How can DNA determine morphology?

[pellman]

The following comes from a rudimentary understanding of DNA and biology. Correct me if I have wrong assumptions.

Each cell of an organism contains the same set of DNA as all the other cells of the body. Exceptions occur from imperfect copying but these are pathological.

Since not all cells are identical, so there exists some process whereby the same DNA functions differently to produce different kinds of cells. However, at some level we can identify sets of cells which are for all practical purposes identical. Maybe, for example, all human femur bone cells are identical. Or maybe within the femur the cells at one end are different than the cells at the other. However, at some level, large compared to the size of an individual cell, the cells are all identical to their neighbors.

Consider such a section of tissue. It has a morphology, a shape that is essential to its proper function. Whatever this shape is, at the scale of an individual cell, the shape cannot be "known." If we speak for a moment as if the cell has a brain and that brain is following the instructions of the DNA "map", I don't see how the cell can "know" how to behave relative to its neighbors in such a way that they collectively produce the necessary large scale shape. Because the cell can only "know" what its nearest neighbors are doing. It doesn't "know" where it is in the overall shape it is working to produce.

Even if each cell had a map of the organ detailing the specific role of every cell in that organ, it would be impossible for the cell to be able identify on the map which cell it corresponds to. And each cell has the same instructions as every other cell. There is no central director communicating separate instructions to the cells.

Take a femur for instance. Bone cells on the surface of a femur might be able to "know" they are on the surface of the bone and so they don't reproduce in such a way that the bone would grow deformed. But how do the cells on the ball of the femur know to arrange themselves nearly spherically while those on the shaft of the femur are arranged nearly cylindrically. They can't possibly know that. And there is no central director with a "big picture" telling them how to do it.

So I don't see how in principle that DNA alone can determine gross morphology. If anyone here can explain it, I'd be very interested.

 

[SteamKing]

So I don't see how in principle that DNA alone can determine gross morphology. If anyone here can explain it, I'd be very interested.

The answer is simple: DNA is not the sole determinant of gross morphology of an organism.

At best, think of DNA as a biological filing cabinet which contains all the plans and specifications for making a human or a turtle. Just like the plans for a building don't actually turn into a structure by themselves, so too DNA requires additional mechanisms to turn the plans for building a cell into an actual biological entity.

When a new animal organism is conceived from the mixing of its parents' DNA, a series of biological processes begin to occur as the fertilized egg cell divides and starts to grow into an embryo. When the embryo develops, the basic plan of the organism is created using many different biological processes:

http://en.wikipedia.org/wiki/Developmental_biology

The cells in the developing embryo communicate with one another by a variety of means as different types of tissue begin to differentiate from the original egg cell. Some of the undifferentiated cells are called stem cells:

http://en.wikipedia.org/wiki/Stem_cell

Depending on the exposure to specific biological signalling chemicals, the stem cells turn into bone or muscle tissue or an organ like a kidney. The signalling chemicals with which most people are aware are the various hormones coursing through our bodies, but there are several other lesser known substances. An increase in a specific hormone, like testosterone, can lead to a deeper voice, increased muscle mass, and hair in inconvenient places.

 

[Ygggdrasil]

How do you go from a single cell (the fertilized embryo) to a complex multicellular organism? Well, first, the embryo needs to define the various axes for its body plan (front-back, top-bottom, and left-right). Often these axes are defined by gradients of signaling molecules (called morphogens) that are produced on one side (leading to a high concentration for example at the side that will become the head) and diffuse away while being slowly degraded (leading to a low concentration at the side that will become the tail). This allows cells to know where they are in the organism just by measuring the concentration of a signaling molecule.

Another important concept here is differentiation. Although all cells contain the same DNA sequence, the cells can modify which regions of DNA are accessible to be read. For example, DNA sequences can be chemically modified (methylated) to silence those regions of DNA, and different cell types will have different DNA methylation patterns. DNA methylation and other types of epigenetic marks are important for telling cells which genes to turn on and which genes to turn off, thus determining which type of cell they become.

Thus, once the embryo begins dividing, the morphogen gradients will let cells know where they are in the embryo. Based on their position, they then can then begin differentiating into the proper cell type by turning on and off the proper sets of genes. The process is much more complicated than that, but this at least gives some insight into the basic mechanisms that occur during development.

The organism for which this question has been best studied is the worm C. elegans. Every C. elegans consists of 959 or 1031 cells depending on the sex, and every cell division leading from the fertilized embryo to the full adult has been mapped. Thus for every single cell in the adult worm, we know exactly the sequence of cell divisions that led up to that cell having its particular position in the adult embryo.

 

[gravenewworld]

The problem is that you fall into the trap that many scientists do when it comes to biology--reductionism. DNA is only one part of a much greater machine and in fact, I'd argue there exist much more complex systems within a cell than DNA. Some strains of rice have more genes than humans do, but who'd argue that rice is more complex than a human? The failure in logic is that DNA does *not* encode all of the information needed to produce a cell. DNA only produces proteins, but that tells you nothing about how those thousands of proteins interact as a system. Those proteins then go on to produce a system of millions of metabolites and millions of post translational modifications that truly describe the physiology of a cell. Many biochemical pathways, when viewed holistically, serve as biochemical supercomputing biosensors that tells DNA what to do. These biosensors can respond to everything from mechanical forces to nutrition or hormone gradients. You can not understand the concept of a symphony by examining the conductor alone.

 

[pellman]

Thanks, guys. This is very complex.

 

[Ygggdrasil]

DNA only produces proteins, but that tells you nothing about how those thousands of proteins interact as a system.

But isn't information about how these proteins interact with other proteins and function in the cell ultimately encoded in the DNA? Just because we don't know how to determine that information from a DNA sequence doesn't mean that it isn't there.

 

[DiracPool]

Thanks, guys. This is very complex.

If you're really interested in this question, pellman, I would suggest picking up the book, "Endless forms, most beautiful" by Sean "B." Carroll. The B is to distinguish this guy from the other Sean Carroll, of GR physics fame. Embryology IS very complex. It's complex and magical and wonderful. In "Origin of species", I think that the last line of the book or something, "endless forms most beautiful."

Whatever it is, that book will answer your question. Another good book is "Your inner fish" You can't just look at the DNA of cell in isolation in order to understand how it creates an organism. It very much is interweaved with the environment in the manner in which the DNA expresses itself during development. This is what they call the science of evo-devo. But Carroll's book goes deep into that, and it's a relatively easy, popular read.

 

[SteamKing]

Thanks, guys. This is very complex.

And that's why were still figuring out some of it, we have no idea about a lot of it, but a tiny bit we're still pretty sure of. The amount of new information added in just the last 30-40 years is staggering, particularly since DNA replication techniques were developed.

 

[Torbjorn_L]

If we speak for a moment as if the cell has a brain and that brain is following the instructions of the DNA "map", I don't see how the cell can "know" how to behave relative to its neighbors in such a way that they collectively produce the necessary large scale shape. Because the cell can only "know" what its nearest neighbors are doing. It doesn't "know" where it is in the overall shape it is working to produce.

If you substitute water molecules for cells and a water drop for the organ, the problem with the attempted description may become clearer:

'If we speak for a moment as if the water molecule has a brain and that brain is following the instructions of the force "map", I don't see how the water molecule can "know" how to behave relative to its neighbors in such a way that they collectively produce the necessary large scale shape. Because the water molecule can only "know" what its nearest neighbors are doing. It doesn't "know" where it is in the overall shape it is working to produce.'

Other comments have described how the organ shape ("water drop shape") emerges out of small scale interactions. The cell doesn't know what it is part of any more than the selfish genes that are the core of the evolutionary process today. They only "know" what they have to do to survive and procreate (through the germ line), painfully learned by trial and error or success, differential reproduction, through each generation. It is the environment (for each gene all the other genes and the rest of the nature it lives in, for each cell all the other cells and the rest of the nature it lives in) that contains the rest of the information that allow structure formation.

The problem is that you fall into the trap that many scientists do when it comes to biology--reductionism. DNA is only one part of a much greater machine ...

That depends on what you mean by "reductionism". If you mean the method to divide and conquer, science is based on that and as I showed by a water drop example above biology is no different from physics in that respect.

If you mean a philosophic claim that disallow emergence, that too fails as per the example.

And finally the water drop example shows how emergence, of structures like galaxies, stars, planets, geophysical systems, cells and complex multicellular organisms, is part and parcel of "divide and conquer".

Also, the "reductionism" of the selfish gene vs the environment is a better model than your definition of "holism" which fails miserably here. The organism machine(s)* and their sensory networks aren't all of the environment, and is too small a concept to describe evolution as it forgets to cover the here necessary entirety of the system. Differential reproduction incorporates everything from developmental defects (it better has) to accidents (some of which are avoidable).

*I would quibble with the idea that a set of cooperating machines are "a" machine rather than a system, there is a feature creep in that definition (likely because if it is philosophy it was never intended to be tested), but the idea has worse problems and never gets off the ground.

 

[mes314159]

Another way to think about it is that the genome is neither a blueprint or a map, it's more like a set of Ikea instructions.

1. Make lots of cells by mitosis.
2. Now there are a lot of cells. If you are a cell on the surface of the ball of cells, turn on gene 125. If you are inside the ball of cells, turn on gene 452.
etc etc
4,567. If you are a cell next to some other cells that are next to some other cells that have gene 452 turned on, turn off gene 452 and turn on gene 681.
etc etc.

In other words, there is no way to easily "decode" the primary DNA sequence of a genome and "guess" what the organism would look like. Genes get turned on and off, proteins and other biomolecules get made, they assemble together (or not) and physically modify each other by phosphorylation, glycosylation, etc (or not), and cells end up having certain shapes, which in turn cause organs to end up having certain shapes, which end up making the whole organism have a certain shape. Some of this is encoded in the basic laws of physical chemistry, which even for a single protein cannot easily be applied. So it's no surprise that we are still absolute infants in our understanding of genomes. I consider this to be a project of centuries at least, hardly years or decades. That doesn't mean we can't answer some reasonable questions, like are there specific genes which if mutated cause someone to have high cholesterol (yes), or get lung cancer with high probability (I don't know). But these are just the tip of the real genomic information iceberg.

 

[Pythagorean]

But isn't information about how these proteins interact with other proteins and function in the cell ultimately encoded in the DNA? Just because we don't know how to determine that information from a DNA sequence doesn't mean that it isn't there.

Isn't it a mix of DNA instructions and environmental context? Like wouldn't a human embryo develop a different morphology on the moon than the same organism developing on Earth?

 

[mes314159]

Obviously embryonic development can be affected by the local environment. In placental mammals, where the embryo is inside the mother's body, it can be affected by chemical substances (like alcohol or thalidomide, which cause developmental disorders) or even by a physical trauma. And to say DNA encodes protein sequence doesn't mean that physical chemistry is ignored. At this point we just don't know that much about the relationship between primary amino acid sequence and tertiary protein structure to say how much protein structure/function is directly encoded, and how much simply "unrolls" due to the intrinsic laws of chemistry.

 

[Ygggdrasil]

Isn't it a mix of DNA instructions and environmental context? Like wouldn't a human embryo develop a different morphology on the moon than the same organism developing on Earth?

Yes, but the organism's DNA determines the response to the environment. In other words, even though individuals with identical DNA might develop differently when placed in different environments, the differential response to the environment should be predictable (in theory) from the organisms's DNA.

 

[Pythagorean]

Yes, but the organism's DNA determines the response to the environment. In other words, even though individuals with identical DNA might develop differently when placed in different environments, the differential response to the environment should be predictable (in theory) from the organisms's DNA.

What if it's chaotic? More precisely, what if the final state of the system is sensitive to the initial conditions [1-4]? Then it's not predictable even if we accept that it's a deterministic process [5].

[1] http://einrichtungen.ph.tum.de/E19/AG%20Krischer/Teaching/TuringPatterns.pdf (introductory material)
[2] http://link.springer.com/article/10.1007/BF00046598#page-1 (chaos in plant morphology)
[3] http://jms.org.br/v28n1/05 (chaos in anatomical morphology)
[4] http://www.ncbi.nlm.nih.gov/pubmed/1467450 (chaos in morphological regulatory networks and context)
[5] http://hum.sagepub.com/content/46/7/777.short (ignore the application -sociology- this is just for background on what chaos theory says about unpredictability vs. determinism.)

 

[Ygggdrasil]

What if it's chaotic? More precisely, what if the final state of the system is sensitive to the initial conditions [1-4]? Then it's not predictable even if we accept that it's a deterministic process [5].

Yes, that's certainly true. There are also cases where stochastic processes determine the phenotype of an organism, meaning that the phenotype is essentially random. So, you are correct that an organisms' phenotype cannot be predicted from its genotype in all cases.

 

[Torbjorn_L]

What if it's chaotic? More precisely, what if the final state of the system is sensitive to the initial conditions [1-4]? Then it's not predictable even if we accept that it's a deterministic process [5].

An interesting observation.

I'm sure PF will be tired of me doing this, but first I will have to put "deterministic" aside. It is a philosophical idea and I can't use it. A gene operates through a stochastic process (with mutational variation) in stochastic environment (cellular sparse chemistry), so essentially predictability in the presence of stochasticity isn't a problem.

Then I have to note (and see below) that stochastic effects and chaos are filtered through several means. Above all evolution is darwinian, small steps that were survivable in previous generations (to a degree due to later changes). If a fetus can develop in different gravity, it is because of the continuity of physics and not of evolution.

But it is worse than that expression of the genome has stochasticity (including descriptions of chaos), but the evolutionary environment has it too. E.g. Lotke-Volterra predator-prey models has chaos IIRC, and if not the weather has. But then again, a gene doesn't know that. Its response to the environment sees only fitness, and fitness seems to be predictive enough.

To wit, if we extrapolate to infinite populations, the susceptibility of fitness to stochastic effects disappears. Filter effects are in play, such as generational delay, and extinction puts a sharp constraint on the amount of stochastic effects that drowns out evolution.

When the population isn't infinite, stochasticity creeps into extinction of alleles and their eventual fixation, and I can't quantify that. But the process should again try to erase random fixated less fit alleles. (I'm officially handwaving here, quantitative examples would be good.)

TL;DR: I don't think the extended phenotype can be predicted from the genome at all. The information that makes the bee's hive or the beaver's dam isn't encoded into some simple mapping derived from genome and environment. Differential (fitness) effects may be teased apart, but then again most many genes will affect many traits. (E.g. length depends on thousands of genes, and those genes affects many other traits.)

Isn't it a bit like trying to predict the coastal lines of an ocean from the knowledge that there is a water cycle?

 

[Pythagorean]

I think quantification of morphological differences can be explained (predicted) mechanistically with a discussion of bifurcation theory. That is, some parameter (like gravitational pull) in the system (here, an ensemble of differentiating tissue cells) will have particular critical values in which a qualitatively different outcome emerges. On either side of the critical value, only quantity will change (like the average cellular density will decrease) but at the critical value, the decrease in average cell density will begin to have more significant effects on the resulting tissue structure's ensemble properties (like a different overall morphological shape, not just a bigger or smaller version of it).

An interesting observation.

I'm sure PF will be tired of me doing this, but first I will have to put "deterministic" aside. It is a philosophical idea and I have no use for it. A gene operates through a stochastic process (with mutational variation) in stochastic environment (cellular sparse chemistry).

It is not just a philosophical idea, it is a useful idea in math and science. A system of differential equations is deterministic if, for every state in the system, it's future state is unique. A phenomena that can be predicted with such a system is said to be deterministic. It doesn't mean all of the organism or universe is deterministic. In the quantitative sciences, we talk about phenomena being deterministic or stochastic, not the whole organism/world. Further, a particular phenomena can have both deterministic and stochastic processes involved in it.

The idea is further confused by systems that we model stochastically, but that we could model deterministically if we wanted. For instance, mechanical statistics is based on a system that we can model deterministically, but there's just too much mathematics involved to do it when we can instead get an accurate prediction of the macrostate with the methodology of stochastics. Note that Occam's razor doesn't fit here because the deterministic method actually explains the mechanisms of particle path and interaction, the statistical method only cares about the whole gas, not the individual particles. So some knowledge about minutia (which particle hit which particle) is ignored in favor of the degenerate macrostates (i.e. the same macrostate can arrive from a variety of microstates).

Similar occurences could happen in mutation variation. We model it stochastically, since there's no way to have complete knowledge of initial conditions of environment and genetics. Any sensitivity will render prediction unavailable. That doesn't prove that it can't be modeled deterministically, just that we haven't found a way to do it yet.

 

[aroc91]

Yes, that's certainly true. There are also cases where stochastic processes determine the phenotype of an organism, meaning that the phenotype is essentially random. So, you are correct that an organisms' phenotype cannot be predicted from its genotype in all cases.

X inactivation in cats is similar in this regard.

 

[Ygggdrasil]

X inactivation in cats is similar in this regard.

The more I think about it, the more examples, I can come up with where stochastic processes affect development in humans. X-inactivation is a great example that not only applies to cats, but humans as well. A similar process is that of mosaicism, where mutations or other changes to the genome that occur during development cause different areas of the body to have slightly different genomes. For example, mosaicism within the brain caused by the insertion of retrotransposon sequences within the DNA of neurons has been postulated as a http://www.nature.com/nrn/journal/v15/n8/full/nrn3730.html.

 

[Torbjorn_L]

I think quantification of morphological differences can be explained (predicted) mechanistically with a discussion of bifurcation theory.

Chaotic processes appear in the organism (modes of heart failure, for one). And to predict more of development, it may have its uses.

But the OP was interested in how the genome control morphology. (Not much, and not in detail; just enough to make do until the next generation in the roughly same environment.)

It is not just a philosophical idea, it is a useful idea in math and science. A system of differential equations is deterministic if, for every state in the system, it's future state is unique.

I confess I forgot about math.

The reason I forgot is that we don't use that characteristic, or we wouldn't notice chaos. What we use is stochasticity and chaos, so we know what type of model to use, and predictability: "In the quantitative sciences, we talk about phenomena being deterministic or stochastic". There are "deterministic" models that aren't predictive (e.g. chaos).

Maybe it is history tripping me up. Until the discovery of chaos, it was merely a redundant description. (Deterministic systems were precisely predictive, not statistically so.) Now it is a problematic description, and I pawn its continuing use off on philosophy of Leibniz's clockwork universe. [ http://en.wikipedia.org/wiki/Clockwork_universe ]

I'm sorry for the confusion.

 

[Pythagorean]

Hmmm... perhaps you missed my post #14 :)