Tuesday, October 21, 2008

Theory of Mind vs. Theory of Reality: The tradeoff along the Autism Spectrum

In any great organization it is far, far safer to be wrong with the majority than to be right alone. -- John Kenneth Galbraith

If the only tool you have is a hammer, you tend to see every problem as a nail.
-- Abraham Maslow

In the land of the blind, the one-eyed man is king.

How autism provides resistance to the delusional thinking of groupthink (aka drinking the Kool-Aid).

When your primary competition is with other people, life is a zero sum game. When your primary competition is with reality, there are no limits.

The cause(s) of Autism Spectrum Disorders (ASDs) remain unknown. The complex genetic disorder hypothesis posits ASDs are an emergent disorder of multiple genes, perhaps dozens or more. How a common disorder of so many genes could evolve has not been suggested. One hypothesis is that genes associated with ASDs did confer some advantage resulting in their selection in the past, but are now detrimental. I suggest ASDs are not disorders at all, but normal (even human defining) developmental responses, particularly to stress in utero and early childhood. This fundamental human neurodevelopmental paradigm programs the brain and behaviors mediated by that brain so as to optimize survival and reproduction depending on maternal and infant stress; invoking abilities to understand, interact with and manipulate other humans when times are good, and abilities to understand, interact with and manipulate the environment via tool production when times are hard. It is hypothesized that this trade-off between Theory of Mind (ToM) and Theory of Reality (ToR) is the quintessential trade-off along the Autism Spectrum. Implications for interaction difficulties between individuals with ASDs and Neurologically Typical individuals (NTs) are discussed. Prevention and treatment are also discussed.

Background ASDs

The Autism Spectrum Disorders (ASDs) are defined and diagnosed by difficulties in communication, social interaction and by repetitive behaviors. In ASDs, there is a high (but not absolute) concordance between monozygous twins, [1] moderate concordance between dizygous twins, [2], [3] and lesser concordance between siblings. With no generally accepted environmental cause, ASDs are thought to be primarily genetic in origin with associations of perhaps 135 genes.[4] No doubt the complex effects on brain structure and behavior observed in ASDs are not mediated via a single pathway, but calls to abandon a search for a single explanation are premature. [5]

A number of single mutations have been associated with multiple cases of autism-like symptoms. I call these "autism-like" because it is not clear if the cause and sequelae of these autism-like syndromes are identical to or even similar to the common cases of autism and ASDs (which remain unknown). A good example is Rett Syndrome which is known to be caused by a loss in the MeCP2 gene which is on the X chromosome. Females have 2 copies of the X chromosome, one of which is silenced. Active MeCP2 in some cells rescues the organisms from the fatal loss of MeCP2 which afflicts males (who have only one X chromosome). RS females develop seemingly normally until 6-18 months when they develop the characteristic RS phenotype which includes autism-like symptoms, but is also characterized by non-autistic symptoms of small head size, breathing abnormalities, vascular abnormalities, scoliosis, growth retardation and others. Because the effects of MeCP2 deletion are mediated through aberrant transcription of methylated DNA, then aberrant transcription of methylated DNA is sufficient to lead to autism-like symptoms. Perhaps the symptoms of common ASDs are also caused by DNA methylation, or perhaps via a shared final common pathway triggered by aberrant DNA methylation.

Nitric oxide is a pleiotropic signaling molecule used in thousands of metabolic pathways where it regulates, ATP supply, O2 consumption, steroid physiology, transcription, axon targeting, the cell cycle, epigenetic programming and many other aspects of physiology, development and neurodevelopment. Many of the pathways observed to be abnormal in ASDs are mediated through NO signaling.

I suggest that the final common pathway mediating autism and autism-like symptoms is low NO in utero, during neurodevelopment and as an adult. Low NO is the archetypal stress response. Virtually any type of stress ultimately results in low NO which physiology uses to trigger compensatory responses. For example, oxidative stress occurs when superoxide levels are increased. Superoxide consumes NO at near diffusion limited kinetics so a high superoxide state is necessarily a low NO state. NO inhibits cytochrome c oxidase. To release that inhibition and increase O2 consumption to maximize aerobic ATP production, the NO level must be lowered. Psychosocial stress increases oxidative stress through catecholamine oxidation, ATP stress increases oxidative stress through increased mitochondrial potential necessary to increase ATP flux, xenobiotic stress increases superoxide through the cytochrome P450 pathway, immune system stress increases superoxide through the respiratory burst. Just about any type of metabolic stress would decrease NO levels and according to the present hypothesis would tend to produce a more autistic-like phenotype. This may be the mechanism for the association of copy number variations with autism. This may also be the mechanism behind the ASD-like symptoms produced by MeCP2 deletion. Females with MeCP2 deletion are mosaic. Some of their cells do have appropriate methylation readout and some do not. Presumably this differential regulation of DNA expression causes metabolic inefficiencies and metabolic "stress" due to cells being out of "sync". In a mosaic organ with non-synchronous regulation, some cells would be working harder than others, perhaps even working at cross purposes. The final common pathway of essentially every kind of metabolic stress is decreased NO.

Virtually all ASD physical symptoms are consistent with pathways regulated by NO being skewed in a low basal NO direction. I suggest that the ASD phenotype is a stress compensatory pathway mediated by low NO.

NO/ROS Balance Programs adult physiology in utero

The physiology of virtually all adult organs is known to be programmed in utero in response to a number of fetal stressors including nutrition[6] stress and hormonal factors. [7] NO/ROS balance in utero does lead to epigenetic programming of adult blood pressure in rats. [8] Stress is a low NO state. [9] The most characteristic physical feature of ASD children is a larger brain[10], with smaller and more numerous minicolumns. [11] Low NO is suggested to cause the characteristic minicolumn structure associated with autism[12] and the timing of stressors may be crucial to the development of the autism phenotype. [13] In guinea pigs, brief prenatal stress increases brain/body mass ratio, and changes adult behavior. [14] Prenatal stress increases learning ability in rats. [15] Prenatal stress does program hypothalamo-pituitary-adrenal function. [16] Low NO does cause neuronal hyperplasia. [17] The patterning of many neural structures is determined in part by gradients in NO mediating proliferation, differentiation, or apoptosis. [18] There are increased asymmetries in the brains of ASD individuals[19], suggesting differential regulation of neuronal growth when the sizes of those structures are formed in utero. Stress in utero causes adaptive changes in the adult physiology of multiple organs; it would be beyond surprising if it did not exert adaptive influences on the most important organ, the brain.

I suggest that low NO in utero, brought about by maternal stress leads to the ASD phenotype in affected individuals, and the genotype that leads to the ASD phenotype was adaptive under conditions where humans evolved, in the “wild”, but is perhaps now less adaptive due to environmental change(s). What possible advantages could the ASD phenotype hold?

The Brain is fundamentally the most Human organ

Humans evolved large and complex brains only because such brains conferred survival and reproductive benefits. Human evolution was shaped mostly by events 100k or more years ago. Humans are social animals, as are all primates. Humans are unique in their use of language with syntax and grammar to convey complex ideas. Humans are the only extant hominin that manufactures and uses tools. The first instances of manufactured stone tools date to about 2.5 to 2.7 MYA (million years ago), and was near universal by 2 MYA[20]. Manufactured tools of perishable materials perhaps were earlier. Modern humans are good at tool manufacture and tool use. Tool use has profoundly shaped human evolution and those parts of the human genome that affect brain structures important for tool creation and use. Similarly, communication and language has profoundly shaped human evolution and those parts of the genome that affect brain structures important for language acquisition and use. The major structures of the brain are formed in utero and early childhood, and are then largely fixed throughout adult life. It is only in utero and early childhood that neurons can be epigenetically programmed to form the major structures in the brain with the characteristic neuroanatomy observed in ASDs such as increased asymmetries, larger numbers of neurons[21] and larger brains.

Brain size at birth limited by female pelvis: Brain optimization requires tradeoffs

The size of the newborn brain is limited by the size of the mother's pelvis through which it can be successfully born. In the absence of medical C-section, cephalopelvic disproportion results in significant infant and maternal mortality. What ever advantages a large brain at birth provides it comes to naught if the infant or mother dies. The structure and size of the brain at birth must be an evolved trade-off between the multiple tasks that brain will be called upon to perform at birth and over the life of the individual and the substantial risk during a natural birth. The only time the most fundamental aspects of brain structure can be modified is while those structures are being formed. Much of the formation of those brain structures occurs in utero, much of it in the first trimester following closure of the neural tube (which is when teratogens such as thalidomide can cause autism). Epigenetic programming of cells occurs when those cells are dividing and undergoing differentiation. Much of the differentiation in early human development occurs in the first trimester. Patterns of methylation modify DNA expression and modify the phenotype of that differentiated cell for the lifetime of that cell. Many neurons do not divide over the lifetime of the individual. NO does modify methylation through the folate pathway and so modifies DNA methylation in ways that are quite complex (for at least this one gene system). Presumably multiple genes are epigenetically modified in complex ways by this same mechanism.

Oxidative stress and low NO cause changes in DNA methylation. Presumably this is part of the normal mechanism by which stress (which results in oxidative stress and low NO) causes global epigenetic reprogramming of diverse genes in diverse tissue compartments under diverse circumstances. Psychological stress causes long lasting changes in neurological functioning as for example PTSD. The details of how psychological stresses of what types cause the characteristic neurological changes that manifest as PTSD are mostly unknown. We know it happens, so there must be physiology that supports those characteristic changes.

DNA methylation mediated through NO does influence expression of genes that are involved in some autism-like syndromes, such as Fragile-X mental retardation gene (FMR1). Aberrant readout of DNA methylation is implicated in the autism-like symptoms of Rett Syndrome (RS). Many of the symptoms of RS are characterized by physiology being skewed in the direction of low basal NO.

Fundamental brain optimization tradeoff: Theory of Mind (ToM) for Theory of Reality (ToR)

Humans are unique among animals for their abilities at communication; language making and language using and tools; tool making and tool using. These abilities are highly dependent on a large brain with substantial plasticity for self modification via learning throughout life. The relative importance of these two extremely important human behavioral characteristics is dependant upon the environment the infant is born into. The major neuroanatomy of the brain originates from structures arising during early neuron proliferation and differentiation during and after neurulation in the first trimester. These structures are then elaborated on later in utero. It would be beyond surprising if the relative aptitude of the brain for communication and/or tool making/using (including manual dexterity for manipulating objects) were not to some extent programmed in utero.

That trade-off would show up as a trade-off between abilities to understand and manipulate other humans (Theory of Mind), and abilities to understand and manipulate reality (Theory of Reality). These are the differences that are seen in people along the autism spectrum. I will discuss how a less developed ToM interferes with ASDs communicating with neurologically typical individuals (NTs) and their very well developed ToM. For the most part ASDs don't pick up the nuances of communication, particularly how it relates to motivations, beliefs and other mental states. The NT ToM forces NTs to think in anthropomorphic terms, even when it is inappropriate because they lack a robust ToR.

I spend a lot of time trying to explain this in several different ways using several different analogies. What I am trying to describe is exceedingly complex, as complex as an entire human brain. Something that complex cannot be described simply except in simplistic terms. It would be like trying to describe a library in a single paragraph.

Good times --> Need Good ToM

One hypothesis of this paper is that when times are good, and a woman's first trimester of pregnancy is characterized by low stress, then the "optimum" infant brain will be one optimized for better communication. If times are good, there will be plenty of other humans around, and the infant's primary competition for food and mates will be with other humans. Because times are good, the cultural information the adults have is working well to produce those good times. Copying that good cultural information with high fidelity is important. Good communication, a good ToM with the ability to understand and manipulate other humans is the best strategy when times are good.

Hard times --> Need Good ToR

When times are hard, a woman's first trimester will be characterized by high stress. The optimum infant brain will be skewed away from communication because during evolutionary time, hard times meant fewer humans in the territory. With fewer other humans around the need for communication is reduced. The cultural practices of the adults are not working well to produce good times. If the cultural practices are not working well, they need to be modified until they are working well. When times are hard, there won't be many other humans around because they will die in infancy. Competition will be against reality for food, shelter, and to stay alive. The adults don't know how to make good times, so that is something the infant will have to figure out for him/herself.

Theory of mind, theory of reality and theory of cognition

I will make a distinction between being able to think about something (cognition) and being able to think about that thinking process (meta-cognition) and coin a term Theory of Cognition (ToC), to denote the ability to think about and emulate different types of cognition. The term is an attempt to be analogous to ToM and ToR, which are meta-abilities to think about and compare multiple models of other minds (necessary for communication by emulating other mental states), and multiple models of potential realities. I am making this distinction because the different types of computation that humans do are not necessarily mapable onto each other.

Word processing ToM Theory of Mind Emulating other minds
Spread sheet software ToR Theory of Reality Emulating Reality
Operating system ToC Theory of Cognition Choosing Emulations

Trying to understand human communication with a ToR might be like trying to write a document not with word processing software but with a spread sheet. A document could be written in a spread sheet. It would be slow, cumbersome and the document wouldn't have the right formatting and wouldn't be as polished as something done in a word processor. On the other hand, a large spread sheet calculation simply can not be done using word processing software. The word processing software doesn't support the primitive functions, addition, subtraction, multiplication, etc that a spread sheet calculation requires. Some of the complex word processing functions can be emulated on a spread sheet but spell checking and grammar correction would be very cumbersome and difficult.

This is the sense that I am trying to convey, that people with a robust ToM can do good and robust communication that is nuanced, and well understood by others with a matching and robust ToM. Their shared ToM is analogous to the word processing software, and a well formatted document is analogous to human communication between individuals with a shared ToM. The ToR is analogous to the spread sheet software and the large spread sheet analogous to a highly technical ToR. A theory of cognition (ToC) would be the selection of the proper software type to do the required computations (ToM(English), ToM(French), ToM(ASL), ToR). There are multiple ToMs, each language is different to some extent, although there are other communication modes, body language, cultural signals, gestures. There is only one ToR, the one which accurately describes reality as it actually is. Individuals may have a ToR that is highly specialized and individualistic, physics or medicine for example. But all of the different ToRs all mesh into one (or should) because they all describe a single reality.

In this analogy I am trying to illustrate that a specialized piece of computation machinery may work very well for one task (word processing) and not at all for another (spread sheet calculation). If you tried to input a spread sheet into a word processor, you would get many error codes, many misspelled words; the word processing software would reject it as badly formed. A very well formed spread sheet cannot be read on a word processor. This is analogous the problem that some NTs have with understanding people with ASDs.

Normal background "housekeeping" features can impede communication. Many types of different word-processing software have automatic spell checking. If a word is spelled wrong, the software will change the spelling to match the spelling to one of the words in its dictionary. If the word is not misspelled, but is simply not in the dictionary, the software can't recognize it and will change it anyway. This is a type 1 error, a false positive. The software falsely identifies a character string and modifies it to match its default identity. This is an inherent property of specialized pattern recognition systems. There is a trade off of type 1 errors (false positive) for type 2 errors (non detection). If you don't have access to modify the "error correction function", it may be impossible to type certain strings because the error correction keeps changing them. In certain word processing software this can be extremely annoying and make writing outside the scope of the software impossible. If the software won't let you have certain character strings in your document, you can't write about them. If your ToM won't let you have certain ideas, you can't think about them. Appreciating that your ToM doesn't have the capacity to think certain thoughts is extremely difficult.

Being unable to conceptualize ideas is not uncommon. There are people who believe in the literal truth of ancient texts and are unable to conceive that they do not accurately describe reality, irrespective of what data can be collected today. These beliefs are from a ToM, shared with others of their community. Such beliefs did not arise from observations of reality, they were told to individuals by other individuals who believed them. Those false beliefs are derived by those false beliefs being communicated to the individual and so shaping their ToM. Such beliefs are often extremely resistant to change.

Learning can be looked at as modification of the brain's neural network so the neuroanatomy can support what ever new idea it is that is being learned. Usually this takes a long time and is quite difficult. Learning physics or mathematics is difficult because the normally developing neural patterning doesn't support that type of thinking the way it supports language. I will discuss this more later.

A mother's necessity makes her child an inventor.

Maternal stress --> fetal development along autism spectrum --> Asperger phenotype

The hypothesis of this paper is that low NO in utero causes development along the autism spectrum so as to program the brain in utero to one that supports a better ToR. Precisely the phenotype that is needed when mothers are stressed and so times are hard. What ever technology and cultural practices are being used, they are not working well enough and so new ones need to be developed.

Which individuals are most adept at tool use today? It is people with Asperger’s, people with ASDs. Many scientists and engineers have Asperger’s, and it is suggested that Einstein, Newton, and many brilliant scientists had Asperger’s. [22] Asperger even said “It seems that for success in science or art a dash of autism is essential.” [23] The stereotypical nerd is someone with facility at math, science and with characteristically poor social skills[24]. The mirror neuron system[25] (responsible for understanding the actions of other individuals) exhibits dysfunction proportional to ASD severity. [26]

The major barrier to revolutionary scientific innovation is conventional thinking and existing paradigms. [27] What Kuhn calls “normal science”. Ideas transmitted culturally are difficult to displace even when wrong. It is nearly 150 years since Darwin’s “Origin of Species”, with overwhelming data supporting and no datum inconsistent with evolution, yet in the USA, 40% of the population believes evolution is false. [28] Some on the Nobel Committee were unable to accept relativity as valid and so Einstein received the Nobel Prize for the photoelectric effect, not relativity. [29]

Cultural notions of what is appropriate affect abilities (i.e. what people think they or anyone can do). Women exposed to a hypothesis wrongly attributing mathematical ability to genes on the Y chromosome have impaired mathematics performance. [30] A degree of social isolation from disrupted mirror neurons may insulate ASD individuals from incorrect paradigms of science, technology and the peer pressure associated with cultural practices which must be abandoned to overcome hard times. No doubt 2.8 MYA everyone “knew” stones didn’t make good tools. The first stone tools were not developed after committees of peers reviewed proposals and selected the highest scoring for implementation; they were developed by the “Einsteins” of the time working alone. Acquisition of nut cracking skill by capuchin monkeys (Cebus apella) using stones as hammer and anvil takes about 2 years and requires considerable repetitive nonproductive effort while watching proficient individuals. [31] No doubt repetitive trial and error was needed to acquire de novo skill(s) to manufacture stone tools 2 MYA, and such individuals had to ignore criticism that they were bizarre for “uselessly” banging stones together.

These culture notions are transmitted through the robust NT ToM. To avoid being adversely influenced by potentially incorrect ToMs, ASDs require a weaker ToM, a ToM that perhaps allows some communication, but one that can easily be ignored, resulting in the ability to ignore the "Kool-Aid" which is a stronger version of groupthink. The delusional world views that some NTs have can become extremely compelling to them,, such that they are unable to perceive that it is delusional, particularly when a charismatic leader with a strong ToM (essentially the definition of a charismatic leader) imposes it on his/her followers. More on this later in the discussion of cargo cult science.

ASD individuals developing skills unrecognized as useful by NTs must possess a compulsion to acquire those skills despite peer pressure that such skills are useless. Many ASDs acquire skills that NTs think are useless, fascination with train spotting, bird watching, collecting, virtually every savant ability is acquired to a degree that NTs do not find useful (if they did, then NTs would acquire such skills to that degree and it would not be thought of as savant). Language and communication is the "savant" skill of NTs. NTs possess a skill at communication (with other NTs) that ASDs cannot hope to master. Just as ASDs sometimes have skills that NTs cannot hope to master. Just as elite athletes have skills that non-athletes cannot hope to master. A society with the ability to use the best skills of a highly variable and diverse group would be better able to cope with adversity than a society where all individuals had the same average abilities. Not every individual in the village needs to be proficient at making stone tools, provided there are enough proficient individuals and the tools their skills produce can be traded for other things.

Cognition: Non-algorithmic calculation

Cognition in human brains is done by neural networks, the fundamental details of which are mostly unknown. Some cognitive abilities (such as savant calendar) are known to be non-algorithmic because the errors are not always the same, and the time for performing the calculation is not asymmetric depending on calculation direction the way a computation performed using an algorithm would be. [32] It is likely that most if not all other types of human cognition are non-algorithmic.

I am using algorithm in the sense that an algorithm is what a Turing machine executes. An algorithm is a series of instructions that when acted upon manipulate data and perform a calculation. The computers that people are familiar with are algorithmic. Calculators use a calculating engine (the processor) to operate an algorithm (the software) to manipulate the data. In general the data does not modify the software or the processor while the computation is in process, and given the same data, the same processor running the same software will produce the same output each and every time the calculation is performed.

Neural networks are inherently non-algorithmic in the sense that there is no "algorithm" explicitly being implemented by the neural network. A neural network may be used to implement an algorithm. For example, humans and other animals have the ability to do approximate mathematical operations such as comparison. Two groups of objects can be compared and the one with the larger number can be selected even when the members of each group have not been counted. This selection can be made by individuals unable to count and even by animals. This selection is non-algorithmic. An individual able to count is also able to count the members of each group and then tell which group is larger by comparing the two values. The person, who can count, knows that the counting algorithm produces a more reliable comparison than the non-algorithmic visual comparison. The person who cannot count does not know how to implement the counting algorithm.

Human brains are not optimally configured to run algorithms. A processor that can run algorithms is in essence emulated in a human brain to run the algorithm under consideration, such as counting or multiplication. The form that the data is in may greatly limit what algorithms that data can be manipulated with. For example multiplication using Arabic numerals is easy. Multiplication using Roman numerals is exceedingly difficult. There is essentially no algorithm for multiplying Roman numerals, individuals use a look-up table. Learning algorithms takes considerable time and effort for many individuals.

Communication requires a Theory of Mind

Exactly how neural networks in the brain configure and reconfigure themselves to do the computations that certain cognitive tasks require is unknown. Presumably there is some type of feedback that modifies the network when sub-optimal results are achieved so as to configure the network to produce better results. How this occurs is unknown, but for language acquisition, some conclusions as to how this optimization works can be made which I discuss below. What I want to emphasize the compulsive aspects of language acquisition. People do not choose to acquire the language they acquire as children, their brains acquire it (or synthesize it de novo) for them.

All communication requires two parties, a sender and a receiver. The sender must have a mental concept, translate that mental concept into a communication medium, transmit that message to the receiver, who must receive and then translate that message back into a mental concept. In that sense, all communication is only the transmission of representations of internal mental states. For there to be communication, the mental concept must necessarily be mapable onto the neural structures of both individuals. If one party is not able to represent the mental concept in their brain, the concept cannot be communicated either from them or to them. In a sense, communication is the transmission of data that allows the receiver to identify and map that concept into a mental representation, in effect the receiver is doing pattern recognition on the data stream and generating a mental representation, in effect a pattern of thought either generated de novo, or a familiar pattern previously used.

In this sense, communication can only occur between individuals with a shared Theory of Mind. This is the sense that I am using ToM in this paper, the emulation of the cognition of another individual to achieve a mapping of the mental state of one individual with the mental state of another individual. The possible fidelity of that mapping determines the possible fidelity of that communication. If a mental state cannot be mapped onto another individual's ToM, then that mental state cannot be communicated to that individual.

Pattern recognition is a well recognized ability. All systems encoding pattern recognition are subject to different forms of error. There is the type 1 error, the false positive, the error in wrongly identifying a false instance as positive. There is also the type 2 error, the false negative, the error in missing the correct identification of a correct instance. In a general sense any pattern recognition system can be made more sensitive, that is with a reduced type 2 error, but then there is an increased type 1 error and there are more false positives.

A type 1 error is getting the attempted message wrong; a type 2 error is missing the attempted message. Since communication is a two-party interaction, the "fault" of miscommunication cannot be attributed to either party, the "fault" lies in their interaction.

Many human interactions engender other types of error. There is no generally accepted definition of what is a Type 3 error, but one definition is "the error committed by giving the right answer to the wrong problem". When times are hard, and the culturally transmitted traditional information isn't capable of solving the hard times, that is an example of the right answer to the wrong problem. The time of adolescence and early adulthood is often a time of rebellion against authority, against conventional wisdom, against cultural norms. Young people are testing the limits of their culturally acquired information; testing to see what works and what doesn't. This is somewhat speculative but this might be a mechanism to reduce the cultural transmission of obsolete or dysfunctional practices. In the absence of a written language, the only cultural practices that can be transmitted are those adopted by the next generation. If the older members of the tribe live long enough to transfer their wisdom to adults past their adolescent rebellion period, perhaps the wisdom is worth transferring. If not, then perhaps it isn't and the tribe should try new approaches until that happens.

Communication and language acquisition

Social animals communicate with each other. In humans the ability to develop language is innate and the brain structures to support language and language development must be coded for genetically. Language itself must be learned or is synthesized de novo during certain periods of brain development. This point is quite important. When humans are growing up in a culture, they adopt the language of the culture, provided that the language is "well formed". If the language the adults are using is not "well formed", the children synthesize a new language that is "well formed". That is, when the children of immigrant parents grow up, they do not adopt the pigeon language their parents are speaking, they either adopt the "well formed" dominant language, or synthesize a "well formed" Creole. The various sign languages did not become "well formed" until children grew up with signing as their first language, which they modified into a "well formed" language.

The acquisition of language in this way tells us several things; that the ability to acquire language is innate, that there is a more "primitive" cognitive structure underlying language (by that I mean that the structure of "thought" has a component that is simpler than the linguistic components humans communicate with). Without a simpler and more primitive cognitive structure, the Creole could not be analyzed as it is being formed to ensure the resulting Creole has a "well formed" grammar. However, the ability to form a Creole is lost at a certain age. The immigrant parents of the Creole synthesizing children continue to speak their pigeon language. This implies that the cognitive structure that analyzes language as it is being learned and forces it to be "well formed", i.e. to conform to standard human linguistic patterns is lost (to some extent) with age. It also implies a compulsion to learn the "standard" language and a compulsion to force others to comply with the "standard".

The development of a de novo language, such as a Creole, is a collective outcome produced by a population. It is not produced by a single individual. Another way of describing it is that the population developing the language acquires a shared neural mapping of the medium of the language (sounds, gestures, etc) to neural structures producing the mental states that are the ultimate outcome of communication (that is the ideas being communicated). In this context, there is no arbitrarily correct mapping. The mapping is correct so long as it is the mapping shared by the group. In the sense of the Galbraith quote at the start what ever the majority adopts as the linguistic mapping is the correct mapping. This is a very important point. What ever the majority adopts as correct is correct; everything else is wrong.

For a single majority linguistic mapping to arise spontaneously there must be very powerful mechanism(s) to eliminate deviation from the mapping acquired by the majority. The majority acquire a shared Theory of Mind with respect to linguistic mapping. In other words, the differences between the shared Theory of Mind and that of any individuals in the population are reduced. The deviation is not reduced by changes to the shared theory of mind; the deviation is reduced by individuals adopting the shared ToM as their own. This is an important point. There is no "shared" ToM. There are only individual ToMs which correspond to the shared ToM more or less. The shared TOM can only be shared to the extent that all individuals have the same components and the same structural relationships between those components. The shared ToM reflects the "lowest common denominator"; the ToM that overlaps with everyone else's ToM is all that can be shared. I think this relates to the importance of "peer pressure" in the age group capable of forming a Creole language. If peer pressure were not so compelling, a single coherent language would be difficult to achieve.

The rigidity of an inflexible ToM maintains stability of communication, of information transmitted culturally to the next generation. If your ToM doesn't support an idea, you cannot transmit it, receive it, understand it, or even think it. When times are easy, transmitting the cultural information that led to those easy times is important. It is important to do so with high fidelity because it worked. When times are hard, the culturally transmitted information isn't working, and so needs to be abandoned or modified. The fidelity of transmission must be reduced so what ever is wrong and/or isn't working can be eliminated.

The ToM of NTs that allows them to communicate so easily with each other limits what they can communicate to ideas that are within the shared ToM. This is an extremely important point, but it is a point that NTs have an extremely difficult time understanding because they can only think using ideas that are within their shared ToM. If an individual's ToM is insufficiently flexible to map an idea, that idea cannot be understood unless the ToM changes. But there is tremendous peer pressure to maintain the shared ToM of the group and to not change it.

This rigidity of the NT ToM is what causes ideas to persist even when those ideas are wrong and the rejection of correct ideas even when well supported by incontrovertible data. Many religious ideas have no supporting evidence and are in fact demonstrably wrong. For example the idea that the Earth is less than 10,000 years old and was created in 6 days as described in Genesis. Similarly the idea of evolution is rejected without a single piece of data inconsistent with it.

Most ToM ideas are transmitted from other individuals, not generated de novo.

Conflicting compulsions for ToM and ToR

A specific ToM is only useful for communication in the context of the group of individuals that share it. The mapping of a data stream (i.e. speech or gestures) into ideas and mental states is arbitrary and the only correct mapping is the one that everyone else in the group shares. There must be a tremendous compulsion to modify one's ToM to conform to that of the group. It is this compulsion that forces the emergence of a single language in a group.

In contrast, a ToR is only useful in so far as it actually corresponds to reality. To eventually develop a robust ToR, the individual must have a compulsion to modify his/her own ToR until it does correspond with reality, irrespective of the ToR of others in the group.

Thus developing and maintaining a good ToR is in conflict with developing a good ToM. A ToM needs to remain static for individuals to be able to communicate with each other. A ToR needs to be dynamic and change when ever it is found to be in error or to be dysfunctional.

I think this is the source of much of the resistance to new ideas in human culture but also in the scientific community. These concepts are laid out by Thomas Kuhn in his book, The Structure of Scientific Revolutions. Most scientists do what Kuhn calls "ordinary science", where they work within the paradigm of their scientific field. It is difficult to work outside the paradigms of a scientific field. Any contradiction of an existing paradigm is considered extraordinary and so requires extraordinary evidence. Some individuals are unable to reject paradigms even when they have been shown to be wrong. In these individuals, their rigid ToM has locked them into a perpetual state of error, and they don't have a sufficiently robust ToC to appreciate that their thinking is faulty and in error. It is mechanisms similar to the mechanisms that enforce a rigid ToM during language acquisition that compel adherence to the faulty ToM in later life, peer pressure, appeal to authority, tradition.

Communication and ASDs

Communication in humans encompasses a number of modalities including speech, sign language, body language, written language, music, artistic expression and perhaps pheromones. Most of these have components that are learned, improve with practice and degrade with disuse demonstrating the involvement of neural structures which retain plasticity (positive and negative) even in adulthood.

Autism is defined by behaviors, behaviors related to social interactions where autistic individuals have what are called characteristic deficits which can be reliably measured. However what constitutes a deficit is a matter of perspective. One example is a "deficit" in the ability to impute anthropomorphic motivation and emotion to inanimate objects as in the work of Frith. In this research, triangles were animated and made to move in three different ways, randomly, goal directed and moving interactively with implied intentions. The two sets of purposeful motions were designed to evoke anthropomorphic responses, e.g. chasing, fighting and coaxing, tricking. Individuals were scored on how closely they matched the scripts the animators of the triangles were trying to portray.

The ASD individuals scored lower than the NTs did, and this was described as a "mentalizing dysfunction". This was taken as a confirmation that people with ASDs have an impairment in attribution of mental states. However, whose "mental state" did the ASDs have an impairment in recognizing? The "mental state" of the triangles? Was this error a type 1 error (false positive), or type 2 error (false negative)? One might say the ASDs had a type 2 error, failure to recognize the "mental state" of the triangles, but one could (more correctly I think) say that the NTs had a type 1 error of falsely attributing a "mental state" to obviously inanimate triangles.

There is no intrinsically correct representation of the mental state of triangles. Triangles do not have mental states. The only way that a mental state can be attributed to triangles is via anthropomorphic projection of human-type intentions onto inanimate objects. In most circumstances this would be a Type 1 error; falsely observing anthropomorphic attributes in inanimate objects. It could also be thought of as a type 3 error, wrongly using a human based anthropomorphic model where it is inappropriate. This type of projection is not uncommon. Imputation of motivation and intentions to inanimate objects was at one time the basis for the religious belief that demons and spirits inhabit and animate virtually every object.

Inappropriate invocation of anthropomorphic feelings is a large part of the entertainment industry. Many cartoons are stylized after humans and many humans develop grossly and dangerously wrong ToR based on these erroneous ideas. In regions where bears are endemic, campers feeding bears is a serious problem. People assume that the anthropomorphic representations they have seen on TV are representative of how bears will react in real life. There have even been cases where a parent has applied peanut butter to a child's face so a bear cub would lick it off to obtain cute pictures.

This relates to the second quote, "If the only tool you have is a hammer, you tend to see every problem as a nail." If the only cognitive structures you have to think with are the cognitive structures of human emotions and communication, trying to figure out the properties of inanimate objects would consist of trying to ascribe human motivations and intentions to those inanimate objects and trying to figure out what they would do next in human terms.

Savant cognitive abilities

A striking feature of some people on the autism spectrum is that in some instances they have narrow and highly superior cognitive abilities. Human mental abilities have distributions in the population, with "normal" abilities being distributed "normally". Usually people with autism are somewhat lower on intelligence tests such as WISC, but with somewhat higher scores in block design. When intelligence tests without a communication component such as Raven's Progressive Matricies are used, some autistic individuals score much higher, in some cases as much as 70 percentile points higher (n=7). [33] That is 70 percentile points higher. Such a lack of congruence between tests is sufficient to show they are not measuring the same thing and we shouldn't use the same label to denote what the different tests are measuring even if there is good correlation among NTs. That correlation can only be spurious.

The distribution of intellectual abilities is "normalized", that is differences are measured and then scaled to fit on a distribution. That scale is arbitrary, and does not reflect any sort of absolute scale of difficulty.

As social animals, humans live in societies, larger communities of humans where there can be specialization and division of labor. It is this specialization and division of labor that has allowed humans to collectively master many technologies. Presumably different mental tasks are optimized by different neural structures. Dispersion in mental abilities requires dispersion in neural structures.

Savant abilities are not rare among people on the autism spectrum, and sometimes occur in individuals with profound disruptions in other cognitive abilities. This shows that to some extent, some cognitive abilities are independent of each other. Presumably superior performance in some cognitive tasks and inferior performance in others represents a trade-off of abilities along multiple spectra. The brain is limited in size, its computation capacity is limited, relative cognitive abilities of individuals will depend on the myriad details of the neurodevelopmental path that individual took.

Communication is "savant" ability of NTs

Many ASDs have savant abilities, which demonstrate that what ever part of the brain is providing those cognitive abilities has superior performance to the corresponding part of NT brains with lesser performance. The one area where NTs are universally better than ASDs is in communication. I suggest that communication is the savant ability of NTs, and that NTs have traded reduced abilities on ToR and ToC for enhanced ability in ToM.

The difficulty in relations between ASDs and NTs is that NTs don't appreciate that the ToM they are using for communication is a savant ability that ASDs don't share, and shouldn't be expected to be able to emulate. An ASD can't emulate the NT savant ability to communicate any more than an NT can emulate an ASD savant ability at mathematics. If you don't have the brain structures that can do the computations, you can't emulate the behavior. You might be able to fake how it sounds, but because the fundamental neural structures are not present, it is just an act and can't have the actual content.

Trying to think about Reality with a ToM is like doing Cargo Cult Science

Richard Feynman coined a term, Cargo Cult Science, to describe the practices of people who may be doing what they call experiments, but they are missing the fundamental intellectual honesty to be actually doing science. The term comes from the observation that tribes in the South Pacific would observe westerners arrive and set up landing strips which would bring aircraft laden with cargo, all sorts of goods that seemed to appear like magic. Along the lines of Arthur C. Clarke's observation that "any sufficiently advanced technology is indistinguishable from magic". They tried to understand the source of this cargo and how to get cargo for themselves using their understanding of reality. They generated a Cargo Cult, and proceeded to adopt rituals to try and cause cargo to appear.

This is really an excellent metaphor for trying to think about a subject with the wrong approach. Their thinking was that such good cargo had to come from the Ancestors, but the Ancestors would only bring such good cargo if they were communicated with in the right way, which the westerners knew how to do, so copy them and the cargo would appear. They built landing strips, control towers and populated them with radio control operators, but to no avail.

Explaining that their approach was wrong would fall on deaf ears. They don't have the background to understand where the cargo actually came from. They had anthropomorphized their observations and reduced them to the human terms they could understand using their ToM. They didn't come to their beliefs via facts and logic, facts and logic won't dissuade them from their beliefs.

Obviously there are multiple individuals involved, a leader and followers and the leader may achieve lots of things even if no cargo shows up. Presumably it is the charismatic persuasion of the leader using the leader's ToM that causes the followers to believe the leader. Simply by leading the effort to obtain cargo the leader achieves status over the others. Even when doing something completely useless and wrong, the society holds together if all of the members share the same ToM. Individuals not sharing the conceptualization of obtaining cargo by building airstrips would not fit in.

To people who have savant mathematical ability, those without it who are trying to emulate mathematical abilities can be seen as trying to do cargo cult mathematics. They can go through the motions, but don't have the ability to generate the content. Similarly, ASDs who try to communicate with NTs are doing cargo cult communication. They can go through the motions, but there is a lot of stuff that is being missed.

Neurological structures required to support an idea

The only ideas that an individual can think about are ideas that can be mapped into that individual's neural network. To learn a new idea, either the neural structure present is sufficiently flexible that the new idea can be mapped into it, or the neural structure must be modified until the new idea can be mapped onto it.

The process of learning a new idea must include as the first steps, the process of modifying the neural networks of the brain such that they can support the new idea being learned. Often the first step is "unlearning" ideas that are wrong. I think that this modification of the brain to support new ideas is why learning takes such a long time. New neural structures need to be made.

All mental representations require some level of neuronal "overhead" to be sustained. The details of how the brain does that are not understood. While the capacities of the brain are large, they are not infinite, and at some point trade-offs must be made.

Socially isolated individuals develop on an autism-like pathway.

All important physiological systems are under feedback control (that would be all physiological systems). Presumably an organ as important as the brain is also under feedback control, and this is reflected in the improved efficiency obtained through practice at certain mental tasks.

Presumably if there is a trade-off of ToM vs. ToR, then isolated individuals with no need for a ToM would develop a more robust ToR. This does appear to be the case in multiple organisms including humans, monkeys and rodents.

The classic work on socially isolated monkeys was done by Harlow in the 1960's, [34] and present animal welfare regulations would make such experiments problematic. Some monkeys were raised with no social contact at all, even with their mothers. Such monkeys were profoundly affected and exhibited rocking behaviors, self-injurious behaviors and profound disruption in abilities to interact with other monkeys. They were termed autistic by the experimenters.

Surprisingly, some of these socially isolated monkeys exhibited superior cognitive abilities. What is especially interesting is that these superior abilities were termed "deficits" by the experimenters. [35] Socially reared monkeys were conditioned with a tone and a startle stimulus. A redundant lamp was then paired with the tone. Socially isolated monkeys conditioned to the redundant light, the socially reared monkeys did not. The experimenters characterized the non-conditioning of the socialized monkeys to the redundant signal of the light as "blocking" the isolated monkeys then exhibited what was termed a "deficit" in blocking. Why the experimenters chose to use the term "deficit" to refer to a superior ability tells us something about the experimenters and their expectations about the socially deprived monkeys, not the monkeys.

Rhesus monkeys raised in social isolation have superior learning performance to those raised in social environments. [36]

Involvement of nitric oxide in social interactions and communication

The archetypal social interaction in mammals is the bonding of the mother to her infant. All mammals exhibit this behavior and have exhibited it for as long as mammals have suckled their young. The first social interaction all mammals have is with their mother. Even mammals thought of as primarily non-social do have this social interaction.

NO is involved in the development of the bonding and smell recognition that occurs in ewes within 2 hour of giving birth. Inhibition of nNOS blocks formation of that olfactory memory, and this blockage can be reversed by infusion of NO into the olfactory bulb. [37] Oxytocin is essential in the formation of normal social attachment in mice. [38] Reduction in oxytocin release following epidural anesthesia in heifers preceded a reduction in maternal bonding type behaviors[39]. Activation of the oxytocin receptor causes activation of nitric oxide synthase. [40] The connections that mediate maternal bonding can occur in the space of a few hours[41], limiting the distance over which axons must migrate to form these new connections.

Why NO is the signaling molecule that mediated the neural remodeling to cause maternal bonding makes evolutionary sense. Lactation is extremely energy intensive. If a mother does not have the metabolic resources to generate sufficient milk of sufficient nutritional quality to sustain her infant until it is weaned, she (and her infant) is better off not bonding to her infant and abandoning it. Spending metabolic resources on a reproductive attempt that will fail will reduce the success of potential future reproductive attempts. A failed reproductive attempt has no advantage either to the mother, or to the infant. An infant's best reproductive strategy in those circumstances is to do what ever increases the likelihood that the infant's mother will have a successful reproductive event later, so that the non-surviving infant may have a surviving sibling.

I discuss this at length in my blog on infanticide. Using NO as the positive signaling molecule to mediate maternal bonding directly couples maternal bonding to energy status. The low NO of metabolic stress directly reduces the degree and fidelity of maternal bonding. In extreme metabolic stress (the most important states for maternal bonding to be blocked) the maternal instinct turns from nurturing to infanticide. It needs to be appreciated that infanticide under conditions of extreme metabolic stress is as much a "maternal" instinct as is nurturing when times are better. I see infanticide as the brutally hard state that desperate metabolic stress induces in postpartum women.

Social isolation reduces NO generating neurons in the brain

When rodents are raised in a socially deprived setting, the numbers of NO producing neurons in some parts of their brains are reduced. [42] A reduction in basal NO in the brain due to development under socially isolated circumstances makes sense. Many social interactions are mediated via NO mediated pathways, including bonding and other pathways mediated through oxytocin. If the environment one is growing into is non-social, social neural pathways have little or no survival benefit. Better to develop the neural structures that will be useful.

Socially isolated individuals retain sufficient neural plasticity to partially recover

Social isolation at birth produced monkeys with profoundly disrupted social abilities. Experiments demonstrated that some of the disrupted social abilities could be restored. This involved the use of "therapist" monkeys, usually socially raised normal monkeys that were substantially younger than the isolated monkeys. [43] In females, a socially isolated female could recover somewhat and be an improved mother following pregnancy and raising an infant however many times the first born infants did not do very well but mothering did improve with subsequence births[44] demonstrating plasticity in neural networks mediating mothering behaviors during pregnancy and/or mothering activities. Since maternal bonding is the archetypal communication pathway for mammals, this suggests that other fundamental communication pathways have plasticity also.

With NO being involved in bonding, improved bonding and mothering interactions with subsequent births is consistent with increased neurogenic nitric oxide as a causal mechanism. If a non-social environment becomes social, reconfiguring neural structures to cope with social interactions would be advantageous.

Potential for treatment

I suggest an analogous treatment for ASD individuals may be to incorporate them into play groups with significantly younger NT children that are at similar developmental stages, but with sufficient adult supervision that nothing untoward can happen.

Doing this in the context of increasing NO levels via the techniques I am working on my have important therapeutic effects.


1 Kates WR, Burnette CP, Eliez S, Strunge LA, Kaplan D, Landa R, Reiss AL, Pearlson GD. Neuroanatomic Variation in Monozygotic Twin Pairs Discordant for the Narrow Phenotype for Autism. Am J Psychiatry. 2004 Mar;161(3):539-46.

2 Greenberg DA, Hodge SE, Sowinski J, Nicoll D. Excess of twins among affected sibling pairs with autism: implications for the etiology of autism. Am J Hum Genet. 2001 Nov;69(5):1062-7.

3 Visscher PM. Increased Rate of Twins among Affected Sib Pairs. Am J Hum Genet. 2002 Oct;71(4):995-6; author reply 996-9.

4 Herbert MR, Russo JP, Yang S, Roohi J, Blaxill M, Kahler SG, Cremer L, Hatchwell E. Autism and environmental genomics. Neurotoxicology 2006 Sep;27(5):671-84.

5 Happe F, Ronald A, Plomin R. Time to give up on a single explanation for autism. Nat Neurosci. 2006 Oct;9(10):1218-20.

6 Godfrey KM, Barker DJ. Fetal nutrition and adult disease. Am J Clin Nutr. 2000 May;71(5 Suppl):1344S-52S. Review.

7 Fowden AL, Giussani DA, Forhead AJ. Intrauterine programming of physiological systems: causes and consequences. Physiology (Bethesda). 2006 Feb;21:29-37. Review.

8 http://ajprenal.physiology.org/cgi/content/full/288/4/F626#BIBL

9 Esch T, Stefano GB, Fricchione GL, Benson H. Stress-related diseases – a potential role for nitric oxide. Med Sci Monit. 2002 Jun;8(6):RA103-18.

10. Herbert MR. Large Brains in Autism: The Challenge of Pervasive Abnormality. Neuroscientist. 2005 Oct;11(5):417-40.

11. Casanova MF, Buxhoeveden DP, Switala AE, Roy E. Minicolumnar pathology in autism. Neurology. 2002 Feb 12;58(3):428-32.

12. Gustafsson L. Comment on "Disruption in the inhibitory architecture of the cell minicolumn: implications for autism". Neuroscientist. 2004 Jun;10(3):189-91.

[1]3. Beversdorf DQ, Manning SE, Hillier A, Anderson SL, Nordgren RE, Walters SE, Nagaraja HN, Cooley WC, Gaelic SE, Bauman ML. Timing of prenatal stressors and autism. J Autism Dev Disord. 2005 Aug;35(4):471-8.

14. Kapoor A, Matthews SG. Short periods of prenatal stress affect growth, behaviour and hypothalamo–pituitary–adrenal axis activity in male guinea pig offspring. J Physiol. 2005 Aug 1;566(Pt 3):967-77.

[1]5. Cannizzaro C, Plescia F, Martire M, Gagliano M, Cannizzaro G, Mantia G, Cannizzaro E. Single, intense prenatal stress decreases emotionality and enhances learning performance in the adolescent rat offspring: interaction with a brief, daily maternal separation. Behav Brain Res. 2006 Apr 25;169(1):128-36.

[1]6. Kapoor A, Dunn E, Kostaki A, Andrews MH, Matthews SG. Fetal programming of hypothalamo-pituitary-adrenal function: prenatal stress and glucocorticoids.J Physiol. 2006 Apr 1;572(Pt 1):31-44.

[1]7. Peunova N, Scheinker V, Cline H, Enikolopov G. Nitric oxide is an essential negative regulator of cell proliferation in Xenopus brain. J Neurosci. 2001 Nov 15;21(22):8809-18.

[1]8. Contestabile A, Ciani E. R Role of nitric oxide in the regulation of neuronal proliferation, survival and differentiation. Neurochem Int. 2004 Nov;45(6):903-14..

[1]9. Herbert MR, Ziegler DA, Deutsch CK, O'Brien LM, Kennedy DN, Filipek PA, Bakardjiev AI, Hodgson J, Takeoka M, Makris N, Caviness VS Jr. Brain asymmetries in autism and developmental language disorder: a nested whole-brain analysis. Brain. 2005 Jan;128(Pt 1):213-26.

20. Susman RL. Fossil Evidence for Early Hominid Tool Use. Science. 1994 Sep 9;265(5178):1570-3.

2[1]. Courchesne E, Karns CM, Davis HR, Ziccardi R, Carper RA, Tigue ZD, Chisum HJ, Moses P, Pierce K, Lord C, Lincoln AJ, Pizzo S, Schreibman L, Haas RH, Akshoomoff NA, Courchesne RY. Unusual brain growth patterns in early life in patients with autistic disorder: an MRI study. Neurology. 2001 Jul 24;57(2):245-54.

22. James I Singular scientists. J R Soc Med. 2003 Jan;96(1):36-9.

23 quoted in 22

24. Al Yankovic. White & Nerdy. Music video by "Weird Al" Yankovic from the album "Straight Outta Lynwood" Volcano records. http://www.youtube.com/watch?v=-xEzGIuY7kw (accessed 12/25/2006)

25. Rizzolatti G, Craighero L. The mirror-neuron system. Annu Rev Neurosci. 2004;27:169-92.

26. Dapretto M, Davies MS, Pfeifer JH, Scott AA, Sigman M, Bookheimer SY, Iacoboni M. Understanding emotions in others: mirror neuron dysfunction in children with autism spectrum disorders. Nat Neurosci. 2006 Jan;9(1):28-30.

27. Thomas S. Kuhn. The Structure of Scientific Revolutions. 3d edition. University of Chicago Press, 1996.

28. Miller JD, Scott EC, Okamoto S. Public Acceptance of Evolution. Science. 2006 Aug 11;313(5788):765-6.

29. Friedman RM. Einstein and the Nobel Committee: Authority vs. Expertise. europhysics news July/August 2005 129-133.

30. Dar-Nimrod I, Heine SJ. Exposure to Scientific Theories Affects Women’s Math Performance. Science. 2006 Oct 20;314(5798):435.

3[1]. Ottoni EB, de Resende BD, Izar P. Watching the best nutcrackers: what capuchin monkeys (Cebus apella) know about others' tool-using skills. Anim Cogn. 2005 Oct;8(4):215-9.

32. Mottron L, Lemmens K, Gagnon L, Seron X. Non-algorithmic access to calendar information in a calendar calculator with autism. J Autism Dev Disord. 2006 Feb;36(2):239-47.

33. Dawson M, Soulières I, Gernsbacher MA, Mottron L. The level and nature of autistic intelligence. Psychol Sci. 2007 Aug;18(8):657-62.


35. BEAUCHAMP AJ, GLUCK JP, FOUTY EH, LEWIS MH. Associative Processes in Differentially Reared Rhesus Monkeys (Macaca mulatta) : Blocking. Developmental Psychobioiogy 24(3): 175-189 (1991).

36. YEATON SP, O'CONNELL MF, STROBEL DA. Malnutrition and social isolation: learning in the developing rhesus monkey. PHYSIOL. BEHAV. 20(2) 125-128, 1978.

37. Kendrick KM, Guevara-Guzman R, Zorrilla J, Hinton MR, Broad KD, Mimmack M, Ohkura S. Formation of olfactory memories mediated by nitric oxide. Nature. 1997 Aug 14;388(6643):670-4.

38. Jennifer N. Ferguson, J. Matthew Aldag, Thomas R. Insel, and Larry J. Young. Oxytocin in the medial amygdale is essential for social recognition in the mouse. Journal Neuroscience, October 15, 2001, 21 (20):8278-8285.

39. G. L. Williams, O. S. Gazal, L. S. Leshin, R. L. Stanko, and L. L. Anderson. Physiological regulation of maternal behavior in heifers: Roles of genital stimulation, intracerebral oxytocin release and ovarian steroids. Biology of Reproduction 65, 295-300 (2001).

40. Gerald Gimpl and Falk Fahrenholz. The oxytocin receptor system: structure, function, and regulation. Physiological reviews vol. 81, No. 2, 629-683, April 2001.

41. Okere CO, Kaba H. Increased expression of neuronal nitric oxide synthase mRNA in the accessory olfactory bulb during the formation of olfactory recognition memory in mice. Eur J Neurosci. 2000 Dec;12(12):4552-6.


43. Harlow HF, Suomi SJ. Social recovery by isolation-reared monkeys. Proc Natl Acad Sci U S A. 1971 Jul;68(7):1534-8.

44. Ruppenthal GC, Arling GL, Harlow HF, Sackett GP, Suomi SJ. A 10-year perspective of motherless-mother monkey behavior. J Abnorm Psychol. 1976 Aug;85(4):341-9.

Sunday, October 19, 2008

Role of low basal NO in capillary and vascular abnormalities

Vascular remodeling under low NO conditions.

I have really just skimmed the surface in what I have presented here. There are a lot more details that reinforce the chain of facts and logic that tie all of this together. This is already quite long and making it longer still isn't going to help you readers very much. Read what I have linked to, and if you have questions or want clarification of certain points, ask me.

There are many chronic degenerative diseases that are associated with what is termed vascular damage. The ultimate cause(s) of this vascular damage remains obscure. I suggest that much of what is perceived to be damage is not "damage" per se, but is simply the consequence of normal active remodeling of the vasculature under chronic conditions of low NO which results in the characteristic and dysfunctional morphology observed. In other words, the vascular state is the consequence of the low NO, and that vascular remodeling processes remain intact, active and essentially fully functional. These processes are simply operating in a low NO environment where the remodeling eventually result in dysfunction vascular morphology. Correcting the low NO environment should restore normal vascular remodeling and restore normal vascular function once there has been sufficient remodeling under conditions of normal NO levels.

In other words, to some extent the "damage" is not acute or even chronic "damage", rather it is simply a consequence of long term remodeling and is to a large extent reversible if caught early enough depending on age and the tissue compartment affected. This is an important point. I think it is not appropriate to call it "damage", if the normal remodeling pathways are working properly. Secondary damage to the tissues perfused by that dysfunctional vasculature is more problematic. There may be regeneration in some tissues such as muscle and liver, but less in others such as the brain or retina. Regeneration and regulation of vascularization is a critical aspect of wound healing. If the vasculature could not regenerate itself, wounds could not heal and death would occur fairly quickly. Even in the elderly, wounds do heal, demonstrating that the regulation of vascularization is intact even if it proceeds at a slow rate.

Vascularization is a complex process requiring many coupled and interacting pathways to function successfully. When vascularization "goes bad", how many pathways are affected and in how many different tissue compartments? A few, dozens, hundreds, thousands? Presumably each endothelial cell regulates itself from internal and external signals. For many endothelial cells to "go bad" simultaneously and in characteristic ways such that characteristic abnormalities develop over macroscopic spatial dimensions the regulation in millions of cells would have to simultaneously "go bad" in precisely "the same" way to achieve "the same" pathology over "the entire" tissue compartment (such as the retina) or systemically throughout the organism. That seems implausibly unlikely.

My hypothesis is that the observed vascular dysfunction is actually good regulation around a bad setpoint and that the bad setpoint is set external (to the endothelial cells) (the "setpoint" has to be external because what physiology is trying to "fix" is external levels of O2 and other nutrients in the local vicinity by modulating vascular perfusion) by the local NO level. Mostly the local level of NO is determined by the local NO production "source", eNOS in the endothelium and then taken up by the major "sink", hemoglobin in red blood cells. There is nothing simple or simplistic about this. The control system must have enough degrees of freedom to regulate the vasculature under each and every physiological condition that occurs, or vascular regulation fails and the organism dies. If the organism is still alive, then vascular regulation has not completely failed… yet.

My focus will be more on how low NO generates the characteristic morphology and what the long term consequences of those changes are rather than on other factors. I am not going to go into the detailed mechanisms of what regulates vascularization. I am going to more focus on the macroscopic control system which must exist even though the details are not very well understood. The experimental techniques necessary to measure and understand those details are quite challenging. They must involve gradients of diffusible (and so necessarily small) molecules over the smallest length scales of the vasculature. There are essentially no sensors to measure the small diffusible molecules that must be involved on the length, time and concentration scales involved. So long as those control systems use NO as a signaling molecule, a change in the basal level of NO will affect that control loop by skewing it in a characteristic direction.

The characteristic vascular remodeling I am most concerned about occurs "at rest", that is under the physiological state that organisms are in most of the time, and during sleep. During non-rest there are other pathways that provide additional regulation of the vasculature but because remodeling at rest occurs, the pathways that are involved at rest must be sufficient to mediate the remodeling.

I spend a lot of time talking in generalities, talking of the philosophy of design and control and how that must be applied in physiology. This is not the standard "hypothesis-experiment-conclusion" that is the standard methodology in science. The reason I don't take that approach is that the systems are too complex; they consist of too many coupled non-linear parameters for which there are no techniques to measure, let alone measure on the length and time scales that must be important to regulate vascularization. If we limit ourselves to what is possible to measure, the vast majority of physiology is completely inaccessible. Even though we can't measure parameters, we know that they must have certain properties due to stability considerations; certain properties because they evolved; certain properties because mammals gestate in utero under very different O2 levels and so on. The degree of certainty from such inferences isn't as high as from actual measurement, but the point of this argument is not to "prove" that low NO causes vascular abnormalities, but rather that NO is involved in "enough" pathways that having the "right" basal NO level must be important. How important? That is a good question and one that can only be answered experimentally. My purpose is to demonstrate that there is sufficient a priori justification to test the hypothesis that raising basal NO will improve vascular abnormalities. If my hypothesis about basal NO is correct, then raising basal NO will have no adverse effects because it simply restores a more normal state where all regulatory pathways work better.

We don't have the ability to measure basal NO levels, and even if we did measuring them in vivo would be to invasive and cause too much injury to be done ethically. There are so many different tissue compartments and so many different pathways that utilize NO as a signaling molecule, that the idea of external control through artificial means is completely preposterous. Because there are so many NO mediated pathways, it is very likely that different individuals will have different and to some extent idiosyncratic thresholds for some of them. What this means is that while there may be some generic symptoms from the effects of low basal NO, there will very likely also be idiosyncratic symptoms. Eventually as basal NO gets lower and lower, more and more NO mediated physiological pathways become marginal and eventually will "fail". As more and more NO mediated pathways become disrupted the course of many degenerative diseases gets closer and more similar. For example, as end stage kidney failure gets worse, vascular disease gets worse too and kidney failure is a common complication of vascular disease. I think the commonality of vascular abnormalities with many degenerative diseases is due to low NO being one of the final common pathways involved in them all.

I am a chemical engineer, so I will talk about and outline my reasoning about this in engineering terms, sensors, control systems, feedback, and that sort of thing. This is not to imply any type of design as ID proponents consider it. That is simply how I think of and understand physiology, just the same as any other chemical plant, a very complex chemical plant with exquisite controls most of which we have no idea how they work, what the design parameters are, what parameter is being controlled to do what and under what circumstances. Hacking into the control system of a complex chemical plant and perhaps bypassing the safety systems (when you don't know what is/is not a safety system) is not something to be done casually. That is how I see a lot of medicine, the same as trying to hack into a chemical plant where none of the pipes or wires are labeled as to what they are carrying, where they are coming from or where they are going, you don't know what the control system is based on, or even what is being controlled. Is that high pressure gas line carrying substrate for a process, a pneumatic control line, a pneumatic power line, a heat transfer line, or does it carry pneumatic messages inside tubes or some of all of these? We know that there is a gigantic amount of redundancy, and that there are so many different levels and layers of control that disabling all of them is difficult. Difficult but not impossible. The most important thing to remember in all of this is that these control systems have one and only one goal, the survival and reproduction of the organism. If for some reason we think a control system looks like it is doing something else that is probably a mistake on our part. A mistake based on the arrogance of our virtually complete ignorance of physiology.

I make a big point of how complicated physiology is, and how little we know about it because most people don't appreciate how complex it really is. The human genome has been sequenced (for some individuals), but the function of the vast majority of the DNA remains completely unknown. We don't know the complete function of any single protein; let alone how it interacts with thousands of other proteins under diverse circumstances.

A lot of what I cite is animal and in vitro research. I appreciate that in vitro and animal studies are not sufficient to base treatment in humans on. However there are no clinical trials where NO levels have been measured (there are no techniques that are sufficiently non-injurious to do in humans in vivo). There are no clinical trials where NO levels have been increased systemically (there are no generally recognized techniques to do so). With no techniques to measure basal NO and no techniques to increase it, some might say it can not be important. That would be wrong. 100 years ago there were no techniques to measure insulin levels in blood and no techniques to increase insulin levels either. Diabetes had been known for millennia. Eventually it was determined to be caused by insufficient insulin and that it could be effectively treated by supplying insulin from an external source. Diabetes in ancient Greece was caused by the same lack of insulin that modern diabetes is caused by.

Most of what I am talking about is quite generic to mammals. The "details" might be different, but the broad brush regulatory pathways are essentially the same. Those details are important for developing therapies specific to humans, but that is not my goal here. My goal is to illustrate how improved regulation of basal NO would improve regulation of vascularization. I don’t need to know the precise basal NO setpoint for a particular individual to know that bringing it closer to "normal" will improve the regulation of pathways mediated by NO.

We know that vascularization has not completely "failed". Even in the worst of circumstances wounds still heal although very slowly. The vascularization pathways are still working, still trying to maintain function, but they are having a very difficult time doing so. What ever is causing that must be a systemic agent that is at the heart of the regulation of vascularization. That agent is basal NO. Increasing basal NO won't act as a drug; it will simply allow normal physiology to reassert the control it evolved to do.

Something as important as basal NO is regulated, it is highly regulated. The problem is that our modern lifestyle has broken a critical component of that regulation. That critical pathway in NO/NOx physiology is the generation of NO and nitrite on the skin by a commensal biofilm of ammonia oxidizing bacteria due to release of ammonia in sweat.

Types of vascular effects covered.

General regulation of the vasculature via NO: vascular tone, blood and lymph flow, hypoxia, anemia, vascularization
Capillary rarefaction: hypertension, Raynaud's Rheumatoid arthritis and systemic sclerosis, Fibromyalgia
Acute changes: triggering of ischemic preconditioning à long term remodeling Retinopathy: Diabetic, hypertensive, tortuous vessels
Brain: Migraine, White matter hyperintensities, vascular dementia, reduced perfusion secondary to migraine, brain atrophy secondary to reduced perfusion
Diabetic vasculopathy: Diabetes type 1, metabolic syndrome, cause of vascular damage, cause of peripheral injury, nerve, poor healing, infections.

General Regulation

I have covered white matter hyperintensities in an earlier blog. I see that as the shut down of long range axonal transport due to low ATP as regulated by low NO. I will write about fibromyalgia in a later post. I did write about Morgellons, which I see as very similar to fibromyalgia except more in the skin and with greater systemic effects (likely because the skin is a big organ, and when a big organ is subjected to low NO stress, it affects a lot of things). I plan to write a future blog focused on fibromyalgia, so I won't spend as much effort on it as it deserves.

The engineering truism, "Good, Fast, Cheap; pick any two" is how evolution has configured physiology. Good only means good enough to survive and reproduce. Fast relates to the time scale of the organism's needs, when running from a bear, speed is of the essence. Cheap relates to the overhead in terms of metabolic cost and forgone reproduction. Every aspect of physiology has evolved as a compromise and trade-off between immediate costs and how that affects ultimate reproductive capacity. Regulation of blood flow is no exception. Every different aspect of vascular regulation is part of a unified physiology. The regulation of blood flow occurs over many different time scales (from seconds to months), under each and every different physiological state that a living organism can achieve (if regulation of blood flow ever "failed", the organism would die).

It is wrong to try and think about different aspects of that regulation in isolation, but only as part of the entire system. We know that the regulatory system evolved, and that every ancestor with a vasculature had a viable regulatory system (that would be every vertebrate ancestor). That constrains its properties considerably. This evolution probably explains a lot of the redundancy and robustness of physiological systems. When a new and improved pathway develops, the DNA encoding it becomes common in the gene pool because it provides improved survival and reproduction. The old pathway doesn't get removed; it stays there and is turned off, turned down, or inhibited, or relegated to a secondary, tertiary or later role. The "cost" of maintaining the DNA associated with a pathway now made obsolete is very small, the metabolic cost of maintaining DNA is that of the organic molecules that make it up, the metabolic cost of the enzymes that keep it repaired, the risk that a mutation in it will lead to something bad. Tiny risks compared to the risk that the newly evolved pathway won't always work in every situation that the old pathways worked in and which might be needed in some extreme circumstance, just in case. This is one of the difficulties in studying systems such as the vasculature. They are highly redundant and can tolerate quite large perturbations until they "break". When they do "break" the organisms dies. The transition between seemingly normal function and dysfunction leading to death can appear to be quite abrupt. It may be difficult to tell just how abrupt those changes actually are without knowing in detail exactly what is going on.

It is also wrong to try and apply any kind of "linear" model to physiology. Some times you can, but usually only for a small part of the actual dynamic range that physiology actually works in. Nothing in physiology is linear or continuous. Genes are expressed or not expressed, and discrete numbers of molecules of proteins are produced. Those proteins interact (or not) with discrete other proteins. Sometimes you can use a continuum model, but there are only 2 DNA molecules that the transcription factors can regulate. At the level of gene expression physiology is quite discrete.

The Good, Fast, Cheap tradeoff in human designed systems is somewhat different. Human labor in the design is often the largest part of the cost. Simplifying the design is done to minimize the design labor component. That always increases the cost and the response time. A typical heuristic for reducing design cost is to apply a "factor of safety". It is cheaper to simply use 3 times more structural material (as the ASME Boiler code calls for) than to calculate "exactly" what is needed, to make the system controls sufficiently precise and reliable such that a lower factor of safety is tolerable, or to tolerate occasional failure. Modular design is good engineering practice because it makes for easier design, easier debugging and easier coordination of human design efforts. None of those tradeoffs are important in living organisms, so what is perceived to be modular design is actually an artifact of systems that have evolved. Adopting a "physiology is modular" heuristic in examining evolved systems is fraught with potential error. There is no "evolutionary pressure" for organisms to evolve physiology in modules, only to evolve physiology that preserves the life and reproductive capacity of the organisms. Physiology may have the illusion of being modular because existing pathways can become duplicated and the now redundant pathways can diverge to accomplish different but similar tasks.

As human designed systems "evolve" down their "learning curve", more effort gets put into design. The design effort per unit may go down, but as the number of units goes up, the total design effort can become enormous. For a technology that is fully mature, the major cost is the raw materials. Evolved systems such as living organisms can in some ways be considered a "fully mature" design, that is where the engineering trade-off of "good, fast, cheap" has been "optimized" to a particular evolved value. That value may not be the value that we want, because it represents the optimization that occurred over many thousands of generations, mostly under conditions quite different than modern life. Our DNA didn't evolve for us to live happy lives. It evolved only because each and every one of our ancestors survived and reproduced, sometimes under quite horrific conditions.

General regulation of the vasculature via NO

There is an ok review that covers some of the basics.[1] There are a few misconceptions in this review. They make the very common error that too much NO is bad and that too much NO causes the formation of peroxynitrite (from NO and superoxide). This is simply incorrect. Peroxynitrite only occurs in vitro when there is near stoichiometric formation of NO and superoxide. [2] This is virtually certainly the case in vivo, but in vivo is considerably more complicated because superoxide (and peroxynitrite) is always confined by lipid membranes. They are both anions, and lipid membranes are impermeable to anions except through anion channels. Superoxide from mitochondria is confined to the mitochondrial inner matrix. Peroxynitrite is similarly confined. Peroxynitrite can decompose into NO2, and NO2 can diffuse through lipid membranes.

They are correct that it is a balance. They don't seem to appreciate how much feedback and crosstalk there is between NO and oxidative stress. NO and superoxide are very much complementary physiological principles. They are analogous to the conjugate variables of quantum mechanics, to the complementary principles of Yin and Yang, male and female, hot and cold, light and dark. These are only analogies, and shouldn't be taken literally. NO and superoxide react at near diffusion limited kinetics, as fast as it is possible for chemical species to react with each other. It is not possible to have both NO and superoxide present simultaneously. Which ever one is in excess will destroy the other. Because NO is lipid soluble (~10x over aqueous in isotropic lipid (lipid membranes are not isotropic, so that is more complicated still)) and superoxide is not, lipid membranes confine superoxide but not NO. This allows NO and superoxide to (nearly) co-exist in close proximity, provided the enzymes providing the NO and superoxide are kept separate (although nitric oxide synthase does make both NO and superoxide as discussed below).

A great many sources of both NO and superoxide are co-regulated by NO, superoxide and peroxynitrite. For example nitric oxide synthase generates both NO and superoxide. As the L-arginine level gets low, then NOS generates superoxide and forms peroxynitrite which has the effect of modifying NOS by oxidizing a critical zinc thiol complex so that NOS becomes uncoupled, and instead of making NO and superoxide makes only superoxide. This can be thought of as the "off" switch for NOS generating NO. A pulse of superoxide from another source can drop the NO level, accelerating the production of NO, locally depleting L-arginine, superoxide is formed by NOS, peroxynitrite is generated, this uncouples NOS which generates more superoxide until the NOS in the vicinity is all irreversibly switched to making superoxide instead of NO. This "switch" changes physiology from a low oxidative stress state dominated by NO to a high oxidative stress state dominated by superoxide. This is the generic "stress" response; lower NO levels switch physiology to respond to stress. This is how mitochondria respond, this is how ischemic preconditioning is triggered, this is how the respiratory burst is triggered, this is what mast cells do, release proteases to switch xanthine oxidoreductase to generate only superoxide. Low NO makes the threshold for all of these switches lower. In the limit, the threshold becomes so low the cells are only in the oxidative stress state. This can be sustained for considerable time, depending on the tissue compartment. It cannot be sustained indefinitely in all tissue compartments without adverse effects in multiple systems. Characteristic vascular remodeling is one of those adverse effects in the vasculature.

There is hysteresis when an organisms or tissue compartment enters an oxidative stress state. Usually stress states are conditions of high metabolic demand, it is advantageous to minimize the metabolic resources necessary to maintain the organism in that state, to free up those resources for productive use.

Peroxynitrite damage occurs due to slow turn off of oxidative stress

Peroxynitrite is a normal signaling compound. Peroxynitrite only occurs at near stoichiometric levels of NO and superoxide. Peroxynitrite effects are not observed in healthy individuals. Peroxynitrite damage doesn't occur in low NO states, it also doesn't occur in high NO states. Presumably peroxynitrite effects occur during the transitions of physiology, during the switching transients; from a superoxide dominated state to a NO dominated state and/or from a NO dominated state to a superoxide dominated state. We know the transition from a NO dominated state to a superoxide state is rapid and exhibits hysteresis. The main NO generating enzymes are turned off by peroxynitrite. The zinc thiolate couple in NOS becomes oxidized which decouples NOS so it produces only superoxide, similarly the Mo-thiol couple in xanthine oxidoreductase becomes oxidized so it no longer reduces nitrite to NO but only generates superoxide from O2 and reducing equivalents.

Presumably the damage observed and attributed to peroxynitrite occurs while physiology is attempting to switch from a superoxide dominated state to a NO dominated state. This requires sufficient NO to overcome the hysteresis of the low NO state. If there isn't enough NO, the transition cannot occur crisply, and physiology stays longer in the state where it generates the near stoichiometric levels of NO and superoxide that cause peroxynitrite damage. The solution to this ineffective and slow switching is to increase basal NO levels so that the transition can occur more rapidly and more robustly.

This is an extremely important point. The presence of peroxynitrite damage is not due to too much NO. Peroxynitrite damage is due to there being not enough NO (except under very rare those circumstances and the problem then isn't too much NO but not enough ATP see the blog on mitochondria damage).

The switching from the NO state to the superoxide state can occur very quickly. If you need to run from a bear, release of epinephrine causes acute oxidative stress. Revving up metabolism takes some time. Mitochondria need to disinhibit cytochrome c oxidase, the heart needs to start pumping blood at a high rate, get the liver putting out glucose at a high rate, the pancreas putting out insulin at a high rate and the lungs supplying O2 at a high rate. ATP cannot be stored. There is hysteresis in systems supplying ATP; ATP must be generated as fast as it is used. An analogy would be a pipeline which has inertia. You can't turn on and off a large pipeline instantaneously. The same is true of ATP. When it isn't needed, but might be in a few seconds, mitochondria get ready to generate it but dissipate the mitochondria potential as heat instead of generating ATP. This wastes substrate, but it is more important to be able to ramp up ATP production in a few seconds and escape with injury than ramp it up slowly and get caught. This readying of physiology to supply ATP at high rates is known as the "fight or flight" state.

When ATP is needed at high rates as in "fight or flight", an optimized organism would shut off ATP consuming pathways that are not needed during that time. If that time is brief, those longer term systems can be turned back on. If the time is prolonged, then what ever those pathways are supplying is lost until they are turned back on. Physiology can turn on the fight or flight state in a few seconds, it takes much longer to stand down from it.

A fundamental aspect of the damage that occurs from chronic low NO occurs because of the chronic activation of the "fight or flight" state. The fight or flight state evolved to be a temporary state. An emergency overload state where some necessary metabolic functions are put off until later to save ATP for immediate consumption. A state where damage is tolerable to save the life of the organism.

Modulating ATP demand over time is an extremely important physiological process, and one which is insufficiently appreciated because it is so automatic, so universal, and goes to such deep evolutionary time that all organisms exhibit it. The reason all organisms exhibit it is because is reduces the "overhead" associated with the production of ATP. That overhead includes the molecules that make up the ATP generating apparatus, the additional muscle to carry those extra muscles around. "Just in time" ATP generation allows those extra molecules to be used for reproduction instead. Over evolutionary time ATP allocation has evolved to be very efficient. This allocation of ATP is what occurs during fight or flight, it is what occurs during ischemic preconditioning.

Because peroxynitrite is a normal signaling compound, there will always be peroxynitrite effects, there will always be peroxynitrite "damage". Some amount of "damage" is unavoidable and physiology has evolved systems to deal with unavoidable amounts of damage. That damage isn't repaired during the low NO state because physiology is doing other things with the ATP, such as running from a bear. Repairing damage has too low a priority. The damage accumulates until there is a high NO state during which it can be repaired. Chronic low NO prevents the repair of peroxynitrite and other damage. When ever the damage rate exceeds the repair rate, damage will accumulate. The absolute rates don't matter, only that the damage exceeds the repair. The problem isn't too much damage, the problem is insufficient repair.

Vascular tone, blood flow, and lymph flow

The vasculature is active tissue. Arterial and venous blood is under pressure, the pressure drop between the heart outlet at the aorta and the heart inlet at the vena cava drives the flow of blood. The cross section of the vessels is regulated locally along their length, in (very) complex ways to regulate that flow. Red blood cells carry O2 from the lungs to the peripheral tissues and carry CO2 back to the lungs. Red blood cells are confined to the vasculature. O2 diffuses from red blood cells into the peripheral tissues. All tissues obtain O2 from the blood except for the external skin. The outer few hundred microns of the external skin derive O2 from the external air. All O2 diffusion is passive diffusion down a chemical potential gradient (more on this later).

The usual lack of blood flow in the skin is easily observed because non-pigmented skin is transparent and is not seen to be red except under conditions of hyperemia. Only a small fraction of the body is in direct contact with blood, only the endothelium. All other cells derive nutrients (other than O2) from extravascular fluid, that is from fluid that has "leaked out" of the vasculature (though it is not leakage per se, it is absolutely necessary extravascular flow). It is this extravascular fluid that carries glucose to the cells, other nutrients including protein (mostly as albumin), insulin, and all other nutrients and signaling components of blood. The extravascular fluid moves much slower than blood. Virtually all cells derive glucose from this extravascular fluid. Necessarily the glucose and insulin levels in plasma in contact with cells is lower than in bulk blood because intervening cells have consumed some of it. How much lower is a good question which is difficult to answer because getting samples to analyze is extremely difficult. The glucose level in the extravascular space next to the cells that are taking that glucose up is of course a much more important parameter than what the glucose level is in bulk blood remote from the cells that are using it.

Adequate flow of extravascular fluid is just as important as adequate flow of blood. The time constant for extravascular fluid flow is longer, but is obviously important and so obviously is actively regulated by physiology. If it were not actively regulated either there would be too much, or too little, or both in different tissue compartments.

The importance of extravascular flow of lymph is not always appreciated. Because it cannot be measured easily and is different in every tissue compartment (or even in the same tissue compartment due to gradients between capillaries), it is not as well mixed as blood is, it is not routinely measured and there are no clinical correlates with it. The fluid must "leak" out of capillaries at the proper rate, and then be transported along through the lymph vessels at the proper rate and then fed back into the circulation at the proper rate.

Accumulation of extravascular fluid is known as edema. This occurs for a variety of reasons, because the flow channels are blocked (as in filarial diseases such as elephantiasis), when there is too much fluid because the kidneys can't get rid of it and it has to go somewhere (the edema of congestive heart failure) and for things such as ascites (in the abdominal cavity).

CO2 must be carried back to the lungs also. CO2 transport is pretty complicated and won't be discussed in detail. CO2 as an uncharged gas diffuses pretty well. It is water soluble and forms carbonic acid, H2CO3. There are significant kinetic impediments to the formation of H2CO3, and so there are enzymes, carbonic anhydrase that catalyze it. H2CO3 disproportionates into H+, HCO3-, and CO3(2-) depending on pH. These are charged, and so cannot penetrate lipid membranes except through ion channels. Some of these are actively ported through cell membranes. Other ions must be co-ported to maintain ion neutrality. Chloride is the ion that does that in red blood cells, but nitrate and nitrate are similar to chloride ion in a lot of ways. They are not considered that important in ion channels, so the conductance of ion channels for nitrate and nitrite are not always measured along with other ions. CO2 can diffuse from tissue compartments containing carbonic anhydrase through intervening tissue compartments that don't, and into tissue compartments that do.

Regulation of vascular tone, blood flow, and lymph flow

The primary regulation of blood flow is via regulation of the cross section of vessels carrying that blood. The heart can pump more blood, and at a higher pressure, but for that blood to go where it is important for it to go the vessels have to modulate their cross section. Vessels are dilated where blood is being regulated to go, and constricted where blood is being regulated to not go. There is limited blood and also limited blood pumping capacity, so both types of control are needed, local to increase local blood flow, and non-local to decrease other blood flow.

The major regulator of vascular diameter and vascular tone is nitric oxide. NO is produced in the endothelium by eNOS. It is the NO that diffused into the vessel wall that regulates its tone. NO activates sGC which makes cGMP which relaxes smooth muscle. NO also diffuses into the blood and is taken up by red blood cells via kinetics that are first order in NO and first order in red blood cell concentration. The major passive sink for NO in the body is hemoglobin. Hemoglobin has a very high affinity for NO, and metabolizes it to either nitrite plus nitrate or to nitrosyl heme. Hemoglobin is normally confined to erythrocytes. Free hemoglobin destroys NO ~600 times faster than does Hb in erythrocytes. Free hemoglobin is responsible for the acute constriction and hypertension associated with hemolytic anemia as in sickle cell anemia.

How much NO diffuses into the blood, into red blood cells and is consumed and swept away and how much NO diffuses into the vessel wall and is consumed by superoxide and how much is left to activated sGC and cause vasodilation is a delicate balance between NO production, hematocrit, blood velocity, redox state, lipid vs. aqueous partitioning, ATP level, O2 level, L-arginine levels, asymmetric dimethyl arginine levels, nitrite, R-SNO thiols, NO from the extravascular space and other things. We know that all of those things are important, none of them can be measured on the length and times scales that we know are important in vivo. Neurogenic or receptor mediated production of superoxide can acutely consume NO and cause acute constriction. Superoxide can also be dismutated into H2O2 which can also cause vasodilation (but that is usually at high metabolic rate, not at rest).

When hematocrit is acutely decreased (taking out blood and replacing it with cell-free fluid, plasma or starch solution) as in isovolemic anemia, exhaled NO levels increase.[3] As Hct was decreased by dilution with hydroxyethyl starch (30, 23, 17, 11 %), cardiac output rose (0.52, 0.60, 0.70, 0.76 L/min), and exhaled NO levels rose (30, 34, 38, 43 nL/min). This demonstrates that NO levels in exhaled air are coupled to hemoglobin concentration in blood. This actually makes sense because the hormone that determines when more red blood cells need to be made is erythropoietin (Epo) and Epo is regulated by HIF-1-alpha which is regulated by low O2 (hypoxia) and also high NO. Both low O2 and high NO are signals of "not enough hemoglobin". HIF-1-alpha also causes expression of VEGF (vascular endothelial growth factor) which is one of the major factors that triggers angiogenesis.

There is starting to be some appreciation that the anemia observed in many chronic diseases may be an adaptive response and not solely something pathological. Anemia increases NO levels. High hemoglobin levels will decrease NO because hemoglobin is the sink for NO. It is the product of NO concentration and hemoglobin concentration that fixes the NO destruction rate. That destruction rate equals the production rate because there is no accumulation. The NO concentration (which is what NO sensors react to) then goes inversely with hemoglobin concentration. In a number of disorders associated with anemia (especially end stage kidney failure), increasing hemoglobin levels to "normal" causes increased death rates over increasing it to somewhat less than normal. Not enough hemoglobin is bad, but not enough NO (because a high hematocrit is destroying it) is worse. Increased hematocrit had the largest adverse effect on vascular disorders. The increased death rates are not concentrated in one or a few categories, but spread out over many. I see this as evidence of how many physiological systems are dependant on proper levels of NO, and how closely coupled that level is to hemoglobin levels in blood. Another example is systemic sclerosis, the death rate is ~2x that of standardized death rates after subtracting out deaths due to systemic sclerosis, but the causes of death are spread out over multiple causes. Presumably what ever is causing the systemic sclerosis is also causing the increased death rates.

When more flow is needed, NO levels are increased.

When greater blood flow is needed acutely through a particular vessel, the velocity goes up, that shear then activates eNOS and NO is generated which causes the vessel to dilate. When tissue becomes hypoxic, NO is generated via reduction of nitrite by deoxyhemoglobin and by other enzymes. When the vessel cannot supply sufficient oxygenated hemoglobin via blood, the increased NO level becomes chronic. Increased NO level would then be the ideal signal to trigger generation of more blood vessels through angiogenesis. It turns out that increased NO does trigger angiogenesis, and blocking NOS does inhibit angiogenesis. Supplemental IP nitrite substantially accelerates compensatory angiogenesis around a blocked artery in mice. The positive effects of nitrite were observed over a very broad dose range, something like a factor of 400.

Increased NO mediates increased blood flow over time scales from seconds to weeks. Presumably these multiple mechanisms for regulating blood flow evolved from an archetypal blood flow regulation mechanism which involved NO.

Acute Regulation of blood flow by NO, not by O2

Under conditions of isovolemic anemia, blood flow increases. The "conventional wisdom" is that it is "hypoxia" that causes the increased blood flow; however that cannot be correct because there actually is no hypoxia. There is no reduction in the O2 level in either the arterial blood, or the venous return blood. With no reduction in O2 level, there is no hypoxia. With no hypoxia, hypoxia cannot be a signal for the body to use to regulate blood flow.

At rest, acute isovolemic anemia is well tolerated. A 2/3 reduction in hematocrit has minimal effect on venous return PvO2, indicating no reduction in either O2 tension or delivery throughout the entire body. At 50% reduction (from 140 to 70g Hb/L), the average PvO2 (over 32 subjects) declined from about 77% to about 74% (of saturation). The reduction in O2 capacity of the blood is compensated for by vasodilatation and tachycardia with the heart rate increasing from 63 to 85 bpm. That the compensation is effective is readily apparent. The mechanism is not. The “obvious” explanation is that “hypoxia” sensors detected “hypoxia” and compensated with vasodilatation and tachycardia. However, there was no “hypoxia” to detect. There was a slight decrease in blood lactate (a marker for anaerobic respiration) from 0.77 to 0.62 mM/L perhaps indicating less anaerobic respiration and less “hypoxia” (though lactate production occurs under oxic conditions). The 3% reduction in venous return PvO2 is the same level of “hypoxia” one would get by ascending 300 meters in altitude (which from personal experience does not produce tachycardia). With the O2 concentration in the venous return staying the same, and the O2 consumption staying the same, there is no place in the body where there is a reduction in O2 concentration. Compensation during isovolemic anemia cannot occur because of O2 sensing.

When red blood cells of dogs are replaced with red blood cells that have been fully oxidized to methemoglobin (and so cannot carry O2), compensation for reduced O2 carrying capacity of blood is greatly reduced.[4] While maintaining the same hematocrit Hct (43%) and substituting (0, 26, 47%) fully metHb erythrocytes, the cardiac output (CO) declined (178, 171, 156 mL/m/kg) while the arterial PaO2 (93, 87, 84 mmHg) and PvO2 (55, 46, 38) also declined. In contrast, when acute isovolemic anemia (Hct 40, 30, 22) was induced using plasma, compensation was much better, CO (155, 177, 187), PaO2 (87, 88, 91), and PvO2 (51, 47, 42). When anemia was induced using dextran solution (Hct 41, 25, 15) cardiac output (143, 195, 243), PaO2 (89, 92, 93), PvO2 (56, 56, 51) compensation was better still. As part of their experiments with the metHb tests, a final dilution was done with dextran to lower the Hct to 26% while still maintaining 47% metHb. Compensation was much improved with CO (263 mL/m/kg), PaO2 (86 mmHg), and PvO2 (41 mmHg) all were increased, despite lower Hct, greater O2, and less “hypoxia”. The compensatory mechanisms to deal with this “hypoxia” cannot be due to reduced O2 levels because the O2 levels were not reduced, in fact, the O2 levels were increased. MetHb does bind NO, not quite as well as does Fe(II)Hb, but the presence of metHb erythrocytes clearly adversely effects compensation. The authors attributed the increased cardiac output to reduced blood viscosity in the case of reduced cell concentration. However when viscosity is increased, blood flow does increase.

When blood viscosity is increased during acute anemia, NO levels increase, flow mediated vasodilation increases and flow increases.

The optimum hemoglobin concentration for O2 delivery is as low as 15%. For O2 delivery to the brain it is about 30%. Normal hemoglobin levels are ~44%.

In summary, when the O2 carrying capacity of blood is reduced by removing erythrocytes, there is essentially complete compensation over a wide range by increased blood flow such that reduced O2 levels never occur. When the O2 capacity of blood is reduced by oxidizing hemoglobin to methemoglobin, there is much less compensation and reduced O2 levels do occur. When viscosity of blood is increased, there is increased shear, increased NO production and increased flow.

Long Term Regulation of vascularization

If acute regulation of flow of blood in the vasculature is not regulated by O2 levels, but is regulated by NO levels, should we expect that physiology uses a control system operating over a different time scale utilizing a different control scheme for other regulation?

Using a different control system presents potential difficulties when transitioning from one control scheme to the other. For the control to be stable, there can't be control regimes where one system is calling for more and the other system is calling for less. The control needs to be monotonic when averaged over a period that is long compared to the response time. We know that the control system evolved. In virtually all cases evolution takes an existing pathway or structure and modifies it for improved or different functionality.

If we know that acute increases in blood flow are mediated through increased NO, and we know that some instances of angiogenesis are increased by increased NO, it is likely that increased NO is the generic control system used for regulating vascularization.

Regulation of vascularization is a critically important physiological effect, and it is regulated exquisitely well and exquisitely complexly. Some of the details are known, many (probably most) are not. The vasculature is "well formed", that is it is very closely matched to the physiological needs of the tissue compartment it is in. There is no great excess of vessels and no great deficiency either. Organisms grow from a single cell, and have a well formed vasculature at all sizes. Organs grow in size, organs also shrink.

For vascularization to be regulated over so many orders of magnitude in size in so many different tissue compartments as organs grown and regress, there must be at least two types of regulation. There must be a mechanism that senses when there is not enough perfusion in a tissue compartment and signals the generation of more vasculature. There must also be a mechanism that senses when there is excess vascularization in a tissue compartment and ablates that excess. We know that there must be at least those two mechanisms. No doubt there are others, but I will focus on those two. It may also be a single mechanism operating in different regimes. I think a single mechanism is the most likely, that mechanism being NO with high NO triggering angiogenesis and low NO triggering ablation of vessels.

When are those signals generated?

They must be generated "at rest". Most growth occurs "at rest". The vasculature of organisms remains well formed after long periods of rest. The blood flow through some tissue compartments doesn't change much between periods of activity and inactivity. In utero, the fetus is always "at rest", capillary spacing is and must be regulated equally well in utero. Actually it must be regulated better in utero than after birth. A fertilized egg increases in size by many orders of magnitude very rapidly. An infant increases in size only about 1 order of magnitude as it becomes an adult and over a much longer time period. At rest would be the ideal time to fix the basal capillary spacing. Metabolic demands are low and constant. The appropriate level of vascularization could be established with the proper excess safety factor. At rest is a good time to remodel important physiological systems because metabolic need for those systems is at a minimum. While running from a bear is a bad time to divert resources into remodeling active systems. There may be additional signals that occur at other times (such as during exercise), but I will focus on the one(s) "at rest" which presumably are involved in all tissue compartments and so is likely to be the archetypal signaling system.

We know that hypoxia is involved in regulation of vascularization via HIF-alpha. Cells not getting enough O2 could generate a signal to generate more vasculature to bring more oxygenated blood to that tissue compartment. How can excess vasculature be measured? It cannot be measured by O2 level. The O2 level in arterial blood is very close to the level in air. That is the level in tissues at rest. At rest, the O2 level is essentially independent of capillary density. O2 demand is low, there are no large gradients in O2 concentration. All arterial blood is at near the saturation level in air. Venous blood is also regulated to a fairly constant O2 level. Gradients in O2 concentration between blood and mitochondria (where O2 is consumed) are in the extravascular space, not in the vasculature.

It also needs to be remembered that when embryos regulate capillary density they do so with quite different O2 partial pressures than do ex-utero organisms.

Physiology needs to generate a signal that measures how diffusively close red blood cells are to cells that require O2 and other nutrients in blood, that is, is a particular cell close enough to red blood cells such that it can obtain enough O2. If there are enough red blood cells close to a cell, that cell can indicate when it has sufficient vascularization and when there is not enough.

I suggest that an important component of that signal is NO. NO has physical properties very close to that of O2, the diffusion of NO through tissues is virtually identical to that of O2. O2Hb is the sink of NO, so the vasculature has the lowest NO level in the body (neglecting formation of superoxide (which is generated in mitochondria and microsomes) for the moment). If there was a volume source of NO, the basal NO level would be higher the farther from a capillary that tissue compartment was.

In summary, there must be a signal by which insufficient vascularization triggers angiogenesis, and also a mechanism by which excess vascularization is ablated. NO can signal both instances due to O2Hb acting as the sink for NO and with the extravascular space acting as a volume source.

Hypoxia and NO activate HIF-alpha which causes the expression of VEGF which is important in angiogenesis. NO is known to be important in angiogenesis, expression of iNOS is important in angiogenesis surrounding vascular infarcts. Neurogenic release of NO is what causes vasodilation by activating sGC. If the neurogenic NO were not sufficiently swept away by enough O2Hb, it would be a good signal for angiogenesis.

Regulation Oxygen delivery, Oxygen extraction, Ischemia reperfusion injury

Physiology can only use what are called intensive properties, properties that are proportional to the concentration or chemical potential of a substance, and not extensive properties, properties that are dependant on the amount of substance available. O2 partial pressure is the same as O2 chemical potential. O2 partial pressure is not proportional to the O2 content of blood because hemoglobin has a non-linear O2 dissociation functionality. O2 partial pressure is proportional to O2 content of plasma because plasma does have a linear O2 dissociation functionality (actually it is simple solubility via Henry's Law).

There are numerous misconceptions about this regarding O2 delivery, O2 extraction and the blood. O2 only moves by passive diffusion down a gradient in chemical potential. In homogeneous media this is down a concentration gradient via Fick's law of diffusion. By homogenous media I mean media where the chemical potential is strictly proportional to the concentration. In non-homogenous media such as blood or mixtures of lipid and aqueous phases one has to be careful. The "concentration" of O2 in red blood cells (mL/L), is not the same as the "concentration" of O2 in plasma in equilibrium with those red blood cells. The chemical potential of O2 in red blood cells and in plasma in a sample of blood is the same (and is the same in all fluids in mutual equilibrium that is the definition of equilibrium). In hemoglobin there is a non-linear relationship between O2 partial pressure and O2 concentration. Physiology can't measure O2 concentration in blood, all it can measure is O2 partial pressure, or more precisely the O2 chemical potential of the O2 sensors in equilibrium with that blood. Lipids have ~10x higher solubility of O2 and NO than do aqueous fluids. This can affect rates of chemical reactions a lot, as can the lower dielectric constant inside lipids. Ions can't enter lipids. Highly polar compounds like water or H2O2 can diffuse through lipid, but not as quickly as something nonpolar like NO or O2.

Some of the physiology literature talks about "oxygen extraction" from blood as if that is a real parameter. It isn't. Oxygen only moves by diffusion. There is no active transport. Tissues don't "extract" O2 from blood, O2 diffuses out of blood if the tissue the blood is flowing through has a lower O2 chemical potential than the blood flowing through it. If tissues are at the same O2 partial pressure as the blood, then they do not extract O2. If the tissues are at a higher O2 partial pressure, O2 diffuses out of the tissues and into the blood. There are no barriers to O2 diffusion. There is nothing that can block O2 diffusion. The vasculature can regulate where blood flows, and bypass less important organs to divert blood to more important organs. Tissues can only regulate O2 consumption by regulating the affinity for O2 of enzymes that consume O2.

Mitochondria are the ultimate sinks of O2; cytochrome c oxidase is the enzyme that reduces one O2 to two H2O's. The binding coefficient (Km) of cytochrome c oxidase for O2 is a sensitive function of the NO level. NO binds to cytochrome c oxidase and inhibits the binding of O2. This is an extremely important regulatory system for mitochondria. It is by regulating the NO level that the affinity of mitochondria for O2 is regulated. High NO, low O2 affinity. Low NO, high O2 affinity. The generation of superoxide by mitochondria under conditions of hypoxia is then seen as a necessary regulatory function. When cells become hypoxic, their mitochondria generate superoxide, that superoxide (confined to the inner matrix!) pulls down the NO level, cytochrome c oxidase is disinhibited, binds O2 at a lower O2 partial pressure, O2 is consumed to a lower partial pressure, the partial pressure gradient between the blood vessel (where it is nearly constant) and the mitochondria (where it is consumed by mitochondria) increases and so the flux of O2 (moles O2 per second) diffusing to the space where the mitochondria is now increases, relieving the "hypoxia". The problem of insufficient "oxygen extraction" is too much NO on the mitochondria. But is that really a problem? Cells don't need "oxygen extraction", they need ATP. If cells have enough ATP, they don't need anything else. Mitochondria are not the only source of ATP. Cells can make ATP via glycolysis which does not consume O2.

It is the attempt to make ATP using O2 under conditions of very high NO during sepsis that causes the mitochondrial damage and the multiple organ failure.

Under conditions of hypoxia, mitochondria first generate superoxide, and pull the NO level down to extract as much O2 as possible. Once that O2 is exhausted, mitochondria have a different need, to prevent the production of a massive amount of superoxide if and when O2 levels are restored. Most of the damage that occurs during ischemia-reperfusion occurs during the reperfusion, not the ischemia. When all the enzymes in mitochondria are primed to make superoxide, it is a bad time to supply a large bolus of O2 because much of it can get turned into superoxide.

Preventing damage following reperfusion is I think where nitrite comes in. There are many different enzymes that reduce nitrite to NO in the cytosol, in mitochondria, in microsomes. The nitrite reductase activity of these enzymes is O2 level dependant. O2 inhibits the production of NO from nitrite by them. When the O2 level drops, they become active nitrite reductases and produce large quantities of NO. This NO binds to heme enzymes and blocks their take-up of O2. This NO blocks the formation of superoxide following reperfusion.

As mentioned earlier, damage due to peroxynitrite from NO and superoxide occurs only when both are generated at near equimolar fluxes. This occurs during the transitions from a state dominated by one to a state dominated by the other.

Regulation of nutritive blood flow

In fMRI BOLD testing, it has been observed that the quantity of blood that flows in the region activated by the neurogenic NO release exceeds the nutritive quantity needed to supply the metabolic activity of that activated region. If the normal mechanism produces blood flow in excess of nutritive needs, there is a "factor of safety" and other mechanisms regulating blood flow are not needed. This is an important point. Blood contains many factors that are needed at different levels, O2, glucose, albumin, insulin, various proteins, hormones, immune cells, cytokines, etc. The levels of many of these factors in blood change week to week, day to day, even minute to minute. The needs for each of them in a particular tissue compartment also change. How many of them can the local regulation of blood vessel tone be determined by? In principle many of them, however if there is more than one control parameter, the control system may become over specified and instabilities may occur. It is much easier for evolution to modify and elaborate on a primitive ancestral trait than to evolve one de novo. The shortest time scale need is for O2. The time scale for O2 need is seconds, the time scale for glucose is minutes, the time scale for angiogenesis is days. If NO is both the shortest time scale control parameter and also the longest time scale, it seems implausible that a different control parameter and/or control system would have evolved to handle an intermediate time scale.

If blood flow in the brain is not acutely regulated to provide for acute nutritive needs, how is it that nutritive needs are met long term? The signals that do regulate acute blood flow either have some dependence on nutritive needs, or those nutritive needs are met by always providing an excess, or the tissue remodels to reduce demand when there is not sufficient excess.

If blood flow to the brain is reduced, how will the brain remodel itself to accommodate? Presumably brain cells self-regulate and reduce their metabolic load until it matches the supply of substrates provided by the blood. Presumably this involves pruning of lesser used cells. Pruning of brain cells is observed following a stroke. As cells necrose, the inhibitory signals they produce are lost, down stream cells become disinhibited, cells become over excited and excitotoxic death occurs. Excitotoxic death strikes cells with compromised metabolic capacity. Compromised due to insufficient blood supply, compromised due to mitochondrial damage, compromised due to other damage.

Capillary rarefaction

I suggest that the capillary rarefaction observed in many disorders, systemic sclerosis, Raynaud's, hypertension, dilative cardiomyopathy is ultimately caused by normal vascular remodeling via the same mechanism that leads to reduced blood flow, that of low NO. Reduced blood flow is observed in many neurological disorders, many even before there are overt signs of neurodegeneration. If blood flow is regulated to be chronically low, presumably efficient regulation of vascularization would ablate the seemingly excess vasculature that is seemingly present, resulting in capillary rarefaction. It needs to be appreciated that this is the completely normal response to a reduced perfusion setpoint. Physiology can't "compensate" because it is the compensatory pathways that are actually doing it.

If there is capillary rarefaction in an organ such as the heart, how will it respond? There isn't sufficient blood supply to maintain normal cell density, the cells "too far" from a capillary become stressed, die, and are cleared. If they are replaced, the replacement cells are insufficiently supplied also. The space could become filled with non-metabolically active fibrotic tissue. The heart still needs to pump sufficient blood, so it gets bigger, but weaker as fibrotic tissue replaces muscle. I think this is what eventually leads to dilative cardiomyopathy.

If the liver doesn’t have enough mitochondria to dispose of reducing equivalents, what does it do with them? Usually it makes fat. I think that is the source of fatty liver from chronic alcohol consumption. Alcohol is metabolized by alcohol dehydrogenase which makes NADH which can only be disposed of in complex I in mitochondria, or by making lipid. Ectopic lipid is an end stage symptom of many degenerative diseases associated with vascular abnormalities, liver failure, kidney failure, dilative cardiomyopathy. Excess NADH makes superoxide, and that superoxide lowers NO levels. Acutely that is adaptive, in that it disinhibits cytochrome c oxidase and allows for more O2 reduction to dispose of the reducing equivalents. In the long term, insufficient NO reduces mitochondria biogenesis resulting in systemic excess reducing equivalents which can only be disposed of by generating lipid. I think this is how the extreme obesity of hundreds of kg occurs. Individuals without enough mitochondria generate ATP via glycolysis; this generates lactate which is disposed of by generating lipid.


Hypertension occurs via increased vascular tone, the stiffness of vessels increases requiring higher pressure to drive the same volume of blood through the vessel bed.

Hypertension is associated with capillary rarefaction, with capillaries getting farther apart then is considered "normal". Capillaries farther apart means fewer capillaries in a tissue compartment and so blood flow through the remaining capillaries must be increased if the same blood delivery is going to occur.

Flow through a capillary and pressure drop across that capillary can be regulated independently. There is no need for a higher pressure drop to drive more blood through a capillary, the capillary could increase in cross section and accomplish the same thing. However bulk flow of blood through a capillary is not the only requirement. As mentioned before, extravascular flow of plasma is equally important (but on a different time scale, minutes as opposed to seconds). Extravascular cells derive nutrients only from local flow of plasma through the extravascular space. This plasma "leaks" out of the capillaries (however it is not "leakage" per se, it is a required flow). That "leakage" is likely proportional to the surface area of capillaries in that tissue compartment and also to the pressure drop from the inside to the outside of the capillaries. As capillary rarefaction reduces the number of capillaries, the total cross section goes down, to supply the same flow of extravascular fluid the pressure would have to go up.

I suspect that increasing extravascular flow is the physiological reason that blood pressure increases. Increased pressure drop through capillaries increases flow through the extravascular space that bypasses those capillaries. I suspect that increased extravascular flow is required to compensate for capillary rarefaction, both to supply the same extravascular flow, but also to increase extravascular flow to compensate for fewer mitochondria and greater ATP from glycolysis (which requires 19x more glucose for the same ATP production). NO is what triggers mitochondria biogenesis, so low basal NO will result in fewer mitochondria and more ATP from glycolysis necessitating increased extravascular fluid through fewer capillaries requiring higher pressure.

Raynaud's Syndrome

Raynaud's occurs when exposure to cold causes acute constriction of blood vessels in the skin, leading to pain, and in some cases necrosis. It is sometimes one of the first symptoms of capillary disorders, and usually accompanies all the others.

Because the output of the heart is limited, control of blood flow to peripheral tissues requires the regulation of both the pressure drop through the specific tissue being perfused, but also the pressure drop in the rest of the vasculature. The skin is a large organ, and constituting the outside surface, the skin is the only place where heat can be dissipated. Heat is brought to the surface by the blood and is easily observed as flushing. The external few hundred microns of skin derive O2 from the external air, they receive all other nutrients from the extravascular flow of plasma.

Conserving heat by constricting blood vessels in the skin when the skin is too cold is an essential part of maintaining the proper internal body temperature. Constricting blood vessels is usually done by generating superoxide, destroying NO and causing a reverse of the vasodilation that NO is producing.

Tortuous vessels

It is the take up of NO by hemoglobin in blood cells that results in the particular morphology of low NO damaged vessels, what is called "tortuous" vessels. There are some cases of familial tortuosity. This tortuosity is produced by essentially the same mechanism that stream meander is produced by. At high velocity there is erosion of the stream bank, and deposition of material at regions of low velocity. The same characteristic mechanism occurs in vessels but via different mechanisms. The crucially important similarity is that the flow of fluid inside the vessel affects the morphology of the flow pattern. In a stream the flow is within the stream banks. In blood vessels the flow is within the vessel.

Red blood cells are denser than plasma, so when there is accelerated flow there is segregation to the outside of the curved flow. This reduces the thickness of the boundary layer along the endothelium and increases the removal of NO at the high velocity outside region by the hemoglobin in the red blood cells that are now closer to the wall (just like in a stream meander). Removal of NO reduces the NO level on the inside and outside of the vessel, and this causes regression of that tissue via apoptosis due to low NO. This regression at high velocity allows for a vessel with an isotropic high velocity to enlarge in diameter. The tissue outside the vessel must regress so as to allow space for the vessel to enlarge in diameter. A series of images that just scream "low NO induced apoptosis" is here.[5] The images are of brain sections at autopsy of an individual with white matter hyperintensities. The vessels show a characteristic "tortuous vessel inside a cavity". The vessel is highly tortuous and the surrounding tissue has regressed leaving a cavity. Since there is no scaring, and there are markers characteristic of apoptosis, presumably the vessel is either a source of a pro-apoptotic diffusible factor, or a sink of an anti-apoptotic factor. It turns out that vessels are sinks for NO, which is an anti-apoptotic signal. The tortuousness of the vessel relates to how the flow changes the local source-sink properties of the vessel.

These blood vessels are often tortuous and appear as a tortuous arteriole in a cavity. These images are quite striking. The vessel is quite corkscrew-like, very tortuous and is in an empty cavity, devoid of white matter. These are sites of apoptosis. Low NO causes the tissues outside the vessel to regress (via apoptosis) and so the vessel migrates in that direction. Tortuous vessels like this are easily seen in retinopathy (where they accompany WMH) where they are caused by the same mechanism. In retinopathy, often when vessels cross there is observed to be "nicking", that is, a reduction in the diameter of the vessels. This is due to the decreased NO at the site of crossing (my hypothesis) where there is more hemoglobin to act as a sink of NO. This migration and remodeling of blood vessels by local NO levels is part of the normal regulation of capillary spacing (my hypothesis).

As cardiovascular risk factors increase there is decreased vascular reactivity; [6] that is there is reduced responsiveness shear induced increased blood flow, of exercise induced hyperemia, and even of NO induced hyperemia. This is what would be expected if basal NO levels are reduced because it is generation of NO by the endothelium that activates sGC and generates cGMP which relaxes the vascular smooth muscle. With a lower background level of NO, it takes more neurogenic NO, more shear generated NO, or more NO donor to achieve the same cGMP level and the same level of vasodilation.

Reductions in NO mediated vasodilation are observed in aged rats.[7]

These observation of long term remodeling of vascular morphology suggest that the remodeling is coupled to NO physiology, and that the basal level of NO is important in that regulation.

White matter hyperintensities

The tortuous vessels and cavities apparent on autopsy around those tortuous vessels indicate loss of neuronal tissue. There is also generalized reduction in capillary density observed in white matter hyperintensities,[8] which can be considered to be capillary rarefaction in the brain.

WMH are also observed during seizures. On acute occlusion of cerebral arteries, WMH occur very rapidly, in 2.7 minutes in the rat. It has been suggested that edema is the cause of WMH, however edema does not occurs this quickly, but ATP depletion does.

There may be other changes that result in WMH too. WMH are associated with markers for hypoxia. In any case, the relevance of this diversion into WMH is only to connect WMH to the ATP status of the brain. The association of WMH with low brain ATP is pretty well established, even if the mechanism(s) for that association is not.

NO and superoxide from iNOS are protective against excitotoxic injury, and this protection can be induced via LPS treatment.

Subjects with WMH have reduced density of blood vessels in regions which show hyperintensity. The reduced blood vessel density is likely due to a remodeling of the vasculature due to chronic low NO. With O2Hb being the sink for NO, if the level of NO is lower, then less hemoglobin is needed to act as a sink. I think this remodeling of the vasculature is the mechanism behind the lower brain blood flow observed in all of the neurodegenerative disorders characterized by WMH. I think it is also the mechanism for capillary rarefaction in non-neuronal tissues observed in hypertension and other disorders.

Ischemic preconditioning in the Brain

Ischemic precondition is a lower ATP state, but more importantly is a lower ATP consumption state. Some aspects of ischemic preconditioning are the same as the fight or flight state. Not enough is known about both to know if they are identical. They might be in some tissue compartments and not in others. All mammalian cells are aerobic and require continuous supply of O2 and substrate for continued ATP generation by mitochondria (except for red blood cells). When cells are deprived of O2 and substrate, they undergo ischemia and become damaged and eventually will die depending on the severity and length of the ischemia. This is the source of the injury when a vessel is occluded in the brain or heart for example. Ischemic preconditioning occurs when a tissue compartment is exposed to brief periods of ischemia prior to a prolonged severe ischemia. In an ischemic preconditioned state, tissues can survive ischemia that would otherwise cause necrosis. It only takes a few brief instances of ischemia to trigger the ischemic preconditioned state, which then persists for variable lengths of time, but it can be as much as a day or longer. The standing down from the ischemic preconditioned state takes longer.

Ischemic preconditioning can be triggered in a few minutes, and persists for hours to days. In the ischemic preconditioned state cells use less ATP. Presumably if cells could survive/reproduce while in the ischemic preconditioned state they would have evolved to do so because it would then free up more ATP for reproduction. Cells did not, so there is something incompatible with long term survival/reproduction with being in the ischemic preconditioned state too long. Presumably the time period that is "too long" depends on the tissue compartment and is probably longer than the normal duration of the normal ischemic preconditioned state.


Migraine is a characteristic episodic type of headache that is often localized to a portion of the head, is sometimes preceded by visual hallucinations called aura or pro droma, and is sometimes triggered by a number of different environmental and/or physiological circumstances. The details of the physiology behind migraine are not well understood.

There has been considerable work on migraine using nitroglycerine because nitroglycerine does reliably induce migraine in susceptible patients. It is unfortunate that the effects of nitroglycerine on migraine have been attributed to nitroglycerine being a "NO donor". Nitroglycerine is not a "NO donor" in the classic sense. The chemistry and physiology behind the effects of nitroglycerine are complex and are not well understood. It can be a source of NO and nitrite via chemistry which is not fully understood, and which is subject to significant changes in fairly brief periods of time (few hours). Nitroglycerine exhibits what is termed "nitrate tolerance", where the dose of nitroglycerine must be increased; other NO donors such as sodium nitroprusside or authentic NO does not cause nitrate tolerance.

Nitroglycerine does irreversibly inhibit aldehyde dehydrogenase which appears to be the main enzyme responsible for generation of NO. This irreversible inhibition is exacerbated by oxidative stress. Nitroglycerine does induce late ischemic preconditioning. Ischemic preconditioning is a state where ATP concentration and consumption is reduced; it is a state that is protective in the short term, but (my hypothesis) detrimental in the long term. Long term treatment with organic nitrates increases cardiac events in patients with healed myocardial infarctions. I suspect that the therapeutic mechanism of nitroglycerine may be to induce ischemic preconditioning pharmacologically. This may be useful at reducing pain, and in reducing acute injury, but may not be helpful in the long term. That may be why some groups see increased cardiac events in long term treatment with nitroglycerine.

Migraine induced by nitroglycerine is not associated with changes in brain perfusion. This article is quite interesting and goes against a lot of conventional thinking and assumptions. It is consistent with the idea that migraine is not caused by vasodilation associated with NO. They did observe the prompt vasodilation associated with acute infusion of nitroglycerine, however there was no vasodilation associated with migraine following the nitroglycerine.

Migraine is pretty reliably triggered by sildenafil (Viagra). Sildenafil inhibits the phosphodiesterase 5 that is the main esterase that removes the cGMP produced by sGC after it is activated by NO. This is the mechanism by which sildenafil potentiates the action of NO through the cGMP pathway. However because there is feedback inhibition of NO production through the cGMP pathway, potentiating the level of cGMP will reduce the level of NO that is produced. Sildenafil thus will reduce the effects of NO mediated through non-cGMP pathways. This is apparent in men with obstructive sleep apnea, where a single dose of sildenafil significantly increases desaturation events. One of the triggers for breathing is S-nitrosothiols and is not mediated through cGMP.

When migraine is visually triggered, there is increased O2 levels by fMRI BOLD. This has been generally interpreted as being due to vasodilation, however the nitroglycerine study shows no vasodilation. I think it is more likely that the increase in O2 levels may be due to reduced O2 consumption due to triggering of ischemic preconditioning. Reduced O2 consumption is also consistent with reduced metabolism. The dynamic range of O2 level is substantially reduced, that is the level between activated and deactivated brain regions.

Migraine has been hypothesized to be associated with spreading depression. Spreading depression is a depolarization of neuronal tissue that propagates at up to a few mm/minute. It is not propagated by axons, but by some other type of signaling. It has many characteristics one would expect of ischemic preconditioning, and likely is related. This review article is interesting because it discusses both spreading depression (SD), and also hypoxia spreading depression like depolarization (HSD). I like the discussion of how they are different and how they are the same and the adherence to precision in naming and discussing the phenomena.

I suspect that they are even more similar than the author suggests, and perhaps are even indistinguishable. The major difference, that SD occurs in normal O2 environments and HSD occurs in low O2 environments simply means that O2 is not the causal factor. I think they are both simply ischemic preconditioning that has been turned on abruptly and hard. Under normal O2 levels, ischemic preconditioning turns off some pathways of ATP consumption, which reduces O2 consumption, so O2 levels go up. Under hypoxic conditions ATP production is reduced, so ATP consumption gets turned off. If the hypoxia is too severe, ATP cannot be produced and cells eventually die during HSD. SD can be tolerated many times with (apparently) little or no damage. I suspect that there is damage, that there is a "pruning" of a few neurons during each instance of SD to reduce the metabolic load and so bring it into better balance with what can be supplied by the vasculature. This "pruning" occurs during any type of seizure or excitotoxic damage. Cells that are firing too much and exceed their metabolic capacity are the ones that succumb to excitotoxic death. The cells that are the "weakest link" in the neural network of the brain. This pruning may not have apparent consequences with each episode, simply due to the redundancy and reserve of neurons present. When that reserve is exhausted, increased dysfunction will occur with each occurrence.

There are reports that people with migraine are at higher risk for lesions of various types visible on MRI. The increased risk due to migraines is additive to other risk factors and is somewhat higher in migraine with aura. Males who experience migraine are at slightly higher risk for cardiovascular disease. Women with migraine are at somewhat greater risk. I see the association of migraine with cardiovascular disease as both being due to and exacerbated by low NO.

Patients with migraine show subtle reductions in grey matter diffusivity compared to controls via high field MRI. Diffusivity relates to ATP levels as discussed above. These same patients also had reductions in grey matter density. The grey matter is where the cell bodies of neurons are, where protein synthesis and mitochondria biogenesis occurs.
The confinement of SD to the gray matter may be the attempt by physiology to spare the cell bodies of neurons. The white matter is mostly axons, which in principle can be replaced if the cell body of that axon remains in tact.

A number of conditions that are associated with mast cell activation are also associated with migraine including allergies, asthma, and irritable bowel syndrome. Elevated levels of histamine are sometimes associated with migraine, suggesting that mast cells in tissues associated with neurons in the brain may be involved in migraine. Agents that sensitize and activate mast cells also increase the sensitivity of intracranial meningeal pain receptors. These are thought to be the source of much of the pain felt during migraine.

When mast cells degranulate they release histamine as well as other agents that cause the production of ROS. ROS destroys NO, and this increases the sensitivity of mast cells to degranulation. Mast cells are responsible for release of ROS that is the inflammatory response to hypoxia.

There is another type of headache that is associated with excessive numbers of immune cells in the CSF, termed pseudomigraine lymphocytic pleocytosis. I see this as the production of superoxide by larger numbers of lymphocytes in the CSF, this superoxide reduces NO levels and triggers ischemic preconditioning and the low NO state.

Migraine is observed more frequently in people with other capillary/connective tissue disorders such as Sjögren's syndrome, Raynaud's and other rheumatic disorders. I think this relates to low NO being the final common pathway in all of these.

In conclusion, migraine is the triggering of ischemic preconditioning in the brain.

Reductions in Brain Blood flow associated with neurodegenerative diseases

Essentially all of the neurodegenerative diseases are characterized by reductions in blood flow, reductions in metabolism, accumulation of damaged proteins, and atrophy and shrinkage of the brain. These changes are not acute, but are progressive sometimes over many years and involve the whole brain.

I see this characteristic decline as the inevitable consequence of low basal NO. Once the NO level gets low enough that basal blood flow is affected, then basal blood flow is reduced, and tissues remodel themselves to accommodate to the now reduced nutritive blood flow. The reduced metabolic demands then set up another round of ablation of excess vasculature, reduced basal blood flow and still more remodeling. This progressive atrophy of tissue due to low NO is what I term the "low NO death spiral". The fundamental problem is the shifted setpoint brought about by reduced basal NO levels. Physiology is still regulating vascularization appropriately, it is simply to the wrong setpoint. The only way to fix the vascular remodeling is by restoring the correct setpoint. The only way to restore the correct setpoint is by restoring the appropriate basal NO level. This level is local to the tissue under consideration, and cannot be measured in vivo. It is on the order of nM/L.

In some ways the "low NO death spiral" is similar to the cell death that occurs during seizures or spreading depression. Those are acute episodes where cells are "tested" and the weakest cells ablated. We know there must be mechanisms for ablating cells because organs can and do shrink. Acute infarcts cause necrosis and scarring, less acute infarcts cause apoptosis and cell removal without scarring.

Diabetic vasculopathy

Vascular abnormalities leading to tissue damage are a common outcome in diabetes type 1 and diabetes type 2. I prefer the term metabolic syndrome over diabetes type 2 because there is a lot more going on than simply high blood sugar, and it is fundamentally different than diabetes type 1. One can have both diabetes type 1 and the metabolic syndrome simultaneously. I won't go into a lot of detail because there is quite an extensive literature on it. Diabetic vasculopathy is a chronic condition, it is not caused acutely. A serious complication is the very slow healing of even minor wounds which then become infected and if not treated adequately that healing occurs, amputation is not infrequently necessary.

There is a lot of thought that it is simply the high blood sugar that causes the damage. This is not strictly correct, but it is certainly related. There have been two recent large trials on standard blood glucose regulation vs. intensive glucose regulation, and with conflicting results. In one trial increased regulation of blood sugar doesn't reduced the death rate, it actually increases it. My interpretation is that in some cases trying to prevent hyperglycemia can be counterproductive. There is another study with a different conclusion, that intensive blood glucose control is beneficial. However in this study the incidence of hypoglycemic events requiring assistance and medical assistance was much higher in the intensive control group.

A recent (1999) "review points out that there is no compartment of glucose in the body at which all glucose is at the same concentration, and that models of glucose metabolism, including effects of insulin on glucose metabolism based on assumptions of concentration homogeneity, cannot be entirely correct." I would be more blunt; such models are wrong.

All cells in all tissue compartments need sufficient glucose. The only place where glucose can easily be measured is in bulk blood which is well mixed and essentially uniform in composition. Most cells derive glucose not from blood, but from plasma in the extravascular space. This plasma has a lower glucose level than bulk blood because cells have removed glucose from it before it reaches the sampling point. Physiology can't regulate extravascular glucose independently of blood glucose because it is plasma from the blood that makes up that extravascular plasma.

Preventing hyperglycemia will be counterproductive if it causes pathologically low glucose levels in the extravascular space (where it cannot be measured). Too much glucose is bad, but not enough glucose is worse. Not enough glucose in the extravascular space can occur even when there is pathologically high glucose in the blood stream. I suspect that to some extent that is the reason that physiology causes hyperglycemia in the first place (in the case of the metabolic syndrome, not diabetes type 1). NO is the signal for mitochondria biogenesis. With low NO, there ends up being not enough mitochondria. This shifts ATP production more to glycolysis, which takes 19 times more glucose per ATP molecule. If 5% of ATP production is shifted from mitochondria to glycolysis, that cell needs twice as much glucose to accommodate it. How can the vasculature deliver twice as much glucose? Only by increasing glucose concentrations in blood. If blood levels of glucose are not allowed to go up, then cells too far from a capillary become starved for glucose.

I suspect that if the groups were stratified for on the basis of capillary density that intensive glucose control would be beneficial for those with high capillary densities and the adverse events occur more in the group with low capillary densities, but it is probably more complicated than that.

In the intensive trial that was stopped, patients averaged 4 years younger and started out ~15 kg heavier and some exhibited larger weight gain since baseline (27.8% gained 10 kg or more compared to 14.1% in the standard group) (the averages are not provided). The starting weight in that trial was 93.5 and 93.6 kg. In the other trial, the starting weights were 78.2 and 78.0 and weight change was smaller, the ending weights were 78.1 and 77.0 kg. The standard treatment leg actually lost weight.

I suspect that weight and weight gain is a marker for degree of ATP production from glycolysis. When ATP is produced by glycolysis, lactate is produced and that lactate must be disposed of. Without enough mitochondria in the liver to recycle lactate into glucose via the Cori cycle, I think the excess lactate gets disposed of as fat. Since mitochondria biogenesis is triggered by NO, low NO will cause fewer mitochondria.

Diabetic vasculopathy is somewhat more complicated than just hyperglycemia. Low NO is a major final common pathway, but the cause is somewhat different. Acute hyperglycemia causes acute production of superoxide which reduces NO mediated regulation of vascular tone. What is interesting in this paper is that a transient elevation of glucose caused a sustained reduction in NO mediated vasodilation. This makes sense from a physiological control sense. When does blood glucose go up? When the body calls for more glucose to deal with an acute event such as running from a bear. The glucose is needed not in the bulk blood, but in the peripheral tissues, in the extravascular space. The only way that pulse of glucose can get to the extravascular space is to increase the pressure drop through the capillary bed and so transiently increase the extravascular flow and the flow velocity in the extravascular tissue compartment.

In obese Zucker rats, flow induced remodeling is characterized by low NO. Treatments that reduce NO decrease vasodilation due to shear, treatments that decrease superoxide (and so increase NO) increase vasodilation.

There have been suggestions that individuals with recurrent diabetic wounds have increased blood NO. This is incorrect. A paper which purports to have found this didn't actually measure NO, they measured the sum of nitrate plus nitrite. This is a common and fundamental error. NO has a very short lifetime in blood (less than 1 second) and is present at only nM/L levels. It is converted into nitrite and nitrate by oxyhemoglobin. Nitrite and nitrate are present at tens of microM/L. NO is extremely difficult to measure, nitrite and nitrate are easy to measure. Nitrite and nitrate are the terminal metabolites of NO, so there is some relationship between NO and nitrite and nitrate levels. Precisely what that relationship is remains largely unknown (and is likely very different in different tissue compartments). NOx levels in blood are more related to NO production rate than to NO concentration. The effects of NO as a signaling molecule are local and are related to the local NO concentration, not the NO production rate averaged over long times and multiple tissue compartments.

NO is one of the cytokines that has major regulatory effects on the immune system. NO attracts immune cells to the site of infection and regulates their function once they are there. This regulation is complex, and is affected by such things as temperature (NO being increased by fever range temperatures). NO causes vasodilation, bringing increased flow of blood. NO inhibits biofilm formation by Pseudomonas and Nitrite inhibited the formation of biofilms by Staphylococcus aureus and Staphylococcus epidermidis, and caused dissociation of biofilms already formed. Biofilm formation is a major virulence factor in infection. Suppression of virulence factor production renders even infectious strains of bacteria non-infectious. This is a point that is not always appreciated. Bacterial strains are infectious only because they produce toxins, proteases, and other virulence factors. Bacteria that do not produce virulence factors are non-virulent. Expression of virulence factors is regulated by bacteria, and until their expression is triggered, bacteria are non-virulent. Raising NO levels locally and systemically will improve the healing of diabetic wounds. Improving vascularization by increasing NO will prevent them from happening in the first place.


NO and NOx physiology is intimately connected with the regulation of vascularization. Capillary spacing is regulated not by gradients of O2, but by gradients of NO. Low NO causes physiology to decrease capillary spacing because low NO mimics the local signal of oxyhemoglobin being diffusively close. Physiology can't compensate because it is the compensatory pathways that are affected.


1 Pechánová O, Simko F. The role of nitric oxide in the maintenance of vasoactive balance. Physiol Res. 2007;56 Suppl 2:S7-S16. Epub 2007 Sep 5. Review.

2 Espey MG, Thomas DD, Miranda KM, Wink DA. Focusing of nitric oxide mediated nitrosation and oxidative nitrosylation as a consequence of reaction with superoxide. Proc Natl Acad Sci U S A. 2002 Aug 20;99(17):11127-32. Epub 2002 Aug 12.

3 Deem, Steven, Richard G. Hedges, Steven McKinney, Nayak L. Polissar, Michael K. Alberts, and Erik R.
Swenson. Mechanisms of improvement in pulmonary gas exchange during isovolemic hemodilution. J. Appl. Physiol. 87(1): 132–141, 1999.

4 MURRAY,JOHN F., AND EDGARDO ESCOBAR. Circulatory effects of blood viscosity: comparison
of methemoglobinemia and anemia. JOURNAL OF APPLIED PHYSIOLOGY Vol. 25, No. 5, 594-599
November 1968.

5 Brown WR, Moody DM, Challa VR, Thore CR, Anstrom JA. Venous collagenosis and arteriolar tortuosity in leukoaraiosis. J Neurol Sci. 2002 Nov 15;203-204:159-63.

6 Silber HA, Lima JA, Bluemke DA, Astor BC, Gupta SN, Foo TK, Ouyang P. Arterial reactivity in lower extremities is progressively reduced as cardiovascular risk factors increase: comparison with upper extremities using magnetic resonance imaging. J Am Coll Cardiol. 2007 Mar 6;49(9):939-45. Epub 2007 Feb 16.

7 Sun D, Huang A, Yan EH, Wu Z, Yan C, Kaminski PM, Oury TD, Wolin MS, Kaley G. Reduced release of nitric oxide to shear stress in mesenteric arteries of aged rats. Am J Physiol Heart Circ Physiol. 2004 Jun;286(6):H2249-56. Epub 2004 Jan 29.
8 Brown WR, Moody DM, Thore CR, Challa VR, Anstrom JA. Vascular dementia in leukoaraiosis may be a consequence of capillary loss not only in the lesions, but in normal-appearing white matter and cortex as well. J Neurol Sci. 2007 Jun 15;257(1-2):62-6. Epub 2007 Feb