The retro-inverso transformation is two operations applied at once: read the sequence backwards, and invert the configuration at every alpha-carbon. Each operation on its own mirrors where the side chains sit. Inversion of amino acid chirality from all-L to all-D mirrors side-chain positions and interactions in conformational space, and a similar mirroring of side-chain positions is independently achieved by reversing the direction of the backbone [1]. Applied together the two mirrorings largely cancel, so a retro-inverso peptide is built from D-amino acids in a reversed sequence and, when extended, assumes a side chain topology similar to that of its parent molecule but with inverted amide peptide bonds [2].
That last clause is where the method lives or dies. The proteolytic machinery is not well equipped to deal with D-amino acid polypeptides, and it is that fact above all else that drove research into stereochemical and directional manipulation of peptide chains [3]. You get the stability. Whether you keep the activity depends entirely on whether the interaction you care about was reading side chains or reading the main chain.
Retro, inverso and retro-inverso are three different molecules
The cleanest experimental statement of the underlying symmetry comes from a four-member set built on IRGERA, the C-terminal residues 130-135 of histone H3: the natural L-peptide, the D-enantiomer, a retro-peptide in which each residue kept its chirality, and a retro-inverso peptide in which each was inverted [4]. The retro-inverso peptide mimicked the structure and antigenic activity of the natural L-peptide but not of the D- and retro-peptides, whereas the retro-peptide mimicked the D-peptide but not the L- and retro-inverso-peptides [4]. The pairing is L with retro-inverso, and D with retro. An all-D analogue of the same sequence is not a retro-inverso analogue and should not be expected to behave like one.
This also separates the transformation from the more common tactic of substituting one or two D-residues at hydrolysis-prone positions. That is a local edit that leaves the chain direction and the rest of the backbone hydrogen-bond network intact. Retro-inversion is systematic inversion of the stereochemistry at every backbone alpha-carbon accompanied by chain reversal [3], and it changes the molecule everywhere at once.
The backbone is not restored, only the side chains
Reversing chain direction while restoring side-chain positions means the amide units themselves are turned around: the carbonyl and the amide N-H that the parent presented at a given point along the chain have swapped. Crystallography on the model retro-inverso dipeptide Bz-S-gAla-R-mAla-NHPh makes this concrete. It adopts an extended backbone conformation in which the NH groups of the gem-diamino alkyl moiety and the CO groups of the malonyl residue face side by side, presenting all carbonyls on the opposite side from the NH groups [5]. In the crystal that extended form is stabilised by interstrand H-bonds running in a single direction through the parallel beta-sheets of the packing, and those sheets differ from the beta-sheets formed by native amino acids only; NOESY experiments suggest a similar conformation prevails in dimethylsulfoxide [5]. Earlier potential-energy calculations on gem-diamino alkyl and malonyl alanine residues had predicted extended forms to be less stable than helical ones because of strong electrostatic repulsion between parallel polar groups, and the structure shows that favourable cooperative interactions among head-to-tail aligned peptide dipoles of facing strands more than compensate, giving two C=O...H-N hydrogen bonds per residue [5].
Read that as the general case rather than a curiosity. A retro-inverso chain can be extended, and it can hydrogen-bond well, but its donor and acceptor register is its own. Anything that recognised the parent's register recognises something else here.
The end-group problem
At the two ends the transformation does not close. Reversal maps the parent's amino terminus onto the analogue's carboxyl end, so the terminal functional groups no longer correspond to the parent's however carefully the interior is matched. An early set of partially modified retro-inverso analogues, enkephalinamides in which the peptide bond between residues 4 and 5 was reversed with or without simultaneous reversal of the carboxyl-terminal amide bond, were the experiment that fixed the point: partial reversal of the peptide bonds of the backbone can result in peptides with enhanced activity compared to a parent compound, provided that the structural complementarity of both the side chains and end groups are conserved [6]. End-group modified retro-inverso was formulated as a transformation in its own right alongside the partially modified retro-inverso concept, and both rest on the same two non-standard building blocks, the gem-diaminoalkyl and the C2-substituted malonyl residues [7].
Those blocks carry real synthetic cost. Gem-diaminoalkyl residues are prepared from peptide amides using [bis(trifluoroacetoxy)iodo]benzene, and N-t-butyloxycarbonyl tyrosine and N-t-butyloxycarbonyl tryptophan decompose on treatment with that reagent unless the side chains are protected, as the t-butyl ether and the N-in-formyl derivative respectively [8]. On the malonyl side, racemic 2-alkylmalonyl diastereomers were found to be separable by HPLC, and the chiral stability of peptides containing optically active malonyl residues had to be investigated under simulated physiological conditions [8]. The malonyl alpha-carbon sits between two carbonyls; treat its configuration as something you measure at the end of the synthesis rather than something you assume, and expect a diastereomeric separation problem if you do not.
Where the approximation holds
Every careful statement of the concept carries the same qualifier. The expectation was that systematic inversion of backbone stereochemistry, accompanied by chain reversal, should yield proteolytically stable retro-inverso isomers whose side chain topology, in the extended conformation, corresponds closely to that of a native sequence [3]. Extended is doing the work in that sentence, and it is the same word that appears in the standard definition [2]. Turns and extended segments survive the transformation far better than anything whose fold is carried by a regular intramolecular hydrogen-bond ladder.
The best-documented successes are antibody recognition of short, non-helical epitopes, an area in which the strategy has been more successful than elsewhere [3]. Solution NMR structures of a 19-mer corresponding to the major antigenic region of foot-and-mouth disease virus (FMDV) and of its retro-inverso analogue show that the two peptides exhibit similar folding features, particularly in their C termini, which may explain in part their cross-reactive properties; the retro-inverso analogue is more rigid than the parent and contains five atypical beta-turns, a feature offered as an explanation for why retro-inverso peptides of this system are often better recognised than the parent peptide by anti-virion antibodies [9]. The same study contains a warning worth more than the result: in trifluoroethanol, where the fold is alpha-helix as confirmed by CD, the antigenic activity of both peptides was completely abolished, indicating that the helix is not the conformation the antibodies read [9]. Run your structural work in the solvent your binding assay uses, not in the co-solvent that gives the tidiest CD trace.
A second near-neutral case sits outside immunochemistry. For a fragment of the salivary protein MUC7, standard enantiomeric exchange and retro-inverso modification had a minimal effect on the secondary structure and biological activity of the peptides and their Cu(II) complexes, yet significantly influenced the thermodynamic stability of those systems [10]. Even when the readout you chose does not move, something else has.
Where it fails, and by how much
Helices are the documented failure mode, and the penalty has been quantified repeatedly. Retro-inverso isomerisation diminished p53(15-29) binding to MDM2 or MDMX by 3.2-3.3 kcal/mol; similar results were replicated with the C-terminal domain of HIV-1 capsid protein at 3.0 kcal/mol and the Src homology 3 domain of Abl tyrosine kinase at 3.4 kcal/mol [2]. CD and NMR spectroscopic as well as X-ray crystallographic study showed that D-peptide ligands of MDM2 invariably adopted left-handed helical conformations in both free and bound states [2] - the analogue does not convert to the parent's handedness on engaging the target, which is the structural account offered for the penalty.
Scaffolding the helix does not rescue it. Stingin is a rationally designed miniature protein of 18 residues with an N-terminal loop and a C-terminal alpha-helix held by two intramolecular disulfide bridges, binding MDM2 through that helix. Measured by fluorescence polarisation and surface plasmon resonance, retro-inverso isomerisation of L-stingin weakened its MDM2 binding by 720-fold, or 3.9 kcal/mol; enantiomerisation drastically reduced binding by three orders of magnitude, and sequence reversal alone abolished it [11]. Note the analyte class before you generalise: that is a disulfide-constrained miniature protein, not a linear peptide.
Simulation and spectroscopy decompose the loss into two separate errors. Replica-exchange molecular dynamics with circular dichroism on the same p53 segment found that the D-peptide analogues adopt a predominantly left-handed helical conformation, and that reversed sequences, as either L- or D-peptides, display a greater helical propensity, a feature reflected by NMR and CD in TFE/water [12]. While approximately similar orientations of the side chains are possible, the ability to mimic the parent peptide is severely compromised by backbone orientation for the D-amino acids and by side-chain orientation for the reversed sequences, so the retro-inverso peptide is disadvantaged as a mimic in both aspects and further chemical modification is required [12]. The two halves of the transformation do not simply cancel at the level of the folded state.
The other failure class is any interface that reads the main chain directly. Retro-inverso analogues of MHC class II restricted peptides, which by construction contain the correct orientation of the side chains but an inverse main chain, were unable to compete in I-E(d) or I-A(d) binding tests, demonstrating a very low if any capacity to bind MHC II molecules, and confirming previous structural data that hydrogen bonds between residues of MHC II molecules and the main chain of antigenic peptides play a major interacting role [13]. That gives a cheap pre-registration test: if the published structure of your complex shows the interface carried by main-chain hydrogen bonds rather than side-chain contacts, retro-inversion is predicted to fail and the binding assay is a formality.
Partial retro-inverso as the compromise
The critical reviews converge on the same escape. Partial rather than global retro-inversion holds much promise, since the loss of crucial backbone hydrogen-bonding through peptide bond reversal can be avoided while still permitting stabilisation of selected hydrolysis-prone peptide bonds; generically applicable synthetic methods for such partially modified retro-inverso peptides are not as yet available [3]. In practice that means identifying the scissile bond, reversing only that one, and leaving the rest of the hydrogen-bond network in its native register, at the cost of the two non-standard residues and their stereochemical bookkeeping [7, 8]. It is a per-sequence synthetic project rather than a lookup.
Protease resistance is the motivation; measure it as its own endpoint
The stability comes from the D-configuration, not from the reversal. The inverted chirality renders retro-inverso peptides less prone to degradation by endogenous proteases, conferring enhanced half-lives, but given their general incapability to adopt the 3D structure of the parent peptides their application should be carefully evaluated and investigated case by case [14]. Both halves of that sentence are load-bearing.
A clean in-cell measurement of the stability side used a collection of SH2 and SH3 interaction motifs of similar size, differing in charge, each carrying an N-terminal fluorescein, introduced into cells by electroporation. The peptides differed strongly in retention of intracellular fluorescence, with half-lives ranging from only 1 to more than 10 h, and residence times were greatly increased for retro-inverso peptides, demonstrating that rapid loss of fluorescence is a function of peptide degradation rather than the physicochemical characteristics of the peptide; fluorescence correlation spectroscopy served as a separation-free way to follow degradation in crude cell lysates and in intact cells [15]. Resistance to trypsin was likewise compared head to head with antigenicity and immunogenicity in the IRGERA series [4]. Run the stability comparison against the parent under the same conditions, and report it separately from the activity comparison.
The analytical work the claim requires
There is a specific cautionary case. An earlier report had concluded, from inhibition enzyme-linked immunosorbent assay, that the retro-inverso isomer of p53(15-29) retained the same binding activity as the wild type peptide, and the discrepancy with the penalty described above was resolved only by applying a battery of biochemical and biophysical tools [2]. One competition format, particularly one with an immobilised or multivalent component, can be blind to a difference of several kcal/mol.
So measure direct affinity for the parent and the analogue under identical buffer and temperature, by two methods that fail differently. The p53 work used surface plasmon resonance, isothermal titration calorimetry and tryptophan fluorescence [2]; the miniature-protein work used fluorescence polarisation and surface plasmon resonance [11]. Report the difference in binding free energy rather than a fold-change in a downstream functional readout, because the functional readout can saturate.
Pair that with structure. CD will tell you whether the analogue populates a helix of the opposite handedness [2]; solution NMR in the assay solvent will tell you what the ensemble actually looks like, and the FMDV work is the reminder that a helix-inducing co-solvent can produce a fold that is irrelevant to recognition [9]; a crystal structure of the complex settles the question when you can get one [2].
Finally, verify the stereochemistry you think you made. A retro-inverso analogue is an isomer of its parent, so accurate mass cannot distinguish the two, and a single residue that survived as L is a mass-silent defect. Chiral amino acid analysis by pre-column derivatisation with Marfey's reagent, 1-fluoro-2,4-dinitrophenyl-5-L-alanine amide, or one of its structural variants carrying a different chiral auxiliary, followed by HPLC separation of the diastereomers so formed and, where required, online HPLC coupled to mass spectrometry, is an established route for biologically relevant synthetic peptides [16]. For partially modified constructs, add the malonyl centre to the list of things to check [8].
What a positive result licenses
If the retro-inverso analogue binds within error of the parent by two orthogonal direct methods in the solvent your assay uses, you have shown that this interface is side-chain dominated under those conditions. You have not shown that the two molecules are conformationally equivalent: the amide inversion is a structural fact of the transformation [2], and a thermodynamic parameter can move while the activity readout does not [10]. Nor does the result transfer to the next sequence. The penalties measured for helical parents fall in a narrow band across structurally unrelated targets [2], and the lack of general success reported in one review [3] and the case-by-case caution in the other [14] summarise that record rather than hedging it.