Site-specific detection and characterization of ubiquitin carbamylation

Abstract

The physiological consequences of varying in vivo CO₂ levels point to a general mechanism for CO₂ to influence cellular homeostasis beyond regulating pH. Aside from a few instances where CO₂ has been observed to cause post-translational protein modification, by forming long-lived carbamates, little is known about how transitory and ubiquitous carbamylation events could induce a physiological response. Ubiquitin is a versatile protein involved in a multitude of cellular signaling pathways as polymeric chains of various lengths formed through one of the seven lysines or N-terminal amine. Unique polyubiquitin compositions present recognition signals for specific ubiquitin-receptors which enables this one protein to be involved in many different cellular processes. Advances in proteomic methods have allowed capture and identification of protein carbamates in vivo, and Ub was found carbamylated at lysines K48 and K33. This was shown to negatively regulate ubiquitin-mediated signaling by inhibiting polyubiquitin chain formation. Here we expand upon these observations by characterizing the carbamylation susceptibility for all Ub amines simultaneously. Using NMR methods which directly probe ¹⁵N resonances we determined carbamylation rates under various environmental conditions and related them to the intrinsic pK_as. Our results show that the relatively low pK_as for half of the Ub amines are correlated with enhanced susceptibility to carbamylation under physiological conditions. Two of these carbamylated amines, not observed by chemical capture, appear to be physiologically relevant post-translational modifications. These findings point to a mechanism for varying levels of CO₂ due to intracellular localization, cellular stresses, and metabolism to affect certain polyubiquitin-mediated signaling pathways.

Graphical Abstract

Introduction

The post-translational attachment of ubiquitin (Ub) through a C-terminal isopeptide bond to a target protein at a designated amine can lead to a variety of signaled outcomes that include marking the protein for proteasomal degradation or initiating DNA repair and maintenance activities. The discrimination between cellular outcomes is achieved by sculpting Ub chains through polymerization at one or more of the seven lysines (K6, K11, K27, K29, K33, K48, K63) or N-terminal amine (M1) of Ub into chains of various composition, branching, and arrangement. The resulting polyubiquitin (polyUb) chains enable unique positions and orientations of the individual Ub units which form binding surfaces that facilitate specific interactions with downstream proteins^1,2 or DNA³. The assembly and attachment of polyUb chains is performed principally by the conjoined activity of three classes of enzymes: the Ub activating enzyme (E1), Ub conjugating enzymes (E2), and Ub ligases (E3)⁴. PolyUb chains can be further shaped by deubiquitinases into their proper frameworks ⁵. Recent studies have begun to elucidate how small-molecule post-translational modifications (PTMs) provide an additional level of modulation to Ub-mediated signaling by modifying Ub itself or the activity of enzymes ^6–8.

Enzymatic small-molecule PTMs by acetylation or phosphorylation of both Ub and Ub-associated enzymes have been observed in vivo^9–11. These PTMs alter Ub-Ub or Ub-enzyme interactions by modifying electronic and steric properties of interaction sites or directly blocking Ub chain elongation. Ub acetylation was observed in vivo at lysines K6, K48, and to a lesser extent K63, and was correlated to decreased polyUb populations¹². These findings were corroborated by an in vitro non-enzymatic acetylation assay with acetyl-CoA which linked the relatively low sidechain pK_as of K6 and K48 to acetylation propensity¹³. The relationship between amine pK_as and reactivity to non-enzymatic acetylation suggests a potential mechanism for in cellulo environmental conditions to passively modulate Ub-mediated signaling.

Similarly to acetylation, protein carbamylation by CO₂ has been observed to affect the function of a few select proteins by forming persistent carbamates at amines in favored CO₂ binding sites. Both the processes of CO₂ and O₂ transport by hemoglobin¹⁴ and carbon fixation by RuBisCo¹⁵ are known to be dependent on carbamylation by CO₂, however the widespread physiological effects of varying levels of pCO₂ indicates a role for transient carbamylation to influence homeostasis^16–18. General interactions of proteins with CO₂ are often difficult to characterize due to the rapid exchange between the carbamates and free CO₂, which precludes studies by conventional biophysical techniques. Recently a carbamylation assay was developed to capture bound CO₂ through ethylation by the triethyloxonium ion, resulting in stable esters which could be assayed by MS/MS proteomics and allowed for the identification of novel CO₂ binding proteins¹⁹. Using this assay it was reported that Ub is preferentially carbamylated at lysines K33 and K48 in vivo, however we were also able to detect carbamates on Ub at three additional sites: M1, K6, and K63 in vitro by 1D ¹³C NMR spectroscopy²⁰. The divergent reactivities of Ub K6 and K33 to in vivo acetylation and carbamylation prompted further investigation into the environmental factors which favor Ub carbamylation. Here we report the use of NMR experiments which directly probe reactions occurring at lysine sidechains and the N-terminal amine to determine the kinetics of Ub carbamylation, and demonstrate how these reactions are affected by temperature, pH, and intrinsic amine pK_a.

Materials and Methods

Protein expression and purification

Human Ub (UniProtKB P0CG48) was expressed and purified from Escherichia coli as reported previously²¹. U-¹⁵N and U-¹³C/¹⁵N Ub were obtained from M9 minimal media supplemented with ¹⁵N ammonium chloride and ¹²C or ¹³C glucose as the nitrogen and carbon sources. Ub E1 (UBA1) (UniProtKB P22314) and E2 (UBE2K, aka E2-25K) (UniProtKB P61086) enzymes that catalyze Ub polymerization via K48 were prepared and used as reported elsewhere ²².

1D ¹⁵N NMR experiments

1D ¹⁵N spectra for Ub at various pH conditions were acquired on a 500 MHz Varian Inova spectrometer with a room temperature Broad Band Observe (BBO) probe at Johns Hopkins University. All other NMR experiments were performed at the University of Maryland. The reaction of Ub with isocyanic acid was monitored using 1D ¹⁵N spectra recorded on a Bruker Avance-III 600 MHz spectrometer equipped with a room temperature BBO probe. Ub was exchanged into a buffer of 10 mM each of 2-(Cyclohexylamino)ethane-1-sulfonic acid (CHES) and 3-(Cyclohexylamino)propane-1-sulfonic acid (CAPS) at pH 9.2 with 10% D₂O. The reaction was started by the addition of a 500 mM NaNCO stock solution to final concentrations of 3 mM Ub and 72 mM NaNCO. Experiments were acquired with 1024 scans and a 220 ppm spectral width centered at 30 ppm, 150 ms acquisition time, continuous wave proton decoupling for NOE enhancement, and a 2 s recycle delay. This method took advantage of strong negative heteronuclear NOEs observed for lysine side chain amines in ubiquitin. ²³ The spectral series was processed and analyzed using Mestrelab MNova within an NMRBox ²⁴ environment.

Ubiquitin amine pK_a measurements

¹³C/¹⁵N Ub was exchanged into a buffer containing 5 mM each of CHES and CAPS at a pH of 10.5 and separated into two 500 μL 1 mM aliquots that were run as two divergent samples for lower or higher pH measurements. After each experiment these samples were consecutively exchanged into the same buffer with either increasing or decreasing pH by the addition of 0.1 M HCl and 0.1 M KOH in approximately 0.3 (H2CN) or 0.4 (HSQC) pH unit steps. D₂O was added to each sample to a final concentration of 5% (v/v). The pK_a values were obtained by fitting the observed changes in ¹⁵N, ¹³C, or ¹H chemical shifts to a Henderson-Hasselbalch equation modified with or without the Hill coefficient; these analyses are discussed further in the Supporting Information (SI). The errors in reported pK_a values for each residue were assessed by fitting 2000 Monte Carlo generated synthetic data sets in which Gaussian noise was added to the chemical shifts according to the resolution in the respective H2CN or ¹H-¹³C HSQC spectra. Both H2CN and ¹H-¹³C HSQC spectra were acquired on a Bruker Avance-III 600 MHz spectrometer equipped with a TCI cryoprobe at 23°C. These spectra were processed with NMRPipe ²⁵ and analyzed using CCPNMR V2 ²⁶. Signals from the H2CN experiments were assigned by a (H2C)N(CC)H-TOCSY experiment²⁷ .

Carbamylation measurements using H2CN spectral series

600 μM ¹³C/¹⁵N Ub was prepared in buffers containing either 20 mM potassium phosphate (pH 7.5) or 10 mM each of CHES and CAPS containing 10% D₂O, 48 mM NaNCO, and at 23°C unless otherwise specified. ¹H/¹⁵N H2CN spectra were acquired sequentially over 12.5 hrs with a 75 ms acquisition time and 2 s recycle delay. The spectra were recorded and processed as described in the previous section unless otherwise specified. The reaction rates were obtained by fitting the percentage decrease in peak volumes to a first order decay. Peak volumes for the magnetically inequivalent epsilon protons of K33 and K29 were combined before analysis.

Determination of the activation enthalpies and entropies for carbamylation reaction

The temperature-dependent rates, k, of carbamylation for Ub lysines K6, K33, and K48 were fit (using linear regression) to the following Eyring equation to determine the enthalpy (ΔH^‡) and entropy (ΔS^‡) of activation:

$$\ln \frac{k}{T}=\frac{-\text{Δ}{H}^{‡}}{RT}+\text{ln}\frac{\kappa k_B}{h}+\frac{\text{Δ}{S}^{‡}}{R},$$ (1)

where R is the gas constant, h is the Planck constant, k_B is the Boltzmann constant; the transmission coefficient κ was set to 1. Due to the small temperature coefficient for the pK_a of phosphate (ΔpK_a /ΔT of −0.0028) ²⁸, the temperature change had negligible effect on the buffer pH, as verified by direct pH measurements for the cyanate-containing phosphate buffer.

Results

As an alternative to transient CO₂ based carbamylation, we chose to study the irreversible reactions of Ub amines with isocyanic acid to generate ɑ-carbamyl and homocitrullene residues (Figure 1a). In the past these carbamylation reactions with isocyanic acid were used to identify the site and populations of hemoglobin carbamylation and were found to recapitulate the known binding mode for CO₂ ^29,30. NMR spectroscopy is uniquely suited to monitor site-specific chemical reactions or ionization states by following changes in signal intensities or chemical shifts of the reporter nuclei. Multidimensional experiments are commonly employed to circumvent issues of low ¹³C / ¹⁵N sensitivities and limited resolution in protein NMR, however the detection of amino or guanidino resonances through one-bond ¹H-¹⁵N couplings is marred by rapid proton exchange with water under conventional conditions. This issue has been classically circumvented by resorting to 1D ¹⁵N spectroscopy, which often suffers from poor resolution and sensitivity, or by employing experiments which use a neighboring ¹H-¹³C pair as a reporter that tend to exhibit limited chemical shift dispersion or are sensitive to secondary effects. As an alternative to these approaches the triple-resonance ‘H2CN’ experiment provides direct observation of ¹⁵N resonances by relaying magnetization through the adjacent methylene group and bypassing the water exchangeable proton³¹. H2CN experiments have been applied to determine lysine and arginine sidechain pK_as ³² but are also suited to study chemical reactions and binding interactions of amines ^33,34.

Figure 1. NMR studies of ubiquitin amines using direct ¹⁵N-detection.

(a) Reaction scheme for amine carbamylation by isocyanic acid. (b) 1D ¹⁵N spectra of ubiquitin amines at several time points after the addition of sodium cyanate at pH 9.2. A slight upfield shift was observed for all of the amines and is reflective of an increase in pH, presumed to be due to the degradation of cyanate. The chemical shift for K63 presented as two signals, denoted as K63 and K63*, which coalesced over time into the downfield chemical shift. (c) Kinetics of carbamylation of ubiquitin amines at pH 9.2 were monitored as the ablation of the respective ¹⁵N NMR signals. (d) Carbamylation rates relative to the reaction rate for K6 are plotted against amine pK_as.

1D ¹⁵N NMR spectroscopy of ubiquitin and the reactions with isocyanic acid

Despite the limitations of sensitivity and resolution, ¹⁵N-detected NMR spectroscopy can be used to directly observe the ionization state, dynamics, and reactions occurring with nitrogen groups undergoing rapid proton exchange with water. The limited chemical shift dispersion of solvent-exposed amine resonances often limits the ability to determine site-specific properties of these groups within proteins. However, 1D ¹⁵N direct detection experiments have been used successfully in special circumstances to determine the amine pKas in a small peptide ³⁵ or to study atypical lysines engineered into the hydrophobic core of staphylococcal nuclease ³⁶. In the case of Ub the seven lysine sidechains and the N-terminal amine are positioned in unique environments owing to electrostatic, steric, and hydrogen bonding interactions. These environmental factors influence unique sidechain dynamics for the Ub lysines ³² and were expected to confer a wide range of sidechain pK_as and spread out the ¹⁵N resonances at elevated pH (Figure S1). At pH values below 9 and above 10 many of Ub lysine sidechain ¹⁵N resonances can be distinctly observed by 1D ¹⁵N NMR experiments but not all can be resolved into isolated peaks. At the intermediate of these pH values, around 9.3, all the lysines and N-terminal amine have unique chemical shifts that could be assigned with ¹⁵N_ζ chemical shifts recorded from H2CN experiments (Figure S2).

The carbamylation reactions of Ub were ascertained by ¹⁵N direct-detection spectroscopy at pH 9.2 by adding 24 molar equivalents of sodium cyanate (3x per amine) to 3 mM U-¹⁵N Ub, and the reaction was followed for two days by monitoring the ablation of the amine signal intensities (Figure 1b). Three distinct reactivity groups were gleaned for the Ub amines: the N-terminal amine resonance disappeared within 10 hours, K6, K33, and K48 signals decayed to ~20% after two days, and the remaining lysines K11, K27, K29, and K63 demonstrated limited reactivity, with K27 and K29 being markedly unreactive (Figure 1c,d). Two resonances were initially observed for K63 which coalesced at a rate proportional to the reaction occurring at the N-terminal amine, likely due to the spatial proximity of these two amines. The susceptibility of Ub amines to carbamylation was expected to be correlated to the amount of unprotonated, reactive species present at a given pH and determined by the sidechain pK_a, and therefore we looked to comprehensively characterize the pK_as for all Ub amines by H2CN and ¹H-¹³C HSQC experiments.

Ubiquitin amine pK_as

We initially determined Ub lysine pK_as by titrating Ub through the pH range of 8.6 to 12.3 over 15 points and following the ¹⁵N_ζ-¹H_ε chemical shifts recorded by H2CN experiments (Figure 2a,b). The pH dependent chemical shifts for the six observable lysines were fit to a Henderson-Hasselbalch equation with (Figure 2c) or without (Figure S5) the inclusion of the Hill coefficient to determine the pK_as. Although a signal for the epsilon amine of K27 could be observed at physiological pH, exchange broadening at elevated pH prevented its reliable detection by this method. In a separate ¹H-¹³C HSQC-based titration over the same pH range sufficiently resolved crosspeaks were monitored to independently assess pH-dependent effects on Ub sidechain resonances. ¹H_ε-¹³C_ε crosspeaks could be resolved for K6, K11, K33, and K27, and the N-terminal ¹H_α-¹³C_α crosspeak could be followed for M1 (Figure S3). These resonances were similarly fit to the Henderson-Hasselbalch models to determine the amine pK_as (Table 1).

Figure 2. H2CN NMR experiments were used to determine the ubiquitin lysine pK_as.

(a) The H2CN spectrum recorded for ubiquitin at pH 8.6. K29 and K33 are observed with magnetically inequivalent protons at low pH. (b) Overlay of H2CN spectra of ubiquitin measured at three pH values. The chemical shift trajectories for K48 and K63 are highlighted. (c) The ¹⁵N_ζ chemical shift recorded at various pH values (dots) was fit to a Henderson-Hasselbalch equation modified with the Hill coefficient (solid lines) to determine the sidechain pK_a. (see also SI)

Table 1.

pKa values for ubiquitin amines derived from pH titration series recorded by ¹H-¹⁵N H2CN or ¹H-¹³C HSQC experiments, using the indicated nuclei as reporters

Amine	15Nζ	1Hε (H2CN)	13Cε	1Hε (HSQC)
M1	--	--	9.19 ± 0.05	9.21 ± 0.01
K6	10.47 ± 0.02	10.57 ± 0.03	10.5 ± 0.1	10.55 ± 0.04
K48	10.53 ± 0.02	10.72 ± 0.06	--	--
K33	10.88 ± 0.03	10.99 ± 0.03	10.8 ± 0.1	10.73 ± 0.06
K63	11.03 ± 0.02	11.06 ± 0.03	11.0 ± 0.1	11.5 ± 0.2a
K27	--	--	11.2 ± 0.1	11.5 ± 0.1b
K11	11.47 ± 0.03	11.46 ± 0.07	--	--
K29	11.68 ± 0.03	11.85 ± 0.03	--	--

^aPeak overlap in the proton dimension of the HSQC experiments for K63 decreased the accuracy in calculating this pK_a.

^bThe proton resonances for K27 are degenerate at lower pH values but begin to split into unique resonances as the pH is increased. This is combined with exchange broadening at elevated pH which limited an accurate proton chemical shift determination and decreased the accuracy in this calculated pK_a.

In principle, both ¹H and ¹³C / ¹⁵N pH dependent chemical shift trajectories can be fit to determine sidechain pK_as. Proton chemical shifts tend to exhibit enhanced perturbations due to broad structural rearrangements and non-localized ionization events ³⁷, and in this work we consider these as secondary reporters. For both ¹H-¹⁵N H2CN and ¹H-¹³C HSQC titrations the pK_as calculated from proton chemical shifts were in good agreement with the primary reporting nuclei, aside from small differences in the pK_a values derived from HSQC experiments due to peak overlap for K63 and pH dependent splitting of the proton resonances for K27(Table 1). Overall, we were able to obtain the pK_as for all the Ub amines using ¹⁵N_ζ and/or ¹³C_ε/ɑ reporters and standard U-¹⁵N / ¹³C labeling procedures. Where applicable we consider the ¹⁵N resonances to be the most accurate reporters due to direct sampling of the ionizable nuclei. However, for three amines with both ¹⁵N_ζ and ¹³C_ε reporters, K6, K33, and K63, pK_as derived from both H2CN and ¹H-¹³C HSQC series were in good agreement. Our complete characterization of Ub amine pK_as concurs with previous reports which sample either lysine side chains¹³ or the N-terminal amine³⁸ individually. The ordering of the Ub amine pK_a values matches the reactivity profiles we observed by ¹⁵N 1D spectroscopy measurements. M1, with a pK_a near the experimental pH, reacted the quickest, followed by K6, K33, and K48 as the group of lysines with sidechain pK_a values below 11. The reaction rates for the other amines are significantly depressed in comparison but remain distributed with respect to pK_a values, aside from the anomalously low reactivity for the substantially buried sidechain of K27.

Ubiquitin carbamylation at variable pH and temperature

The H2CN experiment provided sufficient Ub amine chemical shift dispersions at all pH values to allow characterization of residue-specific carbamylation kinetics at physiological pH and to assess how these reaction rates are affected by environmental factors. The reaction with cyanate was initially probed at pH 7.5, 27.0°C, and 15 molar equivalents of sodium cyanate per amine, and the pseudo first-order reaction kinetics were observed over a 12-hour period (Figure 3a). In contrast to our 1D ¹⁵N data, the amine reactivities under these conditions did not fall into distinct reactivity groups but generally reflected the spread of sidechain pK_a. K33 exhibits an exceptionally quick carbamylation rate relative to its pK_a value, only reacting slower than the N-terminal amine which reacted too quickly for any signal to be observed. K63 (pK_a 11.03) was also carbamylated anomalously slower than K11 (pK_a 11.47) and K29 (pK_a 11.68) with respect to pK_a values. These results were reproduced at a lower sodium cyanate concentration of 10 molar equivalents, however the quick reaction rate for the N-terminal amine could be measured and K63 was unreactive to carbamylation (Figure 3b). We expected a linear relationship between isocyanic acid concentration and the reaction rates, but some of the lysines exhibited a greater sensitivity. For a 1.5x increase in sodium cyanate concentration, K33 reacted 2.4x faster, K29 and K6 2.0x faster, K48 and K11 showed a rate increase consistent with the increase in isocyanic acid concentration of 1.5 fold, and appreciable carbamylation was detected at K63.

Figure 3. H2CN experiments were used to determine the carbamylation rates for ubiquitin amines at physiological pH.

(a) The kinetics for ubiquitin carbamylation by 120 equivalents of sodium cyanate were measured as a decrease in peak volume for crosspeaks recorded with H2CN spectral series. The peak intensities (dots) recorded over a 12 h time course were fit to an exponential decay model (solid lines). (b) Overlay of carbamylation rates recorded with 80 and 120 molar equivalents of sodium cyanate. Comparison of carbamylation kinetics recorded at 27°C, 32°C, and 37°C degrees with (c) 80 or (d) 120 molar equivalents of sodium cyanate. (e) Overlay of carbamylation rates recorded at 27°C and 37°C.

We attempted to assess the reaction enthalpies for the observable lysines by running temperature series. The rate of carbamylation was measured at three temperatures, 27.0°C, 32.0°C, and 37.0°C, for both 10 (Figure 3c) and 15 (Figure 3d) molar amine-equivalents of sodium cyanate. Under both conditions the reaction rates for the low pK_a amines of K6, K33, and K48 increased in a stepwise manner with increasing temperature. This is contrasted with the reaction rates for the other observable lysines which either increase at 32.0°C and then level off (K11) or increase at 32.0°C followed by a decline in reaction rate at higher temperature (K29, K63). From 27.0°C to 37.0°C the rates increased roughly 4-fold for K6, K33, and K48 in contrast to modest overall increases for the others (Figure 3e). The temperature dependent carbamylation rates for these three amines, at both sodium cyanate concentrations, were fit to the Eyring equation to determine the enthalpies of activation (Table 2). The average enthalpies of 23 kcal mol⁻¹ (10x cyanate) and 17 kcal mol⁻¹ (15x cyanate) are similar to the reported activation energies for Ub lysine acetylation of 22.6 kcal mol^{−1 13}.

Table 2.

Thermodynamic characteristics of carbamylation of ubiquitin amines with 80x and 120x NCO derived from the temperature series.^a

	10x sodium cyanate		15x sodium cyanate
Amine	ΔH‡ kcal mol−1	ΔS‡ cal mol−1 K−1	ΔH‡ kcal mol−1	ΔS‡ cal mol−1 K−1
K6	24 ± 10	−34 ± 33	17 ± 5	−56 ± 15
K48	22 ± 5	−38 ± 18	14 ± 2	−63 ± 7
K33	23 ± 6	−36 ± 20	19 ± 7	−50 ± 22

^aThe enthalpies (ΔH^‡) and entropies (ΔS^‡) of activation were determined from linear regression fit of the temperature-dependent carbamylation reaction rates using Equation 1. The errors represent the standard errors of the linear regression.

The pH dependence of the reaction rates was characterized at five additional points above pH 7.5 at 27.0°C. Amine carbamylation by isocyanic acid is known to proceed with the neutral forms of both components and is reported to be a second-order reaction. As the pK_a of isocyanic acid is ~3.5 under standard conditions, the reaction rates are theoretically constant in the pH range above 5.0 and up to ~1.5 units below the amine pK_as. Within a pH range of −1 to +1 units around the amine pK_a the population of the unprotonated, reactive form increases tenfold while the proportion of reactive isocyanic acid to non-reactive cyanate falls by a factor of 100, thus a significant reduction in the reaction rate is expected.^39,40 Mostly consistent with these expectations, for all of the Ub amines the reaction rates increased at each pH step approaching the respective pK_as, leveled off near the pK_as, and significantly decreased at higher pH (Figure 4 a,b). The reaction occurring at the N-terminal amine could be followed at neutral and high pH but was too fast to be observed at intermediate pH values and likely adheres to a similar trend. At pH values below 10.3 the K33 sidechain (pK_a 10.88) is particularly susceptible to carbamylation and reacts faster than the lower pK_a sidechains of K6 (pK_a 10.47) and K48 (pK_a 10.53). As the pH approaches these sidechain pK_as, this disparity vanishes, and the reaction rates become distributed reflective of their pK_as. This is accompanied by an anomalously high reactivity rate observed for K29 at this pH, and possibly indicates a pH-induced rearrangement of the contacts between the β2 strand and the α-helix (Figure S1d).

Figure 4. The effect of pH on the carbamylation rates of Ub amines.

The carbamylation rates at variable pH are shown for both (a) low pK_a and (b) high pK_a ubiquitin amines with 80 molar equivalents of sodium cyanate at 27°C. The anomalously high reactivity of K29 at pH 10.3 is presumed to be a consequence of a pH-induced structural rearrangement.

Ub is particularly resilient to structural destabilization by mutation or PTMs, however complete shielding the sidechain amines by hydrophobic Alloc or Boc protecting groups has been shown to make Ub insoluble in water and decrease the overall structural stability⁴¹. During some carbamylation studies we did observe visible precipitation during extended reaction times (36+ hours) when the number of carbamylated amines per Ub exceeded 5-6 on average, as determined by electrospray ionization mass spectrometry. Within our 12.5-hour assay, where the average degree of carbamylation was similarly measured to be approximately 2 (Figure S4), the concentration of soluble Ub was confirmed by 1D ¹H series where the total amide (6.55 to 9.70 ppm) and hydrophobic methyl (−0.10 to −0.60 ppm) signal integration was monitored after the addition of sodium cyanate. Under conditions of elevated temperature or at an intermediate pH, where maximal carbamylation rates were observed, no decrease in bulk signal intensity was evident. To confirm the viability of Ub after carbamylation, a conjugation assay with the E1 and E2 enzymes that assemble K48-linked Ub chains ²² was performed after an overnight incubation of Ub with 15 molar equivalents of isocyanic acid at 37°C. Aside from the decreased population of higher order chains due to the blocked conjugation for Ub K48_carb, the unmodified Ub was effectively polymerized into chains (Figure S6).

Discussion

Lysine sidechain and N-terminal amines can interact with electrophilic compounds to form covalent PTMs. Some of these modifications are known to require catalysis by enzymes, as evidenced in proteinaceous and small-molecule PTMs such as ubiquitylation ¹, sumoylation ⁴², biotinylation ⁴³, and methylation ⁴⁴. Acetyl-CoA can both act as a cofactor for acetyltransferases in enzymatic acetylation ⁴⁵ or directly react with suitable amines to spontaneously generate the acyl modification ^13,46. Other PTMs of amines occur independently of any enzymatic activity as in carbamylation by isocyanic acid ^47,48 or CO₂ ⁴⁹, oxidation by reactive oxygen species (ROS) ⁵⁰, or glycation by sugars ^51–53. Endeavors in proteomic analyses have begun to elucidate both the breadth of proteins whose activity and function depend on amine PTMs and how structural factors facilitate these types of modifications ^54,55.

Non-enzymatic PTMs have demonstrated a specificity for occurring at certain amines, and this is correlated with the amine pK_a which is influenced by the steric hinderance imparted by neighboring functional groups and the proximity of charged functional groups⁵⁶. The N-terminal amine has often been found to exhibit a significantly lower pK_a, and higher nucleophilicity at physiological pH, than solvent exposed lysine sidechain amines, and this property has been exploited to engineer single-site protein modifications with electrophilic ligands ^57,58. By this same principle proteins which contain “druggable” lysines, those with amines anomalously reactive with specific classes of ligands, have been focused on as targets for small molecule drugs to inhibit deleterious functions and treat diseases^59–61. Similarly, in nature only a few proteins have been observed to bind CO₂ as a functional modifier by the way of stable carbamates on amines, despite the significant physiological concentration range of CO₂⁶². This is confounded by the widespread physiological effects observed by modulating the concentration of CO₂ by in vivo and in cellulo assays and points to an obfuscated importance for direct protein interactions with CO₂^63–66.

In contrast to acetyl and methyl small molecule PTMs which are often catalyzed by acetylases and methyltransferases, respectively, amine carbamylation by CO₂ occurs independently of enzymatic activity and is readily reversible. Long-lived and physiologically relevant protein carbamates are thought to necessitate privileged binding sites with proximal cationic residues to stabilize the charged product^67,68, such as the functional CO₂ binding sites of hemoglobin⁶⁹ and RuBisCO⁷⁰. Temporary carbamylation events conversely occur on all surface amines to varying degrees and lifetimes which are largely dependent on the transition state energies and population of the unprotonated, reactive species. For most proteins these transient carbamylation events are benign; however, for Ub and some ubiquitin-like proteins, which are involved in cell signaling by forming peptide or isopeptide bonds through specific amines, a significant degree of carbamylation could modify their signaling properties by directly blocking chain extension.

In this study we have demonstrated the utility of NMR to directly characterize PTMs occurring at lysine sidechains and the N-terminal amine by the direct observation of ¹⁵N resonances. For the carbamylation of Ub, where the surface lysines are situated in unique chemical environments, we have shown that 1D ¹⁵N spectroscopy could be utilized to obtain site-specific reactivity profiles for the N-terminal amine and all seven lysine residues concomitantly albeit within a constrained pH range and using a high concentration of uniformly ¹⁵N labeled protein. These downsides were circumvented by using the triple-resonance H2CN experiment which affords proton detection in addition to detecting the sidechain ¹⁵N chemical shift. By adding a second dimension, the amine resonances could be resolved at physiological and elevated pH, which enabled the determination of sidechain pK_as as well as reaction kinetics. Using both ¹⁵N-detected 1D and H2CN experiments we have shown that Ub amines display differential susceptibilities to carbamylation by isocyanic acid.

Our recent report that Ub is selectively carbamylated by CO₂ proposes an additional mechanism for modulating Ub chain forming reactions and the population of polyUb chains reversibly and in response to a changing cellular environment. We previously demonstrated that carbamates for Ub lysines K33 and K48 could be detected in vivo by a mass spectrometry assay but three additional sites were revealed by in vitro ¹³C NMR experiments. These five carbamylation sites have the lowest pK_as in Ub: M1, K6, K48, K33, and K63. In this work we have used isocyanic acid as a non-reversible carbamylation agent and were able to directly study the susceptibility of all the Ub amines to carbamylation, which generally reflected lower intrinsic pK_a values. The N-terminal amine and the three lysines with the lowest pK_as, K6, K33, and K48, demonstrated a greater susceptibility to carbamylation by isocyanic acid compared to the other four lysines under all conditions at physiological and intermediate pH values (7.5 to 10.3), and the efficiency for carbamylation of these amines demonstrated a marked sensitivity to temperature. K33, however, demonstrated a higher carbamylation rate than K6 and K48 despite having an appreciably higher pK_a, which suggests the presence of additional factors promoting carbamylation at this site. As K33 was not especially reactive to non-enzymatic acetylation, this points to a favorable environment for interacting with isocyanic acid and CO₂ ^29,71. Contrary to our previously reported CO₂ carbamylation assay, K63 was only modestly sensitive to irreversible carbamylation by isocyanic acid at physiological pH and temperature. These last two observations are being actively investigated.

The specificity for carbamylation at the four Ub amines with the lowest pK_as proposes a mechanism for intensity of specific polyUb signals to be modulated by environmental pH and the flux of CO₂. Evidence has been presented that hyper- or hypocapnic conditions can modulate K48 polyUb signaling in cellulo, and thereby affect the process of protein turnover by proteasomal degradation which is necessary for cellular homeostasis²⁰. By this same token, the innately low pK_as and enhanced susceptibility to carbamylation for the amines of M1, K6, and K33 would render these polyUb chains subject to regulation by passive carbamylation within different organelles, cell types, and localization within the body due to divergencies in pH and CO₂ concentrations. This effect would manifest as a modulation of the population and length of homogeneous-linkage Ub chains, where the direct consequences of these chain types are well known, in addition to regulating the generation of mixed-linkage unbranched and branched Ub chains that have enigmatic physiological responses.

ABBREVIATIONS

Ub

ubiquitin

polyUb

polyubiquitin

E1

ubiquitin activating enzyme

E2

ubiquitin conjugating enzyme

E3

ubiquitin ligase

PTM

post-translational modification

MS

mass spectrometry

RuBisCo

ribulose-1,5-bisphosphate carboxylase-oxygenase

HSQC

heteronuclear single-quantum coherence

Alloc

allyloxycarbonyl

Boc

tert-butyloxycarbonyl