3 recent reports use evolution to study mechanisms of antibody diversification

In Chapter 6 of Darwin's Black Box, Michael Behe listed several immune subsystems that he considered irreducibly complex (IC), and therefore (according to him but no one else) unevolvable. One incredibly complex immune subsystem that Behe neglected to mention was the system that genetically modifies antibody genes during the course of an immune response. This system is largely responsible for our ability to generate stronger and faster resistance to subsequent infections, and is integral to why vaccines work. 3 recent papers used concepts in evolution to help characterize one of the most interesting and novel immunological genes discovered since the RAGs in the late 80s, a gene called activation induced deaminase (AID, pronounced as initials), a gene pivotal to antibody modification. Not only did these papers reveal interesting functional insights into AID, but also helped solidify a model for the origin and evolution of this system. Two of the articles come from labs instantly recognizable to most molecular immunolgists, but are totally unknown in ID/evolution circles. The third comes from a lab that most of the regulars here at the Panda's Thumb would immediately recognize, PT's own Andrea Bottaro.
Background: post-activation modifications of antibody genes Every antibody-producing B cell expresses an antibody of a different specificity. That specificity is determined at two distinct stages of development. During early B cell development, antibody genes are assembled through the somatic rearrangement of gene segments in a process known as V(D)J recombination. The Recombination Activating Genes, RAG-1 and RAG-2, are the only B cell specific genes essential for this reaction. V(D)J recombination was originally identified by Behe as one of his IC systems in Darwin's Black Box. Its mechanics and putative evolution are discussed here. After a B cell is activated by binding to a foreign agent, the B cell proliferates and can further diversify its antibody genes through three processes. In Somatic Hypermutation (SHM), the antibody genes undergo point mutations within their binding regions, and through an undeniably Darwinian process, B cells that produce antibodies with beneficial mutations are selected for and expand. In Class Switch Recombination (CSR), the exons encoding the tail end of the antibody (called the constant region) are swapped with different constant region exons, altering which immune cells can interact with the antibody and changing the way the immune system utilizes it. In Gene Conversion (GC), the binding region of the antibody genes are swapped with regions contained within antibody pseudogenes. This process is superficially similar to CSR, but because it occurs in the binding region, it has the same functional effect as SHM. As GC does not occur in mice or man, it will not be further discussed here. Through a combination of these three processes, antibody genes are optimized for each specific infection. AID enters the picture Mechanistically, SHM, CSR, and GC are quite different, and until recently, were considered unique subsystems. However, at the turn of the millennium, a gene called activation induced deaminase (AID) was discovered to be essential for all three processes. While much is unknown about AID's function, the current consensus is that AID deaminates cytosine nucleotides, converting them into uracils within the antibody genes. Depending on where the deamination occurs and the DNA repair pathway employed to fix it, SHM, CSR, or GC results. SHM has been observed in all organisms which possess an antibody rearranging immune system, from cartilaginous fish to mammals. However, CSR is present only in tetrapods (e.g. amphibians, reptiles, birds, and mammals). Both cartilaginous and bony fish possess a functional AID gene, but not the other components necessary for CSR, namely the additional constant regions and other regions called switch regions, where AID targets. Each constant region has a switch region immediately in front of it, and when the switch region of both the original constant region, and a second constant region undergo AID-mediated cytosine deamination, the two switch regions recombine and the constant regions are swapped. Each switch region has its own promoter. As AID preferentially targets transcriptionally active loci, the choice of which switch region is targeted is determined primarily by their transcriptional activity. AID.png Figure 1 A. Genetic organization of a rearranged antibody heavy chain gene. The binding region, which is the rearranged VDJ exon is shown in red/yellow/blue. The 4 exons of the (mu) constant region are in green and the (gamma) constant region is in purple. The switch regions are represented by circles. B. In a putative model of somatic hypermutation, AID binds to the VDJ exon of the antibody and converts random cytosines to uracils (represented by asterisks). When the mismatch repair machinery repairs the mutated DNA, it utilizes an error-prone polymerase, often leaving a mutation near the site of conversion. C. In class switch recombination (CSR), AID targets the switch regions and converts cytosines to uracils there. The DNA repair machinery uses homologous recombination to repair the damage, causing the region between the two switch regions to be deleted, and changing the constant region exons. Evolution of CSR Like the V(D)J recombination system discussed here, it is not difficult to imagine how a three part irreducibly complex system like CSR could have evolved from another system. AID was already present in early tetropods, as SHM was and still is active in these organisms. The appearance of additional constant regions would only require a duplication of the original constant region, an entirely plausible event given that the bony fish that do not undergo CSR often contain multiple constant regions. In these organisms, the choice of which constant region is used is determined at the level of alterative splicing. Therefore, the only unique feature of the CSR system is the switch regions. Switch regions are extremely heterogenous in sequence, and have very little similarity to one another. The only feature they share in common is a propensity to form loop-like structures when in a single strand (such as when they're undergoing transcription). In fact, in some species of frog, switch regions are merely a region of repetitive AGCT tetramers, which don't even form loops. It isn't difficult to imagine a scenario by which AID could be co-opted from an SHM system for use in a primitive CSR system, swapping the original constant region with a duplicated one. But what about the switch regions? How will AID target the constant regions if no switch regions are present? The binding regions of the antibody genes don't contain sequences that resemble switch regions, yet they are ample targets for AID-mediated cytosine deamination, indicating that switch regions wouldn't even be necessary for a primitive CSR system. It's entirely possible that the switch regions evolved after CSR, providing better targets for AID. Because they don't have a research program of their own, IDists would be quick to play the incredulous card, doubting the feasibility of the scenario and demanding an unreasonable burden of proof to prevent them from dismissing the evolvability of the system altogether. While there are certainly gaps to be filled, no one denies that, the point of science is that the gaps can be filled. This is unlike ID, where no research is ever proposed and certainly never, ever conducted. What they did One issue with the evolution of CSR from SHM is whether the original AID in the early tetropods was capable of mediating a CSR-type reaction. The C-terminal region of AID is required for CSR, but not SHM, suggesting that additional components may be necessary to go from an SHM-capable AID protein to a CSR-capable one. In a recent issue of the Journal of Experimental Medicine, Vasco Barreto and colleagues from Michel Nussenzweig's lab reported that the AID gene from zebrafish was capable of activating CSR when transfected into mouse B cells that lacked their own AID gene. In this simple experiment, Barreto et al. demonstrated that even though bony fish do not possess a CSR system, their AID was fully able to catalyze CSR should the other components of the CSR system appear. They also examined pufferfish AID and found extremely low levels of CSR. In a similar study, published in the journal International Immunology, Wakae and collegues in the lab of Masamichi Muramatsu (who discovered AID while a post-doc in Tasuko Honjo's lab), tested both zebrafish AID as well as catfish AID, and found similar results. With these two studies, there could be no doubt that that particular step in the proposed evolution of the CSR system was entirely feasible. Additionally, it demonstrated that AID's catalytic mechanism for SHM was similar enough to CSR that it would be preserved in bony fish even after 400 million years of differential selection. Travis Ichikawa and colleagues, in the lab of PT's own Andrea Bottaro reported a more detailed analysis of pufferfish and frog AID in the Journal of Immunology. While frog AID was fully capable of mediating CSR in mouse AID-deficient B cells, they found essentially no CSR when pufferfish AID was transfected (remember, frogs have CSR, but pufferfish don't). Ichikawa et al. discovered that the catalytic domain of pufferfish AID was CSR competent, but the non-catalytic domain was inhibiting the reaction. As the C-terminal non-catalytic domain contains a nuclear export signal (NES) required for CSR activity (but not SHM) in mice, it was hypothesized that the pufferfish NES was somehow incompatible with mouse CSR. However, Ichikawa et al. disproved that notion by first observing that pufferfish AID had no difficulties in nuclear export/import in mouse B cells, and second that when replacing the NES sequence on human AID, the pufferfish NES did not negatively impact CSR. Why is this significant? While the details of class-switching and somatic hypermutation are of interest to only a handful of scientists, this research is significant to the ID/evolution debate in several ways. First of all, we now have clear-cut evidence of how a three part IC system could have evolved. No modifications would be necessary to convert an SHM-catalyzing AID gene into a CSR-catalyzing one. This provides yet another crystal clear example of co-option that can be widely cited for years to come. On another level, this research shows that more and more researchers are utilizing concepts in evolution to provide functional insight into their own areas of interest. The standard approach to characterizing a gene is to make mutations in important domains and see how that affects the function of the protein. However, by swapping pieces with another organism, this can identify important functional domains that no one might have thought to make mutations in. Furthermore, it helps to form a model for the evolution of the system, and this can lead to more functional predictions. Function informs evolution, and evolution informs function. Contrast this with ID. Because ID, as currently formulated by its chief advocates, places no constraints on the abilities or tendencies of the designer, all data is equally probable and equally consistent. Therefore, no functional predictions can be made from ID, and no functional data could ever provide evidence for or against ID. This is the primary reason why ID's most vociferous advocates don't do any research on ID. This brings up another reason for the significance of these reports. You may have heard Behe or Wells propose some experiments on ID. Aside from the fact that those experiments have nothing to do with ID, you've never heard either of them actually do the experiments they proposed. It's one thing to sit around and come up with ideas, it's another thing entirely to act on them. It costs nothing to propose an experiment, but to conduct one takes money, resources, and a few years of blood and sweat. That is the real B.S. detector for scientists. If their experiments are so great, why aren't they doing them? Here we have a terrific example of someone who posts regularly at the Panda's Thumb throwing his hat into the ring and doing an evolution-inspired experiment. Let's see Behe or Dembski or Wells do that. References 1. Barreto VM, Pan-Hammarstrom Q, Zhao Y, Hammarstrom L, Misulovin Z, Nussenzweig MC. AID from bony fish catalyzes class switch recombination. J Exp Med. 2005 Sep 19;202(6):733-8. 2. Wakae K, Magor BG, Saunders H, Nagaoka H, Kawamura A, Kinoshita K, Honjo T, Muramatsu M. Evolution of class switch recombination function in fish activation-induced cytidine deaminase, AID. Int Immunol. 2006 Jan;18(1):41-7. 3. Ichikawa HT, Sowden MP, Torelli AT, Bachl J, Huang P, Dance GS, Marr SH, Robert J, Wedekind JE, Smith HC, Bottaro A. Structural phylogenetic analysis of activation-induced deaminase function. J Immunol. 2006 Jul 1;177(1):355-61. Link to PDF



Did Ichikawa et al. attempt to use a fragment of the AID that only contained the catalytic subunit without the inhibitory subunit in mouse B cells? If so, did this fragment catalyze antibody class-switching recombination?

Secondly, what are the similarities of this inhibitory subunit to other proteins? Did the tetrapod lineage simply dump this part of the protein or was this subunit kept and somehow modified so that it no longer inhibited CSR?


Michael: we did not use the catalytic domain alone, but the expectation is that it would not work, because a mutation in the AID gene that causes protein truncation just after the beginning of the non-catalytic domain is known to cause a human immunodeficiency (hyper-IgM syndrome type 2) which is characterized by defective switching and somatic hypermutation.

Also, I am not sure I would call the fugu non-catalytic domain “inhibitory”, as opposed to simply non-functional (or not fully functional) in mammalian cells. Clearly, full functionality of AID requires both the catalytic and non-catalytic domains, and issues of inter-domain compatibility, effects on protein folding and dimerization (AID is thought to act as a dimer - two AID molecules sticking together as in the model shown in figure 2 of this paper by our collaborators), as well as potential interactions with co-factors may all contribute to the decreased activity of the fugu sequence.

That said, we mention in our paper the possibility (based on other data) that AID may have been under selection for decreased functionality. This sounds paradoxical, but there is evidence that AID, if let loose in the nucleus, can mutate some sequences that are not its appropriate targets, and even generate cancer-causing mutations. If that’s the case, it may not be a coincidence that one of the vertebrates with the most compact genome, fugu, also displays the AID molecule with the lowest intrinsic activity, since mistargeting by AID would be more likely to cause deleterious gene mutations in a “denser” genome. So, perhaps direct “inhibitory” effects are not entirely out of the question, but in the absence of direct evidence they are just speculation at this point.

As for the second part of your question, the AID non-catalytic domain is actually quite conserved between fish and tetrapods (which is good for us, because it may make it easier to identify which of the few differences are crucial). It also shares similarities with the corresponding domains of a number of other cytidine deaminases, which belong to the same protein family as AID, but have very different functions (another great story, for another day).

Andrea, do you know if anyone has shown a low level of class-switching in the absence of switch regions? I’m trying to remember if a group like Honjo’s ever made a CSR construct that lacked switch regions. Of course, why would they ever do that?

The ID guys would kill for one experiment like this…

In today’s Science:

Dolezal, P. et al. (2006) Evolution of the Molecular Machines for Protein Import into Mitochondria. Science, 316, 314-318

In creating mitochondria some 2 billion years ago, the first eukaryotes needed to establish protein import machinery in the membranes of what was a bacterial endosymbiont. Some of the preexisting protein translocation apparatus of the endosymbiont appears to have been commandeered, including molecular chaperones, the signal peptidase, and some components of the protein-targeting machinery. However, the protein translocases that drive protein import into mitochondria have no obvious counterparts in bacteria, making it likely that these machines were created de novo. The presence of similar translocase subunits in all eukaryotic genomes sequenced to date suggests that all eukaryotes can be considered descendants of a single ancestor species that carried an ancestral “protomitochondria.”

Andrea, do you know if anyone has shown a low level of class-switching in the absence of switch regions? I’m trying to remember if a group like Honjo’s ever made a CSR construct that lacked switch regions. Of course, why would they ever do that?

Actually, they have, plus people have generated mouse switch region knock-outs or replacement mutants with all sorts of sequences, precisely to study target sequence specificity. Below are some references, but to make a long story short, there are some basic sequence motifs and structural moieties (e.g., R-loops) that are typical of S regions and are preferred targets for the recombination. In their absence, CSR is reduced, but it still works to some, variable extent (depending on the residual or replacement sequences).

Kinoshita K, Tashiro J, Tomita S, Lee CG, Honjo T. Target specificity of immunoglobulin class switch recombination is not determined by nucleotide sequences of S regions. Immunity. 1998 9:849-58.

Tashiro J, Kinoshita K, Honjo T. Palindromic but not G-rich sequences are targets of class switch recombination. Int Immunol. 2001 13:495-505.

Luby TM, Schrader CE, Stavnezer J, Selsing E. The mu switch region tandem repeats are important, but not required, for antibody class switch recombination. J Exp Med. 2001 193:159-68.

Min IM, Schrader CE, Vardo J, Luby TM, D’Avirro N, Stavnezer J, Selsing E. The Smu tandem repeat region is critical for Ig isotype switching in the absence of Msh2. Immunity. 2003 19:515-24.

Yu K, Chedin F, Hsieh CL, Wilson TE, Lieber MR. R-loops at immunoglobulin class switch regions in the chromosomes of stimulated B cells. Nat Immunol. 2003 4:442-51.

Shinkura R, Tian M, Smith M, Chua K, Fujiwara Y, Alt FW. The influence of transcriptional orientation on endogenous switch region function. Nat Immunol. 2003 4:435-41.

Zarrin AA, Tian M, Wang J, Borjeson T, Alt FW. Influence of switch region length on immunoglobulin class switch recombination. Proc Natl Acad Sci USA. 2005 102:2466-70.

(Sorry for not putting in direct links, but comments with too many links are rejected as spam)

“Endless forms most beautiful”, indeed! This is amazing stuff. I can almost see why some investigators get more mystical/spiritual the deeper they probe. This is the kind of discovery that the word “awesome” should be reserved for. Thank you, Andrea & Co.

“In Somatic Hypermutation (SHM), the antibody genes undergo point mutations within their binding regions, and through an undeniably Darwinian process, B cells that produce antibodies with beneficial mutations are selected for and expand.”

I wonder if the ID people don’t think it strange that when the immune system is faced with a novel antigen, nature uses random mutation and selection among B cells to come up with an antibody that fits it.

I suppose in their worldview, the decent thing for the immune system to do would be to LOOK at the antigen carefully, and Intelligently Design a single antibody that would do the job of attacking the nasty parasite. This would avoid the immoral waste of less fortunate B cell lines that is part of the abomination that is darwinian selection. Such an organism will surely one day be found - the search has only just begun.

And here’s another thing that is sure to give them nightmares - the Designer created the parasite that made me sick; darwinian selection made the antibody that made me get better again.

So, tetrapods have CSR but other gnathostomes (cartilaginous and bony fish) lack it. This prompts me to ask, what about the other sarcopterygians, namely coelacanths and dipnoid lungfish? What kind of immune systems do they have?

Here is what Dembski has said about the immune system on an old ISCID thread:

Dembski: It seems clear that organisms employ GAs to solve many of the tasks of living (cf. the immune system). But why should that provide confirmation for organisms being the result of GAs (e.g., through a Darwinian evolutionary process). It seems that the immune system, for instance, is a general purpose GA that targets an interloper, sets up a gradient that tracks the interloper, and then runs a GA adapted to that gradient whose output is a molecular assemblage that vanquishes the interloper. All of this sounds very high-tech and programmed. What’s more, none of this contradicts my work in NFL. GPGAs (General Purpose Genetic Algorithms) show all the hallmarks of design.

So Dembski views the ability of an immune system to create a unique immune response as evidence that the immune system was intelligently designed, but not evidence that the immune system itself is an intelligent designer.

I think the immune system is a terrific example of the ability of genetic algorithms. The immune system does not know what type of pathogen it will encounter before hand, and therefore the information necessary to combat a specific infection is not “snuck in” to the GA.

Hi Andrea, so switch regions aren’t absolutely necessary for a minimal CSR system. But clearly AID and the extra constant regions are, which means the CSR system is still IC. So now we can form a model for the evolution of an IC CSR system from an SHM system.

1. A duplication creates an additional constant region downstream of the original constant region. 2. AID induces a low level of spontaneous switching, allowing selection to occur on the duplicated region (additionally, the duplicated region can be expressed through alternative splicing) 3. Switch regions evolve to better target AID to the 5’ ends of the constant regions.

How do we know these events are feasible? a. AID from bony fish can catalyze CSR (see the articles cited in the original post) b. Switch regions aren’t necessary for CSR (see the refs in Andrea’s comment) c. A lot of organisms already possess extra constant regions that can be expressed through alternative splicing (i.e. the delta region, or IgD)


Per Matt’s question - does this mean that the target sites for CSR depend more upon DNA structure rather than a particular sequence? If R-loops are largely what is necessary, then is there a specific sequence needed, or is it a structure that the proteins utilize?

Secondly, regarding the hyper-IgM syndrome type 2 that affects the carboxyl-terminus of AID and abolishes CSR - do these truncations affect protein stability or is that known?


About this Entry

This page contains a single entry by Matt Inlay published on July 20, 2006 1:16 PM.

Science and politics was the previous entry in this blog.

Gallagher gets it? is the next entry in this blog.

Find recent content on the main index or look in the archives to find all content.



Author Archives

Powered by Movable Type 4.381

Site Meter