Transition-state analysis of the arginine-specific human ADP-ribosyltransferase 1

Significance

Dysregulation of hsART1 activity has been shown to permit immune cell evasion in non–small cell lung cancer through elevated mono-ADP ribosylation (MARylation) at Arg125 of purinergic type 2 receptor (P2X7) in P2X7-positive T cells, resulting in NAD⁺-induced cell death of tumor-penetrating immune cells. The transition-state structure of the human arginine-specific ADP-ribosyltransferase 1 (hsART1) is critical to understanding the enzymatic chemistry of this membrane-associated, mono-ADP-ribosyltransferase selective for MARylation of L-arginine residues. Here, we use kinetic isotope effects and computational chemistry to resolve the transition-state structure. Here, we use kinetic isotope effects and computational chemistry to resolve the transition-state structure for L-arginine-specific ADP-ribosylation catalyzed by hsART1. The information informs on chemistry and provides an important step toward the design of transition-state analogs.

Abstract

The ADP-ribosylation of proteins is a versatile, reversible, posttranslational modification involved in the dynamic regulation of numerous cellular processes. Human ADP-ribosyltransferase 1 (hsART1, EC: 2.4.2.31) is a membrane-associated, GPI-anchored, mono-ADP-ribosyltransferase selective for mono-ADP ribosylation (MARylation) of L-arginine residues. Dysregulation of hsART1 activity has been shown to permit immune cell evasion in non–small cell lung cancer (NSCLC) through elevated MARylation at Arg125 of the purinergic type 2 receptor (P2X7) in P2X7-positive T cells, resulting in NAD⁺-induced cell death (NICD) of tumor-penetrating immune cells. With hsART1 emerging as an immunotherapy target in select cancers, there is a need to develop small-molecule inhibitors. The transition state (TS) for the MARylation of P2X7 peptide was determined from kinetic isotope effect (KIE) measurements of ³H-, ¹⁴C-, ¹⁸O-, and ¹⁵N-labeled NAD⁺ substrates. Quantum mechanical (QM) calculations of the reaction coordinate, mapped with experimental KIEs, identify a TS geometry consistent with a highly dissociative, asymmetric, concerted mechanism with minimal contributions from the leaving group nicotinamide (NAM) and minor contributions from the incoming L-arginine guanidinium. The absence of a normal deuterium solvent isotope effect identifies a positively charged guanidinium nucleophile, leading to a dication N-ribosyltransferase mechanism. Together with the unusual, normal ¹⁸O–O4^′ KIE, we identified unique charge accumulation across the oxocarbenium at the TS caused by an increased bond order between the C1^′–C2^′, and decreased bond order between the C4^′–O4^′ of the nicotinamide mononucleotide ribose. This is the first L-arginine-specific ADP-ribosylation TS to be characterized, a step toward the design of TS analogs.

ADP-ribosylation is a reversible posttranslational modification where ADP-ribose (ADPR) from $\beta$-nicotinamide adenine dinucleotide (NAD⁺) is installed on the target substrate (1). While traditionally studied in the context of proteins, this modification is also installed on a diverse array of biomolecules, including nucleic acids and select small-molecule metabolites (2–4). The ADP-ribosylation of proteins is involved in the coordination of cellular processes, encompassing cellular signaling, stress-related responses, signposting DNA damage, and aiding in protein recruitment for its subsequent repair (5). These processes are regulated by the complementary activities of ADP-ribosyltransferases (ARTs) that install the ADPR, and are removed by specific ADPR glycosylhydrolases that work together to dynamically modulate ADP-ribosylation mediated effects (6–8).

ADP-ribosylation is also leveraged by various bacterial toxins to dysregulate normal cellular function during infection (1, 2, 9). While bacterial ARTs were originally defined as virulence factors targeting eukaryotic hosts, they are now recognized for broader ecological roles in mediating interbacterial competition (10–12) and defense against bacteriophage infection (13, 14).

ADP-ribosylation can occur by the installation of a single ADPR onto a protein surface residue (mono-ADP-ribosylation; MARylation) or through the sequential addition of multiple ADPRs to preestablished ADPR modifications to form long and branched polymers, called (poly-ADP-ribosylation; PARylation) (15). Both MARylation and PARylation have been shown to modulate target function (16–18). While ADP-ribosylation has been observed in vitro and documented in vivo across a variety of residues, arginine-specific MARylation is catalyzed exclusively by members of the ADP-ribosyltransferase C (ARTC) family (1, 19, 20). Of the four ARTCs expressed in humans, there are two catalytically active MARylating ARTCs (hsART1 and hsART5) and two catalytically inactive homologs that lack catalytic glutamates within the conserved R-S-E motif present within arginine-specific ARTCs (hsART3 and hsART4) (21, 22).

hsART1 is a membrane-associated glycosylphosphatidylinositol (GPI)–anchored enzyme that accumulates both intracellularly and extracellularly during maturation and trafficking (18). The exact intra- and extracellular partitioning of hsART1 varies between cell types, tissues, and cancer models. The nascent protein initially localizes to the endoplasmic reticulum (ER) before being transported to the plasma membrane and exported to the cell surface, allowing it to MARylate a diverse range of targets across both compartments (16–18, 24, 25).

hsART1 catalyzes the MARylation of several cellular protein targets and a weak promiscuous NAD⁺ hydrolase activity (Fig. 1A) (18, 23, 26). Targets of arginine-specific hsART1 include the P2X7 receptor (P2X7) (27). P2X7 is a purinergic type 2 receptor family, adenosine 5^′-triphosphate (ATP)–gated cation channel involved in the activation of numerous proinflammatory pathways, as well as being essential for inflammatory responses and antitumor immunity in select T cell lineages (27, 28). hsART1 catalyzed MARylation of P2X7 at Arg125 proximal to the ATP binding pocket leads to constitutive activation (Fig. 1 B and C) (26). Overactivation of MARylated P2X7 triggers large pore formation, uncontrolled calcium influx, and phosphatidylserine efflux (26, 29). Together, the destabilization of the cell membrane and significant osmotic pressure lead to cell death through a process called NAD⁺-induced cell death (NICD; Fig. 1B) (30–32). While typically, extracellular NAD⁺ concentrations are low, within the tumor microenvironment, extracellular NAD⁺ concentrations can fluctuate and become significantly elevated due to cellular stress or death (33–35).

Fig. 1.

(A) hsART1 is an arginine-specific ARTC that catalyzes both the MARylation of cellular protein targets and, in the absence of transferase substrate, weak NAD⁺ hydrolysis (23). (B) Extracellular hsART1 expressing cancers induce NAD⁺-induced cell death (NICD), in tumor-microenvironment recruited P2X7⁺ expressing CD8$+$ T cells. MARylation leads to activation of the P2X7, leading to rapid influx of extracellular calcium (B1). Activation leads to pore-formation (B2), and phosphatidylserine accumulation to the outer leaflet (B3). (C) AlphaFold 3 model of the hsART1:NAD⁺:hsP2X7 complex, hsART1 colored in red, hsP2X7 homotrimer colored in green, and NAD⁺ is colored in white. (Left) hsART1 binds to a single hsP2X7 monomer, and R125 MARylation site is bound within the hsART1 active site (Boxed). (Center) The NAD⁺hsART1 binding pocket with surrounding contacting residues, including hsP2X7 R125 (Green). (Right) hsART1:hsP2X7 protein interface at the transferase reaction center. Details of the plDDT scoring of the reaction center are provided in SI Appendix, Fig. S1. (D) Kinetic characterization of hsART1 catalyzed MARylation of the P2X7 peptide by coupled luminescence assay. Kinetic parameters are summarized in Table 1. Data points and error bars represent mean $\pm$ SD, and solid lines represent either fitting to SI Appendix, Eq. S1. (E) Structure of NAD⁺ indicating positions in the molecule which were isotopically radiolabeled with ³H, ¹⁴C ¹⁵N, or ¹⁸O.

In non–small cell lung cancer (NSCLC), high levels of hsART1 expression were associated with reduced tumor infiltration of P2X7-positive T cells, driven by P27XR MARylation and NICD axis (28, 36). Inhibition of hsART1, therefore, offers the unique avenue of immunotherapy, targeting extracellular MARylation as a mechanism to assist T cell–mediated tumor rejection (28). Elevated hsART1 expression has also been associated with poor prognosis in colorectal cancer (37), glioblastoma (38), and various gastric cancers (39), leading to potentially broader relevance in other cancer models.

Knowledge of the transition-state (TS) structure of an enzyme-catalyzed reaction provides a framework for the rational design of tight-binding transition-state analogs (TSA). The measurement of intrinsic kinetic isotope effects (KIEs) provides the only experimental method to deliver subbond order resolution of the TS structure, including its bond distances, angles, and charges.

The mechanisms of the weak, nonspecific NAD⁺ glycohydrolase activity of cholera toxin subunit A (CTA, the ARTC namesake homolog) (40), pertussis toxin subunit A (PTA) (41), and diphtheria toxin fragment A (DTA) (42) have been studied by KIE, and the TS structures have been proposed. Of the three TS-characterized classes of ADP-ribosylating exotoxins, only the TS of the native ADP-ribosylation for PTA (43, 44) and DTA (45) have been studied with peptide or protein substrates. Across the ADP-ribosylating exotoxins, both the hydrolytic and transferase TSs are characterized by nearly completely dissociated nicotinamide (NAM) groups, only minimal interaction of the incoming nucleophile, and with a pronounced oxocarbenium characteristic (40–45). The canonical L-arginine specific MARylation catalyzed by the CTA was inaccessible due to the high substrate K_m for NAD⁺ and modestly slow turnover (40, 46).

In this study, we establish the first TS structure of L-arginine specific MARylation catalyzed by ARTCs through the measurement of the intrinsic KIEs for the MARylation of a P2X7 peptide analog by hsART1 (Fig. 1A–C). Together with computational exploration of the reaction coordinate, mapped with in silico KIEs, we can establish the TS structure by comparison to experimentally measured KIE values. This provides the first TS structure for an arginine-dependent ART. The TS structure and its electrostatic potential surface (ESPS) give insight into arginine-specific MARylation and provide information for the design of TSAs.

Results and Discussion

hsART1 MARylates P2X7 Peptide.

In vivo MARylation of P2X7 at R125 by hsART1 leads to constitutive activation of the receptor and ultimately NICD (Fig. 1B) (30–32). The ADP-ribosylation of P2X7 by hsART1 was modeled with a peptide spanning the flanking five amino acids surrounding the canonical in vivo MARylation site, R125, P2X7_(120−130), hereby referred to as P2X7 peptide. The wild-type P2X7 sequence contains two adjacent arginines, R125 and R126, a previously reported motif for in vivo ART specificity (47). The substrate specificity for MARylation of the P2X7 peptide was monitored by HPLC and LC-MS against the wild type sequence as well as peptides where arginines were replaced by glycines at either (or both) R125 and R126 positions, P2X7_(R125G), P2X7_(R126G), P2X7_{(R125G/R126G)} (SI Appendix, Figs. S2 A and B and S3). hsART1 can MARylate P2X7 peptide, in preference to P2X7_(R125G) or P2X7_(R126G) variants (Table 1). hsART1 can MARylate R126, but has a weaker substrate preference than for the native R125 site. hsART1 is unable to MARylate P2X7_{(R125G,R126G)} peptide that lacks both arginine sites (SI Appendix, Fig. S2).

Table 1.

Steady-state kinetic constants for hsART1 for MARylation of P2X7 peptides

Peptide	Sequence	kcat (s−1)	Km (mM)	kcat/Km (M−1 s−1)
wt	PEYPTRRTLCS	$11.1\pm 0.3$	$0.103\pm 0.010$	$1.08\pm 0.11\times {10}^5$
R125G	PEYPTGRTLCS	$10.2\pm 0.3$	$0.344\pm 0.029$	$2.96\pm 0.27\times {10}^4$
R126G	PEYPTRGTLCS	$9.19\pm 0.29$	$0.186\pm 0.017$	$4.93\pm 0.49\times {10}^4$

Kinetic constants were acquired from fitting to SI Appendix, Eq. S1. Assays were performed at 1 mM NAD⁺, various concentrations of peptide (0.0219 to 0.7 mM) initiated with the addition of hsART1.

Steady-state kinetic constants were determined for the MARylation of the P2X7 peptides from initial rates following MARylated-peptide formation using a luminescence-coupled assay (Table 1 and SI Appendix, Fig. S2C). hsART1 MARylation rates of P2X7 peptides are comparable to those of previously reported L-arginine analogs, with k_cat ranging between 11.1 $\pm$ 0.3 and 9.19 $\pm$ 0.29 s⁻¹. We observed comparable K_m values for R125 containing peptides increasing from 0.103 $\pm$ 0.010 to 0.186 $\pm$ 0.017 mM for P2X7 and P2X7_(R126G), respectively, with a further increase in K_m for P2X7_(R125G) to 0.344 $\pm$ 0.029 mM for the noncanonical MARylation R126 site. hsART1 shows 10-fold improved substrate specificity, (k_cat/K_m) for the wild-type P2X7 peptide compared to previously characterized L-arginine analogs (23). Peptide fragmentation was performed on the P2X7 peptides (SI Appendix, Fig. S3). For the wild-type peptide under nonsaturating NAD⁺ concentrations, we observed competition for MARylation between both R125 and R126 sites at partial turnover (SI Appendix, Fig. S3C). We observed no catalytic processivity and no double MARylation at both R125 and R126 sites. The lack of selectivity for specifically R125 is not surprising, given the small two- to threefold change in k_cat/K_m relative to the single glycine-substituted peptides (SI Appendix, Fig. S2C and Table 1). The lack of processivity prevents interference with KIE measurements; therefore, KIEs were performed on the wild-type sequence peptide as the RR-motif is the as a preferred substrate in vitro, and in vivo (47).

hsART1 Shows Modest Commitment to Catalysis In Vitro.

KIEs determined by intermolecular competition manifest on V/K and report isotopic discrimination on the rate-limiting steps up to, and including, the first irreversible step of the reaction. This encompasses both chemistry, which is reported on by the intrinsic KIE, KIE_int, and substrate binding. Only the KIE_int can be interpreted to map the TS structure with the potential contributions of substrate binding masking KIE_int and thus must be experimentally established.

The forward commitment (C_f) reflects the probability of the trapped substrate:enzyme Michaelis complex to proceed through catalysis relative to the dissociation back to free substrate and can be used to correct the experimentally observed KIE, KIE_obs, to the KIE_int. Arginine-specific MARylation is chemically stable and irreversible, making reverse commitment (C_r) unlikely for the hsART1 catalyzed reverse reaction over the experimental window (19, 48, 49).

MARylation of the P2X7 peptide was measured by the isotope trapping method (50). Forward commitment was determined by rapidly mixing hsART1 with [H_A1^′–³H]NAD⁺ to form the hsART1:NAD⁺ binary complex, followed by a rapid dilution with excess unlabeled NAD⁺ and P2X7 peptide. Analysis of the amount of substrate within the binary complex that is released or that is converted to product yielded a forward commitment factor of 0.106 $\pm$ 0.005 (SI Appendix, Fig. S4). The observed modest C_f results in a small correction to the KIE_obs to obtain the KIE_int (Eq. 1).

$$\begin{array}{c}\hfill {\text{KIE}}_{\text{int}}={\text{KIE}}_{\text{obs}}\left(1+C_f\right)-C_f.\end{array}$$ [1]

Measurement of Intrinsic KIEs for hsART1 MARylation.

Primary and secondary KIEs at isotopically sensitive positions on the nicotinamide mononucleotide (NMN) moiety of NAD⁺ for MARylation of the P2X7 peptide were measured by intermolecular competition using radiolabeled NAD⁺ (Fig. 2 and Table 2). The KIEs were determined by the change in the isotope ratios traced by ³H and ¹⁴C, comparing the accumulation of radiolabel on the MARylated P2X7 product at partial and complete conversion using Eq. 2,

$$\begin{array}{c}\hfill {\text{KIE}}_{\text{obs}}=\frac{ln(1-f)}{ln\left(1-f\frac{R_f}{R_0}\right)},\end{array}$$ [2]

Fig. 2.

Intrinsic KIEs and TS bonds for hsART1-catalyzed MARylation of P2X7 peptide. [C_A1^′–¹⁴C] on NAD⁺ are remote, control labels (KIE $=$ 1.000). Bond distances d1 and d2 are the initial bond distances scanned to generate the primary KIE landscapes (Fig. 4).

Table 2.

Intrinsic and calculated KIEs for the MARylation of P2X7 peptide catalyzed by hsART1

Label	Type of KIE	Intrinsic KIE∗	TS1 KIEcalc†	Final TS KIEcalc‡
[CN1′–14C$]$NAD+	Primary—14C	$1.039\pm 0.{003}^{(3)}$	$1.038$	$1.041$
[NN1–15N,CA1′–14C]NAD+	Primary—15N	$1.028\pm 0.{002}^{(3)}$	$1.028$	$1.027$
[HN1′–3H]NAD+	$\alpha$-Secondary—3H	$1.174\pm 0.{010}^{(4)}$	$1.165$	$1.168$
[ON4′–18O,CA1′–14C]NAD+	$\alpha$-Secondary—18O	$1.012\pm 0.{005}^{(7)}$	$0.995$	$1.006$
[HN2′–2H,CA1′–14C]NAD+	$\beta$-Secondary—2H	$1.104\pm 0.{002}^{(3)}$	$1.098$	$1.107$
[HA1′–3H]NAD+	Remote—3H	$1.002\pm 0.{003}^{(3)}$	$N.D$	$N.D$

^∗Experimental KIEs have been previously corrected to 0% substrate depletion using Eq. 2. Intrinsic KIEs were obtained by correction for forward commitment factor (Cf $=$ 0.106 $\pm$ 0.005; SI Appendix, Fig. S4 using Eq. 1). Number denotes the number of independent experiments.

^†Theoretical KIEs were determined on the intermediate TS (TS1, d(1^′C-NAM) $=$ 2.68 Å, d(1^′C-Gu) $=$ 2.22 Å).

^‡Final TS (d(1^′C-NAM) $=$ 2.68 Å, d(1^′C-Gu) $=$ 2.18 Å) using B3LYP/6-31G* calculations using the 6-31$+$G(d,p) basis set (Materials and Methods and SI Appendix); remote 1^′-AMP positions are used only as remote labels and were not included in the analysis.

where f is the fraction of substrate conversion, and R corresponds to the ratios ³H/¹⁴C ratios for the partial (20%, R_f) and complete (100%, R₀) conversion.

After correction for the forward commitment (Eq. 1), the observed KIEs are corrected to reflect the intrinsic KIEs, KIE_int.

The isotope effect for [C_A1^′–¹⁴C] is remote from the reaction center and assigned as unity, but the binding isotope effects (BIEs) of hydrogen at remote positions can influence the measurement of competitive KIEs on V/K. The KIE for the remote position [H_A1^′–³H] was measured and found to be unity (1.002 $\pm$ 0.003), therefore, where [H_A1^′–³H]NAD⁺ is utilized as a light isotope competitive reporter, the experimentally measured KIE is representative of the true observed KIE.

The KIE for the C1^′ anomeric carbon is sensitive to the contributions from both the leaving group (NAM) and the incoming nucleophile (guanidinium) at the TS during N-ribosyl transfer. The significant primary KIE for the [C_N1^′–¹⁴C] position, 1.039 $\pm$ 0.003, together with the large [N_N1-¹⁵N], 1.028 $\pm$ 0.002, indicates an asymmetric TS with modest contributions of the incoming nucleophile while the pronounced [N_N1-¹⁵N] is consistent with almost complete dissociation of the leaving group at the TS and loss of the C1^′–N1 bond order.

The modest [H_N1^′–³H] $\alpha$-secondary KIE, 1.174 $\pm$ 0.010, is consistent with weak residual bond order to C1^′ and substantial but incomplete sp³ to sp² rehybridization at the TS. The $\beta$-secondary [H_N2^′–²H] KIE, 1.104 $\pm$ 0.002, is highly sensitive to the H2^′–C2^′–C1^′–N1 dihedral and reports on the degree of the hyperconjugation between the C2^′–H2^′$\sigma$-bond and the 2p_z orbital of the C1^′ at the TS (51).

We determined a normal KIE for [O_N4^′–¹⁸O] of 1.012 $\pm$ 0.005 (established in 7 independent measurements, across 2 separately synthesized batches of [O_N4^′–¹⁸O, C_A1^′–¹⁴C]NAD⁺). Generally, inverse isotope effects for 4^′–¹⁸O near 0.98 have been reported for other N-ribosyl hydrolases (42), transferases (52), and uncatalyzed hydrolysis (53). The normal 4-¹⁸O KIE indicates a decrease in the O4^′ bond order at the TS, and is unique to L-arginine-specific MARylation.

Solvent Isotope Effect.

1,N⁶-ethenoadenine dinucleotide ($ϵ$-NAD⁺) is a fluorescent dinucleotide derivative of NAD⁺ and substrate analog for hsART1 that contains a fluorescent ethenoadenine base that is quenched in free solution by the NAM within the dinucleotide (23, 54). Cleavage of NAM permits the direct measurement of the fluorescent $ϵ$-ADPR product and allows for a direct measurement of MARylation (23). Initial reaction rates for the MARylation of P2X7 peptide were measured at variable concentrations of either $ϵ$-NAD⁺ or peptide in under near saturating concentrations of the secondary substrate in water or 80% D₂O (v/v) (Fig. 3). Corrections for viscosity changes at 80% D₂O, reaction rates were made by comparison to a 5% glycerol control (Fig. 3, Right and SI Appendix, Fig. S5) (55). Assays were performed at pH 8.0, where there is a well-defined plateau for both kinetic constants, V and V/K, to accommodate shifts in the pK_A that may be misinterpreted SKIEs (23, 56). With $ϵ$-NAD⁺ as the variable substrate, no SKIEs were observed on V or V/K. With the P2X7 peptide as the variable substrate, a small inverse SKIE was observed on V/K, 0.61 $\pm$ 0.05 (as fit by SI Appendix, Eq. S2), but this may be attributed to the effect of viscosity as independent fits V, and V/K are within the error of the fit compared to the glycerol control. Inverse SKIEs have been reported for CTA catalyzed NAD hydrolysis (40). Regardless, the absence of a normal SKIE for the hsART1 catalyzed MARylation indicates that there is no proton exchange at the TS. Therefore, any deprotonation that may occur with the attacking guanidinium happens before or afterMARylation.

Fig. 3.

MARylation of P2X7 peptide by hsART1 has no observed SKIE for $ϵ$-NAD⁺ (Top) and a modest inverse SKIE for P2X7 peptide (Bottom). (Left) Initial rates for varying $ϵ$-NAD⁺ under saturating P2X7 peptide (Top) and varying P2X7 peptide under saturating $ϵ$-NAD⁺ (Bottom). Data points and error bars represent mean $\pm$ SD, and solid lines represent fitting to SI Appendix, Eq. S1. (Right) Kinetic constants from fitting independent rate curves using the Michaelis–Menten equation (SI Appendix, Eq. S1) for $ϵ$-NAD⁺ (Top) and P2X7 peptide (Bottom). SKIE for P2X7 peptide was fit using SI Appendix, Eq. S2.

Computationally Modeling the hsART1 MARylation TS.

The ground state for hsART1 catalyzed MARylation used a structural model of the active site generated from an AlphaFold 3 model of the hsART1:NAD⁺:P2X7 trimer ternary complex (57). The generated model placed the canonical R125 in vivo MARylation site presented in close proximity to the C_N1^′ of the NMN moiety of NAD⁺ (Fig. 1C and SI Appendix, Fig. S1).

From the large [N_N1–¹⁵N] KIE (Table 2), and absent SKIE (Fig. 3), a late dissociative TS with little contribution from the departing NAM and modest contribution from the incoming guanidinium is established, with no assistance from acid–base chemistry at the attacking guanidinium is concerted with the TS. Thus, the L-arginine guanidine is likely to be protonated during the attack, with subsequent deprotonation of the product happening immediately after the transfer has occurred.

KIE landscapes are used to compare computed to intrinsic experimental KIEs values and thereby infer the TS structure. Computed KIEs for hsART1 used a ground state (GS) reaction model containing the side chains of E228, E230, R179, and S202 of the R-S-EXE motif of ARTC homologous enzymes. QM DFT and subsequent KIE calculations were performed varying the leaving group NAM d(C_N1^′–N_N1) and attacking nucleophile guanidinium d(N$\eta$1–C_N1^′) (Fig. 4 and SI Appendix, Fig. S6). Calculated and experimental primary [C_N1^′–¹⁴C] and [N_N1-¹⁵N] isotope effects, as well as the $\alpha$-secondary [H_N1^′–³H] KIE, converge within experimental error to give a late dissociative asymmetric TS structure with a weak bond order to the NAM leaving group, and modest bond order to the incoming nucleophile, characteristic of an D_N+A_N mechanism (Fig. 4A) However, these initial KIE landscapes (TS1 KIE_calc^b; Table 2) failed to recapitulate the measured normal [O_N4^′–¹⁸O] KIE. The initial computed ¹⁸O KIE value, 0.995, is significantly smaller than the experimental value of 1.012 $\pm$ 0.005, but a value larger than measured ¹⁸O KIEs for other ribosyltransferases (Table 2 and SI Appendix, Fig. S6) (42, 52, 53).

Fig. 4.

Comparison of intrinsic and calculated KIE values as a function of leaving group NAM (C1^′–N1) and attacking guanidinium nucleophile (N$\eta$1-C_N1^′) bond distances across the reaction coordinate. (A) Overlapping KIE landscapes for over 0.1 Å(Left), and a subsection calculated over 0.01 Å(Right) for the primary KIEs for 1^′–¹⁴C, 1–¹⁵N, and $\alpha$-secondary 1^′–³H, colored in red, green, and blue, respectively. The Right panel demonstrates the TS bond lengths in agreement with the 1–¹⁴C–C_N1^′ 1–¹⁵N–N_N1, and ³H–H_N1^′ KIE values, with the overlap region highlighted in gold. The shaded regions represent the experimental value for each respective KIE with one SD. (B) Representative TS1 taken from the overlapping region of A, representing a feasible TS (2.68, C_N1^′–N_N1 and 2.22, N$\eta$1-C_N1^′, with a single imaginary frequency). (C) Representative QM DFT bond scans and computed KIEs of feasible TS taken from the overlapping region, varying the C1^′–C2^′ distance as a function of varying C1^′–O4^′ distances for the 1^′–¹⁴C (Left), 1^′–³H (Middle), and 4^′–¹⁸O (Right). Points represent computed KIEs and the shaded area represents the experimental region (KIE_int$\pm$ SD).

Mapping the Unique Normal 4^′–¹⁸O KIE.

The influence of the delocalized ribocation in relation to the measured normal [O_N4^′–¹⁸O] $\alpha$-secondary KIE, was investigated by scans of the d(C_N1^′–O_N4^′), d(C_N1^′–C_N2^′), and d(C_N4^′–O_N4^′) bonds on a geometry within feasible TS structures (TS1; 2.68, C_N1^′–N_N1 and 2.22, N$\eta$1–C_N1^′, with a single imaginary frequency, Fig. 4B). Increasing the C_N1^′–O_N4^′ bond length decreased the oxocarbenium characteristic (Pauling Bond Order, BO; from 1.623 to 1.170) and increased the KIE_calc into the experimentally measured $\alpha$-secondary ¹⁸O KIE range (Fig. 4C). This, however, caused an increase in the calculated primary [C_N1^′–¹⁴C] and [H_N1^′–³H] $\alpha$-secondary KIEs (SI Appendix, Fig. S6). Elevation of these KIEs is inconsistent with converging on an S_N1, or S_N2-like mechanism, where as one of the KIEs is pulled toward unity, the other is increased (58, 59). Varying the C_N1^′–O_N4^′ bond as a function of the C_N1^′–C_N2^′ distance maintained the computed KIEs within the experimentally measured range when the ratio of the bond lengths was retained, but increased the $\beta$-secondary [H_N2^′–²H] KIE, and generated multiple imaginary frequencies when the C_N1^′–C_N2^′ was compressed (Fig. 4C and SI Appendix, Fig. S7). Decreasing the C_N4^′–O_N4^′ (BO: 0.819 to 0.693) and compression of the C_N1^′–C_N2^′ bond (1.143 BO at the TS) selectively increased the $\alpha$-secondary ¹⁸O KIE, gave agreement between the experimental and computed KIEs (SI Appendix, Fig. S8). The final TS structure of hsART1-catalyzed MARylation satisfied all experimentally measured KIE values and exhibited a single imaginary frequency corresponding to ADP-ribosyl transfer (Fig. 5 and Table 2).

Fig. 5.

Comparison of electrostatic potential surface (ESPS) calculated for the GS, and the TS that satisfies all experimentally measured KIE. Blue coloring corresponds to a relative electron deficiency, while red defines a relative electron enrichment. The natural charges from NBO calculations for O4^′, C1^′, C2^′, C4^′, N$\eta$1, N$\eta$2, N_E, and the summed charges for the NAM leaving group are annotated.

TS Bond Order and Charge Distribution.

A charge analysis of the TS relative to the GS gave a positively charged NAM partially delocalized to the anomeric C1^′ and between the C1^′–O4^′, with a $+$0.255 natural charge on C1^′ and partial oxocarbenium between the C1^′–O4^′ having an elevated BO of 1.131 (Fig. 5, SI Appendix, Fig. S9). As the C1^′–N1^′ bond is lost at the TS, electron density increases in the leaving NAM, with $+$0.387 natural charge at the C1^′ anomeric carbon at the TS. The NAM is essentially neutral at the TS. At the TS, the ribocation O4^′ increases in charge from $-$0.593 in the GS to $-$0.470 in the TS. Significant oxocarbenium character develops with a Pauling BO of 1.516 for the C1^′–O4^′ at the TS. The development of oxocarbenium-ribocation character is supported by the modest accumulation in electron density at C_N4^′, with partial loss of the C_N4^′–O_N4^′ bond. The delocalized ribocation is also uniquely characterized by the compression of the C1^′–C2^′ bond, leading to an increase in electron density and a decrease in charge at the C2^′ position from 0.062 to $-$0.007 from the GS to the TS. The positive charge of the attacking guanidinium is delocalized across the N$\eta$1 attacking nitrogen, N$\eta$2, and N_E. At the GS, there is electron enrichment at the E228 contact between the N$\eta$2 and N_E with a natural charge of $-$0.840, and $-$0.648, and electron deficiency on the N$\eta$1, with a natural charge of $-$0.896. At the TS, there is an increase in electron deficiency in the attacking N$\eta$1 to $-$0.848, the solvent-exposed proton increasing from 0.448 to 0.469. The electron deficiency across the E228 and N$\eta$2, N_E contact increase to $-$0.821 and $-$0.624, respectively.

Mutagenesis of E228 and E230.

E230 makes contact with the 2_N^′-hydroxyl of NMN and is essential for hsART1 activity in vitro, with mutation to glutamine completely abolishing activity (SI Appendix, Fig. S10). Homologous mutations of E230 to glutamine completely abolish both the MARylation and the weak promiscuous NAD⁺ hydrolase activities in ART homologs (60–64). Mutation of E228 to glutamine selectively reduces ART1 MARylation activity, but modest NAD⁺ hydrolase activity is retained, similar to other homologous mutations in arginine-dependent ARTs (SI Appendix, Fig. S10 and Table S1) (60–64). E230 facilitates an essential and unique charge at the TS where the generated charge is delocalized between the O_N4^′ and C_N1^′ but also partially with the C_N2^′ position of the NAD⁺.

Conclusions

Dysregulation of hsART1 function within the tumor microenvironment is emerging as a valid immunotherapy target by preventing tumors from killing CD8⁺ T cells via overactivation of P2X7 (28, 33–36). Experimental KIE measurements and DFT calculations permitted the determination of the first TS structure of L-arginine-specific MARylation. The lack of normal SKIEs, identified hsART1 achieves a distinct dication TS, where the incoming nucleophile guanidinium is protonated, attacking the positively charged NMN moiety. Previously characterized TS of MARylating toxins have a similar late dissociative TS but have attacking residue nucleophiles that are neutral or negatively charged, either a deprotonated histidine or cystine for the DTA (43, 44) or PTA (45), respectively. While unexpected, there is precedent for positively charged, protonated guanidinium of L-arginine acting as the preferred nucleophile across alternate reactions, including in select protein methylarginine transferases (PRMTs) (65, 66) and arginine-specific glycosyltransferases (GTs) (67). The absence of a normal SKIE in PRMT1, corroborated by QM/MM calculations, identified that arginine deprotonation occurs postmethyl transfer (65, 66). Similar to hsART1, PRMTs contain two glutamates that flank the guanidinium and provide sufficient activation for the protonated N$\eta$1 of L-arginine to attack the incoming donor methyl from S-adenosyl-l-methionine (SAM) (65, 66, 68). The L-arginine-specific GT NleB and SseK have been suggested to proceed through an S_Ni like mechanism from structural work (69, 70). A recent QM/MM study on NleB1 supported a late dissociative mechanism that is stabilized by an oxocarbenium in the transferring sugar but identified that the attacking guanidinium is protonated during initial transfer (67). A single glutamate sits in plane with the attacking guanidinium and activates the nucleophile through contacting both the N$\eta$1 and N$\eta$2 (67, 70). While strictly not a dication reactant TS similar to arginine-specific hsART1 catalyzed MARylation or PRMT1 catalyzed methyl transfer, NleB-like GTs highlight that simultaneous activation of both N$\eta$1 and N$\eta$2 may be key for facilitating the protonated guanidinium nucleophile.

The intrinsic KIEs are consistent with a late dissociative, asymmetric D_N$+$A_N reaction mechanism with minimal contributions from the leaving group NAM, at 2.68 Å, and minor contributions from the incoming L-arginine guanidinium, at 2.18 Å, similar to previously characterized NAD⁺-dependent enzymes (Fig. 6). The positive charge of NAM within NAD⁺ is delocalized across the C1^′ and C1^′–O4^′ bond, leading to a partial oxocarbenium characteristic in the GS. At the TS, the positive charge is further delocalized between the C1^′–O4^′, leading to an increase in the oxocarbenium characteristic. The resulting charge is, uniquely, stabilized across the C1^′–C2^′ bond by hsART1 E230 contacting the 2^′-hydroxyl of the reactant, leading to compression of the bond with loss of bond order between the C4^′–O4^′ bond.

Fig. 6.

More O’Ferrall–Jencks plot showing the traditional reaction space for NAD⁺ (or NMN) N-glycoside hydrolysis or transfer for previously characterized enzymes, measured by KIE. The axes are the bond orders from C1^′, the anomeric carbon, to the leaving group and the nucleophile. Bond orders in the reactant (Bottom Left) and products (Top Right) are shown. In a stepwise D_N$+$A_N (S_N1) mechanism, an oxocarbenium ion intermediate is traditionally formed (Top Left) by dissociation of the leaving group. The nucleophile attacks C1^′ in a separate step. A concerted, synchronous A_ND_N (S_N2) mechanism would follow the dashed diagonal line. The selected experimental Pauling Bond orders are shown for NAD⁺ glycohydrolase CTA (40), PTA (41), and DTA (53), pyrophosphorylase activity of NAMPT (75), ADP-ribosylation of acetyllysine by Af2Sir2 (52), MARylation for PTA (43, 44), DTA (45), and hsART1 (this work).

Select N-ribosyltransferases and hydrolases have been proposed to proceed through an endocyclic mechanism where the C1^′–O4^′ is first broken prior to the cleavage of the C1^′–N1 bond. This too would manifest as a normal ¹⁸O KIE; however, the absence of a normal SKIE also precluded this mechanism for hsART (58, 71–73).

Despite the mechanistic similarities between ARTCs and ARTDs, there lacks a tight-binding small molecule inhibitor for ARTCs, as seen in other ARTD family members such as PARPs (74). MARylation by the ARTD homolog, DTA has been previously characterized as having a completely dissociated NAM group and a pronounced oxocarbenium characteristic with an inverse $\alpha$-secondary ¹⁸O (42, 45). Previous TSAs based on the late dissociative, oxocarbenium TS of MARylating exotoxins CTA, PTA, and DTA have only shown modest binding affinities in the low micromolar range (45). While hsART1 exhibits a similar late dissociative oxocarbenium TS, the present subangstrom geometry and unique electrostatics of the hsART1 TS will assist the design of mechanism-based TSA inhibitors for arginine-specific MARylation. The unique TS features, such as loss of bond order between C4^′–O4^′, as well as inclusion of an analogous fragment to the positive nucleophile mimicking the dication TS, can be leveraged to push binding potency higher for future TSAs.

Materials and Methods

Supplementary materials and methods are available in SI Appendix. These methods encompass a detailed breakdown of materials; synthesis and purification of all isotope-labeled intermediates and NAD⁺; the measurement of forward commitment for the hsART1:NAD⁺ binary complex; further computational details.

Synthesis of Isotopically Labeled NAD⁺ Substrates.

[H_A1^′–³H]NAD⁺ & [H_N1^′–³H]NAD⁺, and [C_A1^′–¹⁴C]NAD⁺ & [C_N1^′–¹⁴C]NAD⁺ were synthesized enzymatically from [1–³H]ribose, and [1–¹⁴C]ribose, respectively (American Radiolabeled Chemicals, Inc.). [O_N4^′–¹⁸O,C_A1^′–¹⁴C]NAD⁺ was synthesized from [5–¹⁸O]glucose (Omicron Biochemicals, Inc.) and [1^′–¹⁴C]ribose. [H_N2^′–²H,C_A1^′–¹⁴C]NAD⁺ was synthesized from [2– ²H]ribose (Omicron Biochemicals, Inc.) and [1–¹⁴C]ribose. [N_N1–¹⁵N,C_A1^′–¹⁴C]NAD⁺ was synthesized from [1–¹⁵N]NAM (Advanced Molecular Technologies) and [1–¹⁴C]ribose.

Detailed protocols for the synthesis of isotopically labeled NAD⁺ are available in SI Appendix, Methods S3 and Figs. S11–S15.

Commitment to Catalysis.

The forward commitment factor (C_f) to catalysis was measured by substrate trapping (50) using [H_A1^′–³H]NAD⁺ with a rapid mixing Quench Flow (RQF-3, KinTek). The formation of radiolabeled MARylated-P2X7 peptide was quantified after the pulse into excess unlabeled NAD⁺, and P2X7 peptide, first separating the starting material from product by HPLC and then quantifying by liquid scintillation counting (LSC) of the product compared to a rapidly mixed but unchased control. A detailed procedure is available in SI Appendix.

Kinetic Isotope Effect Measurements: Experimental KIEs.

KIEs were measured under competitive conditions using the isotope ratio method. The isotopically sensitive label of interest was mixed with the corresponding remote label; when ¹⁴C was the reporting label for the isotopically sensitive position, [H_A1^′–³H]NAD⁺ was used, and when ³H was the reporting label, [C_A1^′–¹⁴C]NAD⁺ was used. The ³H to ¹⁴C cpm ratio was equal, with no less than 2 $\times$ 10⁵ cpm of ¹⁴C per experiment. Radiolabeled NAD⁺ was repurified by HPLC the day before use to remove any ADPR generated from degradation as outlined in SI Appendix.

The reaction mixture (0.5 mL) contained 50 mM HEPES (pH 8.0) and 0.1 mM NAD⁺ (including ³H and ¹⁴C labels) and 0.25 mM P2X7_(120−130). The reaction was initiated by adding hsART1 to a final concentration of 5 nM, and 0.4 mL of the reaction was quenched with 0.5% TFA (final concentration) after 15–20% depletion of the substrate. The remaining 0.1 mL was driven to 100% competition with the addition of 50 nM hsART1; the reaction was then allowed to proceed for 60 min and quenched with TFA (0.5% final). Unreacted substrates and products were separated using reverse-phase HPLC as outlined in SI Appendix.

Four replicate injections were purified by RP-HPLC for both the partial and complete reactions (SI Appendix, section S2C). Peaks corresponding to the MARylated P2X7 peptide product were collected and dried overnight. The dried samples were redissolved in 200 $\mathrm{\mu }$L of water and mixed with 10 mL of scintillation fluid (Ultima Gold, Revvity).

The ³H/¹⁴C ratios were measured by scintillation counting (5 min/cycle, 10 cycles; Tricarb 4910 TR Revvity). The channels 1 and 2 were set to encompass all ³H signal ($>$99.9%) in channel 1, with counts in channel 2 solely from ¹⁴C. After calibration of the ratio (r) between those two channels using ribose, ([1–¹⁴C]ribose, $>$1 $\times$ 10⁵ cpm), the exact amount of ³H and ¹⁴C in each of the samples was determined using

$$\begin{array}{c}\hfill {\text{cpm}}^{{(}^{3}H)}={\text{cpm}}_{\text{channel1}}-{\text{cpm}}_{\text{channel2}}r,\end{array}$$ [3]

$$\begin{array}{c}\hfill {\text{cpm}}^{{(}^{14}C)}={\text{cpm}}_{\text{channel2}}(1+r).\end{array}$$ [4]

To correct the observed KIEs, KIE_obvs, to 0% depletion of substrate, Eq. 2 was used. When the isotopic enrichment was $<$99%, observed KIEs were corrected using the following equation:

$$\begin{array}{c}\hfill {\text{KIE}}_{\text{obs}}=({\text{KIE}}_{\text{obs*}}-1+e)/e,\end{array}$$ [5]

where KIE_obvs∗ is the apparent uncorrected observed KIE and e is the fractional isotope enrichment of the label at the isotopically sensitive position. When [H_N2^′–²H,C_A1^′–¹⁴C]NAD⁺ and [O_N4^′–¹⁸O,C_A1^′–¹⁴C]NAD⁺ were the heavy label, e was 0.980 and 0.903, respectively.

Solvent Deuterium Kinetic Isotope Effects.

To probe the rate-limiting nature of proton transfer steps, solvent deuterium KIEs (SKIE) on hsART1 catalyzed MARylation of P2X7 were determined using a fluorescent analog of NAD⁺, nicotinamide 1,N⁶-ethenoadenine dinucleotide ($ϵ$-NAD⁺) (23). A detailed summary of SKIEs performed with $ϵ$-NAD⁺ is available in SI Appendix.

Computationally Modeling the TS Structure.

For hsART1 catalyzed MARylation of P2X7 peptide, we followed the general procedures previously established for solving TS structures through quantum mechanical (QM) calculations and computational TS analysis. Detailed computational methods can be found in SI Appendix (59, 75, 76). All subsequent geometry optimizations and frequency calculations were performed with Gaussian 16 (77). In our final QM models, we simulated the GS using B3LYP functional and 6-31$+$G(d,p) basis set. The calculated vibrational frequencies were scaled by 0.9613 to reproduce the true vibrational zero-point energies for calculations of theoretical KIEs in the ISOEFF07 program at the experimental temperature (25 ^°C) (78, 79). ESPSs were calculated by using the CUBE program from Gaussian 16 (77) and visualized in Pymol 3.07 (80) (isovalue $=$ 0.04). More details of the computational methods are available in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Accession Codes: hsART1 (UniProt: P52961) (81), hsP2X7 (Uniprot: Q99572) (82), hsARH1 (UniProt: P54922) (83), hsNUDIX5 (UniProt: Q9UKK9) (84), and CsPPDK (UniProt: P22983) (85). All other data are included in the manuscript and/or SI Appendix.