibility under one condition but may increase fitness under another condition. Otherwise these variants will not survive selection in evolution and will not be so common. That means many of us have these disease susceptibility variants as well. The association of a specific variant in a specific gene to increased susceptibility to a specific disease is usually discovered in “association studies”. In the last decade, genome wide association studies (GWAS) are also becoming very affordable. Such studies look for increased frequency of specific variants in the disease population compared to the control population. However, such studies may have reached their limit since new discoveries increasingly require ever larger sample size. In addition, all genes are related to each other more or less due to gene networks. Therefore, all genes may contribute more or less to the susceptibility to a specific disease. All these factors contribute to the difficulty of association studies and reduce the significance of some of the discoveries made by association studies. 

While almost all nervous system diseases (e.g. depression, anxiety, schizophrenia, bipolar disorder, stroke, migraines etc.) belong to the category of GxE plus GxG, there are some exceptions. Huntington’s disease is one extreme example. Some diseases have both familial form (purely genetic) and sporadic form (GxE, GxG, or non-genetic). For example, there are different types of Alzheimer’s disease (AD). About 5-10% of AD patients are familial form and the rest are sporadic form. Familial AD is purely caused by one of the causal mutations. For example, one type of mutation in the APP gene can lead to overproduction of the Aβ42 peptide which is thought to cause AD (Fig. 5-2). The age of onset for familial AD is usually 40-60. Sporadic AD affects ~7% of people older than 65 and ~40 % of people over the age of 80. AD is predicted to affect 1 in 85 people globally by 2050, making it the most common neurodegenerative disease. Sporadic AD does not mean that there is no genetic contribution. As we discussed above, most sporadic AD is caused by GxE plus GxG. For example, the ApoE gene has at least three variants. About 14% of the general population has the ApoE4 variant. Although ApoE4 does not always causes AD, it more than doubles AD risk. People with the ApoE4 varaint on both chromosomes have 10X more increased risk for AD. 

 Similar to AD, 5-10% of Parkinson’s disease (PD) patients are familial PD while the majority are sporadic PD. Overall, PD affectcs 1% of the population over 60 and 4% of the population over 80. One environmental risk factor for PD is pesticides while smoking and coffee/tea drinking reduces PD risk. For ALS, it’s also true that a small percentage is familial ALS while the majority is sporadic ALS.

Different variants of a same gene are caused by mutations passed on from generation to generation as well as by new mutations (de novo mutations) because mistakes happen during DNA replication. Some de novo mutations can cause genetic diseases. In this case, the affected person will be the first one in the family. For example, almost all Rett syndrome cases are caused by de novo mutations simply because it is a dominant disease (that means there are no symptom-free carriers of the mutation) and the patient is almost never fertile. De novo mutations happen all the time in us and mutations accumulate through generations. Some may be passed on to the next generation and the next if they increase fitness (happens very rarely). Some may be eliminated in natural selection if it negatively affects fitness (e.g. the mutation that causes Rett syndrome). Some may stay in the population with a high or low frequency (increase or decrease fitness) depending on the environment as we discussed above. One important example is the thrifty gene hypothesis. According to this hypothesis, animals and ancient humans lived in environment with very limited food. They had accumulated gene mutations that made them eat as much as possible if there was food even if they were not hungry; and made them conserve energy if they didn’t have to move around much. Unfortunately we have inherited these thrifty genes and they make us love to overeat and sit around. 

 
Most de novo mutations are neutral variants in non-coding DNA, and they accumulate in the genome through generations. Variants that become so common that they affect more than 1% of a population are called polymorphisms. They may become useful in estimating the age of ancient DNA isolated from ancient humans or other organisms. In human evolution studies, such variants in mitochondria DNA (only inherit from mother) and in the Y chromosome (only inherit from father) are especially useful because they do not undergo recombination and they faithfully transmit from one generation to another. 

In very rare cases, a severe deleterious mutation may also introduce a change that is advantageous. For example, in the case of the sickle cell gene, when a person inherits two copies of the mutant gene, the person will develop sickle cell disease (recessive inheritance). However, when a person inherits only one copy of the mutant gene (called a carrier who will not have the disease), the person develops some protection against malaria. It is possible that similar phenomena exist in nervous system diseases. However, it has not been reported yet.

By understanding the genetics of diseases, it may become possible to treat diseases with gene therapy (e.g. using the efficient CRISPR/Cas9 for genome editing). However, as discussed in Part IV, somatic genome editing in the adult brain may not be able to effectively treat a genetic nervous system disease if the brain is already miswired.