An Essential Role of N-Terminal Arginylation in Cardiovascular Development

See allHide authors and affiliations

Science  05 Jul 2002:
Vol. 297, Issue 5578, pp. 96-99
DOI: 10.1126/science.1069531


The enzymatic conjugation of arginine to the N-termini of proteins is a part of the ubiquitin-dependent N-end rule pathway of protein degradation. In mammals, three N-terminal residues—aspartate, glutamate, and cysteine—are substrates for arginylation. The mouseATE1 gene encodes a family of Arg-tRNA-protein transferases (R-transferases) that mediate N-terminal arginylation. We constructed ATE1-lacking mouse strains and found thatATE1 −/− embryos die with defects in heart development and in angiogenic remodeling of the early vascular plexus. Through biochemical analyses, we show that N-terminal cysteine, in contrast to N-terminal aspartate and glutamate, is oxidized before its arginylation by R-transferase, suggesting that the arginylation branch of the N-end rule pathway functions as an oxygen sensor.

Substrates of the ubiquitin (Ub)–dependent N-end rule pathway include proteins with destabilizing N-terminal residues (1–4). A set of amino acids that are destabilizing in a given cell yields a rule, called the N-end rule, that relates the in vivo half-life of a protein to the identity of its N-terminal residue (1–3, 5–8). The N-end rule has a hierarchic structure. Specifically, N-terminal Asn and Gln are tertiary destabilizing residues in that they function through their deamidation, by N-terminal amidohydrolases (7), to yield the secondary destabilizing residues Asp and Glu, whose activity requires their conjugation, by ATE1-encoded Arg-tRNA-protein transferases (R-transferases) (5), to Arg, one of the primary destabilizing residues. The latter are recognized by the Ub ligases (E3 enzymes) of the N-end rule pathway (Fig. 1A) (3, 4, 9).

Figure 1

The N-end rule pathway and phenotypes ofATE1 −/− embryos. (A) N-terminal residues are indicated by single-letter abbreviations for amino acids (22). Yellow ovals denote the rest of a protein substrate. The area highlighted in green describes the understanding gained in the present work (see text). C* denotes an oxidized Cys residue. Type 1 and 2 primary destabilizing residues are recognized by functionally overlapping E3s that include UBR1 (E3α) and UBR2 (3, 9). N-terminal Ala, Ser, and Thr are recognized by an unidentified E3. (B to E) Whole mounts of +/+ (B) andATE1 −/− (C to E) E15.5 embryos. (Fand G) Left ventricular wall in E14.5 +/+ (F) andATE1 −/− (G) embryos. (H toL) Transverse sections of hematoxylin and eosin–stained +/+ (H and K) andATE1 −/− (I, J, and L) E14.5 hearts. TV, tricuspid valve; MV, mitral valve; RV, right ventricle; LV, left ventricle; Ao, aorta; PA, pulmonary artery; RA, right atrium; LA, left atrium; CZ, compact zone; t, trabeculae. Scale bars, 2 mm (B to E), 10 μm (F and G), 200 μm (H to L).

In mammals, the set of destabilizing residues that function through their arginylation includes not only Asp and Glu but also Cys, which is a stabilizing (nonarginylated) residue in the yeastSaccharomyces cerevisiae (5, 10, 11). ATE1-1 and ATE1-2, the isoforms of mammalian R-transferase, are produced through alternative splicing of ATE1 pre-mRNA and have the same specificity as the yeast R-transferase: They arginylate N-terminal Asp or Glu but not Cys (5). This raises the question of how N-terminal Cys is arginylated in mammalian cells. To address this issue and the physiological functions of arginylation, we constructedATE1 −/− mouse strains (12).

Whereas ATE1 +/− mice were apparently normal, the ATE1 −/− genotype conferred embryonic lethality (12). The ATE1 allele was marked with NLS-β-galactosidase (βgal) (12). During embryonic day (E) 9.5 to 12.5, the expression of βgal was high in the neural tube and other specific (often sharply delineated) regions of developing embryo (12). ATE1 −/−embryos were pale and had thinner blood vessels and frequent edemas of the skin (Fig. 1, B and C; Fig. 2, A and B) (12). Hemorrhages were a consistent feature ofATE1 −/− embryos and were the likely proximal cause of their death (Fig. 1, D and E). Of 22ATE1 −/− hearts (E13.5 to E15.5) examined, ∼85% had a ventricular septal defect (VSD) (Fig. 1, I and J). The atria of ATE1 −/− hearts were thin walled, with sparse trabeculae and a large atrial septal defect (ASD) (Fig. 1J). About 90% of examined ATE1 −/−hearts exhibited hypoplasia of both right and left ventricular myocardium (Fig. 1, G, I, and J). The compact zone of left ventricular myocardium was two or three cells thick, in contrast to 7 to 10 cells in +/+ embryos (Fig. 1, F and G). Whereas the aorta and pulmonary artery were completely separated in E13.5 +/+ hearts (Fig. 1K), ∼70% of the examined ATE1 −/− hearts (E13.5 to E15.5) had persistent truncus arteriosus, with the common root of aorta and pulmonary artery straddling a large VSD (Fig. 1L).

Figure 2

Perturbation of angiogenic remodeling but not of vasculogenesis in ATE1 −/− embryos. (A) +/+ and (B) ATE1 −/−E13.5 embryos with yolk sacs. Black arrowheads, major artery; white arrowheads, major vein; arrow, a local hemorrhage. (C andD) Dorsal views of the trunks of PECAM-1–stained E9.5 +/+ (C) and ATE1 −/− (D) embryos. (E and F) Similarly located areas (rectangles in A and B) of +/+ (E) and ATE1 −/− (F) PECAM-1–stained E13.5 yolk sacs. (G andH) E9.5 +/+ (G) and ATE1 −/− (H) embryos stained with antibody to PECAM-1. Scale bars, 1 mm (A), 250 μm (C and H), 100 μm (E).

Vasculogenesis, the de novo formation of uniformly sized blood vessels that yields the primary capillary plexus, was apparently normal in E9.5ATE1 −/− embryos (Fig. 2, G and H) (12). In contrast, the process of angiogenic remodeling—such as, for example, the formation of vessels that normally sprout from intersegmental artery and cross the dorsal midline—was suppressed in E9.5 ATE1 −/−embryos (Fig. 2, C and D) (12). The vessels in E13.5ATE1 −/− sacs often terminated prematurely (Fig. 2F, arrow), and many small vessels remained as a honeycomb-like meshwork of the primary-plexus capillaries (Fig. 2F, arrowhead), indicating a perturbation of angiogenic remodeling inATE1 −/− embryos.

To examine the N-end rule pathway in ATE1 −/−cells, we transfected +/+ and ATE1 −/−embryonic fibroblast (EF) cell lines with plasmids encoding X-nsP4f, a set of 69-kD Flag-tagged Sindbis virus RNA polymerase proteins bearing different N-terminal residues (3,12) (Fig. 1A). The R-transferase substrates Asp-nsP4fand Glu-nsP4f were short-lived in +/+ cells but were completely stabilized in ATE1 −/− cells (Fig. 3, A and B). Strikingly, although R-transferases cannot arginylate the N-terminal Cys (5), the normally short-lived Cys-nsP4f also became long-lived inATE1 −/− cells (Fig. 3, A and B), indicating that N-terminal Cys, perhaps as a result of a preceding chemical modification, was also a substrate of ATE1-encoded R-transferases. To measure arginylation directly, we added purified Ub-X-βgal proteins (X = Met, Arg, Glu, or Cys) or human α-lactalbumin (bearing N-terminal Glu) to +/+ andATE1 −/− EF extracts supplemented with ATP, total tRNA, and aminoacyl-tRNA synthetases (Fig. 3C) (10,12). Ub-X-βgals are rapidly deubiquitylated in vivo and in cell-free extracts, yielding X-βgals (10). Asp-βgal, Glu-βgal, and α-lactalbumin were arginylated in +/+ EF extracts, in contrast to Arg-βgal (bearing a primary destabilizing residue) and Met-βgal (bearing a stabilizing residue) (Fig. 3C). No arginylation of Asp-βgal, Glu-βgal, and α-lactalbumin was detected inATE1 −/− EF extracts (Fig. 3C) or in extracts from ATE1 −/− embryos (fig. S5A), consistent with the in vivo results (Fig. 3, A and B). Surprisingly, the N-terminal Cys of Cys-βgal was not arginylated in +/+ orATE1 −/− extracts (Fig. 3C), suggesting that the demonstrated ATE1 dependence of the in vivo degradation of Cys-bearing N-end rule substrates (Fig. 3, A and B) may involve a modification of N-terminal Cys before its arginylation. A protease called Asp-N cleaves peptide bonds N-terminal to the Asp and CysO3 residues (13), and aspartate aminotransferases use both Asp and oxidized Cys as substrates (14). Thus, either the Cys sulfinic acid residue (CysO2) or the cysteic acid residue (CysO3) may be sufficiently close in structure to Asp (Fig. 3E) to serve as a substrate of R-transferases.

Figure 3

Stabilization of N-end rule substrates inATE1 −/− cells, and oxidation-arginylation of N-terminal cysteine. (A) +/+ andATE1 −/− EF cells were transfected with plasmids expressingfDHFRh-UbR48-X-nsP4f, whose cleavage yielded thefDHFRh-UbR48 reference (DHFR) and X-nsP4f reporter (X = Met, Arg, Asp, Glu, or Cys) (3, 12). Cells were pulse-labeled for 10 min with [35S]methionine and chased for 1 and 2 hours (12). (B) Quantitation of results in (A). For each time point, the ratio of 35S in X-nsP4f(X = Arg, Asp, Glu, or Cys) to 35S in thefDHFRh-UbR48 reference, at the same time point, was plotted as the percentage of this ratio relative to that for Met-nsP4f at time 0 (3, 23). Open and closed symbols: +/+ andATE1 −/− cells, respectively. Triangles, Met-nsP4f; circles, Arg-nsP4f; squares, Asp-nsP4f; inverted triangles, Glu-nsP4f; diamonds, Cys-nsP4f. (C) Cell-free assay for R-transferase using [3H]Arg, S105 extracts from +/+ andATE1 −/− EFs, and unlabeled X-βgals (X = Cys, Asp, Glu, Met, or Arg) or α-lactalbumin (12). Asterisks indicate arginylated endogenous proteins in the extracts. (D) Determination, through Edman degradation, of N-terminal sequences (22) of RGS4-Flag-His6isolated from mouse L cells. Calk, alkylated Cys residue. (E) Chemical formulas of N-terminal Cys, Cys sulfinic acid, cysteic acid, and Asp residues.

One prediction of the Cys-oxidation hypothesis was that the arginylated Cys residue should exist as CysO2 or CysO3. This was verified and confirmed with mouse RGS4, a guanosine triphosphatase–activating protein that bears N-terminal Cys and is arginylated and degraded by the N-end rule pathway in rabbit reticulocyte extract (11). RGS4-His6 was expressed in mouse L cells, purified, treated with iodoacetamide to alkylate Cys residues (thereby making them identifiable by the sequencing procedure used), and N-terminally sequenced by Edman degradation. Both arginylated and unarginylated RGS4s were detected, the former being a major species (Fig. 3D). Remarkably, whereas the expected Cys at position 12 of arginylated RGS4 could be identified as alkylated Cys, the expected (alkylated) Cys at position 2 (position 1 in the unarginylated RGS4) could not be identified by the Edman procedure (Fig. 3D), indicating that N-terminal Cys of RGS4 had been modified before alkylation. Alkylated RGS4 was cleaved with cyanogen bromide (CNBr); HPLC fractionation and online mass spectrometric sequencing of CNBr-produced peptides identified residue 2 of RGS4 as cysteic acid (CysO3) (12).

Another prediction of the Cys-oxidation hypothesis was that the yeast R-transferase should be able to rescue the destabilizing activity of Cys in mouse ATE1 −/− cells, owing to the presence of Cys-oxidation activity in these cells. This was verified using pulse-chase assays with ATE1 −/− EFs expressing X-nsP4f and either S. cerevisiae ATE1 or enzymatically impaired ATE1C23A mutant (12,15). Both Asp-nsP4f and Cys-nsP4f, which were long-lived in mouse ATE1 −/− EFs (Fig. 3, A and B), became short-lived in the presence of yeast ATE1 (12). The rescue by yeast R-transferase required its enzymatic activity, because ATE1C23A, a catalytically impaired mutant (12, 15), had a significantly weaker effect (12). These findings, in conjunction with stoichiometric oxidation of N-terminal Cys in mouse cells (Fig. 3D) (12), indicate that the oxidation is an enzymatic (rather than uncatalyzed) reaction, because the intracellular solvent conditions, including redox potential, are likely to be similar in mammalian and yeast cells.

We have identified a physiological function—cardiovascular development—for the posttranslational conjugation of Arg to N-termini of proteins, a reaction first described 40 years ago (16). We have also shown that the N-terminal Cys undergoes two (rather than one) covalent modifications—oxidation and arginylation—by the N-end rule pathway (Fig. 1A). Met-aminopeptidases cleave off the N-terminal Met of a newly formed protein if a second residue is small enough. Among the arginylatable residues (Fig. 1A), only Cys satisfies this condition (2, 3, 10). Fumagillin and related suppressors of angiogenesis have been shown to inhibit the Met-aminopeptidase MetAP2 (17, 18). The effect of these antiangiogenic reagents may stem from inhibition of the N-terminal Met-Cys cleavage in a normally short-lived regulator of angiogenesis that is targeted by the N-end rule pathway through its N-terminal Cys residue.

The oxidation (and subsequent arginylation) of N-terminal Cys may compete with its other known modifications, including acetylation and palmitoylation. N-end rule substrates with the arginylation-dependent destabilizing N-terminal residues (Asn, Gln, Asp, Glu, and Cys) (Fig. 1A) can also be produced through cleavages anywhere in a protein's polypeptide chain. For example, the conditional cleavage of a subunit of the mammalian cohesin complex at the metaphase-anaphase transition is predicted to produce a putative N-end rule substrate whose degradation would require N-terminal arginylation (8, 19).

HIF1α, a subunit of hypoxia-inducible factor 1 (HIF1) that functions as a key regulator of angiogenesis, is a conditionally short-lived protein. The degron of HIF1α, recognized by a distinct Ub-dependent proteolytic pathway, is activated through the oxygen-dependent hydroxylation of a specific Pro residue (20, 21). By analogy to prolyl-4-hydroxylases that regulate the degron of HIF1α, the currently unknown enzyme that oxidizes N-terminal Cys may also function as an oxygen sensor. If so, the formation and maintenance of the cardiovascular system may involve a battery of distinct, conditionally short-lived regulators such as HIF1 and the currently unknown substrate of the N-end rule pathway that bears N-terminal Cys.

  • * These authors contributed equally to this work.

  • Present address: IGEN International Inc., 16020 Industrial Drive, Gaithersburg, MD 20877, USA.

  • To whom correspondence should be addressed. E-mail: avarsh{at}


View Abstract

Navigate This Article