Why does the highest-affinity clone sometimes fail on Western blot?
What was made, and what was measured
The target was trimethylated Lys260 of MAP3K2, a non-histone methylation site placed there by the oncogenic methyltransferase SMYD3. Methylsite-specific monoclonal antibodies are, in the authors’ words, rare to nonexistent — the antibody has to discriminate the addition of three methyl groups on a lysine, and it has to do so in a specific sequence context rather than merely recognizing “a methylated lysine” anywhere in the proteome.
Rabbits were immunized with the KLH-conjugated methylated peptide. A Fab phage library was built from bone marrow and spleen, and four rounds of panning were run — with an excess of soluble non-methylated peptide present during binding, so clones that could not discriminate were competed away before screening. Six unique methylsite-specific Fabs came out; four were taken through SPR, Western blot, crystallography and molecular dynamics.
| Clone | KD, trimethylated peptide | Detects full-length protein on Western blot? | Methyl-lysine buried surface |
|---|---|---|---|
| E6 | 0.90 nM — tightest in panel | Faint band only | 34.0 Ų exposed |
| C9 | 14 nM | Yes — clean, SMYD3-dependent | 3.8 Ų exposed |
| D6 | 54 nM | Yes — clean, SMYD3-dependent | 0.0 Ų — fully buried |
| F9 | 300 nM | No — off-target bands only | 45.3 Ų exposed |
Affinities for the corresponding non-methylated peptides fell in the millimolar to 0.3 µM range. Specificity was confirmed against a K260A mutant, where the band disappeared, and against unrelated trimethylated histone peptides (H4K5me3, H3K27me3), which bound far more weakly or not at all.
The structural reason the tight binder failed
Four Fab–peptide co-crystal structures (PDB 6LDV, 6LDW, 6LDX, 6LDY) make the mechanism visible. Every clone grips the trimethylated lysine in an aromatic cage — the same principle used by natural histone reader proteins — but the cages sit at different depths and are built from different residues: F9 uses two tyrosines and a tryptophan, C9 two phenylalanines and a tyrosine, E6 three tyrosines, D6 three tryptophans. The chemistry is interchangeable; the geometry is not.
The clones that worked on protein, C9 and D6, bury the methylated lysine deep in the cleft between heavy and light chain — completely, in D6’s case. The clones that failed, F9 and E6, instead form several hydrogen bonds and salt bridges to the peptide’s free C-terminal carboxyl group. That carboxyl does not exist in a full-length protein, where the sequence simply continues. The authors confirmed this directly: amidating the peptide C-terminus nearly abolished binding for F9 and E6, while C9 and D6 bound the amidated peptide with unchanged affinity.
So the tightest binder was tight for a reason that evaporates the moment you leave the peptide. This is not bad luck. It is a predictable failure mode of screening on peptide and ranking on affinity.
What we do differently because of this
- Screen in the assay the antibody has to survive in. If the deliverable has to work on endogenous protein in a Western blot or an IHC section, that is the screen that decides the panel — not the peptide ELISA and not the SPR ranking.
- Deliver a panel, not a winner. Four clones that grip the same modification through different geometry give a sponsor real options when an application changes. One clone gives none.
- Design the immunogen for the endpoint. Conjugation chemistry and C-terminal capping determine whether the resulting antibodies can cross-react to protein at all. Those choices are made before the first immunization, and they are hard to undo afterwards.
One further detail worth noting for anyone comparing platforms: E6 and F9 differ by only a single amino acid in each of CDR H1 and H2, and share nearly identical hydrogen bonding — yet E6 binds roughly 300-fold more tightly, traced to one aspartic acid near the cage that electrostatically attracts the methylated lysine. Single-residue differences of that size are exactly what a large selected repertoire finds and a small screened panel misses.
Why phage display rather than hybridoma for this target class
Hybridoma and B-cell cloning sample on the order of 104–105 cells and cannot apply selective pressure after the fact; phage display samples orders of magnitude more and lets negative selection be built into the campaign, which is what removed the non-discriminating clones here. Historically, rabbit phage display was blocked by a real biochemical problem: the rabbit Cκ1 light chain — which accounts for most of the serum response — carries an extra disulfide bond absent in mouse and human kappa chains and expresses poorly in E. coli. Systems that worked around it were limited to the underused Cκ2 chain and therefore to a fraction of the repertoire.
The library construction method and phagemid vector used in this study display fully rabbit Fabs, Cκ1 included, without resorting to chimeric constructs. The paper’s conflict-of-interest statement records the technology by patent number: the authors used patented technology WizAmp (US 9,890,414), invented by CJ. O. and T. M., for antibody acquisition.
Need an antibody that has to work on endogenous protein?
Tell us the modification, the sequence context, and the assay it has to survive in. A PhD scientist reads it and replies within two business days. No NDA needed to start.
Talk to a scientist →Frequently asked questions
Are methylation-specific monoclonal antibodies actually feasible?
For non-histone sites they are difficult but demonstrably possible. This study returned six unique methylsite-specific Fabs against MAP3K2 K260me3, two of which detect the methylated protein in cell lysate in a SMYD3-dependent manner and lose the signal against a K260A mutant.
Will the antibody cross-react with other trimethylated lysines?
It has to be tested rather than assumed. Here the working clones were tested against histone H4K5me3, which shares a GKme3GG motif with the target, and against H3K27me3 in a different sequence context. Binding to H4K5me3 was detectable but far weaker, and binding to H3K27me3 was around the millimolar level — making these antibodies methylsite-specific, not merely methyl-lysine-specific.
Should I pick clones by KD?
Not on its own. In this panel the 0.90 nM clone was nearly unusable on Western blot while the 54 nM clone worked cleanly, because the tighter clone was bound to a peptide C-terminus that does not exist in the intact protein. Rank by performance in the assay that matters, and use affinity as one input among several.
Can you distinguish mono-, di- and trimethylation?
To a degree, and it varies by clone. Affinity increased with the degree of methylation for all antibodies tested, with two clones showing the clearest preference for the trimethylated form. If methylation state has to be resolved, say so at the start — it changes the screening cascade.
Do I get the sequences?
Yes. Standard engagements return the antibody sequences to the sponsor, with no downstream royalties in standard scope.