Are there many non-functional B cell receptors which have experienced the effect of selection?


What is the history of memory BCR sequences that have V and J out of frame from one another? There appears to be some disagreement in the literature on this point. For the purposes of this discussion, let’s call them non-functional receptors. There are two potential sources.

Non-functional BCRs matched with a functional receptor

We have two copies of the BCR locus. Thus, when the first V(D)J rearrangement results in an out of frame receptor, B cells can rearrange the second locus. If this one succeeds in affinity maturation and its B cell is exported from the germinal center, both loci can be sequenced from DNA. There seems to be broad agreement that this is the dominant source of out of frame sequences. However, there’s another option…

BCRs rendered non-functional from somatic hypermutation (SHM) indels

Insertion/deletion (indel) mutations happen in the course of affinity maturation, and if one gets an indel between the V and J templated regions that is of length not divisible by 3, this will lead to V and J going out of frame. This matters for people working to understand the neutral processes underlying affinity maturation, as said in [CD2016]:

some mutations in nonfunctional sequences may be subject to selection pressures if the sequences were rendered nonfunctional during the affinity maturation process

Because of this, [CD2016] develops a very clever but rather complex mouse system in which they are able to sequence lots of non-functional receptors, which are dominated by sequences with V and J in different frames. You can read the paper to hear how it works, but the outcome is that these sequences are enriched in BCRs that were always non-functional, rather than those that became out of frame after indel.

Thinking a little more I’d like to understand better exactly how sequences which were originally in frame but later had disruptive indels would make it into the blood and thus onto the sequencer. The situation seems like it must proceed as follows:

  1. In frame BCR enters germinal center
  2. Mutation-selection
  3. Indel, leading to out of frame sequence
  4. The indel-containing B cell is exported from the germinal center
  5. That cell is sampled in the blood as a plasma or memory cell.

How frequently are B cells with faulty BCRs exported from the germinal center? Well, there are two ways out of the germinal center that don’t involve death: export as plasma cells or memory cells.

Plasma cells

[VN2012] describe how positive selection is required for plasma cell differentiation:

In addition to cyclic re-entry, positively selected GC cells are exported from the GC as plasmablasts-- the precursors of the plasma cells that will secrete antibody into serum and secretions

For this they cite the work of [PP2006], which shows that there is a strong selective sieve for export as a plasma cell. This is a complex process that isn’t completely understood, but certainly requires BCR signaling or T cell help, both of which require a functioning BCR.

Memory B cells

For memory B cells, the situation seems a little less clear. [VN2012] summarize

… whereas very high-affinity cells have an increased probability of being directed to the plasma cell fate, lower-affinity cells that pass positive selection can be directed into either the memory or recycling GC cell pools…

So although the bar isn’t as high as for plasma cells, they do indicate that cells have to pass positive selection to exit the germinal center. I’d think that a cell without a functional antibody would be unable to pass positive selection.

Something else

Another possibility could be that after an indel event, the second BCR locus could rearrange. However, there is no evidence for peripheral RAG recombination. Furthermore, even if the second rearrangement could happen in the GC it’s unlikely to bind well to whatever is being presented on the FDCs.

Other options include that when the GC dissolves some cells with non-functional receptors survive and make their way into the memory pool. Or perhaps indels could arise by themselves during the decades that memory B cells need to keep up a minimum expression level. But…

How are these cells maintained?

For us to sequence these cells in a memory sort, they have to be kept alive by homeostatic mechanisms. Athough there has been some disagreement in the literature, [BE2009] provide evidence that in vivo, memory B cells need BCR stimulation to proliferate even at low levels. On the other hand, [ML2000] show that switching a high affinity receptor for a low affinity receptor doesn’t reduce lifespan, so high affinity engagement of the BCR isn’t essential. Having a low affinity receptor is different than having a non-functional receptor, though.

Final thoughts

What else is missing here? Are there experiments that could directly address this question? Are there ways in which we can filter out-of-frame sequences that arise from indels? Although the system of [CD2016] is super cool, the resulting neutrally evolving sequences are limited to a single light chain of a transgenic mouse. If we could confidently treat out-of-frame sequences as coming from faulty rearrangments, or even a good estimate of the frequency of non-functional sequences coming from other sources, we would have many more sequences with which to learn about the neutral somatic hypermutation process.

Thank you to Laura Noges, @bussec, @krdav, and @wsdewitt for discussions. Please add your thoughts below! I’m especially hoping to hear from team @steven.kleinstein.


  • [BE2009] Benson, M. J., Elgueta, R., Schpero, W., Molloy, M., Zhang, W., Usherwood, E., & Noelle, R. J. (2009). Distinction of the memory B cell response to cognate antigen versus bystander inflammatory signals. The Journal of Experimental Medicine, 206(9), 2013–2025.

  • [CD2016] Cui, A., Di Niro, R., Vander Heiden, J. A., Briggs, A. W., Adams, K., Gilbert, T., … Kleinstein, S. H. (2016). A Model of Somatic Hypermutation Targeting in Mice Based on High-Throughput Ig Sequencing Data. Journal of Immunology , 197(9), 3566–3574.

  • [ML2000] Maruyama, M., Lam, K. P., & Rajewsky, K. (2000). Memory B-cell persistence is independent of persisting immunizing antigen. Nature, 407(6804), 636–642.

  • [PP2006] Phan, T. G., Paus, D., Chan, T. D., Turner, M. L., Nutt, S. L., Basten, A., & Brink, R. (2006). High affinity germinal center B cells are actively selected into the plasma cell compartment. The Journal of Experimental Medicine, 203(11), 2419–2424.

  • [VN2012] Victora, G. D., & Nussenzweig, M. C. (2012). Germinal centers. Annual Review of Immunology, 30, 429–457.


I think a lot depends whether you are looking at heavy chain or light chain. Genes may appear to be non functional through sequencing error, homopolymer tracts are prevalent in D regions and so CDRH3 regions are prone to error, ie they might appear to be non functional but not really.

Data on heavy chains could contain a second rearranged allele that is out of frame (one would sequence genomic DNA to find these). As you say, difficult to imagine how a B cell without a functional Heavy chain at all could survive the affinity maturation selection processes.

Light chains are a different matter, because they can undergo receptor editing to change light chain, there have been reports of reexpressed RAG in the periphery. If a cell had undergone somatic hypermutation and then receptor editing then you could find rearranged light chain sequences as passengers which may have previously been expressed (and therefore selected).


These are important matters that in my view may be difficult to approach in cases when such sequences are rare occurrences. Sequencing/PCR errors may create reads that appear to be non-functional but which are artefacts of the process. Will it be possible to differentiate between these possibilities if the frequency of cells of the B cell lineage not producing Ig, if they exist, is low? I agree with @DDW that there indeed have been reports of peripheral RAG activity (e.g. in relation to receptor revision):

  • Gevaert P et al. Local receptor revision and class switching to IgE in chronic rhinosinusitis with nasal polyps. Allergy 2013;68:55-63. doi: 10.1111/all.12054.
  • Rochas C et al. RAG-mediated secondary rearrangements of B-cell antigen receptors in rheumatoid synovial tissue. Autoimmun Rev 2007;7:155-9.
  • Hillion S et al. Expression of RAGs in peripheral B cells outside germinal centers is associated with the expression of CD5. J Immunol 2005;174:5553-61.
  • Wilson PC et al. Receptor revision of immunoglobulin heavy chain variable region genes in normal human B lymphocytes. J Exp Med 2000;191:1881-94.


I hope @steven.kleinstein accepts your invitation to jump in, because it seems to me you might be reading too much into their paper. I read the line you quote as referring to sequences recovered from GC cells (which are what they used from their mouse system), when deletion may not yet have occurred.

I would think you could look to see if these cells exist by using a single-cell technology like @bussec et al have published: if you sequence 100,000 cells, can you find any with two different nonfunctional heavy chains? You could even go a step further: we expect 40% of naive B cells should have nonfunctional “passenger allele” heavy chains from a previous unsuccessful rearrangement. So if you bin cells into clonal lineages, you could look for a statistically significant increase on that baseline. (NB - both assume that cells make enough mRNA from nonproductive loci that they can be recovered, something that doesn’t appear to be true in our protocol. But you could imagine doing this type of experiment from gDNA, if need be.) Of course, this won’t tell you which is which, but setting an upper bound on the phenomenon is probably a useful first step.


I also asked Gabriel Victora, and here’s what he had to say:

Luka has been thinking a lot about this lately. I would say nonfunctional BCRs are mostly dead ends (see the most recent nussenzweig science paper). Unless we are talking about the passenger allele. We often get things that look like that, whole clones of nonfunctional BCRs that can’t possibly come from the functional allele.

The paper he refers to is the awesome

Although this does help clarify the selective process in the germinal center, I think the hard part is to quantify Gabriel’s “mostly.”


If a non-functional BCR was produced as a result of SHM, you might expect to also be able to find other members of the clone from which it was produced. If you could recapitulate the lineage of that clone and see the non-functional branch (presumably a dead end) I imagine that would help speak to that mode of production.

Another consideration is the existing of dual light chain expressing B cells, which might provide a means for a once-functional (light chain) rearrangement to persist after being scuppered by SHM.


Thanks, @JamieHeather! I agree, and think that’s a good idea. This isn’t without its challenges, though! My question was about sequences that are non-functional because of V and J being in different frames, which means that the CDR3 length is going to differ between them. Clustering sequences with different CDR3 lengths is hard computationally and hard to have much confidence in, because it’s easy to confuse SHM indels with different rearrangements.

If I understand correctly, I think that @DDW made your point about light chains.


I did have a chance to talk to @steven.kleinstein about this last week and he seemed less concerned about it than I would have expected. That is, he thought that if V and J are indeed in different frames then this has probably always been a passenger allele (note that having stop codons or indels in the V is a different story).

So, unless someone tells me otherwise, we’re happily going to assume that out of frames evolve as passengers. We have some thoughts about how we could further validate this, but that’ll take some time and work.


In (Cui et al., 2016), we found that targeting patterns on nonfunctional sequences resemble those on functional sequences. As you say, to increase confidence of these being true nonfunctional rearrangements the indel putting V/J out-of-frame should not be in V, D or J segment part of the junction (one could develop a statistical test on this). It is also critical that the nonfunctional data be produced using a sequencing technology that is not prone to indel errors, and there are appropriate error-correction steps (e.g., UMIs).