Highly stable Meldrum’s acid derivatives for irreversible aqueous covalent modification of amines

Abstract

This work describes the development of a highly stable, and pH responsive probe for lysine modification. The scaffold has marked stability in the presence of several biological nucleophiles and across a wide pH range (2–12). Several functional analogs showed robust labeling of a protein at pH > 9. Taken together, our system displays versatility and can be easily adapted for variety of applications, while demonstrating stability suitable for a wide range of biologically compatible systems.

Graphical Abstract

Protein bioconjugation techniques are critical to multiple areas of investigation; from probing the topography and reactivity of enzyme active sites¹, to modifying antibodies with imaging agents or cytotoxic drugs.^1,2 Despite the wide variety of available techniques, there is no such thing as a perfect bioconjugation reaction because each protein or system has unique needs for its reactivity.³ Often, new mechanistic insights into the reactivity and side-reactivity of small molecule probes leads to a heightened understanding of the parameters in which these probes can be used. Herein we report a modular Meldrum’s acid amine-reactive Michael acceptor (or MaMa) probe that was inspired by a desire to reversibly protect troublesome lysine side chains (Scheme 1), but became a uniquely stable tool for high pH environments and bioconjugations. Where common lysine-reactive probes such as N-hydroxysuccinimide (NHS) esters are not hydrolytically stable, and isothiocyanates face synthetic challenges including functional group incompatibility, MaMa fills a niche in applications while also bringing stability, synthetic ease, and thermodynamic selectivity.⁴

Scheme 1.

Conjugation of amine to conjugate acceptor 1, followed by subsequent “declick” and release of amine upon addition of DTT was previously shown. This work describes a second amine addition at pH >9 as a stable bioconjugation

Lysine is highly abundant on the surface of proteins, and, as one of the more nucleophilic amino acids, can often wreak havoc when attempting to modify proteins with electrophiles targeted to other amino acids. Inspired by Anslyn’s initial “click declick” work with Meldrum’s acid derivative 1 (Scheme 1), we set about protecting, and unmasking lysine residues on proteins, hoping to leave native lysine intact after deprotection.^5,6 In order to accelerate the reaction, we predicted that a higher pH would be beneficial, but were met by surprising results. To our consternation, we observed covalent protein aggregation that was wholly irreversible (Supplemental Figure S1). Based on established sulfur reactivity and reversibility,^5,7 we hypothesized that our crosslink resulted from a second, non-sulfur nucleophile. We set about elucidating the nuances of this reaction mechanism. To get a better picture of what was happening, we turned to small molecule studies.

As a proof of concept and to gain mechanistic insight into the observed protein aggregation, we simplified the system, replacing lysine with propyl amine. We confirmed previous reports showing that a 1:1 ratio of 1 to free amine in buffer garners a single substituted product, 2 (Figure 1a), at a range of pH values. Notably, with an excess of free amine in buffer, a double amine conjugation (3) forms. This led us to the conclusion that the aggregation we were observing could be due to the crosslinking of two lysine residues to the same electrophile, resulting in oligomers. At first pass, these data appeared at odds with previously published worked. The nitrogen-bound proton on MaMa was previously reported to have a pK_a of 8.9 and once deprotonated, the compound should no longer be electrophilic.⁵ Upon investigation, other groups have observed this second addition in ethanolic solutions and as a minor side-product under prolonged forcing reaction conditions.^5,8 Consistent with our observations regarding the irreversibility on proteins, we have found that this second addition negates the reversibility of the amine conjugation. The double amine adduct, 3, is no longer responsive to subsequent nucleophiles, rendering this an irreversible reaction, and leading to the development of a new class of lysine probes.

Figure 1.

a) Schematic conversion of the single amine adduct (2) to the double amine adduct (3) used to measure the kinetics of the second addition. b) Concentration of 2 relative to the conversion of 2 to 3 at various pH (7–10) over the course of 60 minutes.

We observed the addition of propyl amine to MaMa 2 at constant pH values, and studied their rates to optimize utility for the conjugation reaction. We monitored the addition of propyl amine to 2 via UV-Vis, following the disappearance of a unique starting material peak. Data shows a pH dependent conversion of 2 to 3, being fastest at pH 10, and very slow at pH 7 and pH 8 (Figure 1b). At pH 9 and 10, this addition appeared to follow second order kinetics. We hypothesized that in order for the reaction to occur, both the nucleophilic amine and electrophilic MaMa need to exist in neutral forms. That said, the observed rate is mitigated by the low concentrations of relative reactive species at given pH values. When we used a Henderson-Hasselbalch analysis (utilizing the reported pK_a of the N-H bond in MaMa = 8.9 and propyl ammonium = 10.5, Figure 2a and b, respectively), to calculate relative percent of reactive nucleophile and electrophile as a function of pH, a clear picture emerged of the reactivity profile for these probes (Figure 2c).^5,9 When these percentiles are treated as concentrations and multiplied, a rate profile emerged showing the optimal rate to be found at around pH 9.8, matching our qualitative observations nicely. We measured the rate at pH 10 under pseudo-first order conditions and found the second-order kinetic rate to be 0.14 +/− 0.02 M⁻¹s⁻¹ (Supplemental Figures S2–S5).

Figure 2.

a) Equilibrium of the single amine adduct (2) existing in the active neutral form and the unreactive anionic form upon deprotonation. b) Equilibrium of propyl amine in its unreactive protonated form and its nucleophilic neutral form. c) Plot of relative reactive species of the neutral 2 (blue) and neutral propyl amine (red) as a function of pH, overlaid with an optimal reactivity curve.

We exploited this double addition reactivity, rendering it advantageous in the highly modular synthesis of new bioconjugation probes. This led to the development of a small series of MaMa probes, including a PEG azide (4a), and a nitrobenzoxadiazole (NBD) fluorophore derivative (5a, Figure 3a). Synthesis of these compounds resulted from the corresponding free amine to 1 in pH 7 buffer at a 1:1 ratio. Most of the reactions proceeded in nearly quantitative yield, providing the pure products after simple acetone extraction. Control compounds were synthesized by reacting the respective single adducts with propyl amine at pH 10 (4b and 5b, Figure 3a). The inability of the double adducts to label proteins allowed for easy recognition of nonspecific binding compared to covalent modification in subsequent protein labeling experiments.

Figure 3.

a) General synthesis of MaMa functional analogs (azide and NBD) and their respective controls. b) SDS-gel of BSA treated with 4a and 4b at pH 7–10 after 2 hours and being treated with a Cy3 cyclooctyne (DBCO). c) SDS-gel of BSA treated with 5a and 5b at pH 7–10 after 2 hours.

Utilizing bovine serum albumin (BSA) as a model protein system, we tested the utility of these compounds for protein labeling at a range of pH 7–10. We analyzed covalent modification of BSA via SDS-PAGE electrophoresis and the labeling results were consistent with our small molecule kinetic data and our understanding of the reactivity. Control samples showed minimal labeling of BSA across the range of pH tested, confirming their inability to covalently modify proteins. Additionally, reactive compounds showed negligible labeling at pH 7, but increased labeling from pH 8 up to pH 10 (Figure 3 a,b) after 2 hours. Additionally, longer time point studies (18/22 hour) showed more labeling at the lower pH, but still minimal background. The presence of the azide was confirmed using a cyclooctyne fluorophore (Cy3-DBCO) (Supplemental Figures S6–S11).¹⁰

One remaining characteristic we wanted to better understand was to examine the long-term stability of these MaMa probes at high pH. UV/vis data had suggested that an unspecified change occurred on compounds over prolonged exposure to high pH in the absence of amines, but an NMR experiment helped to solidify our understanding. Over 42 hours, 2 slowly disappeared and in its place a set of peaks with similar relative integration grew in (Supplemental Figure S12). Interestingly, the molecule did not appear to be falling apart, and we hypothesized that it might be existing as the anionic hydroxylate, 6a (Scheme 2). A recent mechanistic study on a similar compound corroborates this hypothesis.¹¹ Furthermore, when the solution was neutralized, 2 was immediately restored. At pH 7, 6 is not formed, even after 24 hours, and the molecule appears to be stable (Supplemental Figure S13). Additionally, 3 shows similar stability, with no change observed at pH 7, but the appearance of a hydroxylate (6b) forming over time at pH 12 (Supplemental Figures S14 and S15). Importantly, the rate of the reaction between MaMa and lysine residues is much faster than that of the hydroxylate formation. This protective feature of the probe bolsters its utility for bioconjugations, as it can exist in aqueous stock solutions for prolonged times prior to exposure to proteins or amines without fear of hydrolytic degradation. Additionally, 3 was subjected to acidic conditions (pH 2) for 24 hours, but no changes to the structure were noted (Supplemental Figure S16). Lastly, we aimed to test the stability of the double amine adduct to various nucleophiles, including DTT and cysteine. To measure the stability of 3 we treated the double-propyl scaffold with 1 equivalent of DTT or cysteine at both pH 7 and 10. Subsequent NMR showed that 3 was stable after 24 hours of treatment at either pH (Supplemental Figures S17 and S18). These results are consistent with our original observations of irreversibility on proteins. This also follows with our understanding of why MaMa is selective of amines. We hypothesize that other adducts can form, but they are reversible. We have not ruled out the possibility of initial adducts with the free cysteine on the surface of BSA, followed by a hand-off to a nearby lysine residue. To test this possibility, we performed a competition study between the addition of N-acetyl cysteine and propyl amine to 2 (Supplemental Figure S19). We observed an initial mixture of products that resolved to 3 over time. Once an amine is added, those adducts are rendered irreversible under the biologically relevant conditions tested.

Scheme 2.

Tetrahedral hydroxylate intermediate hypothesized to form at high pH over extended periods of time. Neutralization causes quick reversion back to original 2 or 3, showing reversibility of hydrolytic product.

Based on first-principle assumptions of pH and pK_a we expect reactivity of future MaMa analogs to be predictable, and in turn they can be customized to the chemical, or biochemical environments in which they are needed.¹² Moreover, the highly modular nature of the first amine addition lends this system to rapid probe development with a range of commercially available amine-containing fluorophores, natural products, or even radioisotopes. While not meant to supplant NHS-ester chemistry, or other lysine conjugates, these compounds are new additions to the bioconjugation tool chest, and enhance the scope of what is possible with lysine modification.