This is number two in my series of blogs on epigenetic control of genome restructuring and hereditary transmission of traits modified by life history events. We are going to take a detour through some classic bacterial genetics history, but it will ultimately bring us back, with new insights, to the kind of epigenetic regulation covered in the previous blog.
One of the strictest dogmas in conventional evolutionary thinking is the explicitly anti-Lamarckist assertion that conditions cannot induce a related genetic change. To quote an unimpeachable orthodox spokesman, my University of Chicago colleague Jerry Coyne:
In fact, we know of no evidence for mutations occurring nonrandomly or "adaptively", i.e., that the occurrence of mutations is somehow biased in a direction that makes them more likely to be favorable when they arise, particularly when the environment changes in a way that requires favorable mutations to fuel adaptive evolution.
The empirical evidence against experience-related genome change is remarkably thin. Textbooks present the famous 1943 Luria-Delbrück experiment as the definitive demonstration that mutations must occur prior to selection.
Salvador Luria and Max Delbrück, who received Nobel Prizes for their pioneering roles in molecular genetics, set out to test whether virus infection could induce mutations to resistance.
The analysis was statistical in nature. (Their paper even includes an elaborate mathematical appendix by Delbrück, a converted theoretical physicist.) The argument was as follows. If infection induced mutations to virus resistance with a certain low probability, then both different samples from one culture and samples from independent cultures should display similar numbers of colonies that would fit some form of normal (Gaussian) distribution.
They set up a large number of independent cultures of bacteria, divided them each into multiple samples, infected all samples with viruses that killed infected cells, and measured the frequency (proportion) of resistant mutant bacteria in each sample that could produce a colony in the presence of the lethal viruses.
What they found was that the replicas of single cultures did produce normal distributions of resistant colony numbers, as expected, but the colony numbers from one culture bore no statistical relationship to the other independent cultures. Some cultures produced very low numbers of colonies, while others displayed "jackpots" of high frequencies of resistant cells.
They correctly interpreted this deviation from a normal culture-to-culture distribution of mutant frequencies to provide evidence for the stochastic occurrence of rare mutations to virus resistance at different times in the growth of each culture prior to selection. Early mutations would multiply into many resistant progeny (high mutant frequency), while later-occurring mutations would produce smaller populations of resistant progeny (low mutant frequency).
Given the lethal nature of the selecting virus Luria and Delbrück used, there was in fact no other possible outcome. Infection was invariably lethal, and only preexisting resistant mutants could survive. Nonetheless, this experiment was cited for over six decades as proof that virus infection could not induce a genetic change to resistance.
One has to be careful with the word "proof" in science. I always said that conventional evolutionists were hanging a very heavy coat on a very thin peg in the way they cited Luria and Delbrück. The peg broke in the first decade of this century.
Bacterial genomicists noted some curious structures in the DNA sequences of many different bacterial species. They were called CRISPRs, for "clustered regularly interspaced short palindromic repeats." The structures had groups of inverted repeat sequences ("palindromic repeats") arranged near each other in a short region of the bacterial DNA.
Although their meaning was not known at first, the regularity of the CRISPRs made them literally pop out of the computer analysis. The functional significance of the CRISPRs became clearer when the non-repeat "spacer" segments separating the inverted repeats were analyzed. They contained short sequences from viruses, plasmids, and other types of invasive DNA. It looked like the CRISPR might serve as a kind of DNA memory bank for past infections of the bacterial cell.
Perhaps the CRISPR system provided not only memory but defense against infectious DNA. This idea was tested by infecting bacterial cells with viruses that are not invariably lethal, and it was found to be correct. The surviving bacteria had new repeats and spacers in their CRISPRs, and the spacers contained DNA sequence fragments from the infecting viruses.
The "impossible" had happened. By a still-unknown mechanism, the bacterial cell managed to insert a short sequence from the infecting virus in its genome's CRISPR region and use that newly acquired information to block the infection. There is now evidence that the CRISPR defense works by encoding a small interfering RNA molecule (siRNA) and using that siRNA to guide cleavage of the invading DNA.
This remarkable CRISPR genome immunity system is an example of the general, recently discovered process known as "RNA interference," or RNAi. RNAi regulates genome expression in eukaryotic cells by targeting the destruction of transcripts from coding sequence, but RNAi also targets the formation of silenced chromatin and so blocks transcription in the first place.
For examples, RNAi can target silencing of specific genetic loci in genetically modified plants. A striking case involves the FWA ("flowering arrested") locus in Arabidopsis. Normally, FWA expression is imprinted to be off in all tissues of the plant except the seeds. But a stable FWA epimutation can occur so that FWA is expressed in all plant tissues, which delays flowering. By inserting a transgene encoding an siRNA that recognizes a retrotransposon at the start of FWA, expression can be silenced, so that the normal flowering pattern is restored in a stable heritable fashion, even when the siRNA transgene is removed.
RNAi serves as an acquired antiviral defense system in eukaryotes, as well. In Drosophila and other animals, there are analogues to the bacterial CRISPRs called "piRNA loci." These loci contain acquired fragments of retroviruses and other mobile genetic elements and encode germline-specific piRNAs that keep the genomic copies of those elements in a silent state. As we shall see in the next blog, this germline-specific piRNA-targeted epigenetic control is key to understanding how life history events activate natural genetic engineering.
This contradicts the normal meaning of "random mutation" - those insertions are obviously not random copying errors. But it clearly contradicts Jerry Coyne's definition of random as well, namely that those changes are not deliberately adaptive.
It seems the next natural question to ask is: How do cells incorporate this information about past invasions when presented with a new invader that they have not seen before?
Humans and animals in that situation look for similarities. Therefore it seems natural to hypothesize that cells do as well.
We have real time natural and laboratory experience with how Drosophila and mice deal with novel invaders. This is known as "hybrid dysgenesis" (http://shapiro.bsd.uchicago.edu/Hybrid_dysgenesis_interspecific_hybridization.html).
Hybrid dysgenesis occurs when males with certain mobile elements in their chromosomes mate with females that do not have those elements and do not encode piRNAs to direct their epigenetic silencing. As the name "dysgenesis" implies, the germlines of the progeny flies or mice experience great instability, show many new insertions of the mobile elements, and acquire chromosome rearrangements. Often, but not always, the dysgenic progeny have so much germline chromosome disruption that they are sterile.
As far as I know, one has done an experiment to ask how Drosophila or mice might respond to mobile elements that are similar but not identical to ones previously sampled for piRNA templates.
Not clear what or who you mean. Please clarify.
Good ol' natural selection.
Once again, you seem to be astonished to note that cells are very complicated, as only billion-year old entities might rather be expected to be.
It's interesting that you reference the original paper for Luria-Delbruck, and wikipedia for CRISPRs, given that the wikipedia entry for Luria-Delbruck gives a rather more useful discussion than your post.
Just checked the Wikipedia entry for the Luria-Delbrueck experiment. No mention of the points made here. Do you have something positive to contribute?
I replied earlier, but that message seems to have gone astray in the blogosphere.
There are many ways that viruses kill cells. Some leave time for a cell response but serve equally well to select resistant mutants, as the work on CRISPRs demonstrates. Some kill as soon as they attach to the cell or very shortly after they inject their genome. So a CRISPR type response is not possible.
I expect that the distribution of resistant mutants will differ significantly in the two cases. But as far as I am aware, no one does the statistical analysis now. They go straight for the DNA to see directly what is going on.
I gather that a fair amount of investigation has been done with regard to the interaction of streptococcus thermophilus CRISPRs with invading phages. In particular, invading phages can result in new spacers 30 nucleotides long being inserted in the CRISPR loci. These genetic loci in the phages are the proto-spacers.
Although the mechanism whereby the bacteria are able to transfer the proto-spacer genetic loci to the CRISPR spacers have not yet been identified, is the identity of the proto-spacer in the phage known? That is to say, can investigators predict in advance which genomic sequence in the phage will constitute a proto-spacer and thus be transferred to the CRISPR loci, or does it seem to be a 'random' sequence which is transferred?
If it is not 'random' how does the bacterium figure out what portion of the phage genome constitutes a proto-spacer and thus should be transferred to the CRISPR spacer?
We know that many different viral sequences end up as CRISPR spacers. As far as I know, no one has yet found a pattern in how these segments are chosen for incorporation into the CRISPR. There may be no pattern, but I would like to see what happens when there are more extensive sets of experimental data. I may also be ignorant of recent work that has found a pattern.
Berthajane Vandegrift
A few Autistic Questions about Freud, Marx and Darwin
CRISPRs provide a clear refutation of the above generalization: Bacteria exposed to some viruses end up with a portion of the of the viral genome inserted into the 'spacer' section of the CRISPR. This new spacer in turn aids the bacteria in overcoming the viral attack. Here we have an example of a mutation that IS 'somehow biased in a direction that makes [it] more likely to be favorable when [it] arise[s].'
You have summarized accurately the first part of what I tried to explain.
The detour about over-interpretation of the Luria-Delbrueck experiment was included as a cautionary tale about exaggerating the significance of particular results to bolster untested arguments.
The real point for epigenetics is the ability of RNA to target the formation of chromatin domains. The next blog will discuss how that is relevant to the control of natural genetic engineering.
PART I
What follows is my attempt to summarize the main points of your blog. Let me know if I've made some mistakes.
In the Luria-Delbruck experiment bacteria were exposed to a lethal virus. At the end of the experiment it was found that some mutant bacteria had survived. The question to be resolved was: Did the mutations which led to survival occur before exposure to the virus or did they occur after exposure? If the former, the mutations were obviously in no way influenced by exposure to the virus. If the later, perhaps exposure to the virus in some way triggered or influenced the mutations.
A careful statistical analysis of the results demonstrated that all the mutations had occurred before exposure to the virus. The conclusion was that exposure to the virus did not in any way affect mutations which would help the bacteria deal with the viral threat. This conclusion has been generalized to the following: 'mutations do not occur nonrandomly or "adaptively", i.e., that the occurrence of mutations is somehow biased in a direction that makes them more likely to be favorable when they arise.' This generalization however is invalid; the virus chosen for the experiment was so lethal that the non-mutated infected bacteria were killed before there was any opportunity for the infecting virus to influence their mutations.