D-Amino Acids and Peptidomimetic Backbones: What Each Modification Changes

A protease does not cleave a sequence. It cleaves a backbone that presents the right atoms in the right geometry inside its active site, and the side chains are what steer the chain into that geometry. Every strategy in this guide attacks the backbone rather than the sequence: invert a stereocenter, alkylate a nitrogen, insert a carbon, or move a side chain from carbon to nitrogen. Each one changes protease susceptibility, each one changes what the molecule can fold into, and each one changes what an analytical method has to do to see it. Those three consequences do not track together, which is the whole reason there are four strategies instead of one.

Two framing points before the chemistry. First, protease resistance is always resistance to the enzymes that were in the tube. Every result below was measured against a named panel, and none of them establishes a general immunity. Second, backbone modification does nothing about non-enzymatic chemistry — hydrolysis at labile linkages, oxidation, deamidation. A backbone that no peptidase will touch can still degrade on the shelf.

D-substitution: the same formula, the other hand

Replacing an L-residue with its D-enantiomer changes no atom count and no bond count. What changes is the configuration at one α-carbon, and the consequence depends entirely on how many centers you invert.

Invert all of them and you have the enantiomer of the parent, not an analog of it. Its conformational preferences are the mirror image of the parent's, and circular dichroism reports that directly: for the all-D enantiomer of polybia-MPI, Zhao and colleagues observed a CD spectrum symmetrical to the parent's and read it as a left-handed α-helix ^[2]. Jia and colleagues describe the all-D derivative of the related peptide polybia-CP in the same terms, as a left-handed α-helix ^[1]. Both groups measured protease susceptibility by taking the HPLC profile after incubation with proteases, and both reported improved stability toward trypsin and chymotrypsin for the D-substituted derivatives ^{[1, 2]}.

Invert only some centers and you have something categorically different: a diastereomer, whose conformation is not the mirror image of anything. The published examples show the fold degrading rather than reflecting. In the partial D-lysine analog of polybia-CP, Jia and colleagues report loss of some α-helical content alongside the improved protease stability ^[1]; Zhao and colleagues report loss of α-helix content in the partially substituted D-lysine analog of polybia-MPI, whose stability was nonetheless increased ^[2]. Two peptides is not a law, but it makes the point that partial D-substitution buys protease stability at the expense of the parent's fold rather than by mirroring it.

The local effects can be sharp even at one position. Halab and Lubell coupled both L- and D-amino acids to (2S,5R)-5-tert-butylproline and to L-proline and followed the resulting prolyl amides by NMR. Within the tert-butylprolyl series, the intramolecular hydrogen bond seen in the major conformer of the L-configured dipeptide became less favorable when the N-terminal residue was D; for the D-configured compound, the coupling constants of the cis isomer indicated a dihedral angle characteristic of a type VIb β-turn, and X-ray analysis of the D-leucyl compound showed that turn geometry with a cis amide ^[3]. Changing one residue's handedness re-sorted which turn the dipeptide occupied and cost it the intramolecular hydrogen bond of the L-configured major conformer, which the crystal geometry of the D-leucyl compound precluded altogether.

N-methylation: deleting a hydrogen bond donor

N-methylation replaces a backbone amide N–H with N–CH3. Net addition is CH2, a monoisotopic increment of 14.0157 Da per site. Two structural consequences follow from the substitution itself, without any measurement: the modified position can no longer donate a backbone hydrogen bond, so it cannot participate in a helical or sheet hydrogen-bonding network as a donor; and the amide becomes tertiary, which narrows the energy gap between the cis and trans rotamers relative to a secondary amide and makes appreciable cis populations possible where they were negligible before.

The proteolytic effect is real, and the cited work reports it for particular positions rather than for N-methylation in general. Janecka and colleagues synthesized analogs of a C-terminally amidated tetrapeptide bearing an N-methylated residue consecutively in each position and tested them against carboxypeptidase Y, aminopeptidase M and a rat brain homogenate. The position-2 analog, which carries sarcosine — N-methylglycine — there, is described by the authors as highly resistant to enzymatic degradation ^[4]. Dong and colleagues went further on magnitude: in a hexapeptide conjugate series, derivatives methylated at selected sites — a glutamate position among them — showed 50- to 140-fold higher half-life values in diluted serum and an intestinal homogenate ^[5]. That is a large number attached to a specific site in a specific sequence, and the authors present it as such. It is not a coefficient you can apply to a different peptide.

The costs are worth stating plainly. An N-methyl amino acid is acylated as a hindered secondary amine, so couplings into and out of the methylated position are slower and less complete than standard couplings; this is routine synthetic practice rather than a finding. And because N-methylation removes a donor, it can destabilize the very fold that made the parent interesting. Neither cost is measured by a protease assay.

β-peptides: one more carbon per residue

A β-amino acid inserts a methylene between the amino group and the carboxyl. A beta-3 residue carries the side chain on the carbon adjacent to nitrogen; a beta-2 residue carries it adjacent to the carbonyl. Each homologation adds CH2, the same 14.0157 Da increment as an N-methylation.

The folding behavior is not a degraded version of α-peptide folding — it is a different set of structures. Appella and colleagues reported that β-peptides built from carefully chosen β-amino acids adopt a stable helix defined by interwoven 14-membered-ring hydrogen bonds, the 14-helix, and then showed that a different choice of β-residues gives a distinct 12-helix, concluding that the nature of the residues affords rational control over which secondary structure forms ^[6]. The hydrogen-bond ring size, not just the handedness, is a design variable.

On proteolysis the β-backbone is the most decisive of the four modifications in the studies cited here. Gopi and colleagues built internally quenched fluorescent substrates around a proteolytically susceptible six-residue stretch of the hemoglobin α-chain and inserted β-residues at the potential cleavage sites. By fluorescence and mass spectrometry, that insertion completely abolished the action of the proteases tested, and the rate of cleavage of the peptide bond immediately preceding the modified site was also considerably reduced ^[7]. The protection extends past the modified bond itself.

Hook and colleagues then probed whether that stability could be rationalized electronically, by altering the electron availability of a given amide carbonyl bond through the introduction of polar substituents at the α-position of a single β-amino acid. They prepared β-amino acids bearing one or two fluorines or a hydroxy group at the α-position, built a series of β-peptides carrying such a residue at the N-terminus or embedded in the chain, and challenged them with a selection of enzymes: all ten tested were resilient to proteolysis, and the authors concluded that introducing a polar, sterically undemanding α-substituent does not appear to facilitate localized or general enzymatic degradation ^[8]. Their hedge is worth keeping — this is a negative result over one substituent class and one enzyme selection, not a demonstration that no enzyme can act.

Peptoids: the side chain moves to nitrogen

A peptoid residue is an N-substituted glycine: the side chain hangs off the backbone nitrogen instead of the α-carbon. Two things vanish in that move. The backbone stereocenter is gone, because the α-carbon now carries two hydrogens. The backbone hydrogen bond donor is gone as well, because every amide is tertiary. Sanborn and colleagues state both properties directly — the peptoid backbone is achiral and lacks hydrogen-bond donors — and show what has to substitute for them ^[9].

What substitutes is the side chains. In the same work, an amphipathic 36-residue N-substituted glycine oligomer bearing α-chiral side chains gave intense, α-helix-like CD spectra in water, trifluoroethanol and methanol, with no significant concentration dependence; the helical structure was reported as virtually unaffected by heat even in aqueous buffer containing 8 M urea, which the authors took as consistent with a dominant role of steric forces in the stabilization ^[9]. Note the form of that last claim: it is the authors' reading of a denaturation result, not a measurement of the forces.

The shape those helices take was proposed computationally first. Armand and colleagues used molecular mechanics, supported by semi-empirical quantum calculations, on peptoids whose N-substituents have chiral centers at the carbon attached to nitrogen; the results suggested that an octamer of (S)-N-(1-phenylethyl)glycine should form a right-handed helix with cis amide bonds, similar to the polyproline type I helix, a model they described as consistent with CD studies of these molecules ^[10]. That is a model consistent with a spectroscopic observable, which is a weaker claim than a solved structure.

Because peptoids cannot hydrogen-bond along the backbone, their conformation is set by the side chains and by the cis/trans amide equilibrium those side chains bias, and the field is still finding new ways to pull on that equilibrium. Gimenez and colleagues showed by NMR and X-ray that non-chiral, neutral fluoroalkyl monomers influence the Kcis/trans equilibria of peptoid amide bonds in model systems, and called the resulting cis preference highly unprecedented given that neither chirality nor charge was used to obtain it ^[11]. Neither that work nor the denaturation study isolates which contribution dominates; both are reported as observations in defined model systems, and what generally governs the peptoid amide equilibrium is still an open question.

How each modification is detected

The four modifications sort into two analytical classes, and the split is not the one people expect.

• Mass-visible. N-methylation and β-homologation each add CH2, so both shift the intact monoisotopic mass by 14.0157 Da per site. Three such modifications give +42.0470 Da — distinguishable at high resolution from an acetylation at +42.0106 Da, a 0.0364 Da gap. But a mass shift says how many, not which positions, and it does not distinguish an N-methylation from a homologation. Fragmentation or independent synthesis of the candidate isomers has to place them.
• Mass-invisible. D-substitution adds nothing at all. Neither, in the general case, does peptoid substitution: an N-substituted glycine residue and the α-amino acid residue carrying the same side chain are constitutional isomers with identical molecular formulas — an N-benzylglycine residue and a phenylalanine residue are both C9H9NO. Readel and colleagues make the point for stereochemical isomers explicitly: such peptides are indistinguishable by mass alone, and selective liquid chromatography tandem mass spectrometry is required to tell them apart. Examining tryptic fragments carrying D-Asp, L-iso-Asp, D-iso-Asp and D-Ser modifications, they found ionization and fragmentation to vary among identical-mass peptides in ways that affect the analysis ^[12].

For stereochemistry specifically, the established route runs through hydrolysis and chiral derivatization, and it carries a confound that has to be controlled. Goodlett and colleagues addressed it directly: peptides were hydrolyzed in deuterated hydrochloric acid and deuterated acetic acid, derivatized with Marfey's reagent to convert the released amino acids into diastereomers, and resolved by reversed-phase HPLC with electrospray MS. Deuterated solvent is necessary, the authors explain, because the original D/L ratio in a peptide changes during acid hydrolysis through racemization; a residue that racemizes in the workup picks up one deuterium at the α-carbon and so is separated by mass from one that did not. They applied it to determine chiral purity at each position of a hexapeptide by-product isolated from a batch of a synthetic pentapeptide ^[13]. A chiral amino acid result reported without such a control includes the sample preparation in its number.

The conformational readouts follow the modification. CD reports the enantiomeric relationship for full D-inversion as a mirrored spectrum ^{[1, 2]} and reports peptoid helicity as an intense α-helix-like signature ^[9], but it is a population average and assigns no residue. NMR is what resolves cis/trans populations and turn geometry at named positions, as in the prolyl amide work ^[3] and the peptoid equilibrium measurements ^[11]. Protease susceptibility itself is measured, not inferred: incubate with a defined enzyme and read the chromatogram ^{[1, 2]}, or use a substrate whose cleavage produces a signal and confirm the products by mass ^[7].

What remains open

Three gaps are worth naming. Partial D-substitution has no predictive rule — whether a given diastereomer keeps or loses the parent fold has been determined case by case, and the two examples above both lost helical content ^{[1, 2]}. N-methylation's proteolytic benefit is reported for particular sites in particular sequences, and no cited work supplies a way to choose the position in advance ^{[4, 5]}. And the cause of peptoid ordering is not pinned down: a steric account is offered as consistent with the denaturation behavior ^[9], while a non-chiral, uncharged monomer class was separately shown to bias the amide equilibrium ^[11], and no cited work isolates the dominant contribution. In all four modification classes, "protease resistant" is shorthand for resistant to a named panel under named conditions, and nothing in the cited literature licenses dropping the panel from the claim.