Why are there so few full-length antibody structures?

Question

I’m a student in the Biochemical Engineering field and the professor at the department just told us in a lecture that if we want to use a full-antibody structure for simulation purposes there aren’t very many of them. He suggested that we start with 1hzh.pdb.

So I wondered whether he was correct and, if so, why there wasn’t anyone using crystallisation or NMR to look into the structures of the antibody drugs that have come onto the market. Perhaps it costs too much and it’s very time-consuming? But, even so, I imagine that it would still be worthwhile.

Answer 1

There IS a problem!

In a previous version of this answer I focussed on various historical, technical, goal-related and commercial reasons for the predominance of structures of antibody fragments (Fab and Fc) rather than complete immunoglobulins (two copies of heavy and light chains) — the “full-antibody structures” of the question. I retain this as an appendix as I think it still of interest, however I had missed the main reason.

The main reason can be found in the two papers on 1HZH — the anti-HIV antibody mentioned by the poster. In a nutshell it is:

Full length antibodies have regions that convey functional flexibility, resulting in a variety of conformational forms in the same pure protein.

To quote from the paper describing the crystallization:

The Fab and Fc domains are connected by hinge regions of exceptional flexibility that facilitate changes in shape for bivalent attachment to antigen and for effector interaction. Electron microscopy of rabbit and mouse IgG has provided evidence for flexibility and arm rotation at the Fc–Fab and Fab–Fab interfaces, most of which is associated with the central hinge region.
The inherent molecular flexibility has complicated crystallization of intact antibodies.

And lest there be any doubt about the relative numbers:

While well over 200 structures of antibody fragments, mainly Fab and Fab0, have been determined [in the year 2000], crystals of intact antibodies have only been reported eight times… Only Dob, Mcg, Kol, Mab231 and Mab61.1.3 have yielded structures or partial structures.

And the paper points out the limitations of even these structures:

These early results suggested that for antibodies with normal hinges the relative placement of the Fc with respect to the Fab arms is not fixed and is free to assume a wide range of configurations.

Solving the problem

The general approach adopted by the authors is summarized in their statement:

Nevertheless, a well ordered antibody crystal can be obtained if a single conformer is incorporated into the crystal lattice owing to stabilization by crystal contacts. In this manner, a single ‘snapshot’ of an antibody structure can be obtained from the plethora of possible conformers.

Achieving this was far from simple — as far as a non-expert can judge from reading their full account, and the fact that the details of their achievement was regarded as justifying publication in the top journal, Science the following year. I reproduce below an image showing what they refer to as

a “snapshot” of the broad range of conformations available in solution.

where the hinge regions are marked.

Reflection: Utility of full-length high-resolution structures

Obtaining a high-resolution full-length immunoglobulin structure is obviously important in itself, but as the answer from @Fizz documents, there had been only one other subsequent structure of this type by the time of a review in 2018. Why is this? Granted it is intrinsically difficult, but one the basis of the difficulty — the range of different structures caused by the hinge flexibility — is also the answer. No one specific structure can be regarded as definitive, so that, as the other answer illustrates, lower resolution images of a range of structures became of greater interest, and these can be related to the high-resolution full-length structures.

Appendix: Minor reasons (from my original answer)

Historically, a fragment containing the constant domain of immunoglobulins could be crystallized from polyclonal antibodies because it was identical for antibodies of different specificities (within a class such as IgG), whereas the different variable regions prevented crystalization. One had a mixture of molecules. It was for this reason the abbreviation Fc arose — it stands for fragment crystallizable. region.

Polyclonal antibodies were available from tumours before artificial means of generating them arose, but the point of interest was the antibody-binding site, so that the Fab fragment — which was easy to generate — would serve the purpose. It is much easier to crystallize and analyse such fragments than the whole immunoglobulin, so that is why they were (and are used).

Finally, one studies the areas of immunoglobulins that are the focus of your scientific interest (and for which you can convince grant authorities are worth funding). For example, if you are interested in interaction with antigen, you solve the structure of the Fab region; if you are interested in the interaction of the constant region with other molecules (or the difference between different classes) you solve the structure of the Fc region; and only if you are interested in a problem like the relative orientation of the different regions under particular situations do you make the effort required to solve the structure of whole antibody.

Answer 2

The more flexible a molecule, the more difficult it is to get a crystallization that actually tells you something about the molecule's free-phase arrangement/behavior. Basically (the best) crystallography requires you get a molecule to grow (crystals) in the same conformation.

The hinge in IgG1 for instance appears to be able to rotate from 0 to 180 degrees (although range may be more limited in actual molecules, depending on the "bulk" of what's attached to the hinge). If you somehow manage to grow a crystal of a full IgG you'll just get a snapshot of that at one angle. (That 2011 paper says there was still a single such image for a full-length [human] IgG1 for instance--the one binding HIV (IgG1 b12) found in your 1HZH file.)

According to a 2018 review only 4 full structures of IgG antibodies had been determined; by PDB entry with year of paper publication:

• 1IGT [1997] -- "a mouse IgG2 with 3 hinge disulfide (SS) bonds, while human IgG2 has 4 SS bonds"
• 1IGY [1998] -- a mouse IgG1 with 2 SS bonds
• 1HZH [2001] -- a human IgG1 with 2 SS bonds
• 5DK3 [2015] -- a human IgG4; "the hinge SS was more stabilized due to the conformational alteration of S228P mutation"

Of these, only the last one, pembrolizumab, is used therapeutically, as far as I know, but the first one, was (interstingly) a [proposed] veterinary product at one point (Mab231--against canine lymphoma); the USDA suspended its license after it failed in larger trials. The 2nd one is an antibody against phentobarbital. The 3rd/HIV one (1HZH = IgG1 b12) is not the one that's been (recently) approved for treatment as Ibalizumab.

The first of these papers (1997) notes some earlier attempts at full structure crystallography; mostly successful in special cases where there was reduced hinge mobility:

Other intact antibody structures have been studied by X-ray crystallography (Silverton et al., 1977; Guddat et al., 1993; Marquart et al., 1980). In two myeloma proteins, the flexible hinge regions connecting Fab and Fc segments were deleted (Silverton et al., 1977; Guddat et al., 1993). The molecules were structurally restrained and, perhaps for this reason, appeared as compact T-shapes, the angle between Fabs close to 180°. A third antibody, Kol, had an intact hinge, but the Fc was so disordered that it was not possible to orient it with respect to the Fabs (Marquart et al., 1980). The two Fabs and a portion of the hinge (upper and core) were, however, visualized.

The first two [fully successful] papers (1997-1998) came from the same lab/group. As I'm not expert in this area (or even in biochemsitry) it's hard for me say what was the breakthrough that enabled the full structures IgG crystal analysis in the late 1990s; the papers aren't incredibly explicit on that. There was certainly a heavy amount of computation involved on the X-ray diffraction data, conducted on a Cray C-90 supercomputer at that time; mentioned in these late 1990s papers. And the diffraction experiments were not conducted in the university where the authors were employed but at the Stanford Synchrotron Radiation Laboratory (SSRL) and Brookhaven National Synchrotron Laboratory, respectively; almost certainly because they used the large synchrotrons available at these labs, as opposed to smaller X-ray sources. There are entire papers on the subtelties of cryportection needed in order to leverage high-energy X-ray sources.

The 2001 short paper in Science on IgG1 b12 is actually devoid of much method details, but the same (Scripps) group has published a longer paper in 2002 in JMB; there they say they (also) ran their diffraction at SSRL (beamline 7-1). No details on computational resources though.

The 2015 paper on IgG4 acknowledges the Canadian Light Source where the diffraction was conducted on beamline 08ID-1 with synchrotron radiation. Interestingly, this paper also used TROSY (NMR) to confirm packing details and 1D NMR to compare hinge behavior in solution of a serine-to-proline replacement in the hinge (more on NMR below).

The 1997 and (to a lesser extent) the 2015 paper have most details on the computations involved (well, in this group of 4 papers). Basically prior Fc and Fab crystal diffraction "signatures" (ideally for the same antibody) are highly desirable and the software basically does a search trying to locate them in the larger "picture". But this can fail and manual hints based on parts of Fc/Fab (e.g. from similar but not the same antibody) may have to be entered to guide the search etc.

The 2002 paper has some hints why the automated search can fail (beyond the hinge problem) even when the Fab's structure has been previously analyzed separately; when the whole antibody is crystallized asymmetries are observed:

Importantly, all three of these intact antibodies crystallize with one whole IgG molecule per asymmetric unit, even though their respective space groups and crystallization conditions vary. Thus, although the two halves of the antibody are identical in sequence, they are not identical in structure.

A (single) crystal structure will not--in itself--let you experimentally verify how much hinge freedom there is. Tomography techniques such as IPET are more sensible for the latter. Resolution is (much) lower, but if just want to capture the angles available in the hinge with something attached, it's far less time consuming. Feed the results into a targeted molecular dynamics (TMD) package together with the crystal structure of Fab & Fc, and you get something like this

---

As for [3D] NMR (which you also suggested), it cannot be easily used because of the too greater size of full-length antibodies (IgG: 145-160 kDa) vs 40 kDa typical limit for NMR. Basically with 3D NMR, the bigger the molecule, the "blurrier" the image your're going to get. While Wikipedia mentions a couple of proposals/experiments for larger molecules, it seems those didn't really take off in everyday practice.

There are actually some papers that have done spectroscopy NMR on full-length antibodies, but present the results with caveats like:

It should be noted that mAb2 used in this study [...] is an example of an intrinsically stable and soluble antibody. Despite this, at a very high mAb2 concentration the measurable NMR parameters registered quite significant differences as solvent conditions were varied, highlighting the inherent sensitivity of this NMR technique. It can be anticipated that other, less stable mAbs, which require more careful formulation to achieve satisfactory solubility and stability profile, would show even greater variation in NMR measurables.

There are also several papers e.g. Marino et al. (2017), Brinson et al. (2019) on 2D NMR on MAb, but again for the purpose of determining the higher-order structure (in solution); although the resolution may be a bit better than IPET (but still below crystallography), it requires more complex math.

Answer 3

There are thousands of structures in the PDB of neutralizing antibodies against all sorts of viruses using all sorts of methods (NMR,x-ray,electron microscopy) as complex or stand alone. I guess your professor has certain requirements that are rarely met, e.g. specifically anti-HIV-1 IGG of certain quality; then he/she only found one structure that suffices.