Bad Lamarckism, Good Lamarckism
Much ink has been spilled defending evolution clearly and forcefully against Biblical creationists. These debates make a lot of noise, but the irrationalism of conservative Christianity is losing ground. The shame is that this distracting discourse makes it hard to discuss real potential scientific issues with Neo-Darwinism. It is the clarity of some Neo-Darwinist arguments like I found here, from Dr. PZ Myers at UMN that makes it possible to discuss the important scientific controversy that is currently happening in evolutionary biology.
The real problem is a controversy between an incomplete Neo-Darwinism and an equally incomplete Neo-Lamarckism. Each side can accuse the other of failure to explain life, because both of them are guilty. Just to set the stage, a reminder of what Neo-Darwinism “replicator theory” says: It says that life is the emergent consequence of self-replicating DNA, and no mechanism other than DNA replication is responsible for functional change over evolutionary time. Importantly, it says all changes in DNA are blind mutations that have no connection to function, they only happen to randomly arrive at better solutions. By this account the only thing that matters in evolution is genetic variation and natural selection. So, it sees organisms as “survival machines” controlled by their DNA strictly as a result of being replicators for these molecules.
Lamarckism is the idea that characteristics acquired during an organism’s lifespan can influence heredity and thereby, evolution. Neo-Darwinists like Dr. Myers attack Lamarckism by pointing out that there is a bottleneck that all multicellular organisms pass through at some stage of their lives where they are reduced to a single cell, and often this cell appears to be stripped of most of its epigenetic markings before becoming an embryo. This happens in humans when our gametes unite and form a zygote. This appears to dispel the notion that evolution could be shaped by anything other than random mutation of DNA, because without extensive epigenetic markings there is no vehicle to carry any other influences forward. This is called the Weissmann barrier.
This conclusion doesn’t necessarily follow from those facts at all, and recent discoveries in epigenetics show that even in large organisms like us, there are clearly some epigenetic markers that do indeed pass through the gametes and get inherited. Recently, biologists have begun to document how non-genetic changes in the cell, and even changes in the behavioral environment around the cell, can also be heritable. This actually does have the effect of shattering the logic of current evolutionary theory. Although it does not reinforce Biblical creationism one iota, it makes Lamarckism more plausible.
I think it is important to first state what is not at issue in this debate:
1. Whether genes influence inherited characteristics (they do)
2. Whether blind variation occurs in genetic material (it does)
3. Whether natural selection can act on genetic frequency (it can)
4. Whether this process of selection can result in evolution (it does)
What is at issue here is a more basic question. The problem is that the Neo-Darwinist theory of evolution presupposes self-organization rather than explaining it, and Lamarckists point to this evasion without explaining it any better. Nothing could have been naturally selected until it was both self-organizing and reproductive. So we can’t assume that natural selection explains self-organization because that is a circular argument. Even if it did, we still need to know how self-organization works for our theory of biology to be complete.
If organisms grew from something like viruses, naked DNA capsules, then the Neo-Darwinist theory would make perfect sense. But instead, all organisms develop from a vastly larger capsule, the cell, which already has a very complex ongoing mechanism for expressing particular genes appropriately at particular times. The DNA molecule is about 2 nanometers wide, but a human cell is at least 20 micrometers across. This is roughly the same size relationship that the contents of your coat pocket have to the contents of your entire residential neighborhood for a square mile around your home.
It’s also not that there isn’t enough information in the genome; it’s that there is far too much. That’s why there is so much interpretive material involved in inheritance. 30,000 sequences of protein-coding DNA can be expressed in many patterns. In fact, the number of ways 30,000 units can be combined uniquely is a number much, much larger than the number of atoms in the observable universe. This means that the hard part is explaining why these numerous genes are expressed in just the right way to produce functional life forms.
Although Myers is correct that there is a great loss of epigenetic information as the multicellular organism goes through its gamete phase, what remains is still an enormously fine-tuned gene regulatory network that sustains a living functional cell. It is this controlling process that is critical to the inheritance process of life, which is why all viruses must hijack a living cell in order to reproduce.
Good Neo-Lamarckists are not arguing that all epigenetic markers are inherited, that would be nonsense. They are pointing out that most organisms do not respect the Weissman barrier, either because they are unicellular so the entire organism is inherited by the daughter cells (which is the vast majority of life) or because they have some degree of totipotency and carry no separate germline (such as all plants,) or because, (like us) they do inherit some small degree of epigenetic marking. This observation firmly disproves the idea that evolution requires a step that strips away all epigenetic markings and removes all variability in the gene regulatory network.
If evolution can and does occur without this stripping-down step, then these rare and minute variations in epigenetic marking become much more significant. This is not because they must account for all heritable changes instead of random mutation. It is because if they were not driving positive functional changes, they would be destroying whatever function had already accumulated in the genome by disrupting the correct gene expression pattern. It would be a runaway process, because there is nothing in the genome itself that prevents changes in gene regulatory mechanisms if these changes can be inherited even very rarely.
This means that these epigenetic changes must be significant drivers of function, rather than insignificant weak effects. If they were not positive drivers, then lineages of organisms with this characteristic would soon die out, because even minute, weak epigenetic effects would accumulate faster than (very rare) positive genetic mutations could control them. It would be like a game of “telephone,” where in each generation the interpretation of what is passed on becomes more nonsensical.
One could use the same argument Dr. Myers articulates against his own position. After all, a beneficial random single-point mutation of DNA would also be a very weak and rare effect. One could criticize the existing theory on the grounds that nearly all random mutations are likewise stripped away by the process of DNA repair, which reduces the error rate in genetic replication more than 1000-fold.
Neo-Lamarckism doesn’t propose a new mechanism for evolution. It just points out that this mystery of self-organization is an inescapable essential ingredient to a coherent theory of life, and we have not solved it. Life is a dance between the genes and the gene interpreters. Which leads and which follows? As Denis Noble likes to point out, a cell stripped of its genome can sometimes live for months (red blood cells do this) before it dies from lack of new proteins. Yet DNA alone in a petri dish will do nothing worthy of the name “life.”
The Weissmann barrier is a straw man that confuses the issue. In his talk, Myers (like Weissmann himself) did a good job of exposing why there is much accumulated information in organisms’ body cells that cannot possibly make it into the next generation. Weissmann’s example was tail amputations on mice, Myers’s was the bulk of epigenetic imprinting on somatic cells. But this proves very little. The bar we have to get over is actually much higher. In order for Neo-Darwinism to be safe, we have to prove that absolutely no information at all is leaking through, and we can’t do that for the reasons stated above. That’s why it’s now clear that the Neo-Darwinist theory, like all good scientific theories (sooner or later) must be overturned.
There is another possibility here that few have considered, probably because the distracting creationist arguments have muddied the waters so much. As Myers mentions, gene expression is controlled by regulatory loops from gene to gene. So are these gene regulatory loops confined to the interior of the cell? No, because the environment exerts a causal influence on gene expression at every level, even for a unicellular organism. It is possible that all organisms are set up with the same universal mechanism for bringing this exterior causal influence to bear on the genome in a controlled, reliable way. Myers hints at this in his talk with his Dobzhansky reference. It may be these mechanisms that are truly naturally selected. In this case, the genome would be a side-effect of an expanding and functional regulatory mechanism that made the genome useful. The cell would be operating the genes, rather than the other way around. This mirrors my epistolution conjecture; the organism organizes itself at the level of the organism.
The genome cannot competely control gene expression because the environment also controls gene expression. This is an obvious fact if you think about it. Of course organisms change adaptively to match the circumstances. But how could the environment enter into the process of gene regulation without introducing disorder? That is the most important question in biology, and it is unsolved.
The core insight of twentieth-century evolutionary theory was to propose that the mutation of DNA amounted to a blind search through the space of many possible genetic forms to find functional ones. The filter sorting good from bad was natural selection, which only happens between generations. This is still true, but in order for epigenetic inheritance to be true also, there must be another blind search and another filter sorting all the cell’s structures during the current generation. This is the only way each version of what is passed on in the game of telephone could make more sense each time.
The filter between generations is survival and reproduction. What could be the filter applied during active life? How could living cells be conducting a blind search during development? Since all known organisms control DNA, the mechanism must have enough universality across the entire tree of life to be fitted to the DNA molecule as an information storage organ.
The filter must be some sort of Lamarckian process because we see that environments do provoke a Lamarckian response in organisms. If you lift weights your quadriceps get larger and if you stay home on bed rest they atrophy. The same goes for neurons and many other malleable biological structures. And what other explanation is there for sleep, where our damaged structures and processes are repaired in a holistic, all-encompassing mode of existence in which we are completely paralyzed and helpless for eight hours a day? Someone could argue that these processes are simply emergent high-level effects that only happen when organisms are sufficiently complex, but then they would be saying that there is not only one logic by which the environment controls gene expression, but multiple such logics. This is harder to imagine and a less parsimonious explanation. Imagine a human adult, as Michael Levin argues, and then wind back the clock day by day until you have a blastocyst. At what moment does the organism switch from one form of genetic control to another?
That is why Lamarckism is still highly relevant. Even Lamarck never argued, as Myers suggests, that all acquired characteristics were inherited. He wasn’t a fool. He was a great biologist of the early nineteenth century, who knew nothing about genes just as Darwin knew absolutely nothing about genes. In fact Darwin postulated that there were objects that moved through the bloodstream that he called “gemmules,” which did what we would now describe as epigenetic marking. And Darwin agreed that use and disuse drove functional changes; he mentions it twelve times in the Origin of Species. Jean-Baptiste Lamarck was the person who coined the term biology. He was a great scientist. He argued that some functional changes that were acquired during life could be inherited, and that this could be among the sources of evolutionary novelty. Both of these points are completely consistent with our current facts. The problem is that this truth falsifies our current best explanation for life.
In principle, we are able to put genetic material from any organism into any other organism, and it can be interpreted without issue. It may not be advantageous for survival, but that is a separate problem. We are all the same type of workers, and we all have the same type of tools in our toolbox. It is estimated that 8% of our own human genome actually has its origins in viral DNA. This suggests that whatever the mechanism of downward control is, it is likely to have been inherited from the very first ancestor of all life, and it is likely to be used by every single cell on earth.
The good news is that if people were focused on the right problem, it is probably highly tractable. We have not discovered any physical mechanism yet that forms a unit that is sensitive to its environment, and uses changes in that environment to organize itself and its own function. But we only have to discover it once. If we just lean into the challenge and try to devise this mechanism experimentally then we will surely make progress. Strangely, instead of embracing this technical challenge, nearly all biologists are insisting that there is nothing more to explain. If current evolutionary theory was sufficient to explain life, then we could make life forms out of non-living materials. We cannot do that yet.
This peculiar intellectual situation is exactly like if in 1890 people looked up at birds and said, “flight is an evolved mechanism,” and they stopped right there and never tried to invent the airplane. We got to airplanes by following clues that the birds gave us, not by taking birds apart under a microscope. Suppose we had focused on that extensive dissection of birds instead of building things like propellers, do you think we would have found the part that made the birds fly? Do you think that aviation would be where it is today if it had stuck to a strictly reductionist approach like this? In order to solve this problem we have to use our creativity. We have to try to invent and build sensitive, self-organizing systems and see if they can solve problems in intelligent ways, in the ways that life forms do, by grappling physically with and then understanding their own contexts. We also have to believe we can do it. The Wright brothers were not self-assured, but they at least entertained the possibility that machines could fly. If we assume we cannot solve this problem, how could we?
One of our most pressing global problems is how to build moral AI. It is possible that the mechanism that gives cells the ability to sense and respond appropriately to their surroundings is what we know, on the level of an organism, as morality. Morality is our general sense of where we are and what is safe to do given the circumstances. Every organism has this sense, and no machines have it. Animals that are socially and cognitively similar to us, like chimps, dolphins, and elephants, have it in spades. They easily work out which behaviors are likely to lead to good things or bad things for themselves and those they care about. Our much more distant cognitive horizon can make this sense very abstract in us; we think about wars and poverty and racism, but all organisms must have this sense to some degree. We have made progress in building machines that are less likely to hurt us and others. For example, we have good anti-lock brakes on cars, and this helps us use them without causing harm. But the sense of the consequences is not intrinsic in the car’s braking mechanism. The car still has to be driven by something that understands the moral context the car is operating in. As our machine-based systems become more powerful and vast in their reach, we urgently need to discover what mechanism creates that contextual awareness. All we know for sure is that it emerges from a large group of cells.
