Oxidation of N-terminal cysteine is required for its arginylation

We wished to determine whether the presence of CysO₃H (instead of Cys) at position 2 of the RGS4 protein arginylated in vivo¹² reflected the requirement for oxidation of Cys before its arginylation, as distinguished from oxidation of Cys after its arginylation. Three eight-residue XHGSGAWL peptides (where X represents Asp, Cys or CysO₃H) were incubated with mouse ATE1-1 R-transferase¹¹ in the presence of ATP, S. cerevisiae Arg-tRNA synthetase and tRNAs. Both the Asp-peptide and the CysO₃H-peptide were efficiently arginylated, whereas the Cys-peptide was not arginylated, as determined by capillary electrophoresis (compare Fig. 1c–f with Fig. 1g, h), and confirmed, with regard to the identities of products, by matrixassisted laser desorption ionization–timeofflight (MALDI–TOF) MS (Fig. 1i, j). The same results were obtained with S. cerevisiae ATE1 R-transferase³³, a strong 'sequelogue'³⁴ of mammalian R-transferases¹¹ (Supplementary Fig. S1). We conclude that the oxidation of N-terminal Cys is essential for its arginylation (Fig. 1c–j).

Increased levels of RGS proteins in ATE1^-/- embryos

RGS4 is a GTPase-activating protein (GAP) for specific G subunits of G proteins and is a member of the family of RGS proteins that downregulate signalling by G proteins^{30, 31, 32, 35, 36, 37}. Our earlier work showed that RGS4 (which begins, as a nascent protein, with Met-Cys) was a substrate of the N-end rule pathway in reticulocyte extract³⁸. The levels of endogenous RGS4 were also increased by proteasome inhibitors³⁷ but the pathway of RGS4 degradation in vivo remained to be identified.

We compared the levels of endogenous RGS4 between 12.5-day-old (E12.5) wild-type mouse embryos and ATE1^-/- embryos, which lacked R-transferases and therefore lacked arginylation¹². The level of RGS4 in E12.5 ATE1^-/- embryos was strikingly higher than in E12.5 wild-type embryos (Fig. 2a). Similar results were obtained in pairwise comparisons of RGS4 in specific tissues from E14.5 wild-type and ATE1^-/- embryos (Fig. 2b). Two other members of the RGS family, RGS5 and RGS16, also bear N-terminal Cys. Experiments analogous to those with RGS4 revealed strong increases in RGS5 and RGS16 levels in ATE1^-/- embryos (Fig. 2c–f). These protein patterns (Fig. 2a–f) were not caused by increased levels of the corresponding messenger RNAs, as indicated by both complementary DNA microarray comparisons (J.S., R.G.H., S. Choi., M. Simon and A.V., unpublished observations) and northern analyses (Supplementary Fig. S2).

Figure 2: Strongly increased levels of RGS4, RGS5 and RGS16 proteins in ATE1^-/- embryos.

a, Lanes 1 and 2, equal amounts of total protein in extracts from wild-type (+ / + ) and ATE1^-/- E12.5 embryos were fractionated by 12% SDS–PAGE, followed by immunoblotting with anti-RGS4 antibody. b, As in a but with extracts from the indicated tissues of wild-type and ATE1^-/- E14.5 embryos. c, As in a but with anti-RGS5 antibody and E14.5 embryos. d, As in b but with anti-RGS5 antibody. e, As in a but with anti-RGS16 antibody. f, As in b but with anti-RGS16 antibody. g, 3T3^toffRGS4_fh cells expressing RGS4_fh were grown in the presence of either ambient air (lane 1) or low (0.5%) oxygen concentration (lane 2). Equal amounts of total protein in extracts were subjected to immunoblotting with antibody against RGS4. h, Lanes 1–3, equal amounts of total protein in extracts from wild-type (lane 1), [UBR1^-/-UBR2^-/-] (lane 2) and UBR1/2^dnR2 (lane 3) cell lines were subjected to immunoblotting with antibody against RGS16. Lanes 4–6, as in lanes 1–3, respectively, but immunoblotting with antibody against serine racemase (SRR). Lanes 7, 8, as in lanes 4–6 but with extracts from wild-type (lane 7) and ATE1^-/- (lane 8) E12.5 embryos. The bands of 23K RGS4 (apparent molecular mass 28 kDa), 20K RGS5 (apparent molecular mass 21 kDa), 23K RGS16 (apparent molecular mass 25 kDa) and 37K SRR are indicated in a–h.

High resolution image and legend (44K)

To address the above issues differently, we employed mouse cells termed UBR1/2^dnR2. These [UBR1^-/-UBR2^-/-] cells, lacking two of several E3 ubiquitin ligases of the N-end rule pathway^{4, 14}, stably express UBR2¹⁰⁴¹, the N-terminal fragment of mouse UBR2 that functions as a dominant-negative inhibitor of the (residual) N-end rule pathway in parental [UBR1^-/-UBR2^-/-] cells (J.S., Z. Xia, Y. T. Kwon and A.V., unpublished observations). The level of endogenous RGS16 was negligible in wild-type cells, barely detectable in [UBR1^-/-UBR2^-/-] cells and much higher in UBR1/2^dnR2 cells (Fig. 2h, lanes 1–3). These results indicated yet again, in a setting independent from that of ATE1^-/- cells, that RGS16 was a substrate of the N-end rule pathway in vivo.

Decreasing NO in vivo stabilizes RGS4, RGS5 and RGS16

To examine the possibility that the oxidation of N-terminal Cys might involve NO, we constructed a mouse cell line, termed 3T3^toffRGS4_fh, that expressed RGS4-Flag-His₆ (RGS4_fh) from a doxycycline-repressible promoter. Treatment of 3T3^toffRGS4_fh cells with N^G-monomethyl-l-arginine (LMMA), an inhibitor of NO synthases (NOSs), markedly increased the level of RGS4 in vivo (Fig. 3a, compare lanes 2 and 4). When 2-(4-carboxyphenyl)-4,4,5,5- tetramethylimidazoline-1-oxyl-3-oxide (cPTIO), a cell-penetrating NO scavenger, was used to decrease NO levels, the increase in RGS4 was even more striking (Fig. 3a, compare lanes 2 and 3, and lanes 7 and 8). In the experiments of Fig. 3a, lanes 1–4 and 9–11, anti-RGS4 antibody detected two bands: the upper band's position was that expected for RGS4_fh; the lower was apparently a proteolytic fragment of RGS4_fh, because changes in its levels paralleled those of full-length RGS4_fh. The lower band was not observed in an otherwise identical but independent experiment of Fig. 3a, lanes 7 and 8. In addition, the same anti-RGS4 antibody did not detect RGS4 in parental 3T3^toff cells (Fig. 3a, lanes 5 and 6), and detected one RGS4 band in ATE1^-/- embryos (Fig. 2a).

Figure 3: Decreasing NO concentration in vivo stabilizes RGS4 and RGS16.

a, 3T3^toffRGS4_fh cells, expressing RGS4_fh, were either untreated or treated with compounds that decrease the levels of intracellular NO, followed by immunoblotting with antibody against RGS4. Lanes 7–8 and 9–11 show two sets of experiments independent of those in lanes 1–4, with independently grown 3T3^toffRGS4_fh cells. The concentrations of N3411 used in lanes 10 and 11 were 0.5 and 1 mM, respectively. Also indicated is the band of 37-kDa glyceraldehyde-3-phosphate dehydrogenase (GAPDH), a loading control, detected on the same membrane with antibody against GAPDH. b, Determination, through Edman degradation, of the N-terminal sequence of RGS4_fh isolated from 3T3^toffRGS4_fh cells that had been treated as described in the panel and in the text. C^alk, alkylated Cys residue. The asterisk after the N-terminal, post-translationally conjugated Arg (R) residue indicates a 'sequenceable' but unidentified residue, at the position of oxidized Cys residue in RGS4, in contrast to alkylated (identifiable) Cys (C^alk) at position 12. c, Lanes 1 and 2, immunoblotting, with antibody against RGS16, of extracts from NIH-3T3 cells that were either untreated (lane 1) or treated with cPTIO (lane 2). Lanes 3 and 4, as in lanes 1 and 2 but with ATE1^-/- EF cells¹² lacking arginylation. d, Lane 1, ¹⁴C-labelled protein markers, of molecular masses 66, 45 and 30 kDa (asterisks). Lanes 2–4, RGS4_fh-expressing 3T3^toffRGS4_fh cells were labelled for 10 min with ³⁵S-methionine/cysteine and chased for 1 and 2 h, followed by immunoprecipitation of extracts with antibody against RGS4, then by SDS–PAGE and autoradiography. Lanes 5–7, as in lanes 2–4 but with cPTIO-treated 3T3^toffRGS4_fh cells. Lanes 8 and 9, as in lanes 2 and 5, respectively, but with 3T3^toff (RGS4_fh-lacking) cells.

High resolution image and legend (42K)

Reverse-transcriptase-mediated polymerase chain reaction showed that 3T3^toffRGS4_fh cells expressed NOS1 (neuronal NOS) mRNA but little if any NOS2 (inducible NOS) and NOS3 (endothelial NOS) mRNAs (Supplementary Fig. S3). Given these findings, we also used an NOS inhibitor (N3411) that is highly selective for NOS1. In agreement with other NO results (Fig. 3a, lanes 1–8), and despite the near-confinement of inhibition by N3411 to NOS1, this inhibitor substantially increased the levels of RGS4 (Fig. 3a, lanes 9–11). We also examined whether RGS4 was elevated in NOS1^-/- adult mice²⁴. Despite the presence of NOS2 and NOS3 in these mice, RGS4 was strongly increased in the NOS1^-/- lung, relative to the wild-type lung, and to a smaller extent in other tissues as well (Supplementary Fig. S4).

To determine whether a decrease in NO concentration could affect the arginylation of RGS4 in vivo, we used N-terminal (Edman) sequencing. RGS4_fh purified from control 3T3^toffRGS4_fh cells was completely or nearly completely (more than 90%) arginylated in vivo (Fig. 3b). In contrast, most RGS4_fh from cells that had been treated with cPTIO could not be sequenced by Edman degradation (Fig. 3b). The identity of the blocking modification of RGS4_fh remains to be determined. Thus, RGS4_fh from cells with decreased NO was largely unarginylated (Fig. 3b) and therefore not a target of the N-end rule pathway, in agreement with other results (Fig. 3a). In a pulse–chase assay, a negligible amount of ³⁵S-labelled RGS4_fh was present at the end of a 10-min pulse in untreated 3T3^toffRGS4_fh cells (Fig. 3d, lanes 2–4). In contrast, a strongly labelled band of RGS4_fh was observed with cPTIO-treated cells at the end of a 10-min pulse (Fig. 3d, compare lanes 2 and 5). Although stabilized in cPTIO-treated cells, RGS4_fh remained partly unstable even in the presence of cPTIO (Fig. 3d, lanes 5–7), which is consistent with the incomplete elimination of NO by cPTIO (refs 21, 22 and references therein).

RGS16 (but not RGS4) was naturally expressed in NIH-3T3 cells and in EF cell lines. Anti-RGS16 antibody detected trace amounts of RGS16 in untreated 3T3 cells but a much higher level of RGS16 after treatment with cPTIO (Fig. 3c, compare lanes 1 and 2). Crucially, this effect of cPTIO was nearly absent when the same experiment was performed with ATE1^-/- EF cells¹², which lacked arginylation (Fig. 3c, compare lanes 3 and 4), yielding independent evidence that the degradation of RGS16 requires both NO and arginylation. We also found that the level of RGS4 was much higher in 3T3^toffRGS4_fh cells grown under conditions of low (0.5%) oxygen concentration (Fig. 2g; see Discussion).

NO-dependent arginylation of RGS4 in vitro

Cys-RGS4, and also the otherwise identical Asp-RGS4 and Val-RGS4 (Fig. 1a), were produced in Escherichia coli with the ubiquitin fusion technique². X-RGS4 proteins were incubated with mouse ATE1-1 R-transferase in the presence of ATP, [³H]arginine, tRNA and other components of the analogous assay with eight-residue peptides (Fig. 1c–j), except that arginylation of X-RGS4 was detected by SDS–polyacrylamide gel electrophoresis (SDS–PAGE) and fluorography. In addition to untreated controls, X-RGS4 proteins were also preincubated with one of two NO donors, either diglutathionyl-dinitroso-iron (DNIC-(GSH)₂, a physiologically relevant NO carrier^{39, 40}) or diethylenetriamine/nitric oxide adduct (DETA-NO). The results indicated an almost complete dependence of the arginylation of Cys-RGS4 in vitro on its previous exposure to a donor of NO (Fig. 4, compare lane 4 with lanes 5 and 6). (Dissolved oxygen and other gases were at levels that are normally present in buffers.) In contrast, Asp-RGS4 was efficiently arginylated irrespective of pretreatment with NO, and Val-RGS4 was not arginylated (Fig. 4). Thus, the NO dependence of in vivo arginylation of proteins bearing N-terminal Cys (Figs 2 and 3) can be reconstituted in an in vitro system (Fig. 4).

Figure 4: In vitro reconstitution of nitric oxide-dependent arginylation of RGS4.

Purified Asp-RGS4 (lanes 1–3), Cys-RGS4 (lanes 4–6) and Val-RGS4 (lanes 7–9) were incubated with [³H]arginine under conditions of the arginylation assay described in the text, followed by SDS–PAGE and fluorography. Lanes 1, 4 and 7, no pretreatment of X-RGS4 proteins. Lanes 2, 5 and 8, pretreatment of X-RGS4 proteins with DETA-NO. Lanes 3, 6 and 9, pretreatment of X-RGS4 proteins with DNIC-[GSH]₂. The asterisk denotes a minor ³H-labelled species whose apparent molecular mass and relative levels indicate that it might be a dimer of X-RGS4.

High resolution image and legend (19K)

Sequence motif of NO-dependent N-degron

The advances described above ((Figs 1–4) accounted for the following, previously unexplained, observation: the conversion of the position 3 residue of RGS4 (position 2 after Met removal) from basic Lys to uncharged Ser stabilized the resulting RGS4_K3S against degradation in a reticulocyte extract³⁸. In contrast, RGS4_K3R, in which Lys 2 was replaced by Arg, remained short-lived³⁸. Moreover, in contrast to wild-type RGS4, which was arginylated and bore CysO₃H at position 2 (Fig. 3d and ref. 12), RGS4_K3S that was transiently expressed in mouse cells was found to have a blocked N terminus (I. Davydov and A.V., unpublished observations). These results can now be explained, because S-nitrosylation of internal (non-N-terminal) Cys by NO in polypeptides depends, in particular, on the presence of a Cys-proximal basic residue, which facilitates the abstraction of H⁺ from the cysteine's thiol group^{21, 22}. In agreement with this explanation, the second residues of MetAP-processed RGS4, RGS5 and RGS16 are Lys, Lys and Arg, respectively. In a different examination of the consensus motif (N-terminal Cys-(basic residue)), we used an antibody against serine racemase²⁰, a protein with a Cys-Ala N-terminal sequence, to examine whether the levels of racemase were increased in ATE1^-/- embryos or UBR1/2^dnR2 cells. In contrast to results with RGS4, RGS5 and RGS16, no significant changes in racemase levels were observed (Fig. 2h, lanes 4–8), as would be expected from the requirement for a basic residue at position 2 for efficacious NO-mediated oxidation of N-terminal Cys.

Discussion

We have shown here that the oxidation of N-terminal Cys in a polypeptide is essential for the arginylation of Cys by ATE1-encoded R-transferases (Fig. 1c–j). We also discovered that the arginylation branch of the N-end rule pathway (Fig. 1a) is a sensor of nitric oxide (NO) that functions through its ability to destroy specific regulatory proteins bearing N-terminal Cys, at rates controlled by NO and apparently by oxygen as well ((Figs 1–4). The first examples of these regulators, RGS4, RGS5 and RGS16, are also the first physiological substrates of mammalian N-end rule pathway. These proteins downregulate specific subunits of G proteins by increasing their GTPase activity^{30, 31, 32, 35, 36, 37}. Through the conditional destruction of RGS4, RGS5 and RGS16, the N-end rule pathway is therefore involved in the regulation of signalling by G-protein-coupled receptors.

The above 'unification' of a ubiquitin-dependent proteolytic pathway and NO signalling opens up new vistas for understanding both. Most of the previously known regulation by NO was based on changes in the functional (for example enzymatic) activity of NO-modified proteins^{19, 20, 21, 22, 23, 27, 28, 29}. In contrast, a Cys-containing N-degron of a protein makes possible the NO-mediated control of circuits that contain this protein, through its NO-dependent degradation by the N-end rule pathway. The observed stabilization of RGS4 in cells grown at low oxygen concentration (Fig. 2g) suggests a possible involvement of oxygen or its derivatives in this NO-dependent regulation. That would also be expected from the NO results alone (Figs 3 and 4), given multiple links between the chemistries of NO and oxygen in vivo^{19, 20, 21, 22, 23}. Our assay in vitro for NO-dependent arginylation of N-terminal Cys (Fig. 4) should eventually yield a detailed understanding of chemical transformations that result in oxidation of this uniquely positioned Cys residue.

The pathway of control by NO (Fig. 1a) targets proteins that bear N-terminal Cys followed by a basic residue. This motif is present in about 30 proteins encoded by the mouse (and human) genome, including RGS4, RGS5 and RGS16. More than half of non-RGS proteins in this set have entirely unknown functions, and the rest are barely characterized. The N-terminal Cys residues of RGS4, RGS5 and RGS16 can also be modified through palmitoylation³⁶. The two modifications are expected to be mutually exclusive and also act in opposite ways, in that palmitoylation increases the activity of RGS proteins as downregulators of G proteins³⁶, whereas NO-dependent arginylation and destruction of RGSs reduce their activity by decreasing their levels in a cell.

RGS4 is a physiological inhibitor of angiogenesis and other tubulogenesis pathways³². Upregulation of RGS4 perturbs cardiovascular homeostasis in mice³¹ and is a molecular correlate of human heart failure³⁰. Mouse ATE1^-/- embryos, which lack R-transferases and therefore lack arginylation, die before E17 with cardiovascular defects¹². Our findings that the levels of RGS4, RGS5 and RGS16 are greatly increased in the hearts and other organs of ATE1^-/- embryos (Fig. 2), and that NO is required for the proteolytic downregulation of these RGSs (Figs 3 and 4), are likely to account in part for the known role of NO, at physiologically optimal levels, in suppressing pathological changes in the heart^{25, 26, 41}. The functions of NO in cardiovascular homeostasis include the stimulation of cyclic GMP formation by guanylyl cyclase and the regulation of cardiac contractility through S-nitrosylation of the calcium release channel^{21, 23, 41}. Our results revealed an entirely different, mutually nonexclusive mechanism of NO signalling in the heart and other organs: the control of regulatory proteins bearing N-terminal Cys through their NO-dependent, arginylation-mediated degradation by the N-end rule pathway. Thus, pharmacological manipulation of activities or expression of R-transferases may provide an alternative, more selective route to clinically beneficial effects that are currently achieved through drugs that alter the levels of NO.

Mammalian R-transferases are strong sequelogues³⁴ of yeast (fungal) ATE1 R-transferases. However, whereas the inactivation of mouse ATE1 results in embryonic lethality¹², a deletion of S. cerevisiae ATE1 renders cells unable to degrade reporters with N-terminal Asp or Glu but has not been found to cause any other abnormal phenotype^{2, 33}. Our findings indicate that one function of arginylation in this and other organisms might be to serve as a sensor of nitrosative/oxidative stress (Supplementary Information). It remains to be determined whether the discovered signalling by NO proceeds exclusively through the oxidation of Cys-containing N-degrons (Fig. 1a) or whether NO can also function at other steps of the N-end rule pathway, for example through S-nitrosylation of its ubiquitin ligases or R-transferases.

A preponderance of circuits relevant to the findings of this work involve arginine: first, Arg is a direct precursor of NO (Fig. 1a); second, the levels of Arg are tightly controlled in a cell and are often downregulated by invading pathogens; third, Arg is a part of Arg-tRNA (a co-substrate of R-transferase), indicating a possible connection between the N-end rule pathway and regulation of translation; fourth, Arg is a primary destabilizing residue and is also conjugated to N-end rule substrates bearing N-terminal Asp, Glu or oxidized Cys (Fig. 1a); and last, some Arg residues in proteins undergo methylation or deimination, the latter a conversion of positively charged Arg to uncharged citrulline^{42, 43}. It remains to be determined whether methylation or deimination of Arg in vivo involves N-terminal Arg, and whether a set of circuits that has now been shown to connect the N-end rule pathway and the signalling by NO holds yet another Arg-linked surprise.

Methods

Full technical details are provided in Supplementary Information.