Assessing Squarates as Amine-Reactive Probes

Abstract

Probes that covalently label protein targets facilitate identifying ligand-binding sites. Lysine residues are prevalent in the proteome, rendering them attractive substrates for covalent probes. However, identifying electrophiles that undergo amine-specific, regioselective reactions with binding site lysine residues is challenging. Squarates can engage in two sequential conjugate addition-elimination reactions with amines. Nitrogen donation reduces the second reaction rate, making the mono squaramide a mild electrophile. We postulated that this mild electrophilicity would demand a longer residence time near the amine, affording higher selectivity for binding site lysines. Therefore, we compared the kinetics of squarate and mono-squaramide amine substitution to alternative amine bioconjugation handles. The data revealed that N-hydroxy succinimidyl esters react four orders of magnitude faster, consistent with their labeling promiscuity. Squarate reactivity can be tuned by substitution pattern. Electron-withdrawing groups on the vinylogous ester or the amide increase reaction rates. Dithionosquarates react more rapidly than squarates, while vinylogous thioester analogs, dithiosquarates, react more slowly. We assessed squarate selectively using the UDP-sugar processing enzyme GlfT2 from Mycobacterium tuberculosis, which possesses 21 surface-exposed lysines. The reaction predominately modified one lysine proximal to a binding site to afford covalent inhibition. These findings demonstrate the selectivity of squaric esters and squaramides, a critical feature for affinity-based chemoproteomic probes.

Graphical Abstract

Covalent protein modification can be exploited to understand protein structure and function, identify new ligands, and pinpoint ligand binding sites.^1–11 The lack of effective small-molecule ligands for many human and microbial proteins motivates the search for functional groups that can turn a ligand into a covalent probe..^7,8 Attenuated electrophiles that react with rare residues like cysteine are commonly employed in probes.^5,9 Unfortunately, many target proteins lack exposed cysteines. Moreover, many cysteine residues engage in disulfide bonds, rendering them inaccessible.

The need for more options has spawned efforts to target less rare amino acids, including lysine.¹⁰ Lysines are over three times more common in proteins than cysteines.^11,12 Moreover, lysines are often located near active sites or at protein–protein interaction sites. Still, the prevalence of lysine residues raises a tension between reaction rate and selectivity.

For electrophiles to report on ligand binding sites, they must react chemoselectively and regioselectively. A proteome investigation of lysine reactivity highlighted the range of electrophiles that function as tools; however, many have constraints.¹³ Some cross-react with other nucleophilic side chains. Others require high affinity to hit their targets or fail to react selectively near the binding site.^14,15

Squaric esters, or squarates, are small electrophiles characterized by a cyclobutenedione core and two vinylogous esters.^16,17 Squarates react preferentially with amine nucleophiles over alcohols and thiols, a valuable property for protein conjugation and generating bifunctional molecules.^18–22 Moreover, squarates can react with two distinct amine nucleophiles. An initial addition-elimination sequence affords a mixed squaramide ester, which is less electrophilic than its precursor. This entity can react with a second, distinct amine to yield asymmetric bis-squaramides. This property has been used to construct enantioselective organocatalysts, generate protein-glycan conjugates, and recruit antibodies to target cells..^23–27

We postulated the attenuated electrophilicity of squaramides would render them excellent covalent probes: squarates and squaramide products are stable in complex biological mixtures. We further hypothesized that squarate reactivity could be tuned to maximize selectivity for binding site lysines.^28,29 In proteome-wide lysine reactivity studies, squarates were among the electrophile classes that liganded the greatest number of lysine residues overall, yet the specific residues targeted by each squarate derivative depended on the probe structure.^13,30 Therefore, squarates should balance reactivity and selectivity, making them effective covalent probes.

Deploying squarates as covalent probes demands a fundamental understanding of their reactivity. We therefore synthesized a panel of squarates with varying electronic properties. We benchmarked their reaction kinetics against other aminophiles commonly used for bioconjugation: a succinimidyl ester and a dichlorotriazine.³¹ The squarate core is ~100-fold less reactive than a succinimidyl ester and can be tuned through substituent modification. Changes in squarate derivative reactivity were modeled through computational analysis.

Our studies indicate that squarates are useful linchpins for affinity labeling, even with ligands of modest affinity. We generated a squarate-containing uridine diphosphate (UDP) derivative and found that it selectively modifies the mycobacterial enzyme galactofuranosyl transferase 2 (GlfT2), which possesses 21 surface lysines. Even though the UDP moiety is a low-affinity inhibitor (K_d in the mM range), the squaramide analog afforded covalent inhibition. Subsequent analysis revealed modification of a single lysine adjacent to the binding site. The squarate derivative exhibited higher specificity than the corresponding dichlorotriazine.⁴ These findings demonstrate the utility of squarates for selective labeling.

Results and Discussion

Squarate Reaction Kinetics.

We assessed the reaction rate of a model squarate to compare with that of other aminophiles used in biological studies. Specifically, we monitored benzylamine addition to dibutyl squarate 1 and quantified rate constants for each amidation step (Figure 1A). Dibutyl squarate reacted with benzylamine at room temperature with a second order rate constant, k₁, of 6.3 x 10⁻² M⁻¹s⁻¹. As expected, the second substitution was slower. When benzylamine was exposed to squaramide ester 2, the rate k₂ was 5.6 x 10⁻⁴ M⁻¹s⁻¹, two orders of magnitude slower.

Figure 1.

Amine modification reactions and second-order rate constants: (A) first and second amidation of squarate, 1, with benzylamine, (B) acylation of benzylamine with succinimidyl ester 4, and (C) addition of benzylamine to dichlorotriazine 6. Rates are also shown in Table S1.

A succinimidyl ester analog of 2 reacts with benzylamine with a rate constant k₃ (Figure 1B) that is ~20,000-fold increase compared to squaric ester-amide 2 and a ~200-fold enhancement over the more reactive squaric ester 1. We next evaluated dichlorotriazine reactivity.⁴ Substitution of dichlorotriazine 6 was faster than succinimidyl ester 4 with a rate constant, k₄, of 8.3 x 10¹ M⁻¹s⁻¹ (Figure 1C). The comparatively attenuated electrophilicity of squarates should promote more selective lysine affinity labeling.

Fine-Tuning Squarate Reactivity.

We explored how perturbation of the squarate core affects its reactivity. Squarate substitution involves two steps: addition and elimination. The former should be rate-determining and therefore have the greatest influence on squarate reactivity. To this end, we evaluated the sensitivity of squarates to electronic perturbation. Replacing the benzylamine substituent in 2 with aniline afforded derivative 8a that underwent reaction with a second-order rate constant of 4.7 x 10⁻² M⁻¹s⁻¹. Thus, the aniline derivative is 40-fold more reactive than squaramide 2.

We next assessed substrates with aniline substituents bearing electron-withdrawing or electron-donating substituents (Table S1). A Hammett plot shows a linear relation between reactivity and substitution constant, σ (Figure 2). These changes span one order of magnitude and are consistent with electron-withdrawing substituents enhancing reactivity of electrophiles. Because the variation is moderate, we conclude that appending typical protein ligands to the squarate core will have little impact on its intrinsic electrophilicity.

Figure 2.

Hammett plot from squarates 8a-p (Table S2) reacting with benzylamine in d₆-DMSO to afford derivatives 9a-p. The substitution constant, σ, is plotted with the log of the relative reactivity of a given compound to phenyl-squaramide 8a. ρ = 0.71.

Electronic perturbation of the vinylogous ester should also alter amine addition rates. We therefore subjected phenyl ester squarate to reaction with benzylamine (Scheme S1, Table S1) and compared the kinetics of substitution of the resulting phenyl ester product to those of butyl ester 2. As predicted, the phenyl ester was 20-fold more reactive than the corresponding butyl ester 2 (1.2 x 10⁻² M⁻¹s⁻¹). When the butyl amide was replaced with aniline substituent, the resulting phenyl ester amide was 190-fold more reactive (1.1 x 10⁻¹ M⁻¹s⁻¹). Thus, replacing either the amide or the ester substituent with a more electron withdrawing group enhances squarate reactivity.

Sulfur Substituted Squarates.

Given the influence of heteroatom substitution on squarate reactivity, we tested the replacement of oxygen atoms with sulfur. We initially synthesized dithio- and dithiono- analogs of dibutyl squarate 1; however, the dithiono analog degraded rapidly. We synthesized the more stable dicyclopentyl dithionosquarate, 10.³² Since compound 10 contains aliphatic leaving groups, we compared its substitution rate with dibutyl squarate 1.

When thionosquarate 10 was subjected to benzylamine treatment, only the bis-amide was observed. Thionosquarates underwent addition-elimination within milliseconds. To explicate changes in reactivity, we calculated energies of the squarate analogs at the B3LYP/6-31G* level of theory.³³ We compared the LUMO energies and electrostatic charges for the squarate 1 and its monoamidated product to the sulfur-substituted versions. (Figure S1). The calculations support our experimental findings that reactivity of product 2 is attenuated compared to squarate 1. They are consistent with our findings that electron-donating or withdrawing substitutions predict reactivity. In thionosquarates, the LUMO energies of the mono-amidation product are lower than even the unreacted non-sulfur analog (Figure S1). Thus, while the LUMO for the monosubstituted product is higher in energy than its unsubstituted counterpart, the second amidation can still occur rapidly. These calculations elucidate why bis-amidated product is obtained exclusively.

Exposing dithiosquarates like compound 11 to amines resulted in a monoadduct. Treatment with excess amine failed to yield addition, indicating that the monosubstituted product is unreactive. This result is consistent with a buildup of negative charge on C4 (Scheme S2), an analysis supported by the calculations. Previous studies indicate that heating squaric thioester-amide with excess amine leads to nucleophilic attack at the carbonyl and ring opening.^35,36 Substitution of the remaining thioether group is unfavorable (Scheme S2). The sulfur substituents biases squarate reactivity toward the mono-addition product (dithiosquarates) or the bis-addition products (thionosquarates).

Squarate-mediated covalent targeting of GlfT2.

Our kinetic data and previous findings¹³ suggest squarates are useful lysine reactive probes; however, most lysine residues on proteins have a higher pKa (10.7) than benzylamine (9.3).³⁷ Squarates attached to high affinity ligands (K_d ~10⁻⁹ M)³⁸ can label, but selectivity with even modest affinity ligands would facilitate chemoproteomics and ligand discovery. Thus, we examined labeling and inhibition of the galactofuranosyl transferase GlfT2 (Figure 4), a substrate that lacks tight-binding ligands.^39,40 GlfT2, an essential enzyme in Mycobacterium tuberculosis, catalyzes polymerization of galactofuranose residues to yield the galactan.^34,40,41 The substrate uridine diphosphate galactofuranose (UDP-Galf) has an apparent K_m of 0.4 mM.⁴² Derivatives of UDP can bind and inhibit GT-A glycosyltransferases like GlfT2 with K_d values similar to those of the substrate K_m.^43–45 The weak affinity of GlfT2 for its ligands provided a challenge for squarate selectivity. We synthesized UDP-squarate derivative 12 (Figure 3A) to determine whether it would react selectively and if covalent tethering improved inhibition by UDP. Intact mass spectrometry (MS) data demonstrated the protein was labeled by squarate 12 predominantly at a single site. (Figure S4). Pretreatment with UDP resulted in lower occupancy (Figure S5).

Figure 4.

(A) Relative activity of GlfT2 samples treated with compound 12 at various concentrations and durations. (B) GlFT2 treated with compounds 12 and UMP to obtain 80% inhibition with enzyme activity assessed before and after dialysis.

Figure 3. Compounds used to examine the binding of UDP-derivatives to the glycosyltransferase GlfT2.

(A) Native substrate donor, UDP-Galf and analogs: electrophilic squaric ester 12 or dichlorotriazine 13. Model of GlfT2 (PDB: 4FIY) with lysines covalently modified by 12 (B) highlighted in blue and 13 (C) highlighted in gray.³⁴

To compare 12 with a more reactive electrophile, we synthesized UDP derivative 13 and performed proteomics on samples labeled with 10-fold higher concentrations of 12 or 13. Under these conditions, squarate 12 remained selective (Figure 3B and Table S3), while dichlorotriazine 13 reacted with eight lysine residues. The eight sites hit by 13 were scattered across the surface of GlfT2, including six additional lysines not labeled by the squarate (Figure 3C and Table S3).³⁴

Noting that 12 covalently modified residues near the active site’s UDP binding motif, we hypothesized it would inhibit enzyme activity. We treated GlfT2 with UDP-squarate 12 for varying durations and then added the native substrate UDP-Galf and synthetic acceptor.⁴¹ GlfT2 activity diminished, indicating inhibition by 12. The extent of inhibition correlated with probe concentration and duration of labeling (Figure 4A). We hypothesized inhibition resulted from the UDP moiety of 12 binding in the active site, blocking substrate binding. We assessed activity of shorter linker length analogs of 12. Shorter lengths showed less inhibition (Figure S3). Modeling the squarate probe into the active site showed shorter linkers were unable to engage all relevant active site residues unlike longer linkers (Figure S3).

To assess whether inhibition arises from covalent modification, we compared 12 with the non-covalent ligand uridine-monophosphate (UMP).^44,46 GlfT2 treatment with 12 or UMP depressed enzyme activity by ~80%. The sample exposed to UMP recovered activity after dialysis, while that treated with 12 did not (Figure 4B). These findings provide further insight into chemoproteomic analysis of lysine reactivity.¹³ Together, the data indicate that squarate analogs selectively modify lysines and function as covalent inhibitors, highlighting their utility as affinity probes.

Conclusions

Our data indicate the reactivity of squarates complements other amine bioconjugation handles. Their reaction rates occupy the “Goldilocks” zone: not so rapid they lack selectivity nor so slow they fail to label targets. Consequently, their reactivity correlates with binding site occupancy.¹³ Their moderated reactivity indicates squarate-modified small molecule probes would be highly selective. Additionally, our data demonstrate the utility of predictions from computational modeling for optimizing squarate reactivity. We anticipate that squarate-containing compounds can be used for understanding protein function and discovering novel ligands and their targets.