Insulin Structure: A6-A11, A7-B7 and A20-B19 Disulfides

Human insulin is two polypeptide chains held together by three disulfide bonds. One is intrachain, linking cysteines A6 and A11 within the A-chain. Two are interchain, A7 to B7 and A20 to B19, covalently joining the A-chain to the B-chain (2003, PMID 14744022; 2018, PMID 30339677). That count of one intrachain plus two interchain bonds is the standard description of insulin structure in the A-chain and B-chain literature, and it is the whole of the molecule’s covalent cross-linking. Nothing else holds the two chains together.

Which residues do insulin’s three disulfide bonds connect?

Each bond has a fixed pair of cysteine positions, and the naming convention states the chain letter followed by the residue number in that chain.

Bond	Type	What it joins	What is known about it	Source
A6-A11	Intrachain	Two cysteines inside the A-chain	Replacing it with a rigid, non-reducible C=C “dicarba” linkage altered the conformational flexibility of both the A-chain N-terminal helix and the B-chain helix	PMID 14744022; PMID 29899115
A7-B7	Interchain	A-chain to B-chain	One of the two covalent links between the chains	PMID 14744022
A20-B19	Interchain	A-chain to B-chain	The second covalent link between the chains	PMID 14744022

Table 1: the three disulfide bonds of human insulin by residue position and type. Sources: 2003 report on the role of disulfide bonds in insulin structure and activity, PMID 14744022, for the bond positions; 2018 dicarba A6-A11 study, PMID 29899115, for the substitution work. The one-intrachain, two-interchain split is stated independently in PMID 30339677.

The two independent descriptions agree on the arrangement without using the same vocabulary. The recombinant-expression report describes insulin as composed of A- and B-chain containing three disulfide bonds, one intrachain and two interchain (2018, PMID 30339677); the bond-position report assigns those same three to A6-A11, A7-B7 and A20-B19 (2003, PMID 14744022). Neither figure contradicts the other, so the positions can be read as the specific case of the general count.

How many cysteines is that, and which chain carries them?

Six, and they are not distributed evenly. Each disulfide bond consumes two cysteine residues, so three bonds require 3 × 2 = 6 cysteines. Sorting the positions from Table 1 by chain gives A6, A7, A11 and A20 on the A-chain and B7 and B19 on the B-chain: 4 + 2 = 6 (positions from PMID 14744022).

Two-thirds of insulin’s cystine content therefore sits on the A-chain, which is the shorter of the two. The B-chain contributes only the two partners it needs for the interchain bonds and carries no intrachain bond of its own.

The residue numbering also sets a floor on chain length that the bond positions alone will support. A disulfide at A20 requires an A-chain of at least 20 residues, and a disulfide at B19 requires a B-chain of at least 19; that is an inference from the bond nomenclature in PMID 14744022, not a measured chain length, and neither source in this set reports the total residue counts.

Why does the A6-A11 bond get replaced in analogue work?

Because it is the one bond that can be removed without breaking the two chains apart, which makes it the only tractable target for testing what a disulfide contributes beyond simple tethering. Substituting the A6-A11 cystine with a rigid, non-reducible C=C dicarba linkage produced an analogue whose cis-alkene form permitted the conformational flexibility of the A-chain N-terminal helix needed for high-affinity insulin receptor binding, and which showed surprisingly rapid activity in vivo (2018, PMID 29899115).

The result that came out of that work was not the one the design predicted. Rapid-acting insulin is usually attributed to reduced self-association, but cis-dicarba insulin is not inherently monomeric, unlike the LysB28 ProB29 analogue (KP insulin), and cis-dicarba KP insulin lowered blood glucose even faster than KP insulin alone (2018, PMID 29899115). An inability to oligomerize was therefore ruled out as the explanation for the rapid onset.

The same study reports a cross-chain effect that a purely local view of A6-A11 would miss. MALDI analyses and molecular dynamics simulations showed the A6-A11 linkage controls the conformational flexibility of the B-chain helix, and through it the overall structural stability of insulin, separately from its regulation of A-chain N-terminal helix flexibility (2018, PMID 29899115). One intrachain bond in the A-chain is doing work on the other chain.

Speed came at a stability cost that the authors state plainly: neither dicarba species is stable, as assessed by fibrillation and thermodynamics assays (2018, PMID 29899115). High-affinity receptor binding, rapid in vivo activity and stability were all traced back to the specific conformational arrangement of that single linkage, which is why it remains a design lever rather than a solved problem. Insulin is one of the oldest entries on any list of peptide drugs cleared by regulators, and its analogue program is still built around this bond.

What makes chain combination the hard step in making insulin?

Regiochemistry. Six free cysteines can pair in more ways than the one native arrangement, and the combinatorial ceiling is easy to compute. For 2n cysteines the number of complete pairings is the double factorial (2n − 1)!!, so six cysteines give 5 × 3 × 1 = 15 possible complete sets of three bonds, of which one is native.

Scaffold	Disulfide bonds	Cysteines (2 × bonds)	Possible complete pairings, (2n − 1)!!	Share that is the native set	Bond count source
Native insulin	3	6	5 × 3 × 1 = 15	1/15 = 6.7%	PMID 30339677; PMID 14744022
Native INSL5	3	6	5 × 3 × 1 = 15	1/15 = 6.7%	PMID 24188028
Minimized INSL5 analogue	2	4	3 × 1 = 3	1/3 = 33.3%	PMID 24188028
Single-disulfide reference	1	2	1	1/1 = 100%	arithmetic only

Table 2: pairing combinatorics computed from the disulfide-bond counts reported in PMID 30339677, PMID 14744022 and PMID 24188028. These are combinatorial ceilings assuming every cysteine is free to pair with every other, not measured folding yields; none of the cited reports gives a random-pairing yield figure.

Dropping from three bonds to two cuts the pairing space by 15 / 3 = 5-fold. That is the arithmetic behind describing a two-disulfide construct as easier to assemble than a three-disulfide one.

Chemistry has attacked the problem from two directions. A 2016 review of synthesis within the insulin superfamily divides the milestones into disulfide bond formation driven by protein folding, and chemical reactivity-directed sequential disulfide bond formation (2016, PMID 26910514). The first lets the peptide find the native set; the second forces each bond in a chosen order.

Both approaches run into the same two obstacles, and the review names them: the hydrophobic nature of the individual A-chain and B-chain, and the need for selective disulfide formation under mildly oxidative conditions (2016, PMID 26910514). Solubility and selectivity, not chain assembly, are the limiting factors. This is the kind of manufacturing burden that separates peptides from small-molecule drugs at the process-chemistry stage.

Recombinant routes inherit the same folding step. One expression system builds the A-chain and B-chain as separate human αB-crystallin fusion proteins in Escherichia coli, releases each chain by cyanogen bromide cleavage, then runs chain combination under oxidative conditions with αB-crystallin acting as a molecular chaperone, before purifying natively folded insulin by phenyl sepharose hydrophobic interaction chromatography (2018, PMID 30339677). The chains are made separately; the disulfides are formed afterwards.

That report validates the product by function rather than by structure alone, using an insulin tolerance test in mice alongside biophysical methods to compare the recombinant hormone with authentic insulin (2018, PMID 30339677). It describes the yield as appreciable and the purity as high without publishing a numeric value for either, so the efficiency gain over an unchaperoned combination cannot be quantified from the abstract. The hormone being folded here is the same one whose release GLP-1 receptor agonists potentiate at the beta cell.

Does a two-disulfide version of the insulin fold still work?

In at least one member of the superfamily, yes. Insulin-like peptide 5 (INSL5) is a two-chain hormone constrained by three disulfide bonds in a pattern identical to insulin, and it is extremely difficult to make by either chemical or recombinant means (2013, PMID 24188028).

The specific failure modes are chain-specific and they mirror insulin’s. The INSL5 A-chain is very poorly soluble and the B-chain is highly aggregating, which together make post-synthesis handling and purification very difficult (2013, PMID 24188028). Solubility problems on one chain, aggregation on the other.

The response was to delete a bond rather than improve the process. A highly active INSL5 analogue with a simpler two-disulfide structure proved easier to assemble than the native three-disulfide peptide and now serves as a mimetic for investigating INSL5’s functional role, including its reported involvement in insulin secretion and beta-cell homeostasis (2013, PMID 24188028).

For insulin itself, the 2016 superfamily review argues that synthesis is no longer the limiting factor in structural and biological investigation of these hormones, and names the refinement of structure toward a glucose-sensitive insulin as the target that the chemistry is now being pointed at (2016, PMID 26910514).