Antibody/Immunoglobulin complex/B cell receptor complex

Please see thread at http://groups.google.com/group/obi-biomaterial/browse\_thread/thread/f6a137b3c7748600

Reported by: alanr

Original Ticket: "geneontology/ontology-requests/5649":https://sourceforge.net/p/geneontology/ontology-requests/5649

Alex,

I hope you don't mind my assigning this item to you; I'm very happy to help with implementing any changes that we might make in light of the email thread Alan refers to, but I really don't want to act on it without your input.

I'll also try to distill out the issues, but it's worth reading the emails for context.

I see from SF 885513 that you composed the current definition for GO:0042571 (immunoglobulin complex, circulating), which is one of the terms of particular interest in the linked emails. Its parent and sibling, GO:0019814 (immunoglobulin complex) and GO:0019815 (B cell receptor complex) also come in for scrutiny.

One point raised is that the definitions for two of these terms include the phrase "functions as an antibody", but in MF we have "antigen binding" with "antibody" as a related synonym. My best recollection is that we've done it that way so that the MF term name describes an action rather than a gene product or complex of gene products, but we should make the BP definitions consistent with MF. The simplest thing to do would be to change the BP defs to use "functions in antigen binding".

Another question refers to the subunit composition of the complexes, so if you could double-check all three defs, it will be much appreciated.

I think I spotted a suggestion that GO either add a CC term for "antibody complex", or add a synonym to "immunoglobulin complex" or its child(ren). I think I prefer a synonym -- unless you know of some distinction between antibody complexes and any other immunoglobulin complexes. There's also a concern that intracellular Ig complexes that act as antibodies aren't represented well.

Midori

Original comment by: mah11

assigned_to: nobody --> aledie

Original comment by: mah11

Midori,

Reading through the email thread was a bit confusing. I'm not sure about who stated "either GO changes, or we don't use GO," but of course it's good to remind people that we always want to improve the GO when we can, but that the GO cannot be all things to all people.

I did not originate the definition of 'GO:0042571 ! immunoglobulin complex, circulating,' but added the sentence "Some forms are polymers of the basic structure and contain additional components such as J-chain and the secretory component," tweaked a few other words, and added a reference. My point at the time was to allow for the fact that some immunoglobulin complexes contained additional protein components. In any case, this was in early 2004, when I was relatively new to the GO and was a bit more reluctant to shake things up. People are obviously more sensitive now to perceived faults in the definition, which I really haven't thought about much for a while.

My proposal:

Eliminate the "functions as an antibody" phrase entirely from all three immunoglobulin terms (GO:0042571, GO:0019814, and GO:0019815). Not all immunogloblins have corresponding antigens to which to bind, and we want to define these terms by their structures in any case. I also added a comment referring to the possibility of antigen binding to the terms.
Add "antibody" as a synonym to these terms, as 'circulating antibody' (exact synonym) to 'GO:0042571 ! immunoglobulin complex, circulating,' as 'membrane-bound antibody' (exact synonym) to 'GO:0019815 ! B cell receptor complex', and as 'antibody' (exact synonym) to the parent term 'GO:0019814 ! immunoglobulin complex'.

The terms "antibody" and "immunoglobulin" are used as more or less exact synonyms by immunologists, medical doctors, and the general scientific community, but the usage pattern is a bit skewed. Immunoglobulin complexes of known specificity are almost always called antibodies. However, people commonly talk about antibodies even when specificities are unknown. Immunoglobulin is more commonly used when we are talking about the structural nature of an antibody, the genes that encode the antibody, or the fraction of blood proteins that are antibodies. One could insist that we shouldn't call an immunoglobulin an antibody unless we know that it has a corresponding antigen -- however in this case we are talking about instances of immunoglobulins and not the canonical form.

Judging from the email thread, I'm guessing some participants would like to impose a difference in meaning between the two words that's not really supported in actual usage. Maybe OBI can define antibody as "an immunoglobulin complex of known specifity", but this term is not appropriate for the GO CC as the protein structure is the same, and as I said above, the distinction fails to take into account the fact that most people use the terms as exact synonyms.

I also agree with Midori that "GO must accurately reflect the canonical in vivo localizations of protein complex types, and we will review the definitions and paths for the immunoglobulin complexes to ensure that they are correct, or make any necessary changes. It is not the purview of GO to capture the locations of protein complexes -- antibodies or otherwise -- in experimental contexts such as prepared fractions or administration to samples."

Regarding Alan Ruttenburg's comment, I'm not sure how the word "circulating" is not compatible with the possibility of internalization of antibody within cells. I'm hoping he's not confusing the use of antibodies for intracellular staining in permeabilized cells in an experimental setting with the canonical localization of extracellular space for circulating immunoglobulins in the GO. In any case, "circulating" is perhaps a bit misleading, since many antibodies do not circulate in the sense of traveling around in the blood stream, but rather act locally within tissues or mucosal areas. But they are still secreted into the extracellular space and circulate in the sense of moving somewhat freely within the body. And the word "circulating" matches usage in the literature.

The B cell receptor is a form of immunoglobulin complex that is found in the plasma membrane of B cells. It is still an antibody, it still can bind with a specific antigen if one is available. In binding with an antigen, it transduces a signal to the B cell causing it to react in a variety of ways depending on the maturation state of the cell, its location relative to other cells like T cells and follicular dendritic cells, and other factors such as locally available cytokines. One of the possible outcomes is differentiation into an antibody secreting cell such as a plasmablast or plasma cell.
The GO CC ontology should be used to describe complexes, not the PRO! Why reinvent the wheel, especially when funding is so limited.
These are my suggested revisions to the GO terms and definitions in full. Note there are changes in some synonym scopes and references as well:

[Term] id: GO:0019814 name: immunoglobulin complex namespace: cellular_component def: "A protein complex that in its canonical form is composed of two identical immunoglobulin heavy chains and two identical immunoglobulin light chains, held together by disulfide bonds and sometimes complexed with additional proteins. An immunoglobin complex may be embedded in the plasma membrane or present in the extracellular space, in mucosal areas or other tissues, or circulating in the blood or lymph." [GOC:jl, GOC:add, ISBN:0781765196] comment: Note that an immunoglobulin complex has the function of antigen binding if a suitable antigen is available. subset: goslim_pir synonym: "antibody" EXACT [] is_a: GO:0043234 ! protein complex

[Term] id: GO:0019815 name: B cell receptor complex namespace: cellular_component alt_id: GO:0042570 def: "An immunoglobulin complex that is present in the plasma membrane of B cells and that in its canonical form is composed of two identical immunoglobulin heavy chains and two identical immunoglobulin light chains and a signaling subunit, a heterodimer of the Ig-alpha and Ig-beta proteins." [GOC:add, ISBN:0781765196] synonym: "B cell receptor accessory molecule complex" RELATED [] synonym: "B lymphocyte receptor complex" EXACT [] synonym: "B-cell receptor complex" EXACT [] synonym: "B-lymphocyte receptor complex" EXACT [] synonym: "BCR" EXACT [] synonym: "immunoglobulin complex, membrane bound" EXACT [] synonym: "membrane-bound antibody" EXACT [] comment: Note that an immunoglobulin complex has the function of antigen binding if a suitable antigen is available. is_a: GO:0019814 ! immunoglobulin complex is_a: GO:0043235 ! receptor complex is_a: GO:0044459 ! plasma membrane part relationship: part_of GO:0001772 ! immunological synapse

[Term] id: GO:0042571 name: immunoglobulin complex, circulating namespace: cellular_component def: "An immunoglobulin complex that is secreted into extracellular space and found in mucosal areas or other tissues or circulating in the blood or lymph. In its canonical form, a circulating immunoglobulin complex is composed of two identical heavy chains and two identical light chains, held together by disulfide bonds. Some forms of are polymers of the basic structure and contain additional components such as J-chain and the secretory component." [GOC:add, ISBN:0781765196] synonym: "circulating antibody" EXACT [] comment: Note that an immunoglobulin complex has the function of antigen binding if a suitable antigen is available. is_a: GO:0019814 ! immunoglobulin complex is_a: GO:0044421 ! extracellular region part relationship: part_of GO:0005615 ! extracellular space

I may have missed the point of the email thread, but I would like these changes implemented in any case following additional comments.

Thanks,

Alex

Original comment by: addiehl

Thanks for your comments, Alex.

Good.
Good. Regarding "an immunoglobulin complex of known specificity", this would make the definition dependent on the state of knowledge of the observer, so I don't think it's a good idea.
You write: > Regarding Alan Ruttenberg's comment, I'm not sure how the word > "circulating" is not compatible with the possibility of internalization of > antibody within cells.

The word "circulating" isn't the problem, it's the fact that it is part_of extra-cellular space. The part_of is read as All GO:0042571 are part_of some GO:0005615, which isn't true if the GO:0042571 is internalized.

> I'm hoping he's not confusing the use of antibodies for intracellular staining > in permeabilized cells in an experimental setting with the canonical localization > of extracellular space for circulating immunoglobulins in the GO.

Nope, but the fact that antibodies can be used for intracellular staining was the hint that made me look for natural processes that internalize. I have to review Darren's responses, but it seems to me that there are natural processes involve antibodies realizing antigen binding internal to the cell.

Moreover there is still the question of how OBI should represent such cases even if they are unnatural. Having a definition in the GO that is structural, without attaching it to localization information would make it easier to build off the GO term in OBI.

Understood.
There is an issue that I think should be ameliorated with go CC, namely that there is inadequate distinction between things in which other things can be located, and complexes. I'd like to see these clearly distinguished. Also, which GO has general definitions for complexes there is a need for identifiers for more specific classes such as species specific subclasses, and, for complexes which have "pluggable parts", such as the SCF complex, identifiers for each of the variants of the complex. It seems PRO can have a role in solving at least the second issue, and possibly the first.
GO:0019814: I would move "An immunoglobin complex may be embedded in the plasma membrane or present in the extracellular space, in mucosal areas or other tissues, or circulating in the blood or lymph." to the comments. It isn't essential to the definition, and still doesn't address the issue of internalized antibodies. Re: "comment: Note that an immunoglobulin complex has the function of antigen binding if a suitable antigen is available", the function is present even if the antigen isn't available, but is only realized if the antigen is available. GO:0019815: Although I'm obviously not an immunologist, http://biology.plosjournals.org/perlserv/?request=get-document&doi=10.1371%2Fjournal.pbio.0040200 would suggest that it is part of the normal function of B cell receptors to be internalized. Possibilities: a) Incorrect understanding on my part b) The paper discussed antigen bound receptors as being internalized - this might be considered a different class c) the definition should not imply that the receptors are always membrane bound d) The class should be considered a "defined class" rather than a "universal". Similar comments apply to GO:0042571. In general I think it's only a good idea to include localization information in the definition of complexes (other than when using defined classes) when the lifecyle of the complex happens entirely at that location. In the case of antibodies this isn't the case. It seems that a story/representation more akin to those in developmental biology is more appropriate than using the time unqualified part_of relationships.

-Alan

Original comment by: alanr

I just now found the comment thread (thank you sourceforge).

I think every immunoglobulin complex has the function to bind antigens. It only realizes the function if suitable antigen is available. Functions exist without that ever happening. If functions shouldn't be part of the definition in GO complexes, that is fine with me as well. (I just noted that Alan made the same point under 6).
I disagree that it is common usage to say that the B cell signalling subunit (CD79) is part_of an antibody. That's why I don't like that synonym there. I believe the 'part of the B cell receptor that is secreted' is commonly referred to as antibody, even when it is membrane bound. I would even argue that CD79 is a separate complex, and that common usage of BCR would not include it, but I can live with how you have it now, specifically as I am not sure if you want to deal with complexes of complexes.

As it is hard to argue about 'common usage', I asked colleagues, and will report back what they think.

It seems that the 'extracellular space' location is used in GO synonymously with 'secreted'. Everything found in the extracellular space is also found in some cells, minimally when it is gobbled up by macrophages. I have no major problem with this, as it seems all location information is meant to indicate that the majority of X is at that location. If GO would consider it, I would prefer a 'secreted' annotation. The internalized antibodies, which Alan has convinced me exist naturally, have all been previously secreted.
see 2. I only have a problem with the synonym antibody for the complex including the signalling subunit.
PRO will apparently subclass under GO complexes. Or do you want to deal with our camel antibodies?

Original comment by: bjoernpeters

Here is the response I got from Shane Crotty (http://www.liai.org/pages/faculty-crotty)

"Gee you ontology guys sure know how to have fun.... Antibody is not accepted as a synonym for B cell receptor complex. Antibody is an accepted synonym for immunoglobulin (circulating)."

I also talked to Howard Grey (http://www.liai.org/pages/faculty-grey)

Who confirmed the above, and was even more conservative, saying that 'antibody' should only truly be used for circulating immunoglobulins with a known antigen binding partner.

Original comment by: bjoernpeters

Alan,

(I will address Bjoern's comments separately.)

Thanks for your comments here and in the email thread, which I will address now.

As you know, the GO CC is described in the GOC documentation in the following way: "The cellular component ontology describes locations, at the levels of subcellular structures and macromolecular complexes," and "...GO cellular component terms describe locations where a gene product may act..." Personally, I prefer the description of location as being "relative to the cell," since extracellular space is hardly a subcellular structure, but in any case brighter minds than mine have already shone their light upon this issue, so I will leave it alone.

It's fair to say that CC protein complexes are assigned locations where they are found in "normal" conditions. However, it will always be easy to find exceptions to the localizations assigned in the GO CC. First of all, all proteins are synthesized within cells, within the cytoplasm or cotranslationally inserted into the ER. For most plasma membrane proteins and secreted proteins and protein complexes, it is readily possible to find them in the ER, the golgi, the transgolgi network, as those proteins transit the constitutive secretory pathway, on their way to the cell surface and beyond. Yet we don't typically assign protein complexes to these locations unless the protein complexes actually perform some function in these locations.

Furthermore, transmembrane cell surface receptors are typically recycled into the endosomal-lysosomal pathway for degradation with varying halflives. For many such protein complexes, such as the BCR, such internalization is triggered by ligand binding events, as described in the article you cite. For the BCR, such internalization delivers the bound antigen to a compartment where the antigen will be proteolytically degraded to allow for antigen processing and presentation to T cells. Certainly, such delivery of antigen is one role of the BCR, probably a fairly important one. However, the localization of the BCR in the endosome or lysosome is brief, although detectable, and thus only the major localization of the BCR in numbers and duration is considered when placement in the GO CC is made. Certainly, GO annotation of data showing the BCR in an endosome is always appropriate.

Indeed, the fate of many fluid phase proteins or protein complexes, those in the "extracellular space," is to be eventually taken up by some cell, often a macrophage, via pinocytosis, phagocytosis, or receptor-mediated endocytosis, and get degraded within the endosomal-lysosomal pathway. Yes, under the right conditions we can detect such proteins within an endosome, but again we would not use this information as the basis of a GO CC placement or even as the basis of an annotation unless we thought it was especially biologically significant for a particular protein or protein complex. The people who collect the garbage are interesting, and even how and when they do the collection is interesting, but the garbage itself is mostly boring.

So let's look at the three papers cited in the email:

PMID:1536138 (Salonen et al.) describes the use of anti-telomere antibodies as a diagnostic tool for lupus (SLE). The paper itself presents no discussion of the mechanism of how those antibodies came to be, or their physiological action within a diseased individual. The existence of such antibodies is simply a biomarker, a proxy for the disease. The important point here is just because an antibody exists that has a specificity for a nuclear or cytosolic protein does not say one bit about the ability of immunoglobulins to access the nuclear or cytosolic compartments of intact cells, nor is it proof that such antibodies play any role at all in a pathological process. In fact in healthy individuals antibodies to dsDNA are often readily detected. Cells die, most often in a controlled fashion, but sometime not. Occasionally, cells do "spill their guts," releasing cytosolic and nuclear components directly into extracellular space. The B cell receptors, ie. the membrane bound antibodies of nearby B cells that happen by chance to have specificity for particular cytosolic or nuclear proteins or even DNA, may bind those components, and trigger the B cells to proliferate and differentiate into plasma cells that secrete antibodies of the exact same specificity. A variety of pathogenic mechanisms most likely contribute to initiation of lupus, including faulty B cell tolerance to self-antigens, inappropriate handling of apoptotic cell fragments, and variations in T cell activation and tolerance. Excess autoantibodies and associated immune complexes (crosslinked antibody-antigen complexes) are a hallmark of lupus and lead to kidney failure, but again, the immune complexes are formed in extracellular space, not by antibody crossing the plasma membrane into a topologically distinct compartment from the cytosol.

PMID:14531798 (Fujii et al.) describes a novel mechanism for the uptake of immune complexes in the kidney. Excess "deposition" of immune complexes in the kidney in lupus lead to kidney failure, and this paper seeks to explain how such complexes are internalized by endothelial cells. It is clearly stated in the abstract as well as the body of the paper that the antibodies employed did not have any antigenic specificity for the mouse or human cells used in the study. The authors demonstrate that the internalization of antibody complexes is unrelated to their specificity, but rather due to the spontaneous self-aggregation of the mouse IgG antibodies used and subsequent crosslinking of cell-surface integrins through an interaction with fibronectin, presumably through the IgG3 constant region portion of the immunoglobulin complex. The authors propose, quite reasonably, that this represents a normal mechanism for removal of antibody aggregates that is overwhelmed in lupus because of the high amount of autoantibody-antigen immune complexes present. The immunoglobulin complexes themselves have no role or function once internalized. (In the email, someone states that this paper talks about internalization of B cell receptors, but that is not true at all.)

PMID: 10837074 (Ghetie and Ward) is a review of the functions of the so-called neonatal Fc receptor, FcRn. Fc receptors in general function to bind immunoglobulins via their Fc portion, the non-polymorphic region of the antibody away from the antigen-binding region. Typically Fc receptors bind immunoglobulins and transduce a signal into a cell dependent on whether the antibody being bound has itself bound an antigen. The FcRn receptor has a somewhat different role. It functions to bind immunoglobulins of IgG isotypes at low pH and release them at neutral pH. In some cell types, having bound an antibody (IgG), the FcRn-IgG complex crosses the cytoplasm within a vesicle from the luminal to basolateral membrane in a process known as transcytosis. The IgG is then released at the plasma membrane on the other side of the cell from where it was picked up. This allows, for instance, newborn animals and babies to pick up IgG immunoglobulin complexes from milk in the intestine and transport it to their own interior extracellular space. Yes, antibody is internalized, but in an antigen-independent fashion. Yes, we can detect such internalized immunglobulin complexes, but these complexes are still topologically on the opposite side of membranes from the nucleus and cytosol.

You state "... the fact that antibodies can be used for intracellular staining was the hint that made me look for natural processes that internalize." Yet the intracellular staining of cytosolic and nuclear antigens utilizes a completely artificial method of using gentle detergent essentially to poke holes in cellular membranes to allow antibodies to leak inside the cells to find their antigens. There is simply no relation between this method and any natural system of antibody internalization.

None of these three papers or the artificial intracellular staining of cytosolic or nuclear antigens validates the statement "it seems to me that there are natural processes involve antibodies realizing antigen binding internal to the cell." In experimental systems, however, one can most likely raise antibodies to endosomal-resident proteins and use such antibodies to bind those proteins following endocytosis. However, this is still an artificial system. Probably you can find even a natural example, if you look hard enough.

The primary problem here is the all-or-none nature of the 'part_of' relationship. It just doesn't really work well for the CC ontology for many complexes, because the exceptions are easy to find. Immunoglobulin complexes are produced in the ER, mature in the golgi, and once on the plasma membrane or floating in extracellular space, are often internalized again. Yet their role of binding antigen happens primarily in extracellular space in natural situations. Even membrane bound forms of immunoglobulins (BCR) bind antigens that themselves are external to the B cell. I can't solve the part-of problem myself, but I do suggest that imposing excess strictness to this relationship in the CC will under close examination cause the unraveling of portions of the CC ontology. Perhaps this will lead to improvements in the CC ontology, but it may destroy the usefulness of the ontology for describing the canonical localizations of protein complexes.

My earlier point 5 was primarily an aside, although an expression of my fear of the PRO taking on more than it can reasonably accomplish while at the same time undermining the GO in some people's eyes. This is obviously not an appropriate forum for this discussion, however.

I have no particular objection to moving the localization information within the definition to the comment for the term GO:0019814, immunoglobulin complex, I will suggest it is appropriate to remain within the definition for GO:0019815, B cell receptor complex, and GO:0042571 immunoglobulin complex, circulating.

Thanks,

Alex

Original comment by: addiehl

Bjoern,

Thanks for your comments.

A. Based on the responses of your experts, I would be happen to use "antibody" as a synonym in the following ways:

As an EXACT synonym for "GO:0042571, immunoglobulin complex, circulating."

As a NARROW synonym for "GO:0019814, immunoglobulin complex."

And as a RELATED synonym for "GO:0019815, B cell receptor complex."

This ensures that all three terms will come up in a search for "antibody". I am unwilling to accept Howard Gray's very narrow definition of antibody, since I believe it is at odds with the fact that secreted immunoglobulins are produced from the plasma cell progeny of B-2 cells in response to antigenic stimulus and therefore must be "antibodies," and even natural antibodies produced by B-1 B cells have antigens that evolutionarily selected and can be determined upon study.

Please ignore my earlier synonym suggestions.

B. Also CD79a and CD79b (Ig-alpha and Ig-beta) are stable components of the B cell receptor complex and will continue to be part of the definition of that complex.

C. The definition of immunoglobulin is intended to allow for non-canonical forms, including camel immunoglobulins.

-- Alex

Original comment by: addiehl

Midori,

When you do finally implement the changes here, please delete the part_of relationship between "B cell receptor complex" and "immununological synapse." The BCR is not always part of the immunological synapse.

Similarly, please delete the part_of relationship between "T cell receptor complex" and "immununological synapse." The TCR is not always part of the immunological synapse.

I put these relationships in the GO before I fully appreciated the proper use of "part_of" relationships,

Thanks,

Alex

Original comment by: addiehl

I am happy with the synonym assignments that Alex suggested; that takes care of all concerns I had.

Original comment by: bjoernpeters

Alex,

Thank you for your thorough and well-informed comments here. I think I've now made all of the changes you've recommended -- definition edits from the 2008-12-11 22:07 comment, synonyms from 2008-12-15 18:05, and part_of parent removal from 2008-12-15 18:11 -- but as usual please let me know if you spot anything I've missed or misinterpreted. If all is well, I'll close this item.

geneontology / go-ontology

Antibody/Immunoglobulin complex/B cell receptor complex #5631