Reconstitution of protein arginylation pathways in bacteria for robust identification and quantification Authors: Xin Lan, Richard M Searfoss, Daniel Lee, Sahil Bhaskaran, Thilini Abeywansha, Benjamin A Garcia, Zongtao Lin, Yi Zhang Published: 2026 ABSTRACT AbstractATE1 is a conserved enzyme that catalyzes the covalent addition of arginine to proteins bearing N-terminal or mid-chain Asp and Glu residues. N-terminal (Nt) arginylation can also occur on Cys, Asn, and Gln following enzymatic conversion, often marking proteins for degradation. Essential for development, this pathway contributes to protein quality control and stress responses. Despite growing insight into ATE1 structure and function, the mechanisms governing its substrate selectivity and coordination with upstream oxygenase and deamidase remain poorly defined. Here, we reconstitute the human processing cascades that generate Nt-arginylated proteins in E. coli, enabling step-resolved analysis of arginylation outcomes in a cellular context. By co-expressing human ADO, NTAN1, or NTAQ1 with ATE1 in a modular system, we achieved efficient conversion of Nt-Cys, Asn, and Gln into arginylation-permissive forms, recapitulating key features of upstream processing. Using this platform, we demonstrated that N-terminal processing is efficient and that ATE1 preferentially modifies protein N-termini over internal acidic residues. Mid-chain arginylation of α-synuclein was detectable but occurred at low frequency, with no major differences in site selectivity observed across the ATE1 isoforms tested. Together, this bacterial reconstitution system provides a scalable experimental platform for quantitative, protein-level analysis of ATE1 substrate specificity under defined conditions.Subject terms: Enzymes, Post-translational modifications Introduction Protein arginylation is catalyzed by a highly conserved enzyme ATE1 (arginyl-tRNA-protein transferase 1) in eukaryotes and is increasingly recognized as a crucial player in normal development and human diseases1–4. While dispensable for yeast viability, homozygous deletion of Ate1 in mice is embryonically lethal, marked by defects in cardiovascular development and angiogenesis2,5. Conditional deletion of Ate1 in vertebrate tissues leads to diverse phenotypes, including infertility, growth retardation, neurodevelopmental abnormalities, and cardiomyopathy6–10. At the molecular level, protein arginylation involves the covalent transfer of arginine from arginyl-tRNAArg to exposed acidic residues, typically N-terminal aspartate (Asp) or glutamate (Glu), on substrate proteins11. This modification is required for the Arg/N-degron pathway, where many arginylated proteins are subsequently recognized by the UBR family of E3 ubiquitin-protein ligases, poly-ubiquitinated, and then degraded12. It is therefore generally accepted that protein arginylation broadly regulates cellular functions through controlling the turnover of its protein substrates. Beyond degradation, ATE1 also mediates non-degradative arginylation events that modulate cytoskeletal dynamics, myosin phosphorylation, and cellular stress responses, highlighting its broader role in cellular regulation8,13. Over the past decade, despite the discovery of several hundred putative arginylated proteins, only a handful of physiological ATE1 substrates have been confidently identified and subsequently validated in vivo. For instance, proteolytic fragments of TDP43 found in brains of patients with frontotemporal lobar degeneration, such as TDP43219 and TDP43247 (bearing Nt-Asp219 and Asp247, respectively), are substrates for ATE1-mediated arginylation, targeting them for degradation14,15. Similarly, endoplasmic reticulum (ER) chaperone GRP78/Bip (bearing Nt-Glu) is arginylated to facilitate the clearance of misfolded proteins under stress conditions16–18. Other Nt-Asp/Glu bearing physiological substrates of ATE1 include β-actin and calreticulin, although in a non-degradative manner19–21. Notably, protein substrates bearing an oxidized N-terminal cysteine are also robust ATE1 substrates in plants and animals (Fig. 1a). Briefly, a thiol oxidase named cysteamine (2-aminoethanethiol) dioxygenase (ADO) catalyzes the conversion of Nt-Cys to cysteine sulfinic acid (Nt-Cys-SO2H, or Nt-Cys*), which is subsequently arginylated by ATE1, leading to its poly-ubiquitination and degradation. Human ADO regulates Nt-Cys containing proteins such as the RGS4/5, key modulators of G-protein signaling22. In plants, cysteine oxidases regulate several transcription factors related to seed germination and oxygen sensing 23–25. Reconstitution of human Cys/N-degron generation pathway. a Schematic overview of the Arg/N-degron generation pathways. N-terminal Cys, Asn, and Gln residues are enzymatically converted to oxidized Cys (C*), Asp, and Glu by ADO, NTAN1, and NTAQ1, respectively, and subsequently arginylated by hATE1-2 to generate RC*, RD, or RE at the N-terminus. b Modular bacterial system for reconstituting Nt-Cys arginylation. A dual-plasmid strategy co-expresses an SMT3-C-Ub fusion substrate and the modifying enzymes ADO, Ulp1, and hATE1-2. c SDS-PAGE analysis of C-Ub purification. Lanes correspond to samples collected during the purification procedure: whole-cell lysate (WCL), soluble fraction (Sup), pellet, flow-through (FT), Ni-NTA beads (Ni-NTA), and eluate (Elu). A distinct band at ~11 kDa corresponds to purified C-Ub. d Size-exclusion chromatography of purified C-Ub. e Bottom-up LC-MS/MS analysis of N-terminal peptide (CDIAALVVDNGSGMGQIFVK) of model substrate C-Ub. MS1 spectrum (left) showing peaks corresponding to unmodified cysteine with carbamidomethylation C(CAM), di-oxidized cysteine C(O2), and di-oxidized cysteine with N-terminal arginylation R-C(O2). MS2 spectrum (right) with ion annotations for N-terminal peptide starting with di-oxidized cysteine or arginylated di-oxidized cysteine. Sequence coverage is shown above. f Top-Down LC-MS/MS analysis of C-Ub expressed with ADO alone (top) or co-expressed with ADO and hATE1-2 (bottom). Peaks 1, 2, and 3 correspond to unmodified C-Ub, the di-oxidized Cys species C(O2), and the arginylated di-oxidized species R-C(O2), respectively. Vector maps indicate the expression constructs used in each condition. g Quantification of the deconvoluted MS1 spectra shown in (f). Relative abundance (%) was calculated based on the peak area. Aside from the N-terminal Asp, Glu, and Cys*, Nt-asparagine (Asn) and glutamine (Gln) can be deaminated by dedicated enzymes NTAN1 (N-Terminal Asparagine Amidase 1) and NTAQ1 (N-Terminal Glutamine Amidase 1) in higher eukaryotes, converting them into Nt-Asp and Glu susceptible to ATE1-mediated arginylation and downstream degradation (Fig. 1a). For example, the degradation of Drosophila IAP1, a key regulator of apoptosis, depends on sequential N-terminal Asn deamination by NTAN1 and arginylation by ATE1, targeting the caspase-generated fragment for proteolysis in the fly eye26. Similarly, in mammals, an autocleaved fragment of the deubiquitylase USP1 (Gln-Usp1Ct) is targeted for degradation following NTAQ1-dependent deamination and subsequent arginylation by ATE1, contributing to the maintenance of genomic stability27. Besides Nt-arginylation, native α-synuclein (αSyn) purified from the mouse and human brain was arginylated in vivo at two internally located Glu residues, E46 and E83, detected by mass spectrometry28,29. Both E46 and E83 in αSyn are important for αSyn function, and their arginylation strongly decreases with age and negatively impacts αSyn pathological intracellular aggregation29. For instance, the E46K mutation in αSyn enhances its fibrillization, cellular toxicity, and pathological aggregation in vitro and in cultured cells, and is associated with early-onset familial Parkinson’s disease and Lewy body dementia30–33. Similarly, the E83A mutation was shown to facilitate αSyn intracellular aggregation and the formation of fibers34,35. Beyond αSyn, ATE1-mediated mid-chain arginylation has been identified on other cytosolic and structural proteins, including actin, myosin, tubulin, and translation-related factors, where it modulates protein function and cellular stress responses without promoting degradation 13,36. Interestingly, except for β-actin, none of the ATE1 substrates and modification sites mentioned above is conserved in all species from yeast to human, highlighting the evolutionary divergence and substrate diversity of the Arg/N-degron pathway. This substrate variability is further supported by recent unbiased, mass spectrometry–based proteomic analyses, which systematically identified ATE1 targets and revealed over 200 distinct arginylation sites across the human proteome37. As the list of candidate arginylation sites grows, rigorous validation of actual ATE1 substrates becomes increasingly critical yet technically challenging. Currently, two main validation strategies are employed: in-cell degradation and in vitro arginylation assays38. In-cell approaches offer greater physiological relevance but are often indirect, as arginylation is inferred from substrate degradation rather than measured directly, largely due to the absence of specific anti-arginylation antibodies and short half-lives of many arginylated proteins38. In contrast, in vitro assays utilize purified components of the reaction, including arginine, arginyl-tRNA synthetase (ArgRS), properly folded tRNAArg, and synthetic peptides or recombinant proteins to reconstitute the arginylation reaction, allowing direct monitoring of ATE1 activity through detection of products39–41. Despite this advantage, the approach is technically demanding and often suffers from inconsistent efficiency and yield, making it less suitable for probing site specificity or substrate selectivity in native proteins. To overcome these technical constraints, we have previously established an in-bacteria arginylation assay, in which substrate proteins were co-expressed with ATE1 in E. coli and then recombinantly purified for downstream biochemical characterization42. Using a model substrate containing a D1DIAALVVDNGSGMG sequence fused to ubiquitin (referred to as DDIAA-Ub hereafter), we demonstrated robust N-terminal arginylation at the Nt-Asp site. A key advantage of this system is its use of endogenous bacterial arginyl-tRNA as cofactors, closely recapitulating the native cellular context of protein synthesis and arginylation. In addition to its reproducibility, this assay enables the direct production and characterization of mid-chain arginylated proteins without the need for complex chemical synthesis43,44. In this study, we expand our in-bacterial assay to fully reconstitute the Arg/N-degron generation pathways, enabling direct assessment of N-terminal arginylation on substrates bearing Nt-Asn, Nt-Gln, and oxidized Nt-Cys*. We demonstrate that human NTAN1, NTAQ1, and ADO are highly efficient enzymes, achieving near-complete conversion of Nt-Asn, Gln, and Cys substrates into arginylation-permissive forms. Notably, N-terminal arginylation occurs at significantly higher efficiency than mid-chain arginylation, even in well-characterized substrates such as αSyn. Mid-chain arginylation was preferentially detected within the α-helical regions of αSyn, rather than in its disordered segments, and no discernible residue-specific selectivity was observed across human and mouse ATE1 isoforms for internal Asp or Glu sites. Results Reconstitution of the human Cys/N-degron pathway in bacteria The Cys/Arg N-degron pathway is a specialized protein degradation route that plays a significant role in cellular responses, particularly in sensing oxygen levels and oxidative stress45,46. It primarily targets proteins that have a cysteine (Cys) at their N-terminus, which is typically exposed following the removal of the initiator methionine or proteolysis. Following exposure, the Nt-Cys is oxidized to Nt-Cys-SO2H by ADO in humans or non-enzymatically by reactive oxygen species (ROS), generating a functional mimic of the carboxylate of Nt-Asp and enabling downstream arginylation by ATE1 (Fig. 1b). Given the variability of ROS-dependent oxidation, we reconstituted the Cys/N-degron branch by introducing ADO into our bacterial co-expression system. In brief, we utilized a dual-plasmid co-expression strategy in which one plasmid carried the SMT3-tagged Nt-Cys substrate, with or without ADO, and conferred kanamycin resistance. The second plasmid constitutively expressed Ulp1 protease and optionally human ATE1-2 (isoform 1B7B, unless otherwise specified), under ampicillin selection (Fig. 1b). This modular design allowed controlled progression of the pathway, facilitating stage-specific isolation of intermediates for biochemical and structural characterization. To assess its utility in analyzing cysteine-dependent processing, we modified the DDIAA-Ub model substrate by introducing an Nt-Cys, yielding the CDIAA-Ub variant (referred to as C-Ub hereafter). We first evaluated the expression and purification of the C-Ub substrate. A distinct band at the expected molecular weight (11 kDa for the CDIAA-Ub-His8 fusion protein) was observed following affinity purification from bacterial lysates, confirming successful expression and purification (Fig. 1c). Subsequent gel filtration chromatography revealed a symmetric elution peak consistent with its theoretical molecular size (Fig. 1d), indicating the protein is structurally homogeneous and well suited for downstream analyses. Next, we examined the in-cell modification state of C-Ub upon co-expression with both human ADO and hATE1-2 using liquid chromatography-tandem mass spectrometry (LC-MS/MS) and bottom-up proteomics. Three clearly distinct N-terminal peptides of C-Ub were identified, corresponding to the unmodified substrate (WT), di-oxidized cysteine (di-ox), and the arginylated di-oxidized form (di-ox + R). The MS1 and MS2 spectra provided well-resolved precursor masses and fragment patterns supporting the confident identification of these three states (Fig. 1e, Supplementary Figs. 1-3). These species reflect the sequential enzymatic steps of Ulp1 cleavage, cysteine oxidation, and ATE1-mediated arginylation. We then sought to quantify the progression of this multistep modification on proteins, which represents a major challenge when detecting arginylation in animal cells. To this end, we employed intact protein mass spectrometry, in which the molecular weight of the full-length protein is directly measured (Fig. 1f, g and Supplementary Fig. 4). C-Ub purified upon co-expression of ADO and ATE1 exhibited a predominant peak (~93%) with an observed mass of 11,340 Da, aligning with the anticipated arginylated di-oxidized product. Minor peaks representing the unmodified (4%) and di-oxidized-only (3%) forms were also detected. In contrast, C-Ub co-expressed with ADO was predominantly in the oxidized form (~99%), with only approximately 1% remaining unmodified. Of note, we roughly estimated the relative stoichiometry of human ATE1-2 and X-Ub substrate to be approximately 1:40 ~ 60, based on the observed purification yields of 7 nmol (0.4 mg ATE1) and 400 nmol (~5–6 mg X-Ub) per liter of bacterial culture, through dozens of purifications performed. These results demonstrate successful reconstitution of the Cys/N-degron modification cascade, where Nt-Cys-containing substrates proceed efficiently and selectively toward the fully modified, arginylated form. Reconstitution of the human Nt-Asn/Gln degron pathway Given the highly efficient reconstitution of the Nt-Cys degron processing cascade, we sought to expand the in-bacteria arginylation system to the Nt-Asn/Gln degron to evaluate their processivity as well. In human cells, proteins bearing Nt-Asn and Nt-Gln are deaminated by enzymes NTAN1 and NTAQ1, respectively, generating Nt-Asp and Nt-Glu residues that are susceptible to arginylation (Fig. 1a). To model this process, we modified the DDIAA-Ub model substrate by substituting the Nt-Asp with Asn or Gln, producing the NDIAA-Ub and QDIAA-Ub variants (referred to as N-Ub and Q-Ub, respectively). These substrates, along with their corresponding deaminases (NTAN1 for N-Ub and NTAQ1 for Q-Ub), were co-expressed from a single plasmid and introduced into the bacterial system alongside Ulp1 and human ATE1-2 (Fig. 2a). Like the C-Ub substrate, we expected to detect unmodified, deaminated (de-amine), and arginylated (de-amine + R) forms of both N-Ub and Q-Ub substrates. Reconstitution of human Nt-Asn/Gln degron pathway. a Schematic of bacterial reconstitution of Nt-Asn/Gln arginylation pathways. b Bottom-up LC-MS/MS analysis of the N-terminal peptide (NDIAALVVDNGSGMGQIFVK) of model substrate N-Ub. MS1 spectra (left) showing peaks corresponding to the deamidated form N(deami) and its arginylated counterpart R-N(deami). MS2 spectrum (right) with ion annotations for the arginylated, deamidated N-terminal peptide. Sequence coverage is shown above. c Top-down LC-MS/MS analysis of N-Ub expressed with NTAN1 alone (top) or co-expressed with NTAN1 and hATE1-2 (bottom). Peaks 1 and 2 correspond to the deamidated species N(deami) and arginylated species R-N(deami), respectively. d Quantification of the deconvoluted MS1 spectra shown in (c). Relative abundance (%) was calculated based on the peak area. e,f Top-down LC-MS/MS analysis and quantification (f) of Q-Ub expressed with NTAQ1 alone (top) or co-expressed with NTAQ1 and hATE1-2 (bottom). Peaks 1 and 2 correspond to the deamidated species Q(deami) and arginylated species R-Q(deami), respectively. Relative abundance (%) was calculated based on the peak area. However, LC-MS/MS analysis of the N-Ub substrate co-expressed with ATE1 detected only the deamidated intermediate and arginylated product, with no evidence of the unmodified form (Fig. 2b, Supplementary Figs. 5-6). Intact protein analysis further showed that deamination followed by arginylation (de-amine + R) occurred in approximately 56-86% of N-Ub substrates. Although intact mass analysis cannot resolve the 1 Da difference between the unmodified and de-aminated N-Ub, the absence of the unmodified peptide in the coupled subsequent top-down MS/MS data suggests that the remaining 14-44% corresponds to the deamidated but non-arginylated intermediate (Fig. 2c, d, Supplementary Figs. 7–9, and Supplementary Table 1,2). Consistently, N-Ub co-expressed with NTAN1 was fully deaminated (Fig. 2d, Supplementary Table 1-2). Similarly, Q-Ub substrate co-expressed with hATE1-2 contained 48-96% arginylation following deamination (Fig. 2e, f, Supplementary Fig. 10), revealing that EDIAA-Ub is also an efficient substrate for hATE1-2 comparable to DDIAA-Ub or C*DIAA-Ub (Figs. 1g, 2d). Similar to N-Ub, unmodified peptide was not detected in Q-Ub samples regardless of ATE1 expression, reinforcing the high catalytic efficiencies of NTAN1 and NTAQ1 under these experimental conditions. Together, these findings confirm that all three pathways of precursor N-terminal modifications—cysteine oxidation, and deamidation of asparagine and glutamine—can be functionally reconstituted within this bacterial platform. This extends prior peptide-based assays to a scalable cellular system that enables quantitative, protein-level analysis while minimizing confounding downstream modifications and rapid turnover. By reconstituting each step from N-terminal processing to arginylation, this system provides a simplified and versatile platform for subsequent studies using diverse substrate sequences. N-terminal arginylation exhibits markedly higher catalytic efficiency than the mid-chain Beyond Nt-arginylation, human and mouse ATE1 enzymes also modify mid-chain Asp and Glu residues, with α-synuclein (αSyn) as a prominent substrate featuring arginylation at E46 and E83 sites28,29. We have previously observed weak signals consistent with mid-chain arginylation in DDIAA-Ub co-expressed with human ATE1, inferred from molecular weight shifts in intact mass analysis; however, these signals were at or near the noise level. Taking advantage of the flexibility of our bacterial arginylation system, we substituted DDIAA-Ub with native αSyn, while preserving the C-terminal His-tag for purification. The αSyn N-terminus harbors a natural Asp at position 2, following the initiator Met, making it an ideal substrate to assess the relative efficiency of N-terminal versus mid-chain arginylation by ATE1 (Fig. 3a). In brief, upon co-expression with Ulp1 and ATE1, αSyn lacking Met1 (referred to as D2-αSyn hereafter) undergoes Ulp1-mediated cleavage and exposes Nt-Asp, creating a site for Nt-arginylation, while mid-chain (E46 and E83) sites remain accessible immediately after translation. N-terminal arginylation exhibits markedly higher catalytic efficiency than the mid-chain. a Schematic of bacterial expression of αSyn arginylation. Ulp1 cleavage exposes the native N-terminal Asp in D2-αSyn, enabling ATE1-mediated arginylation at both the N-terminus and internal acidic residues (e.g., E46, E83). b SDS-PAGE analysis of αSyn purification. c,d Top-down LC–MS/MS analysis and quantification (d) of purified D2-αSyn expressed without ATE1 (left) or co-expressed with hATE1-2 (right). Peaks 1 and 2 correspond to unmodified αSyn (WT) and mono-arginylated αSyn (+R). A species corresponding to doubly arginylated αSyn (+2 R) was not detected (N.D.). Relative abundance (%) was calculated based on the peak area. Similar to other substrates, D2-αSyn expressed with and without ATE1 was purified via affinity chromatography (Fig. 3b), showing a distinct band at the expected molecular weight (15 kDa). Following additional purification by size exclusion chromatography, D2-αSyn was subjected to LC-MS/MS and intact protein analysis. Surprisingly, despite the near-complete sequence coverage, no arginylated peptides at E46 or E83 were detected in ATE1-expressing samples. By contrast, arginylation at D2 was confidently identified, suggesting the occurrence of Nt-arginylation. Intact analysis revealed that 97% of D2-αSyn contained a single arginylation site in the presence of ATE1, whereas double- or multi-arginylated αSyn were not detected (Fig. 3c, d, Supplementary Fig. 11). Together, these results demonstrated that Nt-arginylation is substantially more efficient, while mid-chain arginylation was undetectable or below the sensitivity of our assay. Mid-chain arginylation of αSyn is selective for glutamate residues in α-helical regions In brain tissue, αSyn arginylation at the E46 and E83 sites has been reported to occur at levels ranging from 3% to 50% of the total αSyn pool29. We suspected that the presence of Nt-Asp2 in D2-αSyn might supersede mid-chain arginylation by preferentially directing ATE1 to the N-terminus. To enrich arginylation of αSyn toward internal sites, we constructed the WT αSyn that retains the native initiator methionine (M1-αSyn). This construct was co-expressed with Ulp1 and hATE1-2 in E. coli using the same dual-plasmid system and purified as previously described (Fig. 4a). To our surprise, M1-αSyn was largely unmodified, exhibiting the expected molecular weight of 15,283 Da. Neither mono- nor multi-arginylated forms were detected by intact mass spectrometry (Fig. 4b, Supplementary Fig. 12), suggesting that mid-chain arginylation of αSyn was below the detection threshold. Enrichment of mid-chain arginylation products. a Schematic of bacterial expression of Met1-αSyn for mid-chain arginylation. b Top-down LC-MS/MS analysis of purified M1-αSyn expressed with hATE1-2. Peaks correspond to unmodified M1-αSyn (15283 Da) and a + 156 Da species consistent with mono-arginylation (+R, 15439 Da). c Representative arginylated peptides identified by LC-MS/MS and corresponding modification sites. d Bottom-up LC-MS/MS analysis of site-specific arginylation on M1-αSyn. MS1 spectra (left) and annotated MS2 spectra (right) for peptides modified at E20 (top) and E57 (bottom). Sequence coverage and ion annotations are shown. e Immunoblotting of ATE1 and M1-αSyn expression from three independent E. coli clones. f Summary of arginylation sites with peptide counts (#) identified by bottom-up LC-MS/MS proteomics of M1-αSyn co-expressed with human hATE1-2 (1B7B). Bolded residues represent high-confidence arginylation sites that were reproducibly detected in all three biological replicates. g Sequence logos centered on arginylated (top) and non-arginylated (bottom) Asp/Glu residues. Logos were generated using WebLogo 3 based on ±20 amino acid windows. h Mapping of all Asp and Glu residues in the αSyn primary sequence alongside predicted secondary structures and normalized hydrophobicity. Secondary structure was predicted using JPred4, and hydrophobicity was calculated using the Kyte & Doolittle scale in ProtScale. We therefore turned to more sensitive bottom-up proteomic analysis to detect low-abundance arginylation events. Indeed, multiple arginylated peptides were identified across the αSyn sequence, particularly at sites containing E20, E28, E35, E46, and E57 with high confidence (Fig. 4c, d, Supplementary Figs. 13-14 and Supplementary Data 1). Additional sites, including E61 and D135, were also detected with lower confidence (score < 300), likely due to poor fragmentation or increased peptide length. Peptides spanning E83 were readily detected, however, no arginylation at this site was observed under our experimental conditions. To assess whether hATE1-2 preferentially targets specific mid-chain sites, we performed the MS analysis on three independently prepared biological replicates (n = 3), each derived from a distinct E. coli clone. Although expression levels varied slightly among clones, both ATE1 and αSyn were robustly detected in all samples using western blot (Fig. 4e). Corresponding bottom-up proteomic analysis revealed a positive correlation between ATE1 expression levels and the number of detected arginylation sites, with clone 3 showing the most modification events (Fig. 4f). Notably, E20, E28, E46, and E57 were consistently detected across all replicates with high confidence and peptide counts, indicating that these sites represent robust and reproducible targets of mid-chain arginylation by ATE1. To investigate potential determinants of arginylation site specificity, we generated sequence logos centered on frequently modified D/E residues, which were detected in at least two of three biological replicates. Compared to unmodified D/E residues, those undergoing arginylation appeared to be flanked by regions enriched in features characteristic of α-helices, specifically, a repeating pattern of hydrophobic residues spaced every 3-4 positions, consistent with alignment along a single face of the helix due to its helical periodicity (Fig. 4g). This signature was more evident on the C-terminal side of the modification site, indicating a directional bias in the sequence context. Consistent with this observation, secondary structure analysis of αSyn revealed that frequently arginylated sites, such as E20, E28, E46, and E57, are predominantly located within α-helical regions, whereas lower-confidence sites tend to fall within disordered segments displaying low hydrophobicity (Fig. 4h). Notably, the sequence logo derived from unfavorable sites closely resembles that observed in peptide array experiments, both showing enrichment of negatively charged residues (E/D) downstream of the arginylation site47. Collectively, these results suggest that ATE1 isoform 1B7B (hATE1-2) exhibits low catalytic efficiency for mid-chain arginylation, with a preference for Glu residues located upstream of α-helical segments. ATE1 isoforms exhibited similar selectivity toward αSyn Mouse and human ATE1 mRNAs undergo alternative splicing at exons 1 (1 A/1B) and 7 (7 A/7B), resulting in four major isoforms: 1B1A, 1B7B, 1A7A, and 1A7B36,48. These isoforms have been primarily characterized in mouse and are denoted as isoform 1, 2, 3, and 4, respectively. Among them, mouse isoforms 3 (1A7A) and 4 (1A7B) were shown to preferably target mid-chain arginylation sites in vitro, based on reconstituted arginylation assays using E. coli tRNAArg and arginyl-tRNA synthetase39,47. To evaluate isoform-specific substrate selectivity and activity of ATE1, we systematically compared the known splice variants of ATE1 from both human (1B7B and 1B7A) and mouse (1B1A, 1B7B, 1A7A, and 1A7B). Sequence alignment of human and mouse ATE1 isoforms revealed that exon 1 is conserved mainly within secondary structure elements, with sequence differences primarily localized to connecting loops. In contrast, exon 7 shows greater divergence, including variations within predicted α-helical regions (Fig. 5a). We mapped exons 1 and 7 onto our previously solved structure of human ATE1 isoform 2 (1B7B) in complex with substrate and tRNA to analyze the spatial positioning of the alternatively spliced regions (Fig. 5b). Interestingly, exons 1 and 7 locate on opposite sides of the peptidyl-substrate binding cleft, each forming one boundary of the pocket and potentially contributing to substrate binding thus selectivity. However, closer examination of isoform-specific residues revealed that most of them are located outside of the substrate-facing surface of the cleft, except for residues R22 in exon 1B and Q304 and E309 in exon 7B (Fig. 5b). These residues are neither fully conserved between human and mouse isoforms nor directly involved in substrate binding. For example, R22 is a positively charged arginine in human isoforms 1 and 2, while the corresponding residue in the equivalent mouse isoforms is a neutral glutamine. This lack of conservation suggests that these residues are unlikely to serve as determinants of evolutionarily conserved substrate specificity. Furthermore, the Cα atom of R22 is ~ 6 Å away from the nearest backbone Cα atoms of A4 and A5 in the model substrate (D1DIAALVVDNGS), further indicating limited potential for direct interaction, especially in distinguishing between N-terminal and mid-chain arginylation. Comparison of mammalian ATE1 isoforms in mid-chain arginylation. a Multiple sequence alignment of human and mouse ATE1 isoforms highlighting alternatively spliced regions corresponding to exons 1 and 7. Alignment was generated using COBALT and visualized with ESPript3. Red backgrounds and boxes indicate identical and similar residues, respectively. Secondary structure elements (α-helices and β-strands) are annotated based on the hATE1 isoform 2 structure (PDB: 8TZV). b Structural model of the hATE1-2/tRNA/substrate complex (PDB: 8UAU), highlighting the spatial positions of residues encoded by exon 1b (purple) and exon 7b (green), including R22, Q304, and E309. c Immunoblotting quantification of ATE1 isoform expression levels normalized to α-synuclein. Error bar represents s.d. in triplicate measurements. Relative stoichiometry of human ATE1 and αSyn is estimated to be approximately 1:40, based on the observed purification yields of 7 nmol for ATE1 and 300 nmol for αSyn per liter of bacterial culture. d Summary of arginylation sites identified by bottom-up LC-MS/MS proteomics of M1-αSyn co-expressed with various human and mouse ATE1 isoforms. Each column shows detected sites and peptide-spectrum match counts from three biological replicates (rep 1-3) for each ATE1 isoform. Bolded residues represent high-confidence arginylation sites that were detected across multiple samples. e Mapping of major arginylation sites (E20, E28, E46, E57) onto the α-helical structure of αSyn (PDB: 1XQ8). f Top-down LC-MS/MS analysis and quantification M1-αSyn co-expressed with various human and mouse ATE1 isoforms. Relative abundance (%) was calculated based on the peak area. To experimentally assess isoform-specific effects on mid-chain arginylation, we expressed experimentally verified human ATE1-1 (1A7B) and four mouse (1B1A, 1B7B, 1A7A, and 1A7B) isoforms in the bacterial arginylation system alongside the M1-αSyn substrate, in addition to hATE1-2 used previously (Fig. 4f). Bottom-up LC-MS/MS analysis was performed in three independent colonies for each isoform, and relative expression levels of ATE1 and αSyn were quantified using western blot (Fig. 5c). The resulting modification profiles showed high consistency across isoforms, with sites such as E20, E28, E46, and E57 consistently arginylated under all conditions, indicating limited variation in site preference among ATE1 isoforms (Fig. 5d). To further explore the structural context of these frequently modified sites, we mapped them onto a structural model of soluble αSyn based on PDB entry 1XQ8 (Fig. 5e). Consistent with secondary structure predictions, the major arginylation sites E20, E28, E46, and E57 are located within α-helical regions. The enrichment of arginylation sites within helical regions implies that local secondary structure may promote a conformation favorable for ATE1 recognition. Notably, similarly to hATE1-2, hATE1-1 and all tested mouse ATE1 isoforms showed higher activity toward N-terminal sites as evaluated using D2-αSyn (Fig. 5f). Together, our data suggest that, without cofactors, mid-chain arginylation events of αSyn likely stem from structural resemblance between locally structured regions of αSyn and the N-terminus of typical ATE1 substrates, rather than from defined sequence motifs. This is consistent with an opportunistic mode of mid-chain arginylation rather than a canonical, motif-driven enzymatic reaction. Discussion In this study, we successfully reconstituted the complete enzymatic cascade responsible for generating arginylated substrates in the Arg/N-degron pathway, enabling direct and quantitative examination of ATE1-mediated arginylation following upstream modifications. By incorporating ADO, NTAN1, and NTAQ1 into our previously established in-bacteria arginylation assay, we recapitulated the conversion of Nt-Cys, Asn, and Gln into arginylation-competent forms via oxidation and deamination. Quantitative intact mass spectrometry analysis revealed that deamidation of N-terminal asparagine and glutamine occurred more rapidly than cysteine oxidation, suggesting potential kinetic differences in substrate processing. Importantly, this system generates and detects arginylated products in living cells, closely mimicking the native environment where ribosomal translation is fully functional, thus overcoming limitations associated with both in vitro reconstitution and indirect degradation-based assays. The modularity of this platform enables the systematic analysis of site-specific arginylation in any given sequence and allows scalable production of arginylated proteins for biochemical and structural characterization. Despite the robust recapitulation of Nt-arginylation, our results consistently revealed low levels of mid-chain arginylation, even under conditions that solely enriched mid-chain arginylation reactions using the best-described αSyn as a substrate. Bottom-up proteomics detected low-abundance arginylation at several glutamate residues, with E20, E28, E46, and E57 being the most consistently identified sites. No conserved sequence motif was observed around the identified arginylation sites, suggesting that mid-chain arginylation does not depend on canonical sequence determinants. However, these high-confidence modification sites were predominantly located in predicted α-helical regions, indicating a potential structural preference in ATE1 substrate recognition. Of note, we roughly estimated the relative stoichiometry of human ATE1 and αSyn to be approximately 1:40, based on the observed purification yields of 7 nmol (0.4 mg) and 300 nmol (4.5 mg) per liter of bacterial culture, respectively, through dozens of purifications performed. Strategies that increase ATE1 expression or decrease αSyn yield may be explored in the future to enrich naturally arginylated αSyn for mechanistic studies. Nevertheless, these findings support the idea that mid-chain arginylation, particularly in αSyn via ATE1 alone, is a rare, opportunistic event, mechanistically distinct from the highly efficient activity at protein N-termini. In addition, we compared major human and mouse ATE1 splicing isoforms in parallel to evaluate potential differences in their ability to target mid-chain Glu/Asp sites. These isoforms differ primarily in exons 1 and 7, which surround the peptidyl-substrate binding cleft. Surprisingly, their modification profiles on αSyn across all isoforms were largely consistent, with glutamate residues such as E46 reproducibly arginylated in all cases. Notably, the relative expression levels of mouse isoforms 3 and 4 were approximately threefold lower than other isoforms, potentially due to reduced translation efficiency or protein stability (Fig. 5c). While reduced enzyme level may have affected detection sensitivity, the overall similarity in arginylation patterns suggests ATE1 isoforms share overlapping substrate preference when tested alone without additional cofactors. Future studies in mammalian cells will be essential to determine whether specific cofactors contribute to substrate or site selectivity of ATE1 isoforms. Collectively, our study provides a reconstituted platform for directly and quantitatively evaluating ATE1-mediated arginylation in bacteria. We demonstrated that N-terminal processing and arginylation are highly efficient, whereas mid-chain arginylation, particularly in αSyn via ATE1 alone, occurs infrequently and prefers sites showing characteristics of α-helix. ATE1 isoforms alone exhibited little selectivity toward mid-chain arginylation sites on αSyn. Our study establishes a foundation for the mechanistic analysis of the Arg/N-degron pathway and builds a tool for evaluating any given ATE1 substrate in a robust living cell context. Methods Protein expression and purification The model substrate DDIAA-Ub, consisting of an N-terminal D1DIAALVVDNGSGMG15 sequence fused to ubiquitin, was cloned into the pCIOX vector with an N-terminal SMT3 tag for Ulp1 cleavage and a C-terminal His-tag for purification. D2-αSyn and M1-αSyn were amplified by PCR and inserted into the same pCIOX backbone using the identical tagging strategy. Substrate variants C-Ub, N-Ub, and Q-Ub were generated by site-directed mutagenesis to substitute the first Asp of DDIAA-Ub with Cys, Asn, or Gln, respectively, using the Q5 polymerase mutagenesis protocol. Synthetic DNA fragments for ADO, NTAN1, and NTAQ1, each comprising a T7 promoter, lac operator, ribosome binding site, and coding sequence, were subsequently assembled into the respective constructs to enable co-expression with their model substrates as independent expression cassettes. Human and mouse ATE1 isoforms (Uniprot O95260-1, O95260-2, Q9Z2A5-1, Q9Z2A5-2, Q9Z2A5-3, Q9Z2A5-4) were cloned into the pETDuet vector and co-expressed with Ulp1 under separate T7 promoters. The sequences of oligonucleotides and synthetic DNA fragments used were included in Supplementary Data 2. All constructs were expressed in E. coli BL21(DE3)-RIL cells (Agilent). This RIL strain contains a rare codon plasmid encoding extra copies of argU, ileY, and leuW and overexpresses tRNAArg. Protein production was induced with 0.2 mM IPTG overnight at 16 °C in Luria broth (LB). The His-tagged proteins were purified using Ni-NTA beads (Goldbio) in 50 mM Tris-HCl (pH 7.5), 0.5 M NaCl, 5 mM β-mercaptoethanol, and 10 mM imidazole. Bound proteins were eluted with 250 mM imidazole and analyzed by SDS–PAGE. Proteins were further purified by size exclusion chromatography (HiPrep™ Sephacryl™ S-100 10/300 HR Column, Cytiva) in 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, and 5 mM DTT. Proteins were further concentrated in Millipore concentrators. Top-down proteomics analysis of arginylated proteins The LC-MS/MS platform for top-down proteomics was comprised of a Waters Acquity UPLC M-Class eluting into a Sciex ZenoTOF 7600 with the intact protein analysis setting toggled on49. Approximately 100 ng of each sample was analyzed on a Waters nanoEase M/Z Protein BEH C4 column (1.7 µm, 300 Å, 300 µm X 50 mm) held at 50 °C. Samples were separated at 10 µL/min over a 12-minute gradient from 20-40% B, where buffer A consisted of 0.1% formic acid in water and buffer B consisted of 0.1% formic acid in acetonitrile. The Optiflow TurboV microflow ion source was used for microflow separation, and samples were ionized by applying a spray voltage of 4300 V with 15 psi and 60 psi source gas 1 and 2, respectively, and 35 psi and 7 psi curtain gas and CAD gas, respectively. The front end was maintained at 250 °C for desolvation. Two experiments were used for all MS analyses. First, an MS1 TOFMS scan was performed from m/z 500–2000 with a 0.1 sec ion accumulation time, a declustering potential of 80 V, and collection of 8 time bins to sum. Then, an MRMHR MS2 method was applied to MS1 features, selecting a low, medium, and high charge state from each intact protein distribution, with Q1 resolution set to “Low” and the window centered around the most abundant isotopic peak. Target ions were accumulated for 0.25 s before being subjected to EAD fragmentation with an electron kinetic energy of 5 eV and a 5 ms reaction time, analyzed over a m/z 100-2000 MS2 scan window, collecting 8 bins to sum50. For intact protein quantitation, MS1 scans were deconvoluted using BioToolKit in SciexOS (ver. 3.1.6.44) and peak area intensities were calculated. Relative feature abundances were calculated by summing the abundance of all target features and calculating their representative percent abundance. For MS2 fragment mapping, each sample .WIFF file was converted to .MZML using MSConvert software (ver. 3.0.23073) with peak picking option selected51. Then, .MZML files were decharged and deisotoped using the FLASHDeconv deconvolution software (ver. 3.1.0), selecting “Write Detail” parameter as ‘TRUE’ and “Threads” parameter as ‘4’52. The deconvoluted MS2 spectra for each target were summed together, and EAD fragment mapping was performed in ProSightLite (v1.4)53. Settings for ProSightLite include monoisotopic precursor mass and M (neutral) fragment mass mode, fragment ppm error of 10 ppm, and “ECD” setting for EAD analyses since c/z type ions are formed similarly to ECD. Deconvoluted spectra, raw MS1 spectra, and MS2 fragment spectra were plotted in RStudio (v2023.03.0 + 386). Sample preparation for bottom-up proteomics Protein samples were processed using the S-trap micro spin column digestion protocol (PROTIFI). 20 µg of purified protein was transferred to a 1.5-mL Eppendorf tube, added to a final concentration of 5% SDS to denature the protein. The sample was incubated with TCEP (final concentration: 10 mM) for 30 min at 37 °C to reduce disulfides, and iodoacetamide (final concentration: 30 mM) for 30 min at room temperature, protected from light, to alkylate reduced thiols. 2.5 µL of 25% phosphoric acid was added to acidify the sample, followed by 165 µL of binding/wash buffer (final concentration of 100 mM TEAB in 90% methanol). The sample was added to a spin column for centrifugation at 4000 g for 30 s, washed three times with 150 µL of binding/wash buffer, and then spun for 1 min without buffer. The spin column was transferred to a clean tube, added 1 µg of trypsin in 20 µL of 50 mM ammonium bicarbonate buffer, incubated at 37 °C overnight for digestion, centrifuged for 1 min at 4000 g after adding each of the following elution buffers (100 mM TEAB, 0.2% formic acid, and 50% acetonitrile). The pooled solution was dried by SpeedVac and reconstituted in water containing 2% ACN and 0.1% FA. No peptide desalting was required before bottom-up proteomics analysis. LC-MS analysis for bottom-up proteomics A Vanquish Neo UHPLC was coupled to an Exploris 240 or Orbitrap Ascend (Thermo Scientific). Peptide samples were maintained at 7 °C on the sample tray in the LC system. Separation of peptides was carried out on an Easy-Spray™ PepMap™ Neo nano-column (2 µm, C18, 75 µm X 150 mm) at room temperature with a mobile phase consisting of a linear gradient of A (0.1%FA in H2O) and B (acetonitrile containing 0.1% FA) under the following conditions: 0 → 90 → 110 → 110.5 → 116 min, 0% → 28% → 36% → 98% → 98% B. The flow rate was 300 nL/min. The injection volume is 5 µL. The voltage applied to the nano-LC electrospray ionization source was 1.9 kV. The temperature of the ion transfer tube (ITC) was set at 275 °C. Spectra were collected in a data-dependent acquisition mode with an MS1 scan range of m/z 350–2000 in the Orbitrap. For Exploris, the top 20 most intense peaks from a single high-resolution (120,000) full MS spectrum of parent ions were fragmented for MS2 spectra. For Ascend, the time of a scan cycle is 3 seconds. Parent ions assigned as peptides in charge states +2-5 with intensity higher than 1E4 were included for fragmentation. HCD-induced fragmentation (MS2) scans were recorded in an Orbitrap (scan from m/z 110–1500). Dynamic exclusion was set as a repeat count of 1 within an exclusion time of 20 seconds. All other parameters were left as default values. Search parameters of bottom-up proteomics using Byonic software. Each data file (in “.raw” format) was generated by the instrument (Xcalibur software), and searched using the byonic software v4.5.2 (Protein Metrics). Peptides were required to have at least a tryptic C-terminus, and up to two missed cleavages were allowed in the database search. The mass tolerances of precursor and product ions were set to 10 and 20 ppm, respectively. Carbamidomethylation (+57.021464 Da) on Cysteine residues was allowed for 2 per peptide as a “common” variable modification in the byonic parameter file. Oxidation (+15.99492 Da) on Methionine residues was allowed for 2 per peptide as a “common” variable modification. Total “common” modifications were limited to up to 4 per peptide. Arginylation (+156.10111 Da) on peptide N-terminal, Asp, and Glu residues, arginylation after deamidation (+157.08513 Da) on Asn and Gln residues, arginylation after di-oxidation (+188.09094 Da) or tri-oxidation (+204.08585 Da) on Cysteine residues was allowed for 1 per peptide as a “rare” modification. Total rare modifications were limited to 1 per peptide. The false-positive rate was set at 1% or lower. Immunoblotting Bacterial cultures were harvested by centrifugation at 10,000 g for 1 min and resuspended in PBS. The cells were lysed in RIPA buffer (Sigma-Aldrich, R0278-500ML) supplemented with protease inhibitor (Thermo Scientific, A32955), followed by sonication on ice for 5 min. Lysates were clarified by centrifugation at 13,000 rpm for 20 min at 4 °C. The supernatant was mixed with sodium dodecyl sulfate (SDS) loading buffer containing β-mercaptoethanol and boiled at 95 °C for 10 min. Samples were separated by SDS-PAGE and transferred onto polyvinylidene difluoride (PVDF) membranes. Membranes were probed with primary and secondary antibodies, and signals were developed using ECL substrate (VL001-200, R&D Systems). Blots were visualized using a Bio-Rad stain-free imaging system. Uncropped gels were included in the Supplementary Fig 15. Antibodies The primary antibodies used in this study are as follows: Rabbit monoclonal anti-ATE1 (Abcam, ab199423, 1:1000), mouse monoclonal anti-α-synuclein (Santa Cruz Biotechnology, sc-12767, 1:200). The following secondary antibodies were used: HRP-conjugated goat anti-rabbit IgG (Jackson Immuno Research, 111-035-003, 1:10,000), HRP-conjugated horse anti-mouse IgG (Fisher Scientific, 50-195-914, 1:5000). Statistics and reproducibility No statistical methods were used to predetermine sample size. No data were excluded from the analyses. The experiments were not randomized, and investigators were not blinded to allocation during experiments and outcome assessment. Unless otherwise stated, all experiments were performed with at least two to three independent biological replicates. Sample sizes, number of replicates, and any quantification methods are specified in the corresponding figure legends and relevant Methods sections. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Reporting Summary Transparent Peer Review file Acknowledgements We thank Dr. Anna Kashia at the University of Pennsylvania for sharing mouse ATE1 expression constructs. We are grateful to Dr. Ebmeier at the University of Colorado Boulder Mass Spectrometry Facility for help with initial pilot experiments. Author contributions X.L., R.S., D.L., S.B., and T.A. performed experiments, together with B.G., Z.L., and Y.Z. analyzed the data. X.L., R.S., Z.L., and Y.Z. wrote the manuscript with input from all authors. Peer review Peer review information Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editors: Laura Rodríguez Pérez. A peer review file is available. Funding This work was supported by grants from NIH R35 GM150678 to Y.Z., NIH R21 CA292191 to Z.L., NIH R01 HL177113 to Z.L. and B.A.G., Research Education Component (REC) through an NIA grant P30AG066444 to Z.L., Case Comprehensive Cancer Center (P30CA043703 to Y.Z. and T32CA059366 to T.A.), and summer research scholarships provided by CWRU SOURCE. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH. Data availability The MS data have been deposited in the MassIVE database (MSV000098787, MSV000098788, MSV000098790, Supplementary Table 3). Source data are provided with this paper. Other data are available from the corresponding author upon request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Xin Lan, Richard M. Searfoss. Contributor Information Zongtao Lin, Email: tom.z.lin@rutgers.edu. Yi Zhang, Email: yi.zhang26@case.edu. Supplementary information The online version contains supplementary material available at 10.1038/s42003-026-10275-z. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Reporting Summary Transparent Peer Review file Data Availability Statement The MS data have been deposited in the MassIVE database (MSV000098787, MSV000098788, MSV000098790, Supplementary Table 3). Source data are provided with this paper. Other data are available from the corresponding author upon request.