A vendor-published report on work at Novo Nordisk describes how divergent native chemical ligation of peptide hydrazides assembled several Bowman-Birk protease inhibitor analogues from shared building blocks. Grafting a region of a potent trypsin inhibitor onto the inhibitor's…
A divergent native chemical ligation strategy built on peptide hydrazides can assemble analogues of the Bowman-Birk protease inhibitor BBI , and the strongest result publicly reported is a variant with a 4-fold improvement in inhibitory effect. The variant came from grafting a region of a potent trypsin inhibitor onto the α-chymotrypsin-binding loop of the BBI scaffold. The work was led by Christian Tornøe at Novo Nordisk in Denmark, and the record of it is a vendor-published article dated February 6, 2018.
That article reports that several BBI analogues were synthesized, that the divergent route needed fewer reaction steps than a linear synthesis would have, and that the loop-grafted analogue inhibited its target more strongly by a factor of four. It does not give sequences, protocols, assay conditions, or characterization data. The chemistry can still be explained precisely, and the design logic examined, but the claims travel with limits attached.
Native chemical ligation joins two unprotected peptides through a single amide bond. One partner carries a C-terminal thioester , and the other begins with a cysteine. The cysteine thiol attacks the thioester, transferring the acyl group onto the cysteine side chain. A spontaneous intramolecular S-to-N acyl shift then moves the acyl group to the cysteine α-amine, producing a native peptide bond with a cysteine at the junction. The reaction runs in aqueous buffer near neutral pH, tolerates unprotected side chains, and gives the same amide linkage a ribosome would. Its one hard requirement is an N-terminal cysteine on the C-terminal fragment, and that constraint shapes how any target sequence is cut into pieces for assembly.
The awkward part of the chemistry has always been the thioester itself. Peptide thioesters can be made directly in Boc-based solid-phase synthesis, but the Fmoc chemistry most laboratories use relies on piperidine for deprotection, and piperidine destroys thioesters. The peptide hydrazide route solved that problem. A peptide hydrazide is stable through Fmoc synthesis and survives cleavage and storage as a solid. At the moment of ligation, a brief treatment with sodium nitrite at acidic pH converts the hydrazide to an acyl azide, which reacts with an aryl thiol to form the active thioester in situ. The ligation partner is then added, and bond formation proceeds in the same vessel. The practical consequence is that the building blocks are dormant until deliberately activated, which is exactly what a divergent synthesis strategy wants.
In practice the ligation is run with an aryl thiol catalyst such as 4-mercaptophenylacetic acid, which accelerates the transthioesterification step, and with a reducing agent to keep cysteine thiols in their reactive form. The hydrazide activation step is carried out cold and acidic; the coupling step that follows is buffered near pH 7. Because the two operations are separate, each can be optimized independently. For a peptide chemist this means the risky chemistry is confined to a short, well-defined activation, and the fragments themselves never have to survive harsh conditions beyond the synthesis that made them.
BBI is a good match for this chemistry. The soybean inhibitor is a small protein of roughly 70 residues with fourteen cysteines arranged as seven disulfide bonds. A single solid-phase chain of that length is difficult to make reliably: deletions accumulate, and purification becomes the real product. Fragmenting the sequence into pieces of 15 to 30 residues, purifying each piece, and ligating them is the standard answer. The cysteine-rich sequence that gives BBI its stability also supplies ligation junctions, and after assembly the reduced chain can be folded and oxidized to the native disulfide pattern. Where a junction is needed and no cysteine exists, one can be installed for the ligation and desulfurized to alanine afterward; if alanine is the native residue at that position, the junction leaves no trace.
BBI is the founding member of a family of serine protease inhibitors found in legumes and cereals. Its architecture is a tandem duplication: two homologous subdomains, each built around a solvent-exposed reactive-site loop held rigidly by disulfide bonds, connected by a short linker. This gives BBI its double-headed character. In the soybean protein, one loop inhibits trypsin with a lysine at the P1 position, the residue that enters the trypsin S1 pocket. The other loop inhibits α-chymotrypsin with a leucine at P1. The loops engage the protease the way a substrate would, but the disulfide-anchored scaffold locks the scissile bond in place, and cleavage is so slow that the protein acts as an inhibitor.
The double-headed organization is the defining trait of the Bowman-Birk family. Some family members pair a trypsin site with an elastase site or a second trypsin site, and the two loops behave independently: occupying one loop with its target protease does not block the other. For engineering, this means the two halves of the molecule can be modified in parallel, and a panel of analogues can explore both loops at once. The Novo Nordisk work could change one loop while relying on the rest of the scaffold to hold the other loop and the overall fold in place. That independence is a large part of why the scaffold is worth the synthetic effort.
That architecture also makes BBI attractive as a scaffold for protein-protein interaction PPI leads. PPIs regulate many cellular processes and are attractive drug targets, but their interfaces are large and flat, and small molecules struggle to engage them selectively. The alternative is a protein-sized lead, and protein-sized leads are hard to modify and fine-tune with ordinary chemistry. A small, rigid, disulfide-stabilized scaffold such as BBI splits the difference: the folded core supplies stability and a defined geometry, while the reactive loops present a small binding epitope whose sequence can be changed. Because the P1 residue is the dominant specificity switch, replacing it redirects the inhibitor from one protease to another. The residues on either side of P1 refine affinity by adding contacts to the enzyme surface. Loop grafting transfers an entire optimized binding motif at once, rather than a single residue.
The Novo Nordisk report describes exactly that maneuver. A specific region of a potent trypsin inhibitor was grafted onto the α-chymotrypsin-binding loop of BBI, and the resulting analogue showed a 4-fold increase in inhibitory effect. Because the graft sits in the chymotrypsin loop, the gain could reflect stronger inhibition of chymotrypsin by the remodeled loop, or stronger inhibition of trypsin if the grafted region carries its own P1 specificity. The public summary does not say which protease was assayed, nor what the baseline inhibitor was. The direction of the effect is clear; the target of the effect is not.
Divergent assembly is the second half of the strategy, and it is where hydrazide chemistry shows its practical value. In a linear campaign, every analogue is a separate assembly line, with the full-length construct built by its own sequence of fragment syntheses and ligations. Changing a single loop still means rebuilding the conserved parts for that analogue. In a divergent campaign, the conserved core is made once as a set of shared fragments, and the variable loops are made as a small library of short, interchangeable pieces. Each analogue is completed at the final steps by joining the shared core to a different loop fragment. The cost of the platform is paid once and spread across the panel.
Peptide hydrazides fit this workflow because they are stable in storage and reactive only on demand. A batch of core fragments can be synthesized, purified, and characterized once, then held until needed. Activation with nitrite and thiol happens at the moment of ligation. The variable loop fragments are short, inexpensive to make, and easily diversified. The report states that the divergent approach required fewer reaction steps than a linear synthesis strategy. No step counts are given, so the claim is directional rather than quantitative, but the logic is sound: shared fragments are made once, and the number of ligations grows with the length of the longest assembly path, not with the number of analogues multiplied by the full chain length.
A practical point the summary leaves out is the folding step. Ligation yields a reduced linear chain with fourteen thiols, and the native protein is reached by oxidative folding, in which the chain forms its disulfide bonds under controlled redox conditions. Multi-disulfide folding is often where yields are lost and misfolded species appear. A divergent platform helps here as well: once a folding protocol is worked out for the scaffold, the same protocol can be applied across the panel, and the variable loops rarely change the folding problem because they are short and surface-exposed. This is another reason the shared-core design pays off beyond the ligation itself.
Everything the report actually claims fits in a short table. Each entry is stated in the source without the supporting detail that would let a reader audit it.
| Reported claim | Figure as reported | What it measures | Status in the source |
|---|---|---|---|
| BBI analogues synthesized | several | Number of distinct BBI variants assembled by divergent hydrazide ligation | Stated; no inventory given |
| Potency gain after loop grafting | 4-fold | Inhibitory effect of the grafted analogue relative to a comparator BBI | Stated; no assay given |
| Synthesis efficiency | fewer reaction steps than linear synthesis | Step count of the divergent route versus a linear route for the same panel | Stated; no counts given |
| Date the report appeared | February 6, 2018 | Publication date of the source article | Exact |
An inhibitory effect is typically quantified as an equilibrium dissociation constant Ki for the inhibitor-protease complex, or as an IC50 from a fixed-concentration assay. A four-fold shift in either number is a real but modest change. It can come from a single new contact, a better-fitting P1 side chain, or a loop that is more preorganized for binding. Without the baseline value, the absolute significance is unknowable: a shift from 1 nM to 0.25 nM is a different result from a shift from 10 µM to 2.5 µM, and the same factor of four describes both. The reported gain is a directional signal, not a quality statement.
What the result does establish is that the platform works end to end. Several analogues were assembled from hydrazide fragments, the products could be folded into active, disulfide-bridged inhibitors, and a sequence transplant from another inhibitor measurably changed function. That is a meaningful proof of concept for scaffold-directed inhibitor engineering: a small, rigid inhibitor can absorb a foreign loop sequence and behave as a tunable lead, and the divergent synthesis can produce a panel without rebuilding the conserved core for each variant.
What it does not establish is equally important. The result does not establish generality beyond a single inhibitor family and a single grafting example. It does not establish selectivity: a loop that binds its intended target more tightly may also bind related serine proteases more tightly, and no selectivity data are in the report. It does not establish a thermodynamic or kinetic mechanism. It says nothing about stability in serum, cell entry, oral exposure, or any property that would decide whether the analogue is a drug lead. The four-fold figure rests on a single published summary, and no primary publication describing this work has been identified. It should be treated as reported but not verified.
Plan the fragment map around cysteines. Every ligation junction requires an N-terminal cysteine on the C-terminal fragment, so the native cysteine positions are the natural cut points. Where a cut is needed and no cysteine exists, install one, ligate, and desulfurize to alanine. If the native residue at that position is alanine, the operation is traceless; if not, the analogue carries a substitution, and that change must be tested. Put the junctions in the conserved core, not inside a loop that will be varied, so all analogues share the same core fragments. That single design choice is what makes the divergent route cheaper than the linear one.
Verify the disulfide pattern before trusting any potency number. A protein with fourteen cysteines has an enormous number of possible pairing arrangements, and only the native pattern produces the rigid loop geometry BBI uses to inhibit. An activity assay will not reliably distinguish a correctly folded inhibitor from a partially oxidized mixture. Pair each new analogue with disulfide mapping, or at minimum with a purified, folded species whose inhibition is reproducible between batches, before comparing numbers.
Add a selectivity screen early. Serine proteases share a conserved catalytic architecture, and a loop that binds one active site tightly will often bind its close relatives. A panel of related proteases, trypsin, chymotrypsin, and elastase where relevant, is inexpensive to run and will reveal whether a potency gain is a specificity gain or just indiscriminate tightening. For a PPI-oriented lead, off-target protease inhibition is a liability, not an acceptable trade.
Read every potency claim against four questions. What is the baseline Ki or IC50 of the parent, and what is the value of the variant? Which protease was assayed, and was it the intended target or a counter screen? Under what conditions, with what substrate? And was the analogue pure and correctly folded? A claim that stops at a fold-improvement, without the absolute numbers, is a lead to follow up, not a number to quote. The same discipline applies here: the four-fold figure comes from a vendor summary with no protocols, no sequences, and no characterization data. Chase it to a primary source or reproduce it in the laboratory before building a program on it.
The record left by the report is thin, and the questions a synthetic peptide chemist would ask immediately are unanswered. The identity of the potent trypsin inhibitor that supplied the grafted region is not stated. The exact sequences of the BBI analogues are not given. The assay that produced the four-fold number is not described, and neither is the protease it was measured against. The source is a brief article in a vendor publication, written to illustrate what the chemistry makes possible rather than to report a complete study, and it is the only record of this work available. Nothing here has been independently verified.
The open questions cluster into four groups, and each one determines whether the approach transfers to another problem:
These questions matter because the scientific value of the report sits between a chemistry demonstration and a therapeutic claim. The chemistry demonstration, that divergent hydrazide ligation can build a panel of disulfide-rich inhibitor analogues, is plausible and mechanistically well grounded. The therapeutic claim is not established by a four-fold number with no context. A verification study would be straightforward to design: name the donor inhibitor and the grafted segment, give the sequences of parent and analogues, report disulfide connectivity, and measure inhibition of the target protease and related proteases with Ki values for both the parent and the variant. That package would place the four-fold factor on an absolute scale. Its absence from the public record is the reason the claim cannot yet be called a finding.
For a researcher, the defensible takeaway is the combination: a cysteine-rich scaffold, an Fmoc-compatible ligation method, and a divergent fragment plan make the systematic fine-tuning of protein-sized PPI leads practical, and a modest, well-measured potency improvement from a single loop graft is exactly the kind of increment such a platform is meant to produce. The next step is a fuller primary report with sequences, kinetics, selectivity, and folding data. Until it appears, the four-fold improvement stands as a reported result, not a verified one.
Related reading: The GABRD-P2X2 Complex Behind Progesterone's Calcium Signal in Sperm, Synthesis and Mass Spectrometry of Phosphorylated Cysteine (pCys), How Neuropeptide Y Binds the Y2 Receptor for Obesity Drug Design, FGFR2-Targeting Peptide for Early Esophageal Adenocarcinoma Detection.